Demonstration of a free-space optical communication link in the 10-meter range using electro-absorption modulator arrays

K UNGLIGA T EKNISKA H ÖGSKOLAN

M ASTER T HESIS

Demonstration of free space optical communication link in 10-meters

range using electro-absorption modulator arrays.

Author:

Adrien C ^HOPARD

Supervisor:

Dr. Qin W ^ANG Examiner:

Dr. Martin V ^IKLUND Delegated to:

Dr. Katia G ALLO

March 1, 2017

iii

KUNGLIGA TEKNISKA HÖGSKOLAN

Abstract

RISE ACREO Sensorsystem

Demonstration of free space optical communication link in 10-meters range using electro-absorption modulator arrays.

by Adrien C HOPARD

This work has been focused on the implementation of an indoor asymmet-

ric Free Space Optical (FSO) link using a Quantum Well (QW) based Electro

Absorption Modulator (EAM) as core component for communication. The

use of such a technology is allowing the user to convey information with a

high data rate and a low power consumption via a compact and lightweight

Modulated Retro Reflector (MMR) perfectly suitable for satellite commu-

nication and space applications. Optical and electrical characterizations of

such modulators have been carried out in order to reinforce existing knowl-

edge about those components and set a suitable communication link. Prac-

tical challenges and solutions are discussed here and implemented on the

functional test bench link. At last, space focused investigations have been

led to estimate the performances and robustness of components when con-

fronted to low temperature environments and radiations.

v

Acknowledgements

I would first like to thank my thesis supervisor Dr. Qin W ANG for the op-

portunity she gave me and her enthusiasm toward this project and Dr. Ka-

tia G ALLO , my examiner at KTH who supported me. I am also particularly

grateful to Mikael Karlsson and Ingemmar Petermann for sharing their ex-

pertise and helpful advices all along my stay. It has been a pleasure to work

in collaboration with Darius Jakonis and Michael Salter among all the staff

from ACREO. Thank you to Pr. Jan Pallon from Lund University, for his

time and involvement that enabled us to perform radiation exposures.

vii

Contents

Abstract iii

Acknowledgements v

1 Introduction 1

1.1 Background . . . . 1

1.1.1 Free space optical communication . . . . 1

1.1.2 Asymmetrical optical link . . . . 1

1.1.3 FSO operational windows . . . . 3

1.1.4 State of the art . . . . 3

1.2 Aims . . . . 4

2 EAMs and characterisation 5 2.1 Modulator structure . . . . 5

2.2 Quantum wells . . . . 6

2.2.1 Optical characterisation of MQW . . . . 7

2.2.2 Theoretical principle . . . . 8

2.2.3 Operational wavelength tuning . . . . 9

2.3 Device features . . . . 10

2.3.1 P-N junction and contacts . . . . 10

2.3.2 Geometries and bandwidth limitation . . . . 11

2.3.3 Anti-reflective coating . . . . 13

2.4 Overall DC characterization . . . . 15

2.4.1 Optical measurement on modulators by FTIR . . . . 15

2.4.2 Dark current . . . . 18

2.4.3 IV measurement and optical responsivity . . . . 21

3 FSO asymmetrical communication setup 23 3.1 Asymmetrical FSO communication setup . . . . 23

3.2 Electrical driving . . . . 23

3.2.1 Electrical driving setup . . . . 24

3.2.2 Impedance matching . . . . 27

3.3 Short distance FSO link . . . . 32

3.3.1 Optical setup specifications . . . . 32

3.3.2 General performances and on bench characterisation 33 3.4 Long distance FSO link . . . . 35

3.4.1 Short distance optical setup limitations and implemented solutions . . . . 35

3.4.2 Optical setup specifications . . . . 37

3.4.3 High speed performances and overall results . . . . . 38

4 Verification of EAMs functionalities for space applications 43

4.1 Low temperature behaviour . . . . 43

viii

4.1.1 Dark current . . . . 44 4.1.2 Wavelength shift . . . . 46 4.2 Radiation tests . . . . 47

5 Discussion 49

5.1 Summary . . . . 49 5.2 Future work . . . . 50

Bibliography 51

ix

List of Figures

1.1 C3PO asymetric link . . . . 2

1.2 Athmosheric transmittance . . . . 3

2.1 EAM structural diagram . . . . 5

2.2 EAM top views . . . . 6

2.3 Multiple quantum wells reflectance . . . . 7

2.4 FTIR reflectance module . . . . 7

2.5 QW wavefunction profiles . . . . 8

2.6 Coupled QW wavefunction profiles . . . . 9

2.7 Wire bonded modulator . . . . 10

2.8 Probestation and wirebonding station . . . . 11

2.9 Top wiews of EAMs . . . . 11

2.10 Theoretical bandwidth limitation and pixel capacitance . . . 12

2.11 Multiplexing and relaying . . . . 13

2.12 ARC principle . . . . 14

2.13 ARC reflectance . . . . 15

2.14 EAM reflectance . . . . 15

2.15 FTIR reflectance module . . . . 16

2.16 EAM reflectance and contrast ratios . . . . 17

2.17 EAM transmittance and contrast ratios . . . . 18

2.18 Dark current . . . . 19

2.19 Dark current histogram . . . . 20

2.20 Photocurrent and responsivity . . . . 22

3.1 Assymetric FSO link principle . . . . 23

3.2 Typical driving board . . . . 24

3.3 Electrical driving setup diagram . . . . 25

3.4 Typical driving signals . . . . 26

3.5 Experimental PCB . . . . 27

3.6 Electrical matching configuration . . . . 28

3.7 Matching results . . . . 28

3.8 Frequency response for a few connected pixels . . . . 30

3.9 Frequency response with different optical powers . . . . 31

3.10 Short distance link diagram . . . . 32

3.11 Short distance FSO link optical sinusoidal signals . . . . 34

3.12 Short distance FSO link optical PRBS signals . . . . 35

3.13 Laser beam spot size . . . . 37

3.14 Long distance link diagram . . . . 37

3.15 Long distance FSO link optical sinusoidal and LVDS signals 39 3.16 Long distance FSO link optical LVCMOS signals . . . . 40

3.17 Long distance FSO link optical PRBS signals . . . . 41

3.18 Eye diagrams . . . . 42

x

4.1 Low temperature setup . . . . 44

4.2 Dark current evolution with temperature . . . . 45

4.3 Photo current and responsivity evolution with temperature . 46

4.4 Linear particle accelerator . . . . 48

xi

List of Tables

3.1 Q factor and BER estimations on optical PRBS signals. . . . . 40

xiii

List of Abbreviations

FSO Free Space Optical

QW Quantum Wells

EAM Electro-Absorption Modulator RF Radio Frequency

MRR Modulated Retro Reflector

IR InfraRed

UAV Unmanned Aerial Vehicle MQW Multiple Quantum Wells ARC Anti-Reflective Coating

FTIR Fourier Transform InfraRed spectroscopy NIR Near InfraRed

QCSE Quantum Confined Stark Effect CQW Coupled Quantum Wells

Eg Energy Gag

CR Contrast Ratio PCB Printed Circuit Board FOV Field Of View

SMA Sub Miniature version A

LVDS Low Voltage Differential Signaling

LVDCMOS Low Voltage Complementary Metal Oxide Semiconductor PRBS PseudoRandom Binary Sequence

BER Bit Error Rate

LEO Low Earth Orbit

1

Chapter 1

Introduction

1.1 Background

1.1.1 Free space optical communication

Free Space Optical (FSO) links have proven their usefulness through a wide range of communication applications [1] where the propagation media re- mains favourable and the use of waveguides, fibres or cables is either an issue, unwanted, or useless. Optical data transmission has been revealed to be a good and low-cost way to send high rate information data streams without having to deal with the over-allocated and crowded Radio Fre- quency (RF) spectrum via radio communication or to avoid high mainte- nance systems.

Unlike the case where an isotrope or Lambertian source is used to commu- nicate with one or few punctual receivers, using the high directionality of a laser source is commonly recommended and will ensure that most of the emitted power will be useful for communication. It enables then a quite reliable and secure long distance link.

A few drawbacks are nevertheless unavoidable. Indeed, a proper light propagation requires to have a relatively perturbation-free environment.

For ground-to-ground or ground-to-space communication through atmo- spheric environment, a lot of parameters have to be taken into account such as scattering induced by particles or bad meteorological conditions (fog, rain...) that can induce losses up to 100dB/km, atmospheric absorp- tion or turbulences. Beside those aspects, the beam dispersion is the intrin- sic distance-limiting factor even for space applications and a good pointing accuracy, tracking ability and stability must be achieved to have an opera- tional FSO link.

1.1.2 Asymmetrical optical link

A conventional optical link typically connects the emitter (signal source),

where the light source is integrated, to the receiver side where data streams

are collected. The main particularity of the asymmetrical link is that the

Infra Red (IR) light source and receiver (optics + photodetector) are on the

same side of the link to provide an unmodulated laser beam and interro-

gate the other side of it. Only the input data encrypting is performed at the

2 Chapter 1. Introduction

other end of the link and the beam with encoded information is then sent back to the interrogator. Hence, the Modulated Retro-Reflector (MRR) is the key component with the two roles of modulating the signal according to the input data sequence, while ensuring that the beam will be reflected back in exactly the same direction. In order to achieve a good link bud- get, efficient optical solutions are needed. Suitable designs for this kind of retro reflection have already been studied and successfully used (such as the corner cube reflector and the Cat Eye-Optic [2] [3]).

This asymmetric FSO geometry ensures the reduced size and weight on the MRR end of the link which makes it perfectly suitable and advantageous for on-board systems such as nano-satellites uplink/downlink real-time com- munication, planes, drones or spacecrafts intercommunication where space and weight-saving will enable other tools to be implemented. An example of low orbit Satellite to ground communication is shown in Fig. 1.1 as the main goal of the H2020-C3PO project

.

F

1.1: C3PO asymmetric FSO link diagram using an MRR for low earth orbit satellite uplink/downlink commu-

nication system 1

1

Project coordinated by Airbus Defence and Space in collaboration with RISE ACREO AB, University of Oxford (UK), Deutches Zentrum Fuer Luft–und Raumfahrt EV (DE),Astri Polska (PL) and Tematys (FR) and funded by the European Commission H2020 program. http://c3po-h2020.eu/ and https://www.acreo.se/projects/

advanced-concept-for-laser-uplink-downlink-communication-with-space-objects-c3po

1.1. Background 3

1.1.3 FSO operational windows

Among the IR Spectrum, a few wavelength bands have been chosen for be- ing convenient for laser communication in atmospheric environment. The main selection criteria was the low attenuation induced during atmospheric propagation. For low enough attenuation windows (see Fig. 1.2) [4], ef- fective and high power light sources, detectors, and various other optical components such as fibres have been reached a reliable degree of matu- rity in the main communication intervals, centred around 850nm, 1060nm, 1250nm and 1550nm.

850nm 1060nm

1250nm 1550nm

F

1.2: Atmospheric transmittance spectrum

Multiple Quantum Well (MQW) based modulators have already been de- veloped at ACREO in the most commonly used communication bands at operational wavelength around 850nm and 1550nm using GaAs and InP based semiconductor technologies. GaAs materials were use to develop surface-normal EAMs[5] working at 850nm while here, InP-based modula- tors with operating wavelength around 1550 nm are used and will be char- acterised in chapter 2 and 4, while optical links using those EAMs will be detailed in chapter 3.

1.1.4 State of the art

ACREO has been developing and processing those EAMs within its clean

room facilities through a few projects. Previous versions of such asymmet-

rical links have been demonstrated [6] with UAV flight tests allowed by the

lightweight and low power-consumption of EAMs driving circuits. A data

rate of 20Mbit/s has been achieved at the time and the project covered a

laser tracking system development, long distance tests (up to 1.2km) and

data treatment. A few problems such as the atmospheric limitation and

data rate limitation have already been encountered.

4 Chapter 1. Introduction

The C3PO project goal is to increase the communication speed up to 1Gbit/s while decreasing the consumed power and reducing the mass and volume taken by the modulator and its driving electronics. ACREO plays then a central role within this project since the design of the modulator and the driving electronic boards are key challenges and require expertise in those domains.

1.2 Aims

In order to facilitate analyses and characterisation on MRRs and surround- ing electronic driving boards, setting up an optical test-bench is one of the focused task of this project. It has been established at first for short dis- tance communication, allowing performances characterisations and con- crete physical studies of EAMs without fulfilling the true FSO communi- cation criteria.

Starting with a phase of simple characterisations on modulators in order to

get to know better the physical principles behind those and the working en-

vironment, my main role has then been to set up a longer range optical test

bench for FSO communication. In order to reach higher driving frequencies

and so higher transmission data rates, efforts have been made to overcome

the encountered important electrical problems. Beside those practical im-

plementations of EAMs in an operational link, a few extra physical charac-

terisations have been led such as a succinct study of the low temperature

behaviour and radiation exposures.

5

Chapter 2

EAMs and characterisation

2.1 Modulator structure

ACREO has been developing QW based EAM technologies since 2000 via clean-room processing in their facilities operated jointly with KTH (Kung- liga Tekniska Högskolan) in Kista. Surface normal EAMs, used for FSO communication, have been one of the many applications to be matured in this area. their high performance improvement potential and the large de- mand in the communication domain were the main drivers for this evolu- tion.

Transmissive and reflective modulators have been developed. The struc- tural diagram of a typical surface normal reflective EAM is shown in Fig.

2.1.

InP Substrate

Golden back mirror N-doped Layer (n+ InAlAs) N-contact

Intrinsic layer MWQ InAlAs/InGaAs stacking structure P-doped Layer ( p+ InAlAs)

P-contact ARC

InGaAs Layer

F

2.1: Structural diagram of a reflective EAM

On top of the InP substrate wafer, a stacking of multiple InAlAs and In- GaAs layers is sandwiched in-between a P-doped InAlAs and a N-doped InAlAs layer, and plays the role of the MQW structure. This stacking is then considered as the intrinsic slab of the p-i-n junction which is the base structure of those the modulators. Electrodes are then processed to allow access for electrical connections.

The MQW section is then the key part which, similarly to a photodiode intrinsic layer, will interact with the incoming light beam and determines most of the optical proprieties of the EAM.

A highly doped InGaAs layer is then deposited on op to ensure a good

ohmic contact with the top electrodes.

6 Chapter 2. EAMs and characterisation

An anti-reflective coating (ARC) is placed on top of the active area in order to maximize the amount of incident light propagating through the modu- lator and avoid the typical reflection at the first interface (see subsection 2.3.3).

For a reflective modulator, a golden mirror is deposited on the back. Nev- ertheless, MRR can be generated using a transmissive modulator, where the backside mirror is replaced by an ARC, in association with an optical system and a reflective element. For example, the use of a cat-eye optic geometry is optimised with the use of a transmissive EAM and a spherical mirror on its focal plan.

By using Bragg mirrors (stacking of a few more layers on top and back side of the component), it is also possible to insert the structure in a cavity in order to maximise the optical modulation efficiency, resulting in higher contrast ratio. However, this reduces and limits the wavelength window.

As a consequence, the operational wavelength of EAMs being highly tem- perature dependent, a slight operational condition change could drastically annihilate any modulation.

Typical geometries of fabricated pixelated reflective EAMs are shown in Fig. 2.2. They will be discussed more precisely in subsection 2.3.2.

(

) (

)

F

2.2: 2.2a Top view of an EAM : 6*6 pixelated (250µm*250µm) square array with sided pads on pixels,2.2b

Photo of a wire-Bonded EAM

2.2 Quantum wells

In this structure, the main feature is of course the QW stacking which is the

optically active area. InGaAs/InAlAs structures have been demonstrated

to be suitable for optoelectronic components operating from 1.3µm to 1.5µm

[7]. MQWs structures are manufactured via epitaxy on top of an InP sub-

strate.

2.2. Quantum wells 7

2.2.1 Optical characterisation of MQW

Figure 2.3 shows the optical reflectance spectra of the MQW stacking with- out any applied electric field.

F

2.3: Reflectance spectrum of the MQW structure measured by FTIR

This measurement has been conducted using a Bruker Vertex70v FTIR ma- chine. The NIR source was set and a interchangeable Bruker I26312 reflec- tive accessory has been installed for this measurement (see picture 2.4).

F

2.4: Bruker I26312 interchangeable FTIR reflectance module

Applying then an electric field will change the optical response of the MQW

structure and so should induce intensity modulation onto a monochromatic

light beam. A shift of the observable excitonic peaks toward longer wave-

length should be observed as well as a broadening and decrease of their

amplitudes.

8 Chapter 2. EAMs and characterisation

2.2.2 Theoretical principle

The previous effects can be described by the Quantum Confined Stark Effect (QCSE) and is induced in the Coupled Quantum Wells (CQW) structure of the modulator. In the case of an unbiased single Quantum Well (QW) structure, we would have the energy profile illustrated in Fig. 2.5a. For photon energies lower than the energy gap (E

), the absorption is of course limited and so the reflectance is higher for longer wavelengths (see Fig.2.3).

E

E

E

E

z E

E=0

(

)

E

E

E

E

E≠0

(

)

F

2.5: 2.5a Non perturbed wave function in single QW, 2.5b Wave function in biased single QW

Stark effect

Applying an electric field (E) will first bring the 2 energy levels closer to- gether (see analogy with single QW on Fig. 2.5b) which is called the Stark effect (analogue to the Zeeman effect through magnetic field) and will im- mediately induce a shift of the absorption exciton spectra toward longer wavelengths. In combination with the steepest part of the reflectance curve, it will then generate a large contrast ratio around the working wavelength.

Those results can be deduced from a second order perturbation approach.

Quantum confinement

The second effect is induced by the quantum confinement among the QW structure. Applying the electric field will tend to pull the electrons and hole in opposite directions which will result in a significant change of their respective wave function shapes (see analogy with single QW on Fig. 2.5b) affecting their overlapping. Since the absorption probability of a photon is related to the overlapping of those wave functions, the amplitudes of the excitonic peaks will decrease and they should also experience a certain broadening.

In the case of the CQW structure [8], the mechanism remains similar but the

contribution of heavy holes in both wells (hh1 and hh’1) have to be taken

2.2. Quantum wells 9

into account as shown in Fig. 2.6. It will then gives two Contrast Ratio (CR) peaks under bias that might be useful for modulation.

F

2.6: Wave function profile in CQW structure

2.2.3 Operational wavelength tuning

In order to centre the optical response (see Fig. 2.3) around the desired working wavelength, it is necessary and useful to fine tune a few parame- ters during the fabrication.

The energy gaps in the MQW stackings will of course be the decisive fac- tors. Fortunately, by adjusting the III-V compound proportion in each alloy of the stacking, we are allowed to select the energy gaps in a given layer by applying the following formula 2.1[10] :

Eg(A

B

) = (1 − x)Eg(A) + xEg(B) − x(1 − x)C (2.1) Where x is the proportion of B compound, Eg(A) and Eg(B) are the bulk energy gaps of A and B compound respectively and C is the bowing pa- rameter which accounts for the depart from the linear approximation. Tab- ulated values of those parameters can easily be found [9] and so the energy gap of In

Ga

As and In

Al

As can easily be computed and tuned al- most freely. The main limitation in this selection process is that the lattice mismatching has to be taken into account in order to avoid too strong stress or strain that might weaken the structure.

The Layer width will also require investigation since they determine the position of the energy levels in the wells (E

and E

in Fig. 2.5). Nev- ertheless, the quantum confinement needs to be preserved and so the wells can not be too wide nor the barriers too thin.

Tuning the energetic profile of each CQW by fixing the appropriate material

proportions and geometrical dimensions will then give a lot of freedom in

order to obtain the desired operational wavelength.[10]

10 Chapter 2. EAMs and characterisation

The final modulator stacking consists of 80 periods set to the following arrangement in order to get a good CR around 1550nm: In

Ga

As /In

Al

As /In

Ga

As /In

Al

As (64Å/15Å/64Å/48Å). The number of periods results from a trade off between optimising the interac- tion length and increasing the distance between the p- and n-doped layers.

Indeed, stacking more CQW structures would of course improve the optical efficiency, and therefore the CR, but the modulator would require a higher driving potential due to the thicker intrinsic layer.

2.3 Device features

2.3.1 P-N junction and contacts

On each side of the MQW stacking, a doped layer is used in order to apply the driving electric field to the intrinsic region and so changes its optical response. The top InAlAs layer is P-doped using Be as the dopant while the bottom layer is N-doped using Si as a dopant.

The modulator will hence adopt an electrical behaviour similar to a diode.

The structure should then always be operated under reverse bias in order to avoid large current density generation that would destroy the component.

On Fig. 2.2a, we can see the golden ring and the large electrical pads on Fig. 2.2b that surround the modulator. They will be used as the ground electrodes since they are connected to the anode of the modulator (P-doped layer). The negative electric potential is then applied to the cathode (N- doped layer) via the small electrodes pads (70µm*70µm) placed on top of each pixel.

Electrical connections are either performed using a probe station for ex- perimental characterisations, or wire bonding in order to link the golden contacts to an external and more accessible electrical track via a 15µm thin golden wire. Those bond connections can be seen on Fig. 2.2b and 2.7 where they are connected from the main ground contacts and pixel contacts toward a driving Printed Circuit Board (PCB). Fig. 2.8 shows the probe sta- tion and the wire bonding machines typically used for those manipulations.

F

2.7: Wire bonded modulator to electrical contacts

2.3. Device features 11

(

) (

)

F

2.8: 2.8a Probe station, 2.8b Wirebonding station

2.3.2 Geometries and bandwidth limitation

As shown on pictures 2.2 and 2.9, a few different types of modulator ge- ometries have been developed at ACREO depending on requirements and specifications. Each individual component is basically built with the same structure but its geometry will of course be advantageously shaped.

(

) (

)

(

) (

)

F

2.9: 2.9a 6*6 pixelated (250µm*250µm) square ar- ray with centred pads, 2.9b Round shaped modulator, 2.9c 400µm*400µm square pixelated modulator, 2.9d Round sin-

gle pixel modulator φ=150µm

Bandwidth limitation

The main concern when it comes to design is of course the size of an indi-

vidual pixel. Indeed, the surface area of a pixel will be the intrinsic limita-

tion for high data rate transmission since it will restrict the bandwidth due

to the induced parasitic capacitor. Under a reversed electric field, the p and

n parallel planes of the pixels, combined with the intrinsic layer, will act

12 Chapter 2. EAMs and characterisation

like a capacitor (see equation 2.2) and will play a limiting role for high fre- quency transmission (see equation 2.3 and subsection 3.2.2). The parasitic capacitance of the junction is deduced from the equation 2.2:

C

= 

A

d

(2.2)

where 

,  are the vacuum permittivity and intrinsic media permittivity (

= 8.85 10

F.m

and  = 12.6F.m

) respectively, A

is the area of a pixel and d

is the distance between the two doped planes and in this case the width of the intrinsic layer.

The theoretical bandwidth limitation is then given by equation 2.3:

f

= 1

2π(r + Z

)C

(2.3)

where r (around 15 ohms) is referred to as the sum of the two p and n con- tact resistances and Z

is the characteristic impedance of the driving circuit.

In most of the cases, 50Ω characteristic impedance electronic components were used and so r can be neglected. Scaling down the pixel would then im- prove,because of the impact of C

, the maximum modulation frequency.

Of course a few parasitics such as the inductance induced by the wires or capacitance induced by the PCB should slightly decrease the performances of the setup and the non ideal electronic components used for the setup also have to be taken in consideration.

The theoretical bandwidth limitation and the induced capacitance as a func- tion of the edge size of a square pixel are displayed in Fig. 2.10.

F

2.10: Cut off frequency and induced capacitance as a function of the pixel size

In order to keep an adequate surface area, pixelating the modulator is then

a viable solution as long as all the pixels are not driven in parallel. Never-

theless, in order to restrict the consumed power of the electrical board, it is

hard to imagine, for example in the case of a 36 pixel modulator (Fig.2.2a),

to drive each pixel with its own amplification stage. A trade off has then

to be found between the number of pixels driven by an amplification stage

2.3. Device features 13

and the required bandwidth (see subsection 3.2.2 and Fig.3.8). The next de- velopment stage would then be to track the position of the laser spot on the modulator array and send an electrical driving signal only the surrounding pixels, decreasing then the number of simultaneously working drivers at a given time and so the power consumption.

Advantageous geometrical features

Golden tracks from the side of the array to each pixel (see Fig.2.2b) are use- ful features when it comes to connecting those pixels. Indeed, the wire bonding connection might damage the pad and what is underneath it. It is then safer and easier to connect the pixels and the golden wires will not shadow any pixel.

The position of pads on the pixels (see Fig.2.9a and 2.2a) also requires some attention since, depending on where the outer electrical contacts are, it will be easier or not to connect them and will affect the chance to deflect the beam with the golden wires.

Beside this, combined with an optical system, the pixelated geometry al- lows the system to achieve simultaneously high speed transmission and large Field Of View (FOV). Using some ingenuity, it could also enable spa- tial multiplexing which would allow communication with multiple inter- rogators or enable information relaying [3](see Fig.2.11) since a pixel can be used for detection like a photodiode as well (see subsection 2.4.3 ).

(

) (

)

F

2.11: 2.11a Spatial multiplexing using a pixelated modulator,2.11a Data relaying using pixelated modulator

partially in detection mode

2.3.3 Anti-reflective coating

In order to maximize the amount of light penetrating the active area of the modulator, an anti-reflective layer is deposited on top of the structure (Fig.

2.1). Inserting such a thin layer will prevent a large amount of light from being reflected at the air-modulator interface. At normal incidence without the ARC, the fraction of reflected power is given by equation 2.4:

R =    

n

− n

n

+ n

2

(2.4)

Using n

= 1 and n

≈ 3.5 for the air-substrate (modulator) interface, it

becomes important to avoid those first 30% reflection losses. Of course, in

14 Chapter 2. EAMs and characterisation

order to calculate this value more precisely in the case where we do not have the ARC, we should take into account all the layers stacked under- neath the first InGaAs layer sould be taken into account. As shown on Fig.

2.12, the principle is to introduce a thin layer where n

< n

< n

in order to make the two out of phase reflected waves to destructively interfere.

n

n

n

out of phase reflected waves

F

2.12: ARC principle

For a normal incidence, we then consider the following optimum thickness in order to introduce the π phase shift :

t = λ 4n

(2.5) Nevertheless, even by fixing the thickness this way, it is necessary to select a material with a suitable refractive index in order to have two interfering waves with the same amplitudes. The refractive index of the ARC media should then obey the following equation:

n

= √

n

n

(2.6)

SiN has then be selected to be the ARC material since its refractive index is around n

= 2 in the NIR domain.

From a process point of view, those calculations do not reflect perfectly the reality since all the stacked structure should be considered. A few deposi- tion tests have then been run in order to converge toward a suitable thick- ness and deposition rate to optimise the wavelength range. Figure 2.13 rep- resents the reflectance spectra for those different ARCs measured by FTIR.

This top ARC is also used as a passivation layer for the whole modulator

and will prevent any unwanted electrical contact and physical damage that

a modulator could receive while being handled.

2.4. Overall DC characterization 15

F

2.13: Reflectance spectra of different ARCs mea- sured by FTIR

2.4 Overall DC characterization

In order to use suitable parameters for the operational communication link, and to understand the way EAMs work, a few typical DC characterisation have been led. The optical as well as the electrical behaviour of coupled quantum wells based EAMs are presented in this section.

2.4.1 Optical measurement on modulators by FTIR

Using a final reflective chip, one can obtain the optical profile at 0V bias resulting from the overall stacking which is shown in figure 2.14.

F

2.14: EAM reflectance optical spectrum at 0V bias voltage measured by FTIR

One can still clearly see the profile similarity with the MQW stacking spec-

trum that we already observed on figure 2.3. On most of the FTIR mea-

surement, one can see that the Y scale do not display any values due to the

16 Chapter 2. EAMs and characterisation

fact that, using those modulators in reflections, the amplitude of the signal depends drastically on the alignment of the sample and the reflective mod- ule. One part of the FTIR beam might also be lost by an aperture limiting element on the setup (sample holder in the case of a mounted modulator) which would scale down the curve.

This measurement has also been performed using a Bruker Vertex70v FTIR machine. A combination of the NIR source and a PIKE Technologies verti- cal reflective module (Fig.2.15) has been used for the measurement in order to facilitate the positioning of the sample on top of it.

F

2.15: PIKE interchangeable FTIR vertical re- flectance module

Applying then a bias voltage to the modulator will result in a few changes in this optical profile and will induce optical modulation to a monochro- matic beam at a suitable wavelength according to subsection 2.2.2. Figure 2.16 displays the resulting optical reflection profiles obtained when apply- ing different DC Bias Voltages.

The Bruker Vertex70v FTIR machine has been used with the associated Bruker I26312 interchangeable FTIR Reflectance module with a mounted reflective modulator.

At a first sight, one can observe a few interesting results from the optical profile curves. As expected, one can clearly see the shift toward longer wavelength of the QW optical reflection dip when applying the voltage.

Indeed, at 0V bias, one of them is centred around 1530nm and shifts up to 1580nm at -6V bias (Fig. 2.16) of similarly from 1540nm up to 1590 nm on Fig. 2.17. Combined with the steep transition around 1545nm, it induces a shift of the whole curve around this wavelength and give rise to a strong contrast ratio peak.

Another CR window can be observed around 1520 nm and comes from the bump inversion around this wavelength, still due to the exciton dip shifts.

In practice, we could select a working wavelength in either of those two

CR peaks but the one induced by the steep part shift presents a major ad-

vantage: compared to the other peak, this one occurs in a spectrum re-

gion where the modulator is more transparent (higher reflectance) and so

2.4. Overall DC characterization 17

F

2.16: EAM reflectance optical spectra at different bias voltages measured by FTIR and related contrast ratio

the amount of sent back light will be larger, making the reflected signal stronger. For the other peak, the modulator is almost opaque, the collected signal then will be way weaker and a larger photocurrent might disturb the electrical setup.

The noisy aspect of the curves is due to the fact that reflectance measure- ment signal is weak as a consequence of the really small reflective modula- tor area and misalignments in the FTIR chamber. Moreover, only 13 pixels were controlled on a 36 pixels EAM totally covered by the collimated beam of the FTIR light source, which results in a weaker evolution of the optical spectrum.

However, some data have previously been recorded on a similar but trans- missive sample by a previous master student with more substantial evolu- tion and are displayed in Fig.2.17.

On those measurement, it is easier to notice the exciton dips shifts toward longer wavelength (Stark effect) and their widening (quantum confinement).

Moreover, the two available CR windows are really clear and even if the

second one is more wavelength selective, the amount of reflected light will

be larger. Knowing that we will use a laser with the wavelength tuned

around 1545 nm for this modulator, one can see that applying a voltage will

actually increase the transmittance/reflectance. The optical signal will then

be flipped compared to that of driving voltage signal. [see Fig.3.12 and3.17

where the displayed driving voltage amplitude (right axis) corresponds to

the absolute value of the negative driving voltage]

18 Chapter 2. EAMs and characterisation

F

2.17: EAM transmittance optical spectra at differ- ent bias voltages measured by FTIR and related contrast ra-

tio

2.4.2 Dark current

The first simple electrical measurement that can be performed on such de- vices is the dark current I-V characteristic. Indeed, using a diode like com- ponent under reverse bias, as long as the breakdown voltage of the diode is not reached, the pixel should generate a small leakage current. Forward bi- asing the modulator would end up in a large current generation that might quickly destroy it, and the same would be observed when reverse bias- ing the modulator close or below its breakdown voltage. No attempt to test these limits have been made and so this subsection will be focussed on characterisations in a safe voltage range.

Using a probe station (Fig. 2.8a) connected to a HP4156A precision semi- conductor parameters analyser allows quick and reliable I-V measurements.

Without any IR light source, and a minimum amount of ambient light (even if the modulator is not that sensitive to those wavelengths), one can obtain reliable curves such as the one displayed on Fig. 2.18.

This figure represents the characterisation of a 12 pixels modulator (K1-16) where each pixel has dimensions of 400µm*400µm ( Fig. 2.9c) and generates a few nA of current.

The first goal of such a systematic measurements is the detection of leaking

pixels and so defective or damaged pixels. One can clearly see on this figure

that pixel K1-16-pix12 has a higher dark current level compared to all the

others.

2.4. Overall DC characterization 19

F

2.18: I-V measurements of the dark current for all pixels on Sample K1-16

A few elements might cause this kind of leaking. Material defects might be generated if the modulator has not been stored in proper conditions for a long time or if it comes from the edges of the wafer where the processes are a bit less accurate. Physical damage, after successive or careless probing or wire bonding which might be violent processes for pixels, will also end up in the generation of a larger dark current level. Depending on the defect or damage that the pixel endured, the leaking level will be more or less pronounced. We can see on figure 2.18 that pixel 12 is leaking moderately.

Pixel K1-16-pix10, on the other hand, has been removed from this figure due to its very high leaking level. This can occur when the damages are too important (pixel destroyed after a violent probing on electric contacts for example), so that leaks might become severe (up to several µA for a small bias voltage).

After a certain leaking level, pixels will not be light sensitive anymore and so they become useless and even problematic from an electrical point of view. They should then be disconnected from the electrical driving setup.

In practice, such measurements can be done in a comparative way in order for example to show that the wire bonding process increases slightly the dark current level if it does not destroy the pixel.

One can also observe that all the dark current curves, except for the leaking pixel K1-16-pix12, are quite similar. This tells us about the good uniformity of the structure over the whole modulation since in this case the sample comes from the central part of a processed wafer.

Theoretically, this leaking current is generated by two processes. The main one is linked to the thermal enhanced generation of electron-holes pair in the intrinsic region. This current will then be proportional to the surface of the pixel and will present a large temperature dependence (see subsection 4.1.1).

The second source of current in absence of any optical shining will be in-

duced by non terminated bonds at the edges of the pixels created during

20 Chapter 2. EAMs and characterisation

the dry etching process. A passivation via wet etching might decrease its contribution.

The total amount of generated current can then be expressed as follows:

I = J

(T ) ∗ A + J

(T ) ∗ L

(2.7) Where J

is the surface current density induced by the thermal current, J

the linear current density induced along the edges of the pixel, A the area of the pixel and L

the contour length of a pixel.

The small pixel size prevents us from neglecting the edge generated current with respect to the thermally-enhanced surface current. At this scale the dark current can no longer be considered to be proportional to the surface area of the pixels. This result can be deduced from figure 2.19:

F

2.19: Dark current histograms for different modu- lators at -4V bias voltage

Those measurements at room temperature have been conducted on mod- ulator K1_16 which is a 400µm*400µm square pixels modulator (fig: 2.9c), modulator K1_11 is a modulator with 250µm*250µm square pixels (fig:2.2a) and modulator K1_13 is a round modulator with pixels surface areas equiv- alent to a 400µm*400µm square pixel (fig:2.9b).

From modulator K1_11 and K1_16, since each pixel is a square surface, equation 2.7 becomes:

I = J

(T ) ∗ L

+ J

(T ) ∗ 4L (2.8) where L is the edge length of a pixel. A quick calculation gives us the fol- lowing values: J

(T = 300K)

= 65mA.m

and J

(T = 300K)

= 2µA.m

.

The dark current levels of K1_16 and K1_13 can not be differentiated since

the pixel area is identical and the contour length difference remains very

weak even with the different geometries.

2.4. Overall DC characterization 21

2.4.3 IV measurement and optical responsivity

Contrary to the previous subsection where we looked at the electrical be- haviour of modulators without any light interaction, we are now looking at their characterisation when exposed to a monochromatic laser source.

Intuitively, we can clearly state that the generated photocurrent will be di- rectly linked to the absorption of the modulator, depending on the wave- length. Indeed, this current is generated by formation of electron-hole pairs in the intrinsic region when a photon is absorbed. The applied electric field will then drive them apart before a recombination occurs, resulting in a current.

This current can be calculated with the quantum efficiency of the modulator in detection mode as stated below:

I(λ, V ) = µ(λ, V ) ∗ Φ (2.9)

I(λ, V ) = µ(λ, V ) ∗ P λ

hc (2.10)

where P is the optical power, and µ, the quantum efficiency, corresponds to the ratio of generated electrons to the number of incoming photons.The generated photocurrent is thus proportional to the optical power that the pixel receives.

The typical result for the characterisation of one pixel can be seen in figure 2.20.

We can observe on the top figure that from nA for the dark current, pixels generated up to a several µA with only a few mW of optical power ( a fraction of 2mW for this measurement).

The bottom figure represents the responsivity of the pixel, which indicates the amount of current generated per unit of optical power for different wavelengths. The responsivity is then also tightly linked to the absorbance characteristic of the modulator and its quantum efficiency. In our case, the representation of the absorbance and so the responsivity would basically be similar to a inverted version of the reflectance/transmittance spectrum of the EAM (see Fig.2.16 and 2.17).

Unfortunately, due to the limited range of the tunable laser that has been used, the responsivity curve is not displayed on a large spectral range like for the optical characterisation done by FTIR. Nevertheless, one can cor- relate the displayed part of the responsivity to the upper portion of the reflectivity profile shown previously (from 1550nm to 1600 nm on fig. 2.16):

The wavelength range seems to be shifted between those two ways of char-

acterisation simply because different modulators from different process batches

have been used resulting in slightly different operational wavelengths. How-

ever, we can also clearly see the Stark Effect shift of the QW bump when a

bias voltage is applied as explained in 2.2.2. Centred around 1545nm at -2V

bias, it shifts toward 1570nm for a -6V bias similarly to its behaviour previ-

ously observed on the reflectance and transmittance spectra in figure 2.16

and 2.17.

22 Chapter 2. EAMs and characterisation

F

2.20: I-V measurement of the photo current for dif- ferent wavelength shining and related responsivity

For this measurement, a collimated beam was sent through a glass window

toward the characterised pixel. The transmission spectrum of the glass was

taken into account after measurements for corrections. Nevertheless, the

numerical values given for the responsivity are not significant since only a

fraction of the collimated laser beam (≈ 2mm wide) was collected by the

400µm*400µm square pixel.

23

Chapter 3

FSO asymmetrical

communication setup

3.1 Asymmetrical FSO communication setup

This chapter will be focussed on the discussion of practical solutions imple- mented on an indoor test bench of an asymmetrical FSO link. The principle of such a network has already been discussed and introduced in subsection 1.1.2 and a simple principle diagram is shown in figure 3.1.

Signal generation and driving electronics

MRR Tunable laser source

Tracking system Detection setup Signal processing circuit

Interrogator Transmitter

F

3.1: Asymmetric FSO link principle diagram

All the heavy and power consuming hardware elements are set on the in- terrogator side while just as few and as lightweight elements as possible on the transmission side of the link. On the transmitter side, the driving electronics play a large role in the performances and the maximum com- munication speed of the system. Section 3.2 will be dedicated to this matter while section 3.3 and 3.4 will be focused on the optical aspect, specifica- tions, evolutions of the setup and the achieved performances. At first, a 0.5m link has been set on the test bench and will be discussed in section 3.3.

The final operational 10m communication test bench will be described in section 3.4 and the main high speed AC results will be treated then.

3.2 Electrical driving

A typical design for driving board electronics is shown in figure 3.2:

24 Chapter 3. FSO asymmetrical communication setup

F

3.2: Typical driving board where the modulator, in the centre, is bounded to electrical contacts and surrounded

by the driving electronic components

The EAM array in the middle is wire bonded to electrical channels (see Fig.

2.7) on the boards that connect it to the surrounding driving components.

The board dimensions of 6cm*4.5cm makes it quite compact. Included in a housing with the optical retro reflector, those would be the only elements on the transmitter side of the link where the main mass comes from the optical elements and the housing.

From a development point of view, a more rudimentary and larger scale electronic setup has been used in order to test the components behaviours and performances with more freedom. A few suitable elementary compo- nents such as various types of SMA connectors and cables have been pur- chased in order to furnish the laboratory and ensure the self-sufficiency of the setup concerning basic supplies.

For communication speeds above 50 MHz, electronic problems appear on such circuits and had to be progressively overcome. Indeed, since the wave- lengths of the generated signals start to shrink when higher frequencies are reached, non static behaviours can be encountered. Around 300MHz (elec- trical wavelength around 1m) the size of the electrical setup, including ca- bles, becomes comparable to the signal wavelength, a few effects such as re- flections, built in standing waves or interferences might be observed in the electrical driving circuit and will drastically disturb it and reduce its high speed bandwidth. Antenna effects, as well as induced crosstalk in-between cables, also have to be prevented to avoid any misinterpretations.

In order to have a flat frequency response of the driving electronic circuit and so a large and neat operational bandwidth from 50MHz up to 500MHz, all those parameters had to be assimilated and improved.

3.2.1 Electrical driving setup

In order to have a good optical contrast ratio and enable the optical detec-

tion of a modulated signal, a 6V peak to peak input temporal signal has to

be delivered to the modulator. Due to its diode electrical profile, it should

always be fed under reversed bias. A final signal centred on -3V with a 6V

3.2. Electrical driving 25

peak to peak amplitude for a large bandwidth would be the perfect driving voltage signal and would ensure a decent optical CR on the optical reflected beam.

The following setup ( Fig.3.3) is then used in order to generate such a signal at the EAM terminals:

Signal generator

Bias Tee

VDC=-3V VDD

ZMatch ZPara

EAM Amplification Biasing Stage Matching

C_DC

CDC CBias BiasL

Modulator

F

3.3: Diagram of the electrical driving setup

Signal generation

The signal generator is chosen depending on the kind of signal we want to transmit. For sine wave signals, a HP 8657B generator is convenient and allows the generation of waves up to 2GHz with adjustable amplitude.

Square wave signals are generated via a computer controlled IDT UFT 3G 8T49N242I test board that delivers 0.3V peak to peak LVDS signal up to 3GHz and a 2.5V peak to peak LVCMOS signal up to 250MHz. Pseudo- Random Binary Sequences (PRBS) are very nicely generated via an AN- RITSU pulse pattern generator MP1701A with adjustable amplitude and frequency.

Signal amplification

The amplification stage, which is the major power consuming step, should generate a suitable voltage swing (around 6 V amplitude) from the previous signals, centred on 0V due to the two DC blocking capacitors (C

on Fig.

3.3).

A Hittite HMC471MS8G dual amplifier is used for sine waves and square waves amplifications. Its 20 dB gain allows us to convert with low added noise and distortion a 0.5V peak to peak signal into a nearly 6V peak to peak signal without saturation or undesirable effects. Indeed, using such component carelessly by saturating its input will result in distorted sig- nal generation and so in strongly frequency-dependent signals which will then reduce drastically the operational bandwidth. A V

=11-12V power (0.13A) supply has been used in order to prevent such saturation and al- low a 6V output. Nevertheless, the large bandwidth (from DC up to 5GHz) of this amplifier makes it safe but power consuming for this application.

In our case, it is suitable even for square waves applications since it will

amplify the few harmonics that the IDT board can deliver. A low gain pre-

amplification stage could be required for LVDS signals in order to reach the

0.5V amplitude as an input signal and so obtain the best voltage swing for

26 Chapter 3. FSO asymmetrical communication setup

the modulation. Contrariwise, a 16dB attenuation is introduced between the IDT SMA output and the Hittite amplifier to turn the 2.5V amplitude output of the LVCMOS signal into a suitable 0.5V peak to peak signal.

For PRBS signals, the ANTRISU pulse pattern generator MP1701A creates a very neat and steep signal that contains very high frequency harmonics.

Even with a proper 0.5V amplitude signal at its input, the Hittite dual am- plifier is unable to deliver a proper amplified signal due to the too high frequency components of the input. Indeed the wide bandwidth of the am- plifier will ensure the proper amplification of the high speed components and neglect the amplification of the lower order harmonics, resulting in an unusable signal. For those pseudo random signals, a more robust and high gain Minicircuit ZKL 1R5 amplifier is used under a V

=13V power sup- ply to generate a proper pseudo random driving signal. The 40dB of this amplifier gain requires a 10dB attenuation of the 0.5V input amplitude for safety reasons even if this amplifier deals really well with saturation.

Taking every precaution to avoid any reflections ( adding an external 50 ohm oscilloscope channel for example or use mismatched propagation medium) or saturation of the amplifier, it is however still impossible to drive the mod- ulator with this signal since it would be forward biased during part of the time. A negative biasing stage is then used to centre the signal around neg- ative voltages and ensure the reverse biasing of the modulator at all times.

Biasing

In order to shift the signal to negative voltages, a Picosecond Pulse Labs bias tee 5555 is used to apply a -3V DC component to the signal. This bias tee is simply equivalent to a capacitor that will let high frequencies go through and block DC components, and an inductor that will let the low frequency component in and will prevent the AC signal to leave the circuit via the DC supply(see Fig.3.3). The resulting signal is then simply the sum of the amplified AC component and the negative bias voltage.

Typical driving signals are displayed in figure 3.4 for a sinusoidal signal and a PRB signal at 100 MHz and 300 M bit.s

respectively.

(

) (

)

F

3.4: 3.4a Driving signal for sine wave at 100MHz, 3.4b Driving signal for pseudo random bit communication

(on-off keying) at 300M bit.s

3.2. Electrical driving 27

The visualisation and recordings of electrical signals are performed via a 4 channel TEKTRONIX TDS 754A oscilloscope and a developed python pro- gram in order to retrieve signals from the oscilloscope.

Once those kind of signals are generated, they can safely be applied to the modulator. The insertion of a matching impedance between this driving circuit and the PCB on which the modulator is mounted and wire bonded is then required and is be investigated in 3.2.2.

3.2.2 Impedance matching

For the sake and proper functioning of the general setup, the EAM (round single pixelated modulator φ=150µm (see Fig.2.9d) on the small blue chip on picture 3.5) is glued to a RF PCB and properly wire- bonded to the board’s two channels as shown in figure 3.5. The surrounding channel will be considered as the ground plane whereas the middle track will carry the previously generated signal. A female SMA connector has been soldered (see in Fig.3.5) in order to respect those specifications and facilitate the de- vice operation. The ground (outer shell of an SMA cable) and signal (core of an SMA cable) of the driving stage of the circuit (detailed in subsection 3.2.1) will then be directly connected to the two channels of the PCB and so with the EAM electrodes.

F

3.5: Experimental PCB with SMA connection, matching impedance and wire bonded EAM

This simple connection could work properly in this configuration if the treated signals remained at low frequencies. Yet, for RF frequencies, if two propagation media do not have the same characteristic impedance, re- flected waves might be generated.

In our case, from the transmission lines (simple 50Ω SMA cable and SMA connector) to the unmatched PCB, the impedance changes suddenly, result- ing in a large reflection coefficient according to equation 3.1.

Γ

= Z

− Z

Z

+ Z

(3.1)

28 Chapter 3. FSO asymmetrical communication setup

Γ

is in this case the voltage reflection coefficient at a Z

/Z

impedances interface. An analogy with the optical reflection between two media with different refractive indexes can be observed.

In order to prevent this reflection, setting the impedance of the board to a more appropriate value has been demonstrated to be a good solution. More specifically, the impedance of the total load should be in theory set to the complex conjugate impedance of the source ( Z

= 50Ω in our case) in order to maximize the power transfer.

Z

Z

Z

Transmission Line ( SMA) Matching Board and modulator

EAM

F

3.6: Electrical matching configuration

Maximising the signal transfer, to the PCB media and so to the EAM, is per- formed in practice by soldering a predominant load (Z

) in-between the ground plate and the signal track on the RF board ( see Fig.3.6 and picture 3.5 ). A few resistors values has been tested and the S

reflection coefficient has been measured using a network analyser. The results are dis- played on figure 3.7. This measurement has been conducted with a single round 150µm pixel wire bonded to the RF board where the SMA connector has been soldered with the matching resistance in-between the ground and the signal tracks.

F

3.7: S

coefficient for different matching resistors

3.2. Electrical driving 29

Knowing that the parasitics impedance (Z

) remains negligible in this case, as expected, we can observe that the least reflection matching is given for the case where the load resistance is the closest to the characteristic impedance of the source Z

= 50Ω.

In this configuration, Z

accounts for the induced capacitance of the modulator (see subsection 2.3.2), and all the induced parasitics components such as the capacitance introduced by the board, wires and tracks induced inductances, and parasitic resistors. In this experimental setup configura- tion, the parasitics induced by the board are minimised since this kind of RF PCB have been specially designed for this purpose. However, the amount of parasitics in the case of a final board (see Fig. 3.2) might need to be quan- tified.

Moreover, this kind of matching works quite well in our case simply be- cause the 56Ω termination predominates over Z

. Indeed, as long as Z

remains electrically hidden behind the 56Ω characteristic impedance imposed to the board, the signal transfer should not be disturbed. Nev- ertheless, if Z

reaches a state where it is no longer negligible compare to the characteristic impedance, for example increasing the number of pix- els driven by the same circuit, the matching would have to be rethought, otherwise some unwanted frequency-dependent behaviours might appear, reducing the general bandwidth, the signal quality and performances of the setup. Of course, some finer matching using a Pi-matching or T-matching network could be required in those cases.

The maximum number of pixels that can be driven in parallel by a single driver matters since it would drastically impact the power consumption of a board when it comes to drive a whole 36 pixels EAM. The final decision would of course come from a trade off with the desirable operational band- width which should be slightly lower than the theoretical limitation (see Fig.2.10). Increasing the number of driven pixels in parallel scales up the used surface and so the induced capacitance of the driven group. At one point we should be able to see that the modulator itself becomes a limiting factor for the bandwidth due to its decreasing impedance. This group of pixels will then not be negligible any more compared to the 56 Ω matching on the electrical setup and its bandwidth will then be reduced.

In order to quantize the impact of this induced mismatch, a measurement of the bandwidth with 0 to 4 connected 250µm*250µm square pixels have been performed. A sine signal was generated with an amplitude of -1dBm with the HP 8657B signal generator in order to ensure a proper voltage swing and avoid distortion and saturation after amplification. The Hittite dual amplifier under 11V bias has been used followed by the bias tee and connected to the 56Ω matched boards with 0 to 4 bonded pixels. The driv- ing amplitudes at different frequencies have then been measured using a Tektronix P6243 active probe in order to minimise the probing impact and are displayed on figure 3.8.

The "No pixel Connected" curve corresponds to the unperturbed driving

voltage where the output of the bias tee was directly connected to a 50Ω

channel of the oscilloscope. Its decreasing trend is related to a combination

30 Chapter 3. FSO asymmetrical communication setup

0 1 2 3 4 5 6

0 200 400 600 800 1000

V

Frequency (MHz)

No pixel Connected 1 pixel connected 4 pixels connected

F

3.8: Frequency response for different numbers of connected pixels

of the limited performances of the signal generator and the used oscillo- scope.

One can perfectly see that for 1 connected pixel on a 56Ω matched board, the signal seems to be quite unperturbed up to 500MHz. Nevertheless, the more pronounced frequency dependence can be related to the small mis- match induced by a simple pixel. The cut-off frequency is then already slightly affected. Introducing then a 56Ω matched board, where 4 pixels have been connected, clearly shows a strongly problematic behaviour. In- deed, even in the lower frequency range (bellow 300MHz), the driving sig- nal amplitude applied to those pixels is heavily perturbed.

Similar measurements have been led with the round single pixel EAM which has been mainly used for the operational FSO link (see subsection 3.3 and 3.3). Its 150µm diameter becomes quite suitable since the really small in- duced capacitance remains very weak and keeps the 56Ω matching reliable up a larger frequency range (see subsection 3.4.3).

For this setup, the impedance was set due to the use of conventional elec- tronics components such as the SMA cables, the signal generator, the bias tee and the generation test board, and fixed to a characteristic impedance of 50Ω. Matching impedances was then a necessity. The next step would be to test a "home made" driving board (similar the the one shown in fig- ure3.2), where nothing fixes the impedance to 50 Ω, resulting in a lower power consumption and heating of the setup and expected better perfor- mances for even higher driving frequencies. Indeed, this 56Ω matching, even if it seems quite robust in our range of interest, remains relevant only up to a certain frequency and for small enough driven areas.

For further improvement, this work would of course necessitate a more

proper matching through more complex structures such as T and Pi struc-

tures. Nicer and more precise driving circuits will definitely induce better

performances for high speed data transfer on paper but they might never-

theless be more fragile and selective when it comes to small characteristics

changes. All the parasitics from the board should then be finely quantized

3.2. Electrical driving 31

in order to avoid surprises. One other main parameter that would have to be taken into account is the fact that when light is shined on the modulator, a photocurrent is induced. Remaining quite bias dependent (see Fig. 2.20), this generated current might completely disturb this sharper matching.

This result has already been observed using the 56Ω matched setup with different laser powers used for communication . The evolution of the mod- ulated optical response (direct image of the electrical driving setup) shows that the shined optical power and so the amount of induced photocurrent already disturbs the electrical circuit (see Fig.3.9).

0 0.2 0.4 0.6 0.8 1 1.2

50 100 150 200 250 300 350 400 450 500

V /V

Frequency (MHz)

12dBm optical power 6dBm optical power

F

3.9: Frequency dependency of the optical response with different optical powers

Using the the setup that will be described in section 3.3, the optical mod- ulated amplitudes have been measured for different laser power outputs:

from 6 dBm (around 1.5 mW ) up to 12 dBm (around 7 mW). For this mea- surement, the round single pixel Φ = 150 µm EAM has been used at its operational wavelength of 1562 nm. The same sinusoidal input electrical driving signal has been sent in both cases and the optical retro-reflected signal peak to peak amplitude (V) has been recorded at different frequen- cies for the two optical powers.

The first result we can observe is that the relative signal amplitude evolu- tion is more affected by larger laser powers. The optical response becomes more frequency dependant (ripples starts to appear) and the general rela- tive amplitude remains lower when increasing the optical power.

Due to its bias dependency, the generated photocurrent will obviously con-

tain a DC component as well as an AC component, image of the driving

signal. On one hand, the DC current should not be able to travel all the

way through the circuit due to the DC blocking capacitors. It would then

be drained away via either the bias tee inductor or the matching load. In

the last case, it will obviously slightly disturb the matching in-between this

load and the 50Ω characteristic circuit. On the other hand, the high speed

component of this photocurrent certainly interferes with the driving signal,

resulting in a more frequency-dependent profile.