Optical Wireless: an emerging
solution for the next-generation
communications
D265ModTPhD/EN01
Phd Course Emerging Digital Technologies
ISBN: XXXXXXXXXXXX
Optical Wireless: an emerging
solution for the next-generation
communications
Author
Alessandro Messa
Supervisor
Contents
List of Figures xv
List of Acronyms xxi
Abstract xxiv
1 Optical Wireless: Devices and Communications Systems 1
1.1 Optical systems categories . . . 2
1.1.1 Optical Wireless Communication (OWC) link configurations 4 1.2 Optoelectronic devices . . . 6
1.2.1 Optical sources . . . 6
LED . . . 6
Electrical characteristics . . . 8
Optical properties . . . 11
Vertical Cavity Surface Emitting Laser (VCSEL) . . . 13
1.2.2 Receivers . . . 15
positive-intrinsic-negative (PIN) Photodiode . . . 15
Avalance Photodiode . . . 17
1.3 Optical channel modeling . . . 18
1.3.1 Transmitter model . . . 19
1.3.2 Receiver model . . . 20
1.3.3 Trasmission medium model . . . 21
1.3.4 Optical noise . . . 21
1.4 Modulation techniques . . . 22
1.4.1 On-Off Keying (OOK) . . . 23
1.4.2 DC-balanced . . . 23
1.4.3 Pulse Modulation (PM) . . . 23
1.5 OWC applications overview . . . 24
2 Analog equalization for high speed OWC 29 2.1 Analog equalization . . . 30
2.1.1 Bridged-T amplitude equalizer . . . 30
2.2 Discrete Multitone modulation . . . 33
2.2.1 Discrete Multitone (DMT) modulation and demodulation schemes . . . 33 2.2.2 DMT design considerations . . . 34 Number of subcarriers . . . 34 Channel equalization . . . 35 Adaptive algorithm . . . 36 Cyclic Prefix (CP) . . . 36 System synchronization . . . 36 2.3 Experimental setup . . . 37 2.4 Results . . . 39
3 High Throughput OWC for Data Center 43 3.1 State of the art . . . 43
3.2 System Design . . . 46
3.3 Experimental Results . . . 49
3.3.1 10 Gbit/s OWC link . . . 49
3.3.2 24 Gbit/s OWC link . . . 52
3.3.3 40 Gbit/s OWC link . . . 55
4 High Energy Physics applications 59 4.1 Introduction . . . 59
4.2 Particle interaction with matter . . . 62
4.3 Communication System Design . . . 63
4.4 Proton Irradiation Experiment . . . 64
4.5 Results . . . 68
5 WDM transmission with a single multi-colour Photo-diode 71 5.1 Introduction . . . 71
5.2 Triple Junction (TJ) photodiode . . . 73
5.3 Experimental setup . . . 76
5.5 Matlab Routine Algorithms . . . 81 5.5.1 System characterization . . . 82 5.5.2 Transmission algorithm . . . 83 5.6 Results . . . 86 6 Conclusions 91 A Noise equations 95 A.1 Shot Noise . . . 95
A.2 Thermal noise . . . 97
List of Publications 99
List of Figures
1.1 OWC system model. . . 2
1.2 OWC applications categorization related to transmission range, bit rate and coverage area. The graph is intended as a qualitative reference. All the reported values are an approximation of real
values. . . 3
1.3 Types of OWC link configurations. . . 4
1.4 Semiconductor p-n junction behavior in different polarization
con-ditions where V0 is the built-in voltage at open circuit, VF is the
applied voltage in forward polarization and VR is the applied
volt-age in reverse polarization. The blue curves represent the voltvolt-age
along the length of the semiconductor x. . . . 7
1.5 Diode i − v curve displaying the three main area of operation:
breakdown in red, reverse bias in blue and forward bias in green. . 8
1.6 Electrical characteristics taken from LXZ1-6565 LUXEON Z Light
Emitting Diode (LED) datasheet. . . 9
1.7 Normalized frequency response of white and blue light emitted by a Luxeon Star LED. . . 10 1.8 Optical power spectral density of the two main configurations to
obtain white light from LEDs. . . 11 1.9 Sharpe, Stockman, Jagla & Jägle (2005) 2-deg V*(l) luminous
ef-ficiency function. . . 12 1.10 Normalized radiant intensity pattern represented in in polar
coor-dinates, the dashed lines represent the 120◦ emission angle. . . 13
1.11 Simple diagram of a VCSEL cross-section (Not to Scale dimension). 14 1.12 VCSEL L-I-V curve. . . 15 1.13 PD equivalent model + i-v curve as a function of incident light. . 16 1.14 Linear baseband transmission model of an OWC link. . . 18
1.15 Scheme of free-space propagation. . . 20 1.16 Optical power spectra of the three main noise sources. . . 21 2.1 Zobel’s network as a balanced bridge drown in two equivalent
rep-resentations. . . 31
2.2 Zobel’s network represented in the bridged-T shape configuration. 31
2.3 The dual networks used in the formation of the bridged-T network. 32 2.4 Analog equalization circuit: (a) circuit schematic; (b) picture of
the realized equalizer on a PCB. . . 32 2.5 Transfer functions of the wireless channel with or without
equal-ization and the implemented pre-emphasis function. . . 33 2.6 DMT signal processing schematic diagram. . . 35 2.7 Experimental set-up. AWG: Arbitrary Waveform Generator; RTO:
Real-Time Oscilloscope; Amp: amplificatory stage; FPGA: Field Programmable Gate Array; PD circuit: Pre-distortion circuit; TX: infrared LED; RX: APD module. . . 38 2.8 Estimated SNR both for RT and OL transmission at r = 0 cm. . . 39 2.9 Preliminary channel evaluation for bit and power loading technique. 40 2.10 Constellation diagrams of real-time and off-line transmission at r
= 0 cm and r = 100 cm. . . 40 2.11 Achievable bit-rate in real-time and off-line transmission at fixed
BER = 2 × 10−3 as a function of the radial distance. . . 41
2.12 Achievable bit-rate as a function of the received optical power. . . 41 3.1 Polygonal rack distribution in a Optical Wireless Cellular (OWCell)
Data Center Network (DCN). . . 46
3.2 Optical Wireless Communication System for DCN physical layer. 47
3.3 High Throughput (HT) OWC link setup realized in our laboratory. 48 3.4 Simulation setup and graphical representation of the tolerance to
misalignment (D). . . 50
3.5 Simulated PRX curves as a function of ρ at various distances.
Hor-izontal black line indicates the sensitivity level at 10 Gbit/s. Dif-ferent curves were taken for difDif-ferent d values: (a) 1.5 m; (b) 2 m; (c) 3 . . . 50
3.6 Simulated and measured PRX values as a function of ρ. The curves were taken for optimal d value at each distance: (a) 1.5 m; (b) 2
m; (c) 3 m. . . 51
3.7 Eye diagrams of the 10 Gbit/s signal on-axis at various distances: (a) 1.5 m; (b) 2 m; (c) 3 m. . . 51
3.8 BER values as a function of ρ at all the link distances. d was optimized at each link distance: (a) 1.5 m; (b) 2 m; (c) 3 m. . . . 52
3.9 BER values as a function of ρ at all the link distances. d was optimized for 3 m distance and used also at 1.5 and 2 m link distances: (a) 1.5 m; (b) 2 m; (c) 3 m. . . 52
3.10 SNR of the OWC system taken at ρ = 0 mm and at ρ = 4.5 mm. 53 3.11 Corresponding bit (a) and power (b) loading in the two significant cases of displacement. The bit rates were 24 Gbit/s and 10 Gbit/s, respectively. . . 53
3.12 (a) and (b) constellation diagrams of all the subcarriers at ρ = 0 mm and ρ = 4.5 mm, respectively. . . 54
3.13 (a) Maximum bit rate as a function of the radial displacement; (b) Received SNR of the data transmission after bit/power loading. . 55
3.14 12.5 Gbit/s transmission: (a) Bit loading; (b) BER measurements. 55 3.15 SNR curves for ρ = and ρ =1 mm. . . 56
3.16 40 Gbit/s transmission: (a) bit-loading allocation for ρ = 0 and ρ = 1 mm; (b) power-loading allocation for ρ = 0 and ρ = 1 mm. . 56
3.17 Overall constellation diagrams (including all subcarriers) for the two cases. . . 57
3.18 40 Gbit/s transmission: (a) Bit-rate as a function of transverse distance ρ; (b) Bit-error ratio vs. transverse distance ρ . . . 57
4.1 Schematic overview of the Compact Muon Solenoid (CMS). . . 60
4.2 CMS Tracker Optical Readout and Control Systems. . . 61
4.3 Optical fibers management outside the first patch pannel. . . 61
4.4 60 GHz Radio Frequency (RF) transmission system. . . 62
4.5 OWC system experimental setup at 10 Gbit/s and 20 cm distance. 64 4.6 (a) Received signal’s eye diagram ρ=0 (b) BER as a function of radial misalignment. . . 64
4.8 The Irradiation (IRRAD) facility inside the Conseil Européen pour la Recherche Nucléaire (CERN) accelerator complex. The irradi-ation chamber with the cool box. The red arrows represent the
proton beam path. . . 66
4.9 Experimental setup for the irradiation test at CERN. Red lines represent the connections for the board under test, here B3. SMU: Source-Meter Unit. . . 67
4.10 (a) Single PCB with VCSEL and photodiode (PD) facing each other; (b) The six boards on the Plexiglas frame to fit the device inside the cool box. . . 67
4.11 IL as a function of time. Different colored areas indicate Area A: cooling ramp; Area B: idle; Area C: p-beam ON. . . 68
4.12 L-I curves of Board 3. . . 69
4.13 PD dark current vs time (B4 to B6). . . 69
4.14 PDs dark current and VCSELs IL as a function of time. Different colored areas indicate Area A: cooling ramp; Area B: idle; Area C: p-beam ON; Area D: cold annealing −21◦C; Area E: temperature ramp −21 ◦C to 20◦C; Area F: warm annealing 20◦C. . . 70
5.1 Typical Wavelength Division Multiplexing (WDM)-Visible Light Communication (VLC) system design. . . 72
5.2 Cross section of a TJ-PD. . . 74
5.3 TJ-PD equivalent electrical circuit. . . 74
5.4 TJ-PD responsivity as a function of wavelength. . . 75
5.5 Electrical bandwidth for each junction of the TJ-PD. . . 76
5.6 Block diagram of the experimental setup. . . 77
5.7 RGB LED white light: (a) Transmitter emission spectrum; (b) Source colour position in the chromaticity diagram. . . 78
5.8 Schematic circuit of the LED driver. . . 78
5.9 Receiver PCB: 3D model of the designed PCB and the real proto-type of the receiver. . . 79
5.10 Communication channel modelled as a MIMO. . . 80
5.12 Characterization of the multicolor PD: the three figures report the
output voltage vector V (Blue data set V1, Green data set V2, Red
data set V3) at different optical power levels when the receiver is
illuminated by P1 (Blue LED), P2 (Green LED), P3 (Red LED) . 86
5.13 (a) Traces of the three outputs of the transimpedance amplifiers (TIAs) connected to the multicolor PD, when illuminated by a WDM RGB signal; (b) Same traces obtained after the MIMO pro-cessing for the three colors. . . 87 5.14 (a) BER vs threshold curves after the MIMO processing blue
chan-nel; (b) Eye diagram of the blue signal in optimal threshold condition. 88 5.15 (a) BER vs threshold curves after the MIMO processing green
channel; (b) Eye diagram of the green signal in optimal threshold condition. . . 88 5.16 (a) BER vs threshold curves after the MIMO processing red
chan-nel; (b) Eye diagram of the red signal in optimal threshold condition. 88 A.1 Block diagram of photodetection . . . 96 A.2 Photodiode frequency response . . . 98
List of acronyms
ADC Analog to Digital Converter
APD avalanche photodiode
AWG Arbitrary Waveform Generator
AWGN Additive White Gaussian Noise
B2B Board to Board
BER Bit Error Ratio
BPSK Binary Phase Shift Keying
CAGR Compounded Average Growth Rate
CAP Carrierless Amplitude Phase
CERN Conseil Européen pour la Recherche Nucléaire
CMS Compact Muon Solenoid
CoR Cell of Rack
COTS commercial off-the-shelf
CP Cyclic Prefix
D- Directed
DAC Digital to Analog Converter
DBR distributed Bragg reflector
DCN Data Center Network
DMT Discrete Multitone
EEL edge-emitting laser
ESD Electrostatic Discharge
EVM Error Vector Magnitude
FCC-ee Future Circular electron-positron Colliders
FDM Frequency Division Multiplexing
FEC Forward Error Correction
FFT Fast Fourier Transform
FOV Field-of-View
FPGA Field Programmable Gate Array
FSO Free Space Optics
GBP gain-bandwidth product
H- hybrid
HEP High Energy Physics
HL High Luminosity
HT High Throughput
ICI Inter-Carrier Interference
IEEE Institute of Electrical and Electronic Engineers
IFFT Inverse Fast Fourier Transform
IM/DD Intensity Modulation with Direct Detection
IoT Internet of things
IR Infra-Red
ISI intersymbol interference
ITU International Telecommunication Union
LASER Light Amplification by Stimulated Emission of Radiation
LD LASER diode
LED Light Emitting Diode
LHC Large Hadron Collider
Li-Fi Light Fidelity
LOS Line-of-Sight
LTI linear time-invariant
MEMS microelectromechanical systems
MIMO Multiple-Input and Multiple-Output
Nd- Non-directed
NLOS Non-Line-of-Sight
NoC network-on-chip
NRZ Non-Return-to-Zero
OCC Optical Camera Communications
OFDM Orthogonal Frequency-Division Multiplexing
OOK On-Off Keying
OPA optical phased array
OWC Optical Wireless Communication
OWCell Optical Wireless Cellular
OWiNoC optical wireless network-on-chip
PCB Printed Circuit Board
PD photodiode
PG Pattern Generator
PIC photonic integrated circuit
PIN positive-intrinsic-negative
PM Pulse Modulation
PPM Pulse Position Modulation
PRBS Pseudorandom Binary Sequence
PS Proton Synchrotron
PSK Phase-Shift Keying
QAM Quadrature Amplitude Modulation
RF Radio Frequency
RIN Relative Intensity Noise
RTO Real-Time Oscilloscope
RX receiver
RZ Return-to-Zero
SC Single-Carrier
SCO sampling clok offset
SM Switching Matrix
SMU Source Meter Unit
SNR Signal-to-Noise Ratio
STO symbol time offset
TJ Triple Junction
TX transmitter
UOWC underwater optical wireless communication
UV Ultra-Violet
V2I Vehicle-to-Infrastructure
V2V Vehicle-to-Vehicle
VCSEL Vertical Cavity Surface Emitting Laser
VLC Visible Light Communication
VLCA Visible Light Communication Associaton
VLCC Visible Light Communication Consortium
WBAN Wireless Body Area Network
WDM Wavelength Division Multiplexing
WiFi wireless fidelity
WPAN Wireless Personal Area Network
Abstract
The new millennium that we are facing is characterized by profound technological and social transformations dictated by a disruptive digital revolution. In this scenario, the telecommunications industry plays a pivotal role in providing the infrastructure for increasingly greedy data services. The main objective of the next generation of communication is to reach everyone everywhere by offering broadband transmission channels. For this reason, wireless devices have spread rapidly, becoming an essential component for the society of the future. Typically, the term “wireless” is associated with RF communication systems both for historical and commercial pervasiveness reasons. But, despite the enormous success, this technology comes up against its biggest limit:
“The Spectrum Crunch”. A finite natural resource like the RF spectrum is not able to
sustain an exponential demand of frequency allocation, in fact governments around the world and the International Telecommunication Union (ITU) are constantly engaged in finding new solutions.
Optical Wireless Communication (OWC) is an emerging solution to these prob-lems. The OWC systems exploit optical radiation as a carrier for signal transmission in free space. The wide spectrum (ranging from Infra-Red (IR) to Ultra-Violet (UV) wavelength) free of licenses, the potential for high transmission capacity and low costs make this technology extremely interesting for the scientific community. In the last decades the OWC has managed to obtain different spaces in the communications sec-tor, proving to be a complementary technology to RF, especially in those applications where RF communications are disadvantaged (interference, security, propagation dif-ficulties, etc.). Furthermore, the diffusion of solid-state lighting systems has been the main driver for the development of VLC (OWC using the visible spectrum), thanks to the Light Emitting Diode (LED) ability of a faster switching compared with conven-tional lighting. All these conditions have shifted the interest also outside the academy. In 2003 the Visible Light Communication Consortium (VLCC) was founded and in 2014 became Visible Light Communication Associaton (VLCA) [1]. Currently the OWC is at the verge of maturity, in April 2019 the Institute of Electrical and Electronic Engi-neers (IEEE) 802.15.7 standard on Short-Range OWC has been published and IEEE 802.15.13 is in developing phase.
In such a dynamic and complex context full of application scenarios, the author intends to present in this manuscript his personal contributions made during the PhD. Presenting cutting-edge applications for next-generation communications. The thesis is organized as follows:
Chapter 1 introduces the basics of OWC systems, providing a general overview of the system architectures, analyzing their various opto-electronic components and functional blocks, modeling the transmission channel and the related noise sources. Then, the main modulation formats used in this manuscript are described. Finally, an overall overview of the OWC applications is provided.
Chapter 2 describes the realization and experimental results of an IR optical con-nection for indoor applications. The proposed Non-directed (Nd-)Line-of-Sight (LOS) OWC system exploits an Orthogonal Frequency-Division Multiplexing (OFDM) mod-ulation technique in combination with an analog equalization in order to achieve a real time 230 Mbit/s transmission.
Chapter 3 reports all the experimental demonstrations and milestones of our High Throughput OWC systems developed for Data Center applications. The chapter presents three OWC systems with different bit rates (10 Gbit/s, 24 Gbit/s, 40 Gbit/s), analyzing the tolerance to misalignment and highlighting how active alignment is not essential whether there is an adequate mechanical precision.
Chapter 4 reports the peculiar OWC application for High Energy Physics (HEP). The results obtained quantify for the first time the opto-electronic components tolerance to a high-energy proton beam irradiation and they outline the way for the next design choices.
Chapter 5 illustrates the work initiated at Vienna University of Technology and finished at Sant’Anna School of Advanced Studies by the author. The carried out research reports, for the first time, a novel VLC WDM system using a single Triple Junction (TJ) photodiode (PD) in combination with a MIMO post-processing.
Chapter 6 summarizes the author’s contributions in this work and it put the obtained results in perspective with respect to the challenges of the future of next-generation communication technologies.
1
Optical Wireless: Devices and
Communications Systems
The most common definition of a transmission system is a chain of elements which allow data to travel within a physical medium between two distinct nodes of a telecommunication network. OWC is a specific case of this definition, the term refers to any optical communication in free space that uses photons as carriers of information. The spectrum portion that is taken into consideration goes from the extreme UV (10 nm) to the far IR (1 mm) offering an unlimited and unlicensed bandwidth (e.g. the visible spectrum alone is around 400 THz of bandwidth). Although OWC systems have numerous types of applications, all of them follow a common scheme, as depicted in Fig. 1.1.
• In the highest level of the model we find the source or the destination of
the information. Usually, in this level the digital signals, coming from the preceding communication layers, are generated or collected.
• The following level deals with the encoding or decoding process of the data.
During transmission, the modulation box converts the incoming data into a new format in order to adapt the signal to the communication channel. While, on the receiver side, the demodulation box performs the inverse operation. This layer is essential to improve the transmission performance in terms of interference, capacity and spectral efficiency.
Figure 1.1: OWC system model.
• At the edge of the transceiver, we have the optoelectronic devices which
transduce the electrical signal into the optic signal and vice versa.
• Finally, on the bottom of this model, we find the transmission medium level.
It defines how the light propagates the free space channel before reaching the destination.
In this chapter all the model layers will be comprehensively treated in the various constituent elements, addressing the specific communication functions and highlighting advantages an disadvantages in comparison with RF devices. Furthermore, a comprehensive overview of the latest OWC technologies will be issued at the end of the chapter in Section 1.5.
1.1
Optical systems categories
The broad number of different OWC systems can be classified into various cate-gories dictated by the main application characteristics.
The first discriminating factor is the transmission environment, where the background light and the transmission medium properties are dominant in the design process. In this case the OWC devices are divided in two groups indoor and outdoor systems.
Another way to classify the application scenarios is by referring to the ser-vices provided by the system. Here, the main drivers are the bit rate and the coverage area. The latter greatly influences the geometry configuration of the OWC link, where the permutation of directionality and visibility generates six different configurations. This topic will be explored in Section 1.1.1.
Figure 1.2: OWC applications categorization related to transmission range, bit
rate and coverage area. The graph is intended as a qualitative reference. All the reported values are an approximation of real values.
Finally, a exhaustive categorization is proposed in [2] where OWC variations are based on the transmission range.
Fig. 1.2 shows a summary of the OWC application in the related cate-gory. In the short range, there are typically point-to-point applications at high bit rate and small coverage. Examples are optical interconnects between chips or Board to Board (B2B) communication or even optical transmission through subcutaneous tissue. Mid-range applications include the so-called Light Fidelity (Li-Fi) technologies and all those optical communication systems for Wireless Body Area Network (WBAN) or Wireless Personal Area Network (WPAN). In the same distance range, but at lower bit rate, we have optical wireless BUS for Internet of things (IoT), Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I) and intra-satellite communication. In addition, other emerging solutions must be mentioned, underwater OWC ( 10 to 100 m distance at tents of Mbit/s) and Optical Camera Communications (OCC) used for indoor positioning using small amount of transmitted data. By increasing the distance, applications go from indoor to outdoor communication. Above the km range we find the terres-trial optical link backhauls and at the extreme distance the space backhauls and inter-satellite communication.
Figure 1.3: Types of OWC link configurations.
1.1.1
OWC link configurations
During the design phase of an OWC system, the first step is to define the link directionality and visibility based on the type of application. The directionality is a qualitative feature defined by the quantitative properties of the transceiver, such as the transmitter (TX) emission angle, the receiver (RX) Field-of-View (FOV) and by the geometry of the propagation. Typically we can define three types of directionality: Directed (D-), Nd- and hybrid (H-). On the other hand, the visibility represents the ability of the electromagnetic field to propagate in the medium with or without obstruction (shadowing) towards the RX. Visibility is therefore classified in two categories: LOS and Non-Line-of-Sight (NLOS).
Fig. 1.3 shows all the possible combinations of those two properties.
1. D-LOS is the geometrical configuration also known as point-to-point link, where the TX/RX normals lie on the same optical path with opposite direc-tions. The light propagates straight to the receiver without any obstruction. Typically, in this configuration, the TX has narrow emission angle and the RX has narrow FOV. Due to this reason, it is the most power-efficient OWC link: de facto all the optical power is constrained in a small portion of space, while the narrow RX FOV rejects most of the background noise.
Often, the optoelectronic components used in this type of OWC system have small active area, thus their bandwidth is very high, in fact they can reach bit rates higher than 1 Gbit/s in direct modulation. However, the main drawback of D-LOS is the low user mobility, this configuration re-quires a fine alignment and can be easily disrupted by obstructions. Due to their communication properties, D-LOS OWC system performs best in outdoor environments, space applications and data-center.
2. H-LOS is the geometrical configuration where the TX normal lies on the optical path towards the receiver, while the RX is oriented in a different direction but with a larger acceptance angle. This system improves the mobility compared to the previous case, although it makes the SNR worse, because the wider FOV enables a signal reception from different angles, but, at the same time, inevitably captures the external background light. Typically, this configuration is used in point-to-point links when a coarse alignment is needed.
3. Nd-LOS is the geometrical configuration where the TX/RX normals do not lie on the same optical path and they aim different directions. The TX has a larger emission that is collected by the RX with a large FOV through an obstacle-free propagation. This link is usually implemented in broadcast topology, when the end-terminals need a wide range of movements. Even though the communication is more robust than others, power budget and SNR can become insurmountable obstacles in the design phase, therefore it is necessary to adopt further technological devices in order to make it work. 4. D-NLOS is the geometrical configuration where the propagation can be partially obstructed by obstacles and the light can reach destination af-ter several reflections. In the directed case, the the propagation path is conveyed towards the receiver in such a way that the normal of the last reflection and the RX normal lie on the same line with opposite directions. The small FOV allows high background light rejection, but this time the systems is less efficient than the LOS. Although the configuration is able to overcome obstacles, it is strongly limited by the power budget, so this is the OWC link that has the most absolute need for a fine alignment.
5. H-NLOS is the geometrical configuration where the propagation can be partially obstructed by obstacle and the light can reach destination after
several reflections. In this case, even if the RX normal is not oriented towards the last reflection direction, a larger FOV grants better robustness to misalignment.
6. Nd-NLOS is the geometrical configuration where the propagation can be partially obstructed by obstacle and the light can reach destination after several reflections. The light emission can be so diffused that radiation can reach destination from any direction, especially if it is combined with a wide acceptance angle at the receiver. This link in known as "diffused" and it is the most robust configuration. Unfortunately, it is, at the same time, the most vulnerable to the multi-path dispersion and background noise.
1.2
Optoelectronic devices
On the front end of OWC model depicted in Fig. 1.1, we have the optoelectronic transceivers level. Here, the optoelectronic components act as an interface be-tween the optical domain and the electrical domain, transducing the photons into electrons and vice-versa. The foundation of the photonic device operation is the generation and recombination of electron-hole pairs. In an OWC system transceiver the TX converts the electric signal in a optical signal, while the RX collects the incoming radiation transforming it into a measurable electrical volt-age.
1.2.1
Optical sources
The optical sources exploits the radiative recombination of electron-hole pairs to emit light. In particular, the optoelectronic devices on the TX side can exploit the spontaneous mechanisms of electroluminescence of the LEDs or the stimulated emission such of the Light Amplification by Stimulated Emission of Radiations (LASERs). Only those TXs used in the research applications will be described in this manuscript.
LED
LEDs are special types of p-n junction diodes which, when polarized directly, al-low the fal-low of current with consequent emission of light. A simple diode consists of a silicon substrate where part of the crystal lattice is doped with acceptor atoms
Figure 1.4: Semiconductor p-n junction behavior in different polarization
con-ditions where V0 is the built-in voltage at open circuit, VF is the applied voltage
in forward polarization and VRis the applied voltage in reverse polarization. The
blue curves represent the voltage along the length of the semiconductor x. (p-junction) and the other part with donor atoms (n-junction). These impurities may or may not make an electron available compared to those required for the chemical bonding. In open circuit conditions, the strong gradient concentration of holes and electrons generates a diffusion current where the electron-hole pair recombination leaves a depletion region at the interface. The inherent potential that is established at the ends of this junction depends both on the doping con-centration and on the temperature. Typically, for silicon, at room temperature,
V0 has a value between 0.6 V and 0.8 V. The diode assumes different behaviors
depending on the polarization applied on its ends, as shown in Fig. 1.4.
• Reverse polarization occurs when the n-junction is connected to a
pos-itive pole and the p-junction is connected to a negative pole. In this con-dition the depletion area becomes wider increasing the potential difference. The diode behaves as an open circuit, the generated current does not flow in the component.
• Breakdown polarizationtakes place when the reverse bias voltage
over-come the breakdown voltage. This condition can trigger the Zener effect and avalanche effect. In both cases the current is enabled to flow with an intensity determined by the external generator while the voltage across the diode remains close to the breakdown voltage. This process is not
necessar-Figure 1.5: Diode i − v curve displaying the three main area of operation:
breakdown in red, reverse bias in blue and forward bias in green.
ily destructive, as long as the maximum value of dissipating power is not exceeded.
• Forward polarization is obtained when the external voltage overcome
the built-in potential barrier. Now, electron-hole pair recombination can be re-established and current is enabled to flow. As a first approximation the driven current increases linearly with the imposed voltage.
Electrical characteristics
Even for what concerns the electrical characteristics, LEDs and diodes share the same behavior. The i − v curve in Fig. 1.5 shows the typical diode current trend as a function of the supply voltage. Usually, only the forward bias area is
consid-ered for LED applications, where the forward voltage Vf wrepresents the threshold
voltage needed to drive enough current to generate photons. This does not mean that LEDs cannot work in the reverse bias and breakdown. In this region they can be used as receiver. The wider depletion zone in reverse polarization can absorb photons and produce an electron-hole pair driving current in the circuit. For the purpose of this thesis we will consider LEDs only as transmitters.
In order to convey the optical signal to the receiver, the LED must be modulated electrically. One way is to vary the supply voltages in the quasi linear
2 ,6 2 ,7 2 ,8 2 ,9 0 ,0 0 ,2 0 ,4 0 ,6 0 ,8 1 ,0 T 2 i v curve F o r w a r d C u r r e n t ( A ) Forwa rd V olt a ge (V ) T 1 T 2 T 1
(a) i − v LED characteristic at
different temperature showing the current increase with same voltage in hotter conditions. 0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 0 ,0 0 ,5 1 ,0 1 ,5 2 ,0 L-I curve R e l a t i v e l i g h t o u t p u t ( A . U . ) Forwa rd C urrent (m A ) Qu as i -l i n ear regi on
(b) LED forward current as a
function of relative light output power, showing the quasi-linear behaviour.
Figure 1.6: Electrical characteristics taken from LXZ1-6565 LUXEON Z LED
datasheet.
LED’s optical power can be directly modulated by adding a current variation in the quasi linear region to a constant current driver. Both the modulations are affected by temperature degradation [3,4], mainly because junction temperature affects radiative recombination promoting the non-radiative one. Typically, in order to stabilize temperature, high brightness LEDs are often equipped with heat sinks. In lightning and VLC applications current led drivers are preferred, because they are more adaptive, secure and linear. They provide a constant current regardless of the applied load and they prevent LED over driving due to eventual voltage variation. Indeed, in Fig. 1.6a even a small change in voltage can lead to a large increase in current that can exceed the maximum operating values. In addition, the current modulation exploits a more linear behaviour between current and optical power, as shown in Fig. 1.6b .
Another important feature of LEDs that must be taken into account in the design of communication systems is the electrical bandwidth of the device. This is mainly determined by the LED parasitic capacitance and resistance, which introduce a delay in the impulse response of the system and it is expressed by the
according to the following equation:
f−3dB = 1
2πτ ∼
0.35
τ (1.1)
Typically, the electrical bandwidth of a commercial off-the-shelf (COTS) LED is in between 10 and 20 MHz. However, in a OWC system the overall transfer function must take in account the effects caused by electro-optic conversion. From the theory, the electrical bandwidth and the optical bandwidth are connected by the relation:
Hel(f) = √1
2Hopt(f) (1.2)
although the optical band is larger than the electric one, situations can arise where optical phenomena can greatly reduce the overall signal band. A striking example is the phosphorescent materials which do not immediately re-emit the absorbed radiation but they release it slowly. This behaviour leads to a deterioration of the optical frequency response when they are modulated by a faster light source. Typically, in the lighting market, white light can be obtained with a phosphor coated blue LED chip (this topic will be explored in Section 1.2.1), where the phosphor layer causes a strong reduction of the frequency response from 10 to a few MHz. In Fig. 1.7 are reported the electrical bandwidth for the blue LED without phosphor coating and the electrical bandwidth of the same blue LED with the phosphor layer.
B lue R esponse N o r m a l i z e d a m p l i t u d e ( d B ) Freq uency (MHz)
Figure 1.7: Normalized frequency response of white and blue light emitted by
4 0 04 5 0 5 0 05 5 0 6 0 0 6 5 0 7 0 07 5 0 8 0 0 0 ,0 0 ,2 0 ,4 0 ,6 0 ,8 1 ,0 Wa rm whit e C ool whit e N o r m a l i z e d p o w e r ( A . U . ) Wa velengt h (nm )
(a) Normalized optical spectra of
Warm and Cool phosphor-coated white LED 4 0 0 4 5 0 5 0 0 5 5 0 6 0 0 6 5 0 7 0 0 0 ,0 0 ,2 0 ,4 0 ,6 0 ,8 1 ,0 N o r m a l i z e d p o w e r ( A . U . ) Wa velengt h (nm )
(b) Typical normalized optical power vs. wavelength of LUXEON Star RGB LED.
Figure 1.8: Optical power spectral density of the two main configurations to
obtain white light from LEDs.
Optical properties
LEDs have the primary objective of illuminating an environment, not transferring information. This is why the main target of light is the human eye, not a pho-todiode. In fact, it is much more common to find the photometric quantities in LED’s data-sheets rather than the radiometric quantities. Radiometric quantities allow us to characterize light in terms of physical quantities such as: the number of photons, photonic energy, optical power. However, they are irrelevant when it comes to light perceived by the human eye. For example, infrared radiation does not cause any luminous sensation in the eye at all. While photometric quantities are physical quantities that derive from the radiometric quantities weighted on the relative spectral photopic luminous efficiency curve of human eye. For design purposes it is fundamental to understand how these quantities are defined and how they relate to each other.
The optical power spectral density p(λ) is the spectral distribution of the optical power, typically expressed in mW/nm or dBm/nm. In fact, it represents the statistical distribution of the number of photons emitted at a certain wave-length. This optical property allows to discriminate the type of light sources.
3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 1 0 - 4 1 0 - 3 1 0 - 2 1 0 - 1 1 0 0 E y e s e n s i t i v i t y p h o t o p i c f u n c t i o n V ( l ) ( A . U . ) Wa velengt h l (nm )
Figure 1.9: Sharpe, Stockman, Jagla & Jägle (2005) 2-deg V*(l) luminous
efficiency function.
For example, the white LEDs can be made both combining a blue Gallium Nitride (GaN) LED with Yittrium Aluminum Garnet (YAG) phosphor layer, or mixing the light coming from a Red Blue and Green LED triplet. The former are among the most common and economic, but the phosphor layer latency limits the bit rate performances. The latter are more expensive, but offer interesting possibili-ties for high-speed VLC systems, where WDM can be exploited. Fig. 1.8 depicts the spectral differences of those two devices. Knowing the optical spectrum of the light source, usually reported in LED data sheet, is essential to calculate the transmitted optical power by integrating the spectral density over the entire
range of the visible wavelengths (λmax= 780nm and λmin = 380nm).
PT X =
Z λmax
λmin
p(λ)dλ (1.3)
In the same way it is possible to relate a radiometric quantity such as the optical power with a photometric quantity such as the luminous flux using the following relation:
Φ = KmZ λmax
λmin
V(λ)p(λ)dλ (1.4)
where V (λ) is the eye sensitivity photopic function ( shown in Fig. 1.9 [5]), Km
is the maximum visibility (Km = 683 lm/W) at λ = 555 nm. In this way it is
possible to easily convert from the emitted optical power to the one perceived by the retina.
90° 80° 70° 60° 50° 40° 30° 20° 10° 0° -10° -20° -30° -40° -50° -60° -70° -80° -90° 0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 1 , 0
Normalized radiant intensity (A.U.)
Figure 1.10: Normalized radiant intensity pattern represented in in polar
coor-dinates, the dashed lines represent the 120◦ emission angle.
Another important photometric property is the illuminance E. It is defined as the ratio between the incident luminous flux on a surface and its area.
E = ∂Φ ∂S = 1 r2 ∂Φ ∂Ω = I(θ) r2 (1.5)
Assuming the light fixture as a point source that radiates with spherical symme-try, the illuminated surface is linked to the solid angle by the following relation
S = Ωr2, as a matter of fact E results to be dependent by the luminous intensity
I in the direction θ. The radiant intensity is closely related to the concept of
directivity, that is defined as the ratio between the radiant intensity in a specific direction and the total radiated power over all directions. However, it is more common to look for the emission angle, where the directivity is between the max-imum value and its half. Fig. 1.10 shows the typical polar radiation pattern with the relative emission angle.
VCSEL
The VCSEL belong to the category of optical sources which emit light through stimulated emission. Since 1965 these devices have gained enormous interest in the industrial sector. The global VCSEL market size was valued at $ 1,977 million in 2017, and is projected to reach $ 4,749 million by 2023, registering a Compounded Average Growth Rate (CAGR) of 15.7% from 2017 to 2023 [6]. The main applications are optical communications and 3D sensing for face recognition in smartphones.
Figure 1.11: Simple diagram of a VCSEL cross-section (Not to Scale dimension).
VCSEL emits light in vertical direction with respect to the wafer plane. Fig. 1.11 represents a simplified diagram of a VCSEL. Basically, it is composed of several stacked layers. On top and at the bottom, we find the electrodes that injects the current into the device. At the center, we have the laser cavity made up of multiple quantum well layers, which is the active region where lasing actually happens. This cavity is confined by two distributed Bragg reflectors (DBRs) mirrors with 99.0% and %99.9 reflection coefficient respectively. Then the oxide layer, which construct an optimized light emitting window, confines the photon beam in a circular aperture with a small divergence angle. The p- and n- prefix in the Fig. 1.11 stands for the type of doping in a specific layer that builds up the current density in the device.
The electrical characteristics of the VCSELs are optimized to produce coherent light while consuming less electrical power compared to the power spent by the edge-emitting lasers (EELs) and LEDs. At the same time, the VCSEL generates a smaller optical power than its competitors. In Fig. 1.12 is possible to see the small current threshold, but also the potential disadvantage of having lower optical power output in the range of mW. Nevertheless, given the small size, the VCSELs have the great advantage of having a high electrical bandwidth (1-20 GHz) and therefore makes them ideal for all those high speed applications.
Regarding optical properties, the single mode VCSELs have a very narrow
emission spectrum and an emission angle lower than 20◦. This allows to manage
the low optical power more efficiently and to reach longer distances.Typically, they are mainly used as transmitters in high speed D-LOS configurations.
0 1 2 3 4 5 6 7 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 V o l t a g e ( V ) 0 0.5 1 1.5 2 2.5 O p t i c a l P o w e r ( m W ) Curre nt (mA)
Figure 1.12: VCSEL L-I-V curve.
1.2.2
Receivers
In optical systems, the most common photodetectors are semiconductor receivers. These, in a completely specular manner to the optical sources already described, convert the optical signal into a useful electrical signal. As in the case of LEDs, they are p-n junction diodes too. But in this case, the lattice is doped asymmet-rically and the junction is arranged parallel to the receiving surface. This devices are not designed to work in direct bias, but to work photovoltaic or photocon-ductive mode. The most common photoreceiver are PIN PDs and avalanche photodiodes (APDs).
PIN Photodiode
PIN-PD gets its name from its structure where a large intrinsic semiconductor region parts the p-doped region from the n-doped. The incident photons are absorbed generating an electron-hole pairs every time the photon energy hν over-come the bandgap energy of the semiconductor’s depletion region, these carriers are swept across the semiconductor by a built-in electrical field, generating a flow of electric current ip. This process it is also known as photoelectric effect and the bandgap is the amount of energy to promote a valence electron bound to an atom
Figure 1.13: PD equivalent model + i-v curve as a function of incident light.
amount of energy per time unit. If we divide it for the energy of a single photon we obtain the number of incident photons per time unit. These are converted in a flow of electrons with some quantum efficiency η. Therefore, the photocurrent is the product of the electron generation rate with the electric charge q.
ip = qη Pin hν = <Pin (1.6) <(λ) = qη hν ≈ η λ[nm] 1.23985 × 10−3[nm × W/A] (1.7)
Eq. (1.6) shows the linear relation between the received optical power and the generated photocurrent. While Eq. (1.7) represents the photodiode responsiv-ity <, that is a function of the wavelength of the incident radiation and measure the electro-optical gain of the device.
A general photoreceiver can be modeled by the equivalent circuit in Fig. 1.13
inset, where id is the current characteristic of an ideal diode, ip is the generated
photocurrent, idk is the dark current, in is the thermal noise current, Cd is the
photodiode junction capacitance, Rsh is the shunt resistance, Rs is the series
resistance and RL is the load resistance. From this schematic, it is possible to
understand the i-v behaviour shown in Fig. 1.13 and all the noise contributions. The output current falling on the load resistance will be the sum of all the current contributions of the equivalent model:
io= I0 eqVkT −1 + <Pin+ s 4kBTe Rtot Be+ q 2q<PinBe (1.9)
The first term represents the classical exponential characteristic of an ideal diode, then, at fixed temperature and fixed electrical bandwidth, the curve is translated by a constant value proportional to the power of incoming light. The dark current and noise current expressions are derived by the shot noise and
thermal noise equations reported in Appendix A. The Rtot related to the thermal
noise is the equivalent resistance of the circuit Rtot = RLRsh
RL+Rs+Rsh. Photodiodes
are generally used in photoconductive mode, i.e. in reverse polarization. From Fig. 1.13, it can be clearly seen that the photocurrent is almost independent of the applied voltage. When there is no light the photocurrent is close to zero, except for the small dark current. As the optical power increases, the photocurrent increases linearly until it reaches the limit dictated by the maximum power that can be handled by the photodiode. The frequency response of a photodiode is closely related to the type of load. If the current is simply dropped onto a resistance then the bandwidth is
f−3dB = 0.35 τr = 0.35 q τ2 RC+ τdif f2 + τdrif t2 (1.10)
where τRC the RC time constant of the diode-circuit combination, τdrif tthe charge
collection time of the carriers in the depleted region of the photodiode and τdif f the charge collection time of the carriers in the undepleted region of the photo-diode. When reverse bias is close to zero the RC component is dominant and can be only reduced decreasing the active area of the PD. When reverse bias is higher the dominant effect is the drifting delay. Since in the passive current to voltage converter the gain depends on the resistance value, in high gain condition
the τRC can be dominant limiting the overall bandwidth
If the device is connected to a active current to voltage conveter such as TIA the bandwith depends mainly on gain-bandwidth product (GBP)
f−3dB =
s
GBP
2πRF(Cd+ CF) (1.11)
Avalance Photodiode
The APDs are quite similar to PIN-PD. They are usually composed of 4 asym-metrically doped semiconductor layers. A p- region, an intrinsic area, a less doped
pregion and a highly doped n section. The third zone represents the
fundamen-tal part of the avalanche photodiode, because it allows the multiplicative or gain effect of the charges. Basically, the generated carriers, accelerated by a strong bias voltage field, collide with the lattice producing an avalanche of carriers. This brings to a photocurrent gain M greater than unity.
Unfortunately, these devices are affected by strong shot noise represented by the excess noise factor F . Since the multiplication process of the carriers that con-tribute to the output current is intrinsically “noisy”. In the absence of other noise
sources, an APD provides a SNR which is√F worse than a PIN detector with the
same quantum efficiency. An APD, however, can produce a better overall system signal-to-noise ratio than a PIN detector in cases where the APD internal gain boosts the signal level without dramatically affecting the overall system noise.
1.3
Optical channel modeling
The OWC applications, which are considered in this manuscript, exploit the In-tensity Modulation with Direct Detection (IM/DD). Fig. 1.14 depicts the equiva-lent linear baseband transmission model where the impulse response h(t) reflects the channel characteristics, x(t) is the incoming optical signal modulated by the
Figure 1.14: Linear baseband transmission model of an OWC link.
intensity, < is the receiver responsivity, n(t) represents both the electrical (see Eq. (1.9)) and the optical noise and y(t) is the optical received current and is expressed as:
y(t) = <x(t) ⊗ h(t) + n(t) (1.12)
i.e. the system output is equal to the convolution of the input with the impulse response of the channel with the contribution of the noise. Since the channel input represents instantaneous optical power, x(t) must be non-negative:
and the average transmitted optical power Pt is given by Pt = lim T →∞ 1 2T Z T −T x(t)dt (1.14)
instead of the |x(t)|2 expression that is used when it represents amplitude. While
the average received optical power is given by
Pr = H(0)Pt (1.15)
The frequency response H(f) of a OWC link is a linea combination of two con-tributions:
H(f) = HLOS(f) + Hdif f(f) (1.16)
where HLOS is the LOS contribution and it is independent of the modulation
frequency, while Hdif f comes from the diffuse reflections in the environment [7].
The directivity of an OWC geometrical configuration can be also quantified with power ratio between the LOS and the diffuse signals:
k = HLSO Hdif f
!2
(1.17) when k is larger than about 13 dB then the optical channel is transparent (i.e., the bandwidth is beyond 300 MHz) [8]. Therefore, in many operational cases, we can assume the frequency response flat with respect to the DC value. So for most purposes, the single most important quantity characterizing a channel is the DC gain H(0). The DC gain is the concatenation of all the transfer functions of the individual components that describe the optical channel:
H(0) = HT X(0)T (λ)HRX(0) (1.18)
where we have the transmitter and receiver transfer functions respectively and the optical medium contribution T (λ). For the following demonstrations we assume a geometric configuration of the link as shown in Fig. 1.15.
1.3.1
Transmitter model
OWC sources, such as LEDs or VCSELs, produce an wide beam with an angular
distribution of the optical power PT X that follows the Lambertian law [9]
I(φ) = I0cosmφ=
m+ 1
Figure 1.15: Scheme of free-space propagation.
considering the geometry in the Fig. 1.15, we choose to normalize the radiation
pattern to I0 = I(0) such that:
Z π 2 0 Z 2π 0 I0cosm(φ)sin(φ)dφdθ = 1 (1.20)
where m is is the Lambertian radiation index of the source
m= − ln(2) ln(cos(φ1
2))
(1.21) Taking into account also the effect of the propagation distance, we obtain:
HT X(0) = 1 + m
2π cosm(φ)
1
d2 (1.22)
1.3.2
Receiver model
The receivers are represented as active areas ARX capable of collecting photons
within a certain acceptance angle. Furthermore in order to improve reception performance filters and optical concentrators are exploited, which must be con-sidered in the overall calculation of the transfer function.
HRX(0) = ARXf(ψ)g(ψ)cos(ψ) (1.23)
where f(ψ) is the optical filter transfer function and g(ψ) the optical gain function of the concentrator.
400 600 800 1000 1200 1400 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Sun Incandescent lamps Fluorescent lamps N o r m a l i z e d A m p l i t u d e ( a . u . ) Wavelength (nm)
Figure 1.16: Optical power spectra of the three main noise sources.
1.3.3
Trasmission medium model
In free space applications the light can travel the atmospheric medium or propaga-tion can be challenged by harsh environment such as underwater scenarios. Like any other energy propagation it is subject to absorption, scattering, and refrac-tion phenomena causing an exponential decrease in luminous intensity. Typically, light propagation can be described well by the Beer-Lambert law [10]:
T(λ) = e−k(λ)d (1.24)
k(λ) is the attenuation coefficient and takes into account all the attenuation
mechanisms involved. It is common sense to neglect this coefficient in the case of short distances link in air, while is important to study its contribution when the medium become more complex.
1.3.4
Optical noise
In order to complete the dissertation on the optical channel modelling, it is nec-essary to describe the part relating to the optical interferences acting on the channel. There are three main sources of optical noise that come to play in OWC systems as shown in Fig. 1.16.
The sun light is the most predominant source of optical noise especially in outdoor environment. It is characterized by an optical broadband emission spectrum with maximum values in the visible region. We can consider the mod-ulation bandwidth of the sun fairly constant. The incoming optical power is
transduced in electrical shot noise, that can be easily rejected by the receiver electrical bandwidth itself of by a specific low pass filter.
Incandescent lamps are still widespread in homes and indoor environments. These are artificial light sources that emit photons thanks to the overheating of a metallic element due to the Joule effect. For this reason they have a large part of the spectrum emitted in the infrared region. Usually, they are supplied by the home power grid (50÷60 Hz) [11], producing a narrow modulation band less than 2 kHz. Those frequencies are neglected in VLC links where the signal spectrum is at higher frequencies, but in few cases, they can be a potential problem in slow speed applications such as indoor positioning and camera communication. Even in this case the unmodulated optical noise is converted in shot noise and is treated as for the sunlight.
Finally, the fluorescent lights are the most invasive for the optical channel. Their electrical power spectrum is mainly constrict to a band between 100 kHz and 300 kHz. Therefore all these spectral components fall into the baseband modulation spectrum of the communication system. This means that filtering them would affect the signal response. Only Single-Carrier (SC) modulation systems can overcome the issue by utilizing transmission frequencies that are located outside the fluorescent noise bandwidth. Usually,this type of optical interference can be rejected shadowing the source of the noise. OWC receivers with narrow acceptance angle will only see the well directed signals, excluding all the unwanted external contributions coming from different directions.
1.4
Modulation techniques
In VLC, the LED bandwidth is quite limited, which is why selecting the appropri-ate modulation format has become crucial to use efficiently the given bandwidth. The spatially incoherent reception in OWC makes it difficult to realize an ef-ficient coherent detection, so the most viable solution is the IM/DD. Among all the several IM/DD modulation schemes, the OOK is the simplest and it can be implemented with different encoding techniques such as Non-Return-to-Zero (NRZ), Return-to-Zero (RZ) and DC-balanced (Manchester). Then we have more power efficient modulation format which however require a higher complexity of realization such as PM and subcarrier modulation such as OFDM and WDM.
1.4.1
On-Off Keying (OOK)
OOK represents digital data through the presence or absence of the analogical signal. NRZ-OOK, as a matter of fact, is not a real coding, since the data is di-rectly transmitted as such in the output. The digital symbol “1” is represented in the bit interval as a high voltage signal, while “0” is zero voltage level. Although NRZ is bandwidth efficient (roughly equivalent to bit rate), its spectral compo-nents go all the way down to 0 Hz. This makes NRZ more susceptible to noise at low frequencies and it prevents its use in AC coupled applications. On the other hand, in RZ the digit is coded with a signal that “return” to zero voltage level at each pulse with a duty cycle D. This guarantees an improvement in terms of power efficiency, but the needed bandwidth increments of a factor 1/D compared to NRZ. Both the encoding suffer of loss in synchronization.
1.4.2
DC-balanced
In order to reject the DC component and improve synchronization, DC-balanced code associates high state a low state to a positive and negative voltage with respect to a reference voltage level. In this way one of the digits is no longer associated with the DC component, shifting the power spectrum at higher fre-quencies.
Manchester encoding enhances the clock recovery implementation. In this type of encoding the digital state “1” is represented with a transition from the high
signal and to the low signal at Tbit/2, while the digital state “0” is represented
with the opposite transition. The analogical signal is generated combining an NRZ bit stream with the Manchester clock signal through a XOR port. Forcing one transition per bit removes all the long sequences of “ones” and “zeros” making clock recovery easier.
1.4.3
Pulse Modulation (PM)
PMs also known as impulsive modulations guarantee high-speed data transmis-sion, improving power efficiency with a relatively limited cost of bandwidth. Among its modulation formats, the most popular in OWC is Pulse Position Mod-ulation (PPM). It is considered by many to be the best modMod-ulation technique for communication systems that use IM/DD and limited power. It consists of
period, there are L slots, one of which is occupied by the impulse and the other
L −1 have zero value. The information is coded by the position occupied by
the impulse within the symbol. Although the PPM offers a good efficiency in terms of power, increasing the number of bits per symbol causes a considerable reduction in spectral efficiency.
1.4.4
Multicarrier modulation
Multicarrier modulation is one of the encoding techniques where information is parallelized on a number of sinusoidal secondary carriers. Each subcarrier in-volves a narrow band signal spaced apart by a ∆f interval. This is equivalent to having a slower and therefore less distorted, signal on each channel, requiring easier receiver equalization. In OFDM, the subcarriers are chosen orthogonal to each other with an adequate guard interval between successive OFDM symbols in order to reduce the effects of intersymbol interference (ISI).
In OWC a further way of multiplexing the signal is to exploit the different wave-lengths over the wide optical spectrum as subcarriers. This multiplexing tech-nique is also known as WDM. Typically, decoding is performed by optical filters at the receiver separating the different “color” components. In the field of visible communication this type of modulation, associated with RGB LED sources, has been very successful, allowing to break several time the transmission speed record in VLC.
1.5
OWC applications overview
The OWCs are a clear example of communication systems complementary to the RF technologies. As presented in this chapter, they have many important characteristics such as wide spectrum, high-data-rate, low latency, high security, low cost, and low energy consumption, addressing the highly demanding require-ments of the next generation communications. OWC systems are very agile and they can fit several application scenarios especially where RF communications are disadvantaged.
State of the art OWC applications, which are closely related to the work presented in this manuscript, are summarized in the following.
Most of the OWC technologies are in the short-range distance (meters), in WPAN indoor networks. They are designed as alternative solutions to the
RF communication. Li-Fi is an emblematic example of how this technology can provide services like those offered by wireless fidelity (wireless fidelity (WiFi)) [12]. The typical OWC system is a bidirectional link which exploits white phos-phor LED as down-links and IR LED as up-links [13]. These types of systems operate in Nd-LOS configuration because they must guarantee certain coverage and mobility to the user [14]. The bit rates obtained with these OWC systems mainly depend on the band of the optoelectronic devices [15], on the modulation formats [16], on the multiplexing [17] and equalization techniques [18]. Our re-search team broke the world record of bit rates realizing 5.6 Gbit/s with a WDM system using commercial RGB LEDs [19].
Currently, higher speeds are achieved with LASER diode (LD) or µLED, which have an intrinsic electrical band higher than 1 GHz [20,21]. Unfortunately, these devices are economically more disadvantageous as they do not exploit the existing lighting infrastructure and, in the case of µLED, the production process is not yet industrialized. Furthermore, eye safety regulations can be critical for LD, while µLEDs have low output optical power.
So far, WDM in combination with OFDM and analog equalization seems the most promising leverage to the transmission speed. To the best of our knowledge, the highest achieved data rate with an RGBY LED in eye-safe condition is 15.73 Gbit/s [22]. However, the system architecture complexity is still an open issue. For each transmitted wavelength, lenses, optical filter, photodiode are needed.
Our contribution to this extend, is to reduce the number of components on the receiver side. Recently we demonstrated for the first time the feasibility of using a single receiver in a WDM RGB transmission. This allowed to eliminate the optical filters and to reduce the number of photodiodes to one [23].
All the abovementioned VLC links are D-LOS configuration systems, there-fore are less useful in WPAN desktop application. In this scenario, the preferred configuration is Nd-LOS, where the RX can operate in a certain coverage area. Few publications are reported in this field and most of the contributions come from our group [24].
D-LOS multi-gigabit links fit better the DCN applications, which mainly exploit coherent IR sources modulated both directly and indirectly to replace the large number of cable connections that limit the expansion of new generation infrastructures. Some solutions offer complex re-configurable transceivers [25] or steerable microelectromechanical systems (MEMS) [26], while others offer
point-to-point configuration [27], which are tolerant to misalignment [28].
High throughput OWC point-to-point link can be also used as optical wire-less interconnect in B2B communication, especially in those cases where the harsh environment and the harness are crucial for the design.
HEP experiments have massive harness due to the large number of optical fiber connected the particle detectors. Here OWC can provide wide bandwidth with significantly reduced mass. For this application is essential to estimate the radiation hardness of the OWC optoelectronic components. Most of the recent studies refer to the radiation level in the space environment [29], which are less stringent than the radiation level of HEP requirements. Therefore, we recently reported the unique evaluation of VCSEL and PDs under strong proton irradiation [30].
Another emblematic harsh environment is the underwater. Here, the wire-less communications are strongly attenuated by the medium. RF communications suffer power losses of a 60 dB/m at 10 MHz [31]. While blue light sources are less attenuated in water with equal frequencies. underwater optical wireless com-munication (UOWC) can enable wireless comcom-munication in a range of distances between 1 and 100 m. The UOWC is a promising technology because offer higher bit rate (e.g.1 Mbit/s to 30 Gbit/s) compared to the conventional acoustic com-munication systems. In the last decade, the experiments conducted in this niche research field have multiplied [32]. Most publications are laboratory demonstra-tions with tanks of water [33,34], while only 3 research groups, among which our own, have tested UOWC devices which actually work in real environment [35–37]. For the sake of completeness, other examples of OWC applications, which are not related to this thesis, are mentioned below.
OCC is a peculiar subcategory of the VLC, the main difference is the receiver type. OCC exploit the rolling shutter effect of the CMOS image sensor of the smartphone’s camera [38]. Typical bit rates are in the range of hundreds of kbit/s, the small amount of information (e.g. lamp ID number) gathered by the receiver can provide a useful framework for indoor localization applications [39].
VLC also finds applications as outdoor terrestrial wireless technology. VLC can provide an integrated system for the management of urban traffic and safety in autonomous driving [40, 41]. Traffic light and car headlight can be used to communicate essential information between vehicles V2V and between the in-frastructure V2I connected to the cloud. In smart cities, fully connected by the
IoT, this application can be widely implemented.
In the niche field of ultra-short-range communication, the OWC can be used as interconnection inside the chip. In kilo-cores processor the conventional pla-nar metal interconnections, which constitute the network-on-chip (NoC) between cores, can arise impairments in terms of latency, power consumption, routing management [42]. optical wireless network-on-chip (OWiNoC) is a promising solution to the scalability of the many-cores architecture exploiting plasmonic antennas embedded with the photonic integrated circuit (PIC) network [43]. op-tical phased array (OPA) of plasmonic antennas are used to enable new routing feature [44]. The OPAs can also provide chip-to-chip communication, in [45] the authors demonstrated data rates of 100 Mbit/s at distances up to 0.5 m.
Finally, the most extreme applications in OWC are the ultra-high through-put Free Space Optics (FSO) systems for terrestrial backhauling and space com-munications. They are generally coherent systems that can reach very long dis-tances by means of precision optics and high power laser sources [46,47].
2
Analog equalization for high
speed OWC
As discussed in the Chapter 1, the bandwidth of the LEDs can limit the transmis-sion capacity of the OWC system. In the case of phosphor-coated (PC)-LEDs, the electrical bandwidth is few MHz, while LED without phosphor coating (e.g. IR LED) reaches 10 - 15 MHz bandwidth. A common way to compensate the losses after the cut-off frequency is to use an equalization circuit, which flatten the modulation bandwidth improving the overall communication performances.
These circuits are also very effective when used with multicarrier modula-tions. In fact, OWC systems that exploit multicarrier modulation together with analog equaliztation, show lower BER and higher SNR compared to the system without it.
In this chapter we deal with a practical implementation of an indoor real-time OWC transmission in Nd-LOS configuration using a low-cost IR LED and Field Programmable Gate Array (FPGA) based DMT processing with an analog
equalization circuit. Achieving a bit rate from 230 to 100 Mbit/s over a 3 m2
area.
In the present work the author contributed with the design of the analog equalization circuit and participated in the realization of the experiment.