Optical Wireless Communications: System and Channel Modelling with MATLAB® [2nd ed] 9781498742702, 149874270X, 9781498742696, 9781315151724

Table of contents :
Content: Introduction: Optical Wireless Communication Systems. Optical Sources and Detectors. Channel Modelling. Modulation Techniques. System Performance Analysis: Indoor. FSO Link Performance under the Effect of Atmospheric Turbulence. Outdoor OWC Links with Diversity Techniques. Visible Light Communications. Ultraviolet OWC & Underwater OWC

Citation preview

Optical Wireless Communications

Optical Wireless Communications System and Channel Modelling with MATLAB® Second Edition

By Z. Ghassemlooy, W. Popoola, and S. Rajbhandari

MATLAB® is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks does not warrant the accuracy of the text or exercises in this book. This book’s use or discussion of MATLAB® software or related products does not constitute endorsement or sponsorship by The MathWorks of a particular pedagogical approach or particular use of the MATLAB® software. CRC Press Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2018 by Taylor & Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business No claim to original U.S. Government works Printed on acid-free paper International Standard Book Number-13: 978-1-4987-4269-6 (Hardback) This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright.com (http:// www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging‑in‑Publication Data Names: Ghassemlooy, Zabih, author. | Popoola, W., author. | Rajbhandari, S., author. Title: Optical wireless communications : system and channel modelling with MATLAB / by Z. Ghassemlooy, W. Popoola, and S. Rajbhandari. Description: Second edition. | Boca Raton, FL : CRC Press/Taylor & Francis Group, 2018. Identifiers: LCCN 2018054015 | ISBN 9781498742696 (hardback : acid-free paper) | ISBN 9781315151724 (ebook) Subjects: LCSH: Optical communications. | Signal processing. | Free space optical interconnects. | Signals and signaling—Mathematical models. | Laser communication systems. | MATLAB. Classification: LCC TK5103.592.F73 G46 2018 | DDC 621.382/7—dc23 LC record available at https://lccn.loc.gov/2018054015 Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com and the CRC Press Web site at http://www.crcpress.com

BK-TandF-9781498742696_TEXT_GHASSEMLOOY-181209-FM.indd 4

16/04/19 2:11 PM

Table of Contents List of Figures ...................................................................................................................................xi List of Tables ..................................................................................................................................xxv List of Abbreviations ...................................................................................................................xxvii Preface ....................................................................................................................................... xxxiii About the Authors ......................................................................................................................xxxvii

Chapter 1 Introduction: Optical Wireless Communication Systems ............................................ 1 1.1 1.2 1.3 1.4 1.5 1.6

Wireless Access Schemes ..................................................................................4 A Brief History of OWC .................................................................................. 10 OWC/Radio Comparison ................................................................................. 13 Link Configuration .......................................................................................... 14 OWC Application Areas .................................................................................. 23 Safety and Regulations ....................................................................................25 1.6.1 Maximum Permissible Exposures (MPE) ..........................................28 1.7 OWC Challenges .............................................................................................28 References .................................................................................................................. 32

Chapter 2 Optical Sources and Detectors ................................................................................... 39 2.1 2.2

2.3

Light Sources ................................................................................................... 39 The Light-Emitting Diode ............................................................................... 42 2.2.1 LED Structure .................................................................................... 45 2.2.2 Planar and Dome LEDs ..................................................................... 45 2.2.3 Edge-Emitting LED ...........................................................................46 2.2.4 LED Efficiencies ................................................................................ 48 2.2.4.1 Internal Quantum Efficiency............................................... 48 2.2.4.2 External Quantum Efficiency.............................................. 48 2.2.4.3 Power Efficiency.................................................................. 49 2.2.4.4 Luminous Efficiency............................................................ 49 2.2.4.5 LED Modulation Bandwidth............................................... 50 2.2.5 White LEDs ........................................................................................ 52 2.2.5.1 Thermal Effects������������������������������������������������������������������ 53 The Laser ......................................................................................................... 53 2.3.1 Operating Principle of a Laser ........................................................... 53 2.3.2 Population Inversion ........................................................................... 54 2.3.3 Optical Feedback and Laser Oscillation ............................................ 55 2.3.4 Properties and Specifications of a Laser ............................................ 56 2.3.5 Basic Semiconductor Laser Structure ................................................ 57 2.3.6 The Structure of Common Laser Types ............................................. 58 2.3.6.1 Fabry-Perot Laser ............................................................... 58 2.3.6.2 Distributed Feedback (DFB) Laser..................................... 59 2.3.6.3 Vertical Cavity Surface Emitting Laser (VCSEL)..............................................................................60 2.3.6.4 Superluminescent Diodes (SLDs)........................................ 61 2.3.7 Comparison of LED and Laser Diodes .............................................. 62 v

vi

Table of Contents

2.4 Photodetectors ................................................................................................. 63 2.4.1 PIN Photodetector .............................................................................. 65 2.4.2 APD Photodetector ............................................................................. 67 2.5 Photodetection Techniques .............................................................................. 67 2.5.1 Direct Detection ................................................................................. 68 2.5.2 Coherent Detection ............................................................................. 68 2.5.2.1 Heterodyne Detection.......................................................... 70 2.5.2.2 Homodyne Detection........................................................... 71 2.6 Photodetection Noise ....................................................................................... 72 2.6.1 Quantum Shot Noise .......................................................................... 72 2.6.2 Dark-Current Shot Noise and Excess Noise ....................................... 73 2.6.3 Background Radiation ........................................................................ 74 2.6.4 Thermal Noise .................................................................................... 75 2.6.5 Relative Intensity Noise (RIN) ........................................................... 75 2.6.6 Signal-to-Noise Ratio (SNR) .............................................................. 76 2.7 Optical Detection Statistics ............................................................................. 76 References .................................................................................................................. 78

Chapter 3 Channel Modelling ..................................................................................................... 81 Indoor Optical Wireless Communication Channels .......................................................................................................... 81 3.1.1 LOS Propagation Model .....................................................................84 3.1.2 Non-LOS Propagation Model ............................................................ 87 3.1.3 Ceiling Bounce Model .......................................................................94 3.1.4 Hayasaka-Ito Model ........................................................................... 95 3.1.5 Spherical Model .................................................................................96 3.2 Artificial Light Interference ............................................................................96 3.2.1 Incandescent Lamp ............................................................................99 3.2.2 Fluorescent Lamp Driven by Conventional Ballast .................................................................................................99 3.2.3 Fluorescent Lamp Model ................................................................. 100 3.3 Outdoor Channel ........................................................................................... 104 3.3.1 Atmospheric Channel Loss .............................................................. 104 3.3.2 Fog and Visibility ............................................................................. 107 3.3.3 Beam Divergence ............................................................................. 115 3.3.4 Optical and Window Loss ................................................................ 119 3.3.5 Pointing Loss .................................................................................... 119 3.3.6 The Atmospheric Turbulence Models .............................................. 121 3.3.6.1 Log-Normal Turbulence Model........................................ 125 3.3.6.2 Spatial Coherence in Weak Turbulence ............................ 129 3.3.6.3 Limit of Log-Normal Turbulence Model.......................... 131 3.3.6.4 The Gamma-Gamma Turbulence Model......................... 131 3.3.6.5 The Negative Exponential Turbulence Model .................. 135 3.3.7 Atmospheric Effects on OWC Test Bed ........................................... 135 3.3.7.1 Calibration of the Test Bed to the Real Outdoor Environment....................................................... 139 3.3.7.2 Demonstration of Scintillation Effect on Data Carrying Optical Radiation.............................................. 145 References ................................................................................................................ 151 3.1

Table of Contents

vii

Chapter 4 Modulation Techniques ............................................................................................ 157 4.1 Introduction ................................................................................................... 157 4.2 Analogue Intensity Modulation (AIM) ......................................................... 160 4.3 Digital Baseband Modulation Techniques .................................................... 162 4.3.1 Baseband Modulations ..................................................................... 162 4.3.2 PAM and On-Off Keying (OOK) ..................................................... 163 4.3.3 OOK Error Performance in AWGN Channel ................................... 167 4.4 Pulse Position Modulation ............................................................................. 171 4.4.1 PPM Error Performance ................................................................... 173 4.4.2 PPM Variants ................................................................................... 176 4.4.2.1 Multilevel PPM................................................................. 177 4.4.2.2 Differential PPM............................................................... 178 4.4.2.3 Differential Amplitude Pulse Position Modulation (DAPPM) ...................................................... 178 4.5 Pulse Interval Modulation (PIM) .................................................................. 178 4.5.1 DPIM Error Performance ................................................................. 183 4.5.1.1 DPIM with No Guard Band ............................................. 186 4.5.1.2 DPIM with One Guard Slot.............................................. 187 4.5.2 Optimum Threshold Level ............................................................... 188 4.6 Multilevel DPIM (MDPIM) .......................................................................... 192 4.7 Comparisons of Baseband Modulation Schemes .......................................... 194 4.7.1 Power Efficiency ............................................................................... 194 4.7.2 Transmission Bandwidth Requirements ........................................... 194 4.7.3 Transmission Capacity ..................................................................... 195 4.7.4 Transmission Rate ............................................................................ 196 4.7.5 Peak-to-Average Power Ratio (PAPR) ............................................. 196 4.8 Subcarrier Intensity Modulation .................................................................... 196 4.8.1 Phase Shift Keying ........................................................................... 197 4.8.2 Quadrature Amplitude Modulation (QAM) ..................................... 198 4.9 Multi-Carrier Modulations ............................................................................ 198 4.9.1 Multiple-Subcarrier Intensity Modulation (MSIM) ......................... 199 4.9.2 Orthogonal Frequency Division Multiplexing (OFDM) .................. 201 4.9.2.1 OFDM High PAPR Reduction Techniques ......................204 4.9.2.2 Pilot Signal Estimation at the Receiver ............................206 4.9.3 Carrierless-Amplitude and Phase Modulation (CAP) ......................209 4.10 Optical Polarisation Shift Keying (PoLSK) .................................................. 211 4.10.1 Binary PoLSK .................................................................................. 213 4.10.2 Bit Error Rate Analysis .................................................................... 217 4.10.3 MPOLSK .......................................................................................... 219 4.10.4 Differential Circle Polarisation Shift Keying (DCPoLSK) .............. 221 4.10.5 Error Probability Analysis ............................................................... 223 4.10.6 Comparison of BPOLSK, OOK, and BPSK-Based FSO Links ......... 224 References ................................................................................................................ 225 Chapter 5 Indoor System Performance Analysis ...................................................................... 229 5.1 5.2

The Effect of Ambient Light Sources on Indoor OWC Link Performance ................................................................................ 229 Fluorescent Light Interference with no Electrical High-Pass Filtering ........................................................................................ 230 5.2.1 Matched Filter Rx ............................................................................ 231

viii

Table of Contents

5.3 5.4

BLW without Fluorescent Light Interference ................................................ 238 Fluorescent Light Interference with Electrical High-Pass Filtering ............................................................................................ 246 5.5 Wavelet Analysis ........................................................................................... 249 5.5.1 The Continuous Wavelet Transform (CWT) .................................... 253 5.5.2 The Discrete Wavelet Transform (DWT) ......................................... 255 5.5.3 DWT Based Denoising ..................................................................... 256 5.5.4 Comparative Study of DWT and HPF ............................................. 261 5.5.5 Experimental Investigations ............................................................. 262 5.5.5.1 On-Off Keying (OO......................................................262 5.6 Link Performance in Multipath Propagations ............................................... 265 5.6.1 On-Off Keying (OOK) ..................................................................... 265 5.6.2 Pulse Position Modulation (PPM) ..................................................... 272 5.6.3 Digital Pulse Internal Modulation (DPIM) ........................................ 274 5.7 Mitigation Techniques ................................................................................... 275 5.7.1 Filtering ............................................................................................ 275 5.7.2 Equalisation ...................................................................................... 276 5.7.2.1 The Zero Forcing Equaliser.............................................. 278 5.7.2.2 The Minimum Mean Square Error Equaliser (MMSE)............................................................................ 279 5.7.2.3 The Decision Feedback Equaliser (DFE) ......................... 281 5.8 Equalisation as a Classification Problem ....................................................... 282 5.9 Introduction to Artificial Neural Network .................................................... 282 5.9.1 Neuron .............................................................................................. 283 5.9.2 ANN Architectures ..........................................................................284 5.10 Training Network .......................................................................................... 285 5.10.1 Backpropagation Learning (BP) ...................................................... 286 5.11 The Ann-Based Adaptive Equaliser .............................................................. 286 5.11.1 Comparative Study of the ANN- and FIR-Based Equalisers ............................................................................................. 292 5.11.2 Diversity Techniques ........................................................................ 294 References ................................................................................................................ 295 Chapter 6 FSO Link Performance with Atmospheric Turbulence ........................................... 301 6.1 6.2 6.3

On-Off Keying .............................................................................................. 301 6.1.1 OOK in a Poisson Atmospheric Optical Channel ............................302 6.1.2 OOK in a Gaussian Atmospheric Optical Channel ..........................304 Pulse Position Modulation .............................................................................307 Subcarrier Intensity Modulation .................................................................... 311 6.3.1 SIM Generation and Detection ........................................................ 312 6.3.2 SIM-FSO Performance in Log-Normal Atmospheric Channel ............................................................................................ 314 6.3.3 Bit Error Probability Analysis of SIM-FSO ..................................... 318 6.3.3.1 BPSK-Modulated Subcarrier............................................ 319 6.3.3.2 M-ary PSK-Modulated Subcarrier.................................... 325 6.3.3.3 DPSK-Modulated Subcarrier ............................................ 325 6.3.3.4 Multiple SIM Performance Analysis ................................ 326 6.3.4 Outage Probability ........................................................................... 328 6.3.4.1 In a Log-Normal Atmospheric Channel ........................... 330

ix

Table of Contents

6.3.5

SIM-FSO Performance in Gamma-Gamma and Negative Exponential Atmospheric Channels ................................................. 332 6.3.6 Outage Probability in Negative Exponential Model Atmospheric Channels ..................................................................... 334 6.4 Atmospheric Turbulence-Induced Penalty .................................................... 335 Appendix A .............................................................................................................. 339 Appendix B ..............................................................................................................344 References ................................................................................................................344 Chapter 7 Outdoor OWC Links with Diversity Techniques ..................................................... 347 7.1 7.2

Time Diversity ...............................................................................................348 Spatial Diversity Techniques ......................................................................... 350 7.2.1 Combining Schemes ......................................................................... 353 7.2.1.1 Adaptive Optics Schemes������������������������������������������������� 353 7.2.1.2 Linear Combining Techniques����������������������������������������� 354 7.2.2 Maximum Ratio Combining (MRC) ................................................ 356 7.2.3 Equal Gain Combining (EGC) ......................................................... 357 7.2.4 Selection Combining (SelC) ............................................................. 359 7.2.5 Effect of Received Signal Correlation on Error Performance .......................................................................360 7.2.6 Outage Probability with Receiver Diversity in a Log-Normal Atmospheric Channel ........................................... 362 7.3 Transmitter Diversity in a Log-Normal Atmospheric Channel ..........................................................................................................364 7.4 Transmitter-Receiver Diversity in a Log-Normal Atmospheric Channel ..........................................................................................................364 7.5 Results and Discussions of SIM-FSO with Spatial Diversity in a Log-Normal Atmospheric Channel ........................................................ 365 7.6 SIM-FSO with Receiver Diversity in Gamma-Gamma and Negative Exponential Atmospheric Channels ............................................... 368 7.6.1 BER and Outage Probability of BPSK-SIM with Spatial Diversity ............................................................................... 369 7.6.2 BER and Outage Probability of DPSK-SIM in Negative Exponential Channels ...................................................................... 372 7.7 Terrestrial Free-Space Optical Links with Subcarrier Time Diversity ............................................................................................... 378 7.7.1 Error Performance with STDD ........................................................ 378 7.7.2 Error Performance of Short-Range Links ........................................ 380 7.7.3 Long-Range Links ............................................................................ 380 7.7.4 Short-Range Link ............................................................................. 381 7.7.5 Long-Range Link ............................................................................. 382 7.8 Aperture Averaging ....................................................................................... 384 7.8.1 Plane Wave ....................................................................................... 384 7.8.2 Spherical Wave ................................................................................. 385 7.8.3 Gaussian Beam Wave ....................................................................... 385 7.9 Hybrid RF-FSO Scheme ................................................................................ 386 Appendix A .............................................................................................................. 389 Appendix B .............................................................................................................. 390 Appendix C .............................................................................................................. 391 References ................................................................................................................ 392

x

Table of Contents

Chapter 8 Visible Light Communications ................................................................................. 397 8.1 Introduction ................................................................................................... 397 8.2 Bidirectional VLC .........................................................................................403 8.3 System Description ........................................................................................405 8.3.1 VLC System Model ..........................................................................408 8.3.2 Channel Delay Spread ...................................................................... 418 8.3.3 Holographic Diffuser ....................................................................... 421 8.3.4 SNR Analysis ................................................................................... 422 8.4 System Implementations ................................................................................ 425 8.4.1 On-Off Keying (OOK) with a Forward Error Correction ................ 426 8.4.2 Bit Angle Modulation (BAM) .......................................................... 427 8.4.3 Pulse Modulation Schemes .............................................................. 428 8.4.4 PWM with Discrete Multitone Modulation (DMT) ......................... 430 8.4.5 Multilevel PWM-PPM ...................................................................... 432 8.4.6 PWM with NRZ-OOK ..................................................................... 434 8.5 Multiple-Input, Multiple-Output (MIMO) VLC ........................................... 434 8.6 Orthogonal Frequency Division Multiplexing (OFDM) ..................................440 8.6.1 Channel Estimation, Equalisations, and Synchronisation ...............440 8.7 All Organic VLC ........................................................................................... 443 8.8 Home Access Network ..................................................................................449 8.9 Indoor Localisation ....................................................................................... 452 References ................................................................................................................ 456 Chapter 9 Relay-Assisted FSO Communications .....................................................................469 9.1

Wireless Networks .........................................................................................469 9.1.1 FSO Network Topologies ................................................................. 470 9.2 Relay-Assisted Communications ................................................................... 472 9.2.1 Serial Relaying ................................................................................. 474 9.2.2 Parallel Relaying .............................................................................. 474 9.2.3 All-Optical Relay-Assisted FSO Communications .......................... 475 9.3 All-Optical Amplify-and-Forward ................................................................ 476 9.3.1 Optical Amplification ....................................................................... 476 9.3.1.1 Erbium-Doped Fibre Amplifier ........................................ 478 9.3.1.2 Semiconductor Optical Amplifier..................................... 478 9.3.1.3 Comparison of EDFAs and SOAs.................................... 478 9.4 All-Optical Regenerate-and-Forward ........................................................... 479 9.4.1 Nonlinear Effects ............................................................................. 479 9.4.2 SPM-Based Optical Regenerator ...................................................... 481 9.4.3 Highly Nonlinear Fibres ................................................................... 482 9.5 All-Optical Aoaf Relay-Based FSO with Turbulence ................................... 482 References ................................................................................................................ 486 Index .............................................................................................................................................. 491

List of Figures Figure 1.1  Global data traffic..........................................................................................................2 Figure 1.2  Power per user for different access technologies..........................................................8  isible light in the electromagnetic spectrum in the context of other Figure 1.3  V communications technologies, adopted from [49]........................................................9 Figure 1.4  N  ormalised power/unit wavelength for optical wireless spectrum and ambient light sources................................................................................................... 11 Figure 1.5  B  andwidth capabilities for a range of optical and RF technologies for (a) long range and (b) short range................................................................................ 15 Figure 1.6  Access network bottleneck.......................................................................................... 16 Figure 1.7  A system block diagram of an OWC system............................................................... 16 Figure 1.8  Link configurations: transmitter and receivers........................................................... 17 Figure 1.9  Cellular OWC system.................................................................................................. 17 Figure 1.10  Multi-cell indoor non-directed cellular OWC systems.............................................. 19 Figure 1.11  Multi-spot diffusing configuration............................................................................20 Figure 1.12  Hybrid diffuse and tracked LOS links...................................................................... 21 Figure 1.13  Optical wireless LAN: (a) diffuse and (b) line of sight............................................. 22 Figure 1.14  The effects of scintillation......................................................................................... 23 Figure 1.15  A system block diagram of an outdoor OWC link....................................................24 Figure 1.16  Response/absorption of the human eye at various wavelengths................................25 Figure 1.17  The eye safety limits for 900 and 1550 nm wavelengths...........................................28 Figure 2.1  W  avelength and energy of the ultraviolet, visible, and IR portions of the electromagnetic spectrum.................................................................................40 Figure 2.2  Optical transmission windows.................................................................................... 41 Figure 2.3  T  wo-level atomic system illustrating the three fundamental processes: (a) absorption, (b) spontaneous emission, and (c) stimulated emission...................... 41 Figure 2.4  L  ED: (a) the luminescence intensity as a function of the energy for spontaneous emission process and (b) spectral profile............................................... 43 Figure 2.5  (a) GaAs diode emission spectrum at 295 K and 77 K and (b) dependence of the emission peak and half-power width on the temperature.................................44 Figure 2.6  A  n illustration of the radiated optical power against the driving current of an LED....................................................................................................................44 Figure 2.7  Planar LED structure showing light emission on all surfaces....................................46 Figure 2.8  Lambertian intensity distribution................................................................................46 Figure 2.9  The basic structure of a dome (hemispherical) LED.................................................. 47 xi

xii

List of Figures

Figure 2.10  The structure of a double heterojunction AlGaAs edge-emitting LED.................... 47 Figure 2.11  LEDs power versus current characteristics............................................................... 49 Figure 2.12  E  ye sensitivity function based on the 1978 CIE data. (http://www.ecse.rpi.edu/~schubert/Light-Emitting-Diodes-dot-org)...................... 50 Figure 2.13  LED luminous efficiency.......................................................................................... 51 Figure 2.14  An illustration of the optical and electrical bandwidth............................................. 52 Figure 2.15  (a) Normalised optical spectrum of a white LED, and (b) actual and projected increases in the efficacy of colour-mixed (CM) and phosphor-coated (PC) LED packages. (http://www.hi-led.eu/wp-content/ themes/hiled/pdf/led_energy_efficiency.pdf)........................................................... 53 Figure 2.16  An illustration of lasing action based on a four-level He-Ne laser............................ 54 Figure 2.17  L  aser: (a) an illustration of optical feedback and (b) a typical spectrum of multi-mode LD...................................................................................... 55 Figure 2.18  L  aser: (a) output power against drive current plot and (b) small signal model..............57 Figure 2.19  DFB laser: (a) structure and (b) spectral profile........................................................ 59 Figure 2.20  The basic structure of a VCSEL............................................................................... 61 Figure 2.21  Relative response of PDs for different materials.......................................................64 Figure 2.22  A PIN PD: (a) schematic diagram and (b) reversed biased....................................... 65 Figure 2.23  R  esponsivity and quantum efficiency as a function of wavelength for PIN PDs. Reproduced from [17]..........................................................................66 Figure 2.24  Block diagram of a direct detection optical Rx........................................................ 68 Figure 2.25  Block diagram of a coherent detection optical Rx.................................................... 69 Figure 2.26  A coherent Rx with balanced detection. BS: beam splitter...................................... 70 Figure 2.27  D  iagram of a front-end photodiode detector along with channel impairments................................................................................................. 72 Figure 3.1  Block diagram of an IM/DD OWC system................................................................. 82 Figure 3.2  Equivalent baseband model of an IM/DD OWC link................................................. 82 Figure 3.3  A block diagram of the OWC link with spatial diversity............................................ 83 Figure 3.4  Geometry of the LOS propagation model................................................................... 85 Figure 3.5  Geometry used to describe the single-reflection propagation model.......................... 89 Figure 3.6  I mpulse responses of a diffused link: (a) with a LOS path and (b) with no LOS paths................................................................................................. 89 Figure 3.7  C  hannel delay spread: (a) mean delay spread, (b) RMS delay spread with LOS component, (c) RMS delay spread with no LOS, and (d) maximum data rate distributions........................................................................... 91 Figure 3.8  Average time between two reflections as a function of the room dimensions............97 Figure 3.9  T  he frequency response of the diffused channel using the first diffused reflection and the sphere models.................................................................................97

List of Figures

xiii

Figure 3.10  Incandescent bulb: (a) time domain waveform and (b) frequency spectrum. No optical filtering....................................................................................99 Figure 3.11  L  ow-frequency fluorescent lamp: (a) time domain waveform and (b) frequency spectrum. No optical filtering........................................................... 100 Figure 3.12  H  F fluorescent lamp time domain waveform: (a) low-frequency component, (b) high-frequency component, and detected electrical spectrum, (c) low-frequency component, and (d) high-frequency component......................... 101 Figure 3.13  (a) Low-frequency interference component and (b) high-frequency interference component........................................................................................... 103 Figure 3.14  A  tmospheric absorption transmittance over a sea level 1820 m horizontal path [57]................................................................................................................... 106 Figure 3.15  D  istribution of the electromagnetic radiation resulting from (a) Rayleigh or molecular scattering and (cb) Mie or aerosol scattering..................................... 107 Figure 3.16  S  chematic depiction of fog formation: (a) radiation and (b) advection. (Data from Robert Tardif (2011))............................................................................. 108 Figure 3.17  P  article size distribution versus particle radius (μm): (a) advection fog and (b) convection fog....................................................................................... 110 Figure 3.18  A  ttenuation versus visibility using the Kruse model for: (a) Tth = 2% and (b) Tth = 5%....................................................................................................... 112 Figure 3.19  Kim model against the visibility for Tth of 2% and a range of wavelengths................113 Figure 3.20  T  ime profiles of (a) visual range, (b) specific attenuation, and (c) differences in specific attenuation during a fog event that occurred in Milan on 11 January 2005. In (b), two profiles are shown: the measured laser attenuation (red curve) and the attenuation as estimated from visual range (blue curve) [75]............................................................................................ 114 Figure 3.21  A  ttenuation versus visibility for Kim and Naboulsi models for fog at λ of 0.83 μm: (a) Tth of 5% (b) Tth of 2%, (c) 1550 nm and (d) 2000 nm.................... 115 Figure 3.22  C  omparison of different models for a range of wavelengths for different visibility values: (a) 0.1 km, (b) 0.25 km, (c) 0.5 km, and (d) 1 km........................ 116 Figure 3.23  M  easured attenuation coefficient as a function of visibility range at λ = 830 nm in early 2008, Prague, Czech Republic [82]........................................ 116 Figure 3.24  Beam divergence..................................................................................................... 117 Figure 3.25  Typical beam expander............................................................................................ 118 Figure 3.26  The receiver aperture and a laser beam footprint................................................... 120 Figure 3.27  An atmospheric channel with turbulent eddies....................................................... 122 Figure 3.28  L  og-normal probability density function with E[I] = 1 for a range of log irradiance variance. σl2 ..................................................................................... 128 Figure 3.29  Plane wave transverse coherence length for λ = 850 nm and a range of Cn2...............130 Figure 3.30  Plane wave transverse coherence length for λ = 1550 nm and a range of Cn2.............130 Figure 3.31  G  amma-gamma probability density function for three different turbulence regimes, weak, moderate, and strong...................................................................... 133

xiv

List of Figures

Figure 3.32  S.I against log-intensity variance for Cn2 = 10 −15 m−2/3 and λ = 850 nm......................134 Figure 3.33  Values of α and β under different turbulence regimes: weak, moderate to strong, and saturation.......................................................................... 134  egative exponential probability density function for different values of I0...............136 Figure 3.34  N Figure 3.35  (a) Block diagram of the FSO experimental setup, (b) the simulation chamber, and (c) the laboratory chamber setup...................................................... 137  omparison of measured mean ROF attenuation for two field experiments Figure 3.36  C [126], [127] and the mean lab-based fog attenuation data at 0.83 μm. ROF: Real outdoor fog...................................................................................................... 140 Figure 3.37  (a) Changes in loss against visibility and (b) visibility versus the change in V. Black (solid line) is the real V for link range of 5.5 m........................ 141 Figure 3.38  The block diagram for the measurement of fog attenuation and visibility............. 142 Figure 3.39  Time dependence of visibility within the FSO chamber........................................ 143 Figure 3.40  M  easured attenuation against: (a) time and (b) V for full fog event at a wavelength of 1550 nm............................................................................................ 143 Figure 3.41  T  he measured fog attenuation against the wavelength (visible – NIR spectrum) for: (a) dense (V = 0.048 km), (b) thick (V = 0.3 km), (c) moderate (V = 0.46 km), and (d) light (V = 0.783 km) fog conditions................ 144 Figure 3.42  Refractive index for glycerine-based smoke against the wavelength...................... 145 Figure 3.43  R  eceived signal distribution without scintillation (mean = 0.0487, σ2 = 1.2e − 5)................................................................................ 146 Figure 3.44  R  eceived signal distribution with scintillation (mean = 0.0451, scintillation index = 0.0164)......................................................... 146 Figure 3.45  T  he measured eye diagram of received NRZ signal in the condition of: (a) no turbulence, 12.5 Mbps and (b) weak–medium turbulence with σ12 = 0.0164..................................................................................................... 147 Figure 3.46  T  he histogram of the received OOK-NRZ signal: (a) without turbulence and with turbulence Rytov variance of (b) 0.00423, (c) 0.0483, and (d) 0.242.......................................................................................... 149 Figure 3.47  T  he predicted and measured Q-factor (OSNR) against the Rytov variance for OOK-NRZ at 12.5 Mbps..................................................................... 150 Figure 4.1  Modulation tree......................................................................................................... 158 Figure 4.2  O  ptical transmission system block diagram: (a) IM, (b) external modulation, and (c) Rx.............................................................................................. 160 Figure 4.3  Time domain waveforms of 4-PAM modulation schemes........................................ 163 Figure 4.4  Transmitted waveforms for OOK: (a) NRZ and (b) RZ (γ = 0.5).............................. 164 Figure 4.5  PSD of (a) OOK-NRZ and (b) OOK-RZ (γ = 0.5)..................................................... 165 Figure 4.6  The block diagram of OOK system........................................................................... 167 Figure 4.7  Matched filter output for detected OOK-NRZ pulse................................................. 168 Figure 4.8  PSD of PPM for L = 4, 8, and 16............................................................................... 172

List of Figures

xv

Figure 4.9  T  he block diagram of the matched filter–based Rx for PPM scheme with soft and hard decision decoding........................................................................ 174 Figure 4.10  BER performance for OOK (NRZ and RZ) and 4-PPM........................................ 177 Figure 4.11  Valid symbols for

( 42 ) MPPM................................................................................. 177

Figure 4.12  M  apping of source data to transmitted symbols for 4-DPIM(NGB) and 4-DPIM(1GS).................................................................................................... 179 Figure 4.13  PSD of (a) DPIM(NGB) and (b) DPIM(1GS) for L = 4, 8, 16, and 32..................... 182 Figure 4.14  T  ypes of error in DPIM: (a) Transmitted 8-DPIM(1GS) signal, (b) erasure error, (c) false alarm error, and (d) wrong slot error.............................. 183 Figure 4.15  The block diagram of the matched filter–based Rx for the DPIM scheme.................184 Figure 4.16  P  ER versus probability of slot error for 16-DPIM(NGB) for a range of packet lengths...................................................................................................... 185 Figure 4.17  Block diagram of a DPIM demodulator.................................................................. 185 Figure 4.18  Using DPIM in a fixed throughput system.............................................................. 186 Figure 4.19  Matched filter output for detected DPIM pulse....................................................... 186 Figure 4.20  C  onditional probability density functions of s0 in the presence of signal independent AWGN.................................................................................. 188 Figure 4.21  Scaled conditional probability density function of s0 for P(0) > P(1)..................... 189 Figure 4.22  N  ormalised optimum threshold level versus PER: (a) DPIM(NGB) and (b) DPIM(1GS)................................................................................................. 191 Figure 4.23  (a) MDPIM Tx and (b) MDPIM Rx system block diagrams.................................. 193 Figure 4.24  Block diagram of multiple-subcarrier modulation system...................................... 199 Figure 4.25  Spectra of FDM and OFDM systems...................................................................... 201 Figure 4.26  B  lock diagram of DCO-OFDM–based optical communication system. S/P: serial-to-parallel converter, P/S: parallel-to-serial converter, DAC: Digital-to-analogue converter, TIA: Transimpedance amplifier, ADC: analogue-to-digital converter.......................................................................202 Figure 4.27  Pilot-assisted OFDM signal frame structure...........................................................205 Figure 4.28  T  he PAPR CCDF A plot of the PAPR CCDF for DCO-OFDM with and without the pilot-assisted PAPR reduction algorithm; 16-QAM, R = 5 iterations, L = 4, U = 5, and Nsub = 127 data-carrying subcarriers................207 Figure 4.29  B  ER performance of DCO-OFDM with and without pilot-assisted PAPR reduction algorithm; 16-QAM, R = 5 iterations, L = 4, U = 5, and Nsub = 127 data-carrying subcarriers................................................................208 Figure 4.30  Block diagram of the: (a) QAM and (b) CAP modulation schemes........................ 210 Figure 4.31  An illustration of (a) frequency selective channel and (b) flat fading channel............211 Figure 4.32  A  n illustration of the m-CAP concept in terms of the frequency response for different values of m: (a) m = 1, (b) m = 2, (c) m = 4, (d) m = 8, and (e) m = 10................................................................................................................. 212

xvi

List of Figures

Figure 4.33  (a) PolSK Tx block diagram and (b) LiNbO3 modulator. LD, laser diode; PBS: polarising beam splitter.................................................................................. 213 Figure 4.34  A  block diagram of the coherent optical PolSK heterodyne Rx. LO: local oscillator; DC: directional coupler; BPF: band-pass filter; LPF: low-pass filter.......................................................................................................... 215 Figure 4.35  SOPs at the output of PolSK Rx.............................................................................. 217 Figure 4.36  B  lock diagram of the MPOLSK Tx for coherent optical communication systems. LD (laser diode); COD (encoder); PBS (polarisation beam splitter); PBC (polarisation beam combiner); AM (external amplitude modulator); PM (LiNbO3 device–based external phase modulator).......................................... 219 Figure 4.37  B  lock diagram of the MPOLSK coherent optical Rx. Optical BPF (optical band-pass filter); BPF (electric band-pass filter); LO (local oscillator); AFC (automatic frequency control); PLL (Costas loop–based phase-locked loop circuit); P/S (parallel-to-serial converter)................................................................ 220 Figure 4.38  B  lock diagram of DCPOLSK Tx for the coherent optical communication system. TL (transmitting laser); POLSK (external polarisation modulator)............................................................................ 222 Figure 4.39  B  lock diagram of a heterodyne Rx for a DCPOLSK-modulated signal. OBPF (optical band-pass filter); QW (quarter-wave plate); LO (local oscillator); AFC (automatic frequency control circuit); PBS (polarisation beam splitter); PD (photodiode); BPF (electric band-pass filter); and LPF (electric low-pass filter)........................................................................................... 222 Figure 4.40  C  omparisons of BER performances of various schemes against the normalised electric SNR E[ℜPr Plo ] = 1. (sq: BPOLSK-FSO using the square-law modulation; ref: BPOLSK-FSO with a reference carrier)....................224 Figure 5.1  Block diagram of the OWC system with FLI............................................................ 231 Figure 5.2  Flowchart for simulation of OOK shown in Figure 5.1............................................. 232 Figure 5.3  N  OPR and OPP to achieve a BER of 10 −6 against the data rate for OOK with and without FLI....................................................................................... 233 Figure 5.4  N  OPR to achieve an SER of 10 −6 against the data rate for 4, 8, and 16-PPM with HDD and with/without FLI.......................................................... 236 Figure 5.5  N  OPR to achieve an SER of 10 −6 against the data rate of 4, 8, and 16-PPM with SDD and with/without FLI.......................................................... 236 Figure 5.6  NOPR against the data rate for 4, 8, and 16 DPIM with and without FLI................ 237 Figure 5.7  T  ime waveforms: (a) transmitted binary signal, (b) high-pass filter output, (c) BLW signal, and (d) for three different high-pass filter cut-on frequencies............................................................................................. 239 Figure 5.8  H  istogram of MF output for OOK with: (a) fc−hpf  /Rb = 10−3, and (b) f−hpfc/Rb = 10−2...................................................................240 Figure 5.9  The normalised optical power requirement versus fc−hpf  /Rb for OOK.......................244 Figure 5.10  T  he normalised average optical power requirement versus fc−hpf /Rb for PPM(HDD)...........................................................................................244

List of Figures

xvii

Figure 5.11  T  he normalised average optical power requirement versus fc−hpf  /Rb for PPM(SDD)................................................................................. 245 Figure 5.12  T  he normalised average optical power requirement versus f−hpfc /Rb for DPIM..................................................................................................... 245  he normalised average optical power requirement versus the Figure 5.13  T normalised HPF cut-on frequency for OOK for a range of data rates.................... 247 Figure 5.14  The normalised average optical power requirement against the normalised HPF cut-on frequency for PPM(HDD) for a range of data rates and for (a) L = 4, (b) L = 8, and (c) L = 16............................................................................248 Figure 5.15  T  he normalised average optical power requirement as a function of the normalised HPF cut-on frequency for PPM(SDD) for various data rates and for: (a) L = 4, (b) L = 8, and (c) L = 16.............................................................. 249 Figure 5.16  N  ormalised average optical power requirement versus the normalised HPF cut-on frequency for DPIM for various data rates and for: (a) L = 4, (b) L = 8, and (c) L = 16............................................................................ 250 Figure 5.17  T  he normalised average optical power requirement versus the bit rate for OOK with FLI and optimised HPF.................................................................... 251 Figure 5.18  T  he normalised average optical power requirement versus the bit rate for PPM(HDD) with FLI and optimised HPF......................................................... 251 Figure 5.19  T  he normalised average optical power requirement versus the bit rate for PPM(SDD) with FLI and optimised HPF.......................................................... 252 Figure 5.20  T  he normalised average optical power requirement against the bit rate for DPIM with FLI and optimised HPF.................................................................. 252 Figure 5.21  The scaling and translation to the Morlet wavelet.................................................. 253 Figure 5.22  The CWT of the signal of: (a) non-stationary and (b) stationary............................ 254 Figure 5.23  Time–frequency representation of (a) CWT and (b) STFT.................................... 255 Figure 5.24  DWT: (a) decomposition of a signal and (b) decomposition process...................... 256 Figure 5.25  DWT-based Rx in the presence of artificial light interference............................... 257 Figure 5.26  T  he normalised optical power requirement against the data rates for OOK with and without DWT denoising in the presence of FLI............................. 258 Figure 5.27  N  OPR versus the data rates for 4, 8, and 16-PPM with the hard decoding scheme with DWT denoising in the presence of FLI..............................................260 Figure 5.28  N  OPR versus the data rates for 4, 8, and 16-PPM with the soft decoding scheme with DWT denoising in the presence of FLI..............................................260 Figure 5.29  N  OPR versus the data rates for 4, 8, and 16-DPIM schemes with DWT denoising in the presence of FLI............................................................................. 261 Figure 5.30  T  he OPP against the decomposition level for the DWT based Rx at the data rates of 20, 50, 100, and 200 Mbps................................................................. 262 Figure 5.31  T  he OPP against cut-off frequencies for an HPF based Rx at the data rates of 20, 50, 100, and 200 Mbps......................................................................... 262 Figure 5.32  The schematic diagram of the experimental set-up for an indoor OWC link............ 263

xviii

List of Figures

Figure 5.33  (a) Time waveform and (b) spectrum of the interference produced by an artificial light source...................................................................................... 263 Figure 5.34  T  he time waveform of the received signal at 10 Mbps in the presence of FLI............................................................................................................ 264 Figure 5.35  (a) The measured and simulated Q-factors against the data rate with and without FLI and eye diagrams of the received OOK-NRZ in the absence of FLI, (b) at 1 Mbps, and (c) 10 Mbps..................................................... 265 Figure 5.36  Eye diagrams of the received OOK-NRZ at 25 Mbps in the presence of FLI: (a) without filter, (b) analogue HPF, (c) digital HPF, (d) DWT, and (e) the measured and simulated Q-factors against the data rate for indoor OWC systems in the presence of FLI with HPF and wavelet denoising.................266 Figure 5.37  The block diagram of the unequalised OOK system..............................................266 Figure 5.38  Discrete-time equivalent system for the communication system............................ 267 Figure 5.39  Discrete-time impulse response ck for a sequence of six taps................................. 267 Figure 5.40  E  xample of multipath distortion on OOK-NRZ at a normalised delay spread of 0.4................................................................................................... 269 Figure 5.41  N  OPR against the normalised delay spread DT for unequalised OOK in a diffuse indoor OWC channel.................................................................. 271 Figure 5.42  T  heoretical and simulated OPPs for the unequalised OOK system against h0................................................................................................................. 271 Figure 5.43  Discrete-time equivalent system for PPM scheme.................................................. 272 Figure 5.44  N  OPR against the normalised delay spread for 4, 8, and 16-PPM with hard and soft decision decoding schemes in a dispersive channel................................................................................................... 273 Figure 5.45  N  OPR against the normalised delay spread for 4, 8, and 16-DPIM with and without guard slots in a dispersive channel.............................................. 274 Figure 5.46  B  ER against the SNR for: (a) 50 Mbps OOK-NRZ and (b) 25 Mbps OOK-RZ for a range of cut-off frequencies....................................... 276 Figure 5.47  B  ER against SNR for OOK-NRZ/RZ with 0.6/Tb and 0.35/Tb low-pass filtering..................................................................................................... 277 Figure 5.48  Structure of the (a) ZFE and (b) adaptive linear transversal equaliser................... 279 Figure 5.49  Structure of the minimum mean square error equaliser.........................................280 Figure 5.50  S  chematic diagram of DFE with an algorithm to update tap coefficients applied at dashed boxes....................................................................... 281 Figure 5.51  T  he schematic diagram of a neuron showing inputs, a bias, weights, and an output........................................................................................................... 283 Figure 5.52  A single-layer feedforward network with two output neurons................................284 Figure 5.53  Fully connected feedforward multilayer network...................................................284 Figure 5.54  Neural network with a feedback connection........................................................... 285 Figure 5.55  ANN based DFE structure...................................................................................... 286

List of Figures

xix

Figure 5.56  N  OPR versus the normalised delay spread for unequalised and ANN equalised (linear and decision feedback) OOK schemes.............................. 290 Figure 5.57  N  OPR against the normalised delay spread DT for unequalised and ANN-based equalisers for the PPM(HDD) scheme......................................... 291 Figure 5.58  N  OPR against the normalised delay spread DT for unequalised and ANN-based linear equalisers for PPM(SDD) scheme..................................... 291 Figure 5.59  N  OPR versus the normalised RMS delay spread for unequalised and ANN equalisers for the DPIM scheme............................................................ 292  ER against the electrical SNR for the traditional and ANN linear Figure 5.60  B equalisers for the OOK scheme at a data rate of 200 Mbps for a channel with DT of 2 with different training lengths (TLs).................................................. 293 Figure 5.61  B  ER against the electrical SNR for the traditional and ANN DF equalisers for the OOK modulation scheme at a data rate of 200 Mbps for a channel with DT of 2 with different training lengths...................................... 293 Figure 5.62  T  he MSE between the actual and target outputs from the linear equalisers for: (a) FIR filter equaliser and (b) ANN equaliser................................ 294 Figure 6.1  B  ER against the average photoelectron count per bit for OOK-FSO in a Poisson atmospheric turbulence channel for σl2 = [0, 0.1, 0.2, 0.5]................... 303 Figure 6.2  T  he likelihood ratio against the received signal for different turbulence levels and for the noise variance of 10 −2................................................. 305 Figure 6.3  O  OK threshold level against the log intensity standard deviation for various noise levels....................................................................................................306 Figure 6.4  B  ER of OOK-based FSO in atmospheric turbulence with σl2 = 0.2 considering fixed and adaptive threshold levels........................................................306 Figure 6.5  B  ER of OOK-FSO with a fixed and adaptive threshold at various levels of scintillation, σl = [0.2, 0.5, 0.7] and I0 = 1..................................................307 Figure 6.6  Time waveforms for 4-bit OOK and 16-PPM...........................................................308 Figure 6.7  B  inary PPM BER as a function of scintillation index for K Bg = 10; Te = 300 K, ζ = 0.028, Rb = 155 Mbps, and g- = 150..................................................309 Figure 6.8  B  lock diagram of SIM-FSO: (a) transmitter and (b) receiver. TIA (trans-impedance), OBPF (Optical band-pass filter)......................................... 312 Figure 6.9  O  utput characteristic of an optical source driven by a subcarrier signal showing optical modulation index............................................................................ 314 Figure 6.10  Q  PSK constellation of the input subcarrier signal without the noise and channel fading................................................................................................... 315 Figure 6.11  R  eceived constellation of QPSK premodulated SIM-FSO with the noise and channel fading for SNR = 2 dB and σl2 = 0.001............................................... 315 Figure 6.12  R  eceived constellation of QPSK premodulated SIM-FSO with the noise and channel fading for SNR = 2 dB and σl2 = 0.5................................................... 316 Figure 6.13  B  ER against the normalised SNR using numerical and 20th-order Gauss-Hermite integration methods in weak atmospheric turbulence for σl2 = 0.12........................................................................................... 321

xx

List of Figures

Figure 6.14  T  he BER against the average received irradiance in weak turbulence under different noise-limiting conditions for Rb = 155 Mbps and σl2 = 0.3...................... 322 Figure 6.15  B  lock diagram of an FSO link employing DPSK-modulated SIM; (a) transmitter and (b) receiver. TIA (trans-impedance amplifier); TT (transmitter telescope); RT (receiver telescope)................................................. 326 Figure 6.16  B  ER against the received irradiance for SIM-FSO with different subcarrier modulation techniques in weak atmospheric turbulence for σl2 = 0.3, λ = 850 nm, and link range = 1 km.................................................... 327 Figure 6.17  B  ER against the normalised SNR for multiple subcarrier FSO system in weak atmospheric turbulence for N = [1, 2, 5, 10] and σl2 = 0.3............. 328 Figure 6.18  S  NR required to attain a BER of 10 −6 against the number of subcarriers for BPSK-modulated SIM-FSO system with σl = [0.1, 0.2, 0.5, 0.7]....................... 329 Figure 6.19  O  utage probability against the power margin for a log-normal turbulent atmospheric channel for σl2 = [0.1, 0.3, 0.5, 1]........................................................ 331 Figure 6.20  B  ER performance against the normalised electrical SNR across all of the turbulence regimes based on gamma-gamma and negative exponential modes................................................................................................... 333 Figure 6.21  E  rror performance of BPSK-SIM and OOK with a fixed and adaptive threshold-based FSO in weak turbulence regime modeled using gamma-gamma distribution........................................................................... 334 Figure 6.22  T  he outage probability against the power margin in saturation and weak turbulence regimes for σl2 = 0.5.............................................................................. 335 Figure 6.23  E  rror rate performance against normalised SNR for BPSK-SIM-based FSO in weak atmospheric turbulence channel for σl2 = [0, 0.1, 0.3, 0.5, 0.7]................................................................................... 336 Figure 6.24  T  urbulence-induced SNR penalty as function of log irradiance variance for BPSK-SIM-based FSO for BER = [10 −3, 10 −6, 10 −9]......................................... 337 Figure 6.25  B  ER of BPSK-SIM against the turbulence strength in weak atmospheric turbulence for normalised SNR (dB) = 5, 20, 25, and 30................... 338 Figure 6.26  P  Out of BPSK-SIM against the turbulence strength in weak atmospheric turbulence for mp (dBm) = 30, 35, 38, and 40......................................................... 339 Figure 7.1  (a) Spatial diversity schemes and (b) a MIMO channel model.................................. 351 Figure 7.2  The mth FSO link in a SIMO scheme. LD and PD are laser diode and photodetector, respectively........................................................................................ 351 Figure 7.3  T  he BER versus SNR for SISO-FSO over clear, weak, moderate, and strong turbulence regimes with and without aperture averaging. ‘Weak’, ‘Mod’, and ‘Str’ refer to weak, moderate, and strong regimes, respectively. ‘+AF’ denotes applying aperture averaging.............................................................. 352 Figure 7.4  Block diagram of a spatial diversity receiver with

detectors................................ 355

Figure 7.5  A  typical SISO FSO Rx. The received signal is shown at optical, electrical, and logical domains for a bitstream of {1, 0, 1, 1, 0}. ORx: Optical Rx; OptAp: Optical aperture........................................................................360

xxi

List of Figures

Figure 7.6  Block diagram of combining schemes: (a) optical domain, (b–c) electrical domains, and (d) logical domain. OTx: optical transmitter....................................... 361 Figure 7.7  C  orrelation coefficient for a weak turbulent field as a function of transverse separation. Reproduced from [69]............................................................ 362 Figure 7.8  T  he exact outage probability and its upper bound against the power margin with EGC spatial diversity in the weak turbulent atmospheric channel for σl2 = 0.22 and = [1, 4]......................................................................... 363 Figure 7.9  B  PSK-SIM link margin with EGC and SelC against a number of photodetectors for various turbulence levels and a BER of 10 −6............................... 365 Figure 7.10  D  PSK-SIM with SelC spatial diversity link margin against turbulence strength for = [2, 4, 6, 8, 10]............................................................................... 366 Figure 7.11  B  PSK-SIM diversity link margin with EGC and MRC against number of photodetectors for various turbulence levels and a BER of 10 −6......................... 367 Figure 7.12  E  GC diversity gain in log-normal atmospheric channel against the number of photodetectors at POut of 10 −6 and σl2 = [0.22, 0.52, 0.72, 1].............. 367 Figure 7.13  E  rror performance of BPSK-SIM at different values of the correlation coefficient for = [2, 3] and σl2 = 0.52................................................................... 368 Figure 7.14  E  rror performance of BPSK-SIM with MIMO configuration in the turbulent atmospheric channel for σl2 = 0.3............................................................. 369 Figure 7.15  B  PSK-SIM error rate against the normalised SNR in gamma-gamma and negative exponential channels for two photodetectors..................................... 371 Figure 7.16  The outage probability as a function of power margin mEGC (dBm) for = [1, 2, 4] in the negative exponential channel................................................... 372 Figure 7.17  D  iversity gain against number of independent photodetectors at BER and POut of 10 −6 in the negative exponential atmospheric channel................. 373 Figure 7.18  The pdf, p (max {Ii}i=1 ) for = [1, 4, 10], and I0 = 1 in a negative exponential channel................................................................................................. 374 Figure 7.19  E  rror rate of DPSK-SIM against the average received irradiance with spatial diversity in negative exponential channel for = [2, 4, 6, 8, 10]............... 375 Figure 7.20  O  utage probability against the average irradiance with SelC spatial diversity in negative exponential channel for I* = 0 dBm and = [1, 2, 4, 6, 10]................................................................................................... 376 Figure 7.21  P  redicted SelC diversity gain per photodetector against POut in saturation regime for = [1, 2, 4, 6, 10]................................................................ 377 Figure 7.22  P  redicted SelC diversity gain (dB) per photodetector for POut = [10 −6, 10 −3, 10 −2] in saturation regime........................................................... 377 Figure 7.23  T  he subcarrier STDD block diagram: (a) transmitter and (b) receiver. TIA-trans-impedance amplifier; OBPF-optical bandpass filter.............................. 379 Figure 7.24  The BER against –γ with and without STDD at 155 Mbps, σ 2 = 0.3, l

Cn = 0.75 × 10 −14 m−2/3.............................................................................................. 382

Figure 7.25  T  he average received irradiance at a BER of 10 −9, Rb (Mbps) = [155, 625] and 1-STDD gain for different strengths of turbulence......................... 383

xxii

List of Figures

Figure 7.26  L  ink BER performance against the normalised SNR with the subcarrier STDD in the negative exponential turbulence–induced fading channel................. 383 Figure 7.27  Aperture averaging factor for a plane wave propagation with the inner scale lo = 0............................................................................................................... 385 Figure 7.28  A  perture averaging factor for a spherical wave propagation with the inner scale lo = 0...................................................................................................... 386 Figure 7.29  A  perture averaging factor for a Gaussian beam wave propagation with the inner scale lo = 0........................................................................................ 386 Figure 7.30  M  easured availability of RF, FSO, and a hybrid system in percentage versus measurement date. The measurement was taken from December 2001 to January 2003 at Graz University of Technology. FSO wavelength was 850 nm with a total transmit power of 8 mW. The link distance was 2.7 km. More details are available in [29]........................................................ 387 Figure 7.31  F  SO and RF link attenuations versus link distance for different channel conditions. The RF and FSO detection threshold shows the achievable level of the distance, within which the link is available with BER of 10 −6..................................................................................................... 388 Figure 8.1  Two approaches to generating white light using LEDs............................................. 398 Figure 8.2  An illustration of the VLC concept...........................................................................400 Figure 8.3  Key features of VLC.................................................................................................405 Figure 8.4  A block diagram of a VLC system............................................................................406 Figure 8.5  (a) VLC link, (b) LED optical spectrum of Osram Ostar white-light LED, and (c) modulation bandwidth with and without blue filtering.................................406 Figure 8.6  A typical WPLED power–current response..............................................................407 Figure 8.7  Type of light sources for VLC systems.....................................................................407 Figure 8.8  Illumination by LEDs...............................................................................................408 Figure 8.9  (a) LED array and illuminance distributions for (b) one transmitter and (c) four transmitters............................................................................................ 410 Figure 8.10  O  ptical power distribution at the received optical plane for an FWHM of (a) 70° and (b) 12.5°............................................................................... 412 Figure 8.11  Modeling of VLC channel with IM/DD................................................................. 414 Figure 8.12  T  he PSD (solid line, which corresponds to the left axis) is compared to the measured spectral reflectance (which corresponds to the right axis) of plaster and plastic wall (dash-dot line), floor (dash line), and ceiling (dot line)[100].......................................................................................................... 415 Figure 8.13  (a) LOS and diffuse propagation model and (b) received optical power distribution for the first reflection............................................................................ 415 Figure 8.14  The distribution of received power with reflections................................................ 417 Figure 8.15  R  MS delay distribution for (a) a single transmitter positioned at (2.5, 2.5) and (b) for four transmitters positioned at (1.25, 1.25), (1.25, 3.75), (3.75, 1.25), (3.75, 3.75)............................................................................................ 419

List of Figures

xxiii

Figure 8.16  A light source with a holographic LSD................................................................... 421 Figure 8.17  S  patial distribution of received power: (a) without LSD, with LSD (b) 10°, (c) 20°, and (d) 30°; (i) simulated and (ii) measured................................... 423 Figure 8.18  (a) Predicted normalised power distribution of a seven-cell link using 30° LSDs (b) predicted power contour plot of a seven-cell link using 30° LSDs......... 424 Figure 8.19  Schematic diagram of the LED control mechanism............................................... 425 Figure 8.20  T  he schematic diagram of an OOK intensity modulated VLC system.................. 426 Figure 8.21  Time waveforms for BAM signals for VLED dimming control............................. 427 Figure 8.22  P  WM/PPM block diagrams: (a) digital, (b) analogue waveforms, (c) PWM waveforms, and (d) PWM frequency spectrum....................................... 428 Figure 8.23  P  WM-DMT block diagram: (a) transmitter, (b) receiver, and (c) DMT (OFDM) spectrum.................................................................................... 430 Figure 8.24  P  WM waveform with 80% dimming level and PWM sampled DMT time waveform......................................................................................................... 431 Figure 8.25  M  ultilevel PWM-PMM scheme: (a) system block diagram and (b) timing waveforms.............................................................................................. 433 Figure 8.26  PWM-OOK system: (a) block diagram and (b) waveforms.................................... 434 Figure 8.27  M  easured signal waveforms: (a) received PWM-OOK with 70% duty cycle for PWM, the eye diagram, and (b) recovered NRZ-OOK data signal.................. 435 Figure 8.28  MIMO: (a) imaging and (b) non-imaging............................................................... 436 Figure 8.29  (a) VLC MIMO system and (b) schematic of VLC MIMO model......................... 437 Figure 8.30  Simulated waveforms at the transmitter and receiver for the MIMO..................... 438 Figure 8.31  B  lock diagram for OFDM visible light communication with frequency domain channel estimation..................................................................................... 442 Figure 8.32  O  FDM-VLC link performance: (a) EVM versus the frame index, (b) EVM versus the transmission distance for QPSK, and (c) BER versus the transmission distance for QPSK........................................444 Figure 8.33  Schematic block diagram for the VLC with PLED................................................ 445 Figure 8.34  L  -I-V curves of the PLED under test; shown in dashed lines are the diodes used in [150].................................................................................... 445 Figure 8.35  (a) The system BER and Q-factor performance as a function of data rate; 3 Mbps can be achieved without the use of an equaliser. At 4 Mbps, the link fails and errors are introduced into the system; eye diagrams are shown inset. (b) The SNR measured throughout the system from 20 kHz – 1 MHz using an Agilent N9010A electrical spectrum analyzer. The SNR is smoothed and fitted exponentially to predict the SNR at higher data rates...............................446 Figure 8.36  B  ER performance of the PLED-VLC system with the FPGA-based LMS equaliser; clearly there is an increase in performance with an increasing number of taps as expected. The key result is that the 10 Mbps link has a BER within the FEC limit; meaning that the data can be recovered with an overhead of just 7%............................................................................................ 447

xxiv

List of Figures

Figure 8.37  (a) The raw BER performance in comparison with a deterioration of 1 Mbps is noted due to the longer PRBS sequence used. Eye diagrams at 3 Mbps are shown inset and at (b) 1 Mbps, and (c) 2 Mbps [187]......................448 Figure 8.38  E  qualised BER performance of the link with a training length of 215 and a varying number of neurons [187].........................................................448 Figure 8.39  ANN BER performance of the link with Tlen of 214 and 104; offering performance up to 19 and 18 Mbps, respectively [187]..........................................449 Figure 8.40  Home access network.............................................................................................. 450 Figure 8.41  A  one-dimensional optical wireless cellular system: (a) block diagram and (b) functional block diagram of a BS/UT module............................................ 451 Figure 8.42  BER versus the link margin for a one-dimensional cellular VLC system..................452 Figure 9.1  Point-to-point, ring, and mesh topologies.................................................................. 471 Figure 9.2  Ring topology extension networks............................................................................ 471 Figure 9.3  Mesh FSO topology network at different heights...................................................... 472 Figure 9.4  Illustration of relay-assisted FSO communications................................................... 473 Figure 9.5  (a) Serial relay transmission and (b) parallel relay transmission............................... 474 Figure 9.6  Basic configuration of AOAF relay-based FSO link................................................. 476 Figure 9.7  An erbium-doped fibre amplifier............................................................................... 478 Figure 9.8  Spectral broadening of a pulse due to SPM-induced temporal variations................480 Figure 9.9  A  signal regeneration scheme illustrating the re-amplification, re-shaping, and re-timing operations......................................................................... 481 Figure 9.10  Schematic diagram of an SPM-based 2R regenerator............................................. 482 Figure 9.11  S  erial AOAF relay-based FSO link: (a) schematic block diagram and (b) relay configuration....................................................................................... 483 Figure 9.12  E  xperimental (exp) and theoretical (theo) result for BER versus SNR for a single-hop (S), dual-hop (D), and triple-hop (T) with no turbulence........................................................................................................... 485 Figure 9.13  M  easured eye diagrams of received FSO system for (a) triple-hop, (b) dual-hop, and (c) single link for BER values of 10 −6, 10 −4, and 10 −2, respectively, over Cn2 = 3.8 × 10 −10 m−2/3................................................... 486

List of Tables Table 1.1  P  ossible Options to Overcome the Spectrum Bottleneck Experienced in Current RF Wireless...................................................................................................3 Table 1.2  Access Technologies for the Last Mile Link...................................................................5 Table 1.3  Current and Future Wireless Technologies..................................................................... 5 Table 1.4  A Brief History of OWC Systems................................................................................. 12 Table 1.5  Properties of RF and OWC Technologies..................................................................... 13 Table 1.6  Features of Directed and Diffuse Links....................................................................... 19 Table 1.7  Classification of Lasers According to the IEC 60825-1 Standard................................ 27 Table 1.8  A  ccessible Emission Limits for Two Wavelengths, 850 nm and 1550 nm......................................................................................................................... 27 Table 1.9  E  xample of MPE Values (W/m 2 ) of the Eye (Cornea) at 850 nm and 1550 nm Wavelengths [78].....................................................................................28 Table 1.10  Challenges in OWC Systems....................................................................................... 29 Table 2.1  Common LED Materials and Their Optical Radiation Wavelengths........................... 45 Table 2.2  Laser Diode Materials and Their Corresponding Radiation Wavelengths................... 58 Table 2.3  Key Features of LDs for Multi-Mode and Single-Mode............................................... 62 Table 2.4  A Comparison of an LED and a Semiconductor Laser Diode...................................... 62 Table 2.5  Operating Wavelength Ranges for Different PD Materials [17]...................................66 Table 2.6  Typical Performance Characteristics of PDs................................................................ 67 Table 2.7  Comparison between IM-DD and Coherent Schemes.................................................. 68 Table 2.8  Dark-Current Values for Different Materials [17]........................................................ 73 Table 2.9  Statistical Parameters of Poisson Random Distribution [20]........................................ 77 Table 3.1  Simulation Parameters..................................................................................................90 Table 3.2  Artificial Ambient Light Sources.................................................................................. 98 Table 3.3  I b and Ipk-pk for an Incandescent Bulb With and Without Optical Filtering............................................................................................................99 Table 3.4  I b and Ipk-pk for a Low-Frequency Fluorescent Lamp With and Without Optical Filtering............................................................................................ 100 Table 3.5  I b and Ipk-pk for a High-Frequency Fluorescent Lamp With and Without Optical Filtering............................................................................................ 101 Table 3.6  Low-Frequency Component Phase Values.................................................................. 102 Table 3.7  High-Frequency Component Amplitude and Phase Values........................................ 103 Table 3.8  T  ypical Atmospheric Scattering Particles with Their Radii and Scattering Process at λ = 850 nm................................................................................ 106 xxv

xxvi

List of Tables

Table 3.9  I nternational Visibility Range and Attenuation Coefficient in the Visible Waveband for Various Weather Conditions [67]......................................................... 109 Table 3.10  Particle Size Distribution for Advection and Maritime Fog [61], [69]...................... 110 Table 3.11  Weather Conditions and Their Visibility Range Values........................................... 112 Table 3.12  Link Budget for FSO Link........................................................................................ 121 Table 3.13  Test Bed System Parameters..................................................................................... 138 Table 3.14  System Characteristics for λ = 830 nm and Ptx = −1.23 dBm................................... 139 Table 3.15  Attenuation of 550 nm Laser at Given Values of V for L = 5.5 m and 11 m..................142 Table 3.16  Measured V and T Values at 0.550 μm...................................................................... 145 Table 4.1  Comparisons of Different Modulation Schemes......................................................... 162 Table 4.2  M  apping of M-Bit OOK Data Format into PPM, DPIM, DPPM, and DAPPM Symbols................................................................................................. 171 Table 5.1  T  he NOPR to Achieve a BER of 10 −6 at Rb of 200 Mbps for DWT Based Rx with Different Mother Wavelets [46].......................................................... 258 Table 5.2  The List of Training Algorithms for ANN [77].......................................................... 287 Table 5.3  ANN Parameters for Equalisations............................................................................. 287 Table 6.1  Values of K1 and K0 for Different Noise-Limiting Conditions.................................... 321 Table 6.2  Simulation Parameters................................................................................................ 323 Table 6.3  Fading Strength Parameters for Gamma-Gamma Turbulence Model........................ 332 Table 6.4  Comparison of Modulation Techniques...................................................................... 338 Table B.1  Zeros and Weights of Gauss-Hermite Integration with n = 20...................................344 Table 7.1  D  iversity Gain at a BER of 10 −6 in Gamma-Gamma and Negative Exponential Channels................................................................................... 371 Table 7.2  Numerical Simulation Parameters............................................................................... 375 Table 7.3  Gain per Photodetector at a BER of 10 −6.................................................................... 376 Table 7.4  Simulation Parameters................................................................................................. 381  ading Penalty and STDD Gain at BER = 10 −9, Rb = 155 Mbps, σl2 = 0.3, Table 7.5  F and ξ = 1/(U + 1)A....................................................................................................... 381 Table 8.1  Historical Development Timeline for VLC................................................................. 401 Table 8.2  C  omparison of VLC, IR, and RF Communication Technologies and Short-Range Wireless Technologies and Standards....................................................404 Table 8.3  Short-Range Wireless Technologies and Standards....................................................404 Table 8.4  System Parameters for a VLC Link............................................................................ 412 Table 8.5  VOOK Code Word...................................................................................................... 427 Table 8.6  Simulated VLC MIMO Systems.................................................................................440 Table 8.7  Technologies for Home Access Networks...................................................................449 Table 8.8  Link Budget for a One-Dimensional VLC Cellular System....................................... 452

Abbreviations 3D 3G 4G 5G

Three Dimensional Third Generation Fourth Generation Fifth Generation

ACO-OFDM ADC ADSL AEL AF AFC AIM Air AlGaAs ALI AM ANN ANSI AOA AORE APD ASE ASK AWG AWGN BAM BER BLL BLW BP BPL BPoLSK BPSK BS CAP CC CCDF CDMA CDRH CENELEC CF CM CMRR CNR CPC CPoLSK CQD

Asymmetrically Clipped Optical-OFDM Analogue to Digital Converter Asymmetric Digital Subscriber Line Accessible Emission Limits Amplifier and Forward Automatic Frequency Control Analogue Intensity Modulation Advanced Infrared Aluminium Gallium Arsenide Ambient Light Interference Amplitude Modulation Artificial Neural Network American National Standards Institute Angle of Arrival All-optical Regenerate-and-Forward Avalanche Photodiode Detector Amplified Spontaneous Emission Amplitude Shift Keying Arbitrary Waveform Generator Additive White Gaussian Noise Bit Angle Modulation Bit Error Rate Beer-Lambert Law Baseline Wander Backpropagation Broadband over Power Line Binary Polarisationisation Shift Keying Binary Phase-Shift Keying Base station Carrier-less Amplitude and Phase Modulation Convolutional Codes Complementary Cumulative Distribution Function Code Division Multiple Access Center for Devices and Radiological Health European Committee for Electrotechnical Standardisation Compressed Forward Colour Mixed Common Mode Rejection Ratio Carrier to Noise Ratio Compound Parabolic Concentrators Circular Polarisation Shift Keying Colloidal Quantum dot xxvii

xxviii

CSI CWT DAB DAC DAPPM DC DCO-OFDM DCPoLSK DD DF DFB DFE DFT DH-PIM DMT DPIM DPLL DPPM DPSK DSL DSSS DTRIC DTV DVB DVB-S2 DWDM DWT EDFA EGC EO EPM ERC ERP eV EVM FC FCC FDM FEC FFT FIR FLI FOV FPGA FPL FPLD FRC FSK FSO FWHM GaAs

Abbreviations

Channel State Iinformation Continuous Wavelet Transform Digital Audio Broadcasting Digital to Analogue Converter Differential Amplitude Pulse Position Modulation Direct Current DC-offset Optical-OFDM Differential Circle Polarisation Shift Keying Direct Detection Decode-and-forward Distributed Feedback Laser Decision Feedback Equaliser Discrete Fourier Transformation Dual-Header Pulse Interval Discrete Multitone Modulation Digital Pulse Internal Modulation Digital Phase Locked Loop Differential Pulse Position Modulation Differential Phase Shift Keying Digital Subscriber Line Direct Sequence Spread Spectrum Dielectric Totally Internally Reflecting Concentrator Digital Television Digital Video Broadcast Digital Video Broadcast Satellite Standard Dense Wavelength Division Multiplexing Discrete Wavelet Transform Erbium Doped Fibre Amplifiers Equal Gain Combining Electrical to Optical Edge Position Modulation Equal Ratio Combiner Emitter Radiation Pattern Electron Volt Error Vector Magnitude Fountain Code Federal Communications Commission Frequency Division Multiplexing Forward Error Correction Fast Fourier Transform Finite Impulse Response Fluorescent Light Interference Field of View Field Programmable Gate Array Fabry-Perot Laser Fabry-Perot Laser Diode Fixed Ratio Combiner Frequency Shift Keying Free Space Optics Full Width Half Maximum Gallium Arsenide

xxix

Abbreviations

GaAsP GaN Ge GPS GS HDD HDTV HNLF HON HPF H-V IEC IF IFFT IID IM IMD IMP InGaN InP IO IoT IPS IR IRC IrDA IS ISC ISFA ISI ISM LAN LCD LD LDPC LED LIA LiNbO3 LMS LO LOS LPF LS LSD m-CAP M2M MCM MF MIMO MISO ML

Gallium Arsenide Phosphide Gallium Nitride Germanium Global Positioning System Guard Slot Hard Decision Decoding High-Definition TV Highly Nonlinear Fibre Heterogeneous Optical Networking High-Pass Filter Hufnagel-Valley model International Electrotechnical Commission Intermediate Frequency Inverse Fast Fourier Transform Independently Identically Distributed Intensity Modulation Inter-modulation Distortion Intermodulation Product Indium Gallium Nitride Indium phosphide Input-Output Internet of Things Indoor Positioning System Infrared Infrared Communications Infrared Data Association Image Sensor Inverse Source Coding Intrasymbol Frequency-domain Averaging Intersymbol Interference International Scientific and Medical Local Area Networks Liquid Crystal Display Laser Diode Low-Density Parity Check Light Emitting Diode Laser Institute of America Lithium Niobate Least Mean Squares Local Oscillator Line of Sight Low-Pass Filter Least Square Light Shape Diffuser Multiband-CAP Machine-2-Machine Multicarrier Modulations Matched Filter Multiple-Input, Multiple-Output Multiple-Input Single-Output Maximum Likelihood

xxx

MLCD MLD MLP MLSD MMF MMSE MMW MPE MPPM MRC MS MSC MSD MSIM MSLD MVR MZI MZM NA NEC NEP NGB NIR NLOS NOPR NRZ OA OBPF OC OE OFDM OIM OLED OLO OMIMO OOK OPLL OPP OPPM OSHA OSNR OSTBC OWC P2P PA-ACO- OFDM PaDF PAM PAPR PBS PC

Abbreviations

Mars Laser Communication Demonstration Maximum Likelihood Decoding Multilayer Perceptron Maximum Likelihood Sequence Detection (MLSD) Multimode Fibre Minimum Mean Square Error Millimetre Wave Maximum Permissible Exposures Multiple Pulse Position Modulation Maximum Ratio Combiner Mobile Station Mobile Switching Centres Multiple Spot Diffusing Multiple Subcarrier Intensity Modulation Maximum Likelihood Sequence Detection Meteorological Visual Range Mach-Zehnder Interferometry Mach-Zehnder Modulator Numerical Aperture Nippon Electric Company Noise-Equivalent Power No Guard Band Near Infrared Non-Line of Sight Normalised Optical Power Requirement Non-return to Zero Optical Amplifier Optical Band-Pass Filter Optimal Combining Optical-to-Electrical Orthogonal Frequency Division Multiplexing Optical Impulse Modulation Organic Light-Emitting Diode Optical Local Oscillator Optical Multiple-Input Multiple-Output On-Off Keying Optical Phase Locked Loop Optical Power Penalty Overlapping Pulse Position Modulation Occupational Safety and Health Administration Optical Signal-to-Noise Ratio Orthogonal Space Time Block Code Optical Wireless Communications Point-to-Point Pilot Aided Asymmetrically Clipped Optical OFDM Partial decode-forward Pulse Amplitude Modulation Peak-to-Average-Power Ratio Polarisation Beam Splitter Phosphor Coated

xxxi

Abbreviations

PD PDAs PDF PDOA PER PIM PLC PMOPR POA POF PoLSK PON PPM PPM+ PRBS PSD PSK PTM PWM QAM QCL QPSK RC RCLED RF RGB RIN RLL RMS RNN ROA RoF ROF RQC RRC RS RSS RSSI Rx RZ S.I SBA SCM SC-PPM SDD SDM SEC SelC SER Si SiC

Photodetector Personal Digital Assistants Probability Density Function Phase Difference of Arrival Packet Error Rate Pulse Interval Modulation Power Line Communications Peak-to-Mean Optical Power Ratio Phase of Arrival Plastic Optical Fibre Optical Polarisation Shift Keying Passive Optical network Pulse Position Modulation Pulse Position Modulation Plus Pseudorandom Binary Signal Power Spectral Density Phase Shift Keying Pulse Time Modulation Pulse Width Modulation Quadrature Amplitude Modulation Quantum Cascaded Laser Quadrature Phase Shift Keying Rateless Codes Resonant Cavity LED Radio Frequency Red-Green-Blue Laser Relative Intensity Noise Run Length Limited Root Mean-Square Recurrent Neural Network Real Outdoor Atmosphere Radio over Fibre Real Outdoor Fog RaptorQ code Root Raised Cosine Reed Solomon Received Signal Strength Received Signal Strength Indicator Receiver Return to Zero Scintillation Index Stimulated Brillouin Amplifier Subcarrier Multiplexing Subcarrier Pulse Position Modulation Soft Decision Decoding Space Division Multiplexing Switch-and-Examine Combining Selection Combining Slot Error Rate Silicon Silicon Carbide

xxxii

SILEX SIM SIMO SINR SIR SISO SLD SLuD SMSR SNR SNRop SOP SPA SPM SRA SSC SSD SSL STBC STDD STFT SWFM TCM TDL TDM TDOA TL TOA TPC Tx UT UV UWB VCSEL VFIr VL VLC VLCC VOOK VSG WCS WDM WiMax WMAN WP WPAN WPLED WPT XGM ZFE ZnSe

Abbreviations

Semiconductor-laser Inter-satellite Link Experiment Subcarrier Intensity Modulation Single-Input Multiple-Output Signal-to-Interference and Noise Ratio Signal-to-Interference Ratio Single-Input Single-Output Semiconductor Laser Diode Superluminescent Diode Side Mode Suppression Ratio Signal-to-Noise Ratio Optical Signal-to-Noise Ratio State of Polarisation Semiconductor Optical Amplifier Self-Phase Modulation Stimulated Raman Amplifier Switch-and-Stay Combining Soft Decision Decoding Solid State Lighting Space-time Block Code Subcarrier Time Delay Diversity Short-Term Fourier Transform Square Wave Frequency Modulation Trellis Coded Modulation Time Delay Line Time Division Multiplexing Time Difference of Arrival Training Length Time of Arrival Turbo Product Code Transmitter User Terminal Ultraviolet Ultra-Wideband Vertical Cavity Surface Emitting Laser Very Fast Infrared Visible Light Visible Light Communications Visible Light Communications Consortium Variable OOK Vector Signal Generator Wireless Communication Systems Wavelength Division Multiplexing Worldwide Interoperability for Microwave Access Wireless Metropolitan Area Network White Phosphor Wireless Personal Area Networks White Phosphor LED Wavelet Packet Transform Cross Gain Modulation Zero forcing Equaliser Zinc Selenide (ZnSe)

Preface In recent years, we have been witnessing a growing demand by users for bandwidth to support broadband wireless services, such as high-definition TV, mobile video phones, video conferencing, and high-speed internet access. With the widespread use of smart devices, as well as the rapid growth in the next generation of internet-of-things (IoT) applications, the quest for sufficient bandwidth is rapidly growing. In the access networks (i.e., last meter and last mile), there are a number of technologies that can address end-users’ communications needs, including copper wire, hybrid coaxial, and optical fibre cables, fibre-to-the-home, and a range of radio frequency (RF)-based wireless communications. However, as the global demand for bandwidth continues to accelerate, it is becoming exceedingly clear that copper/coaxial cables and RF cellular/microwave technologies cannot meet the upcoming needs because of with their limited bandwidth, highly regulated and congested spectrum, and limited accessibility to all. In addition, these technologies require costly licensing fees, suffer from security issues, and incur a high cost of installation. In some countries, the network operators have been deploying new optical fibre-based access networks in order to increase the bandwidth available to their customers. Although it is often thought that optical fibre-based networks offer unlimited bandwidth, in reality, the architectural choices available, compatibility of devices and components, performance constraints of networking equipment, and deployment of the complete system result in limited capacity being offered to end users. Meanwhile, with regard to ubiquitous connectivity between the people and devices at a global level, we have seen a remarkable development in wireless communications technology, which can be naturally extended to provide communications between heterogeneous objects, thus enabling widespread implementation of IoT (i.e., people-to-people, machine-to-machine, terminal-to-terminal, and people-to-machine communications). However, using RF-based wireless technologies, we have (i) spectrum congestion, mostly evident in urban areas, which will lead to limited access to the network; (ii) multipath-induced fading and dispersion, which will affect link performance, especially in highly dense areas; and (iii) insufficient bandwidth for efficient operation of heterogenous devices in indoor environments, considering that more 70% of the wireless data traffic generated is indoor. To reduce the pressure on the RF spectrum, some mobile data traffic can be off-loaded to wireless fidelity (Wi-Fi) and femtocell-based technologies. However, dense deployment of Wi-Fi hotspots is also facing the bandwidth bottleneck. Therefore, to ensure seamless wireless communication with high data rates and low latency, the service provided will have to adopt highly reliable, low-cost and high-speed technologies. The emerging fifth generation (5G) and beyond-5G RF-based wireless technologies are expected to address these issues in the coming years. However, in 5G and beyond-5G wireless networks, there will be additional challenges, such as inter-cell/inter-tier interference, management of spectral resources reuse, etc. Alternatively, optical wireless communications (OWC), which is an innovative complementary technology to the RF wireless systems and has been around for the last three decades, can be adapted to provide high capacity with reduced latency and at a low cost in certain indoor and outdoor applications. It offers flexible and scalable wireless networking solutions that are costeffective, high security at the physical layer, high-speed, license-free, low power usage, immune to RF-based electromagnetic interference, and unconstrained frequency reuse due to a high degree of spatial confinement and ease of deployment for a number of applications, including voice and data, video and entertainment, enterprise connectivity, remote sensing, medical and manufacturing, disaster recovery, illumination and data communications, surveillance, localisation, and many others. In OWC technology, three bands of ultraviolet (UV), visible light (VL), and infrared (IR) can be used to provide the high bandwidth for communication purposes. Due to the unique properties of the optical signal, in an indoor environment, one can precisely define an optical footprint and hence xxxiii

xxxiv

Preface

can accommodate a number of devices within a small area. Thus, optical wireless communications, also referred to as free space optical communication systems for outdoor applications, will play a significant role as a complementary technology to the RF systems in future information superhighways, as well as being able to comply with the 5G Infrastructure Public-Private Partnershipidentified key performance indicators. Having seen the development of OWC systems in the last two decades, we felt there was a need for the second edition of this textbook with additional materials on the technology that was concise and suitable for undergraduate and graduate level courses, as well as researchers and professional engineers working in the field. This book broadly covers five important aspects of OWC systems: (a) the fundamental principles of OWC; (b) devices and systems; (c) modulation techniques; (d) channel models and system performance analysis; and (e) free space optics and visible light communications. In addition, the book covers different challenges encountered in OWC, as well as outlining possible solutions and current research trends. The major attractions of this book are (i) the Matlab simulations and the inclusion of Matlab codes and (ii) experimental test beds to help readers understand the concepts and enable them to carry out extensive simulations, implement OWC links, and evaluate their performance. The book is structured into nine self-contained chapters. To facilitate a logical progression of materials presented and to enable readers to better understand the topics and follow them through, each chapter starts with an introduction followed by background information supported by detailed theoretical analysis, as well as up-to-date supporting references. Any additional supporting materials are included in the end-of-chapter appendices. Starting with a bit of history, Chapter 1 presents an up-to-date review of OWC systems for indoor and outdoor applications, the present state of play, and the future directions for the emerging OWC technology. The wireless access technologies, benefits and limitations, link configurations, eye safety, application areas, and challenges of OWC systems are all covered in Chapter 1. There are a number of light sources and photodetectors (PIN and/or avalanche photodiodes) that could be used for OWC systems. Light-emitting diodes (LEDs) and low-power laser diodes are mainly employed in short-range indoor applications. For long-range outdoor applications, laser diodes are mostly used. Chapter 2 discusses the types of light sources and optical detectors, their structures and optical characteristics, as well as the process of optical detection. Different types of noise encountered in optical detection and the statistics of the optical detection process are also discussed in Chapter 2. To design efficient optical communication systems, it is imperative that the characteristics of the channel are well understood. Characterisation of a communication channel is performed by its channel impulse response, which is then used to analyse and proffer solutions to the effects of channel distortions. A number of propagation models (ceiling bounce, Hayasaka-Ito, and spherical) for a line-of-sight and non-line-of-sight indoor applications are studied in Chapter 3. The artificial light interference that affects the indoor OWC link performance is also outlined in this chapter. As for the outdoor free space optics (FSO) links, the atmospheric channel is a very complex and dynamic environment, which can affect the characteristics of the propagating optical beam, thus resulting in optical losses and turbulence-induced amplitude and phase fluctuations. There are a number of models to characterise the statistical nature of the atmospheric channel, mostly fog and turbulence, which are treated in Chapter 3. A practical test bed for investigating the atmospheric effect on the FSO link, measured data sets and calibration of the indoor data to the outdoor links are also presented in Chapter 3. Most practical OWC systems currently in use are based on the intensity modulation/direct detection scheme. For the outdoor environment, atmospheric conditions, in particular heavy fog, are the major problem, as the intensity of the light beam reduces considerably under thick fog. Increasing the level of transmitted power is one option to improve the link availability. However, eye safety regulations limit the amount of transmitted optical power. For indoor applications, the eye safety limit on transmitted optical power is even more stringent. In Chapter 4, a number of modulation

BK-TandF-9781498742696_TEXT_GHASSEMLOOY-181209-FM.indd 34

16/04/19 10:29 AM

Preface

xxxv

techniques, which are the most popular, in terms of power efficiency, bandwidth efficiency, for both indoor and outdoor OWC applications are discussed. The emphasis is more on the digital modulation techniques including pulse position modulation (PPM), on-off keying (OOK), digital pulse interval modulation (DPIM), etc. The spectral properties, error probability, and the power and bandwidth requirements of a number of modulation schemes are also presented. Advanced modulation techniques, such as the subcarrier intensity modulation, and multi-carrier modulations, such as orthogonal frequency division multiplexing, carrier-less amplitude and phase modulation, and polarisation shift keying, are also covered in this chapter. In indoor scenarios, additional periodic and deterministic forms of noise that degrade system performance is due to the presence of background artificial light sources. The diffuse indoor links suffer from the multipath-induced intersymbol interference, thus limiting the maximum achievable data rates. The performance of the OOK-, PPM-, and DPIM-based systems in the presence of the artificial light interference and intersymbol interference is investigated in Chapter 5. To improve the link performance, possible mitigation techniques using high-pass filtering, equalisation, wavelet transform, and the neural network are also outlined in this chapter. Atmospheric turbulence is known to cause signal fading in the channel. Chapter 6 outlines the outdoor FSO link performance in terms of the bit error rate and the outage probability under atmospheric turbulence for a range of modulation schemes of OOK, PPM, PSK, QPSK, DPSK, etc. A range of channel models are also considered. Primary challenges attributed to outdoor OWC (i.e., FSO) communications are building sway, scattering/absorption-induced attenuation, and scintillation-induced link fading. To address building sway and therefore reduce pointing errors, accurate pointing and tracking mechanisms, multi-array transmitter and receiver, and/or wide beam profiles could be adopted. In FSO links, phase and irradiance fluctuations experienced by the propagating optical beam make optical coherent detection less attractive, simply because it is sensitive to both signal amplitude and phase fluctuations—thus the reasons for adopting the direct detection scheme in terrestrial FSO links. The available options for mitigating the effect of channel fading in FSO links include but are not limited to increased transmit power, diversity (i.e., frequency, spatial, temporal, and polarisation) schemes, including the multiple-input-multiple-output (MIMO), hybrid RF-FSO links, aperture averaging, adaptive optics, and subcarrier time diversity, which are covered in Chapter 7. The link performance using equal gain combining, optimal combining, or equivalently maximum ratio combining and selection combining diversity schemes for log-normal and gammagamma atmospheric channels employing a range of diversity techniques is also outlined in Chapter 7. Chapter 8 is dedicated to VLC, a subject that in the last few years has witnessed an increased level of research activities. Put simply, VLC is the idea of using visible optical sources for illumination, wireless data communications, indoor localisation, and sensing. The main drivers for the VLC technology include the increasing popularity of solid state and organic-based lights and the longer lifetime of high brightness LEDs compared to other existing light sources. The multiple functionalities offered by LEDs has created a whole range of interesting applications, including home networking, car-to-car communication, high-speed communications in aeroplane cabins, in-train data communications, traffic light management, and medical and manufacturing communications, to name a few. The levels of power efficiency and reliability offered by LEDs are superior compared to traditional incandescent light sources. This chapter gives an overview of the VLC communication technology, highlighting the fundamental theoretical background, devices available, modulation and dimming techniques, and system performance analysis. Multiple-input, multiple-output and cellular visible light communication systems are also covered in Chapter 8. As outlined in Chapters 3, 6, and 7, FSO link performance is hampered by the atmospheric channel conditions, thus affecting the link range and availability. However, in order to increase the link availability and the linkspan, the data can be transmitted via relays to end users, which offers many unique advantages and also presents a number of fundamental challenges. Chapter 9 presents an extensive review and discussion on the key aspects of FSO technology with relay focusing on optical networks and their topology, serial and parallel relaying all optical

xxxvi

Preface

relay FSO links. The relay-based FSO link performance under turbulence is also investigated both theoretically and experimentally in Chapter 9. The relevant and necessary Matlab codes are given in each chapter to enable the reader to carry out extensive simulations in order to better understand the topics. Note that it is expected that the readers first study the theory and then use the Matlab codes for their needs. Recent, relevant, and up-to-date references, which provide a guide for further reading, are also included at the end of each chapter. A complete list of common abbreviations used in the text is also provided. Throughout the book, SI units are used. We would like to thank all the authors of all journals and conference papers, articles, and books we consulted in writing this second edition. Special thanks to those authors, publishers, and companies who kindly granted permission for the reproduction of their figures. We would also like to extend our gratitude to all our past and current PhD students for their immense contributions to the knowledge in OWC. Their contributions have enriched the content of this book. Finally, we remain extremely grateful to our families and friends who have continued to be supportive and have provided needed encouragement. In particular, our very special thanks go to Azar, Odunayo, and Kanchan for their continuous patience and unconditional support, which has enabled us to finally complete this challenging task. Their support has been fantastic.

MATLAB® is a registered trademark of The MathWorks, Inc. For product information, please contact: The MathWorks, Inc. 3 Apple Hill Drive Natick, MA 01760-2098, USA Tel: 508-647-7000 Fax: 508-647-7001 E-mail: [email protected] Web: www.mathworks.com

BK-TandF-9781498742696_TEXT_GHASSEMLOOY-181209-FM.indd 36

16/04/19 2:11 PM

About the Authors Professor Zabih Ghassemlooy, CEng, Fellow of IET, Senior Member of IEEE and OSA received his BSc (Hons.) in electrical and electronics engineering from Manchester Metropolitan University, UK, in 1981, and his MSc (1984) and PhD (1987) from the University of Manchester, UK. From 1987 to 1988, he was a Post-Doctoral Research Fellow at City University, UK. In 1988, he joined Sheffield Hallam University as a lecturer, becoming a professor in 1997. In 2004, he joined the University of Northumbria, Newcastle, as an associate dean (AD) for research in the School of Engineering, and from 2012 to 2014 he was AD for Research and Innovation, Faculty of Engineering and Environment. He currently is head of the Optical Communications Research Group. In 2001, he was awarded the Tan Chin Tuan Fellowship in Engineering from Nanyang Technological University, Singapore. In 2016, he was a research fellow and in 2015 a distinguished professor at the Chinese Academy of Science, Quanzhou, China. He became a visiting professor at the University Tun Hussein Onn, Malaysia (2013–2017), and Huaqiao University, China (2017–2018). He has published over 785 papers (309 journals and six books), given more than 92 keynote/invited talks, and supervised over 60 PhDs. His research interests include optical wireless communications, free space optics, visible light communications, radio-over-fibre/free space optics, and sensor networks with project funding from the EU, UK Research Council, and industry. He was the vice-chair of EU Cost Action IC1101 (2011–2016). He is the chief editor of the British Journal of Applied Science and Technology and the International Journal of Optics and Applications, associate editor of a number of international journals, and co-guest editor of a number of special issues. He is a fellow of the IET, a senior member of IEEE, a senior member of OSA, and a chartered engineer. He is a co-editor of four books, including Optical Wireless Communications—An Emerging Technology (Springer, 2016), Visible Light Communications: Theory and Applications (CRC, 2017), Intelligent Systems for Optical Networks Design: Advancing Techniques (IGI Global, 2013), and Analogue Optical Fibre Communications, IEE Telecommunication series 32 (IET, 1995). He is the founder and chair/co-chair of a number of international events, including the IEEE/IET International Symposium on Communications Systems, Networks and DSP, West Asian Colloquium on Optical Wireless Communications, and Workshop on Optical Wireless Communications in ICC since 2015. He is the vice-chair of the OSA Technical Group of Optics in Digital Systems (2018–). From 2004–2006, he was the IEEE UK/IR Communications chapter secretary, the vice-chairman (2004–2008), chairman (2008–2011), and chairman of the IET Northumbria Network (October 2011–2015). Website: http://soe.northumbria.ac.uk/ocr/people/ghassemlooy/ Dr Wasiu O. Popoola h olds a first class (Hons.) degree in electronics and electrical engineering from Obafemi Awolowo University, Nigeria, and an MSc and a PhD degree, both from Northumbria University in Newcastle upon Tyne, UK. During his PhD studies, he was awarded the Xcel Best Engineering and Technology Student of the year 2009. He is currently a Chancellor’s Fellow in the Institute for Digital Communications and LiFi R & D Centre, School of Engineering at the University of Edinburgh, UK. He has published over 100 journal articles, conference papers, and patents, more than seven of which are invited papers. One of his journal articles ranked number 2 in terms of the number of full-text downloads within IEEE Xplore in 2008 from the hundreds of papers published by IET Optoelectronics since 1980. Another paper he co-authored with one of his PhD xxxvii

xxxviii

About the Authors

students won the best poster award at the 2016 IEEE ICSAE Conference. He also co-authored the book Optical Wireless Communications: System and Channel Modelling with MATLAB, published by CRC in 2012 and several other book chapters (one with over 10,000 downloads as of September 2014 since its publication in 2010). Popoola is a senior member of the IEEE, an associate editor of the IEEE Access Journal, guest editor for Elsevier Journal of Optik (special issue on optical wireless communications) in 2017, and a technical program committee member for several conferences. He has been an invited speaker at various events, including the 2016 IEEE Photonics Society Summer Topicals. His primary research interest is in optical communications, including visible light communications, free space optical communication, and fibre communication. Personal webpage: https://www.optical-communications.eng.ed.ac.uk/ Dr. Sujan Rajbhandari (SMIEEE)  is a senior lecturer at Coventry University, where he is working in the field of optical wireless communication. Dr. S. Rajbhandari obtained an MSc in Optoelectronic and Communication Systems with Distinction in 2006 and was awarded the P. O. Byrne prize for the most innovative project. He then joined the Optical Communications Research Lab (OCRG) at Northumbria University as a PhD candidate and was awarded a PhD in 2010. He was with the OCRG at Northumbria University, working as a senior research assistant and research fellow from 2009 until 2012. He joined the communications research group at the University of Oxford in 2012 and worked in the EPSRC’s prestigious Ultra-Parallel Visible Light Communications (UP-VLC) project, which was a collaboration of five of the UK’s leading universities (Oxford, Cambridge, St. Andrews, Edinburgh, and Strathclyde). In 2015, Dr. Rajbhandari joined Coventry University as a lecturer in electrical and electronic engineering and was promoted to senior lecturer in 2017. Dr. Rajbhandari is an active researcher with an international reputation as a leading expert in the field of optical wireless communication. He has published more than 150 scholarly articles with over 3000 citations in the area of optical wireless communications, visible light communication, signal processing, and artificial intelligence, including the book Optical Wireless Communications: System and Channel Modelling with MATLAB®. His work is highly recognised nationally and internationally. Dr. Rajbhandari was the invited tutorial presenter at the 16th International Symposium on the Science and Technology of Lighting, invited speaker ICTF2015, Manchester Metropolitan University. He is the lead guest editor for MDPI Sensors Journal special issue on Visible Light Communication Networks and guest editor for MDPI photonics special issue on Photonic Communications Systems in Access Networks and academic editor for Wireless Communications and Mobile Computing. Dr. Rajbhandari is a co-recipient of the 2015 Winners of the IET Premium (Best Paper) Awards for Optoelectronics Journal, co-author of invited papers in Photonics Research Journal (2013), author of invited papers in CNSDSP2016, SPIE Optics 2014. He has served as a reviewer for EPSRC grants and journals published by IEEE, OSA, IET, Elsevier, IOP, and international conferences, such as IEEE ICC ‘2017 and CSNDSP2016. He has also acted as proceedings co-editor and local organising committee member of NOC/OC&I 2011 and EFEA2012. His research interests lie in the area of optical communication, visible light communication, optical wireless communications, signal processing, modulation techniques, equalisation, applied artificial intelligence, and wavelet transform. He is a senior member of IEEE and an associate member of the Institute of Physics.

1

Introduction: Optical Wireless Communication Systems

We are seeing a growing demand for bandwidth in mobile communications, as the numbers of new data services and applications are emerging continuously. What we are seeing in recent years is an exponential growth in the internet traffic demand, which is scaling much faster than the prediction of Moore’s law, where the data traffic has been increasing by a factor of ten every five years. The global mobile data traffic grew from 70% in 2012 with 820 petabytes (PB) per month to 81% in 2013 with 1.5 exabytes per month and 69% in 2014 with 2.5 exabytes per month at the end of 2014, 30 times the size of the entire global internet in 2000 [1]. The global mobile devices and connections also increased from 6.9 billion in 2013 to 7.4 billion in 2014 with 88% of the growth mainly due to smartphones. It was expected that there would be up to 3.4 billion internet users and 19 billion global network connections (fixed plus wireless) by 2016. The overall mobile data traffic is expected to grow to 24.3 exabytes per month by 2019, which is a 14-fold increase compared to 2012. As for the mobile network (cellular) connection, speeds for the downstream have risen from 1.4 Mbps in 2013 to 1.7 Mbps in 2014. Figure 1.1 shows the distribution of mobile internet services. By 2020, the communication network and service environment will be markedly richer and more complex than that of today. It is expected that in 2020; everything will be connected to the communication network based on a host of application-specific needs—people, internet of things (IoT), machine-to-machine (M2M), knowledge and information—by a truly flexible, mobile, and powerful means. According to the World Bank, increased broadband penetration can create additional GDP growth of up to 1.38% and 1.21% points in low- and middle-income, and high-income economies, respectively. In addition, the migration of mobile services to applications requiring broadband, together with the move to larger numbers of small microcells in the range of 10 Gbps; a few meters within a room

** * (across huge areas a few km) *** ****

6

Optical Wireless Communications

to other devices operating in the same band. The amended IEEE 802.11ac standard, which supports fast Wi-Fi devices, can provide theoretical physical data rates approaching 7 Gbps. This is often orders of magnitude higher than the capacity of the backhaul link. Therefore, in order to expand the growth of Wi-Fi beyond the infrastructure access points, the use of a peer-to-peer communication model such as the WiGig CERTIFIEDTM program is under development by the Wi-Fi Alliance, and it uses the IEEE 802.11ad standard for the 60 GHz unlicensed band. IEEE 802.11ad offers a high-speed (up to 7 Gbps physical rate) wireless connectivity over a short range (i.e., less than 10 meters) for wireless docking and connection to displays, virtually instantaneous wireless backups, synchronisation, and file transfers between computers. Many vendors already have products supporting IEEE 802.11ad, where many antenna elements are fitted on a small platform to create a high-gain antenna array. Antenna arrays of 10–20 elements and a gain of >12–15 dBi are feasible, and if used at the transmitter (Tx) and receiver (Rx), they can offer a link budget gain of >25 dB. However, the path loss of >20 dB at 60 GHz on the flip side is higher than other Wi-Fi frequency bands of 2.4 GHz and 5 GHz. Thus, the need for using directional antennas. However, the IEEE 802.11ad standard cannot meet the data rate requirements (i.e., >10 Gbps or more) of the next generation display and Input-Output (IO) interfaces [20]. In order to meet the requirement for higher data rates of at least 20 Gbps, new mechanisms need to be developed, thus the formation of the IEEE 802.11ay task group [21]. Multiple-input, multiple-output (MIMO) techniques have been effectively used in modern wireless communication technologies (Wi-Fi, 4G, 5G), offering increased throughput and user capacity, and large performance gains, mainly in terms of link/network capacity and link robustness (through diversity) in a number of applications [22], [23]. The inclusion of MIMO antennas in 802.11n operating in both the 2.4 and 2.5 GHz bands uses the MIMO technology, offering net data rates from 54 to 600 Mbps[24]. 802.11ac, which is an extension of the standard, includes wider channels (80 or 160 MHz compared to 40 MHz) within the 5 GHz band and offers 1 Gbps, matching the Gigabit Ethernet with the option of multi-user MIMO [25]. Though MIMO is highly desirable in non-LOS applications, in LOS propagation mode in typical outdoor and indoor scenarios, it performs poorly due to a high correlation of the received signals at the receive antennas. Therefore, in principle, this could create challenges for using the MIMO technology in millimetre wave, since the line-of-sight (LOS) channels are typical in indoor environments [26]. Note that the LOS signal propagation offers a communication channel with a low path loss and low-frequency selectivity compared to the non-LOS. However, MIMO with highly directive phased antenna arrays can effectively exploit the benefits of both MIMO and LOS propagation [27]–[30]. The following are options that could be adopted for fixed wireless access schemes: i. Worldwide Interoperability for Microwave Access (WiMax)—is based on the IEEE 802.16d standard for fixed broadband wireless access with theoretical data rates up to 120 Mbps over a line-of-sight (LOS) link range of 50 km, but it is significantly reduced with non-LOS. ii. Broadband over Power Line (BPL)—piggybacks data communication signals onto existing power cables, already in place at homes and businesses, therefore there is no need to dig up the environment or erect wireless masts to provide internet connections to computers and other devices. BPL operates between 1.7 and 30 MHz and occasionally up to 80 MHz, and uses have included orthogonal frequency division multiplexing (OFDM) and direct sequence spread spectrum (DSSS) [31]. The latest generation of power line equipment offers theoretical bandwidth speeds of up to 200 and 500 Mbps. iii. Ultra-Wideband (UWB) technology—does not require a licence provided the transmitted power density versus the transmission frequency is met (i.e., lower level of transmitted power to avoid interference), thus offering a solution for the bandwidth, cost, power consumption, and physical size requirements of the next-generation consumer electronic devices over a range less than a few tens of meters. UWB differs substantially from

Optical Wireless Communication Systems

7

conventional narrowband RF, Bluetooth, and 802.11a/g, and uses 3.1–10.6 GHz frequency band to provide higher data rates and mobile broadband services to a large number of end users by using nano and microcell. This will require the development of very high-capacity short-range links connecting base stations to MSC, which in turn could be connected to the main trunk network via optical fibre cables [32]. The introduction of microcells has reduced distances between the base station and user, up to 1 km, has enabled greater data rates and mobile broadband services to be provided to a greater number of subscribers. This improved ability to provide broadband mobile services to large numbers of subscribers will require the development of very high-capacity short-range links to connect the cells back to the mobile switching centre. In terms of the mobility in indoor and outdoor environments, there is absolutely no doubt that RF WCS is the preferred technology. In both environments, the data rates available for the mobile users on the move is limited, which can be addressed by the allocation of additional spectrum. In indoor environments with both fixed and mobile terminals, the circumstances are less apparent. For fixed terminals, 10 Mbps and 100 Mbps Ethernet wired LANs are predominantly used with 1 Gbps and higher data rates emerging in the market. The mobile connectivity is provided via the existing cellular networks. The emergence of portable computing devices, such as laptops and smartphones, has fuelled the demand for mobile connectivity and hence led to the development of RF wireless LANs and wireless personal area networks (WPAN). Wireless LANs and WPANs (unlicensed with a bandwidth of 20 MHz) offer users increased mobility and flexibility compared with traditional wired networks and may be classified as either wireless or ad hoc wireless networks [33]. These networks also allow users to maintain network connectivity whilst roaming anywhere within the coverage area of the network. These configurations require the use of access points, or BSs, which are connected to the wired LAN and act as the interface to the wireless devices. In contrast, ad hoc wireless LANs are simple peer-to-peer networks in which each client only has access to the resources of the other clients on the network and not a central server. Ad hoc wireless LANs require no administration or preconfiguration and are created on demand only for as long as the network is needed. The term “wireless” is synonymous with the radio, and there are numerous radio LAN products on the market today. The RF wireless LANs available at present use the unregulated “free” spectrum regions, offering 1–2 Mbps at 2.4 GHz international scientific and medical (ISM). However, the available bandwidth is limited to around 80–90 MHz, and therefore must be shared with numerous other products on the market, such as cordless telephones and baby monitors. The next generation of radio LAN products based on the IEEE 802.11 standards operate in the 2.4 and 2.5 GHz bands offering 100 Mbps—comparable to wired Fast Ethernet—have been allocated exclusively for use by wireless LAN products. Consequently, this allows systems to be optimised in terms of data rate and efficiency, and to be free from the constraints associated with coexisting with other products. There are currently two competing standards in this band, IEEE 802.11a and HiperLAN2, both of which specify maximum data rates of 54 Mbps [34], [35]. As the popularity of wireless LANs and WPAN increases to carry enormous amounts of data for multimedia and other services within the limited frequency spectrum, they will face spectrum congestion, leading to service degradation and reduced viability of the technology in the long run. A possible solution proposed by the Wireless Gigabit Alliance would be to move up the frequency spectrum, exploiting the unlicensed 60 GHz band. It has a system bandwidth of 7 GHz, offering almost 7 Gbps data rate to ensure that a full range of interoperable wireless gigabit solutions reach the end users. The 60 GHz band is also being considered within the IEEE 802.11ad-framework [36]. However, at 60 GHz, the path loss is very high; therefore, these are not suitable as an alternative technology to conventional wireless LANs. They are more attractive for very short-range indoor applications with very high data rates. Figure 1.2 shows power per user for the core, metro edge, fixed access, and wireless access. The fixed access network power usage is expected to remain flat for the next 10 years. Wireless RF

8

Optical Wireless Communications

FIGURE 1.2  Power per user for different access technologies.

access power could grow by a factor of 100 in the next 10 years. By 2020, it is predicted that the wireless RF access power consumption will dominate the global network. Optical fibre with an enormous transmission capacity of about 4 Tbps at the wavelength of 1.55 μm together with the erbium-doped fibre amplifiers (EDFA) can provide the best option for short- to long-distance communication links. However, the deployment cost is still high for the “last mile” and rural areas. A hybrid millimetre waves and optical fibre could be one possible option to achieve the requirements of broadband wireless systems. Therefore, an integration of optical and RF wireless network is an excellent cost-effective option for transmitting a wide range of wideband signals. Radio-over-fibre system (also known as millimetre-wave wireless over optical fibre) plays an important role in wireless communications due to the tremendous growth of end users that has led to higher bandwidth and higher data transfer requirements. Furthermore, RoF system is an alternative solution when the wireless communication systems operate on small cell size, where the utilised carrier frequencies are not adequate to be transported over coaxial cables or when the radio coverage experiences dead zones [37]. Both of these requirements can be fulfilled by utilising the backbone optical fibre communication networks that cross the globe offering bandwidth in excess of 50 THz [38], thus offering data rates in excess of 500 Mbps to both fixed and mobile users [39]–[41]. The RoF system is also known for extending the radio coverage of a central station (CS) for wireless applications [42]. This scenario could be achieved by extending the transmission link between the CS and the BS to bring the access network closer to every mobile user by deploying optical fibre as the medium [37]. However, the overall offered transmission capacity is still limited due to the low carrier frequency. Along with radio, the term “wireless” is also applicable to systems that utilise other regions of the electromagnetic spectrum, such as the infrared (IR) also known as OWC. The enormous growth in the number of information terminals and portable devices in indoor and outdoor environments has accelerated the research and development in OWC technology. The OWC spectrum allocation is depicted in Figure 1.3, showing bandwidth orders of magnitude higher than the RF bandwidth.

Optical Wireless Communication Systems

9

FIGURE 1.3  Visible light in the electromagnetic spectrum in the context of other communications technologies, adopted from [49].

The OWC technology can operate in both indoor and outdoor environments, covering a transmission range from a few meters to several kilometres. Based on their operating wavelength, OWC is split into three categories: ultraviolet (100–400 nm), visible light (380–780 nm), and infrared (IR) (800–2500 nm) [25]. As a complementary access technology to the RF techniques, OWC was first proposed as a medium for short-range WC more than two decades ago [43]. OWC systems are being installed for urban networks, high-speed ground to train networks [44], the last mile FSO links, high-speed indoor OWC links [45], [46], and indoor VLC systems [47], [48]. Additionally, the OWC technology (i.e., VLC or also known as the Li-Fi) can be employed as part of the emerging 5G technologies to overcome the looming spectrum crisis in WCS, particularly in highly populated environments, as well as minimizing the energy consumption in order to achieve greener WCS. Such a technology could be part of the attocell network as the cell sizes are smaller compared to the RF femtocell network, thus offering data rates up to 10 Gbps over a short-range non-LOS path (a few meters or less). The OWC cellular systems potentially can unlock very high area spectral efficiencies and has the potential to revolutionize the next-generation mobile WCS, 5G, and be used as part of the emerging IoT, ultimately achieving up to 1000x higher capacity per area compared to the RF-based Wi-Fi technologies. Most practical OWC systems use light-emitting diodes (LEDs) or laser diodes (LDs) as the transmitter (Tx) and PIN photodiode or avalanche photodiode (APD) as the receiver (Rx). Intensity modulation (IM) with direct detection (DD) is widely used at data rates below 2.5 Gbps, whereas for higher data rates, external modulation is employed. There is wide diversity in the field of OWC applications, including a very short-range (mm range) optical interconnects within integrated circuits, high-volume ubiquitous consumer electronic products, outdoor intra-building links (a few-kilometer range), inter-satellite links (4500 km) and so on. OWC systems offer a number of unique advantages over its RF counterpart, such as [50]–[52]: • Abundance of unregulated bandwidth (200 THz in the 700–1500 nm range) • No utilisation tariffs • No multipath fading when used indoors

10

Optical Wireless Communications

• Highly secure connectivity, requiring a matching transceiver carefully aligned to complete the transmission • The optical beams transmitted are narrow and invisible, making them harder to find and even harder to intercept • Higher capacity per unit volume (bps/m3) due to neighbouring cells sharing the same frequency • Small, light, compact, smaller size components and relatively low cost • Well-defined cell boundaries and no inter-channel interference • Use one wavelength to cover a large number of cells, therefore no frequency reuse problem as in RF • No need to dig up roads and easily installed • Minimal absorption effects at 800–890 nm and 1550 nm • Health-friendly (no RF radiation hazards) • Lower power consumption • Immunity to electromagnetic interference On the other hand, OWC links with an inherent low probability of intercept and has anti-­jamming characteristics that make it among the most secure of all wide-area connectivity solutions. Unlike many RF systems that radiate signals in all directions, thus making the signal available to all within the receiving range, OWC (in particular outdoor FSO links) transceivers uses a highly directional and cone-shaped optical beam normally installed high above street level or on the ceiling within a room with a LOS propagation path. Therefore, the interception of a laser beam is extraordinarily difficult and anyone tapping into the systems can easily be detected as the intercept equipment must be placed within a very narrow optical footprint. Even if a portion of the optical beam is intercepted, an anomalous power loss at the Rx could cause an alarm via the management software. To protect the overshoot energy against being intercepted at the Rx part, a window or a wall can be set up directly behind the Rx [51]. Based on these features, OWC systems developed for voice, video, and broadband data communications are used by the government, military, finance, and for system back-up in large companies and institutions [52]. OWC operating at the near IR region of 750–950 nm has most of the physical properties of visible light, except that it is at the lower part of the optical frequency spectrum, making it invisible to the human eye. However, the eye is very sensitive to this wavelength range and therefore must be protected by limiting the transmission intensity. The eye is also affected by ambient light sources (sun, fluorescent etc.), see Figure 1.4. At the higher wavelength range of 1550 nm, the eye is less sensitive to the light and therefore eye safety requirements are more relaxed, and the interference due to the ambient light sources is considerably reduced. This wavelength is also compatible with the third transmission window of the backbone optical fibre communication networks. IR light, similar to visible light, will not pass through opaque barriers and will be reflected off the walls, ceiling, and most other objects in a room.

1.2  A BRIEF HISTORY OF OWC OWC is an ancient technology that entails the transmission of information-laden optical radiation through the air from one point to the other. Around 800 BC, ancients Greeks and Romans used fire beacons for signalling, and by 150 BC, Native Americans were using smoke signals for the same purpose of signalling. Other optical signalling techniques such as the semaphore were used by the French sea navigators in the 1790s, but what can be termed the first optical communication in an unguided channel was the photophone experiment by Alexander Graham Bell in 1880. In his experiment, Bell modulated the solar radiation with the voice signal and transmitted it over a distance of about 200 meters. The Rx was a parabolic mirror with a selenium cell at its focal point. However, the experiment did not go very well because of the crudity of the devices used and the intermittent nature of the solar radiation.

Optical Wireless Communication Systems

11

FIGURE 1.4  Normalised power/unit wavelength for optical wireless spectrum and ambient light sources.

The fortune of OWC changed in the 1960s with the discovery of optical sources, most importantly, the laser. A flurry of FSO demonstrations was recorded in the early 1960s into the 1970s. Some of these included the spectacular transmission of a television signal over a distance of 30 miles (48 km) using GaAs LED by researchers working in the MIT Lincoln Laboratory in 1962; a record 118 mile (190 km) transmission of a voice modulated He-Ne laser between Panamint Ridge and San Gabriel Mountain, in the United States in May 1963; and the first TV-over-laser demonstration in March 1963 by a group of researchers working in North American Aviation. The first laser link to handle commercial traffic was built in Japan by Nippon Electric Company around 1970. The link was a fullduplex 0.6328 µm He-Ne laser FSO between Yokohama and Tamagawa, a distance of 14 km [53]. From this time on, OWC has continued to be researched and used chiefly by the military for covert communications. The technology has also been heavily researched for deep space applications by NASA and the European Space Agency with programmes such as the Mars Laser Communication Demonstration and the Semiconductor-laser Inter-satellite Link Experiment, respectively. Although deep space OWC lies outside the scope this book, it is worth mentioning that over the past decade, near-Earth FSO systems were successfully demonstrated in space between satellites at data rates of up to 10 Gbps. In spite of the early knowledge of the necessary techniques to build an operational OWC system, the usefulness and practicality of an OWC system was until recently questionable for many reasons [50]. First, existing communications systems were adequate to handle the demands of the time, Second, considerable research and development were required to improve the reliability of components to assure reliable system operation. Third, a system in the atmosphere would always be subject to interruption in the presence of heavy fog. Fourth, use of the system in space where atmospheric effects could be neglected required accurate pointing and tracking optical systems, which were not then available. In view of these problems, it is not surprising that until now, FSO had to endure a slow penetration into the communication networks. With the rapid development and maturity of optoelectronic devices, OWC has now witnessed a rebirth, which can be traced back to the work of Gfeller and Bapst in 1979 demonstrating transmission at 950 nm and 1 Mbps [43]. This work was followed by Kahn and Barry who did pioneering work on OWC (mostly the IR band of 780–950 nm) where systems with high data rates of 50 Mbps and 70 Mbps were was demonstrated in 1996 and 2000 [54], [55]. Since then the range of activities

12

Optical Wireless Communications

has exponentially grown where data rates greater than 1 Gbps have been reported by many researchers [56], [57]. Also, the increasing demand for more bandwidth in the face of new and emerging applications implies that the old practice of relying on just one access technology to connect with the end users has to give way. These forces coupled with the recorded success of its application in military applications have rejuvenated interest in its civil applications within the access network. Several successful field trials have been recorded in the last few years in various parts of the world, which have further encouraged investments in the field. This has culminated in the increased commercialisation and deployment of OWC in today’s communication infrastructures. Full-duplex outdoor OWC systems running at 1.25 Gbps between two static nodes are now common sights in today’s market just like systems that operate reliably in all weather conditions over a range of up to 3.5 km. In 2008, the first 10 Gbps outdoor OWC system was introduced to the market, making it the highest-speed commercially available wireless technology [52]. Efforts are continuing to further increase the capacity via integrated FSO/fibre communication systems and wavelength division multiplexed (WDM) FSO systems which are currently at experimental stages [58]. Table 1.4 gives a concise history of OWC. In last few years, we have seen growing research activities in VLC both within academia [59]–[62], industry [63], and standardisation bodies [64], [65] because of the increasing popularity of solidstate and organic-based lighting technology. The white illumination LEDs, which are highly energy efficient and have longer lifetimes compared to other light sources, will become the dominant light sources in the next decade. Therefore, VLC offers multiple functionalities of illumination, data communication, and indoor positioning, simultaneously. VLC technology is employed in many applications, including indoor data communication, home/office networking, automotive, and TABLE 1.4 A Brief History of OWC Systems Date

Systems/Devices/Standards

800 BC 150 BC 1790s 1800 1880 1960 1970s 1979 1990

Fire beacons—the ancient Greeks and Romans Smoke signal—Native Americans Optical Telegraph—Claude Chappe, France Ship-to-ship communications—France Photophone—Alexander Graham Bell, USA Laser FSO mainly used in secure military applications Indoor OWM systems at 950 nm and 1 Mbps—Gfeller and Bapst Blue LEDs - Japan and USA (awarded the 2014 Nobel Prize for Physics to Professors Akasaki, Amano, and Nakamura) Open standard for IR data communications—The Infrared Data Association (IrDA) OWC for intra-satellite communications, spacecraft, and on-board data communications—NASA (USA) and ESA (Europe) The Visible Light Communications Consortium (VLCC)—Japan Two visible light standards of JEITA CP-1221 and JEITA CP-1222—Issued by Japan Electronics and Information Technology Industries Association Global standards for home networking (infrared and VLC technologies)—“hOME Gigabit Access” (OMEGA) Project—EU • Task Group IEEE802.15.7 on a standard for VLC • VLCC specification standard adopting and expanding the IrDA physical layer was announced P802.15.7 IEEE draft standard—The Task Group IEEE 802.15.7 [64] Organic-based VLC—Northumbria University, UK Gigabit/s VLC—UK, Itlay, and China Task Group on Short-Range OWC, IEEE 802.15.7.r 10 Gbps VLC

1993 2001– 2003 2007 2008 2009 2010 2011 2013– 2015 2016

13

Optical Wireless Communication Systems

underwater communication, due to its low research and development and installation costs. Both FSO and VLC technologies offer a great potential in terms of cost, applications, and reduced energy consumption as part of the greening of global telecommunications, which can be deployed in 5G wireless systems. For transmitting the same high-density data, the energy consumption per LED light source in VLC-based systems is much less than that of RF-based systems [66]. The communication application holds the highest share in VLC application market, whereas the FSO technology market is expected to reach over $3 billion by 2020, with an estimated CAGR of 45% from 2013 to 2020. This emerging green technology is projected to have a compound annual growth rate of 87% between 2014 and 2020 with a market value of almost $10 billion by 2020 [63], [67].

1.3  OWC/RADIO COMPARISON The comparison between radio and IR for indoors wireless communications is shown in Table 1.5. There is very little doubt that RF-based technology will provide mobility both indoor and outdoor over small and large coverage areas. However, the data rates available to end users are limited. Wired LANs are predominantly 10 Mbps and 100 Mbps, and 1 and 10 Gbps Ethernet are used to

TABLE 1.5 Properties of RF and OWC Technologies Properties

RF

OWC

Implications for OWC

Bandwidth regulated

Yes

No

Data rate

Low to medium (Gbps)

Medium to high (>Gbps)

Transmitted beam size

Large (>20 m)

Small (2 m)

Passes through walls

Yes

No

Resilience to Rain

Low

High

Resilience to Snow

High

Low

Resilience to Fog

High (200 dB/km) • Low link availability • Both path loss and phase variation • Reduced data rate due to dispersion • Fading • Main propagation path is LOS • Simple link design • Inherently low—40% lower than OF • Faster delivery of information • Problematic at high data rates. • Short range (a few km) • Shorter range (< km) • Reduced performance • Limited range • Wavelength division multiplexing • x(t) is the input signal with a high peak-average ratio

Resilience to Fog

2

dt

High (37 dB/km at 830 nm)

14

Optical Wireless Communications

connect fixed terminals. Mobile connections are available using the traditional cellular networks. The RF wireless LANs use the unregulated spectrum bands and are defined in the IEEE 802.11 standards, with networks operating at 2.4 GHz scientific and medical (ISM) band (data rate of 1–2 Mbps), 5.7 GHz (data rate of ∼20 Mbps), and proposed 17 and 60 GHz where available data rates are much higher. The RF technology is excellent at providing coverage at lower data rates (i.e., lower carrier frequencies). This is due to the diffraction and scattering of RF waves and the sensitivity of RF Rxs. RF channels are also robust to blocking and shadowing and can provide full coverage between rooms. At higher carrier frequencies (i.e., offering much larger data rates), the RF propagation becomes more LOS, and problems encountered are similar to those using light. At these frequencies, the RF components are expensive, and the key advantages of the RF technology (i.e., the mobility, coverage, and Rx sensitivity) become less clear compared to the OWC systems. In contrast, OWC systems (indoor and outdoor) covering a wide unlicensed spectral band of 700–10,000 nm has the potential to offer a cost-effective protocol-free link at data rates exceeding 10 Gbps for both indoor and outdoor links (see Figure 1.5). For indoor applications through multiple user-sized cells, and because of the intrinsically abrupt boundary of these cells, interference would be negligible, and the carrier reuse would not be an issue in order to increase the system capacity. Indeed, the optical wireless technology is a future-proofed solution, since additional capacity far beyond the capabilities of radio could be delivered to users as their needs increase with time. Perhaps the largest installed short-range wireless communications links are optical, rather than RF. It is argued that OWC has a part to play in the wider 5G WCS. In large open environments where individual users require greater than 100 Mbps or more, OWC is a more sensible solution because of its multiple user-size cells, reduced interference, and improved carrier reuse capabilities due to its intrinsically abrupt cell boundaries. Optical wireless LANs have been utilised on a large scale in a number of applications, such as telemedicine, emergency situations, high-speed trains, laptops, PDAs, museums, and so on [54], [68]. According to the IEEE 802.11 specifications, the OWC physical layer can support data rates of up to 2 Mbps with a potential to migrate to higher data rates. The Infrared Data Association (IrDA) has standardised low-cost and short-range (1–8 m) OWC data links supporting data rates ranging from 2.5 kbps to 16 Mbps at a wavelength between 850 nm and 900 nm [69]. To an extent, RF and OWC technologies may be viewed as complementary rather than competitive. For example, if a wireless LAN is required to cover a large area, where users can roam freely and remain connected to the network at all times, then RF is the only cost-effective medium that can achieve this. If, however, a wireless LAN is required to cover a more modest area but deliver advanced bandwidth-hungry multimedia network services such as video conferencing and video on demand, then OWC is the only medium that truly has the bandwidth available to deliver this. For the “last mile” access networks, the OWC technology could be employed to overcome the bandwidth bottleneck (see Figure 1.6). To take advantage of OWC high-transmission capabilities and yet have a weather-resistant access, the hybrid OWC/MMW (mmW) technology would be desirable in a number of outdoor applications where higher data rates in all weather conditions are desirable [70]. In fact, this, as well as VLC, could be one of the best solutions for the future 5G WCS.

1.4  LINK CONFIGURATION Figure 1.7 shows a simple block diagram of the OWC system. Both LED and LD could be used as the light source. To date, commercially available outdoor OWC systems have come close to delivering high data rates over a link space up to 5 km, which are significantly faster than the latest radio LAN products currently available. We have also seen very high data rates, >5 Gbps indoor OWC system based on the VLC technology, but over very short transmission spans (less than a meter) being reported [62], [71]. Nevertheless, OWC is a challenging medium, and there are numerous considerations that must be taken into account when designing very high-speed OWC links. In an indoor environment, the light will reflect off the ceiling, walls, and most other objects in a typical

15

Optical Wireless Communication Systems

(a)

(b) FIGURE 1.5  Bandwidth capabilities for a range of optical and RF technologies for (a) long range and (b) short range.

16

Optical Wireless Communications

FIGURE 1.6  Access network bottleneck.

room or enclosure but will not go through opaque barriers. Whereas in an outdoor environment, light will be scattered and absorbed due to atmospheric conditions. There are numerous ways that an optical link can be physically configured. These are typically grouped into four system configurations: (i) directed line of sight (LOS); (ii) non-directed LOS; (iii) diffuse; and (iv) tracked. There are also a number of Rx configurations, as well as hybrid variations of these configurations as illustrated in Figure 1.8. Directed LOS—Typically used for point-to-point communication links mainly outdoor and in some cases in an indoor environment, for example in the VLC-based system. The optical signal is concentrated in a very narrow beam, thus exhibiting low power requirements as well as creating a high-power flux density at the PD. Furthermore, a LOS link offers the highest data rate (a few Gbps) over a linkspan from a few meters to a few km [52], [72]–[75]. Furthermore, directed LOS links do not suffer from multipath-induced signal distortion, and noise from the ambient light sources is also largely rejected when used with a narrow field-of-view (FOV) Rx [76]. The data rate is, therefore, limited by free space path loss rather than the effects of multipath-induced dispersion

FIGURE 1.7  A system block diagram of an OWC system.

Optical Wireless Communication Systems

17

FIGURE 1.8  Link configurations: transmitter and receivers.

[76]–[78]. However, there are disadvantages. For indoor applications, the coverage area provided by a signal channel could be very small, so providing area coverage and roaming could become problematic. Directed LOS links cannot support mobile users because of the requirement for alignment of Rx and Tx modules. In LOS, OWC-based LANs cater for single users, the Tx/Rxr BS is usually mounted on the room ceiling. However, in situations where there is a need to offer services to multiple users within a relatively large room, a possible option would be to adopt a cellular topology, where the narrow transmitted optical beam is replaced with a source with a wide optical footprint (see Figure 1.9) [79]. In cellular OWC links, the optical signal is broadcast to all mobile users

FIGURE 1.9  Cellular OWC system.

18

Optical Wireless Communications

located within the cell, and communication between mobile users is established via a BS located at the centre of each cell mounted on the ceiling. The cellular OWC topology offers mobility at the cost of reduced power efficiency. LOS systems employing MIMO, which involves using LED/LD and PD arrays [80], a spatial light modulator (where data is sent by creating a series of two-dimensional optical intensity images) and or an image as a Tx and Rx, offer great gains [81], [82]. Additionally, directed LOS links (mainly for outdoor applications) must be accurately “pointed” or aligned before use and require an uninterrupted LOS path, making them susceptible to beam blocking. In addition to this, by their very nature, they are more suited to point-to-point links rather than point-to-multipoint broadcast-type links, thus reducing their flexibility. Directed LOS is the most well-known link topology. In recent years, we have seen a growing interest in application of LOS links for a number of outdoor applications: university campuses, last mile access networks, centre cellular communications back-haul, inter-satellite communications, satellite uplink/downlink, deep space probes to the ground, disaster recovery, fibre communications back-up, video conferencing, links in difficult terrains, wide and local area networks, ground-to-ground (short- and long-distance terrestrial links), deep space probes to ground, and ground-to-air/air-to-ground terminal (UAV, HAP, etc.), and temporary links. The LOS OWC link offers a number of advantages, including greater security and higher data rates (easily exceeding 100 Gbps) using wavelength division multiplexing (WDM) techniques, as well as small terminal size, light weight, minimal aperture sizes, and low power consumption. In hybrid LOS systems, either the Tx or the Rx has a wide FOV, while the other element has a narrow FOV. A typical hybrid LOS system uses ceiling-mounted Txs that illuminate the wide area. As in directed LOS links, the Tx has a narrow-angle emission, and the Rx has a narrow FOV for directed non-LOS links. However, to overcome any obstacles between them, the Tx is aimed at a reflective surface so that the first reflection can reach the Rx. In addition to overcoming the barrier, the information signal is received after a single reflection, which minimizes the multipath dispersion [51]. However, the alignment is problematic due to highly directional Tx and Rx. The hybrid system incorporates either a wide beam Tx or a narrow FOV Rx or vice versa. This relaxes the need for strict alignment between the Tx and the Rx. Though blocking probability can be significantly reduced using the link design, the system is affected by multipath propagation. Non-directed LOS—For indoor applications, the non-directed LOS is considered to be the most flexible and robust configuration. Non-directed LOS uses a wide beam Txs, a wide FOV Rxs, and scatters from surfaces within the room to achieve a broader coverage area and an excellent mobility without the need for precise alignment or a tracking mechanism, similar to the RF technology [54]. Non-directed links are suitable for point-to-multipoint broadcast applications, which offer robustness to shadowing and blockage. They overcome blocking problem by relying on reflections from surfaces of objects within rooms so that a high proportion of the transmitted light is detected at the PD from a number of different directions. However, they incur a high optical path loss (thus higher transmit power) and must also contend with multipath-induced dispersion. Whilst multipath propagation does not result in multipath fading in indoor OWC systems since PD sizes are huge in comparison with the light wavelength, it does give rise to intersymbol interference (ISI) [55], [83], thus limiting the data rate to hundreds of Mbps in a typical-size room. However, indoor non-LOS (NLOS)-OWC links employing LEDs with small divergence angles offer higher data rates and lower channel distortion than those with large divergence angles [84]. Therefore, to achieve a higher transmission bandwidth and a more uniform optical power distribution, multi-cell NLOS-OWC systems would be the preferred solution. Multi-cell indoor NLOS-OWC systems (see Figure 1.10) employing the optimized Lambertian order of LEDs offer significant improvement in terms of the transmission bandwidth [85]. A NLOS system with a wide acceptance angle is very sensitive to the ambient light (i.e., sunlight and artificial lights), which can induce a very high noise level, thus strongly limiting the link performance and the maximum achievable data rate [43]. A bi-directional NLOS OWC system based on VLC downlink and an IR uplink upstream have been reported with data rates approaching 250 Mbps over a transmission span of up to 2 m [86].

19

Optical Wireless Communication Systems

FIGURE 1.10  Multi-cell indoor non-directed cellular OWC systems.

Diffuse configuration—Also known as non-directed non-LOS proposed in [43], typically consists of a Tx with a wide FOV that points directly towards the ceiling emitting a wide IR or visible light beam and a wide FOV Rx [54], [87]. The diffuse OWC indoor topology is the most convenient for LAN ad hoc networks since it does not require careful alignment of the Tx and Rx modules, nor does it require a LOS path to be maintained and is almost immune to blockage of the transmission path and shadowing. In addition to this, it is also extremely flexible and can be used for both infrastructure and ad hoc networks [43], [76]. Unfortunately, diffuse links experience high path loss, typically 50–70 dB for a horizontal separation of 5 m [88]. The path loss is increased further if a temporary obstruction, such as a person, obscures the Rx and blocks the main signal path; a situation referred to as shadowing [88]. In addition, a PD with a wide FOV normally collects signals that have undergone one or more reflections from ceiling, walls, and room objects. Such reflections attenuate the signal with typical reflection coefficients being between 0.4 and 0.9 [43]. The received signal can also suffer from the severe multipath-induced dispersion, where the transmitted pulses spread out in time over alternative routes of differing lengths, thus limiting the maximum unequalised bit rate achievable with a room volume of 10 × 10 × 3 m3 to typically around 16 Mbps [43]. Dispersion-induced ISI incurs a power penalty and thus bit-error rate (BER) degradation. As a result of these factors, relatively high optical transmit powers are required. For example, at high data rates, the power requirements of pulse position modulation (PPM) are higher than the on-off keying (OOK). However, the average optical power emitted by an IR transceiver is limited by eye safety regulations and electrical power consumption in portable (battery powered) devices. Therefore, the use of power efficient modulation techniques is desirable. In addition, in diffuse systems, the entire room needs to be illuminated by a single or multiple Txs. This is naturally realizable by diffused light propagation after a few reflections but requires a relatively huge amount of transmitted optical power. For instance, 475 mW of transmitting power is required in a diffuse 50-Mbps link at a horizontal link separation of ∼5m between the Tx and the Rx both directed to the ceiling [54]. Table 1.6 shows the key features of the directed and diffuse OWC links. TABLE 1.6 Features of Directed and Diffuse Links Capacity Bandwidth Power Efficiency Mobility Directed link High Diffuse link Low

High Low

Good Average

BK-TandF-9781498742696_TEXT_GHASSEMLOOY-181209-Chp01.indd 19

Reliable Access

Shadowing/Blocking

Yes with tracking Yes with tracking Yes Yes Yes No

16/04/19 9:32 AM

20

Optical Wireless Communications

Diffuse OWC links have been standardized by both IrDA and IEEE. The IrDA standardisation resulted in the Advanced Infrared (AIr) physical layer interface for 4 Mbps within a typical room sizes [89]–[91]. The IEEE 802.11 standard included a physical layer specification for IR OWC at 1–2 Mbps in its 802.11 standards. Replacing the non-imaging Rx in both non-directed and non-LOS links by a single imaging light concentrator and an array of the photodetector can further improve the link performance by way of reducing the multipath distortion and the ambient light noise. Performance of diffuse systems can be substantially improved by means of spatial diversity techniques, which use a combination of multi-beam Tx and multiple element narrow FOV angle-diversity Rx, which is known as multi-spot diffusing (MSD) system (see Figure 1.11) [92]–[94]. MSD combines the advantages of point-topoint links with the mobility of diffuse links, and the light power is projected in the form of multiple narrow beams of equal optical intensity over a regular grid of small spots on to a ceiling [92]. In this configuration, each diffusing spot would be considered a source IM with the same information, and the arrangement and number of spots is optimized in such a way that at least one spot is in the imager for every Rx position. Thus, offering a LOS link to a narrow FOV receiver, reducing path loss, and increasing link SNR by ∼20 dB [94]–[97]. In MSD, the Rx ideally images one or a number of spots and decodes the data from the diffused reflected energy. The same data is modulated for all spots, and the arrangement and number of spots is optimized so that at least one spot is in the imager for every receiver position. In a typical room setting, the channel bandwidth of an MSD system can be in excess of 2 GHz compared to the tens of MHz in diffuse systems [98]. To implement MSD links, one could use a less expensive and compact scheme based on a laser diode and a computer generated hologram to achieve large spot sizes, for example 10 × 10 spots [99], compared to the more complex array emitter and their associated optics [55]. Alternatively, to reduce transmitter complexity, different geometries for MSD spots, such as a line strip and a diamond, have been used [94]. Angle diversity Rxs consist of multiple Rx elements effectively pointing in different directions either because of their physical orientation or due to imaging concentrators [55], [100]. The photocurrent generated by each Rx element can be processed in a number of ways, including the use of digital wavelet-artificial intelligent techniques. This type of Rx can reduce the effects of ambient noise and multipath distortion because these unwanted contributions are usually received from different directions to that of the desired signal [96]. The performance of such schemes has been analysed in [55], [97], [101]–[105]. An angle diversity Rx based on discrete elements operating at

FIGURE 1.11  Multi-spot diffusing configuration.

Optical Wireless Communication Systems

21

70 Mbps over a linkspan of a few meters has been reported [55], whereas MSD systems at 100 Mbps [96], 140 Mbps with a single Rx [106], and a 155 Mbps with a multi-element imaging Rx receiver [107] also have been reported. Therefore, it is fair to say that MSD systems offer a high bandwidth and low path losses, as well as providing mobility compared to the diffuse and NLOS configuration at the cost of a bit more system complexity [108]. Tracked systems—Here the Tx/Rx BS is mounted on the ceiling and the MSs are located at tabletop height (see Figure 1.8). For the LOS link, where the transmitted optical beams for both the uplinks and downlinks are focused onto the Rx, less power is needed than the diffuse link. The combination of a Tx with a narrow beam profile and an Rx with a smaller FOV results in reduced multipath-induced ISI and ambient light interference. In contrast, the diffuse IR link offers improved mobility since it uses reflections from the ceiling, walls, and object within the room to send data from the Tx to the Rx (i.e., MS), thus requiring no alignment. In such configurations, full connectivity between the Tx and the Rx can be maintained even when the LOS paths are blocked. However, the diffuse link suffers from multipath-induced ISI, hence the reduced data rate. As shown in Figure 1.12, this scheme offers high-power efficiency and potentially high bit rates (1 Gbps using mechanically steerable optics [109]) of directed LOS links with the increased coverage area enjoyed by non-directed-LOS links. Unfortunately, mechanically steerable optics are expensive to realise. Electronic tracking schemes have been proposed in [109] followed by a 100 Mbps practical system operating at a wavelength of 1550 nm in [52] and a 155 Mbps link in [110]. Solid state systems have been proposed, which are conceptually similar to angle diversity techniques. Solid state tracked systems, using the multi-element Tx and the Rx arrays along with a lens arrangement have been proposed [76], [111]. Using this arrangement, steering is merely a matter of selecting the appropriate array element. More recently, a single channel imaging Rx has been proposed [112], which could be described as a hybrid tracked and angle diversity system. Optical multiple input multiple output (OMIMO) systems improve channel quality due to spatial diversity, or system speed using multiplexing. Note that, along with diffuse links using multi-beam Txs and angle-diversity Rxs, tracked systems offer the potential to implement space division multiplexing, whereby multiple users can communicate without suffering a loss of per-user capacity since each user is located in a different cell. OWC links employing one-dimensional tracking systems using arrays of lasers and PDs, angle diversity, and imaging Rxs have been developed to provide high data rates, mobility, and coverage area using the line-of-sight channels in an indoor environment. LOS links employing a Rx with a narrow FOV are more efficient, offering a high optical signal gain and a low optical noise. For outdoor applications with LOS links, the main challenges are the atmospheric conditions that limit the link range and thus the link availability during all weathers. In LOS optical wireless LANs, the narrow transmit beam is now replaced by a beam with a much wider “footprint” to cover a large area

FIGURE 1.12  Hybrid diffuse and tracked LOS links.

22

Optical Wireless Communications

FIGURE 1.13  Optical wireless LAN: (a) diffuse and (b) line of sight.

within a room (see Figure 1.13), thus offering optical “cellular” capability but at the cost of reduced power efficiency, since more transmit power is required to ensure the adequate optical power flux density at the receiving end. In optical cellular configurations, the optical signal is broadcasted to all terminals within the cell, and communication between subscribers is accomplished in a star network topology. One advantage of the optical cellular system is that the same wavelength (or colour) could be used to cover a large number of cells within the same area, thus overcoming the need for wavelength reuse (or frequency reuse as in RF cellular systems). A hybrid cellular and NLOS tracked systems (see Figure 1.10) would offer additional flexibility such as the following: i. Tracking and pointing functions—ensuring 100% link availability at all times ii. Power efficiency—by means of switching off the light sources that do not illuminate user terminals [95], [113] iii. Increased system capacity—each source in the emitter array transmits different data streams iv. Increased transmission bandwidth—as a result of PD arrays with low capacitance v. Reduced ambient light—because of a narrow FOV at the Tx and Rx (PDs) In such a system, two-dimensional arrays of Tx and PDs are utilised, where sources in the Tx array emit normally to the plane of the array. Using optical systems, the light beam is deflected into a specific angle depending on the spatial position of the source within the array. Thus, a cell is split into microcells, each illuminated by a single light source of the array. At the Rx, depending on the angle of arrival of the optical signal, the light is focused onto a particular PD in the PD array. To increase the data rate, spectrally efficient modulation formats could be used (see Chapter 4). Multielement optical Tx and Rx, quasi-diffuse OWC links employing discrete and imaging Rxs, and a combination of coding techniques have been investigated to improve optical power efficiency. For the outdoor environment, FSO links are capable of offering the similar performance of optical fibre communications over a link range spanning from of a few hundred meters up to a few kilometres. Commercial outdoor FSO systems are available with full beam alignment and tracking systems, full-duplex connectivity, configurable and upgradable bandwidth capability, and can be implemented in a number of configurations, including point-to-point, hot standby, multi-hope/ repeater, hub-and-spoke, ring, and mesh. The FSO technology with the confined beam provides significant degrees of robustness and covertness, where an Rx needs to be positioned nearly in the direct path of the FSO beam between two points to intercept it. Signal jamming in RF-based technologies is a major problem, whereas jamming an FSO laser beam is rather difficult because of the pointing capabilities required to accurately place optical energy in an Rx. A key property of FSO technology is that the laser beam is highly directional, which with the appropriate optics, is often designed to have a divergence of a few milliradians or less in order to concentrate the optical energy

Optical Wireless Communication Systems

23

FIGURE 1.14  The effects of scintillation.

on to the Rx. Therefore, in such scenarios pointing, acquisition, and tracking is much more challenging than the RF-based WSC, where the Tx and the Rx antennas may only be required to generally point at one another to establish communications. However, in FSO systems for communication to take place, each optical transceiver must be simultaneously pointed at the other. The typical optical transmit power could be up to 100 mW with a sufficient power margin at the Rx to deal with attenuation due to atmospheric and meteorological conditions, such as fog, snow, rain, drizzle, turbulence, and thermal expansion of structures. Atmospheric and meteorological conditions lead to fading of the received signal, sometimes more than the RF waves. Fog is the most difficult to cope with, due to a very high attenuation, compared to the rain and snow. For a link length of more than a few hundred meters, temperature-induced scintillation affecting the laser beam propagation (i.e., defocusing and deflecting the beam, see Figure 1.14) though the channel becomes more problematic [114]. Commercial FSO systems perform well in clear weather between static nodes at distances of up to a few kilometers. In these links, any mobility correction is limited to building sway, and the atmospheric effects are often countered by increasing the transmit optical power.

1.5  OWC APPLICATION AREAS The outdoor OWC systems (i.e., FSO) are very attractive for a number of applications within the access and the metro networks, as well as to fill an optical gap existing between the network core and the network edge. It can conveniently complement other technologies, such as wired and wireless RF communications, fibre-to-the-X technologies, and hybrid fibre coaxial among others, in making the huge bandwidth that resides in an optical fibre backbone available to the end users. The point that most end users are within a short distance from the backbone—one mile or less—makes the FSO technology very attractive as a data bridge between the backbone and the end users. Among other emerging areas of application, terrestrial FSO has been found suitable for use in the following areas: Last mile access—FSO is used to bridge the bandwidth gap (last mile bottleneck) that exists between the end users and the fibre optic backbone. Links ranging from 50 m up to a few km are readily available in the market with data rates covering 1 Mbps to 10 Gbps. Enterprise—The flexibility of FSO allows it to be deployed in many enterprise applications, such as LAN to LAN connectivity, storage area networks, big data centres, and intracampus and healthcare campus network connections. Systems with data rates up to a few Gbps covering a linkspan of 1–2 km are already available in the market.

24

Optical Wireless Communications

Dense WDM Services—With the integration of WDM and FSO systems, independent players wanting to build their own fibre rings, yet only owning part of the ring. Optical fibre backup link—Used to provide the backup link against loss of data or communications breakdown in the event of damage or unavailability of the main optical fibre link. Link range could be up to 10 km with data rates up to 10 Gbps. Cellular communication backhaul—Can be used to backhaul traffic between base stations and switching centres in the 3G/4G and future 5G wireless networks, as well as transporting IS-95 code division multiple access signals from macro- and micro-cell sites to the base stations. Disaster recovery/Temporary links—In a situation where a temporary link is needed be it for a conference or ad-hoc connectivity in the event of a collapse of the existing communications infrastructure due to made or natural disasters. Difficult terrains—FSO is an attractive data bridge in such instances as across a river, a very busy street, rail tracks, rural areas or where the right of way is not available or too costly to pursue. High definition television—In view of the huge bandwidth requirement of high-definition cameras and television signals, FSO is increasingly being used in the broadcast industry to transport live signals from high-definition cameras in remote locations to a central office, an international sporting arena, etc. Military—Ultra-high-capacity crosslinks between processing satellites, satellite-to-air or -ground platforms, airborne networks, air-to-ground links to extend the high-rate RF links. Figure 1.15 illustrates a block diagram of the outdoor OWC system. For such systems, there are a number of challenges that may affect the OWC link performance. i. Ambient light sources—with the spectra well within the Rx bandwidth of silicon PDs, thus resulting in background noise [115]. ii. Blocking and shadowing—the blockage or partial obstruction of optical rays due to multipath propagation in an indoor environment, which also causes pulse spreading, thus leading to ISI [116], [117]. To reduce the impact of ISI, there are a number of mitigation techniques, including the diversity detection and emission [55], [118], equalisation both at the Tx and Rx [119], and adaptive threshold detection.

FIGURE 1.15  A system block diagram of an outdoor OWC link.

Optical Wireless Communication Systems

25

iii. Alignment and tracking—by employing highly directional and narrow beams of light, the changes in mispointing of the transmit beam, as well as the error due to the tracking of the Rx, will introduce signal fading. One dimensional tracking systems using arrays of lasers and PDs, angle diversity, and imaging Rxs have been developed to provide the coverage area using line-of-sight channels. iv. Adverse atmospheric weather—can have a serious impact on the outdoor OWC link availability and performance. The performance of outdoor laser-based OWC systems is highly dependent on the operating transmission windows [120]–[123]. Shorter wavelengths are most widely used because of the availability of low-cost devices and components. Previous studies based on available weather statistics have shown that weather effects can be mitigated by switching from IR bands (700–1000 nm and 1550 nm) to far-IR [123]. Thus, a quantum cascaded laser operating at the mid-wave and long-wave IR band (3–20 micrometres) can offer a solution to atmospheric-induced optical losses [105], [124], [125]. In recent years, we have seen the emergence of visible light technology for both illumination and communications. These devices are mechanically robust with a high energy efficiency, offering simultaneous illumination and intensity modulation at a data rate in excess of 100 Mbps [62], [126], [127]. Such devices are opening up new applications in indoor environments, such as mobile-tomobile links, and building and industrial automation (see Chapter 8).

1.6  SAFETY AND REGULATIONS IR light sources pose a potential safety hazard to a human if operated incorrectly. The optical beams can cause injury to both the skin and the eye, but the damage to the eye is far more significant because of the eye’s ability to focus and concentrate optical energy. The eye can focus light covering the wavelengths around 0.4–1.4 µm onto the retina. Other wavelengths tend to be absorbed by the front part of the eye (the cornea) before the energy is focused. Figure 1.16 shows the absorption of the eye at different wavelengths. At 700–1000 nm spectral range, where optical sources and detectors are low cost, the eye safety regulations are particularly stringent. The maximum measured radiant intensity with a maximum permissible exposure at a wavelength of 900 nm is ∼143 mW/sr [128]. However, at longer wavelengths of 1500 nm and above (i.e., the third transmission window in optical fibre backbone networks), the eye safety regulations are much less strict, but the devices available are relatively costly. Therefore, when designing optical communication systems, efforts must be made to ensure that the optical radiation is safe and does not cause any damage to the people that might come in contact with it.

FIGURE 1.16  Response/absorption of the human eye at various wavelengths.

26

Optical Wireless Communications

There are a number of international standards bodies which provide guidelines on the safety of optical beams; notable among these are: Center for Devices and Radiological Health—An agency within the U.S. Food and Drug Administration (FDA) that establishes regulatory standards for lasers and laser equipment that are enforceable by law (21 CFR 1040). International Electrotechnical Commission (IEC)—Publishes international standards related to all electrical equipment, including lasers and laser equipment (IEC60825-1). These standards are not automatically enforceable by law, and the decision to adopt and enforce IEC standards is at the discretion of individual countries. American National Standards Institute (ANSI)—Publishes standards for laser use (ANSI Z136.1). ANSI standards are not enforceable by law but do form the basis for the U.S. Occupational Safety and Health Administration (OSHA) legal standards, as well as comparable legal standards that have been adopted by various state regulatory agencies. European Committee for Electrotechnical Standardization (CENELEC)—An organisation that establishes electrotechnical standards based on recommendations made by 19 European member nations. CENELEC standards are not directly enforceable by law but, as with IEC standards, they are often integrated into the legal requirements developed by individual countries. Laser Institute of America (LIA)—The LIA is an organisation that promotes the safe use of lasers, provides laser safety information, and sponsors laser conferences, symposia, publications, and training courses. Each of these organisations has developed ways of classifying lasers. Specific criteria vary slightly from one body to the other, but the IEC classifications will be considered in this section. Lasers are generally divided into four groups: Class 1 to Class 4, with Class 1 being the least powerful and Class 4 being the most powerful. Each class is defined by the accessible emission limits (AEL) metric, this depends on the wavelength of the optical source, the geometry of the emitter, and the intensity of the source. Table 1.7 presents the main characteristics and requirements for the classification system as specified by the revised IEC 60825-1 standard [129]. In addition, Classes 2 and higher must have the triangular warning label, and other labels are required in specific cases indicating laser emission, laser apertures, skin hazards, and invisible wavelengths. For outdoor LOS OC links, generally high-power lasers, Class 3B band are used in order to accomplish good power budget. To ensure eye safety, stringent regulations, such as ANSI Z-136 and IEC 825 series, have been established [129]. The AEL values at two wavelengths mostly used for FSO are presented in Table 1.8. Lasers classified as Class 1 are most desirable for optical wireless communication systems since the radiation they emit is at a safe level under all conditions and circumstances. Figure 1.17 shows the eye safety limits for the two most popular wavelengths of 900 and 1550 nm adopted in OWC systems. For a collimated light source, 1550 nm is preferable. Class 1 lasers require no warning labels and can be used without any special safety precautions. As shown in Table 1.8, the power available to Class 1 lasers is limited. Most commercial terrestrial FSO links operating at up to 1.5 Gbps use Class 1M lasers, these are inherently safe except when viewed with optical instruments, such as binoculars. In certain instances, higher class lasers are used for FSO, the safety of these systems is maintained by installing the optical beams on rooftops with safety and warning labels or on towers to prevent inadvertent interruption [130]. As shown in Table 1.8, a Class 1 laser system operating at 1550 nm is allowed to transmit approximately 50 times more power than a system operating in the shorter IR wavelength range, such as 850 nm, when both have the same size aperture lens. It should, however, be noted that no wavelength is inherently dangerous or eye-safe; it is the output power that determines the laser classification. It is, therefore, possible to design eye-safe FSO systems that operate at any wavelength. It also is important to understand that the regulation addresses the power density in front of the transmit aperture rather than the absolute power created by a laser diode inside the equipment. For example, the laser diode inside the FSO equipment can actually be Class 3B even

Optical Wireless Communication Systems

27

TABLE 1.7 Classification of Lasers According to the IEC 60825-1 Standard Category

Description

Class 1

Low-power device emitting radiation at a wavelength in the band 302.5−4000 nm. Device intrinsically without danger from its technical design under all reasonably foreseeable usage conditions, including vision using optical instruments (binoculars, microscope, monocular). Same as Class 1, but there is the possibility of danger when viewed with optical instruments (binoculars, telescope, etc.). Class 1M lasers produce large-diameter beams or beams that are divergent. Low-power device emitting visible radiation (400−700 nm band). Eye protection is normally ensured by the defence reflexes, including the palpebral reflex (closing of the eyelid). The palpebral reflex provides effective protection under all reasonably foreseeable usage conditions, including vision using optical instruments (binoculars, microscope, monocular). Low-power device emitting visible radiation (400−700 nm band). Eye protection is normally ensured by the defence reflexes, including the palpebral reflex (closing of the eyelid). The palpebral reflex provides an effective protection under all reasonably foreseeable usage conditions, with the exception of vision using optical instruments (binoculars, microscope, monocular). Average-power device emitting radiation in the 302.5−4000 nm band. Direct vision is potentially dangerous. Generally located on rooftops. Average-power device emitting radiation in the 302.5−4000 nm band. Direct vision of the beam is always dangerous. Medical checks and specific training required before installation or maintenance is carried out. Generally located on rooftops. There is always danger to the eye and skin, fire risk exists. Must be equipped with a key switch and a safety interlock. Medical checks and specific training required before installation or maintenance is carried out.

Class 1M Class 2

Class 2M

Class 3R Class 3B

Class 4

though the system itself is considered to be a Class 1 or 1M laser product if the light is launched from a large-diameter lens that spreads out the radiation over a large area before it enters the space in front of the aperture [131]. In order to maintain the Class 1 or 1M safety classifications, it is possible to use a higher-power laser with increased lens aperture or to use multiple large transmission apertures [131]. LED sources do not produce a concentrated light beam and therefore cannot be focused onto the retina. This makes the use of LEDs ideal for indoor applications. The penalty, however, is reduced bandwidth compared to the LDs. Recent developments in vertical cavity surface emitting lasers (VCSEL), offering a safer peak wavelength at 1550 nm [132], make them an attractive option for outdoor and even indoor applications. This due to their well-controlled, narrow beam properties, high data rate, excellent reliability, low power consumption, and the possibility of the array configuration.

TABLE 1.8 Accessible Emission Limits for Two Wavelengths, 850 nm and 1550 nm Class

Average Optical Power Output (mW) λ = 850 nm λ = 1550 nm

1 2 3R 3B 4

500

28

Optical Wireless Communications

FIGURE 1.17  The eye safety limits for 900 and 1550 nm wavelengths.

1.6.1  Maximum Permissible Exposures (MPE) To provide better protection for the eyes or skin of anybody that might come in contact with the radiation of laser equipment, the maximum permissible exposure limits have been stipulated by standards organisations. It is the highest radiation power or energy, measured in W/cm2 or J/cm2, that is considered safe with a negligible probability of causing damage. The MPE is measured at the cornea of the human eye or at the skin surface for a given wavelength and exposure time. It is usually about 10% of the dose that has a 50% chance of creating damage under worst-case conditions [79]. In Table 1.9, the MPE values for the eye at two commonly used wavelengths for FSO systems are presented; the values for the skin are much lower since the skin is usually less sensitive to laser radiation. As shown in Table 1.9, the MPE is higher for brief exposure durations than for high exposure times. The MPE for the eye is much higher at 1550 nm than at 850 nm; this is related to the laser radiation absorption at the level of the various eye components. This difference in MPE values can be explained by the fact that at 850 nm approximately 50% of the signal can reach the retina, whereas at 1550 nm, the signal is almost completely absorbed by the cornea and aqueous humour [78].

1.7  OWC CHALLENGES In Table 1.10, the major challenges encountered in the design and implementation of OWC systems are outlined. TABLE 1.9 Example of MPE Values (W/m2) of the Eye (Cornea) at 850 nm and 1550 nm Wavelengths [78] Exposure duration (s) MPE (W/m2) at 850 nm MPE (W/m2) at 1550 nm

1 36 5600

2 30 3300

4 25 1900

10 20 1000

100 11 1000

1000 6.5 1000

10,000 3.6 1000

Challenges in Optical Wireless Communications InterSymbol Interference (ISI)

Causes

Effects

Solutions

Indoor/FSO

Multipath propagation depends on: • Room size • Number of reflections • Reflection coefficient within the room • Modulation schemes • Optical beam scattering (e.g., FSO)

• Quality of the transmission • Multipath distortion or dispersion • Reduce date rates • Increased BER

• Equalisation and predistortion equalisation • Forward error correction (FEC) • Spread spectrum techniques • Multiple-subcarrier Modulations • More bandwidth efficient than single-carrier • OFDM, MSM • m-CAP • Line strip spot diffusing Tx [Indoor] • Diversity detection and emission [Indoor][133] • Adaptive threshold detection • Multi-beam Tx • FOV controlling • Power efficient modulation PPM, DPIM, etc. • Use LED, holographic diffuser • Adaptive optics—Reduces the need for increased power by correcting beam for improved SNR Use 1550 nm Optical and electrical filtering

Indoor/FSO

Pre-amplification FEC Low-noise amplifier

Indoor/FSO Indoor/FSO Indoor/FSO Indoor/FSO Indoor/FSO Indoor/FSO

Safety

Laser radiation

Damage to the eyes and skin

Wavelengths Noise

Source at 800–900 nm Dark current shot noise

Damaging the eye Limit performance of communication systems BER deterioration

Quantum shot noise Background (ambient) noise Thermal noise Relative intensity noise (LDs) Excess noise ASE (optical amplifier)

APD-amplification process Optical filter

Optical Wireless Communication Systems

TABLE 1.10 Challenges in OWC Systems

Indoor Indoor/FSO Indoor

FSO Indoor/FSO

(Continued )

29

30

TABLE 1.10  (Continued) Challenges in OWC Systems Challenges in Optical Wireless Communications Causes

Effects

Solutions

Indoor/FSO

Turbulence

Random refractive index variation

From 0.22—272 dB/km of loss

• Coding, e.g., LDPC, FEC MIMO • Diversity reception (Temporal and Spatial) • Adaptive Optics • Robust modulation techniques • Adaptive optics • Coherent detection not used due to Phase • Adaptive optics • Increase transmit optical power • Hybrid FSO/RF • Diversity • More efficient modulations

FSO/Also indoor if laser diodes are used

Fog

• Irradiance fluctuation (scintillation) • Image dancing • Phase fluctuation • Beam spreading • Polarisation fluctuation • Mie scattering • Photon absorption

Reflection Index

Different materials

Rain and snow

Heavy rainfall (15 cm/hour) >20–30 dB/km of loss Light snow ∼3 dB/km power loss Blizzard - >60 dB/km power loss High winds

Blocking

• Furniture • Moving objects and people Walls Birds

• Loss of signal • Multipath-induced distortions • Low power due to beam divergence and spreading • Short-term loss of signal

• Link loss • Increased BER

Indoor • Adaptive optics • Increase transmit optical power

FSO

• Adaptive optics • Spatial diversity • Mesh architectures: using diverse routes • Ring topology: User’s n/w become nodes at least one hop away from the ring • Fixed tracking (short buildings) • Active tracking (tall buildings) • Hybrid RF/FSO, where RF system facilitates coarse acquisition and tracking. An RF channel serves as a control channel for FSO data link Diffuse link Cellular system Multi-beam Hybrid FOS/RF

FSO

Indoor Indoor/FSO Indoor/FSO Indoor/FSO

Optical Wireless Communications

Pointing stability and building sway

Higher losses due to reflected surfaces Photon absorption

FSO

Bit error

• Noise • Multipath propagation • Interference

Higher link availability at low speed Bit error rate

Networking Aerosols, gases, and smoke Diming Optical source nonlinearity Non-uniform power distribution

Mostly LEDs P(V)-I

• Mie scattering • Photon absorption • Rayleigh scattering • Reduced light intensity • Higher BER • High BER • Limited dynamic range

Indoor (LOS) Indoor /FSO

All FSO

Indoor (mostly VLC) Indoor/FSO

Indoor (diffuse)

31

Tracking Using a narrow beam Tx and a narrow FOV Rx • FEC • Block code • Convolutional Code • Turbo Code • QoS techniques—Differentiated services protocols sort data by priority to counter capacity changes. Application layer QoS algorithms prioritize data • Increase transmit power • Diversity techniques • Hybrid FSO-RF • PPM and PWM—For single carrier modulations [134], [135] • Adjusting the DC bias, and PWM—Multi-carrier schemes [136], [137] • Reduced modulation index • Digital modulation and data formats • Pre- and post-distortion • Spot diffusing Tx and fly-eye Rxs Study power budget, ambient light interference. • Direction diversity in the IR Rx—uses a cluster of arrow FOV detectors oriented in different directions to improve the sensitivity of rotation • Holographic diffusers • Angle diversity—using multiple narrow beam Tx and multiple non-imaging Rxs. High ambient light rejection and reduced multipath distortion due to narrow FOV detector • Imaging diversity reception—A single imaging lens and a photodetector segmented into multiple pixels • Reduced ambient light • Multi-beam Tx and imaging diversity Rxs • Multi-spot diffusing—combine the advantages of LOS and diffuse systems using holographic spot array generators. Decreased delay spread and also reduction in the Ts power requirements • Line strip spot diffusing Tx—A multi-beam Tx located on the floor produces multiple diffusing spots on the middle of the ceiling in the form of a line strip. Reduces ISI and improved SNR • Multi-Tx broadcast system—A number of Txs placed in a grid near the ceiling of the room • A square grid located near the room ceiling

Optical Wireless Communication Systems

Roaming

32

Optical Wireless Communications

REFERENCES 1. Cisco, “Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2016–2021,” Cisco, 2012. [Online]. Available: https://www.cisco.com/c/en/us/solutions/collateral/service-provider/ visual-networking-index-vni/mobile-white-paper-c11-520862.pdf%0A. 2. 5G Forum, “5G New Wave Towards Future Societies in the 2020s,” no. March, 2015. 3. M. Fiorarli et al., “Challenges for 5G Transport Networks,” 2014 IEEE Int. Conf. Adv. Networks Telecommun. Syst. ANTS 2014, pp. 1–6, 2014. 4. X. Xu et al., “Advanced Modulation Formats for 400-Gbps Short-reach Optical Inter-Connection,” Opt. Express, vol. 23, no. 1, p. 492, 2015. 5. M. Song, E. Pincemin, V. Vgenopoulou, I. Roudas, E.-M. Amhoud, and Y. Jaouen, “Transmission Performances of 400 Gbps Coherent 16-QAM Multi-Band OFDM Adopting Nonlinear Mitigation Techniques,” in 2015 Tyrrhenian International Workshop on Digital Communications (TIWDC), 2015, pp. 46–48. 6. T. Mizuno et al., “32-Core Dense SDM Unidirectional Transmission of PDM-16QAM Signals Over 1600 km Using Crosstalk-Managed Single-Mode Heterogeneous Multicore Transmission Line,” Opt. Fiber Commun. Conf. Postdeadline Pap., p. Th5C.3, 2016. 7. S. Zhang et al., “Capacity-Approaching Transmission over 6375 km Using Hybrid Quasi-Single-Mode Fiber Spans,” J. Light. Technol., vol. 35, no. 3, pp. 481–487, Feb. 2017. 8. Petri Mähönen and Zach Shelby, “Broadband Wireless Access over Two-Layer LMDS with an IEEE 802.16 Future.” [Online]. Available: https://www.isoc.org/inet2000/cdproceedings/posters/168/index. htm. [Accessed: 01-Jul-2018]. 9. M. of I. A. and Communications, “Liaison and Coordination Work for International Standardization in the Business Year 2013.” [Online]. Available: https://www.ituaj.jp/wpcontent/%0Auploads/2013/10/ nb25-4_web-3_po-liaison.pdf%0A. 10. M. Marcus and B. Pattan, “Millimeter Wave Propagation: Spectrum Management Implications,” IEEE Microw. Mag., vol. 6, no. 2, pp. 54–62, 2005. 11. L. Hanzo, H. Haas, S. Imre, D. O’Brien, M. Rupp, and L. Gyongyosi, “Wireless Myths, Realities, and Futures: From 3G/4G to Optical and Quantum Wireless,” Proc. IEEE, vol. 100, no. Special Centennial Issue, pp. 1853–1888, May 2012. 12. A. Fehske, G. Fettweis, J. Malmodin, and G. Biczok, “The Global Footprint of Mobile Communications: The Ecological and Economic Perspective,” IEEE Commun. Mag., vol. 49, no. 8, pp. 55–62, 2011. 13. K. Samdanis, Green Communications: Principles, Concepts and Practice. 14. S. Arnon, Visible Light Communication. Cambridge: Cambridge University Press, 2015. 15. Z. Ghassemlooy, L. N. Alves, S. Zvanovec, and M.-A. Khalighi, Visible Light Communications: Theory and Applications. CRC Press, 2017. 16. “Wi-Fi Enabled Device Shipments Will Exceed 1.5 Billion in 2012, Almost Double that Seen in 2010, Says ABI Research | Business Wire.” [Online]. Available: https://www.businesswire.com/ news/home/20121011006017/en/Wi-Fi-Enabled-Device-Shipments-Exceed-1.5-Billion. [Accessed: 01-Jul-2018]. 17. C. Na, J. Chen, and T. Rappaport, “Measured Traffic Statistics and Throughput of IEEE 802.11b Public WLAN Hotspots with Three Different Applications,” IEEE Trans. Wirel. Commun., vol. 5, no. 10, pp. 3296–3305, Nov. 2006. 18. R. Bancroft, “A Commercial Perspective on the Development and Integration of an 802.11a/b/g Hiperlan/ WLAN Antenna into Laptop Computers,” IEEE Antennas Propag. Mag., vol. 48, no. 4, pp. 12–18, Aug. 2006. 19. R. Ahola et al., “A Single-Chip CMOS Transceiver for 802.11a/b/g Wireless LANs,” IEEE J. SolidState Circuits, vol. 39, no. 12, pp. 2250–2258, Dec. 2004. 20. C. Cordeiro, “The Pursuit of Tens of Gigabits per second Wireless Systems [Industry Perspectives],” IEEE Wirel. Commun., vol. 20, no. 1, pp. 3–5, Feb. 2013. 21. “IEEE 802.11ay task group.” [Online]. Available: http://www.ieee802.org/11/Reports/tgay_update.htm. [Accessed: 01-Jul-2018]. 22. A. J. Paulraj and C. B. Papadias, “Space-Time Processing for Wireless Communications,” IEEE Signal Process. Mag., vol. 14, no. 6, pp. 49–83, 1997. 23. A. Maltsev et al., “MIMO and Multihop Cross-Layer Design for Wireless Backhaul: A Testbed Implementation,” IEEE Commun. Mag., vol. 48, no. 3, pp. 172–179, Mar. 2010. 24. C. Aichele et al., Wireless Networking in the Developing World. 2013. 25. A. Burton, Smart White LED Based Lighting and Communications. Northumbria University, 2015.

9781498742696_Ch01.indd 32

17/04/19 6:16 PM

Optical Wireless Communication Systems

33

26. A. Maltsev et al., “Channel Models for 60 GHz WLAN Systems, IEEE Document 802.11-09/0334r8”. 27. E. Torkildson, C. Sheldon, U. Madhow, and M. Rodwell, “Nonuniform Array Design for Robust Millimeter-Wave MIMO Links,” in GLOBECOM 2009 – 2009 IEEE Global Telecommunications Conference, 2009, pp. 1–7. 28. E. Torkildson, U. Madhow, and M. Rodwell, “Indoor Millimeter Wave MIMO: Feasibility and Performance,” IEEE Trans. Wirel. Commun., vol. 10, no. 12, pp. 4150–4160, Dec. 2011. 29. B. Janjua et al., “Going beyond 4 Gbps Data Rate by Employing RGB Laser Diodes for Visible Light Communication,” Opt. Express, vol. 23, no. 14, pp. 18746–18753, 2015. 30. R. Szweda, “Lasers for Free-Space Optical Communications,” III-Vs Rev., vol. 14, no. 8, pp. 46–49, Oct. 2001. 31. M. Courtney, “Whatever Happened to Broadband over Power Line?,” Eng. Technol., vol. 8, no. 10, pp. 72–75, Nov. 2013. 32. Intel, “Ultra-Wideband (UWB Technology) the Future Technology Enabling High-Speed Wireless Personal Area Networks.” [Online]. Available: http://ecee.colorado.edu/∼ecen4242/marko/UWB/ UWB/Intel’ UWB.pdf. 33. F. Halsall, Data Communications, Computer Networks and Open Systems, Fourth. Addison Wesley, 1995. 34. I. of E. and E. Engineers, “IEEE Standard 802.11a-1999.” 1999. 35. M. Johnsson, “HiperLAN/2 - The Broadband Radio Transmission Technology Operating in the 5 GHz Frequency Band,” HiperLAN/2 Glob. Forum, 1999. 36. G. Kalfas, S. Markou, D. Tsiokos, C. Verikoukis, and N. Pleros, “Very High Throughput 60GHz Wireless Enterprise Networks over GPON Infrastructure,” in 2013 IEEE International Conference on Communications Workshops (ICC), 2013, pp. 873–878. 37. T. Kanesan, The Experimental Design of Radio-over-Fibre System for 4G Long Term Evolution. Northumbria University, 2013. 38. H. Al-Raweshidy and S. Komaki, Radio over Fiber Technologies for Mobile Communications Networks. Artech House, 2002. 39. L. Noel, D. Wake, D. G. Moodie, D. D. Marcenac, L. D. Westbrook, and D. Nesset, “Novel Techniques for High-Capacity 60-GHz Fiber-Radio Transmission Systems,” IEEE Trans. Microw. Theory Tech., vol. 45, no. 8, pp. 1416–1423, 1997. 40. H. Ogawa, D. Polifko, and S. Banba, “Millimeter-Wave Fiber Optics Systems for Personal Radio Communication,” IEEE Trans. Microw. Theory Tech., vol. 40, no. 12, pp. 2285–2293, 1992. 41. A. J. Cooper, “‘Fibre/Radio’ for the Provision of Cordless/Mobile Telephony Services in the Access Network,” Electron. Lett., vol. 26, no. 24, pp. 2054–2056, 1990. 42. T. Jia, S. Zheng, X. Zhang, X. Jin, X. Ai, and J. Xu, “Characteristics of Radio Transmission over Polymer Optical Fiber for Indoor Wireless Coverage,” Opt. Commun., vol. 264, no. 1, pp. 142–147, Aug. 2006. 43. F. R. Gfeller and U. Bapst, “Wireless In-House Data Communication via Diffuse Infrared Radiation,” Proc. IEEE, vol. 67, no. 11, pp. 1474–1486, 1979. 44. R. Paudel, Z. Ghassemlooy, H. Le-Minh, S. Rajbhandari, and B. Livingstone, “Investigation of FSO Ground-to-Train Communications in a Laboratory Environment,” in 2011 Second Asian Himalayas International Conference on Internet (AH-ICI), 2011, pp. 1–5. 45. D. Kedar and S. Arnon, “Urban Optical Wireless Communication Networks: The Main Challenges and Possible Solutions,” IEEE Commun. Mag., vol. 42, no. 5, pp. s2–s7, 2004. 46. D. C. O’Brien, L. Zeng, H. Le-Minh, G. Faulkner, J. W. Walewski, and S. Randel, “Visible Light Communications: Challenges and Possibilities,” IEEE Int. Symp. Pers. Indoor Mob. Radio Commun. PIMRC, pp. 1–5, 2008. 47. Y. Z. Y. Zheng and M. Z. M. Zhang, “Visible Light Communications-Recent Progresses and Future Outlooks,” Photonics Optoelectron. (SOPO), 2010 Symp., 2010. 48. J. P. Conti, “What You See Is What You Send - [Comms Visible Light],” Eng. Technol., vol. 3, no. 19, pp. 66–69, 2008. 49. P. Haigh, Using Equalizers to Increase Data Rates in Organic Photonic Devices for Visible Light Communications Systems. Northumbria University, 2014. 50. J. R. Barry, Wireless Infrared Communications. Boston: Kluwer Academic Publishers, 1994. 51. R. Ramirez-Iniguez, S. M. Idrus, and Z. Sun, Optical Wireless Communications: IR for Wireless Connectivity. NW: Auerbach Publications, 2008. 52. E. Ciaramella et al., “1.28 Terabit/s (32x40 Gbit/s) wdm Transmission System For Free Space Optical Communications,” IEEE J. Sel. Areas Commun., vol. 27, no. 9, pp. 1639–1645, 2009. 53. F. E. E. Goodwin, “A Review of Operational Laser Communication Systems,” Proc. IEEE, vol. 58, no. 10, pp. 1746–1752, 1970.

34

Optical Wireless Communications

54. J. M. Kahn and J. R. Barry, “Wireless Infrared Communications,” Proc. IEEE, vol. 85, no. 2, pp. 265– 298, 1997. 55. J. B. Carruthers and J. M. Kahn, “Angle Diversity for Nondirected Wireless Infrared Communication,” IEEE Trans. Commun., vol. 48, no. 6, pp. 960–969, 2000. 56. A. H. Azhar, T. Tran, and D. C. O’Brien, “A Gigabit/s Indoor Wireless Transmission Using MIMOOFDM Visible-Light Communications,” IEEE Photonics Technol. Lett., vol. 25, no. 2, pp. 171–174, 2013. 57. W. Ke, A. Nirmalathas, C. Lim, and E. Skafidas, “4 × 12.5 Gb/s WDM Optical Wireless Communication System for Indoor Applications,” Light. Technol. J., vol. 29, no. 13, pp. 1988–1996. 58. S. Arnon, “Optimization of Urban Optical Wireless Communication Systems,” IEEE Trans. Wirel. Commun., vol. 24, no. 5, pp. 626–629, May 2003. 59. M. Noshad and M. Brandt-Pearce, “Can Visible Light Communications Provide Gb/s Service?,” arXiv:1308.3217, pp. 1–7, 2013. 60. J. Gancarz, H. Elgala, and T. D. C. Little, “Impact of Lighting Requirements on VLC Systems,” IEEE Commun. Mag., vol. 51, no. 12, pp. 34–41, 2013. 61. L. Grobe et al., “High-Speed Visible Light Communication Systems,” IEEE Commun. Mag., vol. 51, no. 12, pp. 60–66, 2013. 62. S. Rajbhandari et al., “A Review of Gallium Nitride LEDs for Multi-Gigabit-per-Second Visible Light Data Communications,” Semicond. Sci. Technol., vol. 32, no. 2, p. 023001, 2017. 63. “Global Visible Light Communication (VLC)/Li-Fi Technology Market Worth $6,138.02 Million by 2018,” MarketsandMarkets, 2013. [Online]. Available: http://www.prnewswire.com/news-releases/globalvisible-light-communication-vlcli-fi-technology-market-worth-613802-­m illion-by-2018-186272641. html. [Accessed: 21-Sep-2016]. 64. S. Rajagopal, R. D. Roberts, and L. Sang-Kyu, “IEEE 802.15.7 Visible Light Communication: Modulation Schemes and Dimming Support,” IEEE Commun. Mag., vol. 50, no. 3, pp. 72–82, 2012. 65. IEEE Computer Society. LAN/MAN Standards Committee, Institute of Electrical and Electronics Engineers, and IEEE-SA Standards Board, IEEE Standard for Local and Metropolitan Area Networks. Part 15.7, Short-Range Wireless Optical Communication Using Visible Light. Institute of Electrical and Electronics Engineers, 2011. 66. C. X. Wang et al., “Cellular Architecture and Key Technologies for 5G Wireless Communication Networks,” IEEE Commun. Mag., vol. 52, no. 2, pp. 122–130, 2014. 67. “Free Space Optics (FSO) and Visible Light Communication (VLC) Market - Global Forecast to 2022.” [Online]. Available: https://www.marketsandmarkets.com/PressReleases/visible-light-communication. asp. [Accessed: 02-Jul-2018]. 68. A. J. C. Moreira, R. T. Valadas, and A. M. de O. Duarte, “Optical Interference Produced by Artificial Light,” Wirel. Networks, vol. 3, no. 2, pp. 131–140, 1997. 69. “Infrared Data Association Releases IrDA Global Market Report 2007,” 2007. [Online]. Available: http://www.irda.org. 70. S. Bloom and W. S. Hartley, “The Last-Mile Solution: Hybrid FSO Radio,” in AirFiber, Inc. White paper. 71. H. Chun et al., “LED Based Wavelength Division Multiplexed 10 Gb/s Visible Light Communications,” J. Light. Technol., vol. 34, no. 13, pp. 3047–3052, 2016. 72. G. Cossu, A. M. Khalid, P. Choudhury, R. Corsini, and E. Ciaramella, “3.4 Gbit/s Visible Optical Wireless Transmission Based on RGB LED,” Opt. Express, vol. 20, pp. B501–B506, 2012. 73. Z. Shuailong et al., “1.5 Gbit/s Multi-Channel Visible Light Communications Using CMOS-Controlled GaN-Based LEDs,” J. Light. Technol., vol. 31, no. 8, pp. 1211–1216, 2013. 74. D. Tsonev et al., “A 3-Gb/s Single-LED OFDM-Based Wireless VLC Link Using a Gallium Nitride µLED,” IEEE Photonics Technol. Lett., vol. 26, no. 7, pp. 637–640, 2014. 75. H. Y. Hsu, W. C. Lu, H. Le Minh, Z. Ghassemlooy, Y. L. Yu, and S. K. Liaw, “2 × 80 Gbit/s DWDM Bidirectional Wavelength Reuse Optical Wireless Transmission,” IEEE Photonics J., vol. 5, no. 4, 2013. 76. A. M. Street, P. N. Stavrinou, D. C. Obrien, and D. J. Edwards, “Indoor Optical Wireless Systems—A Review,” Opt. Quantum Electron., vol. 29, pp. 349–378, 1997. 77. P. S. Peter, L. E. Philip, T. D. Kieran, R. W. David, M. Paul, and W. David, “Optical Wireless: A Prognosis,” in Proc. SPIE, 1995, vol. 2601, pp. 212–225. 78. O. Bouchet, H. Sizun, C. Boisrobert, F. De Fornel, and P. Favennec, Free-Space Optics: Propagation and Communication. London: ISTE Ltd, 2006. 79. D. Wu, Z. Ghassemlooy, H. Le Minh, S. Rajbhandari, and W. Lim, “Optimisation of Transmission Bandwidth for Indoor Cellular OWC System Using a Dynamic Handover Decision-Making Algorithm,” in Proceedings of the 2012 8th International Symposium on Communication Systems, Networks and Digital Signal Processing, CSNDSP 2012, 2012.

Optical Wireless Communication Systems

35

80. D. C. O’Brien et al., “Integrated Transceivers for Optical Wireless Communications,” IEEE J. Sel. Top. Quantum Electron., vol. 11, no. 1, pp. 173–183, 2005. 81. S. Hranilovic and F. R. Kschischang, “A Pixelated MIMO Wireless Optical Communication System,” IEEE J. Sel. Top. Quantum Electron, vol. 12, no. 4, pp. 859–874, 2006. 82. S. Rajbhandari et al., “A Multigigabit per Second Integrated Multiple-Input Multiple-Output VLC Demonstrator,” J. Light. Technol., vol. 35, no. 20, pp. 4358–4365, Oct. 2017. 83. J. B. Carruthers and S. M. Carroll, “Statistical Impulse Response Models for Indoor Optical Wireless Channels,” Int. J. Commun. Syst., vol. 18, no. 3, pp. 267–284, 2005. 84. T. Borogovac, M. Rahaim, and J. B. Carruthers, “Spotlighting for Visible Light Communications and Illumination,” in 2010 IEEE Globecom Workshops, 2010, pp. 1077–1081. 85. D. Wu, M. A. Khalighi, and X. Tang, “Indoor Non-Directed Optical Wireless Communications Optimization of the Lambertian Order,” J. Electr. Comput. Eng. Innov., vol. 1, no. 1, pp. 1–9, 2013. 86. G. Cossu, A. Wajahat, R. Corsini, and E. Ciaramella, “5.6 Gbit/s Downlink and 1.5 Gbit/s Uplink Optical Wireless Transmission at Indoor Distances (≥1.5 m),” in Optical Communication (ECOC), 2014 European Conference on, 2014, pp. 1–3. 87. M. D. D. Kotzin and A. P. P. van den Heuvel, “A Duplex Infra-Red System for In-Building Communications,” in 36th IEEEVehicular Technology Conference, 1986, vol. 36, pp. 179–185. 88. J. M. Kahn, W. J. Krause, and J. B. Carruthers, “Experimental Characterization of Non-Directed Indoor Infrared Channels,” IEEE Trans. Commun., vol. 43, no. 234, pp. 1613–1623, 1995. 89. P. Barker, A. C. Boucouvalas, and V. Vitsas, “Simulation Analysis of Advanced Infrared (AIr) MAC Wireless Communications Protocol,” IEE Proc.-Circuits, Devices Syst., vol. 149, no. 3, pp. 193–197, Jun. 2002. 90. A. C. Boucouvalas, P. Chatzimisios, Z. Ghassemlooy, M. Uysal, and K. Yiannopoulos, “Standards for Indoor Optical Wireless Communications,” IEEE Commun. Mag., vol. 53, no. 3, pp. 24–31, 2015. 91. V. Vitsas and A. C. Boucouvalas, “Performance Analysis of the Advanced Infrared (AIr) CSMA/CA MAC Protocol for Wireless LANs,” Wirel. Networks, vol. 9, no. 5, pp. 495–507, 2003. 92. G. Yun and M. Kavehrad, “Indoor Infrared Wireless Communications Using Spot Diffusing and FlyEye Receivers,” Can. J. Electr. Comput. Eng., vol. 18, no. 4, pp. 151–157, Oct. 1993. 93. M. Kavehrad and S. Jivkova, “Indoor Broadband Optical Wireless Communications: Optical Subsystems Designs and Their Impact on Channel Characteristics,” IEEE Wirel. Commun., vol. 10, no. 2, pp. 30–35, 2003. 94. A. G. Al-Ghamdi and J. M. H. Elmirghani, “Line Strip Spot-Diffusing Transmitter Configuration for Optical Wireless Systems Influenced by Background Noise and Multipath Dispersion,” IEEE Trans. Commun., vol. 52, no. 1, pp. 37–45, Jan. 2004. 95. G. Yun and M. Kavehrad, “Spot-Diffusing and Fly-Eye Receivers for Indoor Infrared Wireless Communications,” in 1992 IEEE International Conference on Selected Topics in Wireless Communications, pp. 262–265. 96. J. M. Kahn et al., “Imaging Diversity Receivers for High-Speed Infrared Wireless Communication,” IEEE Commun. Mag., vol. 36, no. 12, pp. 88–94, 1998. 97. A. Sivabalan and J. John, “Improved Power Distribution in Diffuse Indoor Optical Wireless Systems Employing Multiple Transmitter Configurations,” Opt. Quantum Electron., vol. 38, no. 8, pp. 711–725, 2006. 98. M. T. Alresheedi and J. M. H. Elmirghani, “Adaptive 10 Gbit/s Mobile Optical Wireless Systems Employing Beam Delay, Angle and Power Adaptation with Imaging Receivers,” in 2011 IEEE Global Telecommunications Conference - GLOBECOM 2011, 2011, pp. 1–6. 99. M. T. Alresheedi and J. M. H. Elmirghani, “Hologram Selection in Realistic Indoor Optical Wireless Systems with Angle Diversity Receivers,” J. Opt. Commun. Netw., vol. 7, no. 8, p. 797, Aug. 2015. 100. C. Chen, W. De Zhong, D. Wu, and Z. Ghassemlooy, “Wide-FOV and High-Gain Imaging Angle Diversity Receiver for Indoor SDM-VLC Systems,” IEEE Photonics Technol. Lett., vol. 28, no. 19, pp. 2078–2081, 2016. 101. F. Alsaadi and J. Elmirghani, “Performance Evaluation of 2.5 Gbit/s and 5 Gbit/s Optical Wireless Systems Employing a Two Dimensional Adaptive Beam Clustering Method and Imaging Diversity Detection,” IEEE J. Sel. areas Commun., vol. 27, no. 8, pp. 1507–1519, 2009. 102. A. G. Al-Ghamdi and J. M. H. Elmirghani, “Multiple Spot Diffusing Geometries for Indoor Optical Wireless Communication Systems,” Int. J. Commun. Syst., vol. 16, no. 10, pp. 909–922, 2006. 103. A. T. Hussein and J. M. H. Elmirghani, “10 Gbps Mobile Visible Light Communication System Employing Angle Diversity, Imaging Receivers, and Relay Nodes,” J. Opt. Commun. Netw., vol. 7, no. 8, p. 718, 2015.

36

Optical Wireless Communications

104. R. T. Valadas, A. M. R. Tavares, and A. M. Duarte, “Angle Diversity to Combat the Ambient Noise in Indoor Optical Wireless Communication Systems,” Int. J. Wirel. Inf. Networks, vol. 4, no. 4, pp. 275–288, 1997. 105. S. Jivkova and M. Kavehrad, “Receiver Designs and Channel Characterization for Multi-Spot High-BitRate Wireless Infrared Communications,” IEEE Trans. Commun., vol. 49, no. 12, pp. 2145–2153, 2001. 106. V. Jungnickel, C. V. Helmolt, and U. Krüger, “Broadband Wireless Infrared LAN Architecture Compatible with Ethernet Protocol,” Electron. Lett., vol. 34, no. 25, p. 2371, 1998. 107. P. Djahani and J. M. Kahn, “Analysis of Infrared Wireless Links Employing Multibeam Transmitters and Imaging Diversity Receivers,” IEEE Trans. Commun., vol. 48, no. 12, pp. 2077–2088, 2000. 108. S. T. Jovkova and M. Kavehard, “Multispot Diffusing Configuration for Wireless Infrared Access,” IEEE Trans. Commun., vol. 48, no. 6, pp. 970–978, Jun. 2000. 109. R. D. Wisely, “A 1 Gbit/s Optical Wireless Tracked Architecture for ATM Delivery,” in IEE Colloquium on Optical Free Space Communication Links, 1996, p. 14/1-14/7. 110. V. Jungnickel, T. Haustein, A. Forck, and C. von Helmolt, “155 Mbit/s Wireless Transmission with Imaging Infrared Receiver,” Electron. Lett., vol. 37, no. 5, pp. 314–315, 2001. 111. M. J. McCullagh and D. R. Wisely, “155 Mbit/s Optical Wireless Link Using a Bootstrapped Silicon APD Receiver,” Electron. Lett., vol. 30, no. 5, pp. 430–432, 1994. 112. M. Castillo-Vazquez and A. Puerta-Notario, “Single-Channel Imaging Receiver for Optical Wireless Communications,” IEEE Commun. Lett., vol. 9, no. 10, pp. 897–899, Oct. 2005. 113. V. Jungnickel, A. Forck, T. Haustein, U. Kruger, V. Pohl, and C. von Helmolt, “Electronic Tracking for Wireless Infrared Communications,” IEEE Trans. Wirel. Commun., vol. 2, no. 5, pp. 989–999, Sep. 2003. 114. L. C. Andrews and R. L. Phillips, Laser Beam Propagation Through Random Media, Second. Washington: SPIE Press, 2005. 115. A. C. Boucouvalas, “Indoor Ambient Light Noise and Its Effect on Wireless Optical Links,” IEE Proc. - Optoelectron., vol. 143, no. 6, pp. 334–338, 1996. 116. J. R. Barry, J. M. Kahn, W. J. Krause, E. A. Lee, and D. G. Messerschmitt, “Simulation of Multipath Impulse Response for Indoor Wireless Optical Channels,” IEEE J. Sel. areas Commun., vol. 11, no. 3, pp. 367–379, 1993. 117. K. K. Wong, T. O’Farrell, and M. Kiatweerasakul, “The Performance of Optical Wireless OOK, 2-PPM and Spread Spectrum under the Effects of Multipath Dispersion and Artificial Light Interference,” Int. J. Commun. Syst., vol. 13, no. 7–8, pp. 551–557, 2000. 118. A. Nuwanpriya, S.-W. Ho, and C. S. Chen, “Indoor MIMO Visible Light Communications: Novel Angle Diversity Receivers for Mobile Users,” IEEE J. Sel. Areas Commun., vol. 8716, no. c, pp. 1–1, 2015. 119. S. Rajbhandari, Z. Ghassemlooy, and M. Angelova, “Effective Denoising and Adaptive Equalization of Indoor Optical Wireless Channel with Artificial Light Using the Discrete Wavelet Transform and Artificial Neural Network,” IEEE/OSA J. Light. Technol., vol. 27, no. 20, pp. 4493–4500, 2009. 120. M. Grabner and V. Kvicera, “The Wavelength Dependent Model of Extinction in Fog and Haze for Free Space Optical Communication,” Opt. Express, vol. 19, no. 4, pp. 3379–3386, 2011. 121. B. Flecker, E. Leitgeb, S. Sheikh Muhammad, C. Chlestil, E. Duca, and V. Carrozzo, “Measurement of Light Attenuation in Fog and Snow Conditions for Terrestrial FSO Links,” 15th IST Mob. Wirel. Commun. Summit, 2006. 122. M. A. Naboulsi, H. Sizun, and F. d Fornel, “Wavelength Selection for the Free Space Optical Telecommunication Technology,” in SPIE, 2004, vol. 5465, pp. 168–179. 123. I. I. Kim, B. McArthur, and E. Korevaar, “Comparison of Laser Beam Propagation at 785 nm and 1550 nm in Fog and Haze for Optical Wireless Communications,” SPIE Proceeding Opt. Wirel. Commun. III, vol. 4214, pp. 26–37, 2001. 124. R. Martini et al., “Free-Space Optical Transmission of Multimedia Satellite Data Streams Using MidInfrared Quantum Cascade Lasers,” Electron. Lett., vol. 38, no. 4, p. 181, 2002. 125. S. Blaser, D. Hofstetter, M. Beck, and J. Faist, “Free-Space Optical Data Link Using Peltier-Cooled Quantum Cascade Laser,” Electron. Lett., vol. 37, no. 12, p. 778, 2001. 126. H. Le Minh et al., “High-Speed Visible Light Communications Using Multiple-Resonant Equalization,” IEEE Photonics Technol. Lett., vol. 20, no. 14, pp. 1243–1245, 2008. 127. A. M. Khalid, G. Cossu, R. Corsini, P. Choudhury, and E. Ciaramella, “1-Gb/s Transmission over a Phosphorescent White LED by Using Rate-Adaptive Discrete Multitone Modulation,” IEEE Photonics J., vol. 4, no. 5, pp. 1465–1473, 2012. 128. A. C. Boucouvalas, “IEC 825-1 Eye Safety Classification of Some Consumer Electronic Products,” in IEE Colloquium on Optical Free Space Communication Links, 1996, vol. 1996, pp. 13–13.

9781498742696_Ch01.indd 36

17/04/19 4:34 PM

Optical Wireless Communication Systems

37

129. IEC, “Safety of Laser Products - Part 1: Equipment Classification and Requirements,” no. http://webstore. iec.ch/webstore/webstore.nsf/Standards/IEC%2060825-1?openDocument. International Electrotechnical Commission, 2007. 130. D. Heatley, D. R. Wisely, and P. Cochrane, “Optical Wireless: The Story So Far,” IEEE Commun. Mag., vol. 36, no. 12, pp. 72–82, 1998. 131. S. Bloom, E. Korevaar, J. Schuster, and H. Willebrand, “Understanding the Performance Of Free-Space Optics,” J. Opt. Netw., vol. 2, no. 6, pp. 178–200, 2003. 132. S. Nazhan and Z. Ghassemlooy, “Chaotic Signal Dynamics of VCSEL for Secure Optical Communication,” pp. 4–9. 133. A. P. Tang and J. M. Kahn, “Wireless Infrared Communication Links Using Multi-Beam Transmittersand Imaging Receivers,” in IEEE International Conference on Communications, 1996, vol. 1, pp. 180–186. 134. X. You, J. Chen, H. Zheng, and C. Yu, “Efficient Data Transmission Using MPPM Dimming Control in Indoor Visible Light Communication,” IEEE Photonics J., vol. 7, no. 4, pp. 1–12, Aug. 2015. 135. K. Lee and H. Park, “Modulations for Visible Light Communications with Dimming Control,” IEEE Photonics Technol. Lett., vol. 23, no. 16, pp. 1136–1138, Aug. 2011. 136. Q. Wang, Z. Wang, and L. Dai, “Asymmetrical Hybrid Optical OFDM for Visible Light Communications with Dimming Control,” IEEE Photonics Technol. Lett., vol. 27, no. 9, pp. 974–977, May 2015. 137. H. Elgala and T. D. C. Little, “Reverse Polarity Optical-OFDM (RPO-OFDM): Dimming Compatible OFDM for Gigabit VLC links,” Opt. Express, vol. 21, no. 20, p. 24288, Oct. 2013.

2

Optical Sources and Detectors

There are a number of light sources and photodetectors (PDs) that could be used in OWC systems. The most commonly used light sources are the incoherent light-emitting diodes (LEDs) and coherent laser diodes (LDs). LEDs are mainly used for indoor OWC applications. However, for short links (e.g., up to a kilometer) and moderate data rates, it is also possible to use LEDs in place of LDs. LDs are monochromatic, coherent, and directional, and therefore they are mostly employed for outdoor applications. Particularly for long transmission links, it is crucial to direct the energy of the information to be transmitted precisely in the form of a well-collimated light beam to withstand atmospheric conditions, which can result in high attenuation. In order to limit the beam divergence, ideally, one should use diffraction-limited light sources together with relatively large high-quality optical telescopes. At the receiving end, it is also advantageous to use a high-directionality telescope not only to collect as much of the transmitted power as possible but also to reduce the background ambient light, which will introduce additional noise and thus deteriorate the link performance. As for PDs, both the PIN and the avalanche photodiode (APD) could readily be used, though the latter is costlier. This chapter discusses the types of light sources, their structures, and their optical characteristics. The process of optical direct detection, as well coherent detection, is also covered in this chapter. Different types of noise sources encountered in optical detection will be introduced, and the statistics of the optical detection process is also discussed.

2.1  LIGHT SOURCES For optical communication systems, light sources adopted must have the following key features: (i) the appropriate wavelength and a narrow linewidth (i.e., high bandwidth); (ii) fast response time (wideband) for high-speed links; (iii) narrow radiation pattern (beam width) with high radiance (i.e., power) level and a small emitting surface area mostly for outdoor applications; (iv) high energy efficiency, longer lifespan, stability (mostly for LDs), high reliability, and low cost; (v) linear power– current characteristics, which is important for analogue signal trans; and (vi) ability to be directly modulated by varying driving current and high modulation bandwidth. There are a number of light sources available, but the most commonly used in optical communications and in OWC are the LEDs and LDs, which are solid-state devices that rely on the electronic excitation of semiconductor materials for their operation [1]. The optical radiation of these luminescent devices (i.e., LED and laser) excludes any thermal radiation due to the temperature of the material, as is the case in the incandescent light sources. Both LDs and LEDs have small sizes; lower forward voltage, and thus, drive current; excellent brightness, particularly in the visible wavelengths; and have the option of emission at a single wavelength or a range of wavelengths. Which light source to choose mainly depends on the particular applications and their key features, including optical power–current characteristics, switching speed, and the beam profile. Note that the power supplied by both devices is similar (about 10–50 mW) [2]. Both LEDs and LDs can be fabricated to emit light across a wide range of wavelengths (colours) from the visible to the infrared (IR) region of the electromagnetic spectrum. These wavelengths and their corresponding energies are as shown in Figure 2.1. The visual range of the human eye only extends from 400 to 700 nm. All these wavelengths are of great interest in OWC. The IR region of the electromagnetic spectrum can be classified into the following five segments: • Near IR (IR-A (0.780–1.4 µm)—Commonly used in optical fibre communications • Short wavelength IR (IR-B (1.4–3 µm)—1530 to 1560 nm range is the dominant spectral region for long-distance telecommunications, including optical fibre and free space optics (FSO) 39

40

Optical Wireless Communications

Wavelength (m) Frequency (Hz)

104 104

106

108

102

1010 1

Radio

1014

1012

10-2 Micro wave

10-4

1016

10-6 IR

10-8 UV

1048 10-10

1020 10-12

X-Ray Gamma Ray

Visible

FIGURE 2.1  Wavelength and energy of the ultraviolet, visible, and IR portions of the electromagnetic spectrum.

• Mid-wavelength IR (IR-C (3–5 µm) • Long wavelength IR (IR-C 8–15 µm) • Far IR (15–1000 µm). The IR-A and part of the IR-B are widely adopted in both optical fibre and OWC. The green part of the visible spectrum (495–570 nm) is of particular interest in underwater OWC because of a low attenuation window within this band, whereas ultraviolet radiations are currently being explored for indoor applications, while the IR band has long been widely used in optical wireless applications, including TV remote controls, file sharing between phones and other personal devices, and pointto-point outdoor FSO links. The IR band can be further classified into the following [3]: • Near IR (0.7 to 1.0 µm)—The region closest in wavelength to the radiation detectable by the human eye. The boundary between the IR light and visible bands is not accurately defined. The human eye is noticeably less sensitive to the light above the 700 nm wavelength. • Short-wave IR (1.0–3 µm) • Mid-wave IR (3–5 µm) • Long-wave IR (8–12 or 7–14 µm) • Very long-wave IR (12 to about 30 µm). There is an alternative classification for optical communications based on the availability of light sources, optical fibre attenuation/dispersion, free space channel losses, and detector availability based on the five telecom optical wavelength bands (see Figure 2.2) as defined by the following [4]: • First transmission window: 800–900 nm—Widely used in early optical fibre communications • Second transmission window—For medium- and long-haul optical fibre communications • The O-band: 1260–1360 nm • The E-band • Third transmission window • The S-band: 1460–1530 nm—For medium and long-haul optical fibre communications • The C-band: 1530–1565 nm—The lowest attenuation and the dominant band for longdistance optical fibre communication networks. Transmissions at 1550 nm do not pass through the corneal filter and cannot harm the sensitive retina. This means that at these wavelength bands the emitted optical power could be allowed to reach values up to 10 mW [5] for OWC links [6].

41

Optical Sources and Detectors

FIGURE 2.2  Optical transmission windows.

• Fourth transmission window—For future photonic networks • The L-band: 1565–1625 nm • The U-band: 1625–1675 nm In light sources, the optical carrier is generated via the two fundamental processes of spontaneous or stimulated emission. In general, the generation of light is due to the transition of electrons from the excited state to the lower energy state. The energy difference due to the transition of the electrons leads to a radiative or a non-radiative process. The radiative processes will lead to light generation, thus optical sources, whereas the non-radiative process typically leads to the creation of heat. In both types of light source devices, the recombination of the carrier is used to provide a photon flux. In solids, the photon interacts with an electron in the following three distinct ways (see Figure 2.3) [1]: i. The photons transfer their energy to the electrons in the filled valence band. The electrons are then excited to an empty state in the conduction band. This photon–electron interaction is associated with the solar cells. Exited state E2

Exited state E2

External pump

Photon

Ground state E1 (a)

Photon

Ground state E1 (b)

Exited state E2

Photon

Photon emission

Ground state E1 (c)

FIGURE 2.3  Two-level atomic system illustrating the three fundamental processes: (a) absorption, (b) spontaneous emission, and (c) stimulated emission.

BK-TandF-9781498742696_TEXT_GHASSEMLOOY-181209-Chp02.indd 41

16/04/19 9:36 AM

42

Optical Wireless Communications

ii. An atom can be shifted from the ground state E1 to the higher excited state E2 by absorbing the energy of an incident photon or by being pumped externally. This process is called absorption and occurs in all materials under normal conditions. iii. Electrons in the conduction band E2 can spontaneously return to the lower state E1 in the valence band, releasing photons in the process. This is process is known as the spontaneous emission, which is associated with LEDs. Photons generated by the spontaneous emission will have random emission patterns (i.e., phase and polarisation states) with frequencies within a certain linewidth. iv. An incident photon causes an electron in the excited state to move back to the ground state E1. As a result, the emitted photon is in the same phase as the incident photon, thus producing coherent light. Note that the generation of coherent light is only possible if the number of electrons in E2 exceeds the number of those in E1. This process is known as population inversion and is the basis for laser operation. In a radiative recombination process, it is usually the case to assume that the emitted photons have the same energy (i.e., the energy band gap of the material). However, at temperatures above absolute zero, the additional thermal energy causes the electrons and holes in the conduction band to reside just above the band edge and holes in the valence band just below the band edge. Hence, the photon energy in a radiative recombination will be slightly higher than the band energy [1]. The frequency and the wavelength of the emitted or absorbed photon is related to the difference in the energy E as given by

E = E2 − E1 = hf =

hc (2.1) λ

where h = 6.626 × 10−34 J · s is the Planck constant, f is the frequency, c is the speed of light 3 × 108 m/s, and λ is the wavelength of the absorbed or emitted light. Both LEDs and LDs are semiconductor p-n junction devices and offer several advantages, including ease of integration due to their compact and small size; high conversion efficiency of current to optical power; high reliability; and the possibility of direct modulation at relatively high data rates.

2.2  THE LIGHT-EMITTING DIODE LEDs emit incoherent light through spontaneous emission when subjected to electronic excitation. The electronic excitation is achieved by applying a forward bias voltage across the p-n junction. The radiated photons could be in the ultraviolet (UV), visible, or IR part of the EM spectrum, depending on the energy band-gap of the semiconductor material. In LEDs, the conversion process is fairly efficient, thus resulting in very little heat compared to incandescent lights. For the spontaneous radiative recombination process, the rate of photon emission and its wavelength are given by [1]

E I ( E = hf ) ∝ E − Eg exp  −  , (2.2a)  kT  λ=

hc   µm, (2.2b) E (eV )

where Eg is the band-gap energy of the semiconductor material, k is Boltzmann’s constant, and T is the absolute temperature. Figure 2.4 illustrates the luminescence intensity as a function of the energy for the spontaneous emission process. Note that the spontaneous emission has the

43

Optical Sources and Detectors

Luminescence intensity I (hc)

Boltzmann distribution e -E/kT Density of states (E-E g)

1.8kT

Predicted Emission spectrum Eg

Eg + 0.5kT

Energy (hc) (a)

Intensity

1300–550 nm

800–900 nm 65

45

15 0 15 Wavelength (nm) (b)

45

65

FIGURE 2.4  LED: (a) the luminescence intensity as a function of the energy for spontaneous emission process and (b) spectral profile.

threshold energy Eg and a half-power width of 1.8kT, which translates into a wavelength spectra width given by [1].

∆λ =

1.8kT (2.3) hc

LEDs have a relatively wide spectral width (i.e., Δλ of 30–60 nm and ∼170 nm at λ of 850 nm and 1300 nm, respectively [1]; see Figure 2.4(b)) and transition. Photon radiation, therefore, is random, with no phase correlation between generated photons, resulting in an incoherent light source. An increase in the temperature will decrease Eg of the material. Given that, in the radiative recombination process, the emitted photon energy Eph is strongly related to Eg, the peak Eph decreases as T increases, as is illustrated in Figure 2.5. In LEDs, photons are radiated in arbitrary directions, with very few of them creating light in the desired direction, resulting in non-coherent light output that leads to a low current-to-light conversion efficiency. In Figure 2.6, the relationship between the radiated optical power and the driving current passing through the p-n junction is illustrated, which clearly shows that the radiated optical

44

Optical Wireless Communications

(a)

(b)

FIGURE 2.5  (a) GaAs diode emission spectrum at 295 K and 77 K and (b) dependence of the emission peak and half-power width on the temperature.

power increases as the driving current increases (i.e., a highly linear characteristic) prior to the saturation region at higher levels of current (more on this in Chapter 8). The current–voltage relationship of an LED is best described by the Shockley diode equation given by J = J0 e qVB / kT − 1 (2.4)



where J0 is the saturation current density, q is the charge of an electron, and VB is the bias voltage.

L re inea gi r on

Output power (W)

Saturation region

Drive current (A)

FIGURE 2.6  An illustration of the radiated optical power against the driving current of an LED.

45

Optical Sources and Detectors

TABLE 2.1 Common LED Materials and Their Optical Radiation Wavelengths LED material/substrate AlGaN/GaN InGaN/GaN ZnTe/ZnSe SiC GaP GaAs0.15P0.85 AlGaInP/GaAs GaAs0.35P0.65/GaAs GaAs0.6P0.4/GaAs GaAsP/GaAs Ga1−xAlxAs/GaAs GaAs InGaAsP/InP

Peak wavelength or range (nm) 230–350 360–525 459 470 470 589 625–700 632 650 700 650–900 910–1020 600–1600

As mentioned earlier, the peak wavelength (or colour) of the spontaneous emission depends very much on Eg of the p-n junction, which in turn depends on the semiconductor material(s) making up the junction. Table 2.1 presents the common LED materials and their respective peak radiation wavelengths or range of wavelengths.

2.2.1 LED Structure LEDs are solid-state light-emitting devices that emit less heat and last longer than other light sources, such as conventional incandescent bulbs. Also, because they are compact and can be molded as an entirely solid-state light source, they are highly impact-resistant and afford unprecedented flexibility in the design of lighting equipment. LEDs are similar to other forms of diodes, but what makes them different is that they have a transparent package that allows infrared or visible energies to pass through, and the p-n junction area can be tailored to specific applications. In the structure of an LED, there is nothing like a resonant cavity or gain medium; hence, its radiation is not going to be as intense as that of an LD. There are a number of LED structures, depending on the application and more importantly how the radiated light is generated from the p-n junction. Here, we only outline the basic structure of the planar, dome, and edge-emitting LEDs. A comprehensive discussion of all possible structures and their detailed analysis is beyond the scope of this book. Interested readers are, however, referred to [1], [2].

2.2.2  Planar and Dome LEDs The planar LED has the simplest structure, and it is fabricated by either liquid or vapourphase epitaxial process over the whole surface of a GaAs substrate [7]. The planar LED structure is shown in Figure 2.7. It emits light from all surfaces in a Lambertian emission pattern. In Lambertian emission, the optical power radiated from a unit area into a unit solid angle (otherwise called the surface irradiance) is constant. In terms of intensity, the maximum intensity, Io, in Lambertian radiation is perpendicular to the planar surface but reduces on the sides with the viewing angle θ according to expression (2.5). The Lambertian intensity distribution is illustrated in Figure 2.8.

I li (θ) = I o cos θ (2.5)

46

Optical Wireless Communications Light output

Ohmic contact p-type epitaxial layer

n-type substrate Ohmic contact

FIGURE 2.7  Planar LED structure showing light emission on all surfaces.

The structure of a typical dome LED is shown in Figure 2.9. In this structure, a hemisphere of n-type GaAs is formed around a diffused p-type region. The diameter of the dome is chosen to maximise the external quantum efficiency of the device. The geometry of the dome is such that it is much larger than the recombination area; this gives a greater effective emission area and thus reduces the radiance. Because the dome structure does not suffer as much internal reflection as the planar LED, it has a higher external quantum efficiency [7].

2.2.3 Edge-Emitting LED As discussed previously, the planar structure of an LED emits light in all directions, but there are many instances in optical communications when this is not desirable, and the light is better confined. The edge-emitting LED does just this by confining the light in a thin (50–100 µm), narrow stripe in the plane of the p-n junction. This is best illustrated by the double heterostructure aluminium gallium arsenide (AlGaAs) edge-emitting LED as depicted in Figure 2.10. The confinement is achieved by deliberately making the active layer of a higher refractive index than the surrounding materials. This then creates a waveguide (through the process of total internal

Plane surface

I(θ)

θ

Io

FIGURE 2.8  Lambertian intensity distribution.

47

Optical Sources and Detectors Light output

n-type

p-type

Active layer

Ohmic contacts

FIGURE 2.9  The basic structure of a dome (hemispherical) LED.

reflection) for the light generated at the junction to travel to both ends of the device. The concept of waveguiding results in higher efficiency for the device and narrows the beam divergence to a half-power width of about 30° in the plane perpendicular to the p-n junction. However, the radiation in the junction plane is Lambertian, with a half-power width of about 120° since there is no such waveguiding in the plane of the junction [2]. The emerging optical beam is, therefore, elliptical in shape. The confined light is channelled to both ends of the device by the waveguiding, to ensure that the light is only radiated from one face; the emitting edge is treated with an antireflection coating, while a reflector is situated at the other end. Edge-emitting LEDs have a narrow spectral width, higher internal quantum efficiency, and higher modulation bandwidth than surfaceemitting LEDs.

SiO2

p+-GaAs p-AlGaAs n-GaAs (active layer) n-AlGaAs

Light Ouput

n-GaAs

n-GaAs substrate Contact metalization

FIGURE 2.10  The structure of a double heterojunction AlGaAs edge-emitting LED.

48

Optical Wireless Communications

2.2.4 LED Efficiencies 2.2.4.1  Internal Quantum Efficiency This term relates to the conversion of carriers into photons within the device. It can, therefore, be expressed as the ratio of the number of internally emitted photons Nin-ph to the number of carrier Nca passing through the p-n junction, as given by

ηin =

N in-ph (2.6) N ca

The internal quantum efficiency can also be related to the fraction of injected carriers that recombine radiatively to the total recombination rate, and this is directly linked to the carrier lifetimes as given by ηin =

Rr Rr + Rnr

τr τ   = =  1 + nr  τr + τ nr  τr 

(2.7)

−1



where Rr and Rnr represent the radiative and non-radiative recombination rates, respectively, and τr and τ nr stand for the radiative and non-radiative lifetimes in that order. 2.2.4.2  External Quantum Efficiency This quantum efficiency is different from the internal quantum efficiency previously described in that it relates to the number of photons emitted externally by the device. As such, it is defined as the ratio of the radiative recombination to non-radiative recombination and by the absorption of the generated light by the semiconductor material, which is given by N ex − ph N ca (2.8) = ηin ηt

ηex  =



where Nex-ph is the number of photons emitted externally, and ηt = N ex − ph N in − ph denotes the optical efficiency. Note that ηt is a function of the optics in and around the device. For a typical planar semiconductor LED device, ηt could be as low as 2%. A simple expression for the estimation of the optical efficiency is given by

ηt =

1 1n (1 − cos θc ) ≈ o   (2.9) 2 4 ns

where θc is the critical angle inside the device, while no and ns are the refractive indexes of the LED’s (semiconductor) material and that of the medium, which the device is emitting light into (i.e., the ambient), respectively. Note that both the number of photons emitted externally by the optical source and indeed the external quantum efficiency depend on the loss mechanisms within the device. These losses could be from a combination of i. Critical angle loss—Due to the total internal reflection of photons incident to the surface at angles greater than the critical angle.

49

Optical Sources and Detectors

ii. Absorption loss—This accounts for the loss due to photon absorption by the LED material. For LEDs made using the opaque GaAs substrate, up to 85% of the total generated photons could be absorbed; for transparent substrate like the GaP with isoelectronic centres, the value is much lower, around 25% [1]. iii. Fresnel loss—Fresnel reflection, which is always present at the interface of materials with different indexes of refraction, causes some of the generated photons to be reflected back to the semiconductor material instead of being emitted externally. 2.2.4.3  Power Efficiency This is simply the ratio of the optical power output Po to the electrical power input Pe to the LED. That is Optical output power  Po Electrical input power  Pe N ex − ph × hf = N ca × q × V

ηp =

(2.10)

Assuming the electrical bias qV ∼ Eg = hf, then we have η p ≈ ηex. Figure 2.11 shows the power–current characteristics of both surface and edge-emitting LEDs. 2.2.4.4  Luminous Efficiency This metric is often used to describe the characteristic LED radiation pattern within the visible spectrum. The luminous efficiency ηlu basically normalises η p by a factor that is related to the eye sensitivity shown in Figure 2.12. It is hence the ratio of the luminous flux to the input electrical power. The luminous flux (lumens) is total emitted flux weighted or scaled appropriately to reflect the varying sensitivity of the human eye to different wavelengths of light, which is given by

∫

φ fl = 683 V ( λ ) Pop ( λ ) d λ (2.11)



where V(λ) is the relative eye sensitivity shown in Figure 2.11, with its value being normalised to the unity at the peak wavelength of 555 nm, and Pop (λ) is the radiation power spectrum of the LED.

Power P0 (mW)

5

Saturated region Linear region

SELED

4 3

Temperature

2 1

ELED 50

FIGURE 2.11  LEDs power versus current characteristics.

Current I (mA)

50

Optical Wireless Communications

FIGURE 2.12  Eye sensitivity function based on the 1978 CIE data. (http://www.ecse.rpi.edu/~schubert/ Light-Emitting-Diodes-dot-org)

Therefore, the luminous efficiency is given by ηlu =



=

φ fl Pe

∫

683 V ( λ ) Pop ( λ ) d λ VI

(2.12)  lm W

Figure 2.13 shows the progression of the LED luminous efficiencies with time, including that of conventional lighting [1]. 2.2.4.5  LED Modulation Bandwidth The amount of modulation bandwidth and the frequency response of an LED depends on (i) the injected current; (ii) the junction capacitance; (iii) the parasitic capacitance; (iv) the doping level in the active region; and (v) the injected carrier lifetime in the recombination region τic. The capacitance values are almost invariant while the response increases with the current [8]. As such, the effects of the aforementioned factor can be reduced by superimposing the AC signal onto a constant DC bias current. Knowing that charge carrier has a definite and finite lifetime, then LEDs have a low-pass transfer function, which is modelled as an equivalent first-order RCj low-pass filter, and is given by

H ( jω ) =

1 (2.13) 1 + jωRC j

51

Optical Sources and Detectors

FIGURE 2.13  LED luminous efficiency.

where R and Cj are the resistance and junction capacitance of the LED. Thus the 3 dB bandwidth is given by

fc =



1 (2.14) 2 πRC j

If the drive current of an LED is modulated at a frequency of ω then the relative optical power output at any given frequency is given as [9]



P(ω ) 1 = (2.15) 2 PO 1 + ( ωτic )

Note that, in the optical domain, the electrical current I is directly proportional to the optical power; thus we can define electrical bandwidth and optical bandwidth separately as  I (ω )  Bele = 20 log10    I (0) 

Bopt

 I (ω )  = 10 log10    I (0) 

Figure 2.14 shows both the optical and electrical 3 dB bandwidth.

(2.16)

52

Optical Wireless Communications P( )/Po 1

0.707 0.5

Electrical bandwidth Optical bandwidth

Frequency ( )

FIGURE 2.14  An illustration of the optical and electrical bandwidth.

From (2.12), it follows, therefore, that the inherent modulation bandwidth of an LED is essentially limited by τic. There are two ways to reduce τic and increase the modulation bandwidth. i. To increase the doping level in the recombination region. It should be noted that the lifetime has both radiative τr and non-radiative τnr components to it, which is given as 1 1 1 = + (2.17) τ ic τr τ nr



So increasing the doping level increases the internal quantum efficiency for only as long as τr is reduced. At very high doping levels, however, an excessive number of non-radiative centres is formed; hence, the increase in modulation bandwidth due to increased doping levels is accompanied by a decrease in quantum efficiency. ii. To increase the carrier density. Lasers have considerably higher modulation bandwidth because the radiative lifetime is further shortened by the stimulated emission process.

2.2.5  White LEDs Although every colour can be produced by LEDs within the visible band, white light (480–750 nm) is the most widely used light source for general illumination, which offers higher energy efficiency (up to 85%), longer lifetime, and switching capabilities compared to the existing lighting sources. The last feature opens up the opportunity for data communications, and thus the concept of visible light communications (VLC). White light emission from an LED is by a mixture of multi-colour LEDs or by the combination of phosphors with blue/UV LED emission [10]. There are different types of white LEDs. Some of the important ones follow: • Ultraviolet (UV)–based white LEDs—Fabricated with pre-coating blue/green/red phosphors onto ultraviolet LED to emit white light [10], [11]. • White phosphors (WP) LEDs—By incorporating the phosphor in the body of a blue LED (a peak wavelength of ∼450–470 nm), some of the blue light will be converted to yellow light with a fairly broad spectral power distribution; see Figure 2.15. The remaining blue light, when mixed with the yellow light, will result in white light [10]–[12]. • RGB (red-green-blue) LEDs—Mixing the light from several coloured LEDs to create a spectral power distribution that appears white. An RGB three-chip LED is a combination of three colours to produce white light with little variance in the Kelvin colour temperature [13]. While WPLEDs are a simpler choice, the RGB LEDs offer significantly enhanced features

BK-TandF-9781498742696_TEXT_GHASSEMLOOY-181209-Chp02.indd 52

16/04/19 9:36 AM

53

Optical Sources and Detectors

(a)

(b)

FIGURE 2.15  (a) Normalised optical spectrum of a white LED, and (b) actual and projected increases in the efficacy of colour-mixed (CM) and phosphor-coated (PC) LED packages. (http://www.hi-led.eu/wp-content/ themes/hiled/pdf/led_energy_efficiency.pdf)

that are suitable for communications, including higher data rates due to the unrestricted bandwidths of the individual components in the triplet and the potential for wavelength division multiplexing (WDM). However, providing a constant white colour balance is challenging, since the individual LEDs are switched on at arbitrary and uncorrelated intervals. A comparison of the optical spectra of an RGB LED and a WPLED is shown in Figure 2.15(a). The RGB LED has peaks at 450 (B), 520 (G), and 635 (R) nm. In GaN WPLEDs, emission occurs at 445 nm, while the Ce:YAG phosphor has a wide spectral emission with the peak at ∼555 nm (green). Currently, WPLEDs (phosphor conversion) are the most energy efficient LEDs, providing package efficacy greater than 130 lm/W, and they are by far the most common type currently available. However, due to additional inefficiencies associated with PC, WPLED packages are thought to have a lower potential maximum efficacy than colour-mixed LEDs; see Figure 2.15(b). LEDs can be operated at several different drive currents, with a typical baseline of 150–350 mA, but >350 mA (up to 1 A or higher) drive currents are also commonly available. Driving the LEDs at a higher current level increases the lumen output but results in a commensurate decrease in efficacy; this phenomenon is known as efficiency droop, as well as power–current non-linearity. The latter is undesirable in data transmission. LEDs require a driver, which is comprised of both a power source and electronic control circuitry, to convert line voltage to low voltage and current from AC to DC; it and may also include additional electronics for dimming and/or colour correction. Currently available LED drivers are typically about 85% efficient, with some improvement projected. 2.2.5.1  Thermal Effects A major factor in determining an LED’s lumen output is the junction temperature. As temperature increases, the light generation process becomes less efficient with reduced lumens, thus the need for a thermal management system with LED lamps. Note that, unlike driver and optical losses, the thermal effects are generally associated with LEDs.

2.3  THE LASER 2.3.1 Operating Principle of a Laser In a conventional incoherent light source like a light bulb and an LED, an atom excited to a higher state randomly emits a single photon based on a given statistical probability. This process resulting in radiation in all directions with a broad spectrum and no interrelationships between individual

54

Optical Wireless Communications

photons is best known as spontaneous emission. Einstein predicted that the excited atoms at higher levels also could be excited to return to the lower state and release their stored energy in the form of light by a process known as stimulated emission (i.e., the gain needed for continuous oscillation). In order to describe the basic principle of stimulated emission, a two-level atomic system will be considered. In this simple system, an atom in the upper-level E2 can fall into level E1 (see Figure 2.3), producing a photon by spontaneous emission. When this photon reaches another excited atom, the interaction stimulates that atom to emit a second photon. This process has two important characteristics (i) multiplicative—one photon becomes two, and so on, via multiple interactions of photons with two other excited atoms; and (ii) the two photons generated have an identical wavelength, direction, phase, and polarisation. Therefore, this ability to “amplify” light in the presence of a sufficient number of excited atoms leads to the “optical gain” needed to maintain continuous oscillation—this is the basis of the laser (i.e., Light Amplification by Stimulated Emission of Radiation).

2.3.2  Population Inversion In the two energy levels in the atomic system under consideration, under thermal equilibrium, the number of atoms N1 in the lower level E1 is greater than the number of atoms N2 in the upper level E2 as stipulated by the Boltzmann statistics given by N2 −( E2 − E1 )  = exp    (2.18) N1 kT



To attain the optical amplification necessary for lasing to take place, we require a non-­ equilibrium distribution of atoms such that the population at E1 is lower than that of E 2 (i.e. N2 > N1). This condition is called population inversion. By using an external excitation, otherwise called “pumping”, atoms from the lower level are excited into the upper level through the process of stimulated absorption. A two-level atomic system is not the best in terms of lasing action as the probability of absorption and stimulated emission are equal, providing at best equal populations in the two levels E1 and E2. A practical laser will have one or more metastable levels in between. An example is the four-level He-Ne laser illustrated in Figure 2.16. To attain population inversion, atoms are pumped from the ground state level E 0 into level E3. The atoms there then decay very rapidly into a metastable level E2 because they are unstable at E3. This increases

E3 Rapid decay E2 Lasing

1.15 μm

Pumping E1 Rapid decay E0

FIGURE 2.16  An illustration of lasing action based on a four-level He-Ne laser.

55

Optical Sources and Detectors

the population at E2, thereby creating population inversion between E2 and E1. Lasing can then take place between E2 and E1.

2.3.3 Optical Feedback and Laser Oscillation The laser cavity, or resonator, is at the heart of the system. Light amplification in the laser occurs when a photon collides with an atom in an excited energy level, thus causing the stimulated emission of a second photon, these are then fed back to release two more photons [2]. To sustain this process and maintain coherence, two mirrors (plane or curved) are placed at either end of the amplifying medium; see Figure 2.17. The lasing medium (a crystal, a semiconductor, or a gas enclosed in an appropriate confinement structure) is placed along the optical axis of the resonator. This unique axis with very high optical gain becomes also the direction of propagation of the laser beam. The optical signal is thus fed back many times whilst receiving amplification as it passes through the gain medium. A stable output is obtained at saturation when the optical gain is exactly matched by the losses incurred in the amplifying medium. In an ideal laser, all the photons in the output beam are identical, resulting in perfect directionality and monochromaticity. This determines the unique coherence and brightness of a laser source. Note that the standing wave (modes) exists at frequencies for which the following condition holds LLD =



Mirror 1 Reflectivity Rm1

λi ,  i = 1,  2,   … (2.19) 2ngm Mirror 2 Reflectivity Rm2

Photons

Gain medium

Light output

LLD (a)

Modes

Intensity

Gaussian output profile

5

3

1 1 0 Wavelength (nm) (b)

3

5

FIGURE 2.17  Laser: (a) an illustration of optical feedback and (b) a typical spectrum of multi-mode LD.

56

Optical Wireless Communications

where ngm is the refractive index of the gain medium. The mode separation is given by

δf =

c (2.20) LLD ngm

In terms of wavelength separation (longitudinal mode spacing), we have ∆λ =



λ2 (2.21) cδf

Figure 2.17(b) shows a spectral profile of a typical multi-mode laser. Note that the line-width Δf = 1/τlt, where τlt is the time window for the propagating light signal (i.e., number of cycles in a given time). While population inversion is necessary for lasing, it is not a sufficient condition. A threshold or minimum gain within the amplifying medium must be attained before lasing can occur as given by

g th = α g +

1 1 ln (2.22) 2 LLD Rm1 Rm 2

In (2.22), the loss coefficient of the gain medium in dB/km is represented by α g, and the fraction of incident power reflected by the mirrors by Rm1 and Rm2. In LDs, the forward bias voltage brings about population inversion and the stimulated emission condition necessary for lasing. Optical feedback is achieved by polishing the end faces of the p-n junction to act as mirrors while the sides are deliberately roughened to prevent any unwanted radiation. In a laser, the conversion process is fairly efficient compared to LEDs, and it has a high output power. In comparison, for an LED to radiate 1 mW of output power, up to 150 mA of forward current is required, whereas for a LD to radiate the same power, only 10 mA or less of current is needed.

2.3.4  Properties and Specifications of a Laser The key properties of a laser source follow: i. High monochromaticity—A photon’s energy determines its wavelength through the relationship E = hc/λ. An ideal laser would emit all photons with exactly the same energy, and thus the same wavelength, and it would be perfectly monochromatic. Many applications are dependent on monochromaticity. For example, in telecommunications, several lasers at slightly offset wavelengths can transmit in parallel streams of pulses down the same optical fibre without crosstalk. ii. Narrow spectral width—The spectral (line) width of a laser is extremely narrow. In fact, it is quite common for LDs to have Loss

Gain < Loss

10 pF

380 pF

1.25

7 nH 20

Spontaneous emission

Stimulated emission

Active layer

Package Chip

LED action

ITh Laser action Drive forward current(A)

(a)

(b)

FIGURE 2.18  Laser: (a) output power against drive current plot and (b) small signal model.

BK-TandF-9781498742696_TEXT_GHASSEMLOOY-181209-Chp02.indd 57

16/04/19 9:36 AM

58

Optical Wireless Communications

TABLE 2.2 Laser Diode Materials and Their Corresponding Radiation Wavelengths Laser material/substrate AlGaN GaAlN ZnSSe ZnCdSe AlGaInP/GaAs Ga0.5In0.5P/GaAs GaAlAs/GaAs GaAs/GaAs InGaAs/GaAs InGaAsP/InP InGaAsSb PbCdS Quantum cascade PbSSe PbSnTe PbSnSe

Wavelength (nm) 350–400 375–440 447–480 490–525 620–680 670–680 750–900 904 915–1050 1100–1650 2000–5000 2700–4200 3–50 µm 4.2–8 µm 6.5–30 µm 8–30 µm

photons to encounter and stimulate further emissions. This continues until the drive current is high enough to produce a population inversion, which then results in a cascade of stimulated emission as the laser crosses the threshold with a gain larger than the loss, see Figure 2.18 [14]. The electrical power needed to reach the threshold current winds up as heat in the laser, as well as a fraction of the above-threshold current that is not converted into light. The heat generated does not just go down as wasted power; it degrades the laser performance and shortens its lifespan. This explains why lower threshold lasers tend to have longer lifespans. A typical vertical cavity surface-emitting LD, for example, the OPV314 with 1.5 mW output power has about 3 mA and 12 mA of threshold and drive currents, respectively. As shown in Figure 2.18(a) the laser power–current response is linear above the threshold level. Single-mode LDs exhibit a linear response above the threshold, whereas, in multi-mode LDs, we sometimes observe a slightly nonlinear response above the threshold level due to mode-hopping. The linearity of the light sources (LED and LD) is particularly important for analogue signal transmission. The small signal electrical model of a typical LD is shown in Figure 2.18(b). As previously mentioned, the energy band gap in lasers depends on the atomic composition of the semiconductor material(s). So changing the composition of the band gap is a good way of manipulating the radiation wavelength(s). Table 2.2 shows some common laser diode materials and their corresponding radiation wavelengths. It should be mentioned that in order for the atoms in adjacent layers of a laser to line up properly or to be structurally compatible, the materials making the different layers must have nearly identical lattice constants. This is necessary to avoid strains (or defects) developing in the layers due to lattice mismatch. For LDs, both the internal and external efficiencies can be defined by equations 2.5 and 2.7.

2.3.6 The Structure of Common Laser Types 2.3.6.1  Fabry-Perot Laser The simplest laser structure is that of Fabry-Perot LD (FPLD), the structure is essentially a resonating cavity of the type shown in Figure 2.17(a). The cavity mirrors are formed by the boundary

59

Optical Sources and Detectors

between the high refractive index semiconductor crystal and the lower refractive index air. One facet is coated in order to reflect the entire light incident on it from within the cavity (the gain medium), while the other facet is deliberately made to reflect less in order to emit the light generated within the cavity. The gain curve of the FBLD is quite broad; the implication of this is that the laser emits most of its light on one wavelength—the centre wavelength λo —but also emits at other wavelengths called the longitudinal modes. These modes or sidebands are distributed and equally spaced on either side of λo. When the intensity of the FPLD is modulated, λo is no longer fixed and does shifts depending on the intensity modulation process (i.e., changing the drive current). This effect is called “mode hopping”, which occurs because the refractive index of the device varies with the temperature, where the temperature itself also varies with the drive current. This effect of mode hopping is undesirable in communication systems because it introduces intensity noise and also affects the maximum data transmission rate, since different wavelengths travel at a different velocity in single-mode fibres with high dispersion. Therefore, the FPLDs are used as a light source in short-haul optical communications where a stable, narrow radiation spectrum under variable ambience conditions is desired [15] and in other applications, such as CD players, where some variations in the centre wavelength can be tolerated. Its key characteristics are a wavelength of 850 or 1310 nm, total output power of a few mW, spectral width of 3–20 nm, mode spacing of 0.7–2 nm, highly polarised, coherence length of 1–100 mm, and a small numerical aperture (NA) (good for coupling into fibre). 2.3.6.2  Distributed Feedback (DFB) Laser To reduce the spectral width, the LD should radiate only in one longitudinal mode. The DFB LD, which is a special type of edge-emitting LD, is optimised for single-mode (single frequency) operation. The single-mode operation is achieved by incorporating a periodic structure or a Bragg grating near the active layer as depicted in Figure 2.19(a). The grating is placed so near the active layer so that the oscillating transverse mode will interact with it [5]. The grating structure usually employed is the distributed Bragg diffraction grating. As for the meaning of the term DFB laser diode, the word p-type g

Grating Active layer

Laser output

n-type

(a)

Ppeak

SMSR

(b)

FIGURE 2.19  DFB laser: (a) structure and (b) spectral profile.

60

Optical Wireless Communications

feedback emphasises that we have the means to return stimulated photons to an active medium. This is done by reflecting a portion of the light at each slope of the grating. The word distributed implies that reflection occurs not at a single point—a mirror, say, but at many points dispersed along the active region. The net result of this arrangement is that only one wave with λo is radiated. The waves that satisfy the Bragg condition of equation (2.23) are scattered off the successive corrugations and interfere constructively. The generated light wave that satisfies this Bragg condition is then reflected back to the active region of the device. It is only the reflected light that receives the feedback and optical amplification needed for lasing action.

Λg =

mg λ g (2.23) 2neff

where Λ g is the grating period required for operations at the grating wavelength λg, neff is the effective refractive index of the waveguide, and mg is the integer order of the grating. Since the radiation contains only a single longitudinal mode, the spectral width of the DFB laser is extremely narrow—see Figure 2.19(b)—which is ideal for wavelength division multiplexing for backbone optical fibre communication networks in particular at 1550 nm. A modified version of using a Bragg grating as a reflector is better known as the distributed-Bragg-reflector laser diode, where the active region is located between two Bragg gratings that work as reflectors. The key characteristics of DFB LD are wavelength, around 1550 nm; total power output, 3–50 mw; spectral width, 10–100 MHz (0.08–0.8 pm); side-mode suppression ratio (SMSR), >50 dB; coherence length, 1–100 m; and small NA (good for coupling into fibre). 2.3.6.3  Vertical Cavity Surface Emitting Laser (VCSEL) All the laser diodes discussed previously are based on edge-emitting devices, which are characterised by the significant length of their active region and their asymmetrical radiation pattern. However, there is an alternative LD known as the VCSEL, shown in Figure 2.20, where the resonant cavity is vertical as the name implies and perpendicular to the active layer. Above and below the active layer, in the vertical direction, are narrowband mirror layers. The light beam emerges from the surface of the wafer through the substrate as outlined in Figure 2.20. The narrowband mirrors are made from alternating high- and low-refractive index layers, each of which is designed to a quarter wavelength of the laser’s operating wavelength [5]. In terms of materials, the GaAs VCSELs are very popular because they are relatively easy to make. The refractive index of GaAlAs is known to vary considerably with aluminium content, thereby providing the refractive index contrast needed for the multilayer mirrors. As a result, 850 nm VCSELs developed from this process are very widely used in optical communications. More recently VCSELs have been developed using InGaAsP compounds that emit at the 1300 and 1550 nm windows [14]. The following are some of the features that distinguish VCSELs from other types of lasers: i. Optical oscillation—which is perpendicular to the surface of the thin active layer, thus radiating a circular output beam in contrast to that radiated by edge-emitting lasers. ii. The short resonant cavity (on the order of 2 µm)—ensures single-mode operations (a spectral width of a gain curve of a few nanometers) and direct modulation for data rate well above 1 Gbps [14]. iii. Lower radiated power (>10 dBm)—VCSEL devices can radiate 3 mW of output power at 10 mA forward current with an intrinsic modulation bandwidth up to 200 GHz. iv. Higher switching speed compared to edge-emitting lasers. v. Comparatively low threshold current—translates into higher efficiency and longer lifespan. vi. A very stable lasing wavelength—since it is fixed by the short (1∼1.5 wavelength thick) Fabry-Perot cavity, have a narrow spectral width 0.5Ac [23], expression (2.43) reduces to

i p (t ) = R  AL Ac cos ( ϑ L − ϑ c )  (2.44)

In homodyne detection for correct symbol detection, ϑ L (t ) must track ϑ c (t ) such that the phase noise ϑ n (t ) = ϑ c (t ) − ϑ L (t ) = 0 . This is best achieved using the optical phase-locked loop (OPLL). In coherent detection systems, the noise in the Rx is mainly dominated by OLO-induced shot noise [25]. The shot-noise-limited Rxs offer improved sensitivity, by up to 20 dB, compared to the IM-DD schemes [26]. In IM-DD systems, the optical carrier signals must be aligned at a large spacing in the optical wavelength domain. This is because of the bandwidth of the optical band-pass filter, which is 2–3 nm. Therefore, the spacing between the optical carriers should be no less than several nanometres, which corresponds to hundreds to thousands of gigahertz. By using coherent schemes, the optical carrier signal could be aligned closely at a spacing 10 times or more the data rate in the frequency domain, thus the possibility of employing frequency division multiplexing.

72

Optical Wireless Communications

2.6  PHOTODETECTION NOISE The noise sources, as well as the frequency and distortion performance of an optical wireless link, are decisive factors in determining the link performance. In line with nearly all communication systems‚ identification of the noise sources at the Rx front end, where the incoming signal contains the least power, is critical. The two primary sources of noise at the input of the Rx are signal-dependent and dark-current shot noise. Figure 2.27 illustrates a schematic diagram of a typical Rx front end with a PD as well as the noise sources.

2.6.1  Quantum Shot Noise For an ideal PD, the only significant noise that affects its performance is that associated with the quantum nature of light itself, where the by-product is that the number of photons emitted by a coherent optical source in a given time is never constant. Although for a constant-power optical source the mean number of photons generated per second is constant, the actual number of photons per second follows the Poisson distribution. This results in photon fluctuation or quantum noise. The quantum noise (also termed photon noise) is a shot noise, which is present in all photon detectors due to the random arrival rate of photons from the data-carrying optical source and the background radiation. The quantum fluctuation is also important because it dominates over the thermal fluctuations within the PD, since hf > kT, where h and f are the Planck’s constant and radiation (optical) frequency, respectively, while k and T represent the Boltzmann’s constant and the temperature. The two-sided power spectral density (PSD) of the quantum shot noise is given by Sq (ω ) = q 〈i 〉 (W Hz) (2.45)

The shot-noise current is given as

σ 2q =



∫

Sq (ω )dω (2.46)

Bandwidth

Note that (2.45) and (2.46) can apply to all noise sources. Therefore, for the PIN and APDs, the quantum shot-noise variance are given by, respectively

σ 2q − pin = 2q 〈i 〉 Bef



σ 2q − apd = 2q 〈i 〉 Bef FM 2

(2.47)



(2.48)

Vcc Pr

ishot ip

Rin

icircuit

Output

Cshot Receiver front end

FIGURE 2.27  Diagram of a front-end photodiode detector along with channel impairments.

73

Optical Sources and Detectors

where Bef is the bandwidth of the electrical filter that follows the PD, and M is the APD gain. The shot noise is, in fact, proportional to (Ntx-ph)0.5, where Nrx-ph is the equivalent number of photons incident on the PD. For coherent Rxs, the shot noise due to the optical signal and the optical LO are given by, respectively

σ 2q −Cc = 2q 〈ic 〉 M 2 BcRx (2.49)



σ 2q −CL = 2q 〈iL 〉 M 2 BcRx (2.50)

where BcRx is the bandwidth of the coherent Rx. For the heterodyne-based Rx, the LO shot noise (2.48) is much larger than the signal shot noise (2.47), therefore, the signal shot noise can be ignored.

2.6.2  Dark-Current Shot Noise and Excess Noise Since the photocurrent is proportional to the incident light power, unlike the electrical system where power is proportional to the square of the current, the quantum shot noise increases with the incident optical power (in square root fashion). The lower end of this relationship is limited by the noise due to the dark current, which is present even when there is no input light incident on the PD. It is produced by the transition of electrons from the valence to the conduction band due to causes other than photon-induced excitation; its magnitude is closely related to the energy bandgap of the PD material(s). Large band-gap materials, such as silicon (Si), indium phosphide (InP), and gallium arsenide (GaAs), show very low values of mean dark current, 〈id 〉, while for germanium (Ge), the value could be significant when they are operated at the room temperature [27]. The dark current is a combination of bulk and surface leakage currents, which contains no useful information, and thereby constitutes as a fundamental shot noise whose variances are given by

σ 2db = 2qI d M 2 FBef (2.51) σ 2ds = 2qI l Bef (2.52)



where Id and Il are the detector’s primary unmultiplied dark and the surface leakage currents, respectively. For coherent Rxs, the dark-current noise is given by σ 2D −C = 2qI d M 2 BcRx (2.53)



The dark current consists of diffusion, tunnel, leakage currents, and generation-recombination taking place in the space-charge region and is proportional to the volume of the depletion region [27]. Typical dark-current values for some PD materials are shown in Table 2.8.

TABLE 2.8 Dark-Current Values for Different Materials [17] Dark current (nA) Photodetector material Silicon Germanium InGaAs

PIN

APD

1–10 50–500 0.5–2.0

0.1–1 50–500 10–50 @ gain = 10

74

Optical Wireless Communications

PDs that employ an internal avalanche gain mechanism to boost the signal above the thermal noise of amplifier stages on the Rx exhibit what is referred to as excess noise [28]. According to [28], the excess noise in an APD is due to the multiplication process in the high-field region of the detector where each primary electron-hole can generate an additional electron through impact ionisation of the bound electrons. These additional carriers can then create still additional carriers in a cascading process. The excess noise increases with the gain. Since the gain exhibits wavelength dependence, the excess noise differs according to the incident wavelength. Similarly, if the signalgenerated photocurrent is also amplified by the gain, it illustrates that the best signal-to-noise ratio (SNR) is obtained at a certain gain. If all primary carriers were to be multiplied equally in an APD, the mean-square current gain g 2 would be equal to the mean gain g, and the excess noise factor, defined as F = g 2 g, would be equal to 1. This is the case for PIN PDs [25]. However, due to the statistical nature of the avalanche process, F is always greater than 1 in an APD and other avalanche devices. The amount of optical power incident on the surface of a PD that produces a signal at the output of the PD, which is just equal to the noise generated internally by the PD, is defined as the noise equivalent power (NEP). The NEP is roughly the minimum detectable input power of a PD. Another equivalent way of stating NEP is to take the ratio of the total noise current to the PD responsivity at a particular wavelength. For a PD in which the dark current is the dominant noise source, the expression for NEP is given by [5]

NEP =

hc ( 2qid ) ηqe qλ

0.5

(2.54)

2.6.3 Background Radiation This type of noise is due to the detection of photons generated by the environment. Two types of sources contribute to background radiation (ambient light) noise: localised point sources (e.g., the sun) and extended sources (e.g., the sky). Background radiation from other celestial bodies, such as stars and reflected background radiation, are assumed to be too weak to be considered in terrestrial FSO links; however, they contribute significantly to the background noise in deep space FSO links. The following are the irradiance (power per unit area) expressions for both the extended and localised background sources, respectively [23], [29], [30].

I sky = N(λ) ∆λ OBPF π   Ω 2 4 (2.55)



I sum = W (λ) ∆λ OBPF (2.56)

where N(λ) and W(λ) are the spectral radiance of the sky and spectral radiant emittance of the sun, respectively, ΔλOBPF is the bandwidth of the optical band-pass filter (OBPF) that precedes the PD, and Ω is the PD’s field of view angle in radians. By carefully choosing an Rx with a very narrow FOV and Δλ, the impact of background noise can be greatly reduced. The OBPF in the form of coatings on the Rx optics/telescope with ΔλOBPF < 1 nm are now readily available. The empirical values of N(λ) and W(λ) under different observation conditions are also available in the literature [20], [23], [29]. The background radiation is a shot noise with variance given by [23].

σ 2bg = 2qBef R ( I sky + I sun ) (2.57)

For coherent Rx use (2.57) and change Bef to BcRx. In most practical systems, the Rx SNR is limited by the background shot noise that is much stronger than the quantum shot noise and/or by thermal noise in the electronics following the PD.

75

Optical Sources and Detectors

2.6.4 Thermal Noise The thermal noise, also known as the Johnson noise, occurs in all conducting materials. It is caused by the thermal fluctuation of electrons in any Rx circuit of equivalent resistance R L and temperature T. The electrons are in constant motion, but they collide frequently with the atoms or molecules of the substance. Every free flight of an electron constitutes a minute current. The sum of all these currents taken over a long period of time must, of course, be equal to zero. The thermal noise is regarded as a “white” noise. This is because the PSD is independent of frequency. Moreover, the thermal noise obeys the Gaussian distribution with zero mean and a variance for IM-DD and coherent Rxs defined by, respectively [31],

σ 2th − D =

4κTe Bef (2.58) RL



σ 2th − CS =

4κTe BcRx (2.59) RL

2.6.5 Relative Intensity Noise (RIN) RIN is contributed to the laser diode and is a major specification and cost driver of lasers, which describes fluctuations in the optical power. In the case of DFB, these mainly stem from intrinsic optical phase and frequency fluctuations caused by spontaneous emission. However, the measured RIN value depends strongly on the current noise and the quality of the measurement setup. It is defined as

RIN =

∆P(t )2 P02

(dBw Hz) (2.60)

where ∆P(t )2 is the intensity fluctuations about the average, P0 is the average optical power. Note that electrical power Pele ∝ i2 ∝ Popt, therefore, RIN can be defined as

RIN =

∆Pele Pele − 0

(dBw Hz) (2.61)

where Pele-0 is the average electrical power, and ΔPele is the overall noise (W/Hz), which has three noise components,

∆Pele = σ 2L + σ 2q + σ 2th (2.62)

where σ 2L is the laser noise. For DD and coherent Rx, the noise contributions due to the RIN are given by, respectively,

σ 2in − D = ηRIN ( RMPr ) Bef (2.63)



σ 2in −C = CηRIN ( RMPr ) BcRx (2.64)

2

2

where C is the common mode rejection ratio for the balanced Rx. The level of RIN noise increases with the square root of link Rx bandwidth. For a typical LD, ηRIN is about −160 to −150 dB/Hz. For heterodyne Rxs, the influence of RIN can be avoided by employing a balanced photodiode configuration as shown in Figure 2.26. In this scheme, the signals from the PDs are subtracted, thus resulting in the cancellation of RIN [19]. It is commonly used in applications where there is a requirement for higher SNR. It is very effective in being able to cancel the RIN or “common mode noise” in laser diodes, and it can detect small signal fluctuations on a large DC signal.

76

Optical Wireless Communications

2.6.6 Signal-to-Noise Ratio (SNR) The SNR for the IM-DD system is defined as

SNR IM − DD =

I p2 ( RMPr )2 = (2.65) σ T2 σ q2 + σ d2 + σ 2bg + σ 2th + σ 2in

For shot-noise-limited conditions, also referred to as the SNR with the quantum limit (the ultimate limit), the SNR is defined as

SNR IM − DD =

RPr (2.66) 2qBef

For a coherent Rx with a sufficiently large optical LO power, the shot noise due to the optical LO is the dominant term, thus

SNR IM −C =

RPr (2.67) 2 MBcRx

For PIN PD, M = 1. From the complex amplitude at the IF stage, the carrier-to-noise ratio (CNR) of the heterodyne signal is given as

CNR IM −C =

ηqe Pr = ηqe N rx − ph (2.68) hf BRx

where BRx = 1/Ts is the Rx bandwidth at the IF stage, [20]Tb is bit duration, and N rx − ph is the mean number of photons per symbol. Note that (2.68) is the shot-noise-limited CNR, which also applies to the homodyne Rx.

2.7  OPTICAL DETECTION STATISTICS According to the semi-classical approach, which treats an optical radiation as a wave and prescribes a probabilistic relation to account for its interaction with the atomic structure of the detector surface [20], the probability of a detector with an aperture area Ad, emitting n electrons from impinging photons during a period Tx obeys the Poisson distribution given as

p ( N el ) =

N el

N el

exp ( − N el n!

) (2.69)

The mean count is related to the aperture area and the received irradiance I(t,r) by the following expression

N el =

ηλ hc

∫∫ I (t, r ) dt dr (2.70)

The process of counting the number of electrons generated by the impinging photons is often referred to as photon or photoelectron counting. Table 2.9 summarises some basic statistical parameters of the Poisson random variable. It is worth noting that both the mean and variance of a Poisson random distribution are the same and are given by the mean count in this instance.

77

Optical Sources and Detectors

TABLE 2.9 Statistical Parameters of Poisson Random Distribution [20] Parameter

Definition

Mean

∑N

el

∑N

el

N el

Mean-square value

p ( N el ) 2

nN el

∑e

Characteristic function

p ( N el )

j ωN el

N el

N el

N el

∑N

qth moment

el

q

N el

2

+ N el

(

p ( N el )

∑(1 − z)

Moment generating function

N el

)

exp  e jω − 1 N el   

p ( N el )

exp  − z N el  ∂q  exp − z N el   z =1 ∂z q 

(

p(n )

n

)

The mean current generated from the N el electrons is i = q nN el Tx , and its variance is given by 2

q σ 2 =   σ 2N el (2.71) T



Since for a Poisson distribution the mean and variance are equal, and by choosing the electrical bandwidth of the post-detection filter as 1/2Tx (which is the Nyquist minimum bandwidth requirement), then we have q 2 q 2 〈i 〉Tx σ 2 =   〈 N el 〉 =   T T q (2.72) q = 〈i 〉 = 2q 〈i 〉 Bef Tx



This is the general expression for the variance of any shot-noise process associated with photodetection. It should be mentioned that the photon counts from the desired incoming optical radiation, background radiation, and radiation due to dark current are independent Poisson random variables. Hence, the probability density function (PDF) of the photoelectron emission due to all of these processes occurring together is also a Poisson distribution whose mean is the sum of the means of the individual processes. With large signal photoelectron counts, the generated signal current probability distribution can be approximated to be Gaussian [23]. That is, p(i) =



1 2 2πσ sh

 ( i − i )2  exp −  (2.73) 2 2σ sh  

By taking the additive white Gaussian noise (AWGN) into account, this expression is modified to become

p(i) =

(

1

2 2π σ sh + σ 2bg

)

2  i− i ) ( exp − 2 2  2 σ sh + σ bg

(

)

  (2.74) 

78

Optical Wireless Communications

These expressions (2.69), (2.73), and (2.74) are only valid for a non-varying received optical field. However, if the incoming field is randomly varying, then the received irradiance and i also vary accordingly. Assuming the randomly varying field has a pdf given by p( i ), it follows that the generated photocurrent is now doubly stochastic, and its pdf, which is obtained by averaging (2.74) over the statistics of the varying field, is given by ∞

p(i) =



∫ 0

(

1

2 2π σ sh + σ 2bg

)

2  i− i ) ( exp − 2 2  2 σ sh + σ bg

(

)

  P ( i ) d i (2.75) 

REFERENCES

1. S. M. Sze and K. K. Ng, Physics of Semiconductor Devices, 3rd ed. John Wiley & Sons Inc., 2007. 2. J. M. Senior, Optical Fiber Communications Principles and Practice, 3rd ed. Essex: Pearson Education Limited, 2009. 3. J. L. Miller, Principles of Infrared Technology (A Practical Guide to the State of the Art), vol. 37, no. 2. Chapman & Hall, 1996. 4. R. Ramaswami, “Optical Fiber Communication: From Transmission to Networking,” IEEE Commun. Mag., vol. 40, no. May, pp. 138–147, May 2002. 5. C. C. Davis, Lasers and Electro-Optics: Fundamentals and Engineering, 2nd ed. Cambridge University Press, 2014. 6. D. C. O’Brien et al., “High-Speed Integrated Transceivers for Optical Wireless,” IEEE Commun. Mag., vol. 41, no. 3, pp. 58–62, 2003. 7. J. Byrnes, Unexploded Ordnance Detection and Mitigation. Springer Netherlands, 2008. 8. W. van. Etten and J. van der. Plaats, Fundamentals of Optical Fiber Communications. Prentice Hall, 1991. 9. A. A. Bergh and J. A. Copeland, “Optical Sources for Fiber Transmission Systems,” Proc. IEEE, vol. 68, no. 10, pp. 1240–1247, 1980. 10. Y. Ohno, “Color Rendering and Luminous Efficacy of White LED Spectra,” in Proceedings of SPIE, 2004, vol. 5530, p. 88. 11. J. K. Sheu et al., “White-Light Emission from near UV InGaN-GaN LED Chip Precoated with Blue/ Green/Red Phosphors,” IEEE Photonics Technol. Lett., vol. 15, no. 1, pp. 18–20, Jan. 2003. 12. F. J. Lopez Hernandez, E. Poves, R. Perez-Jimenez, and J. Rabadan, “Low-Cost Diffuse Wireless Optical Communication System Based on White LED,” 2006 IEEE Int. Symp. Consum. Electron., pp. 1–4, 2006. 13. Xin He, Guanying Cao, and Nianyu Zou, “Simulation of White Light Based on Mixed RGB LEDs,” in IET International Conference on Communication Technology and Application (ICCTA 2011), 2011, pp. 961–964. 14. J. Hecht, Understanding Fiber Optics, 5th ed. Prentice Hall, 2005. 15. B. Šaulys, J. Matukas, V. Palenskis, S. Pralgauskait, J. Vyšniauskas, and B. Saulys, “Analysis of ModeHopping Effect in Fabry-Perot Laser Diodes,” in Radar and Wireless, 2010, vol. 9, no. Iii, pp. 3–6. 16. S. F. Tedde, “Fabrication and Characterization of Organic Photodiodes for Industrial and Medical Applications,” Walter Schottky Institut, Technische Universität München, 2009. 17. G. Keiser, Optical Fiber Communications. McGraw-Hill Companies, 2011. 18. K. Kato, S. Hata, A. Kozen, J. I. Yoshida, and K. Kawano, “High-Efficiency Waveguide InGaAs Pin Photodiode with Bandwidth of over 40 GHz,” IEEE Photonics Technol. Lett., vol. 3, no. 5, pp. 473–474, May 1991. 19. H. Ito, T. Furuta, S. Kodama, and T. Ishibashi, “InP/InGaAs Uni-Travelling-Carrier Photodiode with 310 GHz Bandwidth,” Electron. Lett., vol. 36, no. 21, p. 1809, 2000. 20. R. M. Gagliardi and S. Karp, Optical Communications, 2nd ed. New York: John Wiley, 1995. 21. K. Shiba et al., “High Sensitivity Asymmetric Waveguide APD with over -30 dBm at 10 Gbit/s,” in ECOC 2004, 2004, vol. 42, no. 20, p. 1177. 22. K. Kikuchi, “Coherent Optical Communications: Historical Perspectives and Future Directions,” in High Spectral Density Optical Communication Technologies, Berlin, Heidelberg: Springer Berlin Heidelberg, 2010, pp. 11–49. 23. W. K. K. Pratt, Laser Communication Systems, 1st ed. New York: John Wiley & Sons, Inc., 1969.

Optical Sources and Detectors

79

24. H.-G. Bach, “Ultra-Broadband Photodiodes and Balanced Detectors towards 100 Gbit/s and Beyond,” Proc. SPIE 6014, Active and Passive Optical Components for WDM Communications V, vol. 60140B, 2005. 25. T. Okoshi and K. (Kazuro) Kikuchi, Coherent Optical Fiber Communications. KTK Scientific Publishers, 1988. 26. S. Ryu, Coherent Lightwave Communication Systems. Artech House, 1995. 27. S. Betti, G. De Marchis, and E. Iannone, Coherent Optical Communication Systems, 1st ed. Canada: John Wiley and Sons Inc., 1995. 28. G. R. Osche, Optical Detection Theory for Laser Applications, 1st ed. Wiley-Interscience, 2002. 29. N. S. Kopeika and J. Bordogna, “Background Noise in Optical Communication Systems,” Proc. IEEE, vol. 58, no. 10, pp. 1571–1577, 1970. 30. S. Karp, E. L. O’Neill, and R. M. Gagliardi, “Communication Theory for the Free-Space Optical Channel,” Proc. IEEE, vol. 58, no. 10, pp. 1626–1650, 1970. 31. G. Keiser, Optical Communications Essentials, 1st ed. New York: McGraw-Hill Professional, 2003.

3

Channel Modelling

For OWC systems with the full-duplex mode, multipath interference, multi-user access interference and networking schemes and accurate channel models are essential for obtaining the channel characteristics. In order to design, implement, and operate efficient optical communication systems, it is imperative that the characteristics of the channel are well understood. Characterisation of a communication channel is performed by its channel impulse response (CIR), which is then used to analyse the effects of channel-induced distortions. A considerable amount of work has been published on channel characterisation, covering both experimental measurement and computer modelling of both indoor and outdoor systems. The power penalties directly associated with the channel may be separated into two factors, the optical path loss and multipath induced dispersion. Two types of configurations are considered in the optical wireless channel as outlined in Chapter 1. For directed LOS and tracked configurations, reflections do not need to be taken into consideration, and consequently, the path loss is easily calculated from the knowledge of the transmitter beam divergence, receiver size, and separation distance between the transmitter and receiver. However, a non-LOS configuration, also known as a diffuse system (mainly used in an indoor environment), uses reflections from the room surfaces and furniture. These reflections could be seen as unwanted signals or multipath distortions, which make the prediction of the path loss more complex. A number of propagation models (ceiling bounce, Hayasaka-Ito, and spherical) for LOS and non-LOS are introduced in this chapter. The artificial light interference that affects the link performance is also outlined in this chapter. As for outdoor environments, the atmospheric channel is very complex and dynamic, which can affect characteristics of the propagating optical beams, thus resulting in both optical losses and amplitude and phase fluctuations (mostly due to turbulence). There are a number of models to characterise the statistical nature of the atmospheric channel, and these will be discussed in this chapter. A practical test bed for investigating atmospheric effects on the FSO link, as well as measured data, is also covered in this chapter.

3.1  INDOOR OPTICAL WIRELESS COMMUNICATION CHANNELS As shown in Figure 1.8, there are a number of topologies that are commonly used for indoor applications. The configurations can be classified according to (i) the degree of directionality of transmitters and receivers, and (ii) the existence of the LOS path between the transmitters and the receivers [1]–[3]. Intensity modulation with direct detection (IM/DD) is the de facto method of implementing optical wireless systems principally due to its reduced cost and complexity [1], [4], [5]. In IM/DD OWC systems, the drive current ILED of an optical source is directly modulated by the modulating signal m(t), which in turn varies the intensity of the optical source x(t), i.e., the transmit optical power Pt is proportional to ILED; see Figure 3.1. The receiver employs a photodetector, with a response, which is the integration of tens of thousands of very short wavelengths of the incident optical signal, that generates a photocurrent y(t). This photocurrent is directly proportional to the instantaneous optical power incident upon it. That is, the photocurrent is proportional to the square of the received electric field. An IM/DD-based optical wireless system has an equivalent baseband model that hides the high-frequency nature of the optical carrier [6]. The model is shown in Figure 3.2, where R is the photodetector responsivity, h(t) is the baseband channel impulse response, and n(t) is the signal independent shot noise, modelled throughout the book as the additive white Gaussian noise (AWGN) with a double-sided power spectral density (PSD) of N0/2. Non-LOS links, particularly in indoor applications, are subject to the effects of multipath propagation in the same way as RF systems, and these effects are more pronounced. This type of link can suffer from severe multipath-induced performance penalties, as will be discussed in later chapters. Multipath 81

82

Optical Wireless Communications

FIGURE 3.1  Block diagram of an IM/DD OWC system.

propagation causes the electric field to suffer from severe amplitude fades on the scale of a wavelength. The detector would experience multipath fading if the detector size (i.e., the surface area) was proportional to one wavelength or less. Fortunately, OWC receivers use detectors with a surface area typically orders of magnitude bigger than transmission wavelength. In addition, the total photocurrent generated is proportional to the integral of the optical power over the entire photodetector surface; this provides an inherent spatial diversity as shown in Figure 3.3, [1] which is not possible in RF-based systems. Although indoor OWC links do not suffer from the effects of multipath fading as in RF-based systems, they do suffer from the effects of multipath-induced dispersion, which manifests itself in a practical sense as intersymbol interference (ISI), thus affecting link performance. Dispersion is modelled as a linear baseband channel impulse response h(t). The channel characteristic of an OWC link is fixed for a given position of transmitters, receivers, and intervening reflecting objects. The channel characteristic only changes when these components are moved by distances in the order of centimetres [1]. Due to high bit rates Rb, and the relatively slow movement of objects and people within a room, the channel will vary only on the time scale of many bit periods (i.e., the channel is constant for a block of transmission, and this constant within blocks vary independently), and it may therefore be considered quasi-static [2]. The equivalent baseband model of an IM/DD OWC link can be summarized by the following equations: y ( t ) = Goc Rx ( t ) ⊗ h ( t ) + n ( t ) ∞

=



∫ Rhx ( τ) h (t − τ) d ( τ) + n (t )

(3.1)

−∞

where the symbol ⊗ denotes the convolution, and Goc is the gain of the optical concentrator. R depends on the spectral power distribution (SPD) of the light incident on the photodetector and its photodetector’s wavelength-specific responsivity function. Note that h(t), which can be used to analyse or simulate the effects of multipath dispersion in indoor OWC channels, was modelled by Gfeller and Bapst as follows [7]:



 2t0 to   to ≤ t ≤  h ( t ) = f ( x ) =  t 3sin 2 ( FOV ) cos ( FOV )  0 elsewhere 

where to is the minimum delay.

FIGURE 3.2  Equivalent baseband model of an IM/DD OWC link.

(3.2)

83

Channel Modelling

FIGURE 3.3  A block diagram of the OWC link with spatial diversity.

Whilst (3.1) is simply a linear filter channel with AWGN, optical wireless systems differ from the conventional electrical or radio systems since the instantaneous optical power is proportional to the generated electrical current. Note that, in optical systems x(t) represents the power rather than the amplitude signal, and therefore this places two constraints on the transmitted signal. Firstly, x(t) must be non-negative, i.e.,

Pmax − in  ≥ x ( t ) ≥ 0

(3.3)

where Pmax–in is the transmitter’s maximum instantaneous optical power. Secondly, the eye safety requirements limit the maximum Pt that may be used. Generally, it is the average power requirement which is the most restrictive, and hence the average value of x(t) must not exceed a specified maximum power value Pmax, i.e., Pmax =  

Lim T →∞



1   2T

T

∫ x (t ) dt

(3.4)

−T

This is in contrast to the time-averaged value of the signal x ( t ) as in the conventional RF channels. These differences have a profound effect on the system design. On conventional RF channels, the signal-to-noise ratio (SNR) is proportional to the average received power Pr . Whereas in optical wireless links, SNR is proportional to the square of the average received optical signal power given by [1] 2



SNR =

R 2  H 2 ( 0 ) Pr2 Rb N 0

(3.5)

where N0 is the noise spectral density, and H(0) is the channel DC gain given by ∞

H (0) =

∫ h (t ) dt

−∞

(3.6)

84

Optical Wireless Communications

Thus, in optical systems relatively high optical transmit powers are required, and only a limited path loss can be tolerated. The fact that the average optical transmit power is limited suggests that modulation techniques possessing a high peak-to-mean power ratio are most favourable. This is generally achieved by trading off the power efficiency against the bandwidth efficiency. When the shot noise is the dominant noise, the SNR is also proportional to the photodetector area Ad. This is simply because the received electrical power and the variance of the shot noise are proportional to Ad2 and Ad, respectively. Thus, single element–based receivers favour the use of large-area photodetectors. However, as the photodetector area increases, so does its capacitance Cpd, which has a limiting effect on the receiver bandwidth and thus the transmission capacity (i.e., increasing Ad leads to increased Cpd and therefore reduced Rb). This is in direct conflict with the increased bandwidth requirement associated with power-efficient modulation techniques in high-speed OWC links; hence, there is a trade-off between these two factors. The total received current ye ( t ) = Ry ( t ) + RDC PDC ,   where RDC and PDC are the photodetector’s responsivity to the DC signal and DC power, respectively. Note that, in IR systems, based on the regulation, the average optical power must be below a specified maximum level. In visible light communication (VLC) systems, the illumination requirement specifies the average power level, which may be achieved by including a DC bias (detailed discussion about VLC is provided in Chapter 8). In wireless communications with a LOS between the transmitters and the receivers, the received signal can be written as the sum of a complex exponential and a narrowband Gaussian process, which are known as the “LOS component” and the “diffuse components,” respectively. Thus, the channel transfer function is given by: H OW ( f ) = H los + H diff ( f )



(3.7)

where Hlos is the contribution due to the LOS, and Hdiff is the NLOS paths (i.e., diffuse link).

3.1.1 LOS Propagation Model In general, the indoor OWC system uses an LED as a source and large-area photodetectors. The angular distribution of the radiation intensity pattern is modelled using a generalized Lambertian radiant intensity with the following distribution:



 ( ml + 1) cosml ( φ )  for  Φ ∈[ −π / 2,  π / 2 ]  R0 ( φ ) =  2π  0 for  φ ≥ π / 2  

(3.8)

where ml is Lambert’s mode number expressing directivity of the source beam and φ = 0 is the angle of maximum radiated power. The order of Lambertian emission ml is related to the LED semi-angle at half power Φ1/2 by



ml =

− ln 2 ln ( cos Φ1/2 )

(3.9)

The radiant intensity is given by



S ( φ ) = Pt

( ml + 1) cosml φ ( ) 2π



(3.10)

85

Channel Modelling

The photodetector is modelled as an active area collecting the radiation incident at angles ψ smaller than the detector FOV. The effective collection area of the detector is given by:



 Ad cos ψ  0 ≤ ψ ≤ π / 2 Aeff ( ψ ) =  0  ψ > π / 2  

(3.11)

Though ideally a large area detector would be suitable for indoor OWC to collect as much power as possible, it would in practice cause a number of problems, such as increased manufacturing cost, increased junction capacitance and thus a decreased receiver bandwidth, and increased receiver noise. Hence, the use of an optical concentrator is a cost-effective solution in order to increase the overall effective collection area. The optical gain of an ideal non-imaging concentrator with an internal refractive index ni is:



 n2 i   0 ≤ ψ ≤ Ψ c   g ( ψ ) =  sin 2 Ψ c  0  ψ > Ψ c   

(3.12)

where Ψ c ≤ π / 2 is the FOV. From the constant radiance theorem (also known as Etendue limit), the FOV of the receiver is related to the collection area of the lens Acoll and the photodetector area as [1]:



Acoll sin  

FOV  ≤ Ad 2 

(3.13)

From (3.13), it is apparent that the concentrator gain increases when the FOV is reduced. The link length in indoor OWC is relatively short, and hence attenuation due to the absorption and scattering is very low. Considering an OWC link with a Lambertian source, a receiver with an optical band-pass filter of a transmission Ts ( ψ ), and a non-imaging concentrator of gain g ( ψ ), the DC gain for a receiver located at a distance of d and an angle φ with respect to the transmitter (see Figure 3.4) can be approximated as [1]

FIGURE 3.4  Geometry of the LOS propagation model.

86



Optical Wireless Communications

 Ad ( ml + 1) cos ml ( φ ) Ts ( ψ ) g ( ψ ) cos ψ , 0 ≤ ψ ≤ ψ c  H los ( 0 ) =  2πd 2  0 elsewhere 

(3.14)

For the LOS path, the received power is given as Pr − los = H los ( 0 ) Pt



(3.15)

In LOS-based links with the transmitter and the receiver perfectly aligned, the increase of the LOS signal is given by: H los ( ml ) =



( ml + 1) H 2

los



(3.16)

where H los  refers to a Lambertian transmitter with ml = 1. For short-range LOS links, the multipath dispersion is seldom a problem, and therefore the LOS channel is often modelled as a linear system with the attenuation and the delay [6]. The optical LOS channels are considered non–frequency selective, and the path loss depends on the inverse of the square of the distance between the transmitter and the receiver, with the impulse response given as



hlos ( t ) =

Ad ( ml + 1) d cosml ( φ ) Ts ( ψ ) g ( ψ ) cos ψ δ  t −   2πd 2 c 

(3.17)

where c is the speed of the light in free space, δ (.) is the Dirac function, and δ ( t − d / c ) represents the signal propagation delay. The expression assumes that φ < 90° and ψ < FOV and d  Ad . The Matlab codes for simulating the LOS channel gain are given Program 3.1.

Program 3.1:  Matlab codes to calculate the LOS channel gain. theta=70; % semi-angle at half power m=-log10(2)/log10(cosd(theta)); %Lambertian order of emission P_total=20; %tranmistted optical power by individeal LED Adet=1e-4; %detector physical area of a PD %% Optics parameters Ts=1; %gain of an optical filter; ignore if no filter is used index=1.5; %refractive index of a lens at a PD; ignore if no lens is used FOV=60*pi/180; %FOV of a receiver G_Con=(index^2)/sin(FOV); %gain of an optical concentrator %% Room dimension lx=5; ly=5; lz=3; % room dimension in meter h=2.15; %the distance between source and receiver plane Nx=lx*20; Ny=ly*20;% number of grid in the receiver plane XT=0; YT=0;% position of LED; x=-lx/2:lx/Nx:lx/2; y=-ly/2:ly/Ny:ly/2; [XR,YR]=meshgrid(x,y); % receiver plane grid

87

Channel Modelling D1=sqrt((XR-XT(1,1)).^2+(YR-YT(1,1)).^2+h^2); % distance verctor from source 1 cosphi_A1=h./D1; % angle vector %% H_A1=(m+1)*Adet.*cosphi_A1.^(m+1)./(2*pi.*D1.^2); % channel DC gain for source 1 P_rec=P_total.*H_A1.*Ts.*G_Con; % received power from source 1; P_rec_dBm=10*log10(P_rec); meshc(x,y,P_rec_dBm); xlabel('X (m)'); ylabel('Y (m)'); zlabel('Received power (dBm)'); axis([-lx/2 lx/2 -ly/2 ly/2 min(min(P_rec_dBm)) max(max(P_rec_dBm))]);

3.1.2 Non-LOS Propagation Model For non-directed LOS and diffuse links, the optical path loss is more complex to predict since it is dependent on a multitude of factors, such as room dimensions, reflectivity of the ceiling, walls and objects within the room, positions and orientations of the transmitter and the receiver, window size and placement, and other physical matters within a room. The received power is generally defined as Pr − nlos = ( H los ( 0 ) + H nlos ( 0 )) Pt



 =  H los ( 0 ) + 

∑H



refl

refl

( 0 ) Pt

(3.18)



where Hrefl(0) represents the reflected path. The reflection characteristics of object surfaces within a room depend on several factors, including the transmission wavelength, surface material, the angle of incidence θi , and roughness of the surface relative to the wavelength. The latter mainly determines the shape of the optical reflection pattern. Rayleigh criterion [7] is mostly adopted to determine the texture of a surface. According to this criterion, a surface can be considered smooth if the maximum height of the surface irregularities conforms to the following:



hsi
= target?

err>tgt?

No

HPF? Yes

Gen' match filter r(t)

HPF Data stream

HPF?

Match filter data stream

Save Result

BER target met?

Yes HPF Data stream

No

Yes

Add FL interference

Scale data stream for SNR

No

No

Yes

Gen' unity amp' data stream

Yes

BER?

No

Decrease AWGN Value

Yes Plot results

Inform user sim' end

Downsample End

Match filter data stream

Threshold detect

Calculate threshold

Calculate BER

FIGURE 5.2  Flowchart for simulation of OOK shown in Figure 5.1.

where τ is the sampling time, which depends on the modulation schemes as described in Chapter 4, and the symbol ⊗ denotes convolution. Since the interfering signal is periodic, the error probability can be estimated by calculating the bit (slot) error probability over the interference duration and averaging over the bit period [6]. By considering every slot over a 20 ms time interval—i.e., one complete cycle of m fl (t )—and averaging, the Pbe _ OOK in the presence of the AWGN is given by [3], [11]

Pbe _ OOK =

1 2Nb

Nb

 RPavg Tb + mk   RPavg Tb − mk   + Q    (5.4) N0 2 N0 2   

∑Q  k =1

233

Indoor System Performance Analysis 18 16 14

NOPR (dB) OPP (dB)

12 10 8 6 4

NOPR (without interference) NOPR (with interference) OPP (with interference)

2 0

1

10 Data rate (Mbps)

100

200

FIGURE 5.3  NOPR and OPP to achieve a BER of 10 −6 against the data rate for OOK with and without FLI.

where N b is the total number of bits over a 20 ms interval. For a given packet length Dp, the probability of bit error can be converted into a corresponding packet error rate (PER) as given by

Dp PER = 1 − (1 − Pbe_OOK ) (5.5)

The Matlab codes to simulate the effect of FLI in OOK-NRZ are given in Program 5.1. The NOPR and OPP against Rb with and without interference for the OOK-NRZ are given in Figure 5.3. NOPR in the absence of the inference increases linearly with Rb (note the logarithmic scale in the x-axis). However, NOPR in the presence of FLI is almost constant irrespective of Rb (variation of 1E-4) terr=0; % total error tsym=0; % total bits SNR = 10.^(EbN0_db./10); % signal-to-noise ratio P_avg = sqrt(N0*Rb*SNR/(2*R^2)); % average transmitted optical power i_peak = 2*R*P_avg; % Peak Electrical amplitude Ep = i_peak^2 * Tb; % Peak energy (Energy per bit is Ep/2) sgma=sqrt(N0/2/Tsamp); % noise variance %sgma = i_peak/sqrt(2)*sqrt(nsamp/(2*SNR)); pt = ones(1,nsamp)*i_peak; % tranmitter filter rt=pt; % Rx filter matched to pt while(terr50; break; end end

Note that PIM and DPIM encoder and decoder modules should be included in Figure 5.1. Following an approach similar to that adopted for OOK, the average probability of slot error Pse for PPM with hard decision decoding (HDD) and soft decision decoding (SDD) schemes and DPIM with HDD are given as [11] Pse _ PPM _ hard

N sl

 Pavg LTb log2 L / 2 + mk N0 / 2

∑  L1 Q 

1 = N sl

k =1

( L − 1)  Pavg LTb log2 L / 2 − mk Q + L N0 / 2  Pse _ DPIM

1 = N sl



N

∑ k =1

(L +

   

     

 Pavg LDPIM Tb M / 2 + mk   1 Q   L N0 / 2  DPIM  

− 1)  Pavg LDPIM Tb M / 2 − mk   Q  LDPIM N0 / 2   

(5.6)

(5.7)

DPIM

where N sl is the total number of slots over a 20 ms interval and mk  is given by (5.3) with sampling times taken as Ts-PPM and Ts-DPIM for PPM and DPIM, respectively. For PPM with SDD, rather than considering each slot individually, consider symbols composed of L consecutive slots as one. Thus, for each symbol, a vector [m iL+1 m iL+2 … m iL+L] is defined, which represents the MF outputs due to the interference signal. A 1 is then assigned to each of the L-slots in turn, and the corresponding Pse is calculated using the union bound. From these L probabilities, the mean Pse is then calculated. This process is repeated for the next interference signal vector, and so on until all the symbols have been considered. The overall Pse is then found by averaging over all symbols within 20 ms. Thus, the union (upper) bound for Pse for PPM with SDD is given by [3]

Pse _ PPM _ soft =

1 N sl

N sl / L −1 L

L

 Pavg LTb log2 L + miL + j − miL + k   (5.8) N0 

∑ ∑ ∑ Q  k =1

j =1

k =1 k≠ j

The NOPR as a function of the Rb for an SER of 10 −6 and for 4, 8, and 16 PPM(HDD) is depicted in Figure 5.4. Unlike the ideal cases, variations in NOPR for the channel with FLI are small, indicating that FLI is the main source of performance impairment. Note that NOPR increases with Rb for all cases, but the increments are 20 Mbps. The performance improvement at higher Rb is due to the reduced variation of FLI over one symbol duration. (Note that a soft decision is carried out based on the relative

236

Optical Wireless Communications 15 12

NOPR (dB)

9 6 3 4-PPM (with interference) 8-PPM (with interference) 16-PPM (with interference) 4-PPM (without interference) 8-PPM (without interference) 16-PPM (without interference)

0 -3 -6

1

10 Data rate (Mbps)

100

200

FIGURE 5.4  NOPR to achieve an SER of 10 −6 against the data rate for 4, 8, and 16-PPM with HDD and with/without FLI.

amplitude of slots within a symbol.) In SDD, it is the values of FLI samples relative to other samples within the same symbol, which is important, rather than the absolute values. This leads to a lower Pse, thus reducing OPP [1]. As in OOK and PPM with HDD, DPIM also experiences a significant OPP due to FLI (see Figure 5.6). Note that there is a small variation in NOPR for Rb within the range of 1–200 Mbps. For 4 DPIM, NORPS are ∼15 dB and ∼16 dB at 1 Mbps and 200 Mbps, respectively, which are 1 dB and 2 dB higher and lower than OOK and 4 PPM, for respective Rb. As in the case of PPM(HDD), OPPs are minimum for 16 DPIM and increase with decreasing bit resolutions. DPIM has power penalties, which are slightly higher than PPM(HDD), ranging from 14.6–16 dB at 1 Mbps to 4.5–5.4 dB at 200 Mbps. 10 8 6

NOPR (dB)

4 2 0 4-PPM (without interference) 8-PPM (without interference) 16-PPM (without interference) 4-PPM (with interference) 8-PPM (with interference) 16-PPM (with interference)

-2 -4 -6 -8

1

10 Data rate (Mbps)

100

200

FIGURE 5.5  NOPR to achieve an SER of 10 −6 against the data rate of 4, 8, and 16-PPM with SDD and with/ without FLI.

237

Indoor System Performance Analysis 16 14 12

NOPR (dB)

10 8 6 4 2

4-DPIM (without interference) 8-DPIM (without interference) 16-DPIM (without interference) 4-DPIM (with interference) 8-DPIM (with interference) 16-DPIM (with interference)

0 -2 -4 1

10 Data rate (Mbps)

100

200

FIGURE 5.6  NOPR against the data rate for 4, 8, and 16 DPIM with and without FLI.

The Matlab codes for determining the NOPR for DPIM are outlined in Program 5.2. Program 5.2:  Matlab codes to plot NOPR and OPP for DPIM. %% optical power penalty for OOK snr_1Mbps=10.54; data_rate=[1 10 20 40 60 80 100 120 140 160 180 200]; %data rates nf=5*(log10(data_rate)); % normalization factor x=log10(data_rate); % data rate in log scale NOPR_ook_ideal=5*(log10(data_rate)); % optical power penalty in ideal channel; %% *********** DPIM AWGN channel************* NOPR_ideal_4dpim=NOPR_ook_ideal-5*log10(2*2.5/4); NOPR_ideal_8dpim=NOPR_ook_ideal-5*log10(3*4.5/4); NOPR_ideal_16dpim=NOPR_ook_ideal-5*log10(4*8.5/4); semilogx(data_rate,NOPR_ideal_4dpim); hold on semilogx(data_rate,NOPR_ideal_8dpim,’r’); semilogx(data_rate,NOPR_ideal_16dpim,’k’); %% ********************** 4-DPIM in FLI channel *************** snr_4dpim=[41.7 32.2 29.2 26.45 24.96 23.75 23 22.46 21.81 21.4 20.6 20.6]; % snr required to acheive a ber of 10^-6 snr_diff_4dpim=snr_4dpim-snr_1Mbps; % difference in SNR compared to the LOS 1 Mbps NOPR_4dpim=snr_diff_4dpim./2+nf; % optical power penalty p = polyfit(x,NOPR_4dpim,1); % curve fitting f4 = polyval(p,x); %% ********************** 8-DPIM in FLI channel *************** snr_8dpim=[36.6 27 24.1 21.4 20 18.6 18.1 17.5 16.6 16.4 16.1 15.6]; snr_diff_8dpim=snr_8dpim-snr_1Mbps; NOPR_8dpim=snr_diff_8dpim./2+nf;

238

Optical Wireless Communications

p = polyfit(x,NOPR_8dpim,1); f8 = polyval(p,x); %% ********************** 16-DPIM in FLI channel *************** snr_16dpim=[31.1 21.5 18.6 16.1 14.7 13.5 13.1 12.4 11.7 11.2 11 10.5]; snr_diff_16dpim=snr_16dpim-snr_1Mbps; NOPR_16dpim=snr_diff_16dpim./2+nf; p = polyfit(x,NOPR_16dpim,1); f16 = polyval(p,x); %% plots semilogx(data_rate,f4,’b--’); hold on semilogx(data_rate,f8,’r--’); semilogx(data_rate,f16,’k--’); % xlabel(‘Data rate (Mbps)’); % ylabel(‘NOPR’)%; % optical power penalty semilogx(data_rate,f4-NOPR_ideal_4dpim,’b’); semilogx(data_rate,f8-NOPR_ideal_8dpim,’r’); semilogx(data_rate,f16-NOPR_ideal_16dpim,’k’);

5.3  BLW WITHOUT FLUORESCENT LIGHT INTERFERENCE A sequence of pulses passed through an HPF will experience variations in the nominal zero level, which is a form of ISI known as the BLW. BLW has a detrimental effect on the performance of baseband modulation techniques where a significant power is located at or close to the DC region. In this section, the OPP required to overcome the effect of BLW is investigated. For all the analysis, the HPF is modelled as a first-order RC filter with a 3 dB cut-on frequency and an impulse response g(t). The filter response to a single rectangular pulse of amplitude A and duration τ is expressed as

   Ae − t / RC   0≤t ≤τ  gout (t ) =  (5.9) −τ / RC − t / RC t > τ −1 e  − A e

(

)

where the filter time constant RC = 1 / 2 πfc . Due to the principle of superposition in linear systems, if a sequence of such pulses is passed through an HPF, the output is equal to the summation of the individual responses. Thus, for a bit sequence of {A1 A2 … An} where A1…n ∈{0,  1}, the output of a first-order RC HPF at the end of the nth bit may be expressed as [12] n



gout (t ) t = nτ =

∑ A (e i

−2 πfc τ

)(

− 1 e −2 πfc τ

)

i −1

(5.10)

i =1

To illustrate the effect of BLW, consider the OOK-NRZ signal format with a rectangular pulse shape as shown in Figure 5.7(a). The HPF output gout (t ) with a fc-hpf of 0.05 times Rb and the corresponding BLW are shown in Figure 5.7(b) and (c), respectively, with the Matlab codes presented in Program 5.3. Notice the variation in the amplitude for both 0 and 1 levels in Figure 5.7(b). The average value (i.e., the midpoint between the high and low levels) of the high-pass signal shown changes with time—see Figure 5.7(c)—and it is this variation that leads to the BLW. Program 5.3:  Matlab codes to slow the effect of high-pass filter and baseline wander in OOK-NRZ signal. Rb=1e6; % Bit rate Tb=1/Rb; % bit duration sig_length=1e2; % number of bits

239

Indoor System Performance Analysis

(d) FIGURE 5.7  Time waveforms: (a) transmitted binary signal, (b) high-pass filter output, (c) BLW signal, and (d) for three different high-pass filter cut-on frequencies. nsamp=10; % samples per symbols Tsamp=Tb/nsamp; % sampling time Lsym=1e3; % number of bits fc_rb=5e-2; % normalized cut-off frequency of HPF

240

Optical Wireless Communications

A=1; % normalized amplitude %% ***** Calculate impulse response of filters ***** tx_impulse=ones(1,nsamp)*A; % Tx filter fc=fc_rb*Rb; Note, fc = fc-hpf; % actual cut-on frequency of HPF t=Tsamp:Tsamp:10*Tb; % time vector hpf_impulse(1)=1*exp(-2*pi*fc*t(1)); % impulse response (see eq (5.9)) for loop=2:length(t) hpf_impulse(loop)=-1*(exp(2*pi*fc*t(1))-1)*exp(-2*pi*fc*t(loop)); end %% effect of HPF on OOK OOK=randint(1,Lsym); OOK=2*OOK-1; % removing dc components signal=filter(tx_impulse,1, upsample(OOK,nsamp)); % rectangular pulse shaping hpf_output=filter(hpf_impulse,1,signal); % hpf output; %% plots plotstart=(10*nsamp+1); plotfinish=plotstart+15*nsamp; t=0:Tsamp:Tsamp*(plotfinish-plotstart); subplot(311); plot(t,signal(plotstart:plotfinish),’k’); subplot(312); plot(t,hpf_output(plotstart:plotfinish),’k’); % subplot(413); plot(t,signal(plotstart:plotfinish)-hpf_output(plotstart: plotfinish),’k’) subplot(313); plot(t,-signal(plotstart:plotfinish)+hpf_output(plotstart: plotfinish),’k’)

6000

6000

5000

5000

4000

4000

Frequency

Frequency

The effect of BLW on the performance of baseband modulations with no FLI can be analysed with reference to Figure 5.1. The histogram plots of the MF output with and without HPF for the normalised cut-on frequencies of 10 −3 and 10 −2 are shown in Figure 5.8 (Matlab code is

3000

3000

2000

2000

1000

1000

0 -1

-0.5 0 0.5 Normalized matched filter output (a)

1

0 -1

-0.5 0 0.5 Normalized matched filter output (b)

FIGURE 5.8  Histogram of MF output for OOK with: (a) fc − hpf / Rb = 10 −3, and (b) f− hpfc / Rb = 10 −2 .

1

Indoor System Performance Analysis

241

given in Program 5.4). The solid line in the centre of the plot indicates a value of zero, i.e., no difference between expected and actual MF outputs, which would be the case in the absence of BLW. The increase in the variance of the probability distribution as fc − hpf / Rb increases from 10 −3 to 10 −2 is evident from the figure. For relatively low values of fc − hpf / Rb the HPF impulse response spans many bit periods. Accordingly, the ISI introduced by the HPF is comprised of the weighted sum of many IID binary random variables. Therefore, as a result of the central limit theorem, the distribution can be approximated as Gaussian [3], as confirmed by the overall shape of the histograms. In [13], this Gaussian approximation was used to develop closedform expressions for the probability of error Pe due to the BLW and the Gaussian noise for the OOK and Manchester encoding formats. In [12], [14], this work was further extended, using the non-classical Gauss quadrature rules to determine Pe, rather than assuming a Gaussian distribution. Program 5.4:  Matlab codes to plot distribution OOK due to the BWL effect. Rb=1e6; % Bit rate Tb=1/Rb; % bit duration nsamp=10; % samples per symbols Tsamp=Tb/nsamp; % sampling time Lsym=1e5; % number of bits fc_rb=1e-3; %normalized cut-off frequency fc=fc_rb*Rb; % cut-on frequency of HPF p_ave=1; p_peak=2*p_ave; % peak power of TX’d signal for OOK; % ***** Calculate impulse response of filters ***** tx_impulse=ones(1,nsamp)*p_peak; mf_impulse=ones(1,nsamp)*(1/sqrt(Tb)); t=Tsamp:Tsamp:200*Tb; hpf_impulse(1)=1*exp(-2*pi*fc*t(1)); for loop=2:length(t) hpf_impulse(loop)=-1*(exp(2*pi*fc*t(1))-1)*exp(-2*pi*fc*t(loop)); end % ***** calculate overall impulse response***** temp1=conv(tx_impulse,hpf_impulse); temp2=conv(temp1,mf_impulse); temp2=temp2*Tsamp; system_impulse=temp2(nsamp:nsamp:200*nsamp); %discrete impulse response; % ***** Do analysis on a per sequence basis ***** expected_one=0.5*p_peak*sqrt(Tb); expected_zero=-0.5*p_peak*sqrt(Tb); OOK=randint(1,Lsym); OOK=2*OOK-1; % removing dc components mf_output=filter(system_impulse,1,OOK)/(2*expected_one); mf_output_one=mf_output(find(OOK==1)); % output for tranmitted bit of 1; mf_output_zero=mf_output(find(OOK==-1));

242

Optical Wireless Communications

nbin=51; [n_zero,xout]=hist(mf_output_zero,nbin); [n_one,xout]=hist(mf_output_one,nbin); % combined histogram; % both expected outputs are shifted to zero; % note that removing dc value makes energy for zero and one identical expect_one=xout(find(n_one==max(n_one))); n_total=n_zero+n_one; Fig.; bar(xout,n_zero); % histogram for zero bits Fig.; bar(xout,n_one); % histogram for zero bits Fig.; bar(xout-expect_one,n_total); set(0,’defaultAxesFontName’, ‘timesnewroman’,’defaultAxesFontSize’,12) xlabel(‘Normalized matched filter ourput’); ylabel(‘Frequency’)

To analyse the effect of HPF, the discrete-time equivalent impulse response of the cascaded Tx filter, Rx filter, and HPF needs to be determined. The resulting impulse response cj decays rapidly to zero with time, hence it can be truncated without a significant loss of accuracy [3]. The BER is then approximated using the truncated length J and is calculated by averaging the error rate over all possible symbol sequences of length J. The discrete time equivalent impulse response, truncated to have a duration of J-bit, is given as [3]

 p(t ) ⊗ r (t ) ⊗ g(t ) |t = j τ , cj =  0, 

1< j < J otherwise

(5.11)

where g(t) is the impulse response of HPF. Considering K distinct bit sequences of length J, denoted as {a1, a2, … aK}, let ai,J represent the value of the Jth bit in the sequence ai, where ai,J ∈{0,1}. For ai passed through the system, the MF output, sampled at the end of the Jth bit period, is given by

Ai , J = 2 RPavg ai ⊗ c j

j= J

. (5.12)

The average bit error probability Pbe for OOK for the Jth bit is given by [3, 11]

Pbe _ OOK

1 = K

K

 Ai , J   (5.13) n 

∑Q  σ i =1

where σ n is the standard deviation of the Gaussian, non-white, zero mean, shot noise samples at the MF output, which is given as [12], [15]



σn =    

(

)

−2 πfc τ N0 1 − e 1     (5.14) 2 2 πfc − hpf τ

PPM with HDD may be evaluated in the same way as OOK using (5.11) with sampling times replaced with t = jTb M / L. Since the PPM slot sequence is not IID and fc-hpf is not necessarily small compared to Rb, the probability distribution cannot be assumed to be Gaussian. Considering K-distinct PPM slot sequences of length J, denoted as {b1, b2, … bK}, let bi,J represent the value of

243

Indoor System Performance Analysis

the Jth bit in sequence bi, where bi,J ∈{0,1}. For bi applied to the system, the MF output, sampled at the end of the Jth slot period, is given by

Bi , J = LRPavg bi ⊗ c j

j= J

(5.15)

The probability of slot error is then found by averaging over all K-sequence, which is given by

Pse _ PPM _ hard

 Bi , J − α ppm   (5.16) σn 

K

∑Q 

1 = K

i =1

where α ppm is the threshold level, which is set midway between one and zero levels in the absence of any BLW, and is given as

1 1 α ppm = RPavg LTb log2 L    −  (5.17)  2 L

σn is the standard deviation of the zero-mean, non-white Gaussian noise, which is given as

σn =    

(

)

−2 πfc Ts − PPM N0 1 − e 1     (5.18) 2 2 πfc Ts − PPM

For PPM with SDD, the method is similar, but rather than just considering the Jth slot in each sequence, the next whole symbol following the Jth slot is considered. Therefore, slightly longer sequences needs to be generated. For a sequence k, let the next whole symbol after the Jth slot be denoted as {bi,p bi,p+1 … bi,p+L} where p ≥ J. The corresponding system outputs are given by (5.15) with the sampling times replaced by j = p, p+1, … p+L. Assuming that, for the final symbol of each sequence under consideration, the 1 was transmitted in slot (p+w), where 1 ≤ w ≤ L, the probability of symbol error for PPM is given as



Pe _ symb _ PPM _ soft =

1 K

K

L

i =1

j =1 j ≠ p+ w

∑∑

 Bi , p+ w − Bi , p+ j  Q  (5.19) σn  

Using a similar approach to that used in the OOK and PPM schemes, the probability of slot error for DPIM is found by averaging over all K-sequence, which is given by

Pse _ DPIM =

1 K

K

 Bi , J − α dpim   (5.20) σn 

∑Q  i =1

where α dpim is the threshold level, set midway between expected one and zero levels in the absence of any BLW, as given by

1 1 α dpim = RPavg LDPIM Tb log2 L    −  2 LDPIM

  (5.21)

Figure 5.9, Figure 5.10, Figure 5.11, and Figure 5.12 show the normalised average optical power requirement (NOPR) against fc − hpf / Rb for OOK, PPM(HDD), PPM(SDD), and DPIM, respectively.

244 Normalized average optical power requirement (dB)

Optical Wireless Communications 6 5 4 3 2 1 0 -1 -5 10

-4

-3

-2

10 10 10 HPF cut-on frequency/bit rate

-1

10

FIGURE 5.9  The normalised optical power requirement versus fc − hpf / Rb for OOK.

Normalized average optical power requirement (dB)

Figure 5.9 clearly demonstrates the susceptibility of OOK to the BLW. Power penalties are incurred for the normalised fc− hpf above ∼10 −3, and a 3 dB average OPP is introduced for fc− hpf being ∼1% of Rb. However, PPM is much more resistant to the effects of BLW compared with OOK. For both detection methods, the power penalties occur at fc− hpf = ∼10% of Rb, which is several orders of magnitude higher than OOK. In addition to ∼1.5 dB reduction in OPP, which SDD offers over the threshold detection, there is little difference between the two sets of curves. As expected, higher orders are slightly more resistant to the BLW since the bandwidth requirement is greater and consequently, there is less power below fc− hpf i.e., a fraction of the total power. DPIM offers more resistance to the BLW than OOK, with the higher orders achieving the greatest robustness. This is 6 4 2

L=4 L=8 L = 16 L = 32

0 -2 -4 -6 -8 -10 -3 10

-2

10

-1

10

0

10

HPF cut-on frequency/bit rate

FIGURE 5.10  The normalised average optical power requirement versus fc − hpf / Rb for PPM(HDD).

245

Normalized average optical power requirement (dB)

Indoor System Performance Analysis 6 4 2

L=4 L=8 L = 16 L = 32

0 -2 -4 -6 -8 -10 -3 10

-2

-1

10 10 HPF cut-on frequency/bit rate

10

0

FIGURE 5.11  The normalised average optical power requirement versus fc- hpf / Rb for PPM(SDD).

Normalized average optical power requirement (dB)

mainly due to the increase in the bandwidth requirement. For lower orders, the power penalties are more apparent at fc− hpf of ∼0.01, which is an order of magnitude higher than OOK. Compared with the performance of PPM, DPIM is more susceptible to the BLW. As discussed in the previous chapter, the susceptibility of various modulation schemes to the BLW can be explained by observing the PSD profiles. Since OOK contains a large proportion of its power within the DC region, then it is the most susceptible to the BLW followed by DPIM and PPM. There is a progressive decrement in power at low frequencies with an increase in the order of DPIM and PPM, and hence higher-order PPM and DPIM are less susceptible to the BLW than are the low orders. 6 4

L L L L

=4 =8 = 16 = 32

2 0 -2 -4 -6 -8 -4 10

10

-3

-2

10 10 HPF cut-on frequency/bit rate

-1

10

0

FIGURE 5.12  The normalised average optical power requirement versus f− hpfc / Rb for DPIM.

246

Optical Wireless Communications

5.4 FLUORESCENT LIGHT INTERFERENCE WITH ELECTRICAL HIGH-PASS FILTERING As discussed earlier, an electrical HPF can diminish the effects of FLI, whether this is implemented in the analogue or digital domain. Also, the choice of fc-hpf is a trade-off between the extent of FLI rejection and the severity of the baseline introduced by the HPF and the attenuation of the interfering signal; the optimum choice would be to minimise the overall OPP. Whilst fc-hpf is an important parameter, other filter parameters such as the stop-band attenuation and the roll-off are also likely to have an effect on the performance, which may produce variations in the optimum fc-hpf. Given that the optimum fc-hpf changes with Rb, one solution would be to simulate a system over all envisaged Rb and fc-hpf to determine the correct cut-on frequency. This is obviously less than practical and time-consuming; a compromise method would be to select a limited number of Rb and a range of fc-hpf for a given filter to determine as close to the optimum as possible. In this section, the optimum fc-hpf for HPF which minimises the overall OPP is estimated using a method which combines the analysis carried out in the previous two sections. Using the optimum cut-on frequencies, the OPRs are calculated, and the effectiveness of HPF as a means of mitigating the effect of FLI is assessed. To determine the optimum fc-hpf, Figure 5.1 is adopted with the inclusion of both the FLI signal and HPF. The sampled output due to the FLI signal after being applied to the MF and HPF is given by m j = m(t ) ⊗ r (t ) ⊗ g(t ) t = jTb (5.22)



The sampled output due to the Jth bit is given by (5.12), and the overall output signal is a superposition of the c j and m j . Hence, the overall Pe is found by averaging over M-bit and J-sequence [3], [11] Pbe _ OOK =



1 2 M .2 J

M

  Ai , J + mk   − Ai , J − mk    Q  σ   (5.23)  + Q  σn   n  k =1 a j ∈{0,1}J 

∑∑

Similarly, the probability of slot error for PPM can be approximated as Pse _ PPM _ hard =

1 M .2 J

M

 1  Bi , J + mk − α ppm  ( L − 1)  α ppm − Bi , J − mk   (5.24)  L Q   + L Q    σn σn    k =1 b j ∈{0,1}J 

∑∑

As in the previous section, the probability of symbol error can be obtained by averaging, which is given by



Pe _ symb _ PPM _ soft =

1 1 N/L K

( N / L ) −1 K

L

∑∑ ∑ n=0

i =1

j =1 j ≠ p+ w

 Bi , p+ w + mnL + w − Bi , p+ j − mnL + w  Q  (5.25)  σn

Similarly, the slot error probability of DPIM in the presence of FLI with HPF is given by Pse _ DPIM =

1 M 2J

 1  Bi , J + mk − α dpim  ( LDPIM − 1)  α dpim − Bi , J − mk   Q Q    . (5.26)  + L   σn   σn DPIM  k =1 b j ∈{0,1}J   LDPIM M

∑∑

In order to determine the optimum fc-hpf and to achieve an SER of 10−6 for a range of fc − hpf / Rb, the NOPR is calculated as shown in Figure 5.13, Figure 5.14, Figure 5.15, and Figure 5.16 for OOK, PPM(HDD), PPM(SDD) and DPIM, respectively. It is evident that HPF gives virtually no reduction in NOPR for Rb of 1 and 10 Mbps. For the purpose of determining power requirements, the optimum normalised fc-hpf of 1.4 × 10−4 and 2 × 10−4 were used for 1 and 10 Mbps, respectively, which represent the

247

Normalized average optical power requirement (dB)

Indoor System Performance Analysis 20 19

1 Mbit/s 10 Mbit/s 100 Mbit/s

18 17 16 15 14 13 12 -5 10

-4

-3

-2

10 10 10 HPF cut-on frequency/bit rate

-1

10

FIGURE 5.13  The normalised average optical power requirement versus the normalised HPF cut-on frequency for OOK for a range of data rates.

maximum values that can be used without introducing baseline wander–induced power penalties (BLW). At Rb of 100 Mbps and at fc-hpf of ∼7 × 10−3, high-pass filtering can yield a reduction in the average PPR. However, with electrical HPF we notice reductions in the average OPR of PPM(HDD), even at 1 Mbps. The narrowness of the troughs for 1 Mbps curves in Figure 5.14 suggest that little variation can be tolerated for fc-hpf if the optimum performance is to be achieved. At 10 Mbps (i) the troughs are broader, making the actual choice of fc − hpf / Rb not quite so critical; and there is a floor in the power requirement curves, thus indicating that there is a region in which fc-hpf is high enough to attenuate the interference signal without introducing BLW–induced performance detritions. Within this region, the FLI-induced OPP is due to the additional shot noise, and selecting fc-hpf anywhere within this region will give approximately the same level of performance. It is evident from Figure 5.15 that electrical HPF yields reductions in the average OPR of PPM(SDD) operating at 1 Mbps. At 10 Mbps, from the analysis carried out in the Section 5.2, PPM(SDD) is found to suffer only small power penalties without the use of electrical HPF. Consequently, the reduction in the average OPR of Figure 5.15 appears modest, and the broad troughs suggest that the fc-hpf does not have to be very precise in order to minimise the average OPR. At 100 Mbps, PPM(SDD) is immune to FLI without the use of electrical HPF, and consequently there is no reduction in the average OPR. From Figure 5.16, it is evident that when operating at 1 Mbps in the presence of FLI, electrical HPF introduces no reduction in the average OPRs for DPIM. At 10 Mbps, high-pass filtering is more effective, thus resulting in reduced average OPRs. Compared with PPM(HDD), DPIM has lower optimum HPF cut-on frequencies. This is due to DPIM’s higher susceptibility to BLW. At 100 Mbps, electrical high-pass filtering once again results in reduced average OPRs. The broad troughs suggest that the HPF fc does not have to be exact in order to achieve the near maximum reduction. Using these optimum normalised fc-hpf, the average OPRs for OOK, PPM(HDD), PPM(SDD), and DPIM at Rb of 1, 10, and 100 Mbps are shown in Figure 5.17, Figure 5.18, Figure 5.19, and Figure 5.20, respectively. Also shown in these figures are the OPRs with no FLI or HPF. By comparing Figure 5.17 with Figure 5.3 it is clear that, for Rb of 1 and 10 Mbps, filtering is not effective in reducing the FLI-induced OPP. This is due to the fact that OOK—with a high level of DC—is highly susceptible to the BLW and consequently, only a low normalised fc-hpf can be used, which at low to medium Rb is not effective in attenuating FLI. However, at Rb of 100 Mbps, HPF is much more effective, with a typical reduction in the average OPR of ∼4.4 dB. But this still leaves an OPP of ∼3 dB compared with the same Rb with no FLI.

20 18

Optical Wireless Communications Normalized average optical power requirement (dB)

Normalized average optical power requirement (dB)

248

1 Mbit/s 10 Mbit/s 100 Mbit/s

16 14 12 10 8 6 4 2 -5 10

-4

10

-3

-2

-1

10 10 10 HPF cut-on frequency/bit rate

0

10

20

1 Mbit/s 10 Mbit/s 100 Mbit/s

18 16 14 12 10 8 6 4 2 0 -5 10

-4

10

-3

-1

0

10

(b)

(a) Normalized average optical power requirement (dB)

-2

10 10 10 HPF cut-on frequency/bit rate

16 14

1 Mbit/s 10 Mbit/s 100 Mbit/s

12 10 8 6 4 2 0 -2 -5 10

-4

10

-3

-2

-1

10 10 10 HPF cut-on frequency/bit rate

0

10

(c)

FIGURE 5.14  The normalised average optical power requirement against the normalised HPF cut-on frequency for PPM(HDD) for a range of data rates and for (a) L = 4, (b) L = 8, and (c) L = 16.

Unlike OOK, in PPM(HDD) electrical HPF will result in reduced average OPRs even at a 1 Mbps data rate. For 1 Mbps, the narrowness of the troughs in Figure 5.14 suggest that little variation in fc-hpf can be tolerated in order to achieve the optimum performance. Note that at 10 Mbps (i) the troughs are broader, thus making the actual choice of fc − hpf / Rb not quite so critical; and (ii) there is a floor in the OPR curves, thus indicating a region in which fc-hpf is high enough to attenuate FLI sufficiently without introducing the BLW effect. Within this region, the FLI-induced OPP is due to the additional shot noise only, and selecting fc-hpf anywhere within this region will give approximately the same level of performance. In PPM(SDD) with no HPF there is a small OPP at Rb >10 Mbps (see Figure 5.5). Consequently, there is only a modest reduction in the average OPR with HPF. At 100 Mbps, PPM(SDD) with no filtering shows immunity to FLI, and consequently, no reduction in the average OPR; see Figure 5.19. As previously observed for OOK and PPM(HDD) schemes, it is evident from Figure 5.6 that with FLI the average OPRs for DPIM are very similar for all three Rb. Relative to the OPRs with no FLI, the power penalties decrease with the increase of Rb. In DPIM the power penalties are slightly higher than those of PPM(HDD), ranging from ∼14–16 dB to ∼5–7 dB at 1 Mbps and 100 Mbps, respectively.

249

20 18 16

Normalized average optical power requirement (dB)

Normalized average optical power requirement (dB)

Indoor System Performance Analysis

1 Mbit/s 10 Mbit/s 100 Mbit/s

14 12 10 8 6 4 2 0 -5 10

-4

10

-3

-2

-1

0

10 10 10 HPF cut-on frequency/bit rate

10

18 16 14

1 Mbit/s 10 Mbit/s 100 Mbit/s

12 10 8 6 4 2 0 -2 -5 10

-4

10

-3

(a) Normalized average optical power requirement (dB)

-2

-1

10 10 10 HPF cut-on frequency/bit rate

0

10

(b) 12 10

1 Mbit/s 10 Mbit/s 100 Mbit/s

8 6 4 2 0 -2 -4 -5 10

-4

10

-3

-2

-1

10 10 10 HPF cut-on frequency/bit rate

0

10

(c)

FIGURE 5.15  The normalised average optical power requirement as a function of the normalised HPF cuton frequency for PPM(SDD) for various data rates and for: (a) L = 4, (b) L = 8, and (c) L = 16.

5.5  WAVELET ANALYSIS The topic of wavelets is multifaceted and highly mathematical, and it is a subject that is arguably dominated by researchers with a pure or applied mathematical background. Difficulties due to the mathematical complexities of wavelet analysis are alluded to by some tutorials [16] and by the authors of some of the many mathematically dominated texts published on the subject. Discussion of wavelet theory is by necessity limited to its application in this work. Evidently, the first recorded mention of what has become to be known as a ‘wavelet’ was in 1909 in a thesis by Alfred Haar. The concept of wavelets in its present theoretical form was first proposed by Jean Morlet and by Alex Grossmann in France. Since then a number of notable names have worked on the subject, such as Yves Meyer, with the main algorithm being credited to work undertaken by Stephane Mallat in 1988 [17]. Since that time, work on the subject has become widespread and covers disciplines too numerous to mention. In later years, research became particularly active in the United States with fundamental work being undertaken by Ingrid Daubechies, Ronald Coifman, and Victor Wickerhauser. A comprehensive history of the subject can be found in [18]. The feature-rich mapping of signals to

20

18

Optical Wireless Communications Normalized average optical power requirement (dB)

Normalized average optical power requirement (dB)

250

1 Mbit/s 10 Mbit/s 100 Mbit/s

16

14

12

10 -5 10

-4

10

-3

-2

-1

10 10 10 HPF cut-on frequency/bit rate

0

10

20

1 Mbit/s 10 Mbit/s 100 Mbit/s

18 16 14 12 10 8 6 -5 10

-4

10

-3

(a) Normalized average optical power requirement (dB)

-2

-1

10 10 10 HPF cut-on frequency/bit rate

(b)

20 18

1 Mbit/s 10 Mbit/s 100 Mbit/s

16 14 12 10 8 6 4 -5 10

-4

10

-3

-2

-1

10 10 10 HPF cut-on frequency/bit rate

0

10

0

10

(c)

FIGURE 5.16  Normalised average optical power requirement versus the normalised HPF cut-on frequency for DPIM for various data rates and for: (a) L = 4, (b) L = 8, and (c) L = 16.

wavelet coefficients has resulted in an explosion of research articles and potential applications in many fields of science and engineering from stock market applications and human motor behaviour to optimisation of the JPEG2000 standard [19]. However, it is possible to generalise these applications into three main areas: data compression, denoising, and feature extraction. In some cases, it is a subtle combination of these elements that form a particular application. In this work, it is arguably the combined abilities of denoising and feature identification that are employed. The identification and denoising properties of wavelets are also being applied in medical research [20]–[24]. The field of engineering is well subscribed to by researchers with the subject of wavelets being applied in almost every facet. In the field of communications engineering, wavelet packet transform (WPT) has been proposed as a powerful tool of signal representation in multiple accesses technique [25], [26], modulation techniques as an alternative to subcarrier modulation [27], [28], symbol synchronisation [29], signal estimation [22], CDMA systems [30], channel characterisation [31], [32], adaptive denoising [33], mitigate ISI, channel equalisation [30], [34]–[36], traffic analysis [33], and OFDM for VLC [37], [38].

251

Indoor System Performance Analysis 18

Normalized average optical power requirement (dB)

16 14 12 10 8 6 4 2 0 1

10

100

Bit rate (Mbit/s) without interference

with interference and HPF

FIGURE 5.17  The normalised average optical power requirement versus the bit rate for OOK with FLI and optimised HPF.

10

Normalized average optical power requirement (dB)

8 6 4 2 0 -2 -4 -6 -8

1

10

100

Bit rate (Mbit/s) L = 4, without interference

L = 8, without interference

L = 16, without interference

L = 4, with interference and HPF

L = 8, with interference and HPF

L = 16, with interference and HPF

FIGURE 5.18  The normalised average optical power requirement versus the bit rate for PPM(HDD) with FLI and optimised HPF.

252

Optical Wireless Communications 8

Normalized average optical power requirement (dB)

6 4 2 0 -2 -4 -6 -8 -10

1

10

100

Bit rate (Mbit/s) L = 4, without interference

L = 8, without interference

L = 16, without interference

L = 4, with interference and HPF

L = 8, with interference and HPF

L = 16, with interference and HPF

Normalized average optical power requirement (dB)

FIGURE 5.19  The normalised average optical power requirement versus the bit rate for PPM(SDD) with FLI and optimised HPF.

16 14 12 10 8 6 4 2 0 -2 -4 -6 -8

1

10

100

Bit rate (Mbit/s) L = 4, without interference

L = 8, without interference

L = 16, without interference

L = 4, with interference and HPF

L = 8, with interference and HPF

L = 16, with interference and HPF

FIGURE 5.20  The normalised average optical power requirement against the bit rate for DPIM with FLI and optimised HPF.

253

Indoor System Performance Analysis

5.5.1 The Continuous Wavelet Transform (CWT) Although the Heisenberg uncertainty principle applies to any transform, multi-resolution analysis (MRA) minimises its impact since not every spectral component is resolved equally as in the short-term Fourier transform (STFT). A CWT decomposes signals over dilated and translated mother wavelet ψ(t ), which is well localised in time and frequency and has zero mean as given by ∞

∫ ψ(t) dt = 0 (5.27)



−∞

The family of mother wavelets is normalised using the norm ψ s ,τ (t ) = 1, which is obtained by scaling ψ by s and translating by τ as given by ψ s ,τ (t ) =



1  t − τ ψ (5.28) s  s 

To scale a wavelet means to stretch or dilate. As the wavelet is stretched in the horizontal axis, it is squashed in the vertical direction to ensure that the energy contained within the scaled wavelets is the same as the original mother wavelet [39]. The translation moves the wavelet along the x-axis. The scaling and translation operations are shown in Figure 5.21 for the Morlet wavelet [32]. The scale in the wavelet analysis is similar to the scale used in maps [16]. The high and low scales correspond to global information and the detailed information about the signal, respectively. The scale is inversely proportional to the conventional frequency, so the low scale relates to the high frequency and vice versa. Note that • If s  1, the window

1 s

• If s  1, the window

1 s

is large, and WT reacts mainly to low frequencies. ψ ( t −τ s )

ψ ( t −τ s ) is small, and WT reacts mainly to high frequencies.

(a) Mother wavelet

(b) Wavelet scaled by 0.5 (compressed)

(c) Wavelet scaled by 2 (dilated)

(d) Wavelet translated by 4.

FIGURE 5.21  The scaling and translation to the Morlet wavelet.

254

Optical Wireless Communications

As intimated, there is a relationship between scale and frequency; however, many texts such as [39], [40] suggest it is better to think in terms of the pseudo-frequency fp. The relationship between scale and frequency is dependent on the centre frequency of the wavelet Fc-w and the sampling period and ∆,which   is given by [39], [40] s=



Fc − w .∆ (5.29) fp

Fc − w is defined as an approximate measure of the oscillatory nature of the basis function at its centre. The CWT of signal fp and reconstruction formula are given by [17], [41] ∞

F (s , τ) =



∫ f (t)

−∞

f (t ) =



1 cψ

∞

∞

ds

∫ ∫s dτ

−∞

2

1 *  t − τ ψ dt  s  (5.30) s F (s , τ)

0

1  t − τ ψ (5.31) s  s 

where * denotes a complex conjugate and Cψ depends on the wavelet. The success of the reconstruction depends on this constant, called the admissibility constant, in order to satisfy the following admissibility condition [42] ∞

0 < Cψ =



∫ ψˆ (ω )

−∞

2

dω < ∞   (5.32) ω

where ψˆ (ω ) is the FT of ψ(t ). Figure 5.22 Shows the 3D plot of coefficients of the CWT of stationary and non-stationary signals, which have identical frequency components as given by

x (t ) = cos(20 πt ) + cos(60 πt ) + cos(100 πt ) (5.33)



 10 if t < 0.5  x (t ) = cos(2 πft )where =  30 if 0.5 ≤ t < 1 (5.34)  50 otherwise 

(a)

(b)

FIGURE 5.22  The CWT of the signal of: (a) non-stationary and (b) stationary.

255

Frequency

Frequency

Indoor System Performance Analysis

Time

Time

(a)

(b)

FIGURE 5.23  Time–frequency representation of (a) CWT and (b) STFT.

It is intuitive to see how the time-scale map identifies attributes of the original signal, localising the features in time, which is in complete contrast to the FFT spectra. It is this feature of the signal which makes CWT and its derivatives a popular choice in research and in its application in signal analysis. The multiresolution properties of the CWT are pictorially represented in Figure 5.23, where the horizontal and vertical axes represent time and frequency, respectively. Each box represents an equal portion of the time–frequency plane, but in different proportions to time and frequency. Every box has a non-zero area indicating that an exact point in the time–frequency plane cannot be known. Each rectangular box (both in STFT and WT) has an identical area occupying an identical amount of the time–frequency plane. Each area in case of CWT has different dimensions for length (frequency) and width (time). For lower frequencies, this equates to an improved frequency resolution but a poor time resolution, whilst higher frequencies have an improved time resolution but a poor frequency resolution. In contrast, the time–frequency map for the STFT would show areas of equal heights and widths, however, different STFT windows (equivalent to mother wavelets) would result in different regions up to a point of some lower bound determined by the uncertainty principle. For CWT, the resolution in time or frequency is also determined by the choice of the wavelet.

5.5.2 The Discrete Wavelet Transform (DWT) As with the FFT and STFT, CWT analysis of meaningful signals by hand is almost impossible, and computers are used for signal processing. In this case, a discretised or time sampled version of CWT is used. Fortunately, as the mother wavelet is dilated (higher scales and lower frequencies), the sampling rate can be reduced without affecting the results. If synthesis (reconstruction from the wavelet coefficients) is required, the Nyquist sampling rate must be observed. CWT unmixes parts of the signal, which exist at the same instants but at different scales, thus giving highly redundant CWT coefficients [32]. The redundancy, on the other hand, requires a significant amount of computation time without adding any valuable information. CWT has an unmanageable infinite number of wavelet coefficients, and for most functions, CWT has no analytical solutions and therefore is solved numerically only [43]. The redundant coefficients in CWT can be removed by sampling both the scale and time at powers of two (dyadic sampling), thus leading to DWT. DWT provides sufficient information both for analysis and synthesis of the original signal, with a significant reduction in the computation time. The DWT is determined by applying the concept of the multiresolution analysis (MRA) [44]. The practical implementation of MRA can be done using a filter bank [17]. The process involves using successive, complementary low-pass g[n] and high-pass h[n] filters to split the signal under

256

Optical Wireless Communications

(a)

(b)

FIGURE 5.24  DWT: (a) decomposition of a signal and (b) decomposition process.

analysis into its approximation and detail coefficients and downsampling by two. The high-pass parts contain generally no significant information about signal, and hence they are not decomposed further. However, the low-pass parts are further decomposed into two bands. This process will continue until a satisfactory level of information is obtained. Figure 5.24 represents the process of splitting the spectrum using a filter band and a corresponding wavelet decomposition tree with the approximation and detail coefficients. The first level of approximation y1l and detail coefficients y1h are given by [32] y1h ( k ) =



∑y ( n) g ( 2k − n) (5.35) n

y1l ( k ) =



∑y ( n) h ( 2k − n) (5.36) n

The high-pass and low-pass filters satisfy the condition of the quadrature mirror filter, i.e., the sum of the magnitude response of filters is equal to unity for all frequencies. The approximation coefficients can further be decomposed into different DWT coefficient levels with a maximum level of log2Ls, where Ls is the signal length. The original signal can be reconstructed by the inverse of the decomposition process, using the following: [45]

x ′ (n) =

∑(y

kh

( n ) .g ( 2k − n )) + ( ykl ( n ) .h ( 2k − n )) (5.37)

k

5.5.3  DWT Based Denoising For the OOK system, the Rx design based on the DWT is given in Figure 5.25. This is very similar to the Rx structure shown in Figure 5.1 except for the wavelet denoising block. Wavelet denoising involves decomposition of the signal y(n) into different DWT levels, processing of the DWT coefficients, and reconstruction. In the first level of decomposition, y(n) is split into low-pass and high-pass signals as given by (5.35) and (5.36), respectively, whereas in the second stage the low-pass band signal y1l is further

257

Indoor System Performance Analysis

Matched filter r(t)

y(t) y(n) sample

DWT

Wavelet Denoising

Processing

Output bit

x´(n)

IDWT

FIGURE 5.25  DWT-based Rx in the presence of artificial light interference.

divided into low-pass y2l and band-pass y2h components. Filtering and decimation are continued until the required level is reached. The task here is to separate the interference and the modulating signal in the DWT domain and remove the interfering signal from the reconstructed signal. Since the FLI signal spectral component is mostly based at the lower frequency region, the received signal is decomposed until the DWT coefficients of the interfering signal are concentrated on the approximation coefficients. For removing the interfering signal from the received signal, the approximation coefficients, which correspond to the interfering signal, are then made equal to zero so that the reconstructed signal is interference free i.e.,

yγh ( k ) = 0 (5.38)

where γ is the number of the decomposition levels. The signal is then reconstructed back using the inverse of the decomposing process as given by (5.37). DWT decomposes the signal in the logarithmic scale; therefore, it is difficult to define a precise band of frequencies, which corresponds to the lowest level of approximations. On the other hand, due to the spectral overlapping between the interfering and the modulating signals, it is thorny to precisely define the cut-off frequency for the optimal performance. In the previous section, the performance of the system employing an HPF with a given fc-hfp, which resulted in the least BLW effect, was discussed. However, the definition of such precise cut-off frequency is not possible using the DWT scheme. Therefore, the number of decomposition level γ is determined in such a way that the lowest level of approximation signal is within the 0.5 MHz range in order to make fc-hpf ∼0.5 MHz (except for Rb < 20 Mbps, for which fc-hpf is taken as 0.3 MHz). Note that fc-hpf is chosen as 0.5 MHz as studies have shown that it provides near optimal performance in the presence of FLI when using a digital HPF [45]. The number of decomposition levels is given by

  Fs   γ dl = −  log2   (5.39)  fc − hpf     

where .  is the floor function. It is to be noted here that fc-hpf varies with Rb and the decomposition level. DWT offers flexibility in the analysis where different mother wavelets could be adopted as necessary for the particular application. The NOPRs linked to different mother wavelets are listed in Table 5.1. Only selected mother wavelets from the Daubechies (db), Symlet (sym), discrete Meyer wavelet (dmey), biorsplines (bior), and Coiflets (coif) families with the best and the worst performance are listed. The Haar wavelet (db1) and the db8 show the worst and best performance, respectively, followed closely by the db10 and sym7. Hence the db8 is adopted for discussion in this chapter. The Matlab code for simulating the wavelet-based denoising is given in Program 5.5. See Program 5.1 for all the parameters and for calculation of the error probability for a range of SNR. A similar approach can be applied to other modulation schemes, including PPM and DPIM.

258

Optical Wireless Communications

TABLE 5.1 The NOPR to Achieve a BER of 10 −6 at R b of 200 Mbps for DWT Based Rx with Different Mother Wavelets [46] Mother wavelets db1 db2 db5 db8 db10 Coif2 Coif5 Sym3 Sym4 Sym7 dmey Bior2.2

SNR (dB)

NOPR (dB)

22.01 20.9 16.1 15.2 16.2 16.6 15.4 17.2 15.7 15.3 15.7 16.7

15.7 15.2 12.9 12.3 12.8 13.0 12.4 13.3 12.6 12.4 12.6 13.1

The NOPR for OOK schemes with the DWT denoising in the presence of FLI is shown in Figure 5.26. Also shown is the NOPR for the ideal channel without interference. Since no improvement can be achieved below R b of 10 Mbps, lower R b are not considered here. Given that OOK is very susceptible to the BLW, only low normalised cut-off frequencies (normalised to sampling rate) are possible, which are not effective in reducing the interference. At R b above 10 Mbps, the DWT is very effective in reducing the interference. At an R b of 10 Mbps, DWT offers a reduction of ∼5.5 dB in NOPR compared to the system without filtering. A higher reduction level of ∼7.4 dB is observed at 40 Mbps. However, this still leaves an OPP of ∼1.6 dB compared to the same R b with no interference. OPP is further reduced at higher R b, e.g., ∼1.2 dB at 100 Mbps. Note that the optimum HPF still shows an OPP of 3 dB at 100 Mbps. 18 16

NOPR (dB)

14 12 10 8 6 4 10

NOPR (without interference) NOPR (with interference) NOPR (with interference and DWT) 100 200 Data rate (Mbps)

FIGURE 5.26  The normalised optical power requirement against the data rates for OOK with and without DWT denoising in the presence of FLI.

Indoor System Performance Analysis

259

Program 5.5:  Matlab codes to simulate wavelet-based denoising for OOK-NRZ in the presence of FLI. OOK=randint(1,sig_length); % random signal generation OOKm=[zeros(1,Lfilter) OOK zeros(1,Lfilter)]; Tx_signal=rectpulse(OOKm,nsamp)*i_peak; % Pulse shaping function (rectangular pulse) Rx_signal=R*Tx_signal+sgma*randn(1,length(Tx_signal)); % received signal (y=x+n); % %*****************Effect of FL**************** start_time=abs(rand(1,1))*10E-3; end_time=start_time+Tb*nsamp*sig_length; Ib=2E-6; % average current due to the fL i_elect=fl_model(Ib,Tsamp,start_time,end_time); % FLI model Rx_OOK_fl=Rx_signal+i_elect(1:length(Rx_signal)); % Interference due to FL MF_out=conv(Rx_OOK_fl,rt)*Tsamp; % matched filter output MF_out_downsamp=MF_out(nsamp:nsamp:end); % sampling at end of bit period MF_out_downsamp=MF_out_downsamp(1:length(OOKm)); % truncation; %% **************** wavelet denoising of the signal [C,L] = wavedec(MF_out_downsamp,Lev,wname); cA=appcoef(C,L,wname,Lev); C(1:length(cA))=0; Rx_OOK = waverec(C,L,wname); % reconstructed signal Rx_OOK=Rx_OOK(Lfilter+1:end-Lfilter); Rx_OOK=mapminmax(Rx_OOK); Rx_th=zeros(1,sig_length); Rx_th(find(Rx_OOK>0))=1; % thresholding

The NOPR to achieve an error probability of 10 −6 with the DWT denoising for 4, 8, and 16-PPM with the HDD scheme for 1 Mbps < Rb < 200 Mbps is illustrated in Figure 5.27. With reference to Figure 5.4, it is clear that DWT denoising shows marked reduction in NOPR (i.e., ∼11 dB, 12 Db, and 12 dB for 4, 8, and 16-PPM, respectively) at Rb of 1 Mbps. Also notice the significant reduction in power requirements at Rb of 1 Mbps compared to HPF (see Figure 5.18). Since the PPM frequency spectrum is mostly at the higher frequency region with no DC component at all, a higher fc-hpf could be used without experiencing the BLW effect. Hence for the PPM scheme, the DWT denoising technique offers improvement even at low frequencies. Above Rb of 10 Mbps for 4-PPM and 5 Mbps for 8 and 16-PPM, the DWT-based denoising completely eliminates the FLI-induced power penalties. The NOPR for 4, 8, and 16-PPM (SDD) schemes with DWT denoising at an error probability of 10 −6 and for 1 Mbps < Rb < 200 Mbps is depicted in Figure 5.28. Compared with Figure 5.5, DWT offers no improvement above Rb of 20 Mbps as PPM(SDD) is immune to FLI. DWT provides a significant improvement even at 1 Mbps with the power penalties of ∼4 dB, ∼3 dB, and ∼1 dB for 4, 8, and 16-PPM, respectively, compared to the same Rb with no interference. For 8 and 16-PPM, the power penalties without DWT are 11.45 dB and 13 dB, respectively, thus illustrating the effectiveness of the DWT denoising. At 10 Mbps, the FLI-induced OPP is 1.1–2.1 dB, which is completely eliminated when using DWT denoising.

260

Optical Wireless Communications 10 8

NOPR (dB)

6 4 2 0

4-PPM (without interference) 8-PPM (without interference) 16-PPM (without interference) 4-PPM (with interference and DWT) 8-PPM (with interference and DWT) 16-PPM (with interference and DWT)

-2 -4 -6

1

10 Data rate (Mbps)

100

200

FIGURE 5.27  NOPR versus the data rates for 4, 8, and 16-PPM with the hard decoding scheme with DWT denoising in the presence of FLI.

For an SER of 10 −6 with DWT denoising, the NOPR for 4, 8, and 16-DPIM at 1 Mbps < Rb < 200 Mbps is shown in Figure 5.29. Comparing the performance of the OOK scheme in Figure 5.26 and PPM(HDD) in Figure 5.27, it can be observed that the performance of DPIM with DWT is intermediate between OOK and PPM. Low-order DPIM shows performance similar to OOK with constant OPPs compared to the ideal case. However, higher orders of DPIM show almost zero power penalties. The phenomenon is due to the progressive reduction of the DC and low-frequency components with the increasing bit resolutions. Compared to the performance without FLI, the power penalties

8 6

NOPR (dB)

4 2 0 -2

4-PPM (without interference) 8-PPM (without interference) 16-PPM (without interference) 4-PPM (with interference and DWT) 8-PPM (with interference and DWT) 16-PPM (with interference and DWT)

-4 -6 1

10 Data rate (Mbps)

100

200

FIGURE 5.28  NOPR versus the data rates for 4, 8, and 16-PPM with the soft decoding scheme with DWT denoising in the presence of FLI.

261

Indoor System Performance Analysis 12

10

NOPR (dB)

8

6

4

2

0 10

4-DPIM (without interference) 8-DPIM (without interference) 16-DPIM (without interference) 4-DPIM (with interference and DWT) 8-DPIM (with interference and DWT) 16-DPIM (with interference and DWT) Data rate (Mbps)

100

200

FIGURE 5.29  NOPR versus the data rates for 4, 8, and 16-DPIM schemes with DWT denoising in the presence of FLI.

with DWT are ∼0.7 dB at Rb > 40 Mbps, which reduces to 0.4 dB for 8-DPIM. For 16-DPIM at Rb > 20 Mbps, the power penalties are completely reduced to zero when using the DWT denoising scheme.

5.5.4 Comparative Study of DWT and HPF This section provides the comparative study of adopting the DWT and HPF for reduction of FLI. Because of spectral overlapping of the modulating and the interference signals, it is, in fact, a challenging task to reduce the effect of FLI without incurring OPPs, especially for modulating techniques with the higher level of DC component. Since OOK has the highest DC and the lower frequency components, it suffers most adversely from the application of filtering. Hence the need for high-quality denoising techniques with reduced complexity and minimal information loss. A case study for OOK with HPF and DWT denoising is presented to illustrate the effectiveness of the proposed schemes. It is simple to conclude that a filtering technique which provides the least OPP is also the best option for other modulation techniques. This is because OOK suffers most severely from BLW. Digital HPF shows significantly improved performance compared to its analogue counterpart, and therefore it is adopted here. For more on the performance of OOK and other modulation techniques with analogue HPF, readers should refer to [3], [6], [11]. The OPPs against the decomposition level γ for DWT for Rb of 20, 50, 100, and 200 Mbps are depicted in Figure 5.30, whereas OPPs for different HPF cut-off frequencies and for the same Rb are illustrated in Figure 5.31. For DWT, the lowest threshold level is at γ = 9 for 200 Mbps. Since DWT decomposes signals in a logarithmic scale, the value of γ at which the OPP is at its minimum level increases by one on doubling Rb. Note that (i) for all Rb the optimum fc-hpf is ∼0.2 MHz, and (ii) it is not possible to define a precise fc-hpf for DWT (as only the integer value of γ is permissible). The results shown in Figure 5.30 and Figure 5.31 indicate that the wavelet-based Rx, with no parameter optimisation, displays similar or improved performance compared to the best performance achieved using the HPF. This makes the denoising process faster and more convenient. In addition to the performance improvement, a key advantage of DWT is the reduced system complexity. For the optimum performance, γ varies within the range of 6 to 9 (see Figure 5.31).

262

Optical Wireless Communications 8 20 Mbps 50 Mbps 100 Mbps 200 Mbps

OPP (dB)

6

4

2

0

5

6

7

8 9 Decomposition level

10

11

FIGURE 5.30  The OPP against the decomposition level for the DWT based Rx at the data rates of 20, 50, 100, and 200 Mbps.

For an input signal length of Ls, the total number of floating point operations (only multiplication is taken into consideration here) for the first level of decomposition is LsJ. For the second stage, the length of the input signal is Ls/2, and the total number of operations is LsJ/2, and so on. Hence, the DWT requires a maximum of 2LsJ operations for analysis and synthesis, meaning the maximum number of floating point operations is 4LsJ. Since J = 15 for Daubechies 8 ‘db8’, the maximum number of the operations is 60n. For an HPF of order nf, the total number of floating point operations is Lsnf /2. At a data rate of 200 Mbps and with fc-hfp of 0.2 MHz, the filter order nf is 2148, thus illustrating much-reduced complexity when using DWT. Moreover, the realisation of DWT is also simpler than HPF because a repetitive structure is used at each level of analysis and synthesis. Hence only a 15th-order filter is required to realise DWT compared to a filter of 2148 orders for HPF.

5.5.5  Experimental Investigations 5.5.5.1  On-Off Keying (OOK) The non-directed LOS-OWC system is deployed in a typical 6 × 5 × 3 m3 laboratory room environment with a Rx located at a height of 1 m above the floor. The schematic diagram of the experimental set-up is shown in Figure 5.32 (note that it is not to scale). The pseudo-random bit sequence 8

20 Mbps 50 Mbps 100 Mbps 200 Mbps

OPP (dB)

6 4 2 0

0

0.5

1 Cut-off frequency (MHz)

1.5

2

FIGURE 5.31  The OPP against cut-off frequencies for an HPF based Rx at the data rates of 20, 50, 100, and 200 Mbps.

9781498742696_Ch05.indd 262

17/04/19 4:49 PM

263

Indoor System Performance Analysis

LD Driver

Data Source

LD 90 cm

LSD

Lens Data Acquisition

TIA

PD

FIGURE 5.32  The schematic diagram of the experimental set-up for an indoor OWC link.

(PRBS) of 210 −1 bits is generated using an arbitrary waveform generator (AWG) and is converted into a non-return-to-zero (NRZ) OOK format prior to intensity modulation of a laser diode operating at a wavelength of 830 nm with a maximum optical output power of 10 mW. A holographic diffuser of 10° full-width half-maximum (FWHM) is used to ensure eye safety as well as to increase the optical footprint. The Rx is positioned at the centre of the optical footprint where the received optical power is at its maximum value. The Rx consists of an optical concentrator (a focal length of 5 cm and a diameter of 2.5 cm), and a PD with a daylight filter at 800–1100 nm wavelength range. The peak spectral sensitivity of the PD is 0.59 A/W at 950 nm wavelength with an active area of 7 mm2. The Rx optics (lens and PD) are adjusted to obtain the maximum optical gain. The photocurrent at the PD is amplified using a commercial transimpedance amplifier (TIA), followed by a data acquisition system in order to acquire, process, and analyse the real-time data. The experiment is carried out in two settings: (i) in complete darkness; and (ii) in the presence of the ambient light source, i.e., a fluorescent light lamp located at the ceiling 3 m above the floor and 2 m above the Rx. The Rx is located directly underneath the light source, and the illuminance of the light sources at the Rx position is ∼350 lux. The waveform and the PSD of the fluorescent light are measured using the OSRAM (SFH205F) PD and are given in Figure 5.33. The waveform of FLI is a distorted sinusoidal signal at a fundamental frequency of 500 kHz with harmonics extending up to 0.5 MHz; see Figure 5.33(b). The time waveform of the 10 Mbps OOK-NRZ signal in the presence of FLI is depicted in Figure 5.34. It can clearly be observed that the photocurrent due to the inference is higher than the modulating signal. As a result, it is impossible to achieve an OWC link with a low error probability without incorporating a denoising technique.

Power/frequency (dB/Hz)

6 Voltage (mV)

4 0 -4 -8

0

10

20 30 Time (µs)

(a)

40

50

0 -20 -40 -60 -80

0

0.1

0.2 0.3 Frequency (MHz)

0.4

0.5

(b)

FIGURE 5.33  (a) Time waveform and (b) spectrum of the interference produced by an artificial light source.

264

Optical Wireless Communications 7

Amplitude (mV)

4

0 -4 -7

0

5

10

15 Time (µs)

20

25

30

FIGURE 5.34  The time waveform of the received signal at 10 Mbps in the presence of FLI.

In order to quantify the OWC link performance, the Q-factor is measured with and without FLI. Note that the Q-factor is a measure of the optical signal-to-noise-ratio (OSNR) and hence can be used to assess the quality of the link, which is given by

Q=

vH − vL (5.40) σH − σL

where v H and v L are the mean received voltages and σ H and σ L are the standard deviations for the ‘high’ and ‘low’ level signals, respectively. In order to verify the experimental results, computer simulation of the OWC system under test is also carried out using the measured parameters, including the noise variance, the FLI photocurrent, the average modulating signal photocurrent, and the impulse response of the system (Tx, channel, and Rx). The AC coupling capacitor at the Rx, which induces a BLW effect in baseband modulation schemes [6], [9], is modelled as a first-order analogue RC-HPF with a cut-off frequency of 10 kHz. The measured and simulated Q-factors against Rb with and without FLI are given in Figure 5.35, showing a good agreement with a difference of Ep/2))=1;

Program 5.8:  Matlab codes to simulate the error probability of OOK-NRZ in a multipath channel based on discrete-time equivalent system (Figure 5.38). %% system impulse response Tx_filter=ones(1,nsamp); Rx_filter=fliplr(Tx_filter); c=conv(Tx_filter,h); c=conv(c,Rx_filter); delay=find(c==max(c)); if delay>nsamp; hi(1)=c(delay-nsamp); % taking precursor tap else hi(1)=0; end hi=[hi(1) c(delay:nsamp:end)]; % channel impulse response hi=hi/sum(hi); % normalization OOK=randint(1,sig_length); % random signal generation Rx_signal=conv(OOK, hi); % Received signal, after matched filter (without noise) Rx_signal=Rx_signal(2:sig_length+1); % truncation to avoid precursors MF_out=awgn(Rx_signal,EbN0_db+3,’measured’); % MF output with noise Rx_th=zeros(1,sig_length); Rx_th(find(MF_out>0.5))=1;

The unequalised NOPR for OOK to achieve a BER of 10 −6 in a diffuse channel with DT of [0, 1] is given in Figure 5.41. Note that R b used is irrelevant as the RMS delay spread is normalised to T b. It is clearly apparent from Figure 5.42 that the unequalised NOPR for OOK increases exponentially with DT . An NOPR of 3 dB is incurred when DT is ∼0.23 increasing to 18 dB for DT of 0.51, where the target BER of 10 −6 is practically impossible to achieve with a reasonable optical power.

271

Indoor System Performance Analysis 18 X: 0.51 Y: 17.9

16 14

N O P R (d B )

12 10 8 6 4 X: 0.24 Y: 3.087

2 0 -2 10

-1

0

10 DT

10

FIGURE 5.41  NOPR against the normalised delay spread DT for unequalised OOK in a diffuse indoor OWC channel.

At lower BER, the degradation due to multipath propagation is dominated by the worst case bit sequence, which consists of a single bit preceded and followed by a long string of zeros, and the unequalised OPP can be approximated as given by [49]

(

)

  Q −1 2ζ  ξ  (5.48) unequalised OPP  ( dB) = 10log10  −1  ( 2h0 − 1) Q ( ξ) 



It can be noted that, at low BER, OPP depends only on h0 and ζ. The simulation and approximated (using (5.48)) OPPs for OOK against the channel tap with highest amplitude h0 are given in Figure 5.42.

Theoretical Simulation

Opti cal pow er penal ty (dB )

16 14 12 10 8 6 4 2 0

0.5

0.6

0.7

h0

0.8

0.9

1

FIGURE 5.42  Theoretical and simulated OPPs for the unequalised OOK system against h0.

272

Optical Wireless Communications

OPP increases with decreasing value of h0, which is expected for a channel with severe ISI where h0 reduces. For h0 ≤ ½, the expression (5.48) yields infinite OPP, which can be seen from the asymptotic increment in OPP with decreasing h0.

5.6.2 Pulse Position Modulations (PPM) To analyse the effect of multipath-induced ISI, the discretisation approach discussed in the previous section can be adopted. The discrete-time equivalent system for a PPM scheme is depicted in Figure 5.43. Notice that both HDD and SDD can be adopted as necessary. The detailed descriptions of the MF-based Rx for the PPM system including SDD and HDD schemes are discussed in Chapter 4. To analyse the slot error probability for PPM(HDD), one can adopt the same analysis as in OOK. However, due to the difference in the sampling rate, the discrete-time impulse response ci   of the cascaded system is given as ci = p ( t ) ⊗ h ( t ) ⊗ r ( t ) t = i Tb M (5.49)



L

Hence, the received signal yi in the absence of noise is given by yi = LRPavg bi ⊗ hk



k =ζ

; (5.50)

Comparing (5.43) and (5.49), it can be observed that, for the same R b, the PPM will have higher channel span due to shorter pulse duration, and hence will result in higher ISI for the same normalised delay spread (normalised to bit duration). As in the case of OOK, to estimate the error probability of PPM in a multipath channel, one needs to consider all possible combinations of the PPM symbols within the channel span ζ, calculate the slot error probability of individual slot, and average it over the entire channel span. Hence the SER of PPM is given by



Pse _ PPM _ hard =

1 2ζ

2ζ



 yi − α opt N0 2

∑  L1 Q  i

 ( L − 1)  α opt − yi Q   + L   N0 2

       (5.51)  

Unlike OOK, α opt does not lie in the middle of the 0 and 1 levels due to unequal probability of occurrence of the empty slots and pulses. In fact, α opt is a complicated function of the bit resolution and hi. An iterative approach to determine α opt and approximate Pse _ PPM _ hard is adopted in [11], which requires significant computational time and thus is not practical for ζ > 20 [11]. Here, a constant threshold level α th is utilised, which is given by ∞

∑h (5.52)

α th = 0.5 − 0.25



i

i =1

ak

PPM bi encoder

yi

ci LPavg

ˆ PPM b i decoder

ni

FIGURE 5.43  Discrete-time equivalent system for PPM scheme.

9781498742696_Ch05.indd 272

17/04/19 4:50 PM

273

Indoor System Performance Analysis

The summation ∑ i∞=1 hi provides an estimation of the energy of a pulse, which spreads to its adjacent pulses due to dispersion. For a LOS link, the summation is zero and hence α th = 0.5. The dispersive channel will have a non-zero summation value, therefore α th < 0.5. Using SDD, a pulse is detected depending upon its relative amplitude within a symbol. Assuming that 1 occurs in the slot q of each symbol, the average symbol error probability for unequalised PPM(SDD) is given by [11], [50]

Psymbol _ PPM _ soft



1 = ζ 2

ζ

L

i =1

p =1 p≠ q

 y(i −1) L + q − y(i −1) L + p   (5.53) N0

∑ ∑ Q 

The NOPR to achieve a SER of 10 −6 for unequalised 4, 8, and 16-PPM with SDD and HDD schemes are shown in Figure 5.44. The threshold level given in (5.52) is used for the HDD scheme. Also shown is NOPR for OOK for comparison. For any given order and delay spread, the SDD scheme has a lower power requirement than HDD. For an ideal channel, this difference is ∼1.5 dB, but for multipath channels, the difference increases with DT . The higher-order PPM shows a sharp increment in NOPR with increasing DT for both SDD and HDD schemes. This is due to a decrease in the slot duration for higher order PPM, effectively increasing the ISI. For example, for DT > 0.19, 4-PPM(HDD) offers lower NOPR to achieve a SER of 10 −6 compared to 16-PPM(HDD). For the SDD scheme, the intersection point for 4-PPM and 16-PPM is ∼0.24. It can also be observed that OOK offers the least OPP compared to all orders of PPM(HDD) for a highly dispersive channel (i.e., DT > 0.27), and 4-PPM (SDD) offers the lowest power requirement for higher DT values. Note that the difference in NOPR for the SDD and HDD schemes increases from 1.5 dB for DT = 0 to ∼8.5 dB for DT = 0.4 for 4-PPM with the SDD always requiring lower NOPR. The difference is even higher for 8 and 16-PPM with a noticeable difference of ∼12.5 dB at DT = 0.23 for 16-PPM. Despite the improved power efficiency of SDD, OOK outperforms 16-PPM for DT > ∼0.3 and offers a similar average OPR to 8-PPM for DT ∼0.3. Furthermore, compared with PPM(HDD), OOK offers a lower average OPR than any orders of PPM considered for DT > ∼0.27. This clearly demonstrates the severity of multipath-induced ISI in pulse modulation schemes. 12 10 8

NOPR (dB)

6 4

OOK 4-PPM (hard) 8-PPM (hard) 16-PPM (hard) 4-PPM (soft) 8-PPM (soft) 16-PPM (soft)

2 0 -2 -4 -6 -8 -2 10

-1

10 DT

0

10

FIGURE 5.44  NOPR against the normalised delay spread for 4, 8, and 16-PPM with hard and soft decision decoding schemes in a dispersive channel.

274

Optical Wireless Communications

5.6.3 Digital Pulse Internal Modulation(DPIM) Following the similar approach adopted for PPM, the effect of multipath-induced ISI on the DPIM scheme can be approximated. Similar to OOK and PPM schemes, to determine the error probability, all possible DPIM sequences of at least ζ slots are generated. The SER is approximated by averaging over all possible slot sequences of length ζ and is as follows:



Pse _ DPIM

1 = ζ 2

2ζ



∑  L 1 i

DPIM

 yi + α opt  ( LDPIM − 1)  α opt − yi   Q Q + L   (5.54) DPIM  N0 2   N 0 2  

Normalized optical power requirement (dB)

For each value of DT in the range 10 −3 to 0.4, the average OPR for various orders of DPIM(NGB) is illustrated in Figure 5.45. For comparison, the average OPR for OOK is also shown. With dispersion or with DT being small, PPM(HDD) has a lower average OPR than DPIM for any given order. This is due to the lower average duty cycle of PPM, which results in increased power efficiency. However, due to its lower bandwidth requirement, DPIM has a lower ISI-induced OPP than PPM. Consequently, as DT increases, the average OPR curves for the two schemes intersect, and beyond the point of intersection, DPIM offers the lower OPR of the two schemes. In contrast, for a given order, PPM(SDD) always achieves a lower OPR than DPIM. Compared with OOK, all orders of DPIM considered offer a lower average OPR for DT below ∼0.18. However, beyond ∼0.26 it is the OOK which yields the lowest OPR. For a DPIM slot sequence propagating through a multipath channel, the post-cursor ISI is the most severe in slots immediately following a pulse. Hence, placing a guard slot or slots immediately following the pulse increases the immunity of DPIM to the multipath-induced ISI [5]. At the Rx, upon detection of a pulse, the time slot or slots within the guard band are automatically assigned as zeros, regardless of whether or not the sampled output of the Rx filter is above or below the threshold level. Thus, the postcursor ISI present in these slots has no effect on system performance, provided the pulse initiating the DPIM symbol is correctly detected.

12 10 8

4-DPIM(0 GS) 8-DPIM(0 GS) 16-DPIM(0 GS) OOK

6 4 2 0 -2 -4 0.01

0.1 Normalized delay spread

0.5

1

FIGURE 5.45  NOPR against the normalised delay spread for 4, 8, and 16-DPIM with and without guard slots in a dispersive channel.

9781498742696_Ch05.indd 274

17/04/19 4:50 PM

Indoor System Performance Analysis

275

5.7  MITIGATION TECHNIQUES One of the major challenges in diffuse indoor OWC channels is multipath propagation. A single reflected signal would have the minimum delay, except for the LOS link, and the maximum power provided all surfaces attenuate the signal equally. However, the signal may arrive at the Rx via different paths, thus resulting in delayed and attenuated signals making additional contributions to the dominant reflected signal; this results in a ‘smearing’ of the pulse shape, leading to ISI. Modulation schemes with a low duty cycle are most suitable for indoor OWC systems due to the limited power available to portable devices and to eye safety requirements. In LOS links, reducing the modulation duty cycle can result in an improved SNR performance for a given BER. However, non-LOS configurations with multipath low duty cycle signals (i.e., shorter pulse duration) will experience a higher level of pulse spreading due to the increased number of reflections. The MF Rx with a fixed threshold level detection discussed in the previous section clearly showed that OPP increases exponentially with the channel delay spread DT , and the OPP is higher for modulation schemes with shorter pulse duration. Given the performance constraints of a simple thresholdbased Rx on a noisy multipath channel, one needs to find an alternative mitigation scheme in order to reduce the degrading effects of ISI and noise. Most techniques employed in diffuse indoor OWC systems are implemented in the electrical domain and are borrowed from the more established RF technologies. In this section, a selection of these methods is reviewed and described.

5.7.1 Filtering One of the simplest and most frequently implemented methods used to improve the performance of communication systems is filtering. In OPC systems, an optical filter placed in front of the PD is used to reduce all out-of-band natural and artificial light signals, thus improving the SNR. Additionally, electrical analogue or digital filters improve the SNR by rejecting the high-frequency components, which are not associated with the received signal. One such filter is the MF, which is the optimum filter for digital communications with the AWGN channel. The perfect way of implementing this scheme is by making the normalised gain response of the Rx filter Hr ( f ) identical to the amplitude spectrum of the pulse Gr ( f ) as in (5.55) [51]. This well-known technique is abundantly documented in the literature [51], [52] and can easily be approximated in practical and simulated situations by the simple integrate and dump circuitry; because the average power of the noise is zero, such circuitry maximises the SNR in the same way.

Hr ( f ) = G ( f ) (5.55)

The presence of multipath-induced ISI will severely affect the performance of an MF-based receiver if it is matched to the Tx filter and the channel distortion is not taken into consideration; therefore matching will no longer be accurate. Often, standard low-pass, high-pass, or band-pass filters are used to lessen the effects of noise in communication systems. For simple indoor diffuse OWC systems, a low-pass filter would be used to reject the high-frequency components of the noise. Carruthers et al. [53] suggest a normalised fc of 0.6/Tb for the unequalised OOK-NRZ data format. However, simulations of such filters in a basic threshold-based Rx and a multipath channel show that the optimum normalised fc is nearer to 0.7/Tb. Figure 5.46 illustrates the BER against the SNR for 50 Mbps OOK-NRZ and 25 Mbps OOK-RZ data formats for a range of fc showing almost the same profile except for an fc of 0.5/Tb. Figure 5.47 depicts the BER performance of a threshold-based Rx against the SNR for OOKNRZ/RZ with fc of 0.6/Tb and 0.35/Tb, respectively. Lower Rb NRZ shows improved BER performance compared with higher Rb. Although the filtered RZ case is an improvement over the unfiltered

276

Optical Wireless Communications 10 10

BER

10 10 10 10 10

0

0.5/T 0.6/T 0.7/T 0.8/T 0.9/T

-1

-2

-3

-4

-5

-6

-10

-5

0

5 SNR (dB)

10

15

20

(a)

10 10

BER

10 10 10 10 10

0

0.25/T 0.3/T 0.35/T 0.4/T 0.45/T

-1

-2

-3

-4

-5

-6

-10

-5

0

5

10

15

SNR (dB)

(b) FIGURE 5.46  BER against the SNR for: (a) 50 Mbps OOK-NRZ and (b) 25 Mbps OOK-RZ for a range of cut-off frequencies.

case, its BER performance advantage over NRZ has disappeared completely at lower Rb. Given these results, it is clear that simple low-pass filtering alone cannot compensate for the multipath-induced distortion in short duration-based pulse modulation schemes such as OOK-RZ and PPM.

5.7.2 Equalisation Equalisation is concerned with compensating for imperfections in the channel. In the simplest case the equaliser filter has the inverse characteristics of the channel, i.e., H ( f ) = Y (1f ) , where Y(f) is the Fourier transform of the output of the system response at the output of the MF, essentially

277

Indoor System Performance Analysis 100 10-1

BER

10-2

10Mb/s RZ 25Mb/s RZ 33.3Mb/s RZ 35.7Mb/s RZ 10Mb/s NRZ 25Mb/s NRZ 33.3Mb/s NRZ 35.7Mb/s NRZ 41.66Mb/s NRZ 50Mb/s NRZ 71.4Mb/s NRZ

10-3 10-4 10-5

10-6 -10

-5

0

5

10

15

20

25

30

SNR (dB)

FIGURE 5.47  BER against SNR for OOK-NRZ/RZ with 0.6/Tb and 0.35/Tb low-pass filtering.

restoring the transmitted pulse shape at the Rx. Equalisers are typically used to equalise the channel response, which can be dispersive or have fading properties in both indoor and outdoor environments. Considering OOK, the output of the MF is given by

y ( ts ) =

∞

∑ a p[(i − k ) T ] + n (t ) (5.56) k

s

b

k =−∞

where p ( t ) = s ( t ) ⊗ g’( t ) ⊗ h ( t ) ⊗ r ( t ) , s ( t ) is the Tx response, g’(t) the LED driver response, h(t) is the channel response, r(t) is the Rx response, and ak is the kth OOK symbol. (5.56) can be written as

y ( ts ) = ai +

∞

∑ a p[(i − k ) T ] + n (t ) (5.57) k

b

s

k =−∞ k ≠i

where the first term is the amplitude of the ith received symbol, and the second term represents the contribution of ISI from the preceding and subsequent symbols. Therefore, in order to achieve zero ISI, the following condition must be met [51]

1 p [( i − k ) Tb ] =  0

i=k i≠k

(5.58)

This implies that the Rx output y ( ts ) = ai , i.e., no ISI. In OWC and in particular in VLC systems, the factor that contributes to the deterioration of the overall system response, i.e., introducing ISI, is the low-pass transfer functions of the light sources (LEDs). Equalizing the low-pass response will increase the data rate significantly in the presence of a high SNR. Note (i) equalisation does not equalise the effects of noise, (ii) the system response is found by a pilot signal, and (iii) careful filter design is essential to maximise the effectiveness of the equaliser. Although the characteristics of an indoor OWC channel are quasi-static, they are not uniform for any given Rx-Tx pair. For example, moving an Rx from one part of a room to another will alter h(t) as seen by the Rx.

278

Optical Wireless Communications

Therefore, a common set of filter coefficients cannot be adopted at the Rx. It is common that a training pulse is transmitted to the Rx in order to adjust the weights of the filter taps. Such a filter is usually referred to as a pre-set equaliser since the coefficients are calculated prior to transmission. There are two main types of filters; analogue and digital. Each has intrinsic advantages and disadvantages. In the analogue domain, an HPF RC equaliser, which is simple to implement, is the only real choice with the frequency response given by 1 + j 2 πf τ (5.59) 1 + j 2 πf τk

H RC ( f ) = k −1



where k−1 = R L/(R + R L) is the DC coefficient of the RC equaliser, R is the equaliser resistor value, R L is the load resistor (typically 50 Ω), and τ = RC is the time constant. In order to equalise the low-pass frequency response, the high-pass slope SH must be opposite the low-pass slope SL, i.e., 1 S L = − S H = 6 πτ , which leads to an approximate bandwidth B ≤ 20 logSH10 ( k ) . Note that B is 1− 2 k2

dependent on k, and the selection of k decides the margin of equalisation for the system; a high value of k allows for more equalisation and vice versa. Note that k, however, cannot be increased indefinitely, and the limit is imposed by the Rx’s input dynamic range. The RC equalisers have drawbacks. By introducing an HPF, there is an exponential power penalty around the low-frequency region, which leads directly to the BLW as was discussed. In the digital domain, the zero forcing equaliser (ZFE) and the decision feedback equaliser (DFE) are the most popular. 5.7.2.1  The Zero Forcing Equaliser The ZFE implemented as a transversal filter or the FIR filter consists of a taped delay line with an input Xk, usually tapped at intervals of T seconds, where T is the symbol interval. The output of each tap is weighted by a variable gain factor aN , and each weighted output is summed to form the final output of the filter yk at a particular time kT; see Figure 5.48. There are (2N + 1) taps, where N is chosen large enough to span the ISI with the corresponding weights {a− N … a0 … aN ..}. The minimisation of ISI is achieved by considering only those inputs that appear at the correct sample times. For convenience, x ( kT ) = x k and y( kT ) = yk . The output can be expressed in terms of the inputs and tap weights as given by N

yk =



∑a x

n k −n

(5.60)

n =− N

Therefore  1 when  k = 0  yk =  0 when  k = ±1,  ±2,… ± N 



(5.61)

A transversal equaliser can force the output to go to zero at N-point either side of the peak output and the (2N + 1) equations can be solved in the matrix form as follows:  x0   x N −1    x N     x 2 N −1   x 2 N

x − N +1 x0  x N −1  xN x 2 N −1

 

x− N x − N +1

 

x −2 N +1 x− N



x0



x − N +1

 

x N −1 xN

 

x0 x N −1

x −2 N   a− N  x −2 N +1   a− N +1   x− N   a0    x − N + 1   a N −1  x0   aN

        =          

0 0  1  0 0

        

(5.62)



279

Indoor System Performance Analysis

(a)

(b)

FIGURE 5.48  Structure of the (a) ZFE and (b) adaptive linear transversal equaliser.



a = X Tq



(5.63)

where a is the filter coefficient array, X is the sample point matrix, and q is the output array. Note that the filter coefficients are convoluted into the system and periodically updated in case the system response has been modified in some way. Clearly, a training sequence is required here in order to build up the impulse response of the system; the longer the training sequence is, the better the representation of the system response becomes. A major drawback of the ZFE is that it ignores the presence of additive noise, and its use may result in a significant noise enhancement. This can be understood by noting that the equaliser filter response H ( f ) = Y (1f ) ; so the equaliser introduces large gains where Y ( f ) is small, i.e., boosting noise in the process. Relaxing the zero ISI condition and selecting a channel equaliser characteristic such that the combined power of the residual ISI and the additive noise at the equaliser output are minimised can improve the performance of a ZFE. This is achieved by using an equaliser, which is optimised based on the minimum mean square error (MMSE) criterion [54]. A further variation on the ZFE is the adaptive equaliser, where the tap weights are adjusted with time to compensate for fluctuations in the channel response; see Figure 5.48. There are several adaptive algorithms; most notable are the least mean squares (LMS) and recursive least squares (RLS), and the others are typically variations of these algorithms. In order to find the tap weights, the adaptive algorithm requires training against a header sequence of data symbols that is known at the Rx. 5.7.2.2  The Minimum Mean Square Error Equaliser (MMSE) Based on the classical equalisation theory, the most common cost function is the mean square error (MSE) between the desired signal and the output of the equaliser [55]. An equaliser that minimises this cost function is therefore known as the MMSE as shown in Figure 5.49. This type of equaliser also relies on periodically transmitting a known training sequence in order for the algorithm to minimise the MSE.

280

Optical Wireless Communications

FIGURE 5.49  Structure of the minimum mean square error equaliser.

The output from the equaliser is given as N −1

yk =

∑w x

n k −n

n=0

(5.64)

Taking the expected E or the mean values of the MSE between the output and training sequence, dropping the k for notational clarity, and writing in terms of the correlation matrix R and the crosscorrelation vector p gives E[e 2 ( k )] = E[d 2 ( k )] + w T Rw − 2w T p



(5.65)

where  xk  T R = E[ xx ] = E  x k −1  xk −2 



 ( x k2 )  ( x k x k −1 ) ( x k x k − 2 )   2    x k x k − 1 x k − 2  = E  ( x k −1 x k ) ( x k −1 ) ( x k −1 x k − 2 )  2    ( x k − 2 x k ) ( x k − 2 x k −1 ) ( x k − 2 )

    

(5.66)

And



 dk xk  p = E [ d k x k ] = E  d k x k −1  dk xk −2 

  p0    =  p1   p2  

    

(5.67)

In general, for an N weight filter we have



 p0  p1 p=    p  N −1

     

(5.68)

281

Indoor System Performance Analysis

The optimum weight vector is then found by setting the partial derivative gradient vector to zero (i.e., ∇ = 2Rw − 2 p = 0), therefore (5.69)

w opt = R −1 p



5.7.2.3  The Decision Feedback Equaliser (DFE) The performance of an equaliser is directly related to the severity of the ISI experienced in the system. Under high levels of ISI, linear equalisers will fail due to their inability to produce nonlinear relationships between input and output. Further, if a system transfer function exhibits a deep spectral null, a linear equaliser will struggle to compensate because it will set some of the tap coefficients to be excessively high. Therefore it is necessary to introduce the nonlinear DF equaliser, which operates on the principle of estimating the influence of ISI in the current symbol based upon the previously detected symbol. The DFE consists of two filters, a feedforward filter (pre-filter) and a feedback filter (ISI estimator). The feedforward filter is generally a fractionally spaced FIR filter with adjustable tap coefficients and has a form identical to the linear ZFE [54]. The feedback filter is an FIR filter with symbol spaced taps having adjustable coefficients, its input being the set of previously detected symbols. The output of the feedback filter is subtracted from the output of the feedforward filter to form the input to the detector as follows: N2

N1

qm =

∑

cn ym − n −

i=0



∑b dˆ

(5.70)

n m−n

i =1



where cn is the coefficient value of the ith feedforward tap and ym−n is the current symbol. The estimate of the previous symbol is given by dˆm − n and the feedback filter tap coefficients are given by bn. The block diagram of a DFE is depicted in Figure 5.50. The DFE does suffer from performance degradation due to feedback errors; however, the errors do not cause catastrophic failure, and they occur at relatively higher BER, out of the usual operating area.

yn

cn–2

Z –1 cn–1

Z –1

Z –1

cn

cn+1

Z –1 cn+2

qn

Σ bn–2

bn–1

Z –1

bn

Z –1

bn+1

Z –1

bn+2

Z –1

Z –1

FIGURE 5.50  Schematic diagram of DFE with an algorithm to update tap coefficients applied at dashed boxes.

282

Optical Wireless Communications

5.8  EQUALISATION AS A CLASSIFICATION PROBLEM Consider a baseband communication system as shown in Figure 5.37. The discrete received signal is given by ζ

yi =

∑h a

n n−i

+ ni = bi + ni  

(5.71)

n=0

where bi is the noise-free channel output, ni is the AWGN, and hn is the channel taps given by the ceiling bounce model. The Lth order equaliser has L-tap with an equally spaced delay of τ for the symbol spaced equaliser. The channel output can be written in a vector form as

Yi = [ yi  yi −1 …  yi − L +1 ] T

(5.72)

where T means the transpose operation. Hence the channel has L-dimensional observation space. Depending on the output vector Yi, the equaliser attempts to classify the Rx vector into one of two classes: binary zero and binary one. The equalisation problem is hence forming a decision boundary that corresponds to the transmitted symbol. Therefore, determining the values of the transmitted symbol with the knowledge of the observation vectors is basically a classification problem [56]. The linear decision boundary may be utilised when the patterns are linearly separable. However, practical channels are not linearly separable, and hence a linear boundary region on such a channel is not optimal. In general, the optimal decision boundary is nonlinear, and the realisation of the nonlinear decision boundary can be achieved by using ANN with a nonlinear transfer function. The decision boundaries formed aim to classify the received symbols into groups that belong to the desired symbol value. The boundaries are formed using neurons, which can be thought of as being similar to those found in the human brain and adjust their size in reaction to the training sequence such as tap weights in transversal filters. The major difference between ANNs and transversal equalisers is the structure; the former are arranged into a highly parallel form that allows nonlinear mapping as each input is connected to each neuron. The latter is obviously highly linear (not considering DF) since each input is connected only to its corresponding weight.

5.9  INTRODUCTION TO ARTIFICIAL NEURAL NETWORK An artificial neural network (ANN) is a mathematical and computer model which is loosely based on biological neural networks. ANNs, with a nonlinear statistical modelling capacity, are extensively used for modelling complex relationships between inputs and outputs and for pattern classification. Although introduced in 1948 [57], the extensive studies of ANNs started only in the early eighties after important theoretical results related to ANNs were attained. Now ANNs are applied to such diverse areas as computing [58], medicine [59]–[61], finance [62], control systems [63], statistical modelling [64], and engineering [65]. One of the most common applications is as equalisers in communication systems, which operate based on forming decision boundaries based on a training scheme. This is in contrast to calculating the contribution of ISI from each received symbol as with transversal equalisers. An ANN consists of simple processing units, neurons (or cells) interconnected in a predefined manner. The neurons communicate by sending signals to each other. A neuron is limited to the function of classifying only linearly separable classes [66]. However, an ANN with many neurons can perform a complicated task like pattern classification or nonlinear mapping. In fact, an ANN with a sufficient number of neurons is a universal approximator [67].

283

Indoor System Performance Analysis

5.9.1 Neuron Each neuron in the ANN does a simple task of modifying the input(s) by some predefined mathematical rule. The neuron has N inputs {xi: i = 1, …, N}, a weight wi associating witheach input, and an output y. There may be an additional parameter w0 known as the bias, which can be thought of as a weight associated with a constant input x0 always set to 1. The functional block diagram of a neuron is shown in Figure 5.51. The intermediate output a can be calculated mathematically as a=

∑w x

(5.73)

i i

i

where i = 0, …, N if there is a bias and i = 1, …, N otherwise. The output y is a function of the activation a and is given by

y = f (a)

(5.74)



The activation or transfer function f(.) depends on the application but is generally differentiable. Some popular activation functions are described below. i. Linear function: Defined as y = ma. For m = 1, the function is known as an identity function because the output is the exact copy of the input. ii. Binary threshold function: Limits the activation to 1 or 0 depending on the net input relative to some threshold function. Considering a threshold level of θ, the output is given by



1 y= 0

if a ≥ θ if a < θ

(5.75)

iii. Sigmoid function (logistic and tanh): Commonly used activation functions for nonlinear processing and pattern classification. The output of a log-sigmoid function is a continuous function in the range of 0 to 1 defined as



y=

1 1 + e− a

(5.76)

The tan-sigmoid function is a variation of the log-sigmoid function as output ranges from −1 to +1, given by

y = tanh ( a )



FIGURE 5.51  The schematic diagram of a neuron showing inputs, a bias, weights, and an output.

(5.77)

284

Optical Wireless Communications

FIGURE 5.52  A single-layer feedforward network with two output neurons.

5.9.2 ANN Architectures The neurons can be interconnected in different and complex ways. The simplest way is to arrange neurons in a single layer. However, there are other topologies; the most common are briefly described next. i. Single-layer feedforward network: It contains only an output layer connected to an input layer (Figure 5.52). The number of neurons in each layer varies depending on the number of inputs and outputs required for a particular application, but the network is strictly feedforward, i.e., the signal flows from the input layer to the output layer only. Single-layer networks are limited in capacity because of classification capability in linearly separable classes [63]. ii. Multilayer feedforward network: In order to extract the higher order statistics of data, a higher number of neurons and layers are necessary [68]. Hence multilayer networks with the capability of forming complex decision regions are utilised in different applications. Provided there is a sufficient number of neurons, a two-layer ANN can be used as a universal approximator mapping any input-output dataset [67]. Figure 5.53 shows a fully connected two-layer network. Every multilayer network consists of (i) an input layer with no processing taking place, thus not counted as part of network layers; (ii) a hidden layer; and (iii) an output layer.

FIGURE 5.53  Fully connected feedforward multilayer network.

Indoor System Performance Analysis

285

FIGURE 5.54  Neural network with a feedback connection.

iii. Recurrent neural network (RNN): Unlike the multilayer network, here the output is feedback to the input layer as shown in Figure 5.54. The network is very useful in time-series prediction. The training is more difficult in the recurrent network because of its chaotic behaviour [69]. Elman and the Hopfield networks are common recurrent network architectures.

5.10  TRAINING NETWORK The training of an ANN is accomplished by providing the ANN with the input and desired output data sets. The ANN can adjust its free parameters (weights and bias) based on the input and the desired output to minimise the cost function  E ( n ) (the difference between the desired response d ( n ) and the actual response y ( n ) of the network). The training process depends on two factors: a training data set and a training algorithm. The training data set should be representative of the task to be learnt. A poorly selected training set may increase the learning time. The training algorithm requires a minimisation of the cost function. The convergence of cost function to the local minima instead of the global minimum has been an issue, and algorithms based on the adaptive learning rate can improve the learning rate as well as a convergence to the global maximum [70], [71]. The learning can be classified into supervised and unsupervised learning. In supervised learning, the network is provided with an input-output pair that represents the task to be learnt. The network adjusts its free parameters so that the cost function is minimised. In unsupervised learning or selforganisation, the desired response is not present, and the network is trained to respond to clusters of pattern within the input. In this paradigm, the ANN is supposed to discover statistically salient features of the input population [72]. Here, a multilayer feedforward ANN is used for adaptive channel equalisation for indoor OWC. The supervised learning is more suitable for the equalisation because the training sequence can be used to approximate the channel more accurately. A popular supervised training algorithm known as the backpropagation (BP) is described next.

286

Optical Wireless Communications

5.10.1 Backpropagation Learning (BP) The BP supervised learning algorithm is the most popular training algorithm for multilayer networks. It adjusts the weight of ANN to minimise the cost function  E ( n ) according to

E (n) = d (n) − y (n)

2

(5.78)



The BP algorithm performs a gradient descent on E ( n )  in order to reach a minimum. The weights are updated as wij ( n + 1) = wij ( n ) − η

∂E ( n ) ∂ wij ( n )

(5.79)

where wij is the weight from the hidden node i to the node j and  η is the learning rate parameter. The performance of the algorithm is sensitive to η. If η is too small, the algorithm takes a long time to converge, and if η is too large, the system may oscillate, causing instability [68]. Hence adaptive learning rates are adopted for faster convergence [70], [71]. The BP can be summarised as [73]: Step 1: Initialise the weights and thresholds to small random numbers. Step 2: Present the input vectors x(n) and desired output d(n). Step 3: Calculate the actual output y(n) from the input vector sets, and calculate E (n). Step 4: Adapt weight based on (5.79). Step 5: Go to step 2.

5.11  THE ANN-BASED ADAPTIVE EQUALISER The architecture of the ANN-based channel equaliser is depicted in Figure 5.55. The sample output yi is passed through tap delay lines (TDLs) prior to being presented to the ANN for channel equalisation. The output of the ANN is sliced using α th = 0.5 to generate a binary data sequence. In the DF structure, the decision output is fed back to the ANN. The length of both the forward and feedback TDLs depend on the channel span, which also depends on the delay spread. Since the ceiling bounce model [53] shows an exponential decay in the channel components, few TDLs are adequate for the optimum performance. The ANN needs to be trained in a supervised manner to adjust its free parameters. These are open questions: What should be the size of the network? What architecture should be chosen? How many layers should be in the network? There are not satisfactory answers to these questions, nor is there a predefined rule. The number of neurons in the input and output layers depends on the input vector and the desired outputs. However, the number of hidden layers and neurons can be varied. Increasing the hidden layers results in increased system complexity and training time; however, too few neurons may result in unsatisfactory performance. The goal of optimising

FIGURE 5.55  ANN based DFE structure.

287

Indoor System Performance Analysis

TABLE 5.2 The List of Training Algorithms for ANN [77] Training algorithm

Average training time (seconds)

Levenberg-Marquardt BFGS algorithm Variable learning rate Resilient backpropagation Scaled conjugate gradient Conjugate gradient with Powell/Beale restarts Fletcher-Powell conjugate gradient Polak-Ribiére conjugate gradient

1.88 7.00 4.2 2.64 2.83 3.07 3.90 2.46

ANN is to find an architecture which takes less training time, less storage, and less complexity without degrading performance. There are two major approaches to designing an optimal network structure: (i) build a larger network and prune the number of neurons to reduce connections, and (ii) start with a small network and a limited number of neurons and continue to grow until the desired level of performance is achieved [74]. Theoretically, there is no reason for using more than two hidden layers; practically, the ANN structure with a single hidden layer is sufficient. Using two hidden layers rarely improves the model, and it may introduce a greater risk of converging to a local minimum of the cost function [75]. If enough neurons in a hidden layer are used, then it is not necessary to use more than one hidden layer [76]. Due to the risk of converging to the local minima, a single hidden layer is utilised with a sufficient number of neurons. Theoretically, it is difficult to determine the exact number of neurons required in the hidden layer. Hence here, six neurons in the hidden layer are utilised as simulation results have shown that it could provide near optimum results for all range of delay spread for all modulation techniques, though the number of neurons can be reduced for a less dispersive channel. Another key issue in applying the ANN in the supervised learning method is the learning algorithm. Although the philosophy of all training algorithms is the same, they differ in their approach to minimising the cost function. Some training algorithms converge faster but require a larger memory space, while some guarantee convergence to the global minimum of cost function [70]. The training algorithms available in Matlab 7.4.0 are listed in Table 5.2. For the comparative study, a twolayer ANN with six neurons in the hidden layer and one neuron in the output layer is selected. The SNR is chosen as 24 dB for all simulations with a different combination of transfer functions. For each training algorithm, a total of nine simulations are carried out, and the average time required for successful training is calculated and listed in the second column of Table 5.3. The LevenbergMarquardt algorithm requires the least time to train, whereas the BFG algorithm requires the

TABLE 5.3 ANN Parameters for Equalisations Parameters

Values

ANN type Number of hidden layers Transfer functions (hidden layer) Transfer functions (output layer) Training length Training algorithm

Feedforward BP MLP 1 Log-sigmoid Linear 1000 Scaled conjugate gradient

288

Optical Wireless Communications

longest time. However, it is to be noted that the memory requirements for the Levenberg-Marquardt algorithm are larger than for any other algorithm. Another key factor in selecting the training algorithm is the performance evaluation under different conditions. The BER performance for a range of SNRs and the channel delay spreads are simulated for training algorithms requiring a training time of 0.5))=1;

Indoor System Performance Analysis

289

function [traing_set] =input_ANN_linear(training_input,ff_TDL,sig_length) % function to generate the ANN input start =1; finish=ff_TDL+1; training_set=zeros(ff_TDL+1,sig_length); for i=1:sig_length training_set(:,i)=training_input(start:finish)’; start=start+1; finish=finish+1; end end

Program 5.10:  Matlab codes for the ANN-based DFE for OOK-NRZ. %% ANN parameters ff_TDL=3; % feed forward length fb_TDL=2; % feedback length ff_zeros=zeros(1,ff_TDL); fb_zeros=zeros(1,fb_TDL); nneu=6; % number of neurons in hidden layer; %% Trainign ANNN tlen=1500; %training length OOK=randint(1, tlen); % random signal generation Rx_signal=awgn(filter(hi,1,OOK),EbN0_db+3,’measured’); % Received signal, notice matched filter is not optimum training_input=ANN_input_dfe([ff_zeros Rx_signal],ff_TDL,fb_TDL,tlen, [fb_zeros OOK]); net = newff(minmax(training_input),[6 1],{‘logsig’,’purelin’},’traincgb’); net=init(net); net.trainParam.epochs=1000; net.trainParam.goal = 1e-30; net.trainParam.min_grad=1e-30; net.trainParam.show=NaN; [Net]=train(net,training_input,OOK); %% ANN based DFE OOK=randint(1,sig_length); % random signal generation Rx_signal=awgn(filter(hi,1,OOK),EbN0_db+3,’measured’); % Received signal, notice matched filter is not optimum Rx_signal=[ff_zeros Rx_signal]; Rx_th=fb_zeros; for j=1:sig_length ann_input=[Rx_signal(j:j+ff_TDL) Rx_th(j:j+fb_TDL-1)]’; ann_output(j)=sim(Net,ann_input); if ann_output(j) >0.5, Rx_th=[Rx_th 1]; else Rx_th=[Rx_th 0]; end end % bit error calculation nerr=biterr(OOK,Rx_th(fb_TDL+1:end)); function [training_set] = ANN_input_dfe(Rx,ff_TDL,fb_TDL,sig_len, Rx_th) % function to generate input for DFE

290

Optical Wireless Communications

NOPR ( d B)

6

Unequalized Linear DF

4

2

0 0.01

0.1

DT

1

2

FIGURE 5.56  NOPR versus the normalised delay spread for unequalised and ANN equalised (linear and decision feedback) OOK schemes. training_set=[]; for i=1:sig_len training_set( :,i)=[Rx(i:i+ff_TDL) Rx_th(i:i+fb_TDL-1)]’; end end

The NOPR of OOK-NRZ for the ANN-based linear and DF equalisers for a range of DT are shown in Figure 5.56. Also shown is the NOPR for the unequalised case for an MF-based Rx. The DF scheme offers the lowest NOPR for the range of DT closely followed by the linear case. Unlike the unequalised system, where the irreducible NOPR can be observed at DT > 0.5, the equalised cases do not show such a case even for DT = 2 for both linear and DF equalisers. The linear and DF schemes show identical performance for DT < 1, varying only at DT > 1. At lower DT values, typically 0.2. The NOPR against DT for the 4, 8, and 16-PPM(SDD) with an ANN-based adaptive linear equaliser is depicted in Figure 5.58. As in other equalised cases, the equalised system illustrates significantly improved performance over unequalised cases and shows no infinite NOPR for the DT of 2. Since the soft decision decoding provides natural immunity to ISI even without an equaliser, the reduction in NOPR using the equaliser is higher for the hard decision decoding. It should be noted that the higher 8

NOPR (dB)

6

L=4 L=8 L = 16

3 0

Unequalized Linear DF

-3 -6 0.1

DT

1

2

FIGURE 5.58  NOPR against the normalised delay spread DT for unequalised and ANN-based linear equalisers for PPM(SDD) scheme.

292

Optical Wireless Communications 5 L=4 L=8 L = 16

NOPR (dB)

3

0

Unequalized Linear DF

-3 -5 0.01

0.1

DT

1

2

FIGURE 5.59  NOPR versus the normalised RMS delay spread for unequalised and ANN equalisers for the DPIM scheme.

gain in the equalised hard decision should not be interpreted as the ineffectiveness of the equaliser for the soft decoding. Rather the soft decoding without an equaliser also provides some resistance to ISI, meaning reduced gain with an equaliser. For the equalised systems, the NOPRs are ∼ −0.42 dB and ∼ −1.51 dB for the hard and soft decision decoding respectively. This clearly indicates that the soft decoding offers reduced NOPR for both the equalised and unequalised systems. The simulation confirmed that DFE with soft decoding does not provide any improvement in performance compared to the hard decoding due to the hard decision in the feedback loop, and hence it is not reported here. Note that unlike the unequalised cases, where the higher-order PPM system shows a sharp rise in the OPP, the equalised system show almost a similar profile for all M. The OPP is the highest for 16-PPM with both soft and hard decoding cases due to the shortest slot durations. However, equalised 4-PPM and 8-PPM illustrate identical OPPs. Two factors are involved here: (a) the slot duration and (b) the probability of two consecutive pulses. Since 8-PPM has a shorter pulse duration, the OPP due to this factor should be higher than that for 4-PPM. On the other hand, due to longer symbol length, the probability of two consecutive pulses is significantly lower for 8-PPM than for 4-PPM, hence the OPPs related to two consecutive pulses are higher for 4-PPM (avoiding two consecutive pulses can provide significant performance improvement, as in the case of DPIM [11]). The NOPR to achieve an SER of 10 −6 against the DT for equalised 4, 8, and 16-DPIM with ANNbased equalisers is outlined in Figure 5.59. Similar to other equalised systems, an ANN-based equaliser shows a significant reduction in NOPR compared to the unequalised cases for DT > 0.1. A linear equaliser offers ∼6 dB reduction in NOPR for 4-DPIM compared to the unequalised case at DT = 0.3. Notice that it is not possible to completely remove the ISI, and hence a nominal OPP occurs for the equalised case if DT < 0.1 in comparison to the LOS channel. However, higher OPP is visible at DT > 0.1, and the OPP is ∼6 dB at DT = 2 for all cases. As in the previous cases, the DF structure provides improved performance compared to the linear structure for a highly dispersive channel and a difference of ∼0.6 dB is observed at DT of 1.5.

5.11.1 Comparative Study of the ANN- and FIR-Based Equalisers The comparative studies of the linear and DF-based equalisers based on the ANN and the traditional FIR filters are carried out in this section. To evaluate the performance of the traditional and the ANN-based equalisers, the MSE is calculated between the equaliser outputs and the desired outputs. MSEs are calculated in identical channel conditions and hence can be used to measure for the effectiveness of the equaliser. To calculate the MSE, 1000 random bits are transmitted through

293

Indoor System Performance Analysis 10

10

BER

10

10

10

10

0

-1

-2

-3

Traditional (TL=800) Traditional (TL=1000) Traditional (TL=2000) ANN (TL=200) ANN (TL=1000)

-4

-5

0

5

10 15 Electrical SNR (dB)

20

25

FIGURE 5.60  BER against the electrical SNR for the traditional and ANN linear equalisers for the OOK scheme at a data rate of 200 Mbps for a channel with DT of 2 with different training lengths (TLs).

BER

a diffuse channel with DT of 2 at Rb of 200 Mbps using the OOK modulation scheme. For simplicity as well as for more comprehensive comparisons, the channel is assumed to be noise free, and hence the error in the equaliser outputs is solely due to dispersion in the channel. The noise-free received sequence is used for training of the ANN and the traditional equaliser. The outputs of the equalisers are compared with the desired output to calculate MSEs. The resulting MSEs for both equalisers are given in Figure 5.60. The MSE for the traditional linear equaliser is in the range of 10 −4 to 10 −1, while for the ANN equaliser it is in the range 10 −8 to 10 −4. This indicates the effectiveness of the ANN as an equaliser compared to the traditional equaliser. The bit error probability of the ANN and the traditional linear equalisers for the OOK modulation scheme in a dispersive channel with DT of 2 at Rb of 200 Mbps is given in Figure 5.61. This figure 10

0

10

-1

10

-2

10

-3

10

-4

10

-5

Traditional (TL=1000) Traditional (TL=2000) Traditional (TL=8000) ANN (TL = 500) ANN (TL = 2000) 0

5

10 15 Electrical SNR (dB)

20

25

FIGURE 5.61  BER against the electrical SNR for the traditional and ANN DF equalisers for the OOK modulation scheme at a data rate of 200 Mbps for a channel with DT of 2 with different training lengths.

294

Optical Wireless Communications

MSE

10

10

10

0

-2

-4

0

200

400

Samples

600

800

1000

800

1000

(a)

-5

MSE

10

10

-10

0

200

400 Samples

600

(b)

FIGURE 5.62  The MSE between the actual and target outputs from the linear equalisers for: (a) FIR filter equaliser and (b) ANN equaliser.

reveals that the traditional linear equaliser can match the ANN-based equaliser even in the highly dispersive channel, but the advantage of the ANN is in terms of reduced training length (TL). The simulation results show that the number of training symbols required for the traditional equaliser is significantly higher than that of the ANN equaliser, especially in the highly dispersive channel. BER performance indicates that the ANN trained using 200 bits offers almost identical performance to that of the traditional equaliser trained using 1000 bits for higher SNR values. Lower training length means reduced training time, less complexity, and improved throughput. The bit error probability of the ANN and the traditional DFE for the OOK modulation scheme in a dispersive channel with DT of 2 at Rb of 200 Mbps is depicted in Figure 5.62. As with the linear equaliser, the DF ANN equaliser requires significantly lower training time compared to the traditional equaliser. It is found that the traditional DFE is difficult to train in a highly dispersive channel, and the LMS does not converge for almost the entire range of step sizes. Hence, a normalised LMS [79] is used to train the traditional equaliser. On the other hand, all training algorithms provide nearly similar performance for the ANN-based Rx, with a training length of 500 matching the performance of the traditional equaliser with a training length of 2000. This also simplifies the ANN structure and parameter optimisation compared to the traditional equalisers.

5.11.2  Diversity Techniques Due to the problems of high ambient noise and multipath dispersion inherent in diffuse IR systems and the limitation of traditional techniques, alternative methods of mitigating these effects are being investigated. These are angle diversity [80]–[83], multi-spot systems [84], [85], sectored Rxs [85]–[87], code combing [88], [89], new modulation schemes [90], spatial diversity imaging

Indoor System Performance Analysis

295

receiver [91], self-orienting receiver [92], beam angle and power adaptation with diversity Rxs [93], to name a few. Conventional angle diversity Rxs use a number of receiving elements, which are oriented in different directions. Each element is composed of an optical filter and a non-imaging concentrator such as a hemispheric lens or a compound parabolic concentrator. Imaging Rx can be used to simultaneously reduce both noise and multipath dispersion [94], [95], which offers the following advantages over non-imaging Rxs: (i) using a single imaging concentrator to cover photodetectors, resulting in reduced size and cost compared to angle diversity Rxs; (ii) utilizing a single planar array for all photodetectors, thus enabling the use of a large numbers of pixels and hence minimising the photodetector surface area; (iii) maximizing the SNR by adopting selection and combining schemes at the pixel level. In a sectored Rx, which is a hemisphere, a set of parallels and meridians defines the photodetector boundaries. A self-oriented receiver, which is composed of a single-element photodetector employing an optical front-end based on a lens, is used to point in the direction with the highest SNR received. The option is a modified version of a select-best angle diversity Rx, which can operate with a very narrow FOV. Note that the advantages offered by angle diversity–based Rx depend on how signals received at different elements are detected and processed. In OWC systems that use angle diversity–based Rx, the link performance depends on how the signals are received by the different elements, detected, and processed. For non-LOS links with multipath-induced distortion, there are a number of options for reception schemes, including (i) optimum maximum-likelihood combining (MLC)—also known as also known as matched-filter combining—which is too complex for many applications; (ii) maximal-ratio combining (MRC); (iii) selection best; and (iv) equal gain combining. Note that for the non-LOS link with negligible multipath-induced dispersion, the optimum MLC reduces to MRC. Most of these schemes seek to eliminate noise or ISI by limiting the field of view of the optical Rx, cutting out the noise, and limiting the available paths to the Rx, thus reducing the ISI and reducing the transmission data rates. Such techniques could potentially lead to severe shadowing if only one Tx–Rx pair is used in such systems. To overcome this, it is common to employ several transmitting beams or Rx elements. These schemes undoubtedly have the potential to offer performance improvements, but at the expense of system complexity, more difficult deployment, and additional hardware cost. However, there are alternative mitigation techniques in the electrical domain that are attractive for OWC diffuse links. These alternative techniques could be embodied in many diversity schemes to improve system performance further.

REFERENCES

1. J. M. Kahn and J. R. Barry, “Wireless Infrared Communications,” Proc. IEEE, vol. 85, no. 2, pp. 265–298, 1997. 2. A. J. C. Moreira, R. T. Valadas, and A. M. de O. Duarte, “Optical Interference Produced by Artificial Light,” Wirel. Networks, vol. 3, no. 2, pp. 131–140, 1997. 3. R. Narasimhan, M. D. Audeh, and J. M. Kahn, “Effect of Electronic-Ballast Fluorescent Lighting on Wireless Infrared Links,” IEE Proc. - Optoelectron., vol. 143, no. 6, pp. 347–354, 1996. 4. F. R. Gfeller and U. Bapst, “Wireless in-House Data Communication via Diffuse Infrared Radiation,” Proc. IEEE, vol. 67, no. 11, pp. 1474–1486, 1979. 5. Z. Ghassemlooy, A. R. Hayes, and B. Wilson, “Reducing the Effects of Intersymbol Interference in Diffuse DPIM Optical Wireless Communications,” IEE Proceedings-Optoelectronics, vol. 150, no. 6, pp. 445–452, 2003. 6. A. J. C. Moreira, R. T. Valadas, and A. M. de O. Duarte, “Performance of Infrared Transmission Systems under Ambient Light Interference,” IEE Proc. -Optoelectronics, vol. 143, no. 6, pp. 339–346, 1996. 7. S. Lee, “Reducing the Effects of Ambient Noise Light in an Indoor Optical Wireless System Using Polarizers,” Microw. Opt. Technol. Lett., vol. 40, no. 3, pp. 228–230, 2004. 8. A. R. Hayes, Z. Ghassemlooy, N. L. Seed, and R. McLaughlin, “Baseline-Wander Effects on Systems Employing Digital Pulse-Interval Modulation,” IEE Proc. - Optoelectron., vol. 147, no. 4, pp. 295–300, 2000.

296

Optical Wireless Communications

9. Z. Ghassemlooy, “Investigation of the Baseline Wander Effect on Indoor Optical Wireless System Employing Digital Pulse Interval Modulation,” IET Commun., vol. 2, no. 1, pp. 53–60, 2008. 10. U. Sethakaset and T. A. Gulliver, “Differential Amplitude Pulse-Position Modulation for Indoor Wireless Optical Communications,” EURASIP J. Appl. Signal Processing, vol. 2005, no. 1, pp. 3–11, 2005. 11. A. R. Hayes, “Digital Pulse Interval Modulation for Indoor Optical Wireless Communication Systems,” Sheffield Hallam University, UK, 2002. 12. K. Samaras, A. M. Street, D. C. O’Brien, and D. J. Edwards, “Error Rate Evaluation Of Wireless Infrared Links,” in Proceedings of IEEE International Conference on Communications, 1998, vol. 2, pp. 826–831. 13. A. M. Street, K. Samaras, D. C. Obrien, and D. J. Edwards, “Closed Form Expressions for Baseline Wander Effects in Wireless IR Applications,” Electron. Lett., vol. 33, no. 12, pp. 1060–1062, 1997. 14. S. Khazraei, M. R. Pakravan, and A. Aminzadeh-Gohari, “Analysis of Power Control for Indoor Optical Wireless Code-Division Multiple Access Networks Using On-Off Keying and Binary Pulse Position Modulation,” IET Commun., vol. 4, no. 16, pp. 1919–1933, 2010. 15. K. Samaras, D. C. O’Brien, A. M. Street, and D. J. Edwards, “BER Performance of NRZ-OOK and Manchester Modulation in Indoor Wireless Infrared Links,” Int. J. Wirel. Inf. Networks, vol. 5, no. 3, pp. 219–233, 1998. 16. R. Polikar, “The Engineer’s Ultimate Guide to Wavelet Analysis: The Wavelet Tutorial,” 2005. [Online]. Available: http://users.rowan.edu/∼polikar/WAVELETS/WTtutorial.html. 17. S. G. Mallat, “A Theory for Multiresolution Signal Decomposition: The Wavelet Representation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 11, no. 7, pp. 674–693, 1989. 18. B. B. Hubbard, The World According to Wavelets: The Story of a Mathematical Technique in the Making. Wellesley, Massachusetts: A K Peters, 1996. 19. A. Grinsted, J. C. Moore, and S. Jevrejeva, “Application of the Cross Wavelet Transform and Wavelet Coherence to Geophysical Time Series,” Nonlinear Process. Geophys., vol. 11, pp. 561–566, 2004. 20. A. Jamin and P. Mahonen, “Wavelet Packet Modulation for Wireless Communications,” Wirel. Commun. Mob. Comput., vol. 5, no. 2, pp. 1–18, 2005. 21. A. R. Lindsey, “Wavelet Packet Modulation for Orthogonally Multiplexed Communication,” IEEE Trans. Signal Process., vol. 45, no. 5, pp. 1336–1339, 1997. 22. M. Sablatash, J. H. Lodge, and C. J. Zarowski, “Theory and Design of Communication Systems Based on Scaling Functions, Wavelets, Wavelet Packets and Filter Banks,” in The 8th Intl. Conf. on Wireless Communication, 1996, vol. 2, pp. 640–659. 23. F. Dovis, M. Mondin, F. Daneshgaran, and F. Daneshgaran, “The MODIFIEd Gaussian: A Novel Wavelet with Low Sidelobes with Applications to Digital Communications,” IEEE Commun. Lett., vol. 2, no. 8, p. 208, 1998. 24. B. G. Negash and H. Nikookar, “Wavelet-Based Multicarrier Transmission over Multipath Wireless Channels,” IEE Electron. Lett., vol. 36, no. 21, pp. 1787–1788, 2000. 25. M. M. Akho-Zahieh and O. C. Ugweje, “Wavelet Packet Based MC/MCD-CDMA Communication System,” Electron. Lett., vol. 42, no. 11, pp. 644–645, 2006. 26. H. Zhang, H. H. Fan, and A. Lindsey, “Wavelet Packet Waveforms for Multicarrier CDMA Communications,” in IEEE International Conference on Acoustics, Speech, and Signal Processing, 2002, vol. 3, p. III-2557–2560. 27. H. Zhang, H. H. Fan, and A. Lindsey, “A Wavelet Packet Based Model for Time-Varying Wireless Communication Channels,” in IEEE Third Workshop on SPAWC, 2001, pp. 50–53. 28. M. Martone, “Wavelet-Based Separating Kernels for Array Processing of Cellular ds/cdma Signals in Fast Fading,” IEEE Trans. Commun., vol. 48, no. 6, pp. 979–995, 2000. 29. X. Fernando, S. Krishnan, H. Sun, and K. Kazemi-Moud, “Adaptive DENOISIng at Infrared Wireless Receivers,” Int. Soc. Opt. Eng., vol. 5074, pp. 199–207, 2003. 30. S. Gracias and V. U. Reddy, “Wavelet Packet Based Channel Equalization,” Sadhana, vol. 21, no. 1, pp. 75–89, 1996. 31. S. Mallat, A Wavelet Tour of Signal Processing, 2nd ed. San Diego: Academic Press, 1999. 32. C. S. Burrus, R. A. Gopinath, and H. Guo, Introduction to Wavelets and Wavelet Transforms: A Primer. New Jersey: Prentice Hall, 1998. 33. J. P. Gazeau, Wavelet Analysis in Signal and Image Processing, Northumbria University, 2007. 34. F. Liu, X. Cheng, J. Xu, and X. Wang, “Wavelet Based Adaptive Equalization Algorithm,” in Global Telecommunications Conference 1997, 1997, vol. 3, pp. 1230–1234. 35. A. K. Pradhan, S. K. Meher, and A. Routray, “Communication Channel Equalization Using Wavelet Network,” Digit. Signal Process., vol. 16, no. 4, pp. 445–452, 2006.

Indoor System Performance Analysis

297

36. N. G. Prelcic, F. P. Gonzalez, and M. E. D. Jimenez, “Wavelet Packet-Based Subband Adaptive Equalization,” Signal Processing, vol. 81, no. 22, pp. 1641–1662, 2001. 37. Y. Chen, C. Yang, Q. Yang, W. Liu, C. Li, and D. Zhang, “Wavelet Transform-OFDM in Indoor Visible Light Communication,” in 2015 Optical Fiber Communications Conference and Exhibition (OFC), 2015, pp. 1–3. 38. W. Huang, C. Gong, and Z. Xu, “System and Waveform Design for Wavelet Packet Division MultiplexingBased Visible Light Communications,” J. Light. Technol., pp. 1–1, 2015. 39. C. M. Leavey, M. N. James, J. Summerscales, and R. Sutton, “An Introduction to Wavelet Transforms: A Tutorial Approach,” Insight, vol. 45, no. 5, pp. 344–353, 2003. 40. M. Misiti, Y. Misiti, G. Oppenheim, and J.-M. Poggi, “Wavelet ToolboxTM 4 User’s Guide,” 2008. [Online]. Available: http://www.mathworks.com/access/helpdesk/help/pdf_doc/wavelet/wavelet_ug.pdf. 41. G. Strang and T. Nguyen, Wavelets and Filter Banks, 2nd ed. Wellesley College, 1996. 42. R. Polikar, “The Story of Wavelets,” in Physics and Modern Topics in Mechanical and Electrical Engineering, World Scientific and Eng. Society Press, 1999, pp. 192–197. 43. C. Valens, “A Really Friendly Guide to Wavelets,” 1991. [Online]. Available: http://pagesperso-orange. fr/polyvalens/clemens/wavelets/wavelets.html. 44. R. J. Dickenson and Z. Ghassemlooy, “BER Performance of 166 Mbit/s OOK Diffuse Indoor IR Link Employing Wavelets and neural Networks,” Electron. Lett., vol. 40, no. 12, pp. 753–755, 2004. 45. R. J. Dickenson, Wavelet Analysis and Artificial Intelligence for Diffuse Indoor Optical Wireless Communication. Northumbria University, Newcastle upon Tyne, 2007. 46. S. Rajbhandari, Z. Ghassemlooy, and M. Angelova, “Effective Denoising and Adaptive Equalization of Indoor Optical Wireless Channel with Artificial Light Using the Discrete Wavelet Transform and Artificial Neural Network,” IEEE/OSA J. Light. Technol., vol. 27, no. 20, pp. 4493–4500, 2009. 47. S. Rajbhandari, Z. Ghassemlooy, and M. Angelova, “Signal Detection and Adaptive Equalization Using a Discrete Wavelet Transform—Artificial Neural Network for OOK Indoor Optical Wireless Links,” in 16th ICEE2008, 2008. 48. J. R. Barry, J. M. Kahn, W. J. Krause, E. A. Lee, and D. G. Messerschmitt, “Simulation of Multipath Impulse Response for Indoor Wireless Optical Channels,” IEEE J. Sel. areas Commun., vol. 11, no. 3, pp. 367–379, 1993. 49. J. M. Kahn, W. J. Krause, and J. B. Carruthers, “Experimental Characterization of Non-Directed Indoor Infrared Channels,” IEEE Trans. Commun., vol. 43, no. 234, pp. 1613–1623, 1995. 50. M. D. Audeh and J. M. Kahn, “Performance Evaluation of L-Pulse-Position Modulation on NonDirected Indoor Infrared Channels,” in Communications, 1994. ICC ’94, SUPERCOMM/ICC ’94, Conference Record, “Serving Humanity Through Communications.” IEEE International Conference on, 1994, vol. 2, pp. 660–664. 51. I. Otung, Communication Engineering Principles. Palgrave Macmillan, 2001. 52. F. G. Stremler, Introduction to Communication Systems, 3rd ed. Massachusetts: Prentice Hall, 1990. 53. J. B. Carruthers and J. M. Kahn, “Modeling of Nondirected Wireless Infrared Channels,” IEEE Trans. Commun., vol. 45, no. 10, pp. 1260–1268, 1997. 54. J. G. Proakis, M. Salehi, and G. Bauch, Contemporary Communication Systems Using Matlab, 2nd ed. Boston: PWS Publishing Company, 2003. 55. A. Burr, Modulation and Coding for Wireless Communications. UK: Prentice Hall, 2001. 56. L. Hanzo, C. H. Wong, and M. S. Yee, “Neural Networked Based Equalization,” in Adaptive wireless transceivers, Wiley-IEEE Press, 2002, pp. 299–383. 57. A. K. Jain, M. Jianchang, and K. M. Mohiuddin, “Artificial Neural Networks: A Tutorial,” Computer (Long. Beach. Calif)., vol. 29, no. 3, pp. 31–44, 1996. 58. A. Nogueira, M. Rosario de Oliveira, P. Salvador, R. Valadas, and A. Pacheco, “Using neural networks to classify Internet users,” in Advanced Industrial Conference on Telecommunications/Service Assurance with Partial and Intermittent Resources, 2005, pp. 183–188. 59. A. N. Ramesh, C. Kambhampati, J. R. T. Monson, and P. J. Drew, “Artificial Intelligence in Medicine,” Ann. R. Coll. Surg. Engl., vol. 86, no. 5, p. 334–338, 2004. 60. H. B. Burke, “Evaluating Artificial Neural Networks for Medical Applications,” in International Conference on Neural Networks, 1997, vol. 4, pp. 2494–2495. 61. S.-C. B. Lo, J.-S. J. Lin, M. T. Freedman, and S. K. Mun, “Application of Artificial Neural Networks to Medical Image Pattern Recognition: Detection of Clustered Microcalcifications on Mammograms and Lung Cancer on Chest Radiographs,” J. VLSI Signal Process., vol. 18, no. 3, pp. 263–274, 1998. 62. S. Walczak, “An Empirical Analysis of Data Requirements for financial Forecasting with Neural Networks,” J. Manag. Inf. Syst., vol. 17, no. 4, pp. 203–222, 2001.

298

Optical Wireless Communications

63. Y. Fu and T. Chai, “Neural-Network-Based Nonlinear Adaptive Dynamical Decoupling Control,” IEEE Trans. Neural Networks, vol. 18, no. 3, pp. 921–925, 2007. 64. G. Dorffner, “Neural Networks for Time Series Processing,” Neural Netw. World, vol. 6, no. 4, pp. 447–468, 1996. 65. A. Patnaik, D. E. Anagnostou, R. K. Mishra, C. G. Christodoulou, J. C. Lyke, and R. K. Mishra, “Applications of Neural Networks in Wireless Communications,” IEEE Antennas Propag. Mag., vol. 46, no. 3, pp. 130–137, 2004. 66. D. MacKay, Information Theory, Inference, and Learning Algorithms, 1st ed. Cambridge, UK: Cambridge University Press, 2003. 67. K. Hornik, M. Stinchcombe, and H. White, “Multilayer Feedforward Networks Are Universal Approximators,” Neural Networks, vol. 2, no. 5, pp. 359–366, 1989. 68. S. S. Haykin and Simon, Neural Networks : A Comprehensive Foundation. Prentice Hall, 1999. 69. D. P. Mandic and J. A. Chambers, Recurrent Neural Networks for Prediction. John Wiley and Sons Ltd., 2001. 70. L. Behera, S. Kumar, and A. Patnaik, “On Adaptive Learning Rate that Guarantees Convergence in Feedforward Networks,” IEEE Trans. Neural Networks, vol. 17, no. 5, pp. 1116–1125, 2006. 71. A. Toledo, M. Pinzolas, J. J. Ibarrola, and G. Lera, “Improvement of the Neighborhood Based LevenbergMarquardt Algorithm by Local Adaptation of the Learning Coefficient,” IEEE Trans. Neural Networks, vol. 16, no. 4, pp. 988–992, 2005. 72. U. Windhorst and H. Johansson, Modern Techniques in Neuroscience Research. Springer-Verlag Berlin and Heidelberg GmbH & Co., 1999. 73. R. P. Lippmann, “An Introduction to Computing with Neural Nets,” IEEE ASSP Mag., vol. 4, no. 2, pp. 4–22, 1987. 74. S. Sitharama Iyengar, E. C. Cho, and V. V. Phoha, Foundations of Wavelet Networks and Applications, 1st ed. Chapman & Hall/CRC, 2002. 75. E. A. Martínez–Rams and V. Garcerán–Hernánde, “Assessment of a Speaker Recognition System Based on an Auditory Model and Neural Nets,” in IWINAC 2009, 2009, vol. 5602/2009, pp. 488–498. 76. S. Trenn, “Multilayer Perceptrons: Approximation Order and Necessary Number of Hidden Units,” Neural Networks, IEEE Trans., vol. 19, no. 5, pp. 836–844, 2008. 77. S. Rajbhandari, “Application of Wavelets and Artificial Neural Network for Indoor Optical Wireless Communication Systems,” Northumbria University, 2010. 78. J. G. Proakis, Digital Communications. New York: McGraw-Hill, 2004. 79. S. Haykin, Adaptive Filter Theory. New Jersey, USA: Prentice Hall International, 2001. 80. R. T. Valadas, A. M. R. Tavares, and A. M. Duarte, “Angle Diversity to Combat the Ambient Noise in Indoor Optical Wireless Communication Systems,” Int. J. Wirel. Inf. Networks, vol. 4, no. 4, pp. 275–288, 1997. 81. T. Q. Wang, C. He, and J. Armstrong, “Angular Diversity for Indoor MIMO Optical Wireless Communications,” in 2015 IEEE International Conference on Communications (ICC), 2015, pp. 5066–5071. 82. Y. A. Alqudah and M. Kavehrad, “Optimum Order of Angle Diversity with Equal-Gain Combining Receivers for Broad-Band Indoor Optical Wireless Communications,” IEEE Trans. Veh. Technol., vol. 53, no. 1, pp. 94–105, Jan. 2004. 83. R. L. Aguiar, “Considerations on the Design of Transceivers for Wireless Optical LANs,” in IEE Colloquium Optical Wireless Communications, 1999, vol. 1999, pp. 2–2. 84. A. G. Al-Ghamdi and J. M. H. Elmirghani, “Analysis of Diffuse Optical Wireless Channels Employing Spot-Diffusing Techniques, Diversity Receivers, and Combining Schemes,” IEEE Trans. Commun., vol. 52, no. 10, pp. 1622–1631, Oct. 2004. 85. P. Djahani and J. M. Kahn, “Analysis of Infrared Wireless Links Employing Multibeam Transmitters and Imaging Diversity Receivers,” IEEE Trans. Commun., vol. 48, no. 12, pp. 2077–2088, 2000. 86. C. R. A. T. Lomba, R. T. Valadas, and A. M. de Oliveira Duarte, “Sectored Receivers to Combat the Multipath Dispersion of the Indoor Optical Channel,” in Sixth IEEE International Symposium on Personal, Indoor and Mobile Radio Communication, 1995, vol. 1, pp. 321–325. 87. B. R. Mendoza, R. Pérez-Jiménez, S. Rodríguez, E. Poves, and O. González, “Considerations on the Design of Sectored Receivers for Wireless Optical Channels Using a Monte-Carlo-Based Ray-Tracing Algorithm,” IET Optoelectron., vol. 1, no. 5, pp. 226–232, Oct. 2007. 88. K. Akhavan, M. Kavehrad, and S. Jivkova, “High-speed Power-Efficient Indoor Wireless Infrared Communication Using Code Combining—Part II,” IEEE Trans. Commun., vol. 50, no. 9, pp. 1495–1502, 2002.

9781498742696_Ch05.indd 298

17/04/19 4:50 PM

Indoor System Performance Analysis

299

89. L. Kwonhyung, P. Hyuncheol, J. R. Barry, K. Lee, H. Park, and J. R. Barry, “Indoor Channel Characteristics for Visible Light Communications,” IEEE Commun. Lett., vol. 15, no. 2, pp. 217–219, 2011. 90. T. Fath, J. Klaue, and H. Haas, “Coded Spatial Modulation Applied to Optical Wireless Communications in Indoor Environments,” in 2012 IEEE Wireless Communications and Networking Conference (WCNC), 2012, pp. 1000–1004. 91. T. Chen, L. Liu, B. Tu, Z. Zheng, and W. Hu, “High-Spatial-Diversity Imaging Receiver Using Fisheye Lens for Indoor MIMO VLCs,” IEEE Photonics Technol. Lett., vol. 26, no. 22, pp. 2260–2263, 2014. 92. M. Castillo-Vazquez, A. Garcia-Zambrana, and A. Puerta-Notario, “Self-Orienting Receiver Using RateAdaptive Transmission Based on OOK Formats with Memory for Optical Wireless Communications,” in IEEE Global Telecommunications Conference, 2004. GLOBECOM ’04., vol. 2, pp. 1225–1229. 93. M. T. Alresheedi and J. M. H. Elmirghani, “10 Gb/s Indoor Optical Wireless Systems Employing Beam Delay, Power, and Angle Adaptation Methods with Imaging Detection,” J. Light. Technol., vol. 30, no. 12, pp. 1843–1856, Jun. 2012. 94. M. Castillo-Vazquez and A. Puerta-Notario, “Single-Channel Imaging Receiver for Optical Wireless Communications,” IEEE Commun. Lett., vol. 9, no. 10, pp. 897–899, Oct. 2005. 95. V. Jungnickel, T. Haustein, A. Forck, and C. von Helmolt, “155 Mbit/s Wireless Transmission with Imaging Infrared Receiver,” Electron. Lett., vol. 37, no. 5, pp. 314–315, 2001.

6

FSO Link Performance with Atmospheric Turbulence

This chapter analyses the effect of atmospheric turbulence on the FSO link for a range of modulation techniques. Atmospheric turbulence is known to cause signal fading in the channel where the FSO links may suffer temporary signal degradation or complete system annihilation. The widely adopted mathematical models for describing the fading have already been introduced in Chapter 3. There are many different types of modulation schemes which are suitable for OWC systems, as discussed in Chapter 4. The emphasis in this chapter will, however, be on the effect of atmospheric turbulence– induced fading on OOK, PPM, and phase shift keying premodulated subcarrier intensity modulation schemes. Since the average emitted optical power is always limited, the performance of modulation techniques is often compared in terms of the average received optical power required to achieve a desired bit error rate at a given data rate. It is highly desirable for modulation schemes to be power efficient, but this is not the only deciding factor in selecting a modulation technique. The design complexity of its transmitter and receiver and the bandwidth requirement of the modulation scheme are all equally important. Though it can be argued that the optical carrier signal has an abundant bandwidth, the modulation bandwidth is not abundant because it is limited by the optoelectronic devices. The classical and widely adopted modulation technique used for FSO is OOK [1], [2]. This is primarily because of the simplicity of its design and implementation. It is unsurprising therefore that most of the work reported in the literature [3], [4] is based on OOK. However, the performance of OOK-FSO links with a fixed-threshold based detection under atmospheric turbulence conditions is not the optimal, as will be shown in the following section. In atmospheric turbulence conditions, for the OOK-FSO link to display the optimum performance, the threshold level needs to be variable, changing with the prevailing irradiance fluctuation of the incoming optical radiation and the noise—that is, to be adaptive [5]. Note that adaptive threshold detection is not practically feasible because it requires adaptive optical components as well as continuous monitoring of atmospheric conditions [6]. The PPM-based FSO systems require no adaptive threshold detection scheme [7]–[12]. The subcarrier intensity modulation (SIM), which is immune to turbulence-induced amplitude fluctuation, also requires no adaptive threshold and uses much lower channel bandwidth than PPM but suffers from a high peak-to-average-power ratio (PAPR), which translates into poor power efficiency. PPM, however, requires a complex transceiver design due to the tight synchronisation requirements and a higher bandwidth than the OOK. The performance of OOK, PPM, and SIM schemes adopted for FSO links under atmospheric turbulence conditions (i.e., log-normal, gammagamma, and negative exponential channels) is further investigated in this chapter. There are other modulation techniques, such as polarisation shift keying (PoLSK)—see Chapter 4—which offer improved immunity to turbulence-induced amplitude fluctuation, that could be employed in FSO systems. Choosing a modulation scheme for a particular application, therefore, entails trade-offs among these listed factors.

6.1  ON-OFF KEYING On-off keying (OOK) can use either non-return-to-zero (NRZ) or return-to-zero (RZ) pulse format. In NRZ-OOK, an optical pulse of peak power αePT represents a digital symbol 0, while the transmission of an optical pulse of peak power PT represents a digital symbol 1. The optical source extinction ratio, αe, has the range 0 ≤ αe < 1. The finite duration of the optical pulse is the same as the symbol or bit duration Tb. With OOK-RZ, the pulse duration is lower than the bit duration, giving an 301

302

Optical Wireless Communications

improvement in power efficiency over NRZ-OOK at the expense of an increased bandwidth requirement. In all the analyses that follow, the extinction ratio, αe, is equal to zero and NRZ-OOK, which is the scheme deployed in present commercial FSO systems, is assumed unless otherwise stated.

6.1.1 OOK in a Poisson Atmospheric Optical Channel Given that the received average power Pr = Pt exp ( −α T L ), where α T represents total channel attenuation, the average received photoelectron count is given by [13] 〈n〉 =



ηλTb Pr (6.1) hc

where h and c are Planck’s constant and the speed of light in vacuum, respectively, and η is the quantum efficiency of the photodetector. However, the instantaneous count n, unlike the average count, is not constant. As mentioned in Chapter 2, it varies with time for the following reasons: i. The quantum nature of the light/photodetection process, which suggests that the instantaneous number of counts n follows the discrete Poisson distribution with an associated quantum noise of variance 〈 n 〉 (the mean and variance of a Poisson distribution are the same). ii. The received signal field varies randomly due to the effect of scintillation. This implies that the number of counts is now doubly stochastic, and based on the log-normal turbulence model of Chapter 3, the probability of n counts is derived as [14] ∞



p1 ( n ) =

∫

( ηλTPr / hc )n exp ( −ηλTb Pr / hc ) exp  −

0

1  2 2 σ l 

n! 2 πσ l2 Pr

2  Pr σ l2   ln +  dPr (6.2)  P 2   0

where P0 is the received average power in the absence of atmospheric turbulence and σl2 is the strength of power fluctuation indicator. When an optical pulse is transmitted (that is, a bit 1 is sent), a decision error occurs when the number of counts n is less than a predetermined threshold count, nth. Thus, the probability of detecting bit 0 when bit 1 is transmitted is given by [14] p1 ( n < nth )

(

)

(

)

 ∞ ηλTb ( Pr + PBg ) n exp −ηλTb ( Pr + PBg ) / hc  =  ( hc )n n! 2πσ l2 Pr n=0  0   (6.3)   1  Pr σ l2  2  ×  exp  − 2  ln +  dPr  2     2σ l  P0 nth −1

∑∫

where PBg is the power of the background radiation that falls within the receiver’s field of view and nb = ηλTPBg /hc. Similarly, the probability of detecting bit 1 when bit 0 is transmitted is derived as [15] p0 ( n > nth ) =

∞

∑

( ηλTb PBg / hc )n exp ( −ηλTPBg / hc ) n!

n = nth

nth −1

= 1−

∑ n=0

( ηλTb PBg / hc )n exp ( −ηλTb PBg / hc ) n!

(6.4)

303

FSO Link Performance with Turbulence

It should be noted in (6.4) that atmospheric turbulence has no impact when no optical power is transmitted. If bits 1 and 0 are assumed to be equally likely to be transmitted, then the systemtheoretical bit error rate Pe becomes Pe = 0.5  p0 ( n > nth ) + p1 ( n < nth )  (6.5)



For an optimal performance, nth is the value of n that satisfies expression (6.5) obtained by invoking the maximum likelihood symbol-by-symbol detection condition.

(

 ∞ ( Pr + PBg )n exp −ηλTb ( Pr + PBg ) / hc n P exp −ηλ TP / hc = ( Bg ) ( )  Bg 2πσ l2 Pr  0

∫



)

(6.6)

  1  P σ   × exp  − 2  ln r +   dPr  2    2σ l  P0  2 l

2

The impact of scintillation on the achievable bit error rate (BER) of the system is shown in Figure 6.1. The figure is obtained by combining equations (6.3), (6.4), and (6.5), while the values of nth used are obtained from the solution of equation (6.6). In the figure, the BER is plotted against the average count n0 = ηλTP0/hc. The penalty incurred due to scintillation is quite evident from the plot. For example, with respect to the no scintillation

-1

10

-2

BER

10

-3

10

-4

10

nb = 5 σ 2l = 0 σ 2l = 0.1

-5

10

σ 2l = 0.2 σ 2l = 0.5 10

20 30 40 Average Photoelectron Count , n0

50

60

FIGURE 6.1  BER against the average photoelectron count per bit for OOK-FSO in a Poisson atmospheric turbulence channel for σ l2 = [ 0,  0.1,  0.2,  0.5].

304

Optical Wireless Communications

condition, over 20 additional photoelectron counts per bit are needed to maintain the same BER of 10 −4 in a channel characterised by σl2 > 0.1. Consequently, when designing a terrestrial laser communication link, an adequate margin, based on the results shown in Figure 6.1, should be provided to account for the scintillation effect. Figure 6.1 also shows that increasing the average count can help ameliorate the effect of turbulence and result in improved BER in very weak turbulence, but as the strength of turbulence increases, the increased average photoelectron count has a less significant impact. For example, at σl2 = 0.1 and n0 = 10, the BER is about 10 −1, but this decreases to less than 10 −3 when n0 is increased to 40. In comparison, when σl2 = 0.5, the BER is higher than 10 −1 for n0 = 10 and remains higher than 10 −2 when n0 increases to 40. Note that, in the absence of turbulence, n0 = 〈 n 〉 , and the result presented in Figure 6.1 should be viewed as the theoretical performance lower bound since the photo-multiplication process has been assumed to be ideal.

6.1.2 OOK in a Gaussian Atmospheric Optical Channel With large signal photoelectron counts, and by taking the detection thermal noise into account, the generated signal current probability distribution can be approximated as the tractable Gaussian distribution. Without any loss of generality, the receiver area may be normalised to unity such that the optical power may be represented by the optical intensity, I. If R represents the responsivity of the PIN photodetector, the received signal in an OOK system is therefore given by  i ( t ) = RI 1 + 



(



∞

∑d g (t − jT ) + n (t ) (6.7) j

j =−∞

b



)

where n ( t ) ∼ N 0, σ 2 is the additive white Gaussian noise, g ( t − jTb ) is the pulse shaping function, and dj = [−1, 0]. At the receiver, the received signal is fed into a threshold detector, which compares the received signal with a predetermined threshold level. A digital symbol 1 is assumed to have been received if the received signal is above the threshold level and 0 otherwise. The probability of error, as derived in Chapter 4, is given as i  Pec = Q  th  (6.8)  σn 



(

)

where Q ( x ) = 0.5erfc x / 2 . In the presence of atmospheric turbulence, the threshold level is no longer fixed midway between the signal levels representing symbols 1 and 0. The marginal probability p ( i / 1) is then modified by averaging the conditional pdf of i(t) over the scintillation statistics. Note that scintillation does not occur when no pulse is transmitted. ∞



∫

p ( i / 1) = p ( i / 1, I ) p ( I ) dI (6.9) 0

Assuming equiprobable symbol transmission and invoking the maximum a posteriori symbolby-symbol detection, the likelihood function I , becomes [15] ∞



 − ( i − RI )2 − i 2  L = exp   p ( I ) dI (6.10) 2σ n2   0

∫

Figure 6.2 shows the plot of log (L ) against the average photocurrent i at various levels of scintillation and noise variance of 10 −2. The threshold level, as would be expected, is at the point where L = 1 (i.e., when log (L ) = 0).

305

FSO Link Performance with Turbulence 20 15 10

Log lik elihood ratio

5 Threshold line

0

σ 2 = 10 -2 E[I] = Io = 1

-5

σ 2l = 0.25 2

-10

σ 2l = 0.1 2

-15

σ 2l = 0.02 2 σ 2l = 0.5 2

-20 -25

σ 2l = 0.8 2 0

0.1

0.2

0.3 0.4 0.5 0.6 0.7 Received average photocurrent (A)

0.8

0.9

FIGURE 6.2  The likelihood ratio against the received signal for different turbulence levels and for the noise variance of 10 −2.

Based on the log-normal turbulence model, the plot of ith against the log intensity standard deviation for different noise levels is depicted in Figure 6.3. The threshold is seen to approach the expected value of 0.5 as the scintillation level approaches zero. As an illustration, at a turbulence level σl2 = 0.2, the probability of bit error Pe obtained is plotted against the normalised SNR = (RE[I])2/σn2 in Figure 6.4. The value of ith used for the adaptive threshold level graph is obtained from the solution of equation (6.10). From this figure, the effect of using a fixed threshold level in fading channels results in a BER floor. The values depend on the fixed threshold level and turbulence-induced fading strength. With an adaptive threshold, there is no such BER floor, and any desired level of BER can thus be realised. In Figure 6.5, the BER is again plotted against the normalised SNR at various levels of scintillation, including when the threshold is fixed at 0.5. This is intended to show the effect of turbulence strength on the amount of SNR required to maintain a given error performance level. With a fixed threshold, the BER reaches a floor at a BER that is greater than 10 −4, meaning that a lower BER is not achievable at the specified low scintillation level. From this graph, we can draw two inferences about atmospheric turbulence: (i) It causes a SNR penalty, for example, ∼26 dB of SNR is needed to achieve a BER of 10 −6 due to a very weak scintillation of strength σl2 = 0.252. This, however, increases by over 20 dB as the scintillation strength increases to σl2 = 0.72. (ii) It implies that an adaptive threshold will be required to avoid a BER floor in the system performance. These results illustrate that for the OOK-modulated FSO system to perform at its best, the receiver will require knowledge of both the fading strength and the noise level. This can be resolved by integrating into the system an intensity estimation network, which can predict the scintillation level based on past events. The implementation of this is not trivial, and therefore commercial FSO designers tend to adopt the fixed threshold approach and include a sufficiently large link margin in the link budget to cater to turbulence-induced fading [1].

306

Optical Wireless Communications 0.5

σ 2 = 5*10-3 σ 2 = 10-2

0.45

σ 2 = 3*10-2 σ 2 = 5*10-2

Threshold level, ith (A)

0.4 0.35 0.3 0.25 0.2 0.15 0.1

0

0.1

0.2

0.3 0.4 0.5 0.6 0.7 0.8 Log intensity standard deviation, σl

0.9

1

FIGURE 6.3  OOK threshold level against the log intensity standard deviation for various noise levels.

-1

10

-2

10

-3

BER

10

-4

10

-5

10

-6

10

-7

10

-8

10

2

σ l = 0.2 Threshold Level 0.2 0.5 Adaptive 10

15

20 25 Normalised SNR (dB)

30

35

FIGURE 6.4  BER of OOK-based FSO in atmospheric turbulence with σl2 = 0.2 considering fixed and adaptive threshold levels.

307

FSO Link Performance with Turbulence

-1

10

Fixed threshold

-2

10

-3

Adaptive threshold

BER

10

-4

10

E[I] = I 0 = 1 σ 2l = 0.252

-5

10

σ 2l = 0.5 2 σ 2l = 0.7 2

-6

10

σ 2l = 0.252 5

10

15

20 25 30 Normalised SNR (dB)

35

40

FIGURE 6.5  BER of OOK-FSO with a fixed and adaptive threshold at various levels of scintillation, σ l = [ 0.2, 0.5, 0.7 ] and I0 = 1.

6.2  PULSE POSITION MODULATION This is an orthogonal modulation technique and a member of the pulse modulation family. The PPM modulation technique improves on the power efficiency of OOK but at the expense of an increased bandwidth requirement and greater complexity. In PPM, each block of log2M data bits is mapped to one of M possible symbols. Generally, the notation M-PPM is used to indicate the order. Each symbol consists of a pulse of constant power Pt, occupying one slot, along with M-1 empty slots. The position of the pulse corresponds to the decimal value of the log2M data bits. Hence, the information is encoded by the position of the pulse within the symbol. The slot duration, Ts_ppm, is related to the bit duration by the following expression:

Ts _ ppm =

Tb log2 M (6.11) M

The transmitted waveforms for 16-PPM and OOK are shown in Figure 6.6. A PPM receiver will require both slot and symbol synchronisation in order to demodulate the information encoded on the pulse position. Nevertheless, because of its superior power efficiency, PPM is an attractive modulation technique for optical wireless communication systems, particularly in deep space laser communication applications [16]. Assuming that complete synchronisation is maintained between the transmitter and receiver at all times, the optical receiver detects the transmitted signal by attempting to determine the energy in each possible time slot. It then selects the signal, which corresponds to the maximal energy. In direct photodetection, this is equivalent to ‘counting’ the number of released electrons in each Ts interval. The photo count per PPM slot can be obtained from

Ks =

ηλPr Ts _ ppm (6.12) hc

308

Optical Wireless Communications

FIGURE 6.6  Time waveforms for 4-bit OOK and 16-PPM.

where Pr is the received optical power during a slot duration. An APD could be used to give an increase in the number of photon counts per PPM slot but, unfortunately, the photo-multiplication process that governs the generation of the secondary electrons is a random process. This implies that a large photo-multiplication gain will eventually lead to a large noise factor and an error-prone performance. For a moderately high received signal, as is the case in commercial and short-range FSO systems, the BER conditioned on Ks is given by [17]  Pec = Q   



(Gq )

  (6.13) 2 Gq F ( K s + 2K Bg ) + 2σ th2 

( )

2

K s2

where the parameters are defined as K Bg = ηλPBg Ts / hc Average photon count per PPM slot due to the background radiation of power PBg G Average APD gain q Electronic charge Noise factor of the APD F ≈ 2 + ζG APD ionisation factor ζ 2 σ Th = ( 2κTe q / RL ) ( Ts _ ppm ) Equivalent thermal noise count within a PPM slot duration [17] In the presence of log-normal atmospheric turbulence, the unconditional BER for a binary PPM–modulated FSO obtained by averaging (6.13) over the scintillation statistics can be approximated as [10] Pe ≈



1 π

((

))

 exp 2 2σ k xi + mk   (6.14) wi Q   Fexp 2σ k xi + mk + K n  i =1 n

∑

(

)

where [ wi ]i =1 and [ xi ]i =1 are the weight factors and the zeros of an nth-order Hermite polynomial. These values are contained in Appendix B for a 20th-order Hermite polynomial. mk represents the n

n

309

FSO Link Performance with Turbulence

(

)

(

)

2 mean of ln ( K s ), K n = 2σ Th / ( gq ) + 2 FK Bg, and σ 2k = ln σ 2N + 1 . It is noteworthy that the fluctuation 2

of the mean count, Ks, is brought about by the atmospheric turbulence, and its ensemble average is given by the following [10]:  σ2  E [ K s ] = exp  k + mk  (6.15)  2 



For an M-PPM system, the BER denoted by PeM has an upper bound given by [10] M P ≤ 2 π M e



((

))

 exp 2 2σ k xi + mk   (6.16) wi Q  m K  F exp 2 σ x + +  k i k n i =1  n

∑

(

)

The BER performance of a binary PPM–modulated FSO, using the Matlab code program 6.1, is shown in Figure 6.7 at different levels of scintillation. The extension of the result is straightforward from (6.16) and hence is not presented here. As expected, an increase in the atmospheric scintillation results in an increase in the required signal level to achieve a given BER. Increasing the signal strength can be used to minimise the scintillation effect at a low scintillation index, but as turbulence strength increases, it is observed that the BERs all tend towards a high BER asymptotic value.

-2

10

-3

BER

10

-4

10

-5

10

E[K s ] 140 180 220 260 300

-6

10

-7

10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Scintillation index, σ 2N

FIGURE 6.7  Binary PPM BER as a function of scintillation index for K Bg = 10; Te = 300 K, ζ = 0.028, Rb = 155 Mbps, and g = 150.

310

Optical Wireless Communications

Program 6.1:  Matlab codes to calculate the BER of the binary PPM scheme. %Evaluation of BER of BPPM FSO (using equation (6.16) ) under weak turbulence using the Gauss-Hermite Quadrature integration approach. clear clc %****Hermite polynomial weights and roots************ w20=[2.22939364554e-13,4.39934099226e-10,1.08606937077e-7,7.8025564785e6,0.000228338636017,0.00324377334224,0.0248105208875,0.10901720602, 0.286675505363,0.462243669601,.. 0.462243669601,0.286675505363,0.10901720602,0.0248105208875, 0.00324377334224,0.000228338636017,7.8025564785e-6,1.08606937077e-7, 4.39934099226e-10,2.22939364554e-13]; x20=[-5.38748089001,-4.60368244955,-3.94476404012,-3.34785456738, -2.78880605843,-2.25497400209,-1.73853771212,-1.2340762154,0.737473728545,-0.245340708301,... 0.245340708301,0.737473728545,1.2340762154,1.73853771212, 2.25497400209,2.78880605843,3.34785456738,3.94476404012, 4.60368244955,5.38748089001]; %**************************************************** Ks1 = [140,180,220,260,300]; for i1 = 1:length(Ks1) Ks = Ks1(i1); %*************Simulation Parameters*************************** Rb = 155e6; %Bit rate RL = 50; %Load resistance Temp = 300; %Ambient temperature E_c = 1.602e-19; B_c = 1.38e-23;

%Electronic charge %Boltzmann constant

%************************************************************************ NoTh = (2*B_c*Temp/(2*Rb*RL)); Ioni = 0.028; Kb = 10; gain = 150; F = 2 + (gain*Ioni); Kn = ((2*NoTh)/(gain*E_c)^2) + (2*F*Kb); S_I = 0.1:0.15:0.9; %Scintillation Index for i = 1:length(S_I) SI = S_I(i); Sk = log(S_I(i) + 1); Mk = log(Ks)-(Sk/2); Temp = 0; for j = i:length(x20) ANum(j) = (2*x20(j)*sqrt(2*Sk)) + (2*Mk); Num(j) = exp(ANum(j)); BDen(j) = (sqrt(2*Sk)*x20(j) + Mk); Den(j) = (F*exp(BDen(j))) + Kn; Prod(j) = w20(j)*Q(Num(j)/Den(j)); Temp = Temp + Prod(j); end

FSO Link Performance with Turbulence

311

BER(i1,i) = Temp/sqrt(pi); end end %*********Plot function**************************** figure semilogy(S_I,BER)

6.3  SUBCARRIER INTENSITY MODULATION Subcarrier intensity modulation (SIM) is a technique borrowed from the very successful multiple carrier radio frequency (RF) communications already deployed in applications such as digital television, local area networks (LANs), asymmetric digital subscriber line (ADSL), 4G communication systems, and optical fibre communications [18]. In optical fibre communication networks, for example, the subcarrier modulation techniques have been commercially adopted in transmitting cable television signals and have also been used in conjunction with wavelength division multiplexing [19]. For the seamless integration of FSO systems into the present and future networks, which already harbour subcarrier modulated (or multiple carriers) signals, the study of subcarrier modulated FSO is thus imperative. Other reasons for studying the subcarrier intensity modulated FSO system includes: i. It benefits from already developed and evolved RF communication components such as stable oscillators and narrow filters [20]. ii. It avoids the need for an adaptive threshold required by optimum performing OOK modulated FSO [21]. iii. It can be used to increase capacity by accommodating data from different users on different subcarriers. iv. It has comparatively lower bandwidth requirements than PPM. There are, however, some challenges in the implementation of SIM. These are i. Relatively high average transmitted power due to: a. the optical source being ON during the transmission of both binary digits 1 and 0, unlike in OOK where the source is ON during the transmission of bit 1 only. b. the multiple subcarrier composite electrical signals being the sum of the modulated sinusoids (i.e., dealing with both negative and positive values), requiring a DC bias. This is to ensure that the composite electrical signal, which will modulate the light source (i.e., laser) irradiance, is never negative. Increasing the number of subcarriers leads to the increased average transmit power because the minimum value of the composite electrical signal decreases (becomes more negative) and the required DC bias, therefore, increases [22]. This factor results in poor power efficiency and places a bound on the number of subcarriers, which can be accommodated when using multiple SIM. iii. The possibility of signal distortions due to inherent laser nonlinearity and signal clipping due to over-modulation. iv. Stringent synchronisation requirements at the receiver side. It is, therefore, worthwhile to mention that multiple-SIM is only recommended when the quest for higher capacity/more users outweighs the highlighted challenges or where FSO is to be integrated into existing networks which already contain multiple RF carriers. Several methods have been researched and documented [22]–[24] to improve the poor power efficiency of SIM, but these will not be considered in this book.

312

Optical Wireless Communications

6.3.1 SIM Generation and Detection In optical SIM links, an RF subcarrier signal m(t), premodulated with the source data d(t), is used to modulate the intensity of the optical source—a continuous wave laser diode in outdoor FSO links. Figure 6.8 illustrates the system block diagram of a SIM-FSO system with

a1c

g(t)

X

m1c(t)

cos(ωc1t+φ1) g(t) . . . . .

. . . . . aNc

X

m1s(t)

PT m(t)

sin(ωc1t+φ1) . . .

g(t)

X

Σ

Driver circuit

Serial-to-parallel converter and encoder

d(t)

a1s

Σ

bo DC bias

mNc(t)

Atmospheric channel

cos(ωcNt+φN) aNs

g(t)

X

mNs(t)

sin(ωcNt+φN)

(a)

BPF

X

LPF

g(-t)

Sampler

g(-t)

Sampler

cos(ωc1t+φ1) OBPF

Σ

I n(t) Noise

i(t)

BPF

. . . . .

BPF

X

LPF

sin(ωc1t+φ1)

X

. . . . .

. . . . .

LPF

g(-t)

Sampler

g(-t)

Sampler

Parallel-to-serial converter and decoder

TIA

dˆ (t )

cos(ωcN t+φN)

BPF

X

LPF

sin(ωcN t+φN)

(b) FIGURE 6.8  Block diagram of SIM-FSO: (a) transmitter and (b) receiver. TIA (trans-impedance), OBPF (Optical band-pass filter).

313

FSO Link Performance with Turbulence

N subcarriers. The serial-to-parallel converter distributes the incoming data across the N subcarriers. Each subcarrier carries a reduced symbol rate, but the aggregate must be equal to the symbol rate of d(t). Another obvious possibility, not shown in the figure, is to have different users occupying N different subcarriers. Prior to modulating the laser irradiance, d(t) is first used for modulating RF subcarriers. For M-PSK subcarrier modulation shown in Figure 6.8, the encoder maps each N subcarrier symbol onto the symbol amplitude {aic , ais }i =1, which corresponds to the constellation in use. Since the subcarrier signal, m(t), is sinusoidal, having both positive and negative values, a DC level b0 is added to m(t) before it is used to directly drive the laser diode—to avoid any signal clipping. However, the DC-level shifted signal must operate within the linear region of the P-I characteristics to ensure no clipping and saturation, which will lead to harmonic and intermodulation distortions. The subcarrier signal in the N-SIM-FSO system is defined as m (t ) =



N

∑m (t ) (6.17) i

i =1

During one symbol duration, each RF subcarrier signal is generally represented by

mi ( t ) = g ( t ) aic cos ( ω ci t + ϕ i ) + g ( t ) ais sin ( ω ci t + ϕ i ) (6.18)

where g(t) is the pulse shaping function, and the subcarrier angular frequency and phase are repreN sented by [ ω ci , ϕ i ]i =1, respectively. It follows that each subcarrier can be modulated by any standard RF digital/analogue modulation technique, such as QAM, M-PSK, M-FSK, or M-ASK. Using direct detection at the receiver, the incoming optical radiation, Pr , is converted into an electrical signal, i(t). This is followed by a standard RF demodulator to recover the transmitted symbol as shown in Figure 6.8(b). By normalising the receiver area to unity and representing the received power by irradiance, I, the received signal can be modelled as

i ( t ) = RI [1 + ξm ( t )] + n ( t )

where the optical modulation index ξ =

m(t ) iB − iTh

(6.19)

, as shown in Figure 6.9.

The electrical band-pass filter (BPF) with a minimum bandwidth of 2Rb performs the following functions: selection of the individual subcarrier for demodulation; reduction of the noise power; and suppression of any slow varying RI component present in the received signal. For a subcarrier at ωci the received signal is

i ( t ) = I comp + Qcomp (6.20)

where

I comp = RI ξg ( t ) aic cos ( ω ci t + ϕ i ) + nI ( t ) (6.21a)



Qcomp = − RI ξg ( t ) ais sin ( ω ci t + ϕ i ) + nQ ( t ) (6.21b)

nI(t) and nQ(t) are the independent additive white Gaussian noise (AWGN) with a zero mean and a variance σn2. The quadrature components Icomp and Qcomp are down-converted by the reference

314

Optical Wireless Communications Irradiance I peak

ξ = |m(t)|/iB - iTh

I

iB

iTh

Drive current m(t)

Imax

FIGURE 6.9  Output characteristic of an optical source driven by a subcarrier signal showing optical modulation index.

signals cos ωct and sin ωct, respectively, and applied to the standard receiver architecture. The electrical low-pass filters, which are part of the standard RF receiver, remove any out-of-band (unwanted) signals from the down-converted signal and then pass it on to the decision circuit. In the case of a phase shift keying–modulated subcarrier, the decision circuit estimates the phase of the received signal and decides which symbol has been received. By adopting the approach in [27], the conditional BER expressions can be deduced.

6.3.2 SIM-FSO Performance in Log-Normal Atmospheric Channel Prior to the analytical estimation of the effect of scintillation on system performance, the effect of turbulence and noise on a subcarrier signal constellation will first be presented. The simulation will be based on the block diagram of Figure 6.8, with a QPSK, modulated single subcarrier. The constellation at the input of the transmitter is shown in Figure 6.10. The signal is then transmitted through the atmospheric channel with turbulence-induced fading and noise. For this illustration, the 2 electrical SNR = ( ξ RE [ I ]) / σ 2n is fixed, for instance, at 2 dB. The received constellation under a very low fading of σ l2 = 0.001 is shown in Figure 6.11, while the effect of turbulence-induced fading is made evident by displaying in Figure 6.12 the received constellation with a much higher fading with σ l2 = 0.5. With a very low turbulence-induced fading strength σ l2 = 0.001, the constellations form clusters that are clearly confined within their respective quadrants, and the chance of erroneous demodulation is very low, but as the turbulence level increases to σ l2 = 0.5, the confinement is lost and the constellation points become more staggered and move toward the centre of the plot. This apparently increases the chance of demodulation error during the symbol detection process. To quantify this observation, the bit error and the outage probabilities of SIM-FSO in atmospheric turbulence will be presented next.

315

FSO Link Performance with Turbulence

0.6

0.4

Quadrature

0.2

0

-0.2

-0.4

-0.6

-0.6

-0.4

-0.2

0 In-Phase

0.2

0.4

0.6

FIGURE 6.10  QPSK constellation of the input subcarrier signal without the noise and channel fading. 1 0.8 0.6

Quadrature

0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1

-0.8

-0.6

-0.4

-0.2

0 0.2 In-Phase

0.4

0.6

0.8

1

FIGURE 6.11  Received constellation of QPSK premodulated SIM-FSO with the noise and channel fading for SNR = 2 dB and σ l2 = 0.001.

316

Optical Wireless Communications 1 0.8 0.6 0.4

Quadrature

0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1

-0.8

-0.6

-0.4

-0.2

0 0.2 In-Phase

0.4

0.6

0.8

1

FIGURE 6.12  Received constellation of QPSK premodulated SIM-FSO with the noise and channel fading for SNR = 2 dB and σ l2 = 0.5.

The Matlab codes for Figs. 6.10 to 6.12 are given in Program 6.2. Additional Matlab codes for this section are given in Appendix B. Program 6.2:  Matlab codes used for generating Figure 6.10 to Figure 6.12. %*****Subcarrier modulation simulation for M-PSK intensity modulation******% %*************************************************************************% clear all clc %**********SIMULATION PARAMETERS******************************** N_sub = 1; %No of subcarriers symb = 1e2; %No of symbols Rb = 155e6; %Symbol rate M = 4; %M-PSK Responsivity = 1; %Photodetector responsivity T = 1/Rb; fmin = 1e9; %Starting carrrier frequency Fs = 50*fmin; %Sampling frequency t = 0:1/Fs:T;

317

FSO Link Performance with Turbulence samples_symb = length(t); fmax = 2*fmin; if N_sub > 1 delta = (fmax - fmin)/(N_sub - 1); consecutive subcarriers else delta = 0; end K = 2*pi;

%Number of samples per symbol %Single octave operation %Frequency spacing of any two

%***********GENERATION OF THE SUBCARRIER SIGNAL*********************** [Inphase,Quadrature,Datain] = Basebandmodulation(M,symb,N_sub); Input1 = Inphase + (j*Quadrature); %The input symbols constellation Ac = 1; %Subcarrier signals amplitude M0 = [];M1 = []; for i0 = 1:N_sub j0 = i0-1; XX = Ac.*cos(K.*t.*(fmin + (j0*delta))); XX1 = Ac.*sin(K.*t.*(fmin + (j0*delta))); M0 = [M0;XX]; M1 = [M1;XX1]; end for i1 = 1:N_sub I_phase =[];Q_phase = []; for j1 = 1:symb I_sig = Inphase(j1,i1).*M0(i1,:); I_phase = [I_phase,I_sig]; Q_sig = Quadrature(j1,i1).*M1(i1,:); Q_phase = [Q_phase,Q_sig]; end I_Mod_Data(i1,:) = I_phase; Q_Mod_Data(i1,:) = Q_phase; end PSK_sig = I_Mod_Data - Q_Mod_Data; SIM = sum(PSK_sig,1); %subcarrier multiplexed signal Mod_index = 1/(Ac*N_sub); %Modulation index SCM_Tx = (1 + Mod_index*SIM); %The transmitted signal %**********************THE CHANNEL PARAMETERS************************** %************Turbulence parameters***************** Io = 1; %Average received irradiance I_var = 1e-3; %Log irradiance variance I = Turbulence(I_var,symb,Io,t); %Irradiance using the Log normal turbulence model SNR_dB = 2; %SNR value in dB SNR = 10.^(SNR_dB./10); Noise_var = (Mod_index*Responsivity*Io)^2./(SNR); Noise_SD = sqrt(Noise_var); %**********************RECEIVER DESIGN************************ %************Filtering to separate the subcarriers************ SCM_Rx = (Responsivity.*I.*SCM_Tx); %SCM_Rx = (Responsivity.*SCM_Tx);

318

Optical Wireless Communications

for i0 = 1:length(SNR) Noise =[]; Noise = Noise_SD(i0).*randn(size(SCM_Tx)); if Rb 0 (6.30)

n

∑w f ( x ) (6.31) i

i

i =1

−∞

where [ wi ]i =1 and [ xi ]i =1, whose values are given in Appendix B, are the weight factors and the zeros of an nth-order Hermite polynomial, Hen ( x ) [26]. The degree of accuracy of (6.10) depends n

n

ln( I / I )+σ 2 /2

0 l on the order n of the Hermite polynomial. By invoking a change of variable, y = 2σ l in (6.29a), and combining this with (6.30) and (6.31), the unconditional BER given by (6.29a) can be reduced to the following form:





1 Pe ≅ π

π/2

∫ 0

Pe ≅

1 π

(

 K 0 exp 2 K1  2σ l xi − σ l2 / 2    wi exp  − 2 2sin ( θ )  i =1 n

∑

1 π

∑w Q ( n

i

i =1

(

))

)  dθ (6.32) 

K 0 exp K1  2σ l xi − σ l2 / 2  (6.33)

321

FSO Link Performance with Turbulence

TABLE 6.1 Values of K1 and K0 for Different Noise-Limiting Conditions Noise-limiting conditions Quantum limit K0

ξ 2 RI 0 Pm 2 qRb

K1

0.5

Thermal noise

Background noise

2

2

( ξ RI 0 )

4 kTe Rb

( ξ I 0 ) RPm 2 qRb ( I sky + I sun )

1

1

Pm RL

Thermal and background noise

( ξ RI 0 )

(σ

2 Bg

2

Pm

2 + σ Th

)

1

The values of K1 and K0 are as given in Table 6.1 for different noise-limiting conditions. In Table 6.1, Rb represents the data rate. The BER plots, using both the approximation given by (6.33) and the numerical simulation of the exact solution given by (6.29a) against the normalised 2 SNR = ( ξRE [ I ]) / σ 2n , are illustrated in Figure 6.13. These results hint that using the 20th-order Gauss-Hermite approximation gives an accurate representation of the exact BER. The GaussHermite integration solution of the BER is, however, preferred for its simplicity and compactness. In order to keep the optical source (laser) within its linear dynamic range and avoid signal clipping and saturation-induced distortion, the condition ξm ( t ) ≤ 1 must always hold. For a given value of ξ, this places an upper bound on the amplitude of each subcarrier. The BER given by (6.33) is

Numerical Gauss-Hermite

-2

10

σ 2l = 0.12

-3

10

-4

10

BER

-5

10

-6

10

-7

10

-8

10

-9

10

10

11

12

13

14

15

16

17

18

19

20

Normalised SNR (dB)

FIGURE 6.13  BER against the normalised SNR using numerical and 20th-order Gauss-Hermite integration methods in weak atmospheric turbulence for σ l2 = 0.12.

322

Optical Wireless Communications -2

10

-4

10

-6

10

-8

BER

10

-10

10

-12

10

σ 2l = 0.3 Thermal noise Background noise Thermal+Background Quantum

-14

10

-60

-55

-50

-45 -40 -35 -30 -25 Average received irradiance, I0 (dBm)

-20

-15

FIGURE 6.14  The BER against the average received irradiance in weak turbulence under different noiselimiting conditions for Rb = 155 Mbps and σl2 = 0.3.

plotted against the normalised SNR for different noise-limiting conditions in Figure 6.14, based on the simulation parameters given in Table 6.2. The figure illustrates clearly that, for an FSO link with a suitable optical BPF and a narrow FOV detector, the system performance is limited by thermal noise. Moreover, under this thermal noise–limited condition, the SIM-FSO still requires about an additional 30 dB of SNR compared with the theoretical quantum limit. Matlab codes for generating the BER performance, as shown in Figure 6.14 under different noise conditions, are given in Program 6.3. Additional Matlab codes for this section are given in Appendix B. Program 6.3:  Matlab codes for generating the BER performance under different noise conditions. %Evaluation of BER of SISO BPSK-SIM FSO under weak turbulence using the Gauss-Hermite Quadrature integration approach.; %Different noise sources considered: Background, thermal and dark clear clc %*************Simulation Parameters*************************** Rb = 155e6; %symbol rate R = 1; %Responsivity M_ind = 1; %Modulation index

323

FSO Link Performance with Turbulence

TABLE 6.2 Simulation Parameters Parameter

Value

Data rate Rb Spectral radiance of the sky N(λ) Spectral radiant emittance of the sun W(λ) Optical band-pass filter bandwidth Δλ @ λ = 850 nm PIN photodetector field of view (FOV) Radiation wavelength λ Number of subcarriers N Link range L Index of refraction structure parameter Cn2 Load resistance RL PIN photodetector responsivity R Operating temperature Te Optical modulation index ξ

155 Mbps 10−3 W/cm2µmSr 0.055 W/cm2µm 1 nm 0.6 rad 850 nm 1 1 km 0.75 × 10−14 m−2/3 50 Ω 1 300 K 1

A = 1; RL = 50; Temp = 300; wavl = 850e-9;

%Subcarrier signal amplitude %Load resistance %Ambient temperature %Optical source wavelenght

%********************Background Noise************************ %Considering 1 cm^2 receiving aperture sky_irra = 1e-3; % at 850nm wavelength, in W/cm^2-um-sr sun_irra = 550e-4; %at 850nm wavelength, in W/cm^2-um FOV = 0.6; % in radian OBP = 1e-3; %Optical filter bandwidth in micrometre Isky = sky_irra*OBP*(4/pi)*FOV^2; %Sky irradiance Isun = sun_irra*OBP; %Sun irradiance %**************Rytov Variance******************************************** Range = 1e3; %Link range in meters Cn = 0.75e-14; %Refractive index structure %parameter Rhol = 1.23*(Range^(11/6))*Cn*(2*pi/wavl)^(7/6); %Log irradiance; %variance (Must be less than 1 Varl = Rhol; %Log intensity variance r = sqrt(Varl); %log intensity standard deviation %*****************Physical constants**************************** E_c = 1.602e-19; %Electronic charge B_c = 1.38e-23; %Boltzmann constant %************************************************************************ Pd = A^2/2; K1 = ((M_ind^2)*R*Pd)/(2*E_c*Rb); %Quantum limit K2 = ((R*M_ind)^2)*Pd*RL/(4*B_c*Temp*Rb); %Thermal noise limit K3 = (Pd*R*M_ind^2)/(2*E_c*Rb*(Isun + Isky)); %Background radiation %limit

324

Optical Wireless Communications

Ktemp = (4*B_c*Temp*Rb/RL) + (2*E_c*R*Rb*(Isun + Isky)); K4 = ((R*M_ind)^2)*Pd/Ktemp; %background and thermal; %noise combined %***************Hermite polynomial weights and roots************ w20=[2.22939364554e-13,4.39934099226e-10,1.08606937077e-7, 7.8025564785e-6,0.000228338636017,0.00324377334224,0.0248105208875, 0.10901720602,0.286675505363,0.462243669601,... 0.462243669601,0.286675505363,0.10901720602,0.0248105208875, 0.00324377334224,0.000228338636017,7.8025564785e-6,1.08606937077e-7, 4.39934099226e-10,2.22939364554e-13]; x20=[-5.38748089001,-4.60368244955,-3.94476404012,-3.34785456738,2.78880605843,-2.25497400209,-1.73853771212,-1.2340762154,0.737473728545,-0.245340708301,... 0.245340708301,0.737473728545,1.2340762154,1.73853771212, 2.25497400209,2.78880605843,3.34785456738,3.94476404012, 4.60368244955,5.38748089001]; %**************************BER evaluation******************* Io = logspace(-10,-4,30); %Average received irradiance IodBm =10*log10(Io*1e3); %Average received irradiance in %dBm SNR2 = RL*((R.*Io).^2)./(4*B_c*Temp*Rb); SNR2dB = 10*log10(SNR2); for i1 = 1:length(Io) GH1 = 0;GH2 = 0;GH3 = 0;GH4 = 0; for i2 = 1:length(x20) arg1 = sqrt(K1*Io(i1))*exp(0.5*x20(i2)*sqrt(2)*r - Varl/4); temp1 = w20(i2)*Q(arg1); GH1 = GH1 + temp1; arg2 = sqrt(K2)*Io(i1)*exp(x20(i2)*sqrt(2)*r - Varl/2); temp2 = w20(i2)*Q(arg2); GH2 = GH2 + temp2; arg3 = sqrt(K3)*Io(i1)*exp(x20(i2)*sqrt(2)*r - Varl/2); temp3 = w20(i2)*Q(arg3); GH3 = GH3 + temp3; arg4 = sqrt(K4)*Io(i1)*exp(x20(i2)*sqrt(2)*r - Varl/2); temp4 = w20(i2)*Q(arg4); GH4 = GH4 + temp4; end BER1(i1) = GH1/sqrt(pi); BER2(i1) = GH2/sqrt(pi); BER3(i1) = GH3/sqrt(pi); BER4(i1) = GH4/sqrt(pi); end %*********Plot function**************************** figure subplot(4,1,1); semilogy(IodBm,BER1) %Quantum limit case subplot(4,1,2) ; semilogy(IodBm,BER2) % Thermal noise case subplot(4,1,3); semilogy(IodBm,BER3) % Background radiation case subplot(4,1,4); semilogy(IodBm,BER4) % background and thermal noise case

325

FSO Link Performance with Turbulence

6.3.3.2  M-ary PSK-Modulated Subcarrier Here the data symbols, which are comprised of log2 M binary digits, are mapped onto one of the M available phases on each subcarrier signal, m(t). Based on the subcarrier coherent demodulation and by following the analytical approach given in [27], the following conditional BER expressions are obtained:



Pec ≈

Pec =

2 Q log2 M

(

( log2 M ) γ ( I ) sin ( π / M ))  

for M − PSK,  M ≥ 4 (6.34a)

2(1 − 1 / M  3log2 Mγ ( I )  Q   for M − QAM,  log2 M  even (6.34b) log2 M 2 ( M − 1)  

The unconditional BER, Pe, is thus obtained in a similar fashion, by averaging the conditional BER over the atmospheric turbulence statistics. The resulting BER expression (6.35) for M-PSK has no closed form solution and can only be evaluated numerically. ∞



∫ (

)

2 Q Pe ≅ log2 M

γ ( I ) log2 M sin ( π / M ) p ( I ) dI (6.35)

0

Whenever a subcarrier coherent detection is used, there is always an ambiguity associated with the estimation of the absolute phase of the subcarrier signal [27]. This poses an implementation challenge for the subcarrier coherent demodulation–based systems; this can, however, be solved by considering a differential phase shift keying (DPSK)–based SIM-FSO system as discussed below. 6.3.3.3  DPSK-Modulated Subcarrier This modulation scheme is the most suitable when the absolute phase estimation needed for the subcarrier coherent demodulation is not feasible or too complex to realise. The DPSK premodulated SIM-FSO is demodulated by comparing the phase of the received signal in any signalling interval with the phase of the signal received in the preceding signalling interval [28], as shown in Figure 6.15 for a single subcarrier FSO system. Accurate demodulation of the present data symbol thus depends on whether the preceding symbol has been correctly demodulated or not. The demodulation of DPSK-based SIM-FSO is feasible during atmospheric turbulence because the turbulence coherence time, which is in the order of milliseconds, is far greater than the typical duration of two consecutive data symbols. This implies that the channel properties are fixed during a minimum of two symbol durations—a prerequisite for non-coherent demodulation of a DPSK subcarrier signal. The conditional BER of the DPSK-premodulated subcarrier is given by [27], [28] Pec = 0.5exp ( −0.5γ ( I )) (6.36)



In the presence of scintillation, the following unconditional BER, Pe, is derived using the GaussHermite quadrature integration approximation of Section 6.3.3.1 as ∞

∫

Pe = 0.5exp ( −γ ( I )) p ( I ) dI (6.37)



0



Pe ≅

1 2 π

n

∑w exp ( − K i

i =1

2

)

exp  xi 2 2σ l − σ 2l  (6.38)

326

Input data

d (t )

Optical Wireless Communications

XOR

g(t)

Delay (T)

×

m(t)

Laser driver

+

Carrier cosωct

TT

Atmospheric channel

DC bias

b0

(a) Atmospheric channel

RT

Photodetector/ TIA

i

BPF

x

LPF

g(-t) y(t)

cosωct

Symbol rate Sampler

yk Delay (T)

DPSK demodulator

yk-1

Phase comparator

ˆ d(t) Outputdata

(b) FIGURE 6.15  Block diagram of an FSO link employing DPSK-modulated SIM; (a) transmitter and (b) receiver. TIA (trans-impedance amplifier); TT (transmitter telescope); RT (receiver telescope).

where = Rξ I 0 / 2 σ n2 . In Figure 6.16, the BER of SIM-FSO based on different modulation techniques on the subcarrier are compared at a scintillation level of σl2 = 0.52. The performance in a turbulence-free channel is included in the figure for the estimation of turbulence-induced fading penalty. The figure shows clearly the performance superiority of BPSKmodulated SIM in terms of the amount of SNR required to achieve a given BER. Due to atmospheric turbulence-induced channel fading, a BPSK premodulated SIM-FSO system will incur a power penalty of ∼5 dB at a BER of 10 −6; the penalty rises to ∼10 dB when the error performance level is raised to a BER of 10 −9. This penalty is higher for other modulation techniques, as shown in the figure. The M-PSK (with M ≥ 4) modulated subcarrier is known to be more bandwidth efficient [27], while the DPSK is advantageous in that it does not require absolute phase estimation, but neither is as power efficient as the BPSK-SIM. The choice of modulation technique at the subcarrier level therefore depends on the application at hand and requires a compromise between simplicity, power, and bandwidth efficiencies. Matlab codes for simulating the performance for different modulation techniques, including those shown in Figure 6.16, are given in Appendix B. 6.3.3.4  Multiple SIM Performance Analysis The use of multiple subcarriers is a viable way of increasing system throughput/capacity. To archive this, different data/users are premodulated on different subcarrier frequencies. These are then aggregated and the resulting composite signal used to modulate the intensity of the optical source. For a subcarrier system with N subcarrier signals operating at different frequencies, the inherent nonlinearity of the optical source will result in the transfer of energy among the subcarrier frequencies (i.e., the intermodulation distortion (IMD)). The IMD potentially results in a reduced SNR as it contributes to the number of unwanted signals within the frequency band of interest. The multiple

327

FSO Link Performance with Turbulence BPSK 16-PSK 64-PSK BPSK No fading DPSK 16-QAM

-3

10

-4

10

-5

10

C2n = 75*10-16 m -2/3

-6

BER

10

-7

10

-8

10

-9

10

-10

10

-30

-25 -20 -15 Received average irradiance, I0 (dBm)

-10

FIGURE 6.16  BER against the received irradiance for SIM-FSO with different subcarrier modulation techniques in weak atmospheric turbulence for σl2 = 0.3, λ = 850nm, and link range = 1 km.

subcarrier system will be examined by assuming an ideal optical source with no reference to the IMD and the clipping distortion. By definition, the modulation index/depth is given by

ξ

m(t) = iB′

N

∑

A j g ( t ) cos ( ω cj + θ j ) iB′

j =1

(6.39)

where iB′ = iB − iTh as already shown in Figure 6.9. The peak value of the composite signal m ( t ) occurs when all the individual subcarrier amplitudes add up coherently, that is, N



ξ=

∑ j =1

Aj = iB′

N

∑ξ

j

= N ξsc (6.40)

j =1

In (6.19), all the subcarrier signals have been assumed to have the same individual modulation depth ξsc = ξ / N . Since the SNR on each subcarrier is proportional to the square of the modulation depth, there exists at least a loss of 20 log N (dB) in electrical SNR (equivalent to 10 log N (dB) optical power reduction) on each subcarrier due to the presence of N subcarrier signals with nonoverlapping frequencies. The BER plots of a BPSK premodulated subcarrier with different numbers of subcarriers are depicted in Figure 6.17. The graph is obtained by replacing ξ with ξsc in (6.33). For example, the SNR required to achieve a BER of 10 −6 with five subcarriers is ∼40 dB in an atmospheric channel with a fading strength of σl2 = 0.3. This increases to ∼46 dB with 10 subcarriers under the same

328

Optical Wireless Communications Number of subcarriers

-1

10

1 2 5 10

-2

10

-3

BER

10

-4

10

-5

10

-6

10

-7

10

-8

10

5

10

15

20 25 30 Normalised SNR (dB)

35

40

45

50

FIGURE 6.17  BER against the normalised SNR for multiple subcarrier FSO system in weak atmospheric turbulence for N = [1, 2, 5, 10 ] and σl2 = 0.3.

channel conditions, depicting a 20 log N (dB) increment over the SNR required by a single subcarrier FSO system. In Figure 6.18, these SNR values required to attain a BER of 10 −6 are plotted against the number of subcarriers for turbulent atmospheric channels of different fading strengths. This shows explicitly that for a given BER value, the SNR values increase with an increase in both the number of subcarriers and the fading strength. Based on the foregoing, therefore, the multiple SIM-FSO is only recommended whenever the quest for increased throughput/capacity outweighs its accompanying power penalty.

6.3.4 Outage Probability A system with an adequate average BER can temporarily suffer from increases in error rate due to deep fades, and this ‘short outage’ is not adequately modelled by the average BER [29]. The instantaneous capacity corresponding to the received signal y = RPthx + no and a channel state hk is given by [30]–[32] C ( SNR ( hk )) =

∑P ( x ) ∫ f ( y | x, h ) X

x=0



+∞

1

k

−∞

  (6.41)   f ( y | x , hk )  dy  ; × log2   f ( y | x = m, h ) PX ( m )   m = 0,1 

∑

329

FSO Link Performance with Turbulence 60

Required SNR to attain BER = 10-6 (dB)

55 50 45 40 35 30 2

σ l = 0.1 2

25

2

σ l = 0.2 2 2

σ l = 0.5 2

20

2

15

σ l = 0.7 2 1

2

3

4

5 6 7 Number of subcarriers

8

9

10

FIGURE 6.18  SNR required to attain a BER of 10 −6 against the number of subcarriers for BPSK-modulated SIM-FSO system with σ l = [ 0.1, 0.2, 0.5, 0.7 ] .

where PX ( x ) is the probability of the bit being one (x = 1) or zero (x = 0), with PX ( x = 0 ) = PX ( x = 1) = 0.5, Pt is the average transmit power, and no is the AWGN. f ( y | x , hk ) can be expressed as



   f ( y|x , hk ) =    

 y2  exp  − 2σ 2  ,  n  2 πσ 2n 

x=0

 ( y − 2 Pt Rhk )2  exp −  ,  2σ 2n 2 πσ 2n  

x = 1 

1

1

(6.42)

Note that, in the case of unknown channel at the receiver, the average channel capacity is given by 1



〈C 〉 =

∑ x=0

    f (y | x)  dy  ; (6.43) PX ( x ) f ( y | x ) log2   f ( y | x = m ) PX ( m )  −∞  m = 0,1  ∞

∫

∑

where f ( y | x ) can be expressed as



  y2  1   exp  − 2σ 2  ,   2 πσ 2n n   f ( y|x ) =  ∞  ( y − 2 Pt Rhk )2  1    exp −  fh ( hk ) dhk ,   2 πσ 2 2σ 2n   n 0  

∫

x=0 (6.44) x =1

330

Optical Wireless Communications

Since the optical slow-fading channel is random and remains unchanged over a long block of bits, the time-varying channel capacity will not be sufficient to support a maximum data rate, when the instantaneous SNR falls below a threshold [31], [33]. Such an occurrence, known as system outage, can potentially result in a loss of up to 10 9 consecutive bits at a data rate of 10 Gbps under deep fade scenarios that may last for ∼1–100 ms [34]. Therefore, in this case, the outage probability is an appropriate performance measure of the capacity, which represents the probability that the instantaneous channel capacity C falls below a transmission rate Rb, which is given by the relation

(

)

Pout = Prob C ( SNR ( hk )) < Rb (6.45)



Since C (⋅) is monotonically increasing with SNR = ( RPt ) / 2σ 2n, Pout is the cumulative density 2

function of hk evaluated at h0 = C −1 ( Rd ) σ 2n / 2 Pt2 R 2, and thus it can be determined equivalently from the expression [6.46] h0

∫

Pout = fh ( hk ) dh (6.46)



0

6.3.4.1  In a Log-Normal Atmospheric Channel The probability of the outage can be expressed as

(

)

(

)

Pout = P Pe > Pe* ≡ P SNR < SNR* (6.47)



where Pe* is a predetermined threshold BER, and SNR* is the threshold SNR that corresponds to Pe* in the absence of atmospheric turbulence for a given noise level. By introducing the parameter mp, here called ‘power margin’, to consider the additional power needed to account for turbulenceinduced signal fading, the outage probability is derived as follows:

(

Pout = P m p γ ( I ) < γ

*

)

I0 / m

=



∫ 0

(

 ln I / I 0 + σ l2 / 2  exp − 2 2σ l2 I 2 πσ l  1

)

2

   dI  (6.48)

σ   1 = Q  ln m p − l   σl 2

(

)

Invoking the Chernoff upper bound, Q ( x ) ≤ 0.5exp − x 2 / 2 on equation (6.48) gives an upper bound value for the outage probability. From this an approximate power margin, m, needed to obtain Pout can be obtained as

m p ≈ exp

(

)

−2 ln 2 Pout σ l2 + σ l2 / 2 (6.49)

This extra power needed to obtain a given outage probability is depicted in Figure 6.19 at various levels of irradiance fluctuation. For example, to achieve an outage probability of 10 −6, about 35 dBm of extra power is needed at σl2 = 0.22. This will increase to ∼43 dBm and 48 dBm for σl2 = 0.52 and σl2 = 0.72, respectively. The extra margin can also be viewed as the penalty introduced by the atmospheric turbulence and to reduce it. Diversity techniques will be considered in Chapter 7. The Matlab codes for generating the outage probability for a log-normal turbulence atmospheric channel is given in Program 6.4.

331

FSO Link Performance with Turbulence σ 2l = 0.1 σ 2l = 0.3

-2

10

σ 2l = 0.5

Outage Probability , P

Out

σ 2l = 1 -4

10

-6

10

-8

10

-10

10

35

40

45 50 Power Margin (dBm)

55

60

FIGURE 6.19  Outage probability against the power margin for a log-normal turbulent atmospheric channel for σ l2 = [ 0.1, 0.3, 0.5, 1].

Program 6.4:  Matlab code for Figure 6.19. %Evaluating the additional power (margin, m) needed to achieve a given outage probability, using the Chernoff upper bound. log-normal scintillation model used. clear clc Rhol = [0.5] ro = sqrt(Rhol); %Log intensity standard deviation for j = 1: length(ro) r = ro(j); Pout = logspace(0,-10,50); %Outage probability for i = 1: length(Pout) Po = Pout(i); arg = sqrt(-2*r^2*log(2*Po)) + ((r^2)/2); mp(i) = exp(arg); margin(j,i) = 10*log10(mp(i)*1e3); %Power margin to achieve outage prob.; %in dBm. end end semilogy(margin,Pout) xlabel(‘Power Margin (dBm)’) ylabel(‘Outage Probability’) title(‘SISO’)

332

Optical Wireless Communications

6.3.5 SIM-FSO Performance in Gamma-Gamma and Negative Exponential Atmospheric Channels To obtain the unconditional BER in a gamma-gamma turbulent atmospheric channel, the irradiance fluctuation statistics in the previous section are replaced appropriately by the gamma-gamma pdf. For a BPSK-premodulated subcarrier, the unconditional BER now becomes ∞

∫ (

γ (I )

Pe = Q



0

(α+β ) / 2

) I ) 2Γ(αβ ( α ) Γ (β )

 α+β    −1 2 

(

)

K α−β 2 αβI dI (6.50)

This BER expression can only be evaluated numerically because it does not have a closed-form solution. The values of the parameters α and β, which are used to describe the turbulence strength, are as given in Table 6.3. In the limit of strong turbulence, that is, in the saturation regime and beyond, the BER is obtained by replacing the pdf in the conditional BER with the fully developed speckle (negative exponential) pdf discussed in Chapter 3. By applying the alternative representation of Q(.) given by (6.30), the unconditional BER in the fully developed speckle regime is derived as 1 Pe = πI 0



2  ξRI ) ( I  exp  − 2 2 −  dIdϑ (6.51)  4σ n sin ϑ I 0  0

π/2 ∞

∫∫ 0

The multiple integrations involved in (6.51) can be conveniently circumvented, and doing this reduces the BER expression Pe to the following: 1 Pe = π



π/2

∫

πK o ( ϑ ) exp ( K 0 ( ϑ )) erfc 

(

)

K 0 ( ϑ ) d ϑ (6.52)

0

where K 0 ( ϑ ) = ( σ n  sin ( ϑ ) / ξ R ) and erfc (.) is the complementary error function. The following upper bound, given by (6.26), is then obtained by maximising the integrand with the substitution of ϑ = π / 2 in (6.52). 2

Pe ≤ πK 0 exp ( K 0 ) Q



(

)

2K 0 (6.53)

where K 0 = ( σ n / ξ R ) . From these BER expressions, the error performance of the system can be predicted for any given value of SNR and turbulence strength (or link range). The numerical simulations of the BER expressions (6.50), (6.52) and the upper bound (6.53) are shown in Figure 6.20, where the Pe is plotted against the normalised SNR under different turbulence regimes. 2

TABLE 6.3 Fading Strength Parameters for Gamma-Gamma Turbulence Model Parameter σl2 α β

Weak 0.2 11.6 10.1

Turbulence regime Moderate 1.6 4.0 1.9

Strong 3.5 4.2 1.4

333

FSO Link Performance with Turbulence

Turbulence regime Weak Moderate Strong Saturation (Exact) Saturation (Upper bound)

-2

10

-3

BER

10

-4

10

-5

10

-6

10

20

40

60 80 Electrical SNR (dB)

100

120

FIGURE 6.20  BER performance against the normalised electrical SNR across all of the turbulence regimes based on gamma-gamma and negative exponential modes.

For instance, to achieve a BER of 10 −6 in weak atmospheric turbulence, the required SNR is ∼29 dB, and this rises to ∼65 dB and ∼67 dB, respectively, for moderate and intermediate regimes. While in the saturation regime, a staggering ∼115 dB (the upper bound value is 4 dB higher) is required to achieve the same level of error performance (i.e., BER of 10 −6). Achieving a BER lower than 10 −6 in the saturation regime requires a phenomenal increase in SNR as seen in Figure 6.20. It is noteworthy that the normalised SNR is in the electrical domain, and it is based on the average received irradiance, E[I]. In addition, the ‘kinks’ observed in the curves for strong and moderate turbulence are due to the numerical integration process. In order to compare the error performance BPSK-SIM with an OOK-modulated FSO system of the same average transmitted optical power, the unconditional BER of the OOK-FSO is modified to become



∞ Pe = 0.5    ith

∫

1 πσ 2n

(

)

exp −i 2 / σ 2n di +

∞ ith

∫∫ 0 0

  ( i − 2 RI )2   exp − p ( I ) didI (6.54)   σ 2n  πσ 2n    1

In Figure 6.21, the BER performances of the OOK system employing adaptive (optimum) and fixed threshold values of 0.05 and 0.8 are shown alongside that of BPSK-SIM in a weak turbulent atmospheric fading. Although the optimum OOK is marginally superior to BPSK-SIM under the stated conditions, because it requires 1.6 dB electrical SNR less at a BER of 10 −6, it does require an accurate knowledge of both the additive noise and fading levels to achieve this performance. With the threshold fixed at say 0.05, the OOK requires about 7 dB electrical SNR more than BPSK-SIM

334

Optical Wireless Communications

-1

10

-2

10

-3

BER

10

-4

10

-5

10

Weak turbulence regime OOK (i th = 0.8) OOK (i th = 0.05)

-6

10

BPSK-SIM OOK (Optimum) 5

10

15 20 25 Normalised SNR (dB)

30

35

FIGURE 6.21  Error performance of BPSK-SIM and OOK with a fixed and adaptive threshold-based FSO in weak turbulence regime modeled using gamma-gamma distribution.

at a BER of 10 −6. Also, the BER performance of OOK with a fixed threshold level exhibits a BER floor as shown in Figure 6.21 for ith = 0.8 and in the previous work by [18]. The SIM is therefore recommended in atmospheric turbulence channels as against the fixed threshold OOK currently used in commercial terrestrial FSO systems.

6.3.6 Outage Probability in Negative Exponential Model Atmospheric Channels By following the approach of Section 6.3.4, the outage probability in the fully developed speckle is obtained as follows:

(

Pout = P m p γ ( I ) < γ

*

)=

I0 / m

∫ 0

1  I  exp  −  dI (6.55)  I0  I0

From (6.55) m p, needed to achieve a given Pout saturation regime, is as given by (6.56). This is plotted in Figure 6.22, alongside equation (6.49) with σ l2 = 0.5, which represents the outage probability in weak atmospheric turbulence. A comparison of the results in this figure reveals that the power margin required to achieve a Pout of 10 −6 in the fully developed speckle regime is about 40 dB higher than that required in the weak turbulence regime with σ l2 = 0.5, and this will increase as the required outage probability level is reduced to below 10 −6. The compression of OOK, PPM and subcarrier intensity modulation schemes are given in Table 6.4.

−1

m p =  − ln (1 − Pout )  (6.56)

335

FSO Link Performance with Turbulence Fading strength 2

σ l = 0.5 -2

Saturation regime

Outage Probability , P

Out

10

-4

10

-6

10

-8

10

-10

10

40

50

60

70 80 90 Power Margin (dBm)

100

110

120

130

FIGURE 6.22  The outage probability against the power margin in saturation and weak turbulence regimes for σ l2 = 0.5.

The prohibitive power required in the saturation regime suggests that the establishment of a reliable communication link in this regime is impossible unless the fading effect due to turbulence is mitigated or compensated for.

6.4  ATMOSPHERIC TURBULENCE-INDUCED PENALTY In this section, we will examine the additional power required to achieve a given level of performance due to the presence of turbulence-induced channel fading. Without any loss of generality, the result here will be based on a log-normal turbulence model and the BPSK-modulated subcarrier. This can, however, be extended to other turbulence models and modulation schemes in a straightforward manner. In Figure 6.23, the BER performance is plotted as a function of the normalised SNR for different levels of atmospheric turbulence. As an example, to achieve a BER of 10 −6 in a channel characterised by σ l2 = 0.1, an SNR of about 24 dB will be required. This increases to ∼37 dB as the fading strength increases to σ l2 = 0.5. Also from this plot, the SNR penalty at any given fading strength and BER can be obtained. The SNR penalty is defined as the difference between the SNR required to achieve a specified BER in the presence and absence of atmospheric turbulence. The SNR penalty is shown in Figure 6.24 for the following BER levels: 10 −3, 10 −6, and 10 −9. As we would expect, the SNR penalty increases as the turbulence strength increases and as the benchmark BER value decreases. Beyond σ l2 = 0.1, the SNR penalty increases very sharply—for instance, at a BER of 10 −6, when the fading strength rises tenfold from σ l2 = 0.01 to σ l2 = 0.1, the penalty only rises from 1 dB to 6.5 dB. But when the 2 fading strength increases to σ l = 0.3 , the SNR penalty rises to 15 dB. The Matlab codes for plotting the turbulence-induced SNR penalty as shown in Figure 6.24 are given in Program 6.5.

336

Optical Wireless Communications

-2

10

-4

10

-6

BER

10

-8

10

Fading strength 2

-10

10

σ l = 0.1 2

σ l = 0.3 2

σ l = 0.5 -12

10

2

σ l = 0.7 No fading 15

20

25 30 Normalised SNR (dB)

35

40

45

FIGURE 6.23  Error rate performance against normalised SNR for BPSK-SIM-based FSO in weak atmospheric turbulence channel for σ l2 = [ 0, 0.1, 0.3, 0.5, 0.7 ].

Program 6.5:  Matlab script for Figure 6.24. %Evaluation of BER of SISO BPSK MSM FSO under weak turbulence using the Gauss-Hermite Quadrature integration approach. clear clc %*************Parameters*************************** R=1; %Responsivity Io=1; %Intensity without turbulence N=1; %no of subcarrier r=sqrt(0.1); %log intensity standard deviation Noise=logspace(0,-5,100); %Gaussian noise variance %************************************************** for j=1:length(Noise) No = Noise(j); SNR(j)=10*log10(((R*Io)^2)/(No)); K=(R*Io)/(sqrt(2*No)*N); %****Hermite polynomial weights and roots************ w20=[2.22939364554e-13,4.39934099226e-10,1.08606937077e-7, 7.8025564785e-6,0.000228338636017,0.00324377334224,0.0248105208875, 0.10901720602,0.286675505363,0.462243669601,..

337

FSO Link Performance with Turbulence 45 BER = 10 -3 40

BER = 10 -6 BER = 10 -9

35

SNR Penalty (dB)

30 25 20 15 10 5 0 -2 10

-1

0

10

10 2

Log irradiance variance (fading strength), σ l

FIGURE 6.24  Turbulence-induced SNR penalty as function of log irradiance variance for BPSK-SIM-based FSO for BER = [10 −3, 10 −6, 10 −9].

0.462243669601,0.286675505363,0.10901720602,0.0248105208875, 0.00324377334224,0.000228338636017,7.8025564785e-6,1.08606937077e-7, 4.39934099226e-10,2.22939364554e-13]; x20=[-5.38748089001,-4.60368244955,-3.94476404012,-3.34785456738, -2.78880605843,-2.25497400209,-1.73853771212,-1.2340762154, -0.737473728545,-0.245340708301,... 0.245340708301,0.737473728545,1.2340762154,1.73853771212, 2.25497400209,2.78880605843,3.34785456738,3.94476404012,4.60368244955, 5.38748089001]; %**************************************************** GH=0; for i=1:length(x20) arg=K*exp(x20(i)*sqrt(2)*r – r^2/2); temp=w20(i)*Q(arg); GH=GH + temp; end BER(j) = GH/sqrt(pi); end %*********Plot function**************************** semilogy(SNR,BER) xlabel(‘SNR (R*E[I])^2/No (dB)’); ylabel(‘BER’); %title(‘BPSK E[I]=Io=1 R=1’)

338

Optical Wireless Communications 0

10

-5

10

-10

10

-15

BER

10

-20

10

-25

10

-30

10

Normalised SNR 5 dB 20 dB 25 dB 30 dB

-35

10

-40

10

-2

10

-1

10 2 Log irradiance variance, σ l

0

10

FIGURE 6.25  BER of BPSK-SIM against the turbulence strength in weak atmospheric turbulence for normalised SNR (dB) = 5, 20, 25, and 30.

Moreover, in Figure 6.25 and Figure 6.26 respectively, the BER and Pout metrics of the BPSKSIM scheme are plotted against the turbulence strength for different values of SNR. These plots show that increasing the signal power can help mitigate the fading caused by the atmospheric turbulence, but only in the very weak regime, where σ l2 < 0.1. As the turbulence strength increases beyond this value, the BER and indeed Pout both tend to high asymptotic values that are too high to guarantee a reliable exchange of information via the link. This implies that techniques other than a mere increase in transmitted power will be required to mitigate atmospheric turbulence beyond the very weak regime. To do just that, diversity techniques will be examined in Chapter 7.

TABLE 6.4 Comparison of Modulation Techniques On-off keying

Pulse position modulation

Subcarrier intensity modulation

Simple to implement Synchronisation not required Adaptive threshold required in fading channels Component nonlinearity not an issue Suboptimal with a fixed threshold

Power efficient Synchronisation required Adaptive threshold not required in fading channels Component nonlinearity not an issue

Power inefficient Higher throughput Adaptive threshold not required in fading channels Component nonlinearity an issue with multiple subcarriers Benefits from advances in digital signal processing and matured RF components

High bandwidth requirement

339

FSO Link Performance with Turbulence 0

10

-2

10

Outage probability, POut

-4

10

-6

10

-8

10

Power margin, m

-10

10

30 dBm 35 dBm 38 dBm 40 dBm

-12

10

-2

10

-1

10 2 Log irradiance variance (fading strength), σ l

0

10

FIGURE 6.26  POut of BPSK-SIM against the turbulence strength in weak atmospheric turbulence for m p (dBm) = 30, 35, 38, and 40.

APPENDIX A Matlab Scripts for Sections 6.3.2, 6.3.3.2, And 6.3.3.3 Section 6.3.2 ******************%Function Basebandmodulation%************************** function [Inphase,Quadrature,X] = Basebandmodulation(M,symb,N_sub) X = randint(symb,N_sub,M); H = modem.pskmod(‘M’,M,’phaseoffset’,pi/4,’SymbolOrder’,’gray’); Y = modulate(H,X); Inphase = real(Y); Quadrature = imag(Y); ******************%Function Turbulence %************************** function [I] = Turbulence(Log_Int_var,No_symb,Io,t) %Log intensity Variance Var_l = Log_Int_var; % Number of symbols %No_symb = 1e4; %Log normal atmospheric turbulence l = (sqrt(Var_l).*randn(1,No_symb)) - (Var_l/2); I1 = Io.*exp(l); for i = 1:No_symb a = 1+(i-1)*length(t); b = i*length(t); I(1,a:b) = I1(i); end

340

Optical Wireless Communications

Sections 6.3.3.2 and 6.3.3.3 %Evaluation of BER of SISO M-QAM SIM FSO under weak turbulence using the Gauss-Hermite Quadrature integration approach; %Considering Background and thermal noise. clear clc %*************Simulation Parameters*************************** M = 16; %Number of levels; M-ary QAM; M should be even Rb = 155e6; %symbol rate R = 1; %Responsivity M_ind = 1; %Modulation index A = 1; %Subcarrier signal amplitude RL = 50; %Load resistance Temp = 300; %Ambient temperature wavl = 850e-9; %Optical source wavelength %********************Background Noise************************ %Considering 1 cm^2 receiving aperture sky_irra = 1e-3; % at 850nm wavelength, in W/cm^2-um-sr sun_irra = 550e-4; %at 850nm wavelength, in W/cm^2-um FOV = 0.6; % in radian OBP = 1e-3; %Optical filter bandwidth in micrometre Isky = sky_irra*OBP*(4/pi)*FOV^2; %Sky irradiance Isun = sun_irra*OBP; %Sun irradiance %**************Rytov Variance******************************************** Range = 1e3; %Link range in meters Cn = 0.75e-14; %Refractive index structure parameter Rhol = 1.23*(Range^(11/6))*Cn*(2*pi/wavl)^(7/6); %Log irradiance variance (Must be less than 1 Varl = Rhol; %Log intensity variance r = sqrt(Varl); %log intensity standard deviation %*****************Physical constants**************************** E_c = 1.602e-19; %Electronic charge B_c = 1.38e-23; %Boltzmann constant %************************************************************************ Pd = A^2/2; Ktemp = (4*B_c*Temp*Rb/RL) + (2*E_c*R*Rb*(Isun + Isky)); K1 = 3*log2(M)*((R*M_ind)^2)*Pd /(2*(M - 1)*Ktemp); %*********background and thermal noise combined************************** %****Hermite polynomial weights and roots************ w20=[2.22939364554e-13,4.39934099226e-10,1.08606937077e-7, 7.8025564785e-6,0.000228338636017,0.00324377334224,0.0248105208875, 0.10901720602,0.286675505363,0.462243669601,... 0.462243669601,0.286675505363,0.10901720602,0.0248105208875, 0.00324377334224,0.000228338636017,7.8025564785e-6,1.08606937077e-7, 4.39934099226e-10,2.22939364554e-13]; x20=[-5.38748089001,-4.60368244955,-3.94476404012,-3.34785456738, -2.78880605843,-2.25497400209,-1.73853771212,-1.2340762154, -0.737473728545,-0.245340708301,...

FSO Link Performance with Turbulence

341

0.245340708301,0.737473728545,1.2340762154,1.73853771212, 2.25497400209,2.78880605843,3.34785456738,3.94476404012, 4.60368244955,5.38748089001]; %**************************************************** %*******************BER evaluation******************* Io = logspace(-10,-4,30); %Average received irradiance IodBm =10*log10(Io*1e3); %Average received irradiance in dBm SNR2 = ((R.*Io).^2)./(Ktemp); SNR2dB = 10*log10(SNR2); for i1 = 1:length(Io) GH1 = 0; for i2 = 1:length(x20) arg1 = sqrt(K1)*Io(i1)*exp(x20(i2)*sqrt(2)*r - Varl/2); temp1 = w20(i2)*qfunc(arg1); GH1 = GH1 + temp1; end BER1(i1) = 2*(1 - (1/sqrt(M)))*GH1/(log2(M)*sqrt(pi)); end %*********Plot function**************************** %figure semilogy(IodBm,BER1) xlabel(‘Received average irradiance,E[I] (dBm)’) ylabel(‘BER’) %Evaluation of BER of SISO M-PSK SIM FSO under weak turbulence using the Gauss-Hermite Quadrature integration approach; %Considering Background and thermal noise sources. clear clc %*************Simulation Parameters*************************** M = 4; %Number of phase levels; M-ary PSK Rb = 155e6; %symbol rate R = 1; %Responsivity M_ind = 1; %Modulation index A = 1; %Subcarrier signal amplitude RL = 50; %Load resistance Temp = 300; %Ambient temperature wavl = 850e-9; %Optical source wavelength %********************Background Noise************************ %Considering 1 cm^2 receiving aperture sky_irra = 1e-3; % at 850nm wavelength, in W/cm^2-um-sr sun_irra = 550e-4; %at 850nm wavelength, in W/cm^2-um FOV = 0.6; % in radian OBP = 1e-3; %Optical filter bandwidth in micrometre Isky = sky_irra*OBP*(4/pi)*FOV^2; %Sky irradiance Isun = sun_irra*OBP; %Sun irradiance %**************Rytov Variance******************************************** Range = 1e3; %Link range in meters Cn = 0.75e-14; %Refractive index structure parameter Rhol = 1.23*(Range^(11/6))*Cn*(2*pi/wavl)^(7/6); %Log irradiance %variance (Must be less than 1 Varl = Rhol; %Log intensity variance r = sqrt(Varl); %log intensity standard deviation

342

Optical Wireless Communications

%*****************Physical constants**************************** E_c = 1.602e-19; %Electronic charge B_c = 1.38e-23; %Boltzmann constant %************************************************************************ Pd = A^2/2; Ktemp = (4*B_c*Temp*Rb/RL) + (2*E_c*R*Rb*(Isun + Isky)); K1 = ((sin(pi/M))^2)*log2(M)*((R*M_ind)^2)*Pd/Ktemp; %background and thermal noise combined %Hermite polynomial weights and roots************ w20=[2.22939364554e-13,4.39934099226e-10,1.08606937077e-7, 7.8025564785e-6,0.000228338636017,0.00324377334224,0.0248105208875, 0.10901720602,0.286675505363,0.462243669601,... 0.462243669601,0.286675505363,0.10901720602,0.0248105208875, 0.00324377334224,0.000228338636017,7.8025564785e-6,1.08606937077e-7, 4.39934099226e-10,2.22939364554e-13]; x20=[-5.38748089001,-4.60368244955,-3.94476404012,-3.34785456738, -2.78880605843,-2.25497400209,-1.73853771212,-1.2340762154, -0.737473728545,-0.245340708301,... 0.245340708301,0.737473728545,1.2340762154,1.73853771212, 2.25497400209,2.78880605843,3.34785456738,3.94476404012, 4.60368244955,5.38748089001]; %****************BER evaluation******************* Io = logspace(-10,-3,30); %Average received irradiance IodBm =10*log10(Io*1e3); %Average received irradiance in dBm SNR2 = ((R.*Io).^2)./(Ktemp); SNR2dB = 10*log10(SNR2); for i1 = 1:length(Io) GH1 = 0; for i2 = 1:length(x20) arg1 = sqrt(K1)*Io(i1)*exp(x20(i2)*sqrt(2)*r - Varl/2); temp1 = w20(i2)* qfunc(arg1); GH1 = GH1 + temp1; end BER1(i1) = 2*GH1/(log2(M)*sqrt(pi)); end %*********Plot function**************************** %figure semilogy(IodBm,BER1) xlabel(‘Received average irradiance,E[I] (dBm)’) ylabel(‘BER’) %Evaluation of BER of SISO DPSK SIM FSO under weak turbulence using the %Gauss-Hermite Quadraturen integration approach; %Considering Background and thermal noise sources. clear clc %*************Simulation Parameters*************************** Rb = 155e6; %symbol rate R = 1; %Responsivity M_ind = 1; %Modulation index A = 1; %Subcarrier signal amplitude RL = 50; %Load resistance

FSO Link Performance with Turbulence Temp = 300; wavl = 850e-9;

343

%Ambient temperature %Optical source wavelength

%********************Background Noise************************ %Considering 1 cm^2 receiving aperture sky_irra = 1e-3; % at 850nm wavelength, in W/cm^2-um-sr sun_irra = 550e-4; %at 850nm wavelength, in W/cm^2-um FOV = 0.6; % in radian OBP = 1e-3; %Optical filter bandwidth in micrometre Isky = sky_irra*OBP*(4/pi)*FOV^2; %Sky irradiance Isun = sun_irra*OBP; %Sun irradiance %**************Rytov Variance******************************************** Range = 1e3; %Link range in meters Cn = 0.75e-14; %Refractive index structure parameter Rhol = 1.23*(Range^(11/6))*Cn*(2*pi/wavl)^(7/6); %Log irradiance variance (Must be less than 1 Varl = Rhol; %Log intensity variance r = sqrt(Varl); %log intensity standard deviation %*****************Physical constants**************************** E_c = 1.602e-19; %Electronic charge B_c = 1.38e-23; %Boltzmann constant %************************************************************************ Pd = A^2/2; Ktemp = (4*B_c*Temp*Rb/RL) + (2*E_c*R*Rb*(Isun + Isky)); K1 = 0.5*((R*M_ind)^2)*Pd/Ktemp; %background and thermal noise combined %************************************************************** %****Hermite polynomial weights and roots************ w20=[2.22939364554e-13,4.39934099226e-10,1.08606937077e7,7.8025564785e-6,0.000228338636017,0.00324377334224,0.0248105208875, 0.10901720602,0.286675505363,0.462243669601,... 0.462243669601,0.286675505363,0.10901720602,0.0248105208875, 0.00324377334224,0.000228338636017,7.8025564785e-6,1.08606937077e-7, 4.39934099226e-10,2.22939364554e-13]; x20=[-5.38748089001,-4.60368244955,-3.94476404012,-3.34785456738, -2.78880605843,-2.25497400209,-1.73853771212,-1.2340762154, -0.737473728545,-0.245340708301,... 0.245340708301,0.737473728545,1.2340762154,1.73853771212, 2.25497400209,2.78880605843,3.34785456738,3.94476404012, 4.60368244955,5.38748089001]; %**************************************************** %*******************BER evaluation******************* Io = logspace(-10,-2,30); %Average received irradiance IodBm =10*log10(Io*1e3); %Average received irradiance in dBm SNR2 = RL*((R.*Io).^2)./(4*B_c*Temp*Rb); SNR2dB = 10*log10(SNR2); for i1 = 1:length(Io) GH1 = 0; for i2 = 1:length(x20) arg1 = K1*(Io(i1)^2)*exp(2*x20(i2)*sqrt(2)*r - Varl);

344

Optical Wireless Communications temp1 = w20(i2)*exp(-arg1); GH1 = GH1 + temp1;

end BER1(i1) = GH1/(2*sqrt(pi)); end %*********Plot function****************************; %figure semilogy(IodBm,BER1) xlabel(‘Received average irradiance,E[I] (dBm)’) ylabel(‘BER’)

APPENDIX B

TABLE B.1 Zeros and Weights of Gauss-Hermite Integration with n = 20 ∞

∫

f ( x ) dx =

−∞

∞

∫

2 2 e− x  e x f ( x )  dx ≈  

−∞

n

∑w ( x ) e i

i =1

i

Zeros, xi

Weight, w ( xi )

− xi2

f ( xi ) 2

x Total weight, w ( xi ) e i

1

−5.38748089001

2.22939364554E-013

0.898591961453

2

−4.60368244955

4.39934099226E-010

0.704332961176

3

−3.94476404012

1.08606937077E-007

0.62227869619

4

−3.34785456738

7.8025564785E-006

0.575262442852

5

−2.78880605843

0.000228338636017

0.544851742366

6

−2.25497400209

0.00324377334224

0.524080350949

7

−1.73853771212

0.0248105208875

0.509679027117

8

−1.2340762154

0.10901720602

0.499920871336

9

−0.737473728545

0.286675505363

0.493843385272

10

−0.245340708301

0.462243669601

0.490921500667

11

0.245340708301

0.462243669601

0.490921500667

12

0.737473728545

0.286675505363

0.493843385272

13

1.2340762154

0.10901720602

0.499920871336

14

1.73853771212

0.0248105208875

0.509679027117

15

2.25497400209

0.00324377334224

0.524080350949

16

2.78880605843

0.000228338636017

0.544851742366

17

3.34785456738

7.8025564785E-006

0.575262442852

18

3.94476404012

1.08606937077E-007

0.62227869619

19

4.60368244955

4.39934099226E-010

0.704332961176

20

5.38748089001

2.22939364554E-013

0.898591961453

REFERENCES 1. H. Willebrand and B. S. Ghuman, Free Space Optics: Enabling Optical Connectivity in Today’s Network. Indianapolis: SAMS publishing, 2002. 2. S. Bloom, E. Korevaar, J. Schuster, and H. Willebrand, “Understanding the Performance of Free-Space Optics,” J. Opt. Netw., vol. 2, no. 6, pp. 178–200, 2003.

FSO Link Performance with Turbulence

345

3. X. Zhu and J. M. Kahn, “Free-Space Optical Communication through Atmospheric Turbulence Channels,” IEEE Trans. Commun., vol. 50, no. 8, pp. 1293–1300, 2002. 4. S. M. Navidpour, M. Uysal, and L. Jing, “BER Performance of MIMO Free-Space Optical Links,” 60th IEEE Veh. Technol. Conf., vol. 5, pp. 3378–3382, 2004. 5. H. R. Burris et al., “Adaptive thresholding for Free-Space Optical Communication Receivers with Multiplicative Noise,” in Proceedings, IEEE Aerospace Conference, vol. 3, pp. 3-1473-3–1480. 6. H. R. Burris, C. I. Moore, L. A. Swingen, L. M. Wasiczko, R. Mahon, M. F. Stell, M. R. Suite, W. S. Rabinovich, J. L. Murphy, G. C. Gilbreath, and W. J. Scharpf, “Laboratory Implementation of an Adaptive Thresholding System for Free-space Optical Communication Receivers with Signal Dependent Noise”, Proc. SPIE 5892, Free-Space Laser Communications V, 2005, vol. 5892, p.58920W. 7. I. B. Djordjevic, B. Vasic, and M. A. Neifeld, “Multilevel Coding in Free-Space Optical MIMO Transmission with Q-Ary PPM over the Atmospheric Turbulence Channel,” IEEE Photonics Technol. Lett., vol. 18, no. 14, pp. 1491–1493, 2006. 8. M. Razavi and J. H. Shapiro, “Wireless Optical Communications via Diversity Reception and Optical Preamplification,” IEEE Trans. Commun., vol. 4, no. 3, pp. 975–983, 2005. 9. S. G. Wilson, M. Brandt-Pearce, Q. Cao, and J. H. Leveque, “Free-Space Optical MIMO Transmission with Q-ary PPM,” IEEE Trans. Commun., vol. 53, no. 8, pp. 1402–1412, 2005. 10. K. Kiasaleh, “Performance of APD-Based, PPM Free-Space Optical Communication Systems in Atmospheric Turbulence,” IEEE Trans. Commun., vol. 53, no. 9, pp. 1455–1461, 2005. 11. S. Sheikh Muhammad, W. Gappmair, and E. Leitgeb, “PPM Channel Capacity Evaluation for Terrestrial FSO Links,” Int. Work. Satell. Sp. Commun., pp. 222–226, 2006. 12. H. Hemmati, “Deep Space Optical Communications,” in Deep Space Communications and Navigation Series. Wiley-Interscience, California, 2006. 13. G. Keiser, Optical Fiber Communications. McGraw-Hill Companies, 2011. 14. W. Popoola, Z. Ghassemlooy, M. S. Awan, and E. Leitgeb, “Atmospheric Channel Effects on Terrestrial Free Space Optical Communication Links,” in International Conference on Electronics, Computers and Artificial Intelligence (ECAI 2009), 2009, vol. 3, pp. 17–23. 15. W. O. Popoola, Subcarrier Intensity Modulated Free Space Optical Communication Systems. Northumbria University, Newcastle upon Tyne, 2009. 16. H. Hemmati, “Interplanetary Laser Communications,” Opt. Photonics News, vol. 18, no. 11, pp. 22–27, 2007. 17. S. Karp, E. L. O’Neill, and R. M. Gagliardi, “Communication Theory for the Free-Space Optical Channel,” Proc. IEEE, vol. 58, no. 10, pp. 1626–1650, 1970. 18. T. Obtsuki and T. Ohtsuki, “Multiple-Subcarrier Modulation in Optical Wireless Communications,” IEEE Commun. Mag., vol. 41, no. 3, pp. 74–79, Mar. 2003. 19. I. B. Djordjevic and B. Vasic, “100-Gb/s Transmission Using Orthogonal Frequency -Division Multiplexing,” IEEE Photonics Technol. Lett., vol. 18, no. 15, pp. 1576–1578, 2006. 20. H. Rongqing, Z. Benyuan, H. Renxiang, T. A. Christopher, R. D. Kenneth, and R. Douglas, “Subcarrier Multiplexing for High-Speed Optical Transmission,” J. Light. Technol., vol. 20, no. 3, pp. 417–424, 2002. 21. Z. Ghassemlooy, W. O. Popoola, and E. Leitgeb, “Free-space Optical Communication Using Subcarrier Modulation in Gamma-Gamma Atmospheric Turbulence,” in Proceedings of 2007 9th International Conference on Transparent Optical Networks, ICTON 2007, 2007, vol. 3, pp. 156–160. 22. R. You and J. M. Kahn, “Average Power Reduction Techniques for Multiple-Subcarrier IntensityModulated Optical Signals,” IEEE Trans. Commun., vol. 49, no. 12, pp. 2164–2171, 2001. 23. S. Teramoto and T. Ohtsuki, “Multiple-Subcarrier Optical Communication Systems with Peak Reduction Carriers,” IEEE Global Telecommunications Conference (GLOBECOM), vol. 6. pp. 3274–3278, 2003. 24. S. Teramoto and T. Ohtsuki, “Multiple-Subcarrier Optical Communication Systems with Subcarrier Signal-Point Sequence,” IEEE Trans. Commun., vol. 53, no. 10, pp. 1738–1743, 2005. 25. M. K. Simon and M.-S. Alouini, Digital Communication Over Fading Channels, 2nd ed. New York: John Wiley & Sons Inc., 2004. 26. M. Abramowitz and I. Stegun, Handbook of Mathematical Functions : With Formulas, Graph, and Mathematical Tables. Dover Publications, 1974. 27. J. G. Proakis, Digital Communications. New York: McGraw-Hill, 2004. 28. W. O. Popoola, Z. Ghassemlooy, and E. Leitgeb, “BER performance of DPSK Subcarrier Modulated Free Space Optics in Fully Developed Speckle,” in 2008 6th International Symposium on Communication Systems, Networks and Digital Signal Processing, 2008, pp. 273–277.

346

Optical Wireless Communications

29. V. W. S. Chan, “Free-Space Optical Communications,” IEEE J. Light. Technol., vol. 24, no. 12, pp. 4750–4762, 2006. 30. A. A. Farid and S. Hranilovic, “Outage Capacity Optimization for Free-Space Optical Links with Pointing Errors,” J. Light. Technol., vol. 25, no. 7, pp. 1702–1710, 2007. 31. D. K. Borah and D. G. Voelz, “Pointing Error Effects on Free-Space Optical Communication Links in the Presence of Atmospheric Turbulence,” J. Light. Technol., vol. 27, no. 18, pp. 3965–3973, 2009. 32. Y. Ren, A. Dang, B. Luo, and H. Guo, “Capacities for Long-Distance Free-Space Optical Links under Beam Wander Effects,” IEEE Photonics Technol. Lett., vol. 22, no. 14, pp. 1069–1071, Jul. 2010. 33. E. Biglieri, J. Proakis, and S. Shamai, “Fading Channels: Information-Theoretic and Communications Aspects,” IEEE Trans. Inf. Theory, vol. 44, no. 6, pp. 2619–2692, 1998. 34. E. J. Lee and V. W. S. Chan, “Part 1: Optical Communication over the Clear Turbulent Atmospheric Channel Using Diversity,” IEEE J. Sel. Areas Commun., vol. 22, no. 9, pp. 1896–1906, 2004.

9781498742696_Ch06.indd 346

18/04/19 12:10 PM

7

Outdoor OWC Links with Diversity Techniques

Two primary challenges attributed to outdoor OWC (i.e., FSO) communications are (i) building sway and (ii) scattering/absorption-induced attenuation and scintillation-induced link fading. For the former, one would need accurate pointing and tracking mechanisms, photodetector arrays, and wide beam profiles. Widening the optical beam width will help against building sway, but this will be at the cost of increased geometric loss (see Section 3.3.3). Even under clear atmospheric conditions, FSO links may experience fading due to the turbulence-induced irradiance and phase fluctuations. The signal fluctuation results from a random index of refraction variations along the propagation path as discussed in Chapter 3. The phase and irradiance fluctuations suffered by the traversing beam make optical coherent detection less attractive, simply because it is sensitive to both the signal amplitude and phase fluctuations. This then is one of the reasons behind the popularity of direct detection in terrestrial FSO links and, of course, the fact that direct detection receiver is also much simpler. The scintillation effects could result in deep irradiance fades of up to 20–30 dB that lasts up to ∼1–100 ms [1], [2]. For a link operating at, say, a data rate of 1 Gbps, this could result in a loss of up to 108 consecutive bits (i.e., burst errors). In strong turbulence channels, single-inputsingle-output (SISO) FSO link performance is extremely poor (i.e., not satisfying the typical BER targets) for FSO applications within a given SNR [3]–[6]. Under fog conditions, the atmospheric loss can lead to link attenuation as high as 50 dB/km for 0.2 in Figure 7.5, SelC proves worthwhile with positive link margins, but the gains are still lower than that obtainable from EGC by about ∼1 to ∼6 dB, depending on the number of photodetectors used. Based on the foregoing, SelC spatial diversity will not be recommended for use on short link FSO experiencing weak irradiance fluctuation. The performance of EGC and MRC linear combiners is compared in Figure 7.11. This figure shows very clearly that the link margin obtainable using the EGC is between 0 and ∼2 dB (depending on the turbulence severity) lower than using the complex MRC. Using two photodetectors with optimal MRC in atmospheric turbulence with 0.22 ≤ σ l2 ≤ 1 has the potential to reduce the SNR required to achieve a BER of 10−6 by between ∼2 and ∼12 dB. With up to four independent photodetectors, however, the theoretical link margin for the MRC combiner increases to ∼20 dB as shown in Figure 7.11. Another inference from this figure is that the spatial diversity gain (link margin) becomes more pronounced as scintillation increases; using two detectors with MRC at turbulence level, σ l = 0.2 results in a link margin which is ∼10 dB lower than at σ l = 1. Also for N ≥ 4 , the marginal link margin per unit detector (mN ,σ l − mN −1,σ l ) reduces drastically as the graphs begin to flatten out. For instance, increasing N from 4 to 10 with MRC across the turbulence levels 0.22 ≤ σ l2 ≤ 1 only results in a meagre increase of between 0 and ∼6 dB link margins, while increasing N from 1 to 4 over the same turbulence range results in between ∼3 and ∼ 22 dB diversity gains. The plot of the diversity gain (7.44) at a POut of 10 −6 is shown in Figure 7.12 for a different number of photodetectors and log intensity variance. This plot illustrates the gain based on the outage probability metric. From the plot, using two photodetectors results in ∼4 dB gain at σl2 = 0.72, and

367

Outdoor OWC Links with Diversity Techniques 30

σ 2l = 0.2 2 σ 2l = 0.5 2

Link margin at BER = 10-6 (dB)

25

σ 2l = 0.7 2 σ 2l = 1 EGC MRC

20

15

10

5

0

1

2

3

4

5 6 7 No of photodetectors

8

9

10

FIGURE 7.11  BPSK-SIM diversity link margin with EGC and MRC against number of photodetectors for various turbulence levels and a BER of 10 −6. 16

σ 2l = 0.2 2 σ 2l = 0.5 2

14

σ 2l = 0.7 2 EGC Gain at POut = 10-6 (dB)

12

σ 2l = 1

10

8

6

4

2

0

1

2

3

4

5

6

7

8

9

10

Number of photodetectors

FIGURE 7.12  EGC diversity gain in log-normal atmospheric channel against the number of photodetectors at POut of 10 −6 and σ l2 =  0.2 2 , 0.5 2 , 0.7 2 , 1.

368

Optical Wireless Communications ρ=0 ρ = 0.1 ρ = 0.3 ρ = 0.6 -4

BER

10

3 Photodetectors 2 Photodetectors

-5

10

-6

10

19

20

21

22

23

24

25

26

27

28

Normalised SNR (dB)

FIGURE 7.13  Error performance of BPSK-SIM at different values of the correlation coefficient for N = [ 2, 3] and σ l2 = 0.5 2.

this rises to ∼8 dB with N = 4 . Beyond four photodetectors, the graphs start to plateau, implying a reduction in the amount of gain recorded for each additional photodetector. To illustrate the impact of signal correlation on the error performance, the combination of (7.39), (7.40), and (7.41) is plotted in Figure 7.13 for N = [ 2,  3], ρ = [0, 0.1, 0.3, 0.6], and σ l2 = 0.52. For instance, at a BER of 10 −6, the use of two photodetectors with correlation coefficients ρ = 0.3 and 0.6 requires additional ∼1.6 dB and ∼3 dB of SNR, respectively compared to when ρ = 0. With three photodetectors, the additional SNR required to achieve the same BER of 10 −6 is ∼2.4 dB and ∼4.6 dB for ρ = 0.3 and 0.6, respectively. This shows that the correlation effect results in a power penalty, and this result, by extension, buttresses the need for the photodetector separation to be greater than the spatial coherence length ρ0, in order to get the most from the spatial diversity technique. Furthermore, in Figure 7.14, the plot of (7.49) against the normalised SNR is shown at a turbulence level σ l2 = 0.3 for different values of N and M . It can be inferred from the plot that at a BER of 10 −6, using a 2 × 2 MIMO requires ∼0.4 dB of SNR more than employing a 1 × 4-MIMO configuration. However, spacing four photodetectors to ensure that the received signals are uncorrelated is far more demanding and cumbersome than spacing two photodetectors. Also, to achieve a BER of 10 −6, the use of a 4 × 4-MIMO system requires about 4 dB and 1 dB less SNR compared with using a lone source with 4 and 8 photodetectors, respectively.

7.6 SIM-FSO WITH RECEIVER DIVERSITY IN GAMMA-GAMMA AND NEGATIVE EXPONENTIAL ATMOSPHERIC CHANNELS In this section, the performance of SIM-FSO will be analysed in terms of BER and the outage probability based on the gamma-gamma and negative exponential turbulence models. While the results of the previous sections are valid for short FSO links characterised by weak irradiance fluctuation only, the performance analysis of this section will capture the performance of short to very long FSO links.

369

Outdoor OWC Links with Diversity Techniques MIMO configuration

-3

10

4X4 2X2 1X4 1X8 1X5 1X1

-4

10

-5

BER

10

-6

10

-7

10

-8

10

10

15

20

25

30

Normalised SNR (dB)

FIGURE 7.14  Error performance of BPSK-SIM with MIMO configuration in the turbulent atmospheric channel for σ l2 = 0.3.

7.6.1 BER and Outage Probability of BPSK-SIM with Spatial Diversity The general expression for the unconditional BER of an FSO employing BPSK-SIM with an array of photodetectors can be written as    ∫ Q ( γ ( I ) ) p ( I ) dI (7.50) ∞



Pe =

0

 where γ I represents the post-detection electrical SNR at the BPSK demodulator input and   p I = ∏ iN=1 p ( I i ) is the joint pdf of the uncorrelated irradiance. In evaluating γ I , the EGC linear combining technique is considered because of its simplicity, but other linear combining schemes can be used as well. The total noise variance is given by σ 2EGC = N σ 2Th + σ 2Bg , using a narrowband optical band-pass filter combined with the narrow field of view detectors, the background noise can be 2 2 2 2 reduced considerably to  make the thermal noise dominant. Therefore, σ Bg < Nσ Th and σ EGC ≈ N σ Th. The post-detection γ I for the thermal noise–limited performance at the BPSK demodulator input is thus obtained as

()

()

()

()



 R 2 A2  γ I =  2 N 3σ n 2 

()

N

2

 I i  (7.51)  i =1

∑

The system performance analysis with spatial diversity in weak, moderate, and strong turbulence regimes will be based on the gamma-gamma model introduced in Chapter 4. The gamma-gamma parameters representing each of the stated turbulence regimes are as previously presented in

370

Optical Wireless Communications

Table 6.3. From the numerical solution of (7.50), the BER can thus be obtained. In the limit of strong turbulence, however, the BER expression based on the simple EGC linear combining can be reduced to single integration as follows. Let the random variable Z represent the sum of N independent negative exponential variables, that is, Z = ∑ iN=1 I i . Then the pdf of Z is obtained by adopting the characteristic function method as I 0− N Z N −1 exp ( − Z / I 0 ) ,  Z ≥ 0 (7.52) Γ(N )

p( Z ) =



The unconditional BER with receiver spatial diversity in a negative exponential turbulent atmospheric channel is now derived as

Pe =



1 πΓ ( N

π /2 ∞

) ∫0

∫Z

N −1

(

)

exp − K1 ( θ ) Z 2 − Z dZ dθ (7.53)

0

where K1 ( θ ) = ( RA ) / 4 N 3σ 2 sin 2 ( θ ). Expression (7.53) has no closed form, but the multiple integral involved can be eliminated by invoking equation (3.462) reported in [65]. This leads to the following equation for the BER of BPSK-SIM in a negative exponential fading channel: 2

1 Pe = π



π/2

∫ 0

1

( 2 K1 ( θ ) ) N

(

)

exp (1 / 8 K1 ( θ )) D− N 1 / 2 K1 ( θ ) dθ (7.54)

Since (7.54) cannot be further simplified, its upper bound is obtained by maximising the integrand as Pe ≤



1 2

( 2 K1 ) N

(

)

exp (1 / 8 K1 ) D− N 1 / 2 K1 (7.55)

where K1 = ( RA ) / 4 N 3σ n 2 and Dρ is the parabolic cylinder function whose definition is available in [65]. The outage probability in the fully developed speckle regime is given by the following: 2



m  POut = p  EGC   N 

  Ii  < I 0   (7.56)  i =1 N

∑

= p ( Z < N I 0 / mEGC ) Expression (7.56) is the cumulative distribution function of Z at the specified point, and by combining this with (7.52), the outage probability is obtained as N / mEGC

POut =

∫

p ( Z ) dZ

0



  (7.57)

N / mEGC

=

∫ 0

Z

N −1

exp ( − Z ) / Γ ( N ) dZ

371

Outdoor OWC Links with Diversity Techniques Turbulence regime

-1

10

Saturation (upperbound) Saturation (exact) Weak Moderate Strong

-2

10

-3

BER

10

-4

10

-5

10

-6

10

-7

10

15

20

25

30

35

40

45

50

55

60

Normalised SNR (dB)

FIGURE 7.15  BPSK-SIM error rate against the normalised SNR in gamma-gamma and negative exponential channels for two photodetectors.

Equation (7.57) can be expressed as an implicit function in terms of the incomplete gamma function γ(a, x) [65] or as a series as POut =

γ ( N , N / mEGC ) Γ(N )

1 = Γ(N

∞

)∑ = 0

( −1) ( N / mEGC )

N +

(7.58)

! ( N +  )

where the incomplete gamma function is defined as γ ( a, x ) = ∫ x0 exp ( −t ) t a −1dt . In Figure 7.15, the BER of BPSK-SIM is plotted against the normalised SNR for two photodetectors under different turbulence regimes. When compared to the results obtained with no spatial diversity, the gain of spatial diversity in mitigating the effect of turbulence-induced irradiance fluctuation becomes very clear. A summary of the resulting diversity gains (i.e., reduction in the SNR) with two and three photodetectors at a BER of 10−6 is presented in Table 7.1. As expected, the impact of diversity is least TABLE 7.1 Diversity Gain at a BER of 10 −6 in Gamma-Gamma and Negative Exponential Channels Spatial diversity gain (dB) Number of photodetectors

Weak

Strong

Saturation

2 3

2.6 3.0

11.6 21.2

46.7 63.6

372

Optical Wireless Communications

10

Outage Probability

10

10

10

10

10

10

=1 =2 =4

-1

-2

-3

-4

-5

-6

-7

30

40

50

60

70

80

90

100

Power margin, mEGC (dBm)

FIGURE 7.16  The outage probability as a function of power margin m EGC (dBm) for N = [1, 2, 4 ] in the negative exponential channel.

in weak turbulence regimes, resulting in only 2.6 dB reductions in the SNR at a BER of 10−6. In strong turbulence regimes, however, the gain in SNR is ∼12 dB, increasing to ∼47 dB in the saturation regime. As would be expected, the diversity gain is highest in extreme fading conditions since adding more branches will greatly reduce the chance of a catastrophic fading from happening. Figure 7.16 shows the extra power margin required to achieve a certain outage probability POut with and without spatial diversity in negative exponential atmospheric channels. The gain of employing diversity is apparent from the figure just as the case under the BER performance metric. For instance, at a POut of 10 −6, a diversity gain of almost 30 dB is predicted, and this increases to ∼43 dB with four photodetectors. This significant gain is due to the fact that with multiple photodetectors, more independent irradiances are received. For the sake of comparison, the predicted spatial diversity gains based on both metrics (BER and outage probability) are shown in Figure 7.17 as a function of the number of photodetectors. For both metrics, it is observed that as the number of photodetectors increases beyond four, the diversity gains start to plateau. Although POut and BER are different by definition, they both result in a similar conclusion. It should, however, be mentioned that using an array of photodetectors adds to the cost and design complexity.

7.6.2 BER and Outage Probability of DPSK-SIM in Negative Exponential Channels For a DPSK-premodulated subcarrier with an array of photodetectors, the most suitable linear combining scheme will be the SelC. This is because in DPSK demodulation, the absolute phase information of the subcarrier is not available, and it is only the SelC that does not require the subcarrier phase information. The SNR for a SIM-FSO with SelC is given by

373

Outdoor OWC Links with Diversity Techniques 100 BER Outage

90 80

Diversity gain (dB)

70 60 50 40 30 20 10 0

2

3

4

6 No of Photodetectors

8

10

FIGURE 7.17  Diversity gain against number of independent photodetectors at BER and POut of 10 −6 in the negative exponential atmospheric channel.

(7.46), while the unconditional BER for a DPSK-SIM in a negative exponential channel is given by (7.47): γ SelC ( I ) =



∞

Pe (SelC) =



R 2 A2 I 2 (7.59) 2 N N σ 2Th + σ 2Bg

1

)

 γ SelC ( I )  p( I max )dI (7.60) 2 

∫ 2 exp  − 0

(

(

)

The pdf p( I max ) = p max { I i }i =1 in a negative exponential channel model given by (7.61) is obtained by first finding the cumulative distribution function of I max at an arbitrary point and then differentiating. The resulting pdf is given by (7.61), and the detailed proof is presented in Appendix C. N

p ( I max ) =



N  I   I  exp  −   1 − exp  −    I0    I0   I0

N −1

(7.61)

The plot of the pdf of Imax is shown in Figure 7.18 for different values of N . From the combination of (7.59), (7.60), and (7.61), an expression for the error performance of a DPSK-SIM laser communication system in a fully developed speckle atmospheric channel is obtained as ∞



Pe (SelC) =

∫ 0

N 2I 0

  I   1 − exp  − I 0  

N −1

 I  R 2 A2 I 2 exp  − −  dI (7.62) 2 2  I 0 2 N N σ Th + σ Bg 

(

)

374

Optical Wireless Communications 1 0.9 =1

0.8 0.7

pdf of Imax

0.6 0.5 =4 0.4 = 10 0.3 0.2 0.1 0

0

1

2

3

4

5

6

7

8

9

I

(

)

FIGURE 7.18  The pdf, p  max { I i }i =1 for N = [1, 4, 10 ], and I0 = 1 in a negative exponential channel. N

(

)

Now the received irradiance I0 needed to attain an outage probability POut = P γ ( I ) < γ * in a negative exponential atmospheric channel without diversity is given by (7.62). In arriving at this 2 equation, the threshold SNR has been taken as γ * = RAI * / 2σ n 2, where I * is the receiver sensitivity required to attain the threshold BER*.

(

I0 =



)

I* −1 (7.63) ln (1 − POut )

With an array of PIN photodetectors employing SelC in a negative exponential atmospheric channel, the received irradiance I0SelC needed to attain a given POut is derived from the combination of (7.59) and (7.61) as



I 0SelC =

(

I*

ln 1 − N POut

)

(

)

 N N σ 2Th + σ 2Bg   −1  σ 2Th + σ 2Bg  

1/ 2

(7.64)

From the foregoing, the diversity gain I0/I0SelC can thus be obtained. The numerical simulations presented in this section are based on the parameters of Table 7.2. In Figure 7.19, the error performance of DPSK-SIM in a negative exponential channel, obtained from (7.62), is shown as a function of the average irradiance. This plot brings to bear the potential gain of SelC in reducing the required sensitivity for a given BER under very strong fading conditions. For example, to achieve a BER of 10 −6 with no diversity, about 23 dBm of received irradiance is required while with two photodetectors, about −1.7 dBm is needed to achieve the same level of performance. Moreover, as the number of photodetectors increases, the attained diversity gain per additional

375

Outdoor OWC Links with Diversity Techniques

TABLE 7.2 Numerical Simulation Parameters Parameter

Value

Symbol rate Rb Spectral radiance of the sky N(λ) Spectral radiant emittance of the sun W(λ) Optical band-pass filter bandwidth Δλ @ λ = 850 nm PIN photodetector field of view FOV Radiation wavelength λ

155 – 625 Mbps 10−3 W/cm2µmSr 0.055 W/cm2µm 1 nm 0.6 rad 850 nm

Number of photodetectors N Load resistance RL

1 ≤ N ≤ 10 50 Ω 1 300 K

PIN photodetector responsivity R Operating temperature Te

detector reduces. For instance, for N = 2 , the gain per detector at a BER of 10 −6 is ∼12 dB, and this reduces to about 5 and 4 dB for N = 8 and 10, respectively. These results are summarised in Table 7.3 for up to 10 photodetectors. In discussing the outage probability in negative exponential fading channels, (7.64) is plotted in Figure 7.20. In this plot, the threshold average irradiance I* is assumed to be 0 dBm. This graph shows that for a very long SIM-FSO link, whose channel fading is modelled by the negative exponential distribution, achieving an outage probability of 10 −6 or better will require a minimum of 60 dBm received irradiance without SelC.

-2

10

-3

10

-4

BER

10

-5

10

-6

10

-7

10

=2 =4 =6 =8 = 10 =1 -20

-15 -10 -5 Average received irradiance, E[I] (dBm)

0

FIGURE 7.19  Error rate of DPSK-SIM against the average received irradiance with spatial diversity in negative exponential channel for N = [ 2, 4, 6, 8, 10 ].

376

Optical Wireless Communications

TABLE 7.3 Gain per Photodetector at a BER of 10 −6 N Average Irradiance (dBm) Gain (dB) per N

1

2

4

6

8

10

23.1 0

−1.7 12.4

−12.5 8.9

−15.2 6.4

−16.2 4.9

−16.7 4.0

This power requirement reduces to ∼35 dBm and ∼23 dBm, respectively, with 2 and 4 photodetectors that are combined using the SelC. To further illustrate the gain of using SelC in the saturation regime, Figure 7.21 and Figure 7.22 show the predicted diversity gain at different values of POut and N . With two photodetectors and an outage probability of 10 −6, the maximum predicted gain per detector is about 14 dB. This predicted gain is even observed to be higher at lower values of Pout. This makes sense because the use of diversity in a fading channel increases the received signal strength and, by extension, lowers the outage probability. And in Figure 7.22, it is clearly shown that the gain (dB) per detector peaks at N = 2 and then decreases very rapidly beyond N = 4 . Although up to 10 photodetectors have been shown in the results above, this is mainly for illustration purposes. The use of such a large number of detectors will pose serious implementation difficulties because they all have to be spaced beyond the spatial coherence distance in order to avoid any signal correlation. An interesting point to note from these results is that, in contrast to the shortrange links where the use of SelC is not worthwhile, it is highly recommended in very long-range links, as it results in a significant reduction in the required receiver sensitivity, especially when the photodetector is kept to a maximum of four. This is due to the fact that the irradiance fading

=1 =2 =4 =6 =8 = 10

-2

Outage probability, POut

10

-4

10

-6

10

-8

10

10

20

30 40 50 60 70 80 Required average irradiance, I0SelC (dBm)

90

FIGURE 7.20  Outage probability against the average irradiance with SelC spatial diversity in negative exponential channel for I* = 0 dBm and N = [1, 2, 4, 6, 10 ] .

377

Outdoor OWC Links with Diversity Techniques 25

= = = = = =

SelC Diversity gain (dB) per photodetector

20

1 2 4 6 8 10

15

10

5

0

-5 -10 10

-8

-6

10

-4

-2

10 10 Outage probability, POut

0

10

10

FIGURE 7.21  Predicted SelC diversity gain per photodetector against POut in saturation regime for N = [1, 2, 4, 6, 10 ].

1

SelC Gain (dB) per photodetector

10

POut = 10 -2 POut = 10 -3 POut = 10 -6

0

10

2

3

4

5 6 7 Number of photodetectors

8

9

10

FIGURE 7.22  Predicted SelC diversity gain (dB) per photodetector for POut = [10 −6, 10 −3, 10 −2] in saturation regime.

378

Optical Wireless Communications

experienced in a fully developed speckle regime is dominant over the reduction in the received irradiance due to the reduction in the receiver aperture area. Any scheme that mitigates this dominant irradiance fading will clearly result in improved performance. The predicted average irradiance and diversity gains presented above are valid for as long as the photodetectors’ received signals are uncorrelated. That is, ρ0 < s < θs L where θs is the divergence angle of the optical source in milliradians and L is the link length in kilometers. For s < ρ0, the received irradiances are correlated, and the diversity gain is lower as previously discussed.

7.7 TERRESTRIAL FREE-SPACE OPTICAL LINKS WITH SUBCARRIER TIME DIVERSITY In the previous sections, the use of spatial diversity to lessen the effect of atmospheric turbulence has been discussed. And as highlighted in the spatial diversity results, the best gains are achieved when the detectors are physically separated by a distance greater than the turbulence coherence length. The channel coherence length depends on the turbulence strength and is typically in the order of centimeters. Similarly, the use of aperture averaging requires the receiver aperture to be larger than the turbulence coherence length. This not only makes the system bulky, it is also cumbersome and not always feasible. The use of the turbo product code (TPC) as the channel coding scheme with interleaving has been shown to offer good resistance to burst errors and no error floor due to channel fading in FSO links [67]. This section will examine the subcarrier time delay diversity (STDD) as an alternative or complementary means of mitigating channel fading in SIM-FSO links. The conventional use of multiple subcarriers is to increase throughput/capacity via subcarrier multiplexing. But in this scheme, different subcarriers at different frequencies are used to transmit the delayed copies of the original data. The proposed subcarrier STDD scheme has the advantage of simplicity and low cost for achieving a reasonable diversity gain compared to schemes such as adaptive optics or forward error correction. Moreover, the reduction in throughput associated with temporal diversity can be compensated for through subcarrier multiplexing.

7.7.1 Error Performance with STDD In the STDD scheme, delayed copies of the data are retransmitted on different subcarriers, as shown in Figure 7.19. The STDD scheme relies solely on the statistical temporal variation of the atmospheric turbulence–induced fading. Apart from retransmitting the delayed version of the original data on different subcarriers, other viable options include using different wavelengths or polarisations for the retransmission. Details of how these two other options are done can be found in [70]–[72]. Time delay diversity schemes have the advantage of simplicity and low cost for achieving a reasonable diversity gain compared to schemes such as adaptive optics or forward error correction [70], [71]. These gains are, however, at the cost of retransmission latency and data rate reduction. The system under consideration must therefore be able to trade a low error rate for high latency. According to the Taylor frozen turbulence hypothesis [70], turbulent eddies responsible for the fading are frozen in space and only move across the beam path by the transverse component of the wind. Thus, from the knowledge of the average transverse wind speed, the spatial statistics of turbulence can be translated into temporal statistics. From this transformation, the temporal covariance function of the irradiance fluctuations can be obtained as outlined in [72]. The temporal covariance is useful in determining the correlation time τc. For weak turbulence, τc has been experimentally measured to vary between 3 ms and 15 ms [72]. This suggests the sorts of values required for τ in order for the TDD systems to be efficient in fading channels. In the analysis, we assume the following: τ ≥ τc, and the minimum required buffer size Rbτ is set prior to transmission. To illustrate the performance with the proposed STDD, we consider a BPSK-subcarrier intensity–modulated

379

Outdoor OWC Links with Diversity Techniques

system with a single data-carrying subcarrier and U-STDD paths. During a symbol duration, the data-carrying signal m(t) is given by m (t ) =



U +1

∑d (t − (i − 1) τ) cos(ω t + ϕ ) (7.65) i

i

i =1

where d (.) ∈[ −1,  1]. The standard RF coherent demodulator employed extracts the reference carrier needed to down-convert the received signal to baseband. The sum of the demodulator outputs is then fed into the decision circuit as illustrated in Figure 7.23. From this, the electrical SNR per bit can then be easily derived as γ (I )

d(t)

( Rξ )2 Pm  U +1 I  = σ n2

 

∑ i =1

i

2

 (7.66) 

s1(t)

τ

ω c1

s2(t)

τ

m’(t)

m(t)

ωc2

s3(t)

Optical driver circuit

Light source x(t)

Turbulent atmospheric channel

DC bias

ωc3 . . .

τ

(bo)

s(U+1)(t) (a)

ωc(U + 1)

r(U+1)(t)

ω c (U + 1) O B P F

Photodetector

τ

r3(t) Receiver (TIA)

y(t)

ω c3

τ

Decision circuit

dˆ (t )

r2(t)

ωc2

τ r1(t)

ω c1 (b)

FIGURE 7.23  The subcarrier STDD block diagram: (a) transmitter and (b) receiver. TIA-trans-impedance amplifier; OBPF-optical bandpass filter.

380

Optical Wireless Communications

The noise variance is given by σ n 2 = σ 2Bg + (U + 1) σ 2Th , since there are now (U + 1) demodulation paths associated with every photodetector. It should be noted that with the addition of U-STDD paths, ξ = 1 / (U + 1) A becomes the minimum modulation depth needed to avoid any clipping distortions.

7.7.2 Error Performance of Short-Range Links To evaluate the BER with the subcarrier STDD in a log-normal turbulence channel, expression (7.66) is substituted into (6.8) of Chapter 6 and p ( I ) replaced with the probability density function (pdf) of the sum of (U + 1) independent but identical log-normal variables. Although the sum of independent log-normal variables does not have a closed form [21], it is often approximated as another log-normal variable [65], [66]. There are several approaches to making this approximation; a survey and comparison of these approaches can be found in [73]. In this work, we will use the moment-matching approach (otherwise called the Wilkinson’s method) to make the approximation. This approach is chosen because it is simple, straightforward, and has also been reported [58] to work well for small values of σ l2 , as is the case here. We now make the approximation that the ∑Ui =+11 I i ≡ Z , where Z is a log-normal variable described as Z = exp ( ν ) and ν is Gaussian. The following are the approximate values of the mean and variance of ν obtained by matching the first and second moments of ∑Ui =+11 I i to that of Z ; see [74] for detailed proof.

( )

exp σ l2 − 1  1  µ ν = ln (U + 1) − ln  1 +  2  U +1 



( )

 exp σ l2 − 1  σ 2ν = ln  1 +  U +1  

(7.67)

The resulting BER expression is then simplified using the Gauss-Hermite approximation. The unconditional BER for the BSPK-SIM link with U-STDD thus becomes Pe =

where K 0′ =

( RξI 0 )2 Pm

σ 2Bg + (U +1)σ 2Th

1 π

n

∑w Q ( i

(

K 0′ exp xi 2σ υ + µ υ

i =1

)) (7.68)

. The penalty due to the turbulence fading at a given BER with and without

STDD can therefore be obtained from this equation and the ones in Chapter 6.

7.7.3 Long-Range Links For longer-range FSO links (>1 km), the procedure for obtaining the error rate expression is similar to that of the short-range link above except that p ( I ) now becomes the pdf of the sum of independent negative exponential variables. This pdf is tractable and is easily derived as [69]

p(I = Z ) =

I 0−(U +1) Z U exp ( − Z / I 0 ) (7.69) Γ (U + 1)

where Γ(.) is the gamma function. The resulting BER expression is given by (7.70), which is obtained by combining (6.29a) with the Kolmogorov two-thirds power law (7.66) and (7.69); it does not have a closed-form solution. As a result, the BER will have to be evaluated numerically.

I 0−(U +1) Pe = πΓ (U + 1)

π/2 ∞

∫ ∫Z 0

0

U

(

)

exp − K1′ ( θ ) Z 2 − Z / I 0 dZ dθ (7.70)

381

Outdoor OWC Links with Diversity Techniques

TABLE 7.4 Simulation Parameters Parameter

Value

Symbol rate Rb Spectral radiance of the sky N(λ) Spectral radiant emittance of the sun W(λ) Optical band-pass filter bandwidth Δλ @ λ = 850 nm PIN photodetector field of view Radiation wavelength λ Link range L Index of refraction structure parameter Cn2 Load resistance RL PIN photodetector responsivity R Optical band-pass filter transmissivity Equivalent temperature, Te

155–625 Mbps 10−3 W/cm2µmSr 0.055 W/cm2µm 1 nm 0.6 radian 850 nm 1 km 0.25 × 10−14–2.45 × 10−14 m−2/3 50 Ω 1 90% 300 K

where K1′ ( θ ) = ( Rξ ) Pm / 2σ 2 sin 2 θ. From (7.70) it is observed that within the ( 0 − π / 2 ) limit of integration, the function sin 2 θ is monotonically increasing, and an upper bound of (7.57) can then be obtained by maximising the integrand with θ = π / 2 to obtain 2



Pe ≤

1 2

( 2K1 )U +1

  σ n2 exp   D−(U +1) 1 / 2 K1 (7.71) 2  4 ( Rξ ) Pm 

(

)

where Dρ is the parabolic cylinder function whose definition is available in [65]. The results presented in this section are based on the simulation parameters of Table 7.4 and the BPSK-SIM scheme. Unless otherwise stated, the background, thermal noise, and scintillation are considered to be the system limiting factors.

7.7.4 Short-Range Link From the combination of (6.7), (6.12), and (7.68), we obtain Figure 7.24, which shows the FSO link’s RI 2 P BER with and without STDD at different values of normalised SNR, γ = ( 0σ)2 m . This plot shows γ to be decreasing as the number of temporal diversity paths increases. To determine the optimum number of diversity paths required, we present in Table 7.5 the following parameters: diversity gain (difference between γ with and without STDD), fading penalties (difference between γ in

TABLE 7.5 Fading Penalty and STDD Gain at BER = 10−9, Rb = 155 Mbps, σl2 = 0.3, and ξ = 1 / (U + 1) A γ (dB) (no fading: 15.56 dB) Fading penalty (dB) Diversity gain (dB) (gain/unit path) Effective data rate (Mbps)

0-STDD

1-STDD

2-STDD

3-STDD

5-STDD

35.32 19.76 0 (0) Rb

28.41 12.85 6.91 (6.91) 0.5Rb

25.34 9.78 9.89 (4.99) 0.33Rb

23.54 7.98 11.78 (3.93) 0.25Rb

21.48 5.92 13.84 (2.77) 0.17Rb

382

Optical Wireless Communications No fading 5-STDD 3-STDD 2-STDD 1-STDD No-STDD

-2

10

-4

BER

10

σ 2l = 0.3

-6

10

-8

10

-10

10

10

15

20 25 Normalised SNR, γ- (dB)

30

35

FIGURE 7.24  The BER against γ with and without STDD at 155 Mbps, σl2 = 0.3, Cn = 0.75 × 10 −14 m −2 /3.

fading channel and under no fading), and the effective data rates. This table explicitly shows that at σ l2 = 0.3, 1-STDD has the highest STDD gain per additional path of ∼6.9 dB and the least reduction in data rate. It is interesting to note that the gain per unit path decreases as more paths are added. This makes sense because for a given peak optical power, the modulation index has to be reduced as more paths are added in order to keep the laser within its dynamic range. These findings appear to be independent of how the STDD scheme is implemented. A similar conclusion was previously reported when the delayed data were transmitted on different polarisations/wavelengths [70], [72]. Hence, single re-transmission (1-STDD) will be suggested to mitigate channel fading for shortrange links. Using (7.55) with ξ = 1 / (U + 1) A, it is shown in Figure 7.25 that the average received irradiance and indeed the diversity gain both increase as the fading strength increases. However, the fading penalties and diversity gains are independent of the data rate. For instance, at σl2 = 0.1, the estimated gains and fading penalties with 1-STDD stood at 0.36 dB and ∼4.2 dB for Rb = 155 and 625 Mbps, respectively. For stronger fading conditions with σl2 = 0.5, these values increase to 3 dB and ∼11 dB, respectively.

7.7.5 Long-Range Link The performance of the link in the negative exponential turbulence channel is depicted in Figure 7.26. From the figure it is quite clear that achieving a low-error-rate communication in a long-range FSO link is almost impossible without any diversity technique. To achieve a moderate BER of 10 −6 requires a huge γ of 112 dB at the receiver if no diversity technique is implemented. And this translates into ∼97 dB fading penalty. The implication of this high fading penalty is that at least one fading mitigation technique has to be used in order to establish the link. However, with 1-, 2-, and 3-STDD, the required γ values are estimated to be 62 dB, 43.6 dB, and 35.8 dB, respectively. These gains (reduction in γ with STDD) are achieved at the price of reduced Rb by a factor of 2, 3, and 4, in that order. The STDD technique can therefore be used for the long-range FSO to achieve a low-BER communication.

383

Outdoor OWC Links with Diversity Techniques 10 No STDD

Average irradiance 6 1-STDD

-15

155 Mbps 4

-20

1-STDD gain

625 Mbps

2

-25

-30

1-STTD gain (dB)

8

-10

at BER =10-9 (dBm)

Av erage irradianc e

-5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0

σ 2l

FIGURE 7.25  The average received irradiance at a BER of 10 −9, Rb (Mbps) = [155, 625] and 1-STDD gain for different strengths of turbulence.

1-STDD 2-STDD 3-STDD 5-STDD 0-STDD Exact 0-STDD Upper bound

-2

10

-3

BER

10

-4

10

-5

10

-6

10

-7

10

20

40

60 80 Normalised SNR, γ - (dB)

100

120

FIGURE 7.26  Link BER performance against the normalised SNR with the subcarrier STDD in the negative exponential turbulence–induced fading channel.

384

Optical Wireless Communications

7.8  APERTURE AVERAGING This is one of the simplest forms of spatial diversity, where the aperture of the receiver lens is larger than the fading correlation length [75], [76]. Aperture averaging has been considered to combat scintillation, in particular strong scintillation, by a number of researchers [75]–[77]. With this scheme, averaging is carried out over the relatively fast fluctuations that are due to small-sized eddies; by doing so, the frequency content of the irradiance spectrum is shifted to the low spatial frequencies [20], [75], [76]. The aperture averaging factor, which is widely used to quantify the fading reduction by aperture averaging, is given by [75] A=



σ 2I ( D ) (7.72) σ 2I ( 0 )

where σ 2I ( D ) and σ 2I ( 0 ) denote the scintillation index for a receiver lens of diameter D and a ‘point receiver’ (D ≈ 0), respectively. The strong turbulence is characterised by the way in which the statistical moments of velocity increments increase with the spatial separation Ls. Within the inertial sub-range, the refractive index structure is defined by the Kolmogorov two-thirds power law given as (3.86); for full details see Chapter 3. In the spectral domain, the power spectral density of the refractive index fluctuation is related to Cn2 by Φ n ( K ) = 0.033 Cn2 K −11/3 ;   2 π / L0  K  2 π / l0 (7.73)



7.8.1  Plane Wave For the plane wave propagation with a smaller inner scale lo THz)

Unlimited, 400–700 nm Medium to high (few Mbps to >10 Gbps)

Short

Unlimited, 800–1600 nm High (>10 Gbps)

RF (Wi-Fi) 2.4 GHz and 5–5.9 GHz for WiFi Regulated and limited Medium (150 Mbps for Wi-Fi)

High No

Low Yes

Yes

No Short to long (outdoor)

Short to long (outdoor) High

Excellent—optical radiation is very well confined and does not penetrate walls. High No (indoor environment is highly static with both LOS and non-LOS paths) Poor (incoherent detection) High degree of control with lenses and diffractive elements Medium In progress (IEEE Well developed for indoor 802.15.7 Task Group) (IrDa). In progress for outdoor Illumination, Communications, sensing communications, indoor localisation, sensing Sunlight + other ambient Sunlight + other ambient lights lights Relatively low Limited (indoor) Narrow and wide

Poor High Yes Excellent (coherent detection) Difficulty to constrain on antenna size Low Matured (IEEE 802.11 for WiFi) Communications, localisation (outdoor) All electrical/electronic appliances Medium Good Mostly wide

TABLE 8.3 Short-Range Wireless Technologies and Standards Technology

Speed

Data density

Wi-Fi (IEEE 802.11N) Bluetooth IrDa

Wireless—Current 150 Mbps 3 Mbps 4 Mbps

* * ***

Wi-Gig (IEEE 802.11ad) White Wi-Fi (IEEE 802.11af and IEEE 802.11ah, Giga-IR VLC (IEEE 802.15.7)

Wireless—Future 2 Gbps @ 60 GHz; 10 m within a room 24 Mbps @54 and 790, 900 MHz 1 Gbps >10 Gbps; a few meters within a room

** * (a large area, i.e., a few km) *** ****

Visible Light Communications

405

FIGURE 8.3  Key features of VLC.

IEEE 802.15.7r1 standard was released in 2011, and efforts are currently ongoing to include VLC in the IEEE802.11 group of standards.

8.3  SYSTEM DESCRIPTION It is expected that a VLC should be compatible with light dimming. Most of the research aimed at tackling the dimming problem has focused on the modulating the optical source driving current. Pulse width modulation (PWM) is the standard way to control the average optical power output; by controlling the duty cycle so that the required optical power level can easily be reached. The PWM signal does not carry any information, though, and modulation schemes, such as OOK, PPM, and discrete multi-tone (DMT) have been added to the PWM signal to maintain information transmission at various illumination levels [88]–[91]. Figure 8.4 shows a block diagram of a VLC link. Precise dimming appears to be challenging for incandescent and gas-discharge lamps. Whereas with LEDs, it is quite convenient to accurately control the dimming level and ensure communications even under low dimming level. This is because the LED response time during on-off switch operation is very short (a few tens of nanoseconds). Therefore, by modulating the driver current at a relatively high frequency, it is thus possible to switch LEDs on and off without it being perceivable by the human eyes. Thus, the light emitted from an LED is in the form of a repetitive high frequency and a low average power pulse stream. The average luminous flux emitted by an LED is linearly proportional to the relative width of the dimming signal. Depending upon the application and safety requirements, the transmitter can be an LED or an LD. The LED is preferred over LD if the application is for the multiple propose of illumination, data communication, indoor localisation, and sense, as is the case in a VLC-based system. As for the light source, WPLEDS are the most widely used due to their lower complexity and cost. However, the slow response of phosphor limits the modulation bandwidth of the WPLEDS to a few MHz [92]. A typical VLC system with a WPLED is shown in Figure 8.5(a). The blue

406

Optical Wireless Communications

FIGURE 8.4  A block diagram of a VLC system.

(a)

(b)

(c)

FIGURE 8.5  (a) VLC link, (b) LED optical spectrum of Osram Ostar white-light LED, and (c) modulation bandwidth with and without blue filtering.

Visible Light Communications

407

FIGURE 8.6  A typical WPLED power–current response.

light could be extracted from the received optical beam by using an optical filter at the receiver. Figures 8.5(b) and (c) show the optical spectrum of the emitted white light and the measured modulation bandwidth of Osram Ostar WPLED, respectively. For this type of LED, the 3 dB cut-off frequency is ∼2.5 MHz at a drive current of 300 mA compared to ∼20 MHz for the blue-only response; see Figure 8.5(c). The small signal modulation bandwidth of the WPLED depends on the LED drive current as shown in Figure 8.6 [93]. The types of light sources that could be used in VLC systems with achievable data rates are shown in Figure 8.7.

FIGURE 8.7  Type of light sources for VLC systems.

408

Optical Wireless Communications

8.3.1 VLC System Model Since LEDs could be used for multiple proposes of illumination, data communication, indoor localisation, and sensing, it is necessary to define the luminous intensity and transmit optical power. The luminous intensity is used for expressing the brightness of an LED and is defined as the luminous flux per solid angle; it is given as I=



dΦ (8.1) dΩ

where Ω is the spatial angle and Φ is the luminous flux that can be calculated from the energy flux Φ e as given by [94] 780



∫

Φ = K m V ( λ ) Φ e ( λ ) dλ (8.2) 380

where V(λ) is the standard luminosity curve, and K m is the maximum visibility, which is ∼683 lm/W at a wavelength of 555 nm. The transmit optical power Pt , which indicates the total energy radiated from an LED, is given as Λ max 2 π



∫ ∫Φ

Pt = K m

e

dθ dλ (8.3)

Λ min 0

where Λ min and Λ max are determined from the PD’s responsivity curve. Figure 8.8 shows a typical office environment with LED-based lighting panels, which provides wireless connectivity to the users. In this scenario the dominate transmission mode is the LOS. Assuming a Lambertian radiation pattern by the LEDs, the radiation intensity at a desk surface is given by

I (∅) = I ( 0 )

ml + 1 cosml  ( ∅ ) (8.4) 2π

VLEDs

d

Rx

FIGURE 8.8  Illumination by LEDs.

409

Visible Light Communications

where ∅  is the angle of irradiance with respect to the axis normal to the transmitter surface, I(0) is the centre luminous intensity, and ml is the order of Lambertian emission, which is a measure of the light beam directivity, defined as ml =



ln ( 2 ) (8.5) ln ( cos  Φ1/ 2 )

where Φ1/2 is the semi-angle at a half illuminance of an LED. For the LOS path, the horizontal illuminance/intensity at a point (x, y, z) and the received power at the receiver are given as I hor = I ( 0 ) cosml  ( ∅ ) / d 2 cos(ψ ) (8.6)



Pr = Pt

( ml + 1) APD cosml   ∅ T ψ g ψ cos(ψ )  0 ≤ ψ   ≤   ψ (8.7) ( ) s( ) ( ) con 2 2πd

where ψ is the angle of incidence with respect to the axis normal to the receiver surface, Ts(ψ ) is the filter transmission coefficient, g(ψ ) and ψ con are the concentrator gain and FOV, respectively, and d is the distance between the LED and a PD surface area APD. The gain of the optical concentrator at the receiver is defined by [95]  n2 con  ,  0 ≤ ψ   ≤   ψ con g(ψ ) =  sin 2 ψ con (8.8)  0, 0 ≥ ψ con 



2 is the refractive index of the optical concentrator. where ncon For the non-LOS or diffuse paths, the light must first hit a reflective surface (in this case the wall or other solid objects within a room) and then arrive at the PD; see Figure 8.8. Therefore, the reflectivity of the surface or what percentage of the incident light is reflected back and captured by the PD(s) must be taken into account. Note that the reflections from the surfaces can also be assumed to have Lambertian patterns [96] and have a finite area dAr . Hence, the illuminance from the diffuse path is defined by [96], [97]



dD = ρ

ml + 1 Icos 2π2

ml

(∅ ) cos (α ) cos (β ) cos ( ψ ) d12 d 22

dAr (8.9)

where d1 and d2 are the distances between an LED and a reflective point, and between a reflective point and a PD, respectively. ρ is the reflectance factor, αir and βir are the angle of irradiance to a reflective point and the angle of irradiance to a receiver, respectively. Thus, for the diffuse channel, the total illuminance is given by integrating over the complete reflective surface area and summing over all LED sources, which is given by

EhorDiffuse = 

∑ ∫ dD (8.10) i

i

wall

where i is the index for the LED. To determine the higher order bounces, a recursive operation can be employed as given by [97]

dDb =

∫

wall

dDk −1 ρcos (β ) cos ( ψ ) (8.11) πd 2

410

Optical Wireless Communications

(a)

(b)

(c)

FIGURE 8.9  (a) LED array and illuminance distributions for (b) one transmitter and (c) four transmitters.

where k is the order of reflections. True reflections contain both specular and diffusive mechanisms [98], however experimental evaluation of materials, such as plaster walls, acoustic-tiled walls, carpets, and unvarnished wooden surfaces, has shown that they can be well approximated as Lambertian reflectors [99]. Note that in most VLC systems with NLOS, often only the first reflection is considered, and higher-order reflections are ignored since their contributions are negligible. [100]. In a single-source VLC system, the use of a wide divergence angle LED is a convenient way to achieve wide coverage. However, wide divergence angles can lead to increased multipath-induced ISI, thus significantly limiting the maximum transmission data rate. In order to overcome this problem and to achieve a higher transmission data rate, as well as a more uniform optical power distribution, the divergence angles of LEDs need to be selected with care. Figure 8.9(a) shows a schematic diagram of a distributed LED array (or multi-source) for indoor applications. Here each single LED can be viewed as a point light source, and therefore the radiation pattern of each LED can be viewed as a function of the solid angle in the three-dimensional space with a well-defined radiation footprint. For each cell shown in Figure 8.9(a), the received optical power is maximum at φ = 0 and minimum at φ = φ Max, which are defined as P r − Max = Pt   P r − Min

( ml + 1) APD T ψ g ψ cos ψ ,  0 ≤ ψ ≤ ψ ( ) s( ) ( ) con 2 2πH

( m + 1) APD cosml φ T ψ g ψ cos ψ ,  0 ≤ ψ ≤ ψ = Pt   l ( Max ) s ( ) ( ) ( ) con 2 2πd Max



(8.12)

411

Visible Light Communications

where φ Max is the maximum irradiance angle and d Max is the corresponding maximum distance between the transmitter and the receiver within a cell. Note that, P Min varies with ml . To increase P Min , which is the minimum received optical power at floor level, it should be maximised as a function of ml, as given by

{

}

  ∂ P  r − Min = Kcos ml ( φ Max ) 1 + ( ml + 1) ln ( cos ( φ Max ) ) (8.13) ∂ml

where the parameter K = PTx   Ts ( ψ ) g ( ψ ) cos ( ψ ) 2 πAdPD 2 is independent of ml. The received optical Max power at the cell edge has its maximum value for ∂ Pr∂−mMin = 0 . Therefore, the optimum Lambertian order is given by ml − opt =



−1 − 1  (8.14) ln ( cos ( φ Max ))

and  H  φ Max = cos−1    (8.15)  d Max 



Thus, the optimum transmitter semi-angle ψ 1/ 2 opt at the half-power point can be calculated, which is given by



ψ 1/ 2 opt

      − ln ( 2 ) = cos −1  exp     0 < ψ 1/ 2 opt < 90° (8.16) −1   − 1   ln ( cos ( φ Max ))    

With all LEDs fully switched on, the illuminance distribution produced on the floor level is called the basic illumination pattern, defined in terms of the solid angle Θ as

f A− LED ( x ,  y;  d ) =

(

f (Θ ) . (8.17) x + y2 + d 2 2

)

In order to determine the total horizontal illuminance at point P(x,y) from the LOS path for a multiarray LED, all contributions from each of the LED lights should be considered (i.e., the summation of all illuminance). Distributed multi-source LED array systems are more of a practical nature for two reasons: (i) most rooms use multiple light sources to ensure sufficient illumination; and (ii) offering spatial diversity, thus avoiding blocking and showing. However, as in diffused OWC links, the multiple sources will suffer from multipath-induced ISI, particularly at higher data rates, which will limit the maximum achievable data transmission in a given indoor environment. There are a number of options to overcome this problem. For example, changing the light source layouts, reducing the receiver field of view, adopting coding and equalisation techniques, as well as employing multiplexing schemes. Table 8.4 shows the key parameters for a typical room environment that could be used in VLC system modelling. Figure 8.9(a) shows the illuminance pattern for a single LED transmitter with a semi-angle at a half power of 70°. The luminous flux maximum value of 568.10 lx is observed at the centre of the room. For four LEDs with a semi-angle of 70°, the illuminance distribution is depicted in Figure 8.9(c), showing a value in the range of 315–910 lx and with an average value of 717 lx.

412

Optical Wireless Communications

TABLE 8.4 System Parameters for a VLC Link Parameter Room Source

Receiver

Optical filter A lens at the PD

Size Reflection coefficient Location (4 LEDs)

Location (1 LEDs) Semi-angle at half power (FWHM) Transmit power (per LED) Number of LEDs per array Center luminous intensity Receive plane above the floor Active area (APD) Half-angle FOV Elevation Azimuth ∆t Gain Refractive index

Value 5 × 5 × 3 m3 0.8 (1.25, 1.25, 3), (1.25, 3.75, 3), (3.75, 1.25, 3), (3.75, 3.75, 3) (2.5, 2.5, 3) 70 20 mW 60 × 60 (3600) 300–910 lx 0.85 m 1 cm2 60° 90° 0° 0.5 ns 1.0 1.5

The optical power distribution for at receiver plane for a LOS path (ignoring the reflection of walls) is shown in Figure 8.10(a). The Matlab code to simulate the optical power distribution is given in Program 8.1. Note that there is an almost uniform distribution of optical power at the centre with the maximum and minimum power levels of 2.3 dBm and −2.3 dBm, respectively. However, depending on the half angle, such a uniform power distribution is not possible to achieve. The optical power distribution at the receiver plane with a half angle of 12.5° is depicted in Figure 8.10. There is more than 35 dB of the optical power difference between the minimum and the maximum power level, thus leading to a high SNR in some areas and dead zones in many areas. In order to

(a)

(b)

FIGURE 8.10  Optical power distribution at the received optical plane for an FWHM of (a) 70° and (b) 12.5°.

Visible Light Communications

413

ensure uniform power distribution within the entire room, a holographic light shaping diffuser (LSD) can be used, which will be discussed in the next section. Program 8.1:  Matlab codes to calculate the optical power distribution of a LOS link at the receiving plane for a typical room. %% opto-electical parameters theta=70; % the semi-angle at half power ml=-log10(2)/log10(cosd(theta)); %Lambertian order of emission P_LED=20; %the tranmist optical power by individual LED nLED=60; % the total number of LED array in each cluster (nLED*nLED) P_total=nLED*nLED*P_LED; %the total transmitted power Adet=1e-4; %the detector physical area of a PD Ts=1; %the gain of an optical filter; ignore if no filter is used index=1.5; %the refractive index of a lens at a PD; ignore if no lens is used FOV=70; %the FOV of the receiver G_Con=(index^2)/(sind(FOV).^2); %the gain of an optical concentrator; ignore if no lens is used %% room parameters lx=5; ly=5; lz=3;% the room dimensions in meter h=2.15; %the distance between the source and the receiver plane [XT,YT] = meshgrid([-lx/4 lx/4],[-ly/4 ly/4]); % the position of an LED; it is assumed that all LEDs are located at the same %point for the faster simulation % for one LED simulation located at the centre of the room, use XT=0 and YT=0 Nx=lx*20; Ny=ly*20; % the number of grid in the receiver plane, larger the number, better is the approximation but takes longer time x=linspace(-lx/2,lx/2,Nx); y=linspace(-ly/2,ly/2,Ny); [XR,YR]=meshgrid(x,y); %% simulation for sournce one D1=sqrt((XR-XT(1,1)).^2+(YR-YT(1,1)).^2+h^2); % the distance vector from the source 1 cosphi_A1=h./D1; % the angle vector recevier_angle=acosd(cosphi_A1); H_A1=(ml+1)*Adet.*cosphi_A1.^(ml+1)./(2*pi.*D1.^2); % the channel DC gain for source 1 P_rec_A1=P_total.*H_A1.*Ts.*G_Con;% the received power from source 1; P_rec_A1(find(abs(recevier_angle)>FOV))=0; %% assuming symmetric property, no need to calculate other power % if the transmitter is not symmetrical, you need to calculate power for individual LEDs P_rec_A2=fliplr(P_rec_A1); P_rec_A3=flipud(P_rec_A1); P_rec_A4=fliplr(P_rec_A3); P_rec_total=P_rec_A1+P_rec_A2+P_rec_A3+P_rec_A4; P_rec_dBm=10*log10(P_rec_total); %% figure surfc(x,y,P_rec_dBm); % contour(x,y,P_rec_dBm);hold on; % mesh(x,y,P_rec_dBm);

414

Optical Wireless Communications Optical power Pi

Photocurrent Ip(t)

Channel Rh(t) Noise n(t)

FIGURE 8.11  Modeling of VLC channel with IM/DD.

Unlike conventional RF links, the transmit optical power from an LED must be by its very nature non-negative, i.e., the continuous optical signal Pi ( t ) ≥ 0. Therefore, the time-average power Pt is given by T

1 Pi ( t ) dt (8.18) T →∞ T

Pt = lim



∫ 0

Based on the schematic diagram for a IM/DD VLC system—see Figure 8.11—the generated photocurrent at the receiver (i.e., PD) is given by I p ( t ) = RPt ( t ) ⊗ h ( t ) + n ( t ) (8.19)



where Pt(t) is the instantaneous transmit of an optical signal, h(t) is the CIR, n(t) = (0, σ2) n ( t ) ∼ N 0,  σ 2   is the signal-independent additive white Gaussian noise (AWGN), and the symbol “⊗” denotes the convolution operation. The average received optical power Pr = H(0)Pt, where H(0) is the channel DC gain. Though the channel model in principle is essentially similar to the IR model used in the Chapter 4, the reflectivity of surfaces differs, thus leading to different delay spread and ISI. The reflectivity of walls depends on the wavelength and materials used within a room. Although specular reflections can occur from mirrors or other shiny objects, most reflections are typically diffuse in nature, and most are modelled as Lambertian [100], [101]. The study by Kwonhyung et al. showed that the reflectivity depends on the texture of the reflecting surfaces and wavelength (because of the wide linewidth of the optical source); see Figure 8.12 [100]. However, it is observed that, the reflectivity of the surface in general is less for the visible light compared to the IR radiation. The plaster wall has the highest reflectivity followed by the floor and ceiling, respectively. Considering reflection from walls, the received power is defined in terms of the channel DC gain for both directed H dir ( 0 ) and reflected Href ( 0 ) paths, which is give by [22]:

(

)

N LEDS



Pr =

∑  

    Pt H dir ( 0 ) +  

   Pt dHref ( 0 )   (8.20)  Reflections     

∫

For the LOS channel, the DC gain is defined by [80]



 ( m + 1) APD cos m ( ∅ ) Ts ( ψ ) g ( ψ ) cos ( ψ ) , 0 ≤ ψ < Ψ c  2πdi2 H dir ( 0 ) =  (8.21)  0  ψ > Ψc 

415

Visible Light Communications

FIGURE 8.12  The PSD (solid line, which corresponds to the left axis) is compared to the measured spectral reflectance (which corresponds to the right axis) of plaster and plastic wall (dash-dot line), floor (dash line), and ceiling (dot line)[100].

The DC channel gain of the 1s reflection is given by [96]



 APD ( ml + 1)  ρ dAr cos ml ( φr ) cos ( α ir ) cos (βir ) Ts ( ψ ) g ( ψ ) cos ( ψ r ) ,  0 ≤ ψ r ≤ Ψ c H ref ( 0 ) =  2(πd1  d 2  )2 (8.22)  0  elsewhere  ψ r > Ψ c 

where φr is the angle of irradiance to a reflective point and ψr is the angle of incidence from the reflective surface Figure 8.13(a). Figure 8.13(b) depicts the received optical power distribution of the first

(a)

(b)

FIGURE 8.13  (a) LOS and diffuse propagation model and (b) received optical power distribution for the first reflection.

416

Optical Wireless Communications

reflection, showing the minimum power level of ∼−16.5 dB at the centre of the room and the maximum (∼−11.5 dBm) at the corners. The defined channel gains of a VLC system are considered as non–frequency selective, with the average received power proportional to the detectors active area and inversely proportional to the square of the distance separating source and receiver. The time domain channel characteristic of an optical channel is described by the time-invariant CIR [102], [103]. There are three factors influencing the CIR: (i) the emitting properties of the source, such as divergence angle and collection properties of the receiver (i.e., FOV); (ii) the various positions and orientation of the receiver with respect to the transmitter and the diffuse reflectors; and (iii) the different delays caused by each individual reflection. The CIR for a particular source S(x,y,z) and detector D(x,y,z) can be approximated by a scaled Dirac delta function [104]. For the LOS channel, the CIR is defined by [105]

ψ   d h 0 ( t; S , D ) ≈ H ( 0 ) rect  δ t − i  (8.23)  FOV   c

where c denotes the speed of light and the rectangular function is defined as  1,  x ≤ 1 rect ( x ) =  (8.24)  0, 1 < x < 0



Calculating the impulse response from the first bounce is given by [105]

h1 ( t; S , D ) ≈

N

∑dH j =1

ref

( 0 ) rect 

d + d2  ψ   δ t− 1 (8.25)   FOV c 

with j representing the index of the N segments from the reflective surface. Hence, the total impulse responses as the sum of the LOS and diffuse components is given by [22], [100], [105]

h ( t; S , D ) =  

N

∞

∑∑h (t; S, D ) (8.26) k j

j =1 k = 0

where k simply is the order of reflections. The Matlab codes to plot the power distribution due to reflections from walls are given in Program 8.2. Figure 8.14 shows the distribution of the received power Pr considering the influence of reflections. As shown, the received power is −2.8−4.2 dBm for the entire room. The received average power including reflection is about 0.5 dB larger than the directed received average power. Program 8.2:  Matlab codes to calculate the optical power distribution of first reflection at the receiving plane for a typical room. theta=70; % the semi-angle at half power m=-log10(2)/log10(cosd(theta)); %Lambertian order of emission P_LED=20; %tranmistted optical power by individeal LED nLED=60; % number of LED array nLED*nLED P_total=nLED*nLED*P_LED; %Total transmitted power Adet=1e-4; %detector physical area of a PD rho=0.8; %reflection coefficent Ts=1; %gain of an optical filter; ignore if no filter is used index=1.5; %refractive index of a lens at a PD; ignore if no lens is used FOV=70; %FOV of a receiver

Visible Light Communications

FIGURE 8.14  The distribution of received power with reflections.

G_Con=(index^2)/(sind(FOV).^2); %gain of an optical concentrator; ignore if no lens is used %% room dimension lx=5; ly=5; lz=2.15; % room dimension in meter [XT,YT,ZT] = meshgrid([-lx/4 lx/4],[-ly/4 ly/4],lz/2); % position of Transmitter (LED); Nx=lx*10; Ny=ly*10; Nz=round(lz*10); % number of grid in each surface dA=lz*ly/(Ny*Nz); % calculation grid area x=linspace(-lx/2,lx/2,Nx); y=linspace(-ly/2,ly/2,Ny); z=linspace(-lz/2,lz/2,Nz); [XR,YR,ZR] = meshgrid(x,y,-lz/2); %% %first transmitter calculation TP1=[0 0 lz/2]; % transmitter position TPV=[0 0 -1]; % transmitter position vector RPV=[0 0 1]; % receiver position vector %% %%%%%%%%%%%calculation for wall 1%%%%%%%%%%%%%%%%%% WPV1=[1 0 0]; % position vector for wall 1 for ii=1:Nx for jj=1:Ny RP=[x(ii) y(jj) -lz/2]; % receiver position vector h1(ii,jj)=0; % reflection from North face for kk=1:Ny for ll=1:Nz WP1=[-lx/2 y(kk) z(ll)]; D1=sqrt(dot(TP1-WP1,TP1-WP1)); cos_phi= abs(WP1(3)-TP1(3))/D1; cos_alpha = abs(TP1(1)-WP1(1))/D1;

417

418

Optical Wireless Communications

D2=sqrt(dot(WP1-RP,WP1-RP)); cos_beta=abs(WP1(1)-RP(1))/D2; cos_psi=abs(WP1(3)-RP(3))/D2; if abs(acosd(cos_psi)) βijk 

(8.63)

where Apd − rj is the collection area of the jth receiver, dij2 is the distance from the ith transmitter to the jth receiver, βijk is the angle of incidence on the receiver, and βc is the receiver field of view. The CSI is gathered through the received optical power of the received pilot signals and can be expressed in terms of the channel gain H(0). The DC gains making up the H-matrix is given as



 h  11 h12  h21 h22 H=     h h  N 1 N 2

 h1M  … h2 M      … hNM  

(8.64)

Note that the rank of the H-matrix affects the MIMO system performance. If the H-matrix is not a full rank, the speed of the system must decrease in order to compensate for the H-matrix not

438

Optical Wireless Communications

being a full rank. The imaging system, however, guarantees a full rank channel H-matrix, whereas non-imaging optical MIMO system can have ill-conditioned H-matrix [26]. The jth received signal is given by i = NT

rj = R ⋅ Pt

∑h ⋅ t + ij

i

i =1

σ nj2   

(8.65)

where R is the photodiode responsivity, Pt is the average transmit optical power, and σ 2nj is the meansquare noise current for the jth receiver. The received signals are inserted into the vector y, where y = [r1,…,rj,…,rNR]T. The received data is estimated using the inverse of the H-matrix H’ as

dest = y ⋅ H T

(8.66)

A number of MIMO decoding algorithms, such as zero-forcing (ZF) and minimum mean-squared error (MMSE), maximum-likelihood-decoding (MLD), and vertical Bell Labs layered space-time algorithm (V-BLAST) algorithms, are considered for MIMO-VLC [93]. It is observed that only marginal gain can be obtained by utilising complex algorithm in MIMO-VLC [146]. The estimated signals are then passed through low-pass filters and a parallel to serial converters (P/S) to recover the transmitted serial data. The Matlab code given in Program 8.4 can be used to simulate MIMO system and recover signal. Figure 8.30 shows simulated waveforms observed at the transmitter and receiver.

FIGURE 8.30  Simulated waveforms at the transmitter and receiver for the MIMO.

Visible Light Communications

439

Program 8.4:  Matlab codes to simulate the MIMO system. clear clc close all TxN=4; % No of the transmitter RxN=4; % No of receiver M=1; % bit resolutions bits/Hz/s/channel L=2^M; % Average symbol length nsym=1000; % number of symbols min_BitsToTest = 1e6; max_BitError = 30 ; SNR_dB=1:1:30; % SNR in dB Hmat=[0.4961 0.4936 0.4906 0.4931;... 0.1995 0.6526 0.4115 0.0529;... 0 0.0879 0.5623 0.2001; ... 0.0075 0 0.0538 0.4057]; Hmat=Hmat./sum(Hmat(:)); % Hmat=eye(4,4); for ideal MIMO channel hMod = modem.pammod(‘M’,L, ‘SymbolOrder’, ‘Gray’); hDemod = modem.pamdemod(hMod); % Create a M-PAM based on the modulator for jj = 1:length(SNR_dB) total_nbit = 0; % initialise the loop variables total_nerr = 0; % number of bits tested, errors while total_nbit 10 Mbps, it is necessary to increase the number of ANN inputs and neurons to N = {30, 40}, which can provide 12 and 19 Mbps at a BER of 10−6, respectively. A data rate of 20 Mbps can be achieved at a BER of 4 × 10−4 and 2.5 × 10−5 for N = 30 and 40, respectively. If N > 40, there is no increase in the transmission speed available, and as such, N = 50 was not included in the results. Figure 8.39 illustrates the ANN BER performance for T len of 214 and 104, respectively over the range of neurons. There are a few major differences between the two training sequences shown in Figure 8.39 (and Figure 8.38). The most important difference is that the maximum error-free

FIGURE 8.38  Equalised BER performance of the link with a training length of 215 and a varying number of neurons [187].

449

Visible Light Communications

FIGURE 8.39  ANN BER performance of the link with Tlen of 214 and 104; offering performance up to 19 and 18 Mbps, respectively [187].

data rate achievable is reduced by 1 Mbps to 18 Mbps for T len of 10 4. Clearly, due to the lack of differences in the equalised BER results, the equalised BER performance effectively becomes independent of the training length when a sufficient input-output map and decision boundary is formed. For N = {5, 10}, an error-free data rate of 8 Mbps is possible, while for N = 20, this is increased to 9 Mbps. For higher orders of N, there are slight differences in the achievable transmission speeds where the longer training lengths can offer up to 1 Mbps advantage over the shorter lengths. As for N = 30, error-free data rates of 12 Mbps and 13 Mbps can be obtained for T len of 214 and 215, respectively. Similarly, for N = 40, 18, and 19 Mbps can be attained for T len of 104 and 215, respectively, at a BER of 10 −6. Therefore, it is possible to infer that the ISI spans 30 to 40 symbols at higher data rates. It should be noted that for N = 50 there is no improvement at any training length, and for data rates >19 Mbps, the system becomes noise limited as no further improvement is observed by increasing N.

8.8  HOME ACCESS NETWORK There is the need to deliver multimedia services at gigabits per second or more data rates, in order to address the growing high data traffic at homes, offices, etc. There are a number of data transmission technologies that are used in home access networks, including RF [192], power line communications (PLC) [193], VLC [194], and infrared communications (IRC); see Figure 8.40 and Table 8.7. These technologies are interconnected and controlled by an intelligent MAC layer. The MAC layer selects the most appropriate communications technology in order to attain seamless communications while

TABLE 8.7 Technologies for Home Access Networks Technology

Data rate

RF—60 GHz LOS PLC VLC IRC

1 Gbps—Bidirectional 100 Mbps—Bidirectional >100 Mbps—Broadcast 1 Gbps—Bidirectional

450

Optical Wireless Communications

FIGURE 8.40  Home access network.

maintaining the best link performance [195]. The RF and PLC technologies are used as backbone links to distribute incoming high data rates information to a number of locations within homes and offices. Both wireless technologies (RF and optical) provide the next level of connectivity between the base stations and the end users. The architecture of the VLC system generally consists of three common layers: (i) physical, (ii) MAC, and (iii) application. The MAC layer supports [120] • Mobility • Security • Network beacons generation if the device is a coordinator

• Dimming • Mitigation of flickering • VPAN disassociation and association

• Visibility • Colour function • Providing a reliable link between peer MAC entities

The topologies supported by the MAC layer are peer-to-peer, broadcast, and star. Three different types of physical implementations of VLC are defined in the IEEE 802.15.7. For PHY I, PHY II, and PHY III, the data rates are 11.67–266.6 kbps, 1.25–96 Mbps, and 12–96 Mbps, respectively. The different channel coding schemes supported by 802.15.7 are convolutional codes and Reed Solman (RS) codes for the PHY I and run length limited (RLL) code for the PHY II (intended for indoor use) to address flicker mitigation and DC balance. A one-dimensional optical wireless access network schematic diagram is shown in Figure 8.41(a). The base station (BS) composed of three transmitters providing continuous angular coverage of ∼25 × 8° are connected via a bridge to each other and the gateway hub. Each user terminal (UT) has three receivers, each with a FOV matched to the transmitter channel. Together these create an overall reception field of view matching that of the BS transmitter. As UTs move around within the coverage area, they will be within the FOV of one or possibly more BSs. Of course, some form of control scheme would be required to ensure the correct receiver and transmitter channels are selected with the required link performance metrics.

451

Visible Light Communications

(a)

Tx 1 M A C L a y e r

Transmit data

1-to-3 splitter

Received data

Clock

Tx 2 Tx 3

Clock and data recovery

+Vbias Rx 1 +Vbias

3-to-1 demux

Signal

Rx 2 +Vbias

2-bit address

Rx 3 RSSI RSSI comparator and decoder (b)

FIGURE 8.41  A one-dimensional optical wireless cellular system: (a) block diagram and (b) functional block diagram of a BS/UT module.

452

Optical Wireless Communications

TABLE 8.8 Link Budget for a One-Dimensional VLC Cellular System Parameter

Cell centre

Average transmit power Received intensity at 3 m Estimated collection area (on-axis) Available power Additional receiver loss Available APD received sensitivity Measured system margin

+14 dBm (25 mW) −23 dBm/cm2 0.63 cm2 ∼−25 dBm ∼7 dB ∼35dBm ∼3dB

Figure 8.41(b) shows the functional block diagram of the physical layer of the BSs and UTs [196]. The transmit data from the MAC is split and used to intensity modulate the light sources. The outputs of receivers are applied to the received signal strength indicator (RSSI) comparator and a decoder module the output of which is applied to the 3-to-1 multiplexer in order to select the ‘active’ receiver. A simple power threshold level is adapted to select the particular receiver. In a situation where RSSI values of two receivers are greater than the threshold level, the receiver with the given lower index number (e.g., Rx1) is selected. This technique, called the ‘select good enough’ approach, is inferior when compared with the maximal ratio combining or the select best schemes, but it is rather simple to implement. Table 8.8 shows the link budget for the system shown in Figure 8.41. Figure 8.42 shows the BER performance against the on-axis link range, illustrating error-free operation (BER 40 dB), large output power (>10 W), ultra-wide bandwidth (>10 THz), high power transfer efficiency from pump to signal (>50%), low noise figure, and suitability for long-haul applications, has resulted in wide use of EDFAs in today’s optical communications [58], [64]–[66]. Since in FSO systems signals are not shared by many channels as in fibre optic communications, low-gain and low-noise EDFAs are suitable to be used in relayed-based FSO systems to simultaneously increase the signal quality and decrease the cost. However, EDFAs cannot be integrated with other semiconductor devices because of their physical size. 9.3.1.2  Semiconductor Optical Amplifier Even though EDFAs are widely used for signal amplification, SOAs are intensively studied as an alternative option as preamplifiers, in-line amplifiers, and booster amplifiers in communication systems. Furthermore, SOAs are also investigated for their applications as optical switching elements, optical modulators, frequency converters, and frequency chippers for pulse compression [14]. SOAs are characterised by an extremely strong nonlinearity, high operation speed, low power and energy consumption, and a small size compared to EDFAs [67]. SOAs’ most significant characteristics include high gain (25–30 dB); output saturation power in the range of 5–13 dBm; a wide gain bandwidth of 30–50 nm; input power level, which could be as low as −30 dBm (1μW); and a range of operating wavelengths of 0.8, 1.3, and 1.5 μm. However, SOAs also have some disadvantages when used merely as gain elements. SOAs have a higher noise figure than EDFAs since the gain coefficient of SOAs is very large, higher than 6 dB over 50 nm. Moreover, SOAs have higher crosstalk levels than EDFAs when operating at >10 Gbps since the conduction band lifetime becomes comparable to the bit period. This effect is known as the cross gain modulation (XGM) phenomena [67], [68]. 9.3.1.3  Comparison of EDFAs and SOAs Both EDFAs and SOAs can be used for amplification of the optical signal in optical communications; each has advantages and disadvantages as discussed. However, EDFAs are widely preferred

479

Relay-Assisted FSO Communications

to SOAs mainly for the following reasons. First, EDFAs can have high gains of up to 50 dB with a lower noise figure than SOAs, where the maximum achievable gain is 30 dB. In addition, in EDFAs, the spontaneous emission lifetime is within 5–10 ms [61], which is large enough compared to a bit period of interest for the electron transition (from higher to lower energy level) to respond to the optical signal fluctuations, and hence avoid any significant distortion of the amplified signal. Contrary to the EDFAs, the spontaneous emission lifetime of SOAs is usually within the range of 100–200 ps [61], which implies that the electron transition easily responds to the fluctuation of the optical signal at Gbps rates, thus resulting in potentially major signal impairments and a significant crosstalk level. Even though SOAs can operate at different wavelengths, are compact in size and lower in weight, and have lower pump power levels compared to EDFAs, EDFAs are widely used. This is because EDFAs with power pump lasers, fibres, control electronics, and other cheap optical components are easily integrated with planar waveguide optics and other devices.

9.4  ALL-OPTICAL REGENERATE-AND-FORWARD Although AOAF-based relay systems are an attractive option for practical implementation due to their simplicity, accumulation of the noise at each R ultimately limits the achievable maximum transmission distance. As such, the key component to enable large-scale transparent all-optical networks is the all-optical regenerator at the intermediate node/hop. This process is referred to as the AORF relaying technique, which has witnessed a significant worldwide research activity recently [69]–[72], but with no practical implementation in FSO systems yet. Optical signals propagated through a channel (free space or a fibre) experience signal impairments due to accumulated noise, fading, ASE noise, and nonlinearity of the fibre, which degrade the optical signal-to-noise ratio (OSNR) at the Rx. In order to preserve the signal integrity and ensure that it is transmitted over longer distances while routed via several optical nodes, signal restoration or regeneration can be employed periodically along the transmission path. Signal regeneration includes re-amplification, re-shaping, and re-timing, which are conveniently referred to as 1R, 2R, or 3R. The implementation of signal regeneration depends on the network-case scenario since it is not always beneficial to systematically use the three operations at each optical relay node. Note that the regeneration in 2R and 3R includes re-amplification and re-shaping. For smaller networks, 2R regeneration is more than adequate to maintain a satisfactory signal quality level, which is cost effective. Furthermore, for a long-haul communication link, loss and amplitude fluctuations are usually considered the two major sources of signal impairment compared to timing-jitter, thus there is no need for re-timing on a frequent basis.

9.4.1 Nonlinear effects Nonlinear effects, such as self-phase modulation (SPM), which results from the Kerr nonlinearity, are commonly used terms in fibre optics communication systems. Note that the refractive index of the transmission medium can only depend quadratically on the field, i.e., on the intensity I, as given by [73]:

n = n + n2 I (9.13)

where I = |A2|, A is the pulse envelope, and n2 (cm2/W) is the intensity-dependent refractive index, which depends on the polarisation of the field and is considered a positive value for most transparent materials when the intensity-dependent refractive index is positive. Note that the propagating pulse envelope experiences an additional self-phase shift imposed by n2, which is given as

∅ ( t ) = − k [ n + n2 I ( t )] Leff (9.14)

480

Optical Wireless Communications

where the SPM coefficient δ = k0n2Leff , and Leff is the effective fibre length, which is related to the physical fibre length L given by [74]

Leff = [1 − exp ( −κL )] / κ (9.15)

where κ is the linear loss coefficient. The relationship between ∅ ( t ) and the frequency is given by

ω=

d∅ (9.16) dt

SPM leads only to a phase shift in the time domain and therefore affects the dispersion characteristics of the optical fibre in high-speed optical transmission systems, in which the phase of signals are modulated that results in spectral broadening and pulse chirping (i.e., a temporal variation of the instantaneous frequency, which then runs through a wider range of frequencies) [75]. On the other hand, nonlinear effects can also be exploited in applications such as ultrafast all-optical switching, amplification, optical phase conjugation, pulse compression, and regeneration [74]. The SPM effect results in the leading edge of the propagating signal experiencing a positive refractive index gradient ( dn / dt ) and the trailing edge encountering a negative refractive index gradient ( − dn / dt ) as depicted in Figure 9.8. The effective nonlinear propagation coefficient is given by [76]

knl =

( 2πn2 ) λAeff

(9.17)

where λ is the wavelength of the incident photons, n2 is the nonlinear refractive index, and Aeff is the effective mode area. The equation shows that SPM gives rise to an intensity-dependent phase shift, but the pulse shape remains unaffected. Typically, κ = 0.2 dB/km at λ = 1550 nm and knl = 2.35 × 10 −3 1/mW [76]. In the absence of fibre losses, i.e., κ = 0 and thus Leff = L . Increasing the transmission linkspan between the in-line OAs often necessitates the use of higher transmit optical power levels injected into the optical fibre, which ultimately leads to pulse broadening due to SPM.

FIGURE 9.8  Spectral broadening of a pulse due to SPM-induced temporal variations.

Relay-Assisted FSO Communications

481

FIGURE 9.9  A signal regeneration scheme illustrating the re-amplification, re-shaping, and re-timing operations.

9.4.2 SPM-Based Optical Regenerator In order to increase the transmission distance, both the amplitude and timing-jitters of the propagating pulses must be limited, and the extinction ratio of the signal should be enhanced. The ways this optical regeneration can be achieve can be divided into three categories based on the degenerative effects corrected at each stage: 1R, 2R, and 3R; see Figure 9.9. The 1R regeneration scheme is the simplest regeneration process and only involves re-amplification of the optical signal using OAs such as rare-earth doped optical amplifiers or Raman amplifiers, which are used to overcome power loss due to signal attenuation and connection losses. In other words, 1R can also be referred to as the AOAF technique. 2R regeneration involves two processing steps of re-amplification and re-shaping of the optical signal. Re-shaping is used to suppress the noise accumulated and to improve the extinction ratio in an intensity modulated optical transmissions. The 3R regenerator scheme performs the additional task of re-timing of the signal by extracting the signal clock from the incoming data signal and transferring it to a newly generated pulse train [77]. Among various regenerator configurations, the 2R regenerator based on SPM-based spectral broadening has received significant research attention owing to the ease of its implementation, scalability to high bit rates and multiple channels [78], and the ease with which a re-timing stage can be integrated with it to provide a 3R regenerator [79]. The structure of the regenerator is simpler because it has no pump or probe source within the regenerator module. This type of regenerator achieves its functionality via a nonlinear transfer function, governing the instantaneous output versus input power of the regenerator [80]. Although the basic idea of the SPM-regenerator, which is also known as the Mamyshev optical regenerator, was developed in 1998 [81], the devices were studied extensively only after 2003 [82]–[85]. Figure 9.10 illustrates the schematic diagram of a 2R SPM-based all-optical regenerator. The regenerator normally consists of the three main elements of EDFA, highly nonlinear fibres (HNLF), and an optical band-pass filter (OBPF). At the start of the regeneration process, the degraded optical signal (both in amplitude and time) is applied to an EDFA to amplify the incoming noisy signal to a power level sufficient to produce pulse broadening through the SPM effect. The amplified signal is then passed through an ASE rejection module (filter) to reject the out-of-band ASE noise added by the high-power EDFA. Note that the bandwidth of the OBPF should be chosen to be wider than the signal wavelength. Next, the filtered signal is injected into the HNLF, where it experiences SPM-induced spectral broadening. An OBPF is subsequently used at the output of the HNLF as a re-shaping element, whose centre wavelength is slightly shifted from the input signal wavelength. The OBPF slices a portion of the broadened spectrum with which more power and a widthstabilised output pulse is generated. The bandwidth of the output filter sets the output pulse width and needs to be optimised carefully.

482

Optical Wireless Communications

FIGURE 9.10  Schematic diagram of an SPM-based 2R regenerator.

Note that the OBPF acts as an optical decision gate to discriminate marks and spaces based on the difference in SPM-induced spectral broadening in an optical fibre with nearly zero normal dispersion [6]. If the centre wavelength of the OBPF is offset with respect to the carrier frequency of the signal, then only pulses with sufficient power to produce large enough spectral broadening can pass through the filter. This results in a nonlinear transfer function for the regenerator and, therefore, a suppression in the noise present in the signal. Clearly, this nonlinear transfer function depends strongly on the launch power, filter offset, fibre length, and fibre dispersion [82]. Ultimately, the performance of an SPM-based regenerator depends on three main parameters: (i) the maximum nonlinear phase shift, (ii) the filter-passband offset, and (iii) the filter bandwidth, which must be large enough to accommodate the entire signal so that the width of the optical pulses remain intact [86]. This all-optical 2R regenerator improves the extinction ratio and the optical SNR of the degraded signal, and it has been demonstrated in highly nonlinear silica-based fibres to achieve a million kilometers of error-free transmission without electrical conversion [87].

9.4.3 Highly Nonlinear Fibres Nonlinear fibre optics play an important role in the design of high-capacity lightwave communication systems. More recently, a new family of fibres known as HNLFs, including microstructured fibres, holey fibres, and photonic crystal fibres, share the common property of a relatively narrow core surrounded by a cladding containing a large number of air holes [86]. The HNLF fibres are designed to maximise fibre nonlinearity and minimise optical loss, which makes such fibres optimal for constructing highly efficient fibre amplifiers, dispersion compensators, and various other nonlinear devices. The nonlinear effects associated with optical fibres, including SPM, are governed by a single nonlinear parameter knl defined in (9.15). For most optical fibres, knl has a value of ∼1 W−1/km, which is too small for nonlinear applications. Thus, HNLFs were developed to overcome this issue by having knl  >10 W−1/km [88]. As stated in (9.15), knl depends on wavelength of light λ, the effective mode area Aeff , and nonlinear index coefficient n2, which is a material parameter fixed for each glass material. Hence, the only parameter that can be adjusted to enhance knl is Aeff , which can be reduced with a proper silica-based optical fibre design, which often involves controlling the refractive index profile.

9.5  ALL-OPTICAL AOAF RELAY-BASED FSO WITH TURBULENCE For multi-hop FSO communication systems, the effect of atmospheric turbulence modelled by gamma gamma distribution was investigated in [95], where it was shown that the outage probabilities for both AF and DF degraded with the increasing number of hops. The potential of

483

Relay-Assisted FSO Communications

(a)

(b)

FIGURE 9.11  Serial AOAF relay-based FSO link: (a) schematic block diagram and (b) relay configuration.

a relay-based DSO system employing subcarrier phase shift keying (PSK) under strong turbulence, considering both the channel-state-information (CSI) and fixed-gain relays, was numerically investigated in [42]. It was illustrated that by keeping the hop length fixed and increasing the number of relay base stations the link outage performance deteriorated. For a dual-hop FSO link with strong turbulence, it was shown that at a target outage probability of 10 −6 the SNR gains were ∼12 dB and ∼19 dB for AF and DF serial-based relay nodes, respectively, compared with the P2P link [36]. The schematic block diagram of an AOAF multi-hop relay-based FSO link is shown in Figure 9.11(a), where each node communicates only with the next node, the one before, and the one after. The structure of a typical AOAF relay is depicted in Figure 9.11(b), where convex lenses are used for capturing and launching the optical beam into and out of the optical fibre. With reference to Figure 9.11(a), the received signal at R1, R1, and D are given by, respectively

yR1 ( t ) = hSR1 x ( t ) + nR1 ( t ) (9.18)



yR2 ( t ) = hR1R2 ( Goa1 yR1 ( t ) + n ASE − R1 ( t )) + nR2 ( t ) (9.19) yD ( t ) = hR2 D ( Goa 2 yR2 ( t ) + n ASE − R 2 ( t )) + nR3 ( t ) (9.20)



where x ( t ) and hSR1 are the transmitted data and the channel gain from S → R1, nR1 ( t ), nR2 ( t ) , and nR3 ( t ) is the AWGN sources at R1, R2, and R3, respectively (i.e., a combination of ambient light, thermal, dark, and shot noise sources) [1]. n ASE − R1 ( t ) and n ASE − R2 ( t ) are the OA’s ASE noise at R1, and R2, respectively. In most previous works reported on relay-assisted FSO networks, the OA gain Goa was assumed to be fixed [35], [55], [89]–[91]. Therefore, the received signal at the ith R terminal (i.e., R of D), i = 1, 2,…N + 1 is given as [55]:



 yRi ( t ) =  

i −1

∏ m =1

  Goam hm  hi x ( t ) + nRi ( t ) +   

+ hi n ASEm ( t ) +

i−2

i −1

∑∏ j =1 m = j

i −1

∑∏G

h hn

oam m i ASEi

j =1 m = j +1

i −1

 Goam hm +1nRi ( t ) 

( t ) .

(9.21)

484

Optical Wireless Communications

where m ∈{1,2,…. N }, and hm is the channel gain from m-first to the mth R, which includes the combination of the atmospheric attenuation ha and atmospheric turbulence ht , as given by hi = ha ht . (9.22)

The signal at the output of D is given as

yD′ ( t ) = Goa 3 yD ( t ) + n ASE − R 3 ( t ) (9.23)



Note that for direct transmission from S → D we have ySD′ ( t ) = heq ( t ) x ( t ) + neq ( t ) (9.24)



where heq ( t ) and neq ( t ) are the equivalent channel gain and AWGN, respectively. Note that for two random variables the PDF is [100] ∞

f (z) =



1

∫xf

−∞

X

z ( x ) fY   dx , (9.25) x

where z is the product of random variables x and y (z − xy). The atmospheric turbulence is mainly characterised by Cn2, which is a measure of the turbulence strength; see Chapters 3, 6, and 7. Another parameter that is commonly used to determine the turbulence regime is Rytov variance σ 2R , see Chapters 3, 6, and 7. Here, the gamma gamma distribution, which applies to all turbulence regimes, describes the intensity fluctuations where the probability density function is given as [1]

p(I ) =

2( χ) I ( 0.5ζ )−1 Kζ ′ 2 0.5ζI ,  I > 0. (9.26) Γ ( α ) Γ (β ) 0.5ζ

(

)

where ζ = α + β, ζ′ = α − β, ζ ′′ = β − α, and χ = αβ. Γ (.) and Kζ ′ (.) are the gamma function and the modified Bessel function of the second kind of order (α-β), respectively. Parameters α and β are related to the turbulence conditions; see Chapter 3 [92]. Therefore, the PDF of the channel gains of hSR1 , hR1R2 , hR2 D, and heq are given [93] f ( hSR1 ) =



2( χ) 0.5ζ hSR1 ) Kζ ′ 2 χhSR1 (9.27) ( Γ ( α ) Γ (β ) 0.5ζ

(

)



2( χ) f ( hR1R2 ) = ( hR1R2 )0.5ζ Kζ ′ 2 χhR1R2 (9.28) Γ ( α ) Γ (β )



f ( hR2 D ) =

0.5ζ

(

)

2 (χ) 0.5ζ hR2 D ) K ζ′ 2 χhR2 D (9.29) ( Γ ( α ) Γ (β ) 0.5ζ

(

)

3



ζ−1  ( χ )0.5ζ  h30.5 e f ( heq ) =   0.5ζ  Γ ( α ) Γ (β )  (Goa1Goa 2 Goa 3 )

×G

6,0 0,6

 ( χ )3 h3e  0.5ζ ′,0.5ζ ′′,0.5ζ ′,0.5ζ ′′,0.5ζ ′,0.5ζ ′′    (Goa1Goa 2 Goa 3 

(9.30)

485

Relay-Assisted FSO Communications

For a non-return-to-zero, on-off keying (NRZ-OOK) intensity modulated FSO system, the probability of error as a function of heq is given by:

Pe ( heq ) =

 heq 1 erfc  2 2  2 2σ N

  (9.31) 

The more generalised BER expression for the N-hop FSO link is given as [106] N

(

)

ζ+ 2

 ( χ )0.5ζ  8σ 2 4 Pe =   0.5ζ N N 2  Γ ( α ) Γ (β )  4 2πσ N ( gN ) 2 ( 2π )

 2−ζ 2−ζ 4−ζ , ,  8σ 2 χ 2 N ( ) N 4 4 4 4 N ,3  × G3,4 N +2  24 N F N 2 N −ζ 2 − ζ G , ,  4 4 

 (9.32)     

where F N = (Goa1Goa 2 ,….GoaN , G N ) implies that G is repeated N-times and σ 2N denotes the equivalent variance for N-hop. Figure 9.12 presents a comparison of measured and theoretical BER performance for single-, dual-, and triple-hop FSO links under a clear channel with no turbulence. As shown in the figure, with the serial relay–assisted technique, there is a remarkable improvement in the BER performance compared to the link with no relays. For example, for the measured data, at a BER of 10 −5, there is ∼3.0 dB and ∼6.0 dB of SNR gains for dual- and triple-hops, respectively, compared to the single FSO link. Note that by splitting the overall linkspan into smaller sections, more energy (or power) is conserved. Notice the perfect match between theoretical and measured plots for all cases, particularly for the BER values below 10 −4. The slight mismatch between predicted and measured plots at higher values of BER (i.e., >10 −4) is due to losses associated with the optical components and ASE noise of EDFA in experimental case.

FIGURE 9.12  Experimental (exp) and theoretical (theo) result for BER versus SNR for a single-hop (S), dual-hop (D), and triple-hop (T) with no turbulence.

486

Optical Wireless Communications

FIGURE 9.13  Measured eye diagrams of received FSO system for (a) triple-hop, (b) dual-hop, and (c) single link for BER values of 10 −6, 10 −4, and 10 −2, respectively, over Cn2 = 3.8 × 10 −10 m−2/3.

The eye diagrams for the link with turbulence (i.e., Cn2 = 3.8 × 10 −10 m−2/3) for single-hop, dualhop, and triple-hop for the BER values of 10 −6, 10 −4, and 10 −2 is depicted in Figure 9.13. For a higher turbulence level, the Q-factor decreases as BER increases. For example, for the triple-hop system, at BER values of 10 −6, 10 −4, and 10 −2 the Q-factors are 6.05, 3.91, and 2.12, respectively. While for the single-hop system, the eye diagrams were completely shut and cannot be captured at a BER of 10 −2. This can be explained as follows: When the optical beam propagates through the turbulence channel, the beam experiences scattering (phase fluctuation) and power losses. The longer the link, the more light beams are lost. By shortening the distance, the fading variances along the link decrease, and ultimately improve the performance of the relay-assisted system. This improvement is due to the fact that fading variance is distance-dependent in a FSO channel, and relay-assisted transmission takes advantage of the resulting shorter hops.

REFERENCES 1. Z. Ghassemlooy, W. O. Popoola, and S. Rajbhandari, Optical Wireless Communications—System and Channel Modelling with Matlab, 1st ed., vol. 1, no. iii. CRC Press, 2012. 2. M. A. Khalighi and M. Uysal, “Survey on Free Space Optical Communication: A Communication Theory Perspective,” Commun. Surv. Tutorials, IEEE, vol. PP, no. 99, p. 1, 2014. 3. A. Paraskevopoulos, J. Vucˇi´c, S. H. Voss, R. Swoboda, and K. D. Langer, “Optical Wireless Communication Systems in the Mb/s to Gb/s Range, Suitable for Industrial Applications,” Mechatronics, IEEE/ASME Trans., vol. 15, no. 4, pp. 541–547, 2010. 4. G. Parca, “Optical Wireless Transmission at 1.6-Tbit/s (16 × 100 Gbit/s) for Next-Generation Convergent Urban Infrastructures,” Opt. Eng., vol. 52, no. 11, p. 116102, Nov. 2013. 5. A. Vavoulas, H. G. Sandalidis, and D. Varoutas, “Weather Effects on FSO Network Connectivity,” Opt. Commun. Networking, IEEE/OSA J., vol. 4, no. 10, pp. 734–740, 2012.

Relay-Assisted FSO Communications

487

6. J. Zhuanhong, Z. Qinglin, and A. Faliang, “Atmospheric Attenuation Analysis in the FSO Link,” in International Conference on Communication Technology, 2006 (ICCT ’06), 2006, pp. 1–4. 7. J. Vitásek et al., “Atmospheric Turbulences in Free Space Optics Channel,” 2011 34th International Conference on Telecommunications and Signal Processing (TSP), Budapest, pp. 104–107, 2011. 8. X. Yi, Z. Liu, and P. Yue, “Optical Scintillations and Fade Statistics for FSO Communications through Moderate-to-Strong Non-Kolmogorov Turbulence,” Opt. Laser Technol., vol. 47, 2013. 9. S. Arnon, J. R. Barry, G. K. Karagiannidis, R. Schober, and M. Uysal, Advanced Optical Wireless Communication Systems. New York, USA: Cambridge University Press, 2012. 10. Z. Ghassemlooy, H. Le-Minh, and M. Ijaz, “Free Space Optical Communications,” in Optical and Microwave Technologies for Telecommunication Networks, vol. Chapter 10, no. 2, UK: J Wiley, 2016, pp. 349–377. 11. H. Hemmati, “Deep Space Optical Communications,” Deep Space Communications and Navigation Series. Wiley-Interscience, California, 2006. 12. S. V. Kartalopoulos, Free Space Optical Networks for Ultra-Broad Band Services. 2011. 13. P. C. Gurumohan and J. Hui, “Topology Design for Free Space Optical Networks,” in IEEE International Conference on Computer Communications and Networks, 2003, pp. 576–579. 14. R. Sabella and P. Lugli, High Speed Optical Communications. UK: Springer, 2002. 15. F. S. Vetelino, C. Young, L. Andrews, and J. Recolons, “Aperture Averaging Effects on the Probability Density of Irradiance Fluctuations in Moderate-To-Strong Turbulence,” Appl. Opt., vol. 46, no. 11, pp. 2099–2108, 2007. 16. M. A. Khalighi, N. Schwartz, N. Aitamer, and S. Bourennane, “Fading Reduction by Aperture Averaging and Spatial Diversity in Optical Wireless Systems,” IEEE/OSA J. Opt. Commun. Netw., vol. 1, no. 6, pp. 580–593, 2009. 17. H. Yuksel, S. Milner, and C. C. Davis, “Aperture Averaging for Optimizing Receiver Design and System Performance on Free-Space Optical Communication Links,” J. Opt. Netw., vol. 4, no. 8, p. 462, 2005. 18. A. A. Hashmi, A. Eftekhar, A. Yegnanarayanan, “Analysis of Optimal Adaptive Optics System for Hybrid RF-Wireless Optical Communication for Maximum Efficiency and Reliability,” in International Conference on Emerging Technologies, 2008. 19. Z. Zhao, S. D. Lyke, and M. C. Roggemann, “Adaptive Optical Communication through Turbulent Atmospheric Channels,” in IEEE International Conference on Communications, 2008, pp. 5432–5436. 20. F. E. Zocchi, “A Simple Analytical Model of Adaptive Optics for Direct Detection Free-Space Optical Communication,” Opt. Commun., vol. 248 (2005, no. 248, pp. 359–374, 2005. 21. M. Uysal, J. (Tiffany) Li, and M. Yu, “Error Rate Performance Analysis of Coded Free-Space Optical Links over Gamma-Gamma Atmospheric Turbulence Channels,” IEEE Trans. Wirel. Commun., vol. 5, no. 6, pp. 1229–1233, 2006. 22. M. Uysal and J. Li, “BER Performance of Coded Free-Space Optical Links over Strong Turbulence Channels,” in 2004 IEEE 59th Vehicular Technology Conference. VTC 2004-Spring (IEEE Cat. No.04CH37514), vol. 1, pp. 352–356. 23. S. Rajbhandari, Z. Ghassemlooy, P. A. A. Haigh, T. Kanesan, and T. Xuan, “Experimental Error Performance of Modulation Schemes under a Controlled Laboratory Turbulence FSO Channel,” Light. Technol. J., vol. 33, no. 1, pp. 244–250, 2015. 24. X. Song and J. Cheng, “Subcarrier Intensity Modulated Optical Wireless Communications Using Noncoherent and Differentially Coherent Modulations,” J. Light. Technol., vol. 31, no. 12, pp. 1906–1913, 2013. 25. W. O. Popoola and Z. Ghassemlooy, “BPSK Subcarrier Intensity Modulated Free-Space Optical Communications in Atmospheric Turbulence,” J. Light. Technol., vol. 27, no. 8, pp. 967–973, 2009. 26. A. Jaiswal, M. R. Bhatnagar, and V. K. Jain, “Performance Evaluation of Space Shift Keying in FreeSpace Optical Communication,” J. Opt. Commun. Netw., vol. 9, no. 2, p. 149, 2017. 27. S. M. Navidpour, M. Uysal, and M. Kavehrad, “BER Performance of Free-Space Optical Transmission with Spatial Diversity,” IEEE Trans. Commun., vol. 6, no. 8, pp. 2813–2819, 2007. 28. G. Yang, M.-A. Khalighi, Z. Ghassemlooy, and S. Bourennane, “Performance Evaluation of ReceiveDiversity Free-Space Optical Communications over Correlated Gamma-Gamma Fading Channels,” Appl. Opt., vol. 52, no. 24, pp. 5903–11, Aug. 2013. 29. K. Kumar, D. K. Borah, and S. Member, “Quantize and Encode Relaying through FSO and Hybrid FSO/ RF Links,” IEEE Trans. Veh. Technol., vol. 64, no. 6, pp. 2361–2374, 2015. 30. M. Ahmad and O. Awwad, “Synergies of Radio Frequency and Free Space Optics Communication: New Hybrid Solutions for Next Generation Wireless Mesh Networks,” Int. J. Comput. Networks, vol. 4, no. 4, pp. 135–155, 2012.

488

Optical Wireless Communications

31. W. Gappmair and M. Flohberger, “Error Performance of Coded FSO Links in Turbulent Atmosphere Modeled by Gamma-Gamma Distributions,” IEEE Trans. Wirel. Commun., vol. 8, no. 5, pp. 2209–2213, 2009. 32. A. T. Pham, T. C. Thang, S. Guo, and Z. Cheng, “Performance Bounds for Turbo-Coded SC-PSK/ FSO Communications over Strong Turbulence Channels,” in International Conference on Advanced Technologies for Communications, 2011, pp. 161–164. 33. T. Ohtsuki, “Turbo-Coded Atmospheric Optical Communication Systems,” in IEEE International Conference on Communications (ICC), 2002, vol. 5. 34. I. B. Djordjevic, “LDPC-Coded Optical Communication over the Atmopsheric Turbulence Channel,” in Conference Record of the Forty-First Asilomar Conference on Signals, Systems and Computers, 2007, pp. 1903–1909. 35. M. Safari and M. Uysal, “Relay-Assisted Free-Space Optical Communication,” IEEE Trans. Wirel. Commun., vol. 7, no. 12, pp. 5441–5449, 2008. 36. A. Sendonaris, E. Erkip, and B. Aazhang, “User Cooperation Diversity-Part I: System Description,” IEEE Trans. Commun., vol. 51, no. 11, pp. 1927–1938, 2003. 37. a. Sendonaris, E. Erkip, and B. Aazhang, “User Cooperation Diversity-Part II: Implementation Aspects and Performance Analysis,” IEEE Trans. Commun., vol. 51, no. 11, pp. 1939–1948, 2003. 38. J. N. Laneman, D. N. C. Tse, and G. W. Wornell, “Cooperative Diversity in Wireless Networks: Efficient Protocols and Outage Behavior,” IEEE Trans. Inf. Theory, vol. 50, no. 12, pp. 3062–3080, Dec. 2004. 39. G. Kramer, M. Gastpar, and P. Gupta, “Cooperative Strategies and Capacity Theorems for Relay Networks,” IEEE Trans. Inf. Theory, vol. 51, no. 9, pp. 3037–3063, 2005. 40. E. Bayaki, D. S. Michalopoulos, and R. Schober, “EDFA-Based All-Optical Relaying in Free-Space Optical Systems,” IEEE Trans. Commun., vol. 60, no. 12, pp. 3797–3807, Dec. 2012. 41. X. Tang, Z. Wang, Z. Xu, and Z. Ghassemlooy, “Multihop Free-Space Optical Communications over Turbulence Channels with Pointing Errors Using Heterodyne Detection,” J. Light. Technol., vol. 32, no. 15, pp. 2597–2604, 2014. 42. C. K. Datsikas, K. P. Peppas, N. C. Sagias, and G. S. Tombras, “Serial Free-Space Optical Relaying Communications over Gamma-Gamma Atmospheric Turbulence Channels,” J. Opt. Commun. Netw., vol. 2, no. 8, p. 576, Jul. 2010. 43. S. M. Aghajanzadeh and M. Uysal, “Outage Performance and DMT Analysis of DF Parallel Relaying in FSO IM/DD Communications,” in IEEE Vehicular Technology Conference (VTC Fall), 2012, vol. 2, pp. 1–5. 44. M. Sheng, P. Jiang, Q. Hu, Q. Su, and X. Xie, “End-to-End Average BER Analysis for Multihop FreeSpace Optical Communications with Pointing Errors,” J. Opt., vol. 15, no. 5, p. 055408, May 2013. 45. M. A. Kashani, M. Safari, and M. Uysal, “Optimal Relay Placement and Diversity Analysis of RelayAssisted Free-Space Optical Communication Systems,” J. Opt. Commun. Netw., vol. 5, no. 1, pp. 37–47, Dec. 2013. 46. M. Dohler and Y. Li, Cooperative Communications: Hardware, Channel and PHY. United Kindom: Wiley, 2010. 47. M. A. Kashani, M. Safari, and M. Uysal, “Optimal Relay Placement and Diversity Analysis of RelayAssisted Free-Space Optical Communication Systems,” Opt. Commun. Netw., vol. 5, no. 1, pp. 37–47, 2013. 48. A. Nasri, R. Schober, and I. F. Blake, “Performance and Optimization of Amplify-and-Forward Cooperative Diversity Systems in Generic Noise and Interference,” IEEE Trans. Wirel. Commun., vol. 10, no. 4, pp. 1132–1143, 2011. 49. L. Song, “Relay Selection for Two-Way Relaying with Amplify-and-Forward Protocols,” IEEE Trans. Veh. Technol., vol. 60, no. 4, pp. 1954–1959, 2011. 50. M. R. Bhatnagar, “Average BER Analysis of Relay Selection Based Decode-and-Forward Cooperative Communication over Gamma-Gamma Fading FSO Links,” 2013 IEEE Int. Conf. Commun., pp. 3142–3147, Jun. 2013. 51. C. Abou-Rjeily, “Achievable Diversity Orders of Decode-and-Forward Cooperative Protocols over Gamma-Gamma Fading FSO Links,” IEEE Trans. Commun., vol. 61, no. 9, pp. 3919–3930, 2013. 52. M. R. Bhatnagar, “Performance Analysis of Decode-and-Forward Relaying in Gamma-Gamma Fading Channels,” IEEE Photonics Technol. Lett., vol. 24, no. 7, pp. 545–547, 2012. 53. S. Anees and M. R. Bhatnagar, “Performance Evaluation of Decode-and-Forward Dual-Hop Asymmetric Radio Frequency-Free Space Optical Communication System,” IET Optoelectron., vol. 9, no. 5, pp. 232–240, 2015. 54. J. N. Laneman and G. W. Wornell, “Distributed Space–Time-Coded Protocols for Exploiting Cooperative Diversity in Wireless Networks,” IEEE Trans. Inf. Theory, vol. 49, no. 10, pp. 2415–2425, 2003.

Relay-Assisted FSO Communications

489

55. S. Kazemlou, S. Hranilovic, and S. Kumar, “All-Optical Multihop Free-Space Optical Communication Systems,” J. Light. Technol., vol. 29, no. 18, pp. 2663–2669, Sep. 2011. 56. G. P. Agrawal, Fiber-Optic Communications Systems, 3rd ed. New York, USA: John Wiley & Sons, 2002. 57. U. Adar, T. Kuzniz, I. D. Haber, and N. Budin, “An All-Optical Multi-Hop (Cascaded) High Bit Rate Wireless Communication Field Trial,” in Optical Fiber Communication Conference and Exhibit (OFC), 2002, no. 2001, pp. 446–448. 58. P. C. Becker, N. A. Olsson, and J. R. Simpson, Erbium-Doped Fiber Amplifiers: Fundamentals and Technology. Academic Press, 1999. 59. T. Mukai, Y. Yamamoto, and T. Kimura, “Optical Amplification by Semiconductor Lasers,” in Semiconductors and Semimetals, vol. 22, Academic Press, 1985, pp. 265–319. 60. H. Ghafouri-Shiraz, The Principles of Semiconductor Laser Diodes and Amplifiers: Analysis and Transmission Line Laser Modeling. London: Imperial College Press, 2004. 61. A. Abisayo, Optically Amplified Free-space Optical Communication Systems, University of Nottingham, 2013. 62. D. N. Payne, I. M. Jauncey, L. Reekie, and R. J. Mears, “High-Gain Rare-Earth-Doped Fiber Amplifier at 1.54 μm,” in International Conference on Integrated Optics and Optical Fiber Communication, 1987, vol. 3, p. 167. 63. R. Ramaswami and K. Sivarajan, Optical Networks: A Practical Perspective, 2nd ed. New York: Morgan Kaufmann, 2002. 64. M. YÜCEL, “Increase of Transmission Distance Using EDFA and Module Design for Free Space Optics Applications,” Gazi Univ. J. Sci., vol. 25, no. 1, pp. 119–125, 2012. 65. C. Barnard, S. Member, P. Myslinski, J. Chrostowski, S. Member, and M. Kavehrad, “Analytical Model for Rare-Earth-Doped Fiber Amplifiers and Lasers,” vol. 30, no. 8, pp. 1817–1830, 1994. 66. M. Premaratne and G. P. Agrawal, Light Propagation Gain Media: Optical amplifiers. United States of America: Cambridge University Press, 2011. 67. P. Urquhart, Advances in Optical Amplifiers. 2011. 68. C. Lin, Optical components for communications: Principles and applications. Kluwer Academic Publishers, 2004. 69. Z. Zhu, M. Funabashi, Z. Pan, B. Xiang, L. Paraschis, and S. J. B. Yoo, “Jitter and Amplitude Noise Accumulations in Cascaded All-Optical Regenerators,” J. Light. Technol., vol. 26, no. 12, pp. 1640–1652, 2008. 70. T. I. Lakoba, J. R. Williams, and M. Vasilyev, “Low-Power, Phase-Preserving 2R Amplitude Regenerator,” Opt. Commun., vol. 285, no. 3, pp. 331–337, 2012. 71. T. N. Nguyen et al., “Self-Phase-Modulation-Based 2R Regenerator Including Pulse Compression and Offset Filtering for 42.6 Gbit/s RZ-33% Transmission Systems,” Opt. Express, vol. 17, no. 20, pp. 2900–2902, 2009. 72. M. Rochette, J. N. Kutz, J. L. Blows, D. Moss, J. T. Mok, and B. J. Eggleton, “Bit-Error-Ratio Improvement with 2R Optical Regenerators,” IEEE Photonics Technol. Lett., vol. 17, no. 4, pp. 908–910, 2005. 73. A. Hasegawa and M. Matsumoto, “Optical Solitons in Fibers,” 2003, pp. 41–59. 74. G. P. Agrawal, Nonlinear Fiber Optics, 5th ed. United Kingdom: Academic Press, 2013. 75. V. B. Heidelberg, Nonlinear Optics in Telecommunications. New York, USA: Springer, 2004. 76. S. P. Singh and N. Singh, “Nonlinear Effects in Optical Fibers: Origin, Management and Applications,” Prog. Electromagn. Res., vol. 73, pp. 249–275, 2007. 77. H. Padda, Noise Modelling for 2R Optical Regenerator Systems, University of Nottingham, 2009. 78. L. Provost, F. Parmigiani, P. Petropoulos, and D. J. Richardson, “Investigation of Simultaneous 2R Regeneration of Two 40-Gb/s Channels in a Single Optical Fiber,” IEEE Photonics Technol. Lett., vol. 20, no. 4, pp. 270–272, 2008. 79. C. Ito and J. C. Cartledge, “Polarization independent All-Optical 3R Regeneration Based on the Kerr Effect in Highly Nonlinear Fiber and Offset Spectral Slicing,” IEEE J. Sel. Top. Quantum Electron., vol. 14, no. 3, pp. 616–624, 2008. 80. M. Rochette, L. Fu, V. Ta’eed, D. J. Moss, and B. J. Eggleton, “2R Optical Regeneration: An All-Optical Solution for BER Improvement,” IEEE J. Sel. Top. Quantum Electron., vol. 12, no. 4, pp. 736–743, 2006. 81. P. V. Mamyshev, “All-Optical Data Regeneration Based on Self-Phase Modulation Effect,” in European Conference on Optical Communications, 1998, vol. 1, no. September, pp. 475–476. 82. T.-H. Her, G. Raybon, and C. Headley, “Optimization of Pulse Regeneration at 40 Gb/s Based on Spectral Filtering of Self-Phase Modulation in Fiber,” IEEE Photonics Technol. Lett., vol. 16, no. 1, pp. 200–202, 2004.

9781498742696_Ch09.indd 489

17/04/19 5:58 PM

490

Optical Wireless Communications

83. N. Yoshikane, I. Morita, T. Tsuritani, A. Agata, N. Edagawa, and S. Akiba, “Benefit of SPM-Based All-Optical Reshaper in Receiver for Long-Haul DWDM Transmission Systems,” IEEE J. Sel. Top. Quantum Electron., vol. 10, no. 2, pp. 412–420, 2004. 84. M. Matsumoto, “Performance Analysis and Comparison of Optical 3R Regenerators Utilizing SelfPhase Modulation in Fibers,” J. Light. Technol., vol. 22, no. 6, pp. 108–109, 2004. 85. L. B. Fu, M. Rochette, V. G. Ta, D. J. Moss, and B. J. Eggleton, “Investigation of Self-Phase Modulation Based Optical Regeneration in Single Mode As2Se3 Chalcogenide Glass Fiber,” Opt. Express, vol. 13, no. 19, pp. 7637–7644, 2005. 86. G. P. Agrawal, Applications of Nonlinear Fiber Optics, 2nd ed., vol. 1. Academic Press, 2015. 87. G. Raybon et al., “40 Gbit/s Pseudo-Linear Transmission over One Million Kilometers,” in Optical Fiber Communication Conference, 2002. 88. G. P. Agrawal, Nonlinear Fiber Optics, 4th ed. London: Academic Press, 2007. 89. M. Karimi and M. Nasiri-kenari, “Free Space Optical Communications via Optical Amplify-andForward Relaying,” J. Light. Technol., vol. 29, no. 2, pp. 242–248, 2011. 90. M. A. Kashani, M. M. Rad, M. Safari, and M. Uysal, “All-Optical Amplify-and-Forward Relaying System for Atmospheric Channels,” IEEE Commun. Lett., vol. 16, no. 10, pp. 1684–1687, Oct. 2012. 91. I. S. Ansari, F. Yilmaz, and M. S. Alouini, “Impact of Pointing Errors on the Performance of Mixed RF/ FSO Dual-Hop Transmission Systems,” IEEE Wirel. Commun. Lett., vol. 2, no. 3, 2013. 92. R. Nebuloni and C. Capsoni, “Effects of Adverse Weather on Free Space Optics,” 2016, pp. 47–68. 93. A. P. Prudnikov, I. A. Brychkov, and O. I. Marichev, Integrals and Series: More Special Functions. New York, USA: Gordon and Breach Science Publishers, 1990.

Index Note: Italicized page numbers refer to figures, bold page numbers refer to tables 3 GHz frequency band, 2 10 Mbps, 13, 258, 259, 263, 264, 265, 443, 447, 447–448 100 Mbps, 1, 3, 7, 13, 14, 21, 25, 246–248, 247, 248, 258, 262, 399, 449 20 MHz frequency band, 7 28 GHz frequency band, 2 40 GHz frequency band, 2 60 GHz band, 2, 7 90–110 GHz frequency band, 2

A Absolute temperature, 42 Absorption, 41, 42, 105 Absorption coefficient, 107 Absorption loss, 49 Activation function, 283 Adaptive channel estimation algorithm, 441 Adaptive equaliser, ANN-based, 286–295 diversity techniques, 294–295 versus FIR-based equalisers, 292–294 for OOK-NRZ, 288–289 structure of, 286 Adaptive linear transversal equaliser, 279 Adaptive optics schemes, 353–354 Additive white Gaussian noise (AWGN), 81, 83, 98, 167–169, 175, 188, 231, 233, 314, 414 Advanced Infrared (AIr) physical layer interface, 20 Advection fog, 110 Aerosol refraction index, 107 Aerosols, 106–107 All-optical amplify-and-forward (AOAF), 476–479 basic configuration, 476 with turbulence, 482–486 All-optical regenerate-and-forward, 479–482 highly nonlinear fibres, 482 nonlinear effects, 479–480 SPM-based optical regenerator, 481–482 Ambient light sources, 24, 229–230 American National Standards Institute (ANSI), 26 Amplitude modulation (AM), 160 pulse, 158, 163, 163–166, 428 quadrature, 158, 160, 198, 210, 313, 430 Amplitude shift keying (ASK), 211 Analogue filters, 278 Analogue intensity modulation (AIM), 160–162 Analogue-to-digital converter (DAC), 202, 203, 432 Analytical PSD of OOK-NRZ, Matlab codes for, 165 Analytical PSD of OOK-RZ, Matlab codes for, 165–166 Analytical PSD of PPM, Matlab codes for, 173 ANN-based adaptive equaliser, 286–295 diversity techniques, 294–295 versus FIR-based equalisers, 292–294 structure of, 286 ANN-based DFE for OOK-NRZ, Matlab codes for, 289–290

ANSI Z-136, 26 APD photodetector, 67 Aperture averaging, 384–386 Gaussian beam wave, 385–386 plane wave, 384–385 spherical wave, 385 Arbitrary waveform generator (AWG), 136 Artificial light interference, 96–104 fluorescent lamps, 99–102, 100–101, 100–103 incandescent lamp, 99, 99, 99 sources of, 98 spectrum of, 263 time waveform, 263 Artificial neural network (ANN), 282–285 adaptive equaliser based on, 286–295 architectures, 284–285 backpropagation learning, 286 bit error rate, 449 with feedback connection, 285 multilayer feedforward network, 284 multilayer perceptron, 444, 447 neuron, 283 parameters for equalisation, 286 recurrent neural network, 285 single-layer feedforward network, 284 training, 285–286, 287 Atmospheric attenuation coefficient, 105 Atmospheric channel loss, 104–107 Atmospheric refractive index, 121 Atmospheric turbulence models, 121–135 gamma-gamma turbulence model, 131–135 log-normal turbulence model, 125–129, 131 negative exponential turbulence model, 135 spatial coherence in weak turbulence, 129 Atmospheric turbulence-induced penalty, 335–338 Automatic frequency control unit (AFC), 222 Avalanche photodetectors, 65, 67

B Background radiation, 74 Backhaul bottlenecks, 1, 4 Backpropagation learning (BP), 286 Band-gap energy, 42 Band-pass filter (BPF), 215, 216, 218, 219, 220, 220–223, 222, 312, 313, 326 Bandwidth demand for, 1 efficiency, 159 frequency, 477 Bandwidth requirements and power efficiency of different modulation schemes, Matlab codes for, 195 Base stations (BSs), 1, 4 Baseband modulations, 162–169 comparisons of schemes, 192–196 on-off keying, 163–166 OOK performance in AWGN channel, 167–169 peak-to-average power ratio, 196

491

492 Baseband modulations (Continued) power efficiency, 193–194 pulse amplitude modulation, 163–166 transmission bandwidth requirements, 194–195 transmission capacity, 195 transmission rate, 195–196 Baseline wander (BLW), 159 effect in OOK-NRZ signal, 238–240 on-off keying and, 240–242 without fluorescent interference, 238–245 Basic illumination pattern, 411 Beam diameter, of laser, 57 Beam divergence, 115–119 arrangement of typical LOS FOS link, 117 beam expander, 118 diffraction-limited beam spreading/geometric loss, 118 fraction of the received power to the transmitted power, 119 free-space path loss, 118 geometric loss, 119 of laser, 57 pointing loss, 119–121 received optical power, 118 source diameter, 117 transmitter and receiver effective antenna gains, 118 Beer-Lambert law, 104–105 Bell, Alexander Graham, 10, 397 Beyond 5G networks, 3 Bias voltage, 44 Bidirectional visible light communication, 403 Big data, 2 Binary phase shift keying (BPSK), 158, 197–198 bit error rate, 200–201 comparison with BPOLSK and OOK, 224 modulated subcarrier, 319–322 Binary PoLSK, 213–217, 224 Binary threshold function, 283 Bit angle modulation (BAM), 427–428 Bit error probability analysis, 318–328 advantages of, 22–23 BPSK-modulated subcarrier, 319–322 DPSK-modulated subcarrier, 325–326 M-ary PSK-modulated subcarrier, 325 multiple SIM performance analysis, 326–328 Bit error rate (BER) of ANN classifier, 447–448 binary PPM BER as a function of scintillation index, 308–309, 309 of binary PPM scheme (Matlab code), 310–311 BPSK modulation in AWGN channel (Matlab codes), 200–201 conditioned on the received irradiance, 360 diffuse configuration and, 19 NOPR and OPP, 233, 242 of OOK-based FSO in atmospheric turbulence, 305, 306 of OOK-NRZ (Matlab codes), 170 of OOK-NRZ using matched filter–based Rx (Matlab codes), 170 optical polarisation shift keying, 217–219 and outage probability of BPSK-SIM, 369–372 and outage probability of DPSK-SIM in negative exponential channels, 372–378

Index performance under different noise conditions (Matlab code), 322–324 of PLED-VLC system with the FPGA-based LMS equaliser, 447 and Q-factor performance as a function of data rate, 446, 446–447 in terms of the SNR and energy per bit Eb to noise power N0 ratio, 198 unequalised NOPR for OOK, 270–271 vs. link margin for a one-dimensional cellular VLC system, 452 vs. SNR for OOKNRZ/RZ, 275–276, 276 vs. SNR for optical OFDM with 16-QAM, 207, 208 vs. SNR for SISO-FSO over clear, weak, moderate, and strong turbulence regimes with and without aperture averaging., 352, 352 Blocking, 24 Bluetooth, 5, 5, 453 Boltzmann statistics, 54 Boltzmann’s constant, 42, 72, 422 BPSK-modulated subcarrier, 319–322 Bragg grating, 60 Broadband over Power Line (BPL), 6 Broadband RF, 4 Broadband wireless services, 1

C Cable-based modem technologies, 4, 5 Carrierless-amplitude and phase (CAP) modulation, 209–211 block diagram of, 210 multiband, 210–211, 212 Cauchy inequality, 357 C-band, 40 ceiling bounce model, 94–95, 268 Cellular communication backhaul, 24 Cellular OWC system, 17, 17–18 Center for Devices and Radiological Health, 26 Central station (CS), 8 Channel DC gain, 414–415 Channel delay spread, 91, 418–419 Channel fading, 347 Channel impulse response (CIR), 81, 95, 416 Channel modelling, 81–150 artificial light interference, 96–104 indoor optical wireless communication channels, 81–96 outdoor channel, 104–150 Channel state information (CSI), 437 Circular POLSK (CPoLSK), 222 Class 1-4 lasers, 26, 27 classes, 27 Cloud, 2 Coherent detection, 68–69, 69 Coherent schemes, 68, 68 Coifman, Ronald, 249 Common mode rejection ratio (CMRR), 69 Computer vision, 454 Conditional probability, 349 Continuous wavelet transform (CWT), 253–255 of non-stationary and stationary signals, 254 time-frequency representation of, 255

493

Index Convection fog, 108 Copper cable-base technology, 4 Correlation method, 441 Costa loop, 213 Costas-loop based PLL circuit, 221 Critical angle loss, 48 Cumulative density function, 359

D Dark current, 63–64 Dark-current shot noise, 73–74 Data traffic, 1, 2, 2 Daubechies, Ingrid, 249 DC gain, 414–415 Decision feedback equaliser (DFE), 281, 281 Decomposition, 256 Denoising, 256–261 Dense WDM services, 24 Desert model, 107 Detectability of laser, 57 Dicode PPM, 176 Differential amplitude pulse position modulation (DAPPM), 176, 178 Differential circle polarisation shift keying (DCPoLSK), 221–223 Differential phase shift keying (DPSK), 211 Differential pulse position modulation (DPPM), 178 Difficult terrains, 24 Diffuse configuration, 19, 19–21 Digital audio broadcasting (DAB), 201 Digital filters, 278 Digital phase locked loop (DPLL), 160 Digital pulse interval modulation (DPIM), 158, 162, 178–181 demodulator, 185 error performance, 183–188 in fixed throughput system, 186 link performance in multipath propagations, 274 mapping of source data, 179 multilevel, 192–193 with no guard band, 186–187 normalised optical power requirement, 235–238 with one guard slot, 187–188 probability of slot error for, 244 PSD of, 182 sequence generation, 179 symbol sequence, 180 types of error in, 183 variable symbol length in, 184 Digital subscriber line (DSL), 4, 201 Digital television (DTV), 201 Digital-to-analogue converter (DAC), 202, 203–205, 210 Dimming factor, 429 Direct detection, 68, 68 Direct intensity modulation, 62–63 Directed LOS, 16–18 Directional coupler, 215 Disaster recovery, 24 Discrete Fourier transformation (DFT) channel, 441 Discrete multitone modulation (DMT), 158, 203, 430–432 Discrete wavelet transform (DWT), 255–256 comparative study of HPF and, 261–262

denoising, 256–261 Rx in presence of artificial light interference, 257 Discrete-time equivalent system, 267 Discrete-time impulse response, 267 Distributed feedback (DFB) laser, 59–60 Distributed-Bragg-reflector laser diode, 60 Diversity, 347–393 spatial, 350–363 time, 348–350 transmitter diversity in log-normal atmospheric channel, 364 transmitter-receiver diversity in log-normal atmospheric channel, 364–365 Dome LEDs, 45–46, 47 DPSK-modulated subcarrier, 325–326 Drms of a diffuse channel, Matlab codes for, 92–93 Drms values at different receiver positions, Matlab codes for, 419–421 DWT-based denoising, 256–261

E E-band, 40 Edge position modulation (EPM), 163, 176 Edge-emitting LEDs, 46–47, 47 Efficiencies external quantum efficiency, 48–49 internal quantum efficiency, 48 luminous efficiency, 49–50 modulation bandwidth, 50–52 power efficiency, 49 Electrical bandwidth, 51, 52 Electrical domains, 361 Emitted optical field, 219, 222 Emitter radiation pattern (ERP), 87 Enterprise applications, 23 Equal gain combining (EGC), 295, 348, 354–355, 357–358, 365–366 Equalisation, 276–281 ANN parameters for equalisation, 287 ANN-based adaptive equaliser, 286–295 as a classification problem, 282 decision feedback equaliser, 281 minimum mean square error equaliser, 279–281 zero forcing equaliser, 278–279, 279 Erasure error, 173, 183 Erbium doped fibre amplifiers (EDFAs), 477, 478, 478–479, 481 Erbium-doped fibre amplifiers (EDFA), 8 Error probability analysis, 223–224 OOK-NRZ in a multipath channel, 269–270 OOK-NRZ in a multipath channel based on discrete-time equivalent system, 270 Error vector magnitude (EVM), 443 Ethernet, 7 European Committee for Electrotechnical Standardization (CENELEC), 26 European Space Agency, 11 Excess noise, 73–74 External modulation, 63 External quantum efficiency, 48–49 Eye, 10 safety limits for 900 and 1550 nm wavelengths, 28

494 F Fabry-Perot laser, 58–59 False alarm error, 173, 183, 189 Far IR, 40 Fast Ethernet, 7 Fast Fourier transform (FFT), 210, 431 FET channel noise factor, 424 Fibre to the homes (FTTH), 4, 5 Field programmable gate array (FPGA), 444–445, 447 Filtering, 275–276 Finite impulse response (FIR) filters, 209 FIR-based equalisers, 292–294 5G networks, 1, 3 Fluorescent lamps, 99–102, 100–101, 100–103 Fluorescent light interference (FLI) block diagram, 232 effect on OOK, 233–235 with electrical high-pass filtering, 246–249 matched filter RX, 231–236 Matlab codes for simulation model, 103–104 with no electrical high-pass filtering, 230–236 normalised optical power requirement, 230 optical power penalty, 230–231 time waveform, 264 wavelet-based denoising for OOK-NRZ in the presence of, 259 Fog, 105, 107–115 advection, 110 convection, 108 formation of, 108 maritime, 110 particle size distribution, 110, 110 radiation, 108 received optical density in presence or absence of, 138 time profiles, 114 Forward error correction, 426, 431, 434, 447, 448 Foundation codes, 348 4G networks, 1, 4, 469 Free-space optical (FSO) communications, 1, 22 all-optical amplify-and-forward, 476–479 all-optical AOAF relay-based FSO with turbulence, 482–486 all-optical regenerate-and-forward, 479–482 applications, 23 defined, 470 history, 11 link performance with atmospheric turbulence, 301–347 loss mechanisms in, 104–106 market share, 13 network topologies, 470–472 relay-assisted, 472–476 terrestrial, 23 wireless networks, 469–472 Free-space optical (FSO) links, 9–10 atmospheric turbulence-induced penalty, 335–338 hybrid RF-FSO scheme, 386–389 on-off keying, 301–305 performance with atmospheric turbulence, 301–304 pulse position modulation, 307–309 SIM-FSO performance in gamma-gamma and negative exponential atmospheric channels, 332–334, 334, 368–378

Index SIM-FSO performance in log-normal atmospheric channel, 314 SIM-FSO with spatial diversity in log-normal atmospheric channel, 365–368 subcarrier intensity modulation, 311–335 subcarrier time delay diversity, 378–386 Frequency bandwidth, 477 Frequency division multiplexing (FDM), 201, 201 Fresnel loss, 49 Fresnel zone, 122, 131 Frozen-flow hypothesis, 123

G Gallium arsenide phosphide, 397 Gamma-gamma pdf, Matlab codes for, 131 Gamma-gamma turbulence model, 131–135 Gauss-Hermite approximation, error performance of short-range links, 380 Gauss-Hermite integration, 359, 344 Gaussian atmospheric optical channel, 304–305 Gaussian beam wave, 385–386 Gaussian distribution, 126 Gaussian noise, 238 Giga-IR, 5 Global data traffic, 1, 2, 2 Global positioning system (GPS), 452–453 Greenwood model, 124 Grossman, Alex, 249 Gurvich model, 124

H Haar, Alfred, 249 Hamming distance, 177 Hard decision decoding (HDD), 174–175, 235–236, 246–248 Hayasaka-Ito model, 95 Haze, 105 Heisenberg uncertainty principle, 253 Helmholtz theory, 111 He-Ne laser, 54 Heterodyne detection, 70–71 Heterogeneous optical networking (HON), 470 High definition television, 24 Highly nonlinear fibres (HNLF), 481, 482 High-pass filter (HPF), 159 artificial light interference, 98 BLW without fluorescent light interference, 238–242 comparative study of DWT and, 261–262 effect in OOK-NRZ signal, 238–240 fluorescent light interference with, 246–249 Holographic diffuser, 421, 421–422 Home access networks, 449–452 one-dimensional optical wireless access network, 450, 451, 452 technologies in, 449, 450 Homodyne detection, 71 Hufnagel-Valley (H-V) model, 123 Hufnagel-Valley (H-V)-Night model, 124 Hybrid cellular and NLOS tracked systems, 22, 22 hybrid OWC/MMW (mmW), 14 Hybrid RF-FSO scheme, 386–389

495

Index I

L

IEC 60825-1 standard, 26, 27 IEC 825 series, 26 IEEE 802.11 a/b/g/n/ac (Wi-Fi), 5, 5–6 IEEE 802.11 standards, 7 Impulse response, 88 Incandescent lamp, 99, 99, 99 Indoor localisation, 452–455 Indoor optical wireless communication system, 81–96 block diagram of, 82 ceiling bounce model, 94–95 effect of ambient light sources on, 229–230 equivalent baseband model, 82 fluorescent light interference with no electrical high-pass filtering, 230–236 Hayasaka-Ito model, 95 LOS propagation model, 84–87 non-LOS propagation model, 87–94 schematic diagram of link, 263 spherical model, 96 Indoor system performance analysis, 229–295 baseline wander without fluorescent interference, 238–245 effect of ambient light sources, 229–230 fluorescent light interference with electrical high-pass filtering, 246–249 link performance in multipath propagations, 265–274 mitigation techniques, 275–281 wavelet analysis, 249–265 Infrared (IR), 9, 39–40 vs. VLC and RF, 404 Infrared communications (IRC), 449–450 Intensity modulation (IM), 9 analogue, 160–162 direct, 62–63 subcarrier. See Subcarrier intensity modulation (SIM) Intensity modulation-direct detection (IM-DD), 68, 68, 81, 157–224, 454 Interchannel interference (ICI), 207, 441 Intermodulation distortion (IMD), 326 Intermodulation products (IMP), 431 Internal quantum efficiency, 48 International Electrotechnical Commission (IEC), 26 Internet of things (IoT), 1, 2 Inter-symbol interference (ISI),, 445 Intersymbol interference (ISI), 18, 82, 158 Intra-symbol frequency-domain averaging (ISFA), 441 Inverse fast Fourier transform (IFFT), 210, 430–431 Inverse source coding (ISC), 426 IrDa, 5

LabVIEW script, 445 Lambertian emission, 409 Lambertian intensity distribution, 46 Lambertian radiation, 408, 444 Laser, 53–63 basic semiconductor laser structure, 57–58 classes, 26 distributed feedback, 59, 59–60 Fabry-Perot, 58–59 operating principle, 53–54 optical feedback, 55 oscillation, 55–56 population inversion, 54–55 properties, 56–57 specifications, 56–57 structure of, 58–61 superluminescent diodes, 61 vertical cavity surface emitting laser, 60–61 Laser cavity, 55 Laser diode, 57–58 Laser Institute of America (LIA), 26 Lasers, classification of, 26–27, 27 Last mile access, 1, 2, 3, 5, 8, 9, 14, 18, 23, 387, 469 L-band, 41 Least mean-squares (LMS), 444 Least square (LS) method, 441, 442, 443 Levenberg-Marquardt backpropagation algorithm, 447 Lifetime, of laser, 57 Li-Fi, 5, 9, 397 Light shaping diffuser (LSD), 421–422 Light sources, 39–42 Light-emitting diodes (LEDs), 42–53 common materials and optical radiation wavelengths, 45 current–voltage relationship of, 44 dome, 45–46, 47 edge-emitting, 46–47, 47 efficiencies, 48–52 external quantum efficiency, 48–49 internal quantum efficiency, 48 versus laser diodes, 62, 62–63 luminous efficiency, 49–50, 51 modulation bandwidth, 50–52 organic, 398 Osram Ostar white-light, 406 Osram Ostar WPLED, 407 planar, 45–46, 46 polymer, 443–444, 445 power efficiency, 49 resonant cavity, 398 RGB (red-green-blue), 52–53, 398 spectral width, 43 structure of, 45 thermal effects, 53 ultraviolet (UV)–based white LEDs, 52–53 white, 52–53 white phosphors, 52–53, 398–399, 405–406, 407 Linear combining techniques, 354–356 Line-of-sight (LOS) directed, 16–18 hybrid system, 18 non-directed, 18, 19 propagation, 6, 84–87

J JPEG2000 standard, 250 Junction capacitance, 50–51

K Kalman filter, 454 Kerr nonlinearity, 479 Kim’s model, 112–114, 115, 139 Kolmogorov two-thirds power law, 123 Koschmieder law, 111 Kruse model, 111

496 Link configurations (OWC), 14–23 diffuse configuration, 19–21 directed LOS, 16–18 factor affecting performance of, 24–25 outdoor, 24 tracked systems, 21–23 transmitters and receivers, 17 Link performance in multipath propagations, 265–274 DPIM, 274 OOK, 265–272 PPM, 272–273 Local area networks (LANs), 7 Local oscillator, 215 Logical domain, 361 Log-normal pdf, Matlab codes for, 128 log-normal turbulence model, 125–129, 131 Log-sigmoid function, 283 Long wavelength IR, 40 Long-term evaluation (LTE), 201 LOS channel gain, Matlab codes for, 86–87 Low-density parity-check (LDPC) code, 349 Low-pass filter (LPF), 210, 215, 222, 312, 445 L-PPM symbol, 171–172 Luby Transform (LT) code, 349 Luminous efficiency, 49–50, 51 Luminous intensity, 408

M MAC layer, 449–450 Machine-to-machine (M2M), 1 Mach-Zehnder interferometry (MZI), 213 Mach-Zehnder Modulator (MZM), 160–161 Macro-assisted small cells, 3 Mallat, Stephane, 249 Maritime fog, 110 Maritime model, 107 Mars Laser Communication Demonstration, 11 M-ary PSK-modulated subcarrier, 325 M-ASK, 313 Massive MIMO, 3 Matched filter RX, 231–236 Matched filters (MFs), 221 Matched-filter combining, 295 Matlab codes analytical PSD of OOK-NRZ, 165 analytical PSD of OOK-RZ, 165–166 analytical PSD of PPM, 173 ANN-based DFE for OOK-NRZ, 289–290 ANN-based linear equaliser for OOK-NRZ, 288–289 bandwidth requirements and power efficiency of different modulation schemes, 195 BER of BPSK modulation in AWGN channel, 200–201 BER of OOK-NRZ, 169–170 BER of OOK-NRZ using matched filter–based Rx., 170 BER of the binary PPM scheme, 310–311 BER performance under different noise conditions., 322–324 ceiling bounce model and received signal eye diagram, 268 for determining the NOPR for DPIM, 237–238 DPIM sequence, 179 Drms of a diffuse channel, 92–93

Index Drms values at different receiver positions, 419–421 effect of FLI on OOK, 233–235 effect of high-pass filter and baseline wander in OOK-NRZ signal, 238–240 error probability of OOK-NRZ in a multipath channel, 269–270 error probability of OOK-NRZ in a multipath channel based on discrete-time equivalent system, 270 FLI simulation model, 103–104 gamma-gamma pdf, 131 Kim’s model simulation, 113 log-normal pdf, 128 LOS channel gain, 86–87 mean and standard deviation for practical measurement, 147–149 MIMO system simulation, 439 modeling of channel with IM/DD, 414 OFDM system simulation, 208–209 optical power distribution of a LOS link at the receiving plane for a typical room, 413 optical power distribution of first reflection at the receiving plane for a typical room, 416–418 plot distribution OOK due to the BWL effect, 240–242 plotting optical power distribution in diffuse channel, 90–91 plotting the turbulence-induced SNR penalty, 336–337 PSD of DPIM(0GS), 181–183 PSD of OOK (NRZ and RZ) by simulation, 166 pulse position modulation generation, 172–173 SER of DPIM in AWGN channel, 190–192 SER of PPM based on HDD and SDD, 175–176 values of α and β under different turbulence regimes., 133–135 wavelet-based denoising for OOK-NRZ in the presence of FLI, 259 Maximum likelihood sequence detection (MLSD), 163, 184, 349 Maximum permissible exposures (MPE), 28, 28 Maximum ratio combining, 356–357 Maximum-likelihood combining (MLC), 295 Maximum-likelihood-decoding (MLD), 438 Maximum-ratio combining (MRC), 213, 295, 348 Mean and standard deviation for practical measurement, Matlab codes for, 147–149 Mean delay, 418 Mean delay spread, 88 Mean excess delay, 418 Mean-square error (MSE), 443–444, 447 Mesh topology, 471, 471, 472 Metal semi-conductor metal photodetectors, 65 Meteorological visual range (MVR), 111 Meyers, Ives, 249 M-FSK, 313 Microwave communications, 4 Mid-wavelength IR, 40 Mie scattering, 105, 106, 106–107, 107, 109 Military applications, 24 Millimetre wave technology, 3, 4, 5 Millimetre-wave wireless over optical fibre, 8 Minimum bandwidth requirement, 196 Minimum bit (slot) duration, 194–195 Minimum mean-square error (MMSE) method, 279–281, 438, 441, 443 MIT Lincoln Laboratory, 11

497

Index Mitigation, 275–281 equalisation, 276–281 filtering, 275–276 Mobile backhaul network, 1, 4 Mobile communications bandwidth, 1 speed, 1 Mobile data traffic, 1, 2, 2 Modeling of channel with IM/DD, Matlab codes for, 414 Modulation, 157–224 amplitude modulation (AM), 160 analogue intensity, 160–162 bandwidth, 50–52 bandwidth efficiency, 159 baseband. See Baseband modulations bit angle, 427–428 carrierless-amplitude and phase modulation, 209–211 comparison of different schemes, 162 differential pulse position, 178 digital pulse interval. See Digital pulse interval modulation (DPIM) edge position, 163, 176 external, 63 index/depth, 327 intensity, 9 multi-carrier, 198–211 multilevel DPIM, 192–193 multiple-subcarrier intensity modulation, 199, 199–200 optical polarisation shift keying, 211–219 orthogonal frequency division multiplexing, 201–208 power efficiency, 159 pulse interval, 178–190 pulse position. See Pulse position modulation (PPM) quadrature amplitude, 158, 160, 198, 210, 313, 430 subcarrier intensity, 196–198 transmission reliability, 159 tree, 158 Moore’s law, 1 Morlet, Jean, 249 Morlet wavelet, 253, 253 MPOLSK, 219–221 M-PSK, 313 Multiband-CAP (m-CAP), 210–211, 212 Multi-carrier modulations, 198–211 carrierless-amplitude and phase modulation, 209–211 multiple-subcarrier intensity modulation, 199, 199–200 orthogonal frequency division multiplexing, 201–208 Multilayer feedforward network, 284 Multilayer perceptron (MLP), 444, 447 Multilevel DPIM (MDPIM), 192–193 Multilevel PPM, 177 Multilevel PWM-PPM, 432–433, 433 Multiple PPM (MPPM), 176 multiple SIM performance analysis, 326–328 Multiple subcarriers (MSC), 197 Multiple-input, multiple-output (MIMO), 6, 18, 350, 351 imaging, 436 Matlab codes for simulation of, 439 non-imaging, 436 visible light communication, 434–438, 437 Multiple-input single-output (MISO), 434–435 Multiple-input-single-output (MISO), 350 Multiple-subcarrier intensity modulation (MSIM), 199, 199–200

Multi-pulse pulse position modulation (MPPM), 429. See also Pulse position modulation (PPM) Multi-resolution analysis (MRA), 253, 255 Multi-spot diffusing (MSD) system, 20, 20

N Naboulsi model, 113–114, 115 NASA, 11 Near IR, 39, 40 Near-Earth FSO systems, 11 Negative exponential turbulence model, 135 Network topologies, 470–472 Neuron, 283 Newton-Raphson procedure, 190 Nippon Electric Company, 11 Noise dark-current shot, 73–74 excess, 73–74 figure, 477 Gaussian, 238 photodetection, 72–76 photon, 72 quantum shot, 72–73 relative intensity, 75 signal-to-noise ratio (SNR), 76, 83–84, 150, 161, 229, 275–276, 276, 352, 364, 422–425 thermal, 75, 422 Noise-equivalent power (NEP), 64, 74 Non-directed LOS, 18, 19 Non-directed non-LOS, 19 Nonlinear effects, 479–480 Non-LOS propagation model, 87–94 Non-return-to-zero (NRZ) pulse format, 302 Nonselective scattering, 105 Normalised optical power requirement (NOPR), 230, 236, 237, 243, 244, 245, 251–252, 258, 258, 258–260, 260 for DPIM, 237–238 Nyquist sampling rate, 255

O O-band, 40 On-off keying (OOK), 19, 158, 163–166 block diagram, 167 block diagram of unequalised OOK system, 266 BWL effect, 240–242 comparison with BPOLSK and BPSK-based FSO links, 224 flowchart for simulation of, 232 with forward error correction, 426 FSO link performance with atmospheric turbulence, 301–305 Gaussian atmospheric optical channel, 304–305 link performance in multipath propagations, 265–272 mapping of m-bit data format, 171 NRZ-OOK, 434 performance in AWGN channel, 167–169 in Poisson atmospheric optical channel, 302–304 versus pulse position modulation, 338 versus subcarrier intensity modulation, 338 variable, 426, 427 wavelet analysis, 262–265

498 OOK-NRZ signal, 163–166 bit error rate, 169–170 effect of high-pass filter and baseline wander in, 238 eye diagram, 266 flowchart for simulation of, 232 histogram of, 149, 149 matched filter for, 168, 168–169 multipath distortion, 269 PSD of, 164–166, 165 Q-factor vs. Rytov variance for, 150 transmitted waveforms, 164 OOK-RZ signal, 164, 169 power requirements and duty cycle, 195 PSD of, 164–166, 165 transmitted waveforms, 164 Optical amplification, 476–479 erbium doped fibre amplifiers (EDFAs), 478 frequency bandwidth, 477 noise figure, 477 saturation outpower, 477 semiconductor optical amplifiers, 478 signal gain, 477 Optical atmospheric transmittance, 105 Optical band-pass filter (OBPF), 74, 220, 222, 312, 379, 481–482 Optical bandwidth, 51, 52 Optical depth, 105 Optical detection statistics, 76–78 Optical domain, 361 Optical feedback, 55, 55–56 Optical fibre backup link, 24 Optical modulation index, 203, 313 Optical multiple input multiple output (OMIMO) systems, 21 Optical phase-locked loop (OPLL), 71 Optical polarisation shift keying (PoLSK), 211–219 binary, 213–217, 224 bit error rate analysis, 217–219 block diagram of, 213 circular, 222 differential circle, 221–223 error probability analysis, 223–224 MPOLSK, 219–221 Optical power, 104 Optical power distribution of first reflection at the receiving plane for a typical room, 416–418 of a LOS link at the receiving plane for a typical room, 413 Optical power loss, 119 Optical power penalty (OPP), 230–231, 233, 235–236, 237–238, 244, 246, 247, 248, 258–262, 262, 267, 271, 271–272, 273, 274, 275, 290–291, 292 Optical power requirement (OPR), 246–248, 249, 267 Optical signal-to-noise-ratio (OSNR), 150 Optical sources laser, 53–63 light sources, 39–42 light-emitting diodes, 43–53 Optical transmission system, block diagram of, 160 Optical wireless communications (OWC) advantages of, 9–10 application areas, 23–25

Index brief history of, 10–13, 12 cellular, 17, 17–18 challenges, 29–31 link configuration, 14–23 overview, 3–4 safety and regulations, 25–28 security, 10 spectrum, 8–9, 9 system block diagram, 16 test bed systems, 135–150 vs. RF technology, 13, 13–14 Optical wireless LAN, 22 Optimal combining, 354 Optimum threshold level, 188–190 Organic LEDs (OLEDs), 398 Orthogonal frequency division multiplexing (OFDM), 158, 201–208, 440–443 ACO-OFDM, 202 asymmetrically clipped optical-OFDM, 440 block diagram, 202 channel estimation, 440–443 DC-biased optical-OFDM, 440 equalisations, 440–443 high PAPR reduction techniques, 204–206 Matlab codes for simulation of, 208–209 PAM-DMT, 202 PAPR of, 203–204 pilot signal estimation at the receiver, 206–208 spectra of, 201 synchronisation, 440–443 visible light communication, 440–443, 442, 444 Osram Ostar white-light LED, 406 Osram Ostar WPLED, 407 Outage probability, 328–330 of BPSK-SIM with spatial diversity, 369–372 of DPSK-SIM in negative exponential channels, 372–378 in a log-normal atmospheric channel, 330 in negative exponential model atmospheric channels, 334–335, 335 Outdoor channel, 104–150 atmospheric channel loss, 104–107 atmospheric effects on OWC test bed, 135–150 atmospheric turbulence models, 121–135 beam divergence, 115–119 fog and visibility, 107–115 optical and window loss, 119 Outdoor optical wireless communications (OWC), 347–393 aperture averaging, 384–386 challenges, 347 hybrid RF-FSO scheme, 386–389 spatial diversity, 350–363 terrestrial free-space optical links with subcarrier time diversity, 350–363 time diversity, 348–350 transmitter diversity in log-normal atmospheric channel, 364 transmitter-receiver diversity in log-normal atmospheric channel, 364–365 Output power, of laser, 57 Overlapping pulse position modulation (OPPM), 429

Index P Packet error rate, 184, 185, 233 Packet transmission rate, 195 Parallel relaying, 474, 474–475 Parallel-to-serial converter, 202 Parasitic capacitance, 50–51 Particle size distribution, 110 Passive optical networks (PONs), 5 Peak-to-average power ratio (PAPR), 158, 196, 203–204, 204–206, 207, 431 Phase shift keying (PSK), 197–198, 211 Photoconductors, 65 Photocurrent, 63 Photodetection coherent detection, 68–69, 69 defined, 67 direct detection, 68, 68 heterodyne detection, 70–71 homodyne detection, 71 techniques, 67–71 Photodetection noise, 72–76 background radiation, 74 dark-current shot noise, 73–74 excess noise, 73–74 quantum shot noise, 72–73 relative intensity noise, 75 signal-to-noise ratio (SNR), 76 thermal noise, 75 Photodetectors, 63–67 avalanche, 65, 67 dark current, 63–64 metal semi-conductor metal, 65 noise-equivalent power, 64–65 operating wavelength ranges for different materials, 66 optical feedback, 55–56 performance and compatibility requirements, 65 performance characteristics of, 67 photoconductors, 65 PIN, 65–67 speed of response and bandwidth of, 64 transfer function of, 64 types of, 65 Photon noise, 72 Photons, 41–42 Photophone, 397 PIN photodectors, 65–67 Planar LEDs, 45–46, 46 Planck constant, 42 Planck’s constant, 72 Plane wave, 384–385 plane wave transverse coherence length, 130 Plotting optical power distribution in diffuse channel, Matlab codes for, 90–91 Plotting the turbulence-induced SNR penalty, Matlab codes for, 336–337 Poincaré sphere, 217 Pointing loss, 119–121 Point-to-point topology, 471, 471 Poisson atmospheric optical channel, 302–304 Poisson distribution, 76, 77 Polarisation beam splitter (PBS), 222 Polymer LEDs (PLEDs), 443–444, 445

499 Population inversion, 54 Power efficiency, 49, 159, 193–194 Power line communications (PLC), 449–450 Power spectral density (PSD) of DPIM(0GS), 181–183 of OOK-NRZ signal, 164–166, 165 of OOK-RZ signal, 164–166, 165 of pulse interval modulation, 182 of pulse position modulation, 173 of quantum shot noise, 72 of refractive index fluctuation, 123 PPM generation, Matlab codes for, 172–173 PPM plus (PPM+), 176 Probability density function (PDF), 77, 121, 188–190, 189 Probability of slot error, 243, 244, 246 Probability of symbol error, 243, 246 Propagation line-of-sight (LOS), 84–87 multipath, 265–272 nonl-light-of-sight, 87–94 single-reflection, 89 Proximity algorithms, 454 Pseudorandom binary sequence (PRBS), 444 Pseudo-random binary signal (PRBS), 145 Public telephone networks, 4 Pulse amplitude modulation (PAM), 158, 163, 163–166, 428 Pulse interval modulation (PIM), 178–181, 178–190 demodulator, 185 error performance, 183–188 in fixed throughput system, 186 link performance in multipath propagations, 274 mapping of source data, 179 multilevel, 192–193 with no guard band, 186–187 normalised optical power requirement, 235–238 with one guard slot, 187–188 optimum threshold level, 188–190 probability of slot error for, 244 PSD of, 182 symbol sequence, 180 types of error in, 183 variable symbol length in, 184 Pulse position modulation (PPM), 19, 158, 171–178, 235–236 differential, 178 differential amplitude, 178 error performance, 173–176 FSO link performance with atmospheric turbulence, 307–309 hard decision decoding (HDD), 246–248, 259–260, 260, 260 link performance in multipath propagations, 272–273 L-PPM symbol, 171–172 multilevel, 177 multilevel PWM-PPM, 432–433, 433 multi-pulse, 429 versus on-off keying, 338 overlapping, 429 probability of slot error for, 243, 246 probability of symbol error for, 243 soft decision decoding (SDD), 246–248, 259, 260 versus subcarrier intensity modulation, 338 variants, 176–178

500 Pulse time modulation (PTM), 162 Pulse width modulation (PWM) defined, 162, 405 with discrete multitone modulation, 430–432 with NRZ-OOK, 434 Pumping, 54

Q Q-factor, 150, 264–265, 265 QPSK, 314, 315, 316, 444 Quadrature amplitude modulation (QAM), 158, 160, 198, 210, 313, 430 Quantum efficiency, 63 Quantum shot noise, 72–73

R Radiation fog, 108 Radio frequency (RF), 2, 5, 311, 449–450, 469 coherent detection in, 69 IR and VLC, 404 MIMO channels, 435 vs. optical wireless communications, 13, 13–14 Radio-over-fibre (RoF) system, 2, 8, 470 Radio-over-free space optic (FSO), 2 Rain, 105, 107 loss for, 115 RaptorQ codes (RQCs), 349 Rateless codes (RC), 348 Rayleigh criterion, 87 Rayleigh distribution, 135 Rayleigh scattering, 105, 106, 106 Real outdoor atmosphere (ROA) channels, 135 Real outdoor fog (ROF), 138–140 Received optical signal, 219, 222 Received signal eye diagram, 268 Received signal strength indicator (RSSI), 452 Recurrent neural network, 285 Reed-Solomon (RS) codes, 178, 348 Relative intensity noise (RIN), 75 Relay-assisted FSO communications, 469–486, 472–476 all-optical, 475–476 diagram of, 473 parallel relaying, 474, 474–475 serial relaying, 474, 474 Resonant cavity LEDs (RCLEDs), 398 Resonator, 55 Return-to-zero (RZ) pulse format, 302 RF identification (RFID, 453 RF wireless LANs, 7 RGB (red-green-blue) LEDs, 52–53, 398 Riccati equation, 126 Ring topology, 471 Root mean square delay spread, 88 Root mean-square (RMS) delay spread., 418–419 Root-raised cosine (RRC) filters, 209 Round, Henry Joseph, 397 Rural model, 107 Rytov approximation, 131 Rytov parameter, 126, 133, 134 Rytov transformation, 125 Rytov variance, 149

Index S Satellite communications, 5 Saturation current density, 44 S-band, 40 Scattering, 105–107, 106 Scattering efficiency, 108 Scene analysis, 454 Scintilllation, 355 in atmospheric turbulence, 121 effect on data carrying radiation, 145–150 effects of, 23 index, 131, 139 Selection combining, 348, 354–355, 359–360 Self-phase modulation (SPM), 479–480 Semiconductor laser diode (SLD), 57 Semiconductor optical amplifiers (SOAs), 477, 478–479 Semiconductor-laser Inter-satellite Link Experiment, 11 Sensor signal processing (SSP), 424 Serial relaying, 474, 474 Shadowing, 24 Shockley diode equation, 44 Short wavelength IR, 39, 40 Short-range wireless technologies, 404 Short-term Fourier transform (STFT), 253 time-frequency representation of, 255 Shot noise variance, 422 Shot-noise current, 72 Shout noise power, 477 Sigmoid function, 283 Signal fading, 347 Signal gain, 477 Signal jamming, 22 Signal-to-interference plus noise ratio (SINR), 425 Signal-to-interference ratio (SIR), 424 Signal-to-interference-and-noise ratio (SINR), 162 Signal-to-noise ratio (SNR), 76, 83–84, 150, 161, 229, 275–276, 276, 352, 364, 422–425 Silicon carbide (SiC) semiconductor, 397 Single-input-multiple-output (SIMO), 350, 351 Single-input-single-output (SISO), 350, 352–353, 360, 435 Single-layer feedforward network, 284 Single-reflection propagation model, 89 60 GHz band, 2 Slot error rate (SER), 184, 185 of DPIM in AWGN channel, 190–192 of PPM based on HDD and SDD, 175–176 Slot rate, 195 Snow, loss for, 115 Soft decision decoding (SDD), 174–175, 184, 235–236, 246–248 Solid-state lighting (SSL), 397–399 Space division multiplexing (SDM), 2 Space-time block code (STBC), 348 spatial coherence in weak turbulence, 129 Spatial diversity, 347, 350–363 adaptive optics schemes, 353–354 BER and outage probability of BPSK-SIM with, 369–372 block diagram of receiver with N detectors., 355 effect of received signal correlation on error performance, 360–362 equal gain combining, 357–358 linear combining techniques, 354–356

501

Index in log-normal atmospheric channel, 365–368 maximum ratio combining, 356–357 outage probability with receiver diversity in lognormal atmospheric channel, 362–363 schemes, 351 selection combining, 359–360 Spectral power distribution (SPD), 82 Spectrum bottleneck, 3, 3 Spherical model, 96 Spherical wave, 385 Spontaneous emission, 41, 42, 43 State of polarisation (SOP), 212–213, 216–217 Stimulated Brillouin amplifiers (SBAs), 477 Stimulated emission, 41, 42 Stimulated Raman amplifiers (SRAs), 477 Stokes parameters, 216, 217 Subcarrier intensity modulation (SIM), 196–198 benefits of, 196–197 BER and outage probability of BPSK-SIM with spatial diversity, 369–372 BER and outage probability of DPSK-SIM in negative exponential channels, 372–378 bit error probability analysis of SIM-FSO, 318–328 block diagram of FSO, 312 challenges in implementation of, 311 FSO link performance with atmospheric turbulence, 311–335 generation and detection of, 312–314 issues in, 197 multiple SIM performance analysis, 326–328 versus on-off keying, 338 outage probability, 328–330 outage probability in negative exponential model atmospheric channels, 334–335 phase shift keying, 197–198 versus pulse position modulation, 338 quadrature amplitude modulation, 198 SIM-FSO performance in gamma-gamma and negative exponential atmospheric channels, 332–334, 334, 368–378 SIM-FSO performance in log-normal atmospheric channel, 314 SIM-FSO with spatial diversity in log-normal atmospheric channel, 365–368 Subcarrier time delay diversity (STDD), 378–386 block diagram, 379 error performance of short-range links, 380 error performance with, 378–380 long-range links, 380–381, 382 short-range links, 380–381, 381–382 Submarine laser communication-day model, 124 Sunlight, 96 Super-dense meshed cells, 3 Superluminescent diodes (SLDs), 61 versus LEDs, 62 Switch-and-examine combining (SEC), 355 Switch-and-stay combining (SSC), 355 Switched combining diversity, 355, 359

T Tan-sigmoid function, 283 Taylor frozen turbulence hypothesis, 378 Taylor frozen-flow hypothesis, 122–123

Temporary links, 24 Test bed systems, 135–150 block diagram for the measurement of fog attenuation and visibility, 142 block diagram of, 137 calibration, 139–145 measured attenuation against time and visibility, 143 measured attenuation against wavelength, 144 parameters, 138 scintillation effect on data carrying optical radiation, 145–150 time dependence of visibility within the FSO chamber, 143 Thermal noise, 75, 422 3G networks, 4, 469 Time diversity, 348–350 Time division multiplexing passive optical networks (TDM PONs), 4 Time division multiplexing (TDM), 4 Tine waveforms, 239 Tracked systems, 21–23 Transfer function, 283 Trans-impedance amplifier (TIA), 202, 312, 326 Transmission bandwidth requirements, 194–195 capacity, 195 reliability, 159 windows, 40–41 Transmit optical power, 161, 408 Transmit power, 104 Transmittance, 105 Transmitted 8-DPIM(1GS) signal, 183 Transmitter diversity, in log-normal atmospheric channel, 364 Transmitter-receiver diversity, in log-normal atmospheric channel, 364–365 Transversal filter, 278 Triangulation methods, 453 Turbo product code (TPC), 348

U U-band, 41 Ultraviolet, 9 Ultraviolet (UV)–based white LEDs, 52–53 Ultra-wideband (UWB) signals, 6–7, 470 Urban model, 107

V Values of α and β under different turbulence regimes., Matlab codes for, 133–135 Variable on-off keying (VOOK), 426, 427 Vector signal generator (VSG), 444 Vertical Bell Labs layered space-time algorithm (V-BLAST), 438 Vertical cavity surface emitting laser (VCSEL), 60–61, 61 Very long-wave IR, 40 Visibility attenuation coefficient as function of, 116 versus attenuation using Kim and Naboulsi model, 115 versus attenuation using Kruse model, 112 fog and, 107–115 international visibility code, 109 wavelengths and, 116 weather conditions and, 112

502 Visibility of laser, 57 Visible laser diodes (LDs), 402 Visible light, 9 Visible light communication (VLC), 3, 9, 12, 397–456 all organic, 443–449 architecture of, 450 bidirectional, 403 bit angle modulation, 427–428 block diagram of system, 406 challenges and solutions, 402 channel delay spread, 418–419 defined, 397 historical development, 397–398, 401 holographic diffuser, 421, 421–422 home access networks, 449–452 illustration of, 400 indoor localisation, 452–455 key features of, 405 layers, 450 light sources, 407 link, 406, 412 multilevel PWM-PPM, 432–433, 433 multiple-input, multiple-output (MIMO), 434–438, 437 on-off keying with forward error correction, 426, 426 orthogonal frequency division multiplexing (OFDM), 440–443, 442, 444 polymer, 443 potential of, 403 primary functions of, 399–400 pulse modulation schemes, 428–430 PWM with discrete multitone modulation, 430–432 PWM with NRZ-OOK, 434 schematic diagram of the LED control mechanism., 425 SNR analysis, 422–425 system description, 405–425 system implementations, 425–434 system model, 408–416 system parameters for link, 412 vs. IR and RF, 404 Visible light communications consortium (VLCC), 398 Viterbi decoding, 348

W Wave division multiplexing passive optical networks (WDM PONs), 4 Wavelength division multiplexing (WDM), 2, 398

Index Wavelet analysis, 249–265 comparative study of DWT and HPF, 261–262 continuous wavelet transform, 253–255 discrete wavelet transform, 255–256 DWT based denoising, 256–261 OOK, 262–265 Wavelet packet transform (WPT), 250 Wavelet-based denoising for OOK-NRZ in the presence of FLI, Matlab codes for, 259 Weather, 24 visibility and, 112 Weighted inter-frame averaging (WIFA), 441, 443 White LEDs, 52–53 White phosphors (WP) LEDs, 52–53, 398–399, 405–406, 407 White Wi-Fi, 5 Wickerhauser, Victor, 249 Wi-Fi, 5, 5–6, 404, 453 Wi-Fi Alliance, 6 Wi-Gig (IEEE 802.11ad), 5, 6 Wilkinson’s method, 380 WiMax, 6 Window loss, 119 Wireless access schemes, 4–10 Wireless Gigabit Alliance, 7 Wireless local area network (WLANs), 7, 201 Wireless metropolitan area networks (WMANs), 201 Wireless networks, 469–472 Wireless personal area networks (WPANs), 7 Worldwide Interoperability for Microwave Access (WiMax), 6 Wrong slot error, 174, 183

X xDSL, 5 Xilinx ISE software, 445

Z Zero forcing equaliser, 278–279 structure of, 279 Zero-forcing (ZF) algorithm, 438, 442 Zigbee, 5, 453