Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
Title of Dissertation: TECHNIQUES TO MITIGATE THE EFFECTS OF
ATMOSPHERIC TURBULENCE ON FREE SPACE
OPTICAL COMMUNICATION LINKS
Linda Marie Wasiczko, Doctor of Philosophy, 2004
Dissertation directed by: Professor Christopher C. Davis
Department of Electrical & Computer Engineering
Free space optical communication links are an attractive technology for
broadband communications when fiber optic links are unavailable or simply not
feasible. Atmospheric turbulence, aerosols, and molecular absorption all affect the
propagation of optical waves in the atmosphere. Since atmospheric turbulence is the
major source of errors on free space optical communication links, this dissertation
investigates two techniques to reduce the impact of atmospheric turbulence on such
links. These two techniques are aperture averaging and the incorporation of
nonimaging optical elements into optical receiver systems.
Aperture averaging is the process by which atmospheric turbulenceinduced
intensity fluctuations are averaged across a receiver aperture of sufficient size. We
investigate the behavior of aperture averaging in weak and strong turbulence conditions
by comparing experimental data with available models for plane and spherical wave
propagation. New expressions for the aperture averaging factor in weak turbulence are
given. In strong turbulence conditions, aperture averaging is analyzed with special
attention to the various wavenumber spectrum models. This is the first report of
experimental strong fluctuation aperture averaging data acquired in nonsaturated
conditions.
Nonimaging optical elements are particularly useful for the mitigation of
atmospheric turbulenceinduced beam wander in the focal plane of a free space optical
communication receiver. Experimental results of the bit error ratio enhancement due to
the incorporation of a nonimaging optical element, specifically a compound parabolic
concentrator, are presented. Two link ranges were tested, a 1.7 km link at the
University of Maryland experiencing weak turbulence, and a 32.4 km link at the Naval
Research Laboratory’s Chesapeake Bay Detachment experiencing saturated, strong
turbulence. These results are the first reported experimental test of a nonimaging
optical element integrated into an outdoor free space optical communications system.
TECHNIQUES TO MITIGATE THE EFFECTS
OF ATMOSPHERIC TURBULENCE ON
FREE SPACE OPTICAL COMMUNICATION LINKS
by
Linda Marie Wasiczko
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2004
Advisory Committee:
Professor Christopher C. Davis, Chair
Professor Pamela Abshire
Professor Agisilaos A. Iliadis
Professor Thomas E. Murphy
Professor Owen E. Thompson
©Copyright by
Linda Marie Wasiczko
2004
ii
Dedication
To my Parents, Bernadine & Dennis,
for teaching me perseverance,
humility, and faith
Keep smiling
iii
Acknowledgements
I want to express my appreciation to my advisor, Prof. Christopher Davis, for his
wisdom, friendship, and support during this experience. I also want to thank the
members of my dissertation committee for their service and guidance.
I want to thank two dear colleagues and friends for their continuing
encouragement: Dr. Patricia Mead and Dr. H. Craig Casey.
I am thankful for the technical and nontechnical discussions with my colleagues
in the Maryland Optics Group at the University of Maryland, especially: Dr. Igor
Smolyaninov, Dr. Stuart Milner, Dr. Quirino Balzano, and Dr. Vildana Hodzic. I also
want to thank Dr. Kyuman Cho for introducing me to this project.
I have enjoyed fruitful conversations and collaboration with members of the free
space optics community, especially: Dr. Ray Burris, Dr. Chris Moore, and Mr. Michael
Vilcheck, of the Naval Research Laboratory; Dr. Jennifer Ricklin and Dr. Mikhail
Vorontsov, of the Army Research Laboratory; and Dr. Charles Thompson of the
Lawrence Livermore National Laboratory. I appreciate the equipment loans to this
dissertation research from Prof. Agis Iliadis, Dr. J. “Buzz” Graves at AOptix, Inc., and
Prof. Mario Dagenais.
The American Association of University Women sponsored my dissertation year
work through a Selected Professions Fellowship, for which I am extremely grateful.
Finally, I want to thank my parents, my sister, Pamela, and my friends who have
supported (and antagonized) me throughout my many years in graduate school.
iv
TABLE OF CONTENTS
List of Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Chapter 1: Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 History of Free Space Optical Communication . . . . . . . . . . . . 1
1.2 Free Space Optical Communication Techniques . . . . . . . . . . . 2
1.3 LowOverhead Techniques to Mitigate the Effects of Atmospheric
Turbulence
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Aperture Averaging
. . . . . . . . . . . . . . . . . . . . . 4
1.3.2 Nonimaging Optics
. . . . . . . . . . . . . . . . . . . . . 5
1.4 Useful Terminology
. . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chapter 2: Optical Wave Propagation through Atmospheric Turbulence
. . . 9
2.1 The Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Turbulent Energy Flow
. . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Kolmogorov Turbulence
. . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Wavenumber Spectrum Models
. . . . . . . . . . . . . . . . . . . . 15
2.5 Rytov Approximation . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.6 Strong Turbulence Theory . . . . . . . . . . . . . . . . . . . . . . . 25
2.6.1 Andrews – Prokhorov Asymptotic Analysis
. . . . . . . . . 25
2.6.1.1 Andrews Asymptotic Analysis for the Plane Wave 26
2.6.1.2 Andrews Asymptotic Analysis for the Spherical
Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.6.2 Churnside Asymptotic Analysis. . . . . . . . . . . . . . . . 30
2.6.2.1 Churnside Asymptotic Analysis for the Plane
Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
2.6.2.2 Churnside Asymptotic Analysis for the Spherical
Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.7 Irradiance Variance Models Valid in Both Weak and Strong
Turbulence Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.7.1 Scintillation Index Model for a Plane Wave. . . . . . . . . 32
2.7.2 Scintillation Index Model for a Spherical Wave. . . . . . . 33
2.7.3 Scintillation Index Model using the Atmospheric Spectrum. 34
Chapter 3: Aperture Averaging in Weak Turbulence
. . . . . . . . . . . . . 37
3.1 Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Data Analysis for Weak Turbulence
. . . . . . . . . . . . . . . . . . 38
3.2.1 General Aperture Averaging Form. . . . . . . . . . . . . . 38
3.2.2 Weak Turbulence. . . . . . . . . . . . . . . . . . . . . . . 38
v
3.2.2.1 Plane Wave, Small ℓ
o
. . . . . . . . . . . . . . . 39
3.2.2.2 Plane Wave, Large ℓ
o
. . . . . . . . . . . . . . . 41
3.2.2.3 Plane Wave, ℓ
o
on the Order of the Fresnel Length 42
3.2.2.4 Spherical Wave, Small ℓ
o
. . . . . . . . . . . . . 42
3.2.2.5 Spherical Wave, Large ℓ
o
. . . . . . . . . . . . . 44
3.2.2.6 Spherical Wave, ℓ
o
on the Order of the Fresnel
Length. . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
3.3 Aperture Averaging Experiment. . . . . . . . . . . . . . . . . . . . 44
3.3.1 Aperture Averaging Transmitter and Receiver Systems. . . 45
3.3.2 LabVIEW Data Acquisition . . . . . . . . . . . . . . . . . 50
3.3.3 Fetch Effects . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4 Plane Wave Experimental Results
. . . . . . . . . . . . . . . . . . . 54
3.5 Spherical Wave Experimental Results
. . . . . . . . . . . . . . . . . 57
3.6 New Aperture Averaging Model. . . . . . . . . . . . . . . . . . . . 60
3.7 Determination of the Inner Scale of Turbulence, ℓ
o
. . . . . . . . . . 62
3.8 Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Chapter 4: Aperture Averaging in Strong Turbulence . . . . . . . . . . . . . 68
4.1 Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.2 Data Analysis in Strong Turbulence
. . . . . . . . . . . . . . . . . . 68
4.2.1 General Aperture Averaging Form. . . . . . . . . . . . . . 68
4.2.2 Plane Wave Analysis. . . . . . . . . . . . . . . . . . . . . 68
4.2.2.1 Churnside Asymptotic Analysis for the Plane
Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
4.2.2.2 Andrews Asymptotic Analysis for the Plane Wave 72
4.2.2.3 Scintillation Index Model for the Plane Wave
. . . 73
4.2.2.4 Comparison of Strong Turbulence Models for the
Plane Wave
. . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2.3 Spherical Wave Analysis. . . . . . . . . . . . . . . . . . . 76
4.2.3.1 Churnside Asymptotic Analysis for the Spherical
Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2.3.2 Andrews Asymptotic Analysis for the Spherical
Wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.2.3.3 Scintillation Index Model for the Spherical Wave. 80
4.2.3.4 Comparison of Strong Turbulence Models for the
Spherical Wave . . . . . . . . . . . . . . . . . . . . . . . 80
4.3 Aperture Averaging Experiment
. . . . . . . . . . . . . . . . . . . . 82
4.3.1 Experimental Setup
. . . . . . . . . . . . . . . . . . . . . 82
4.3.2 LabVIEW Data Acquisition
. . . . . . . . . . . . . . . . . 83
4.4 Experimental Results for the Plane Wave Case
. . . . . . . . . . . . 83
4.5 Experimental Results for the Spherical Wave Case
. . . . . . . . . . 88
4.6 Conclusions
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Chapter 5: Free Space Optical Communication using Nonimaging Optics
. . 95
5.1 Introduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2 TurbulenceInduced Beam Motion. . . . . . . . . . . . . . . . . . . 95
vi
5.2.1 Beam Wander
. . . . . . . . . . . . . . . . . . . . . . . . 95
5.2.2 AngleofArrival and Image Dancing
. . . . . . . . . . . . 96
5.3 BER in OnOff Keyed Systems
. . . . . . . . . . . . . . . . . . . . 97
5.4 Nonimaging Optics .
. . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.4.1 Theory of the Compound Parabolic Concentrator . . . . . . 101
5.4.2 The DielectricFilled CPC
. . . . . . . . . . . . . . . . . . 103
5.4.3 The CPCPhotodetector Combination . . . . . . . . . . . . 105
5.5 1.7 km Link Experiment . . . . . . . . . . . . . . . . . . . . . . . . 106
5.5.1 Experimental Setup and Test Range
. . . . . . . . . . . . . 106
5.5.2 ZnSe CPC . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.5.3 Link Budget . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.5.4 C
n
2
Measurements
. . . . . . . . . . . . . . . . . . . . . . 110
5.5.5 Experimental Results and Analysis
. . . . . . . . . . . . . 111
5.6 NRL Link Experiment over the Chesapeake Bay
. . . . . . . . . . . 115
5.6.1 Experimental Setup and Test Range
. . . . . . . . . . . . . 115
5.6.2 CPCPD Combination
. . . . . . . . . . . . . . . . . . . . 117
5.6.3 Experimental Results and Analysis
. . . . . . . . . . . . . 118
5.7 Conclusions
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Chapter 6: Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.1 Summary of Contributions . . . . . . . . . . . . . . . . . . . . . . . 125
6.2 Future Work. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
References
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
vii
LIST OF TABLES
2.1: Typical Rytov variance ranges corresponding to weak, intermediate, and
strong turbulence levels. . . . . . . . . . . . . . . . . . . . . . . . . . .
22
3.1: List of 13 values output per averaging interval in LabVIEW data file.
Here, Channel 0 represents the aperture averaged signal, while Channel 1
represents the scintillometer signal. . . . . . . . . . . . . . . . . . . . .
53
5.1: Link budget analysis for the 1.7 km link, neglecting atmospheric
turbulence.. . . . . . . . . . . . . . . . . . . . . . . .
110
viii
LIST OF FIGURES
2.1: Depiction of the process of turbulent decay, showing the energy cascade
and subsequent division of turbulent eddies in the atmosphere [adapted
from Ref. 13]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
2.2: Three wavenumber spectrum models for refractive index fluctuations.
The three energy ranges relevant to turbulence statistics are indicated. . .
18
2.3: Scaled spectral models for the von Karman spectrum, Eq. (2.24), and
modified atmospheric spectrum, Eq. (2.25), plotted against the
wavenumber scaled by the inner scale. . . . . . . . . . . . . . . . . . . .
19
2.4: The Fresnel zone size, coherence length, and scattering disk size plotted
for a plane wave against propagation distance. For L = 863 m, there are
scale sizes on the link contributing to strong fluctuations. The shaded
area shows scale sizes that do not contribute to strong fluctuations. . . . .
23
2.5: Three scale sizes plotted against the propagation distance for the
spherical wave case with L = 863 m. When C
n
2
= 5×10
15
m
2/3
the
wavefront sees weak scintillations with contributing scale sizes on the
order of the Fresnel zone size. . . . . . . . . . . . . . . . . . . . . . . .
24
2.6: Three scale sizes plotted against the propagation distance for the
spherical wave case with L = 863 m. From the graph, the wavefront
experiences strong fluctuations when C
n
2
= 5×10
13
m
2/3
, where the
contributing scale sizes are outside the shaded area. . . . . . . . . . . . .
24
3.1: Aperture averaging factor for a plane wave plotted against the ratio of the
aperture radius to the Fresnel zone size. The solid line is the exact theory
from Eq. (3.4) while the dashed line is Churnside’s approximation in Eq.
(3.7) [11]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
3.2: Aperture averaging factor for a spherical wave with small inner scale.
The solid line is the exact theory given a Kolmogorov spectrum, while
the dashed line is the approximate formula of Eq. (3.15) [11].. . . . . . .
43
3.3: Aperture averaging transmitter on the roof of the A.V. Williams Building
at the University of Maryland, College Park. . . . . . . . . . . . . . . .
46
3.4: An aerial photograph of the propagation path (dotted line). Satellite
photograph taken in 2002 by GlobeXplorer. . . . . . . . . . . . . . . . .
47
ix
3.5: Diagram of the aperture averaging receiver setup. The scintillometer
channel uses a 5 mm diameter receive lens. . . . . . . . . . . . . . . . .
49
3.6: An optically chopped signal propagated over 863 m and viewed on the
Tektronix oscilloscope. Channel 1 is the signal detected by the
scintillometer, while Channel 2 is detected by the aperture averaged
receiver. The signal detected on Channel 2 shows smaller intensity
fluctuations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
3.7: Front panel of the LabVIEW program designed to calculate and record
irradiance statistics for the aperture averaging experiment. . . . . . . . .
50
3.8: Flowchart of the data acquisition process in LabVIEW for one channel.
The diagram explains how intensity and background levels are
determined for the sampled signal averaged over a 1 min. interval. . . . .
52
3.9: Plane wave aperture averaging factor plotted as the Churnside
approximation Eq. (3.7), the Andrews approximation Eq. (3.8), and the
scintillation index (SI) model for various values of C
n
2
. . . . . . . . . . .
55
3.10: New mean experimental data plotted with error bars, along with
approximations to the aperture averaging factor. . . . . . . . . . . . . .
56
3.11: New mean aperture averaging data plotted with Churnside experimental
data taken over a 500 m path. Neither the plane wave Churnside
approximation nor the SI model are a good fit to the data. . . . . . . . .
57
3.12: Comparison of two different models for the spherical wave in weak
turbulence. The Churnside approximation, Eq. (3.7), predicts a higher
knee in the curve than the SI model. . . . . . . . . . . . . . . . . . . . .
58
3.13: Experimental mean aperture averaging data plotted against the ratio of
aperture radius to the Fresnel zone size. . . . . . . . . . . . . . . . . . .
59
3.14: Comparison of our new mean data with the Churnside data.. . . . . . . . 60
3.15: New weak turbulence fit (Eq. 3.18) shown with new data and the
Churnside model (Eq. 3.15). . . . . . . . . . . . . . . . . . . . . . . . .
61
3.16: New fit to weak turbulence data with an additional square root term,
given by Eq. (3.19), plotted along with the Churnside approximation in
Eq. (3.15). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
3.17: Eq. (3.11) plotted using the curve fit value of ℓ
o
= 5.27 cm, against the
ratio of the aperture diameter to the inner scale. The dashed line
represents Eq. (3.11). . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
x
3.18: Spherical wave aperture averaging factor for ℓ
o
= 1.45 cm using Eq.
(3.17) plotted along with the new data. The dashed line represents Eq.
(3.17). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
64
3.19: New data is plotted using the predicted value ℓ
o
= 4.008 mm. The inner
scale was determined using the atmospheric spectrum, which is
represented by the dashed line. . . . . . . . . . . . . . . . . . . . . . . .
65
4.1: The plane wave aperture averaging factor A vs. the ratio of the aperture
radius D/2 to the transverse coherence length ρ
0
, for L = 863 m and λ =
632.8 nm. The legend indicates different values of irradiance variance
2
I
σ , where turbulence strength increases in the asymptotic limit of
1
2
→
I
σ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
4.2: Aperture averaging factor using the Andrews asymptotic model vs. the
ratio of aperture radius to the transverse coherence length.. . . . . . . . .
73
4.3: Irradiance variance, or scintillation index, of a plane wave versus Rytov
variance. Asymptotic models in the saturation region are shown. . . . .
75
4.4: Transverse coherence length based on strong turbulence models plotted
against irradiance variance,
2
I
σ . . . . . . . . . . . . . . . . . . . . . . .
76
4.5: Churnside approximate aperture averaging factor plotted against the ratio
of the aperture radius to the transverse coherence length, for
2
I
σ = 1.1,
1.25, and 1.5. Curves calculated for an AVW test range path length of
863 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
78
4.6: Andrews asymptotic model for the spherical wave plotted against the
ratio of the aperture radius to the transverse coherence length. . . . . . .
79
4.7: Spherical wave irradiance variance plotted as a function of plane wave
Rytov variance. Asymptotic and weak turbulence relations are indicated.
81
4.8: Behavior of the transverse coherence length in the strong fluctuation
region, plotted as a function of the plane wave Rytov variance. . . . . . .
82
4.9: Average aperture averaged data in strong turbulence using Churnside
asymptotic analysis plotted against the ratio of the aperture radius to the
transverse coherence length. Churnside asymptotic analysis predicts C
n
2
= 9.3×10
12
m
2/3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
84
xi
4.10: New data using the Andrews asymptotic theory plotted against the ratio
of the aperture radius to the transverse coherence length. The solid line
represents the aperture averaging factor given by the Andrews
asymptotic theory for
2
I
σ = 1.1. . . . . . . . . . . . . . . . . . . . . . .
86
4.11: New aperture averaging data calculated using the SI model plotted
against the ratio of aperture radius D/2 to the transverse coherence length
ρ
0
for a plane wave propagating in strong turbulence. The curves
represent the aperture averaging factor given saturated and nonsaturated
strong turbulence values for
2
R
σ . . . . . . . . . . . . . . . . . . . . . . .
87
4.12: New strong turbulence data analyzed using the Churnside asymptotic
method for a spherical wave, with irradiance variances near 1.1. . . . . .
89
4.13: New strong turbulence data plotted using the Andrews asymptotic model
for the spherical wave from Eq. (4.19). The curve represents the
Andrews asymptotic model aperture averaging factor for 1 . 1
2
=
I
σ . . . . .
90
4.14: Mean data with error bars plotted against the ratio of the aperture radius
to the transverse coherence length. The solid line is the aperture
averaging approximation in Eq. (4.17) for 1 . 1
2
=
I
σ . . . . . . . . . . . .
91
4.15: Mean data analyzed with the Churnside asymptotic model, plotted along
with the strong turbulence Churnside data taken over a 1000 m path. . . .
93
4.16: Mean data analyzed using the scintillation index model in the non
saturated strong turbulence region, plotted along with the Churnside data
for a 1000 m path. Saturated and nonsaturated SI model curves for
1 . 1
2
=
I
σ are also shown. . . . . . . . . . . . . . . . . . . . . . . . . . .
94
5.1: Beam wander at the receiver plane is characterized by the short term and
long term beam widths [adapted from Ref. 20]. . . . . . . . . . . . . . .
96
5.2: The BER of a communications link with a gammagamma probability
distribution function (PDF) as a function of signaltonoise ratio. The
Rytov variances plotted are 0.3, 0.6, and 2.0. When 1
4
2
=
L
kD
, there
is no aperture averaging present because the aperture radius is equal to
the Fresnel zone size (FZ). Three more curves are plotted for an aperture
averaged case using a 4 inch diameter aperture. The irradiance variances
used in the calculation are the same in both cases. . . . . . . . . . . . . .
100
5.3: The cone concentrator with cone angle γ and maximum entrance angle θ
i
.
(adapted from Ref. 48). The cone is hollow with reflective (metallic)
edges. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
101
xii
5.4: The design of the CPC surface. The axis of the parabola and the CPC
axis are different. The surface of the CPC is traced out by moving the
vector r through the parametric angle φ. . . . . . . . . . . . . . . . . . .
103
5.5: Plot of the maximum entrance angle in air, θ
e
, and the maximum internal
ray angle, θ
n
for a range of refractive indices. . . . . . . . . . . . . . . .
104
5.6: Setup of the CPC link experiment. Black lines are electrical connections;
orange lines are fiber optic connections. The IR camera is used for
alignment purposes only, and is lowered below the field of view of the
16 in. Meade telescope when not in use. . . . . . . . . . . . . . . . . . .
107
5.7: Profile of the mounted ZnSe CPC directly coupled to an InGaAs PD. . . 108
5.8: BER plotted as a function of C
n
2
for the 1.7 km link with and without the
CPC integrated into the optical receiver. The error bars in C
n
2
represent
its fluctuation over the 1 min averaging interval. . . . . . . . . . . . . .
112
5.9: Same data as in Fig. 5.8, plotted along with a first order linear fit to the
mean data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
113
5.10: Trace 1 shows the data stream after passing through two 20 dB RF
amplifiers. Trace 2 shows the inverted data after passing through a
limiting amplifier. The signal after the limiting amplifier is much cleaner;
however there is still enough turbulenceinduced noise to degrade the
BER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
115
5.11: The NRLCBD laser communication test range [56]. . . . . . . . . . . . 116
5.12: Optical transmitter at NRLCBD. The collimated optical output is
approximately 2 W [56]. . . . . . . . . . . . . . . . . . . . . . . . . . .
116
5.13: The optical receiver at NRLCBD [56]. . . . . . . . . . . . . . . . . . . 117
5.14: Measurements of BER using the CPCPD receiver when an OC1 PRBS
transmission pattern was tested during a 35 minutes test period. The
BER of 1 bit in a 51.84 Mbps transmission is 3.215×10
10
, as indicated
by the dark horizontal line. There were no errors between minutes 1 and
19. The only errors were detected in the 31
st
min. . . . . . . . . . . . .
119
5.15: BER at OC2 over a 12 minute acquisition period using the CPCPD
receiver. A single bit error in a 103.68 Mbps transmission at OC2 gives
a BER of 1.607×10
10
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
120
xiii
5.16: BER at 100Mbps measured over the NRLCBD test range with an optical
fiber coupled receiver, from Ref. 46. Note that data is plotted for 5 sec
intervals. The minimum BER for a 5 sec interval is 2×10
9
, which is
beyond the range of the graph. . . . . . . . . . . . . . . . . . . . . . . .
121
5.17: BER at OC3 measured over a 15 minute period using the CPCPD
receiver. The single error BER for a 1 min acquisition period is
1.072×10
10
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123
5.18: BER at 200Mbps measured over the NRLCBD test range with an optical
fiber coupled receiver, from Ref. 46. Note that data is plotted for 5 sec
intervals. The minimum BER for a 5 sec interval is 1×10
9
, which is
beyond the range of the graph. . . . . . . . . . . . . . . . . . . . . . . .
123
1
Chapter 1
Introduction
1.1 History of Free Space Optical Communication
Since the invention of the Ruby laser in 1960, scientists have attempted to
establish reliable communications channels by modulating and propagating optical
signals over a lineofsight path [1]. Most of the early development of free space
optical communication links was directed towards spacebased applications [2,3,4].
Due to the relative immaturity of laser technology and flawed system demonstrations,
research in free space optical communication began a cycle of declining interest and
funding, only to be resurrected over and over again.
During the mid to late1990s, free space laser communication was once again
resurrected. This time, an increasing demand for high bandwidth communications,
fueled by the explosion of the internet, renewed interest in free space optics (FSO). By
this time, optical fiber communications was well established, and researchers had found
ways to use these efficient, costeffective components in FSO systems.
FSO was voted one of the ten hottest technologies in 2001 [5]. The hype
surrounding the technology was partly due to the lack of optical fiber backbone
connectivity in metropolitan areas and the high cost of laying optical fiber in those
areas. After September 11, 2001, FSO systems were deployed to reconnect parts of the
critical New York City infrastructure; after copper wires, optical fiber, and RF antennas
were destroyed. The advantages of FSO were trumpeted throughout the
2
telecommunications sector. These advantages include: rapidly deployablility; support
for high bandwidth transmissions; high security; the ability to extend the reach of the
optical fiber backbone; and use in disaster recovery situations. Unfortunately, the
technology bubble in the stock market burst. This left FSO vendors without
telecommunication and cable carriers who could purchase and deploy FSO units on a
large scale.
Even with the significant capital investment in FSO during this time, researchers
did not solve the technology’s reliability problems due to atmospheric turbulence,
aerosols, and molecular attenuation. These three environmental properties destroy the
coherence of a propagating optical wavefront. The research presented here
demonstrates techniques to mitigate the distortions induced on the optical beam by
atmospheric turbulence.
1.2 Free Space Optical Communication Techniques
Researchers have pursued a variety of system architectures in an attempt to
improve the reliability of FSO links. A “bare bones” FSO system transmits “1” bits by
turning the optical source on, and “0” bits by turning the optical source off. This
method is called intensity modulation/direct detection (IM/DD), or onoff keying
(OOK). This is a low overhead technique, because there is no attempt made to ensure
the integrity of the signal. This transmission scheme attempts to collect enough photons
at the receiver to correctly detect “1” and “0” bits. Atmospheric turbulence may easily
prevent the detection of a sufficient number of photons to correctly interpret the
received bits.
3
Beyond the basic IM/DD scheme, researchers have tried to come up with
techniques to improve the likelihood that data bits will be received correctly. Scientists
have used adaptive optics techniques, originally developed to improve the quality of
stellar telescopes, to try to restore a distorted wavefront to its original state before it was
destroyed by atmospheric turbulence. Some used phase conjugation techniques, to pre
distort the beam and hope that atmospheric turbulence, by distorting the wavefront, will
actually work to render the wavefront to its unperturbed form. Although these
techniques have shown limited success, they require bulky and computationintensive
systems to achieve wavefront correction. Additionally, systems have not been able to
compute the wavefront correction fast enough to entirely counteract the affects of
atmospheric turbulence.
Diversity techniques are another method of improving the reliability of an FSO
system. In effect, if the optical wavefront propagates in at least two distinct ways, there
is an increased likelihood that the detected signal will be read correctly. Diversity can
occur in the form of spatial diversity (requiring multiple transmitters and/or receivers),
temporal diversity (requiring a signal to be transmitted twice, separated by a time
delay), or wavelength diversity (requiring the transmission of data on at least two
distinct wavelengths). Each of these techniques requires a synchronization of the
received signals. Although the techniques are promising, they do require a significant
electronic overhead in the retiming and synchronization process.
Finally, coding schemes used in RF and wired communications systems have
been adapted for FSO communication. Methods requiring heterodyne detection,
including phase shift keying and quadrature amplitude modulation, have been
4
experimentally demonstrated but do not guarantee enough improvement to be viable
commercially [6,7]. Forward errorcorrection (FEC) codes insert check bits into the
data stream, which contribute an additional power and bandwidth overhead on the
system. Although coding provides an additional layer of information security, studies
have shown that even the best FEC codes cannot negate the affects of atmospheric
turbulence alone [6,8,9].
1.3 LowOverhead Techniques to Mitigate the Effects of Atmospheric Turbulence
The research presented here demonstrates how highperformance FSO
communication systems may be designed using techniques that add nearly no power,
bandwidth, size, or weight overhead to the overall design.
1.3.1 Aperture Averaging
Aperture averaging is a well known concept, which states that the amount of
measured radiation may be increased by increasing the size of the receiver collecting
lens aperture. In effect, any irradiance fluctuations across the collecting lens are
“averaged” by the size of the lens. Therefore, a lens of any diameter will collect more
photons than an ideal point receiver. In a binary communications transmission, such as
OOK, aperture averaging will increase the likelihood of correctly detecting the
transmitted data stream.
Aperture averaging of irradiance fluctuations was first studied by Fried in 1967
[10]. Fried developed a theoretical expression characterizing the amount of aperture
averaging given an incident infinite plane wave. Aperture averaging was
experimentally investigated for both horizontal and spacetoground paths under weak
5
levels of atmospheric turbulence. However, Fried came to an erroneous conclusion that
the amount of aperture averaging is proportional to the inverse square of the aperture
diameter. Later in 1973, Fried came to the correct conclusion that there is a 3 / 7 −
power dependence.
Subsequent experiments on aperture averaging in weak atmospheric turbulence
found very poor agreement with theory. The concept that the amount of irradiance
fluctuations saturates after a certain propagation distance was not understood, so
scientists could not truly differentiate between weak and strong fluctuations. Finally, in
1991, Churnside published experimental results that have shown the best agreement
with theory in weak turbulence conditions until now [11]. The behavior of optical
waves propagating in strong turbulence is not well understood, although new theories
and models have been developed recently. As such, there have been no consistent
experimental results for aperture averaging in strong turbulence.
1.3.2 Nonimaging Optics
On a basic level, imaging optics form an image based on the transformation of
an object by an imaging element. A lens or compound lens is a typical imaging
element. If a converging lens is used, then the object and image are located on opposite
sides of the lens, when the object is placed at or beyond the focal length of the lens. A
diverging lens will form a virtual image on the same side of the lens as the object.
Ideally, the image or virtual image is a perfect replica of the object, although its size
may be magnified or reduced.
Imaging optics are useful when the quality of an optical wavefront must be
preserved. In the case of FSO with IM/DD, we are only concerned with collecting a
6
maximum number of photons. Any information about the phase of the wavefront is
ignored. Atmospheric turbulence will destroy the coherence of a propagating optical
wave, and induce phenomena that will affect the collection of a maximum number of
photons.
Nonimaging optics are not required to preserve the integrity of the object.
Instead, what would be the image may be partially or completely randomized. Only the
conservation of energy is required. A special type of nonimaging optical element is
called a concentrator. These elements collect radiation over a restricted angular range,
and dispense radiation over a wider angular range. The concentrator operates on the
principle of conservation of brightness, preserving the relationship between the
irradiance and its angular distribution.
In this research, a nonimaging concentrator is used to mitigate the affects of
atmospheric turbulence on the focal plane of an imaging optical receiver. The
nonimaging concentrator will not preserve the coherence of the wavefront, but will
improve the detection efficiency of the receiver.
1.4 Useful Terminology
Radiometric terminology has been loosely used in scientific research. Here, we
specifically define radiometric terms used throughout this dissertation.
Solid angle – Ω, the projection of an area onto a unit sphere; measured in steradians (sr).
A sphere has a 4π solid angle.
Projected area – the area projected onto a plane whose normal is the line of sight; the
area is multiplied by the cosine of the angle between the normal to the area and the line
of sight.
7
Optical power – P, also known as radiant flux; the flow of radiant energy per unit time;
expressed in W/m
2
.
Radiance – L, the optical power (or radiant flux) per unit projected area per unit solid
angle; expressed in W/m
2
sr.
Irradiance –
Ω
=
d
dP
I , also known in this work as intensity; the flow of energy per
unit area per unit time; expressed in W/m
2
. The irradiance is proportional to the square
of the amplitude of the electric field of an optical wave.
Spectral radiance –
λ
λ
d
dL
L = , the distribution of radiance per unit wavelength;
expressed in W/m
2
srHz. The spectral radiance is meaningful by integrating it over a
wavelength range.
Frequency – ν , the number of cycles per second; measured in Hz or sec
1
.
Optical wavenumber –
λ
π 2
= k , where λ is the wavelength of the optical wave.
Spatial wavenumber – κ, the inverse of a spatial scale length.
1.5 Organization
This dissertation is organized in four chapters detailing the theory, methodology,
and analysis used to study techniques to mitigate the effects of atmospheric turbulence.
Chapter 2 reviews both the physical nature of atmospheric turbulence and theories
developed to characterize atmospheric turbulence. A framework for studying optical
wave propagation through atmosphere turbulence, in strong and weak turbulence
conditions, is presented. Chapter 3 involves an in depth study of aperture averaging in
weak turbulence conditions. Aperture averaging models, experimental methodology,
8
and analysis of new experimental data are explained. Chapter 4 follows the
organization of Chapter 3, but instead addresses aperture averaging in strong turbulence
conditions. New experimental data in the strong turbulence region are analyzed and
compared with available models. Chapter 5 studies and characterizes the performance
enhancement of free space optical communication systems when nonimaging optics are
integrated into the optical receiver. Chapter 6 summarizes the contributions of this
work and addresses areas for future research.
9
Chapter 2
Optical Wave Propagation through Atmospheric Turbulence
2.1 The Atmosphere
Propagation of optical waves through the atmosphere is affected by atmospheric
turbulence, scattering off aerosols, and atmospheric absorption. This thesis addresses
the predominant cause of distortion of optical waves in the atmosphere, which is
atmospheric turbulence. Atmospheric turbulence is a result of localized variations of
temperature, humidity, and pressure in the atmosphere. These variations result in
localized refractive index fluctuations, where each localized area of lower or higher
refractive index is known as a turbulent eddy [1]. The refractive index of each
individual eddy is not much greater than unity, but the cumulative effect of eddies over
a 1 km path is great.
The refractive index of air at optical frequencies is,
6
2
3
10
10 52 . 7
1 6 . 77 1
−
−
⋅

.

\



.

\
 ⋅
+ = −
T
p
n
λ
(2.1)
where n is the total refractive index, λ is the wavelength in µm, p is the pressure in mb,
and T is the temperature in K. At sea level, n − 1 is typically 3×10
4
[1]. Humidity
effects are typically neglected over land, since humidity affects the value of the
refractive index by less than 1%.
10
2.2 Turbulent Energy Flow
Turbulent flows result when large inertial forces draw together fluid volumes
with very different velocities, and irregular velocity fluctuations are apparent.
Turbulent flows exist when the Reynolds number is greater than 2500 to 5000 [12],
where the Reynolds number is defined by:
υ / UL Re = (2.2)
where U is the characteristic flow velocity, L is the characteristic dimension of viscous
flow, and υ is the kinematic viscosity of the fluid. Since Reynolds numbers in the
atmosphere are large, the associated fluid flows are also highly unstable. Eddies of
scale sizes on the order of flow dimensions move randomly, and eventually give rise to
eddies of smaller scale sizes and lower velocities. Eventually, eddies become small
enough that viscosity forces overcome inertial forces, and the eddies can no longer
decay. The breakdown in eddies is dictated by [12],
( ) ( ) ( ) ( ) j j j j
L U Re υ = (2.3)
where j is the order of the eddy. Higher orders denote smaller eddies.
Richardson first developed a picture of the turbulent energy redistribution in the
atmosphere. The process is shown pictorially in Figure 2.1, with an energy input
region, inertial subrange, and energy dissipation region. The dissipation rate ε is related
to the velocity U
ℓ
of an eddy with characteristic length ℓ by [12],
( )
3 / 1
l
l
ε ≈ U (2.4)
At large characteristic lengths ℓ, a portion of kinetic energy in the atmosphere is
converted into turbulent energy [13]. When the characteristic length reaches a specified
11
outer scale length, L
o
, energy begins to cascade. The energy of one eddy is
progressively redistributed into eddies of smaller scales, until eddies reach a size equal
to the inner scale length, ℓ
o
. The inner scale length, or Kolmogorov microscale, is
defined by,
( )
4 / 1
3
4 . 7 ε υ =
o
l (2.5)
At the surface layer of the Earth, ℓ
o
is typically on the order of 4 mm [12], and υ is
typically 0.148 cm
2
s
1
[13]. Kolmogorov proposed that in the inertial subrange, where
L
o
> ℓ > ℓ
o
, turbulence is isotropic and may be transferred from eddy to eddy without
loss. When the diameter of a decaying eddy reaches ℓ
o
, the energy of the eddy is
dissipated as heat energy through viscosity processes [13].
Figure 2.1: Depiction of the process of turbulent decay, showing the energy cascade and
subsequent division of turbulent eddies in the atmosphere [adapted from Ref. 13].
Energy Input
Energy Dissipation
κ
in
= 1/L
o
κ
d
= 1/ℓ
o
Spatial Wavenumber κ = 1/l
T
u
r
b
u
l
e
n
c
e
S
p
e
c
t
r
u
m
Φ
(
κ
)
L
o
ℓ
o
Inertial Subrange
12
2.3 Kolmogorov Turbulence
Turbulence is by nature a random process, and as such may be described using
statistical quantities. In 1941, Kolmogorov first developed a universal description of
atmospheric turbulence by developing a structure tensor to describe a mean square
velocity difference between two points in the atmosphere. The structure function,
( ) r D
ij
r
, is [1]:
( ) ( ) ( )   ( ) ( )  
1 1 1 1
r v r r v r v r r v r D
j j i i ij
r r r r r r r
− + ⋅ − + = (2.6)
where r
r
is the displacement between two points in space, r r
r r
+
1
and
1
r
v
[1]; and i and j
are two velocity components. To continue the structure function analysis, two
approximations are made given the condition that the displacement is within the inertial
subrange. The first is to assume local homogeneity in the atmosphere, which restricts
the dependence of the velocity statistics in Eq. (2.6) to the vector displacement, r
r
. A
second assumption of local isotropy limits Eq. (2.6) to a dependence on the magnitude
of r
r
, given as r
r
. These assumptions allow the velocity structure function to be
treated as a scalar function [1]:
( ) ( ) ( ) ( )
ij j i ij
r D n n r D r D r D δ
r r r r
⊥ ⊥
− − = ] [

(2.7)
where ( ) r D
r

and ( ) r D
r
⊥
are the structure function components of the wind velocity
field parallel and transverse tor
r
, { } j i j i
ij
≠ ⇒ = ⇒ ∈ 0 ; 1 δ , and n
i
, n
j
are the
components of a unit vector along r
r
[1]. To continue this analysis, the flow is assumed
to be incompressible, giving 0 = ⋅ ∇ v
r
. This assumption is valid when 1
2 2
<< a ν
r
,
13
where a is the velocity of sound [12]. Using the assumption of an incompressible flow,
we can relate the parallel and transverse components of the velocity structure function:
( )

2
2
1
D r D
dr
d
r
=
⊥
. (2.8)
With this relation, we can describe the structure function in terms of one component:
( ) ( )  
2
1 1  
r v r r v D − + = (2.9)
As mentioned previously, Kolmogorov determined that as long as r is within the inertial
subrange, we can assume localized fluctuations and the structure function becomes [1]:
3 / 2 2

r C D
v
= (2.10)
where C
v
2
is the velocity structure constant with units m
2/3
. Eq. (2.10) is the
KolmogorovObukhov “twothirds” power law [12]. To use this relation in
electromagnetic wave propagation problems, the velocity fluctuation form of Eq. (2.10)
must be transformed into a refractive index fluctuation form. This is accomplished by
using the potential temperature, θ, which is the temperature of a parcel of air that is
brought adiabatically (where heat is neither gained nor lost) from a state having
pressure p and temperature T to a state where pressure p
0
= 1000 mbar [14]. For small
values of r, as when r is in the inertial subrange, the potential temperature can be
approximated by [15]:
h T
p
c
g
+ = θ (2.11)
where T is the absolute temperature in Celsius, g is the acceleration of gravity, c
p
is the
specific heat,
p
c g = 9.8 °/km, and h is the height of the parcel above the Earth’s
surface [1,16]. The value
p
c g is also the adiabatic rate of decrease of the absolute
14
temperature. At small heights above the Earth’s surface, the adiabatic rate
p
c g can be
neglected and T is treated as a passive additive [1,16,17]. Passive additives to do not
affect turbulence statistics, and therefore do not alter the twothirds law in Eq. (2.10).
The structure function for the potential temperature is well known [1], and is described
by:
( )
3 / 2 2
r C r D
θ θ
= (2.12)
By taking the derivative of both sides of Eq. (2.1) with respect to n, we find:


.

\
 ∂
−
∂


.

\
 ⋅
+ = ∂
−
T
T
p
p
T
p
n
2
3
10 52 . 7
1 6 . 77
λ
(2.13)
Knowing that pressure fluctuations are relatively small compared to temperature
fluctuations, we can approximate 0 → ∂p , giving:
6
2 2
3
10
10 52 . 7
1 6 . 77
−
−
⋅ ∂

.

\



.

\
 ⋅
+ = ∂ T
T
p
n
λ
(2.14)
Using the relation of potential temperature to absolute temperature from Eq. (2.11), and
the relation of their derivatives θ ∂ = ∂T because T is a passive additive, we find that for
λ = 0.6328µm:
6
2
10 06 . 79
−
⋅ ∂

.

\

= ∂ θ
T
p
n (2.15)
The direct relationship between refractive index fluctuations and potential temperature
fluctuations is shown in Eq. (2.16). We can now infer a relationship between the
structure parameter for refractive index and that of the potential temperature as:
6 2
2
2
10 06 . 79
−
⋅

.

\

=
θ
C
T
p
C
n
(2.16)
15
Optical refractive index fluctuations near the Earth’s surface are a result of temperature
variations [1].
2
n
C is formally known as the refractive index structure parameter, and
varies spatially and temporally. We use this indicator to quantify the strength of
turbulence along the path of the optical propagation. The value of
2
n
C typically varies
from 10
17
m
2/3
under “weak turbulence” conditions to 10
13
m
2/3
in “strong turbulence”
conditions. Using previous information, the refractive index structure function obeys
the same KolmogorovObukhov “twothirds” power law:
( )
3 / 2
2
r C r D
n n
r r
= (2.17)
2.4 Wavenumber Spectrum Models
By using a methodology similar to that used for the characterization of the
strength of various frequency components in a timevarying electrical signal, we can
define a Fourier transform to quantify the ability of different eddy sizes to influence the
refractive index of a random medium. To develop a Fourier spectrum, we must first
determine how refractive index fluctuations vary between different points in the
atmosphere. This relationship is given by a spatial covariance [1,13]:
( ) ( ) ( )
1 1 1 1
, r n r r n r r r B
n
r r r r r r
δ δ + = + (2.18)
where n δ is the fluctuating part of the refractive index, with n n n δ + = . Following the
previous discussion of Kolmogorov turbulence (Sec. 2.3), the assumption of
homogeneity implies that only the separation between two points and not the location of
the two points in a random medium affects the physics, so that ( ) ( ) r B r r r B
n n
r r r r
→ +
1 1
, .
Near the Earth’s surface, the random medium may be treated as isotropic, allowing
16
( ) ( ) r B r B
n n
→
r
. In spherical coordinates, the threedimensional Fourier transform for
the spatial covariance is:
( )   ( )
( )   ( )
( )
( )
r
r
d
r i d d d d
r i d r B
n
n
n n
κ
κ
κ κκ π
κ κ κ φ θ θ κκ
κ κ κ
π π
sin
4
exp sin
exp
0
2
0
2
0
2
3
Φ =
Φ ⋅ =
Φ ⋅ =
∫
∫ ∫ ∫
∫
∞
∞
∞ −
r r r
r r r r
(2.19)
where ( ) κ
n
Φ is the isotropic and homogeneous turbulence spectrum. A Fourier
integral also relates the structure function to the isotropic and homogeneous turbulence
spectrum:
( ) ( )
( )
∫
∞

.

\

− Φ =
0
2
sin
1 8
r
r
d r D
n n
κ
κ
κ κκ π (2.20)
which has been derived from the generic Fourier transform description of the spatial
covariance and the definition of the refractive index structure function,
( ) ( ) ( )  
2
1 1 1 1
, r n r r n r r r D
n
δ δ − + = + . The two descriptions of the refractive index
structure function from Eqs. (2.17) and (2.20) may be compared to determine the
wavenumber power spectrum:
( )
( )
∫
∞

.

\

− Φ =
0
2 3 / 2 2
sin
1 8
r
r
d r C
n n
κ
κ
κ κκ π (2.21)
Using power law relations, the wavenumber spectrum is [1,3]:
( )
3 / 11 2
033 . 0
−
= Φ κ κ
n n
C (2.22)
Eq. (2.22) is the Kolmogorov spectrum. It is only valid in the range
o o
l L
1 1
<< << κ .
( ) κ
n
Φ is an equivalent representation of ( ) r D
n
in inverse space.
17
There are other spectrum models that attempt to describe behavior beyond the
inertial subrange. The Tatarskii spectrum uses a Gaussian function to extend coverage
to the dissipation range (
o
l
1
> κ ), where small eddies are influential:
( )

.

\

− = Φ
−
m
n n
C
κ
κ
κ κ exp 033 . 0
3 / 11 2
(2.23)
where
o
m
l
92 . 5
= κ . The von Karman model is the most widelyused model to describe
characteristics in the energyinput region (
o
L
1
< κ ). Although it does an excellent job
of describing large eddy formation in the troposphere, the model was originally
proposed to describe fluid flow in a circular pipe [18]. The von Karman spectrum
integrates the small eddy dependence in the Tatarskii spectrum with a large eddy
description to give:
( )
( )
( )
6 / 11
2 2
2 2
2
/ exp
033 . 0
in
m
n n
C
κ κ
κ κ
κ
+
−
= Φ for
m
κ κ < < 0 (2.24)
where
o m
l 92 . 5 = κ and
o in
L 1 = κ . In the inertial subrange, the von Karman spectrum
reduces to the Kolmogorov spectrum value.
The Kolmogorov and von Karman spectrums fail to show a “bump” at high
wavenumber values in the inertial subrange, near
m
κ . Hill constructed an exact
spectrum that accounts for the high wavenumber rise [19]. Andrews developed an
analytical approximation to the Hill spectrum, since the Hill spectrum uses a second
order differential equation that has to be solved numerically. Andrews calls his
approximation the Modified Atmospheric Spectrum, which has the representation [20]:
18
( )
( )
( )
2 2
2 2
6 / 7
2
exp
254 . 0 802 . 1 1 033 . 0
in
l
l l
n n
C
κ κ
κ κ
κ
κ
κ
κ
κ
+
−
×


.

\

−


.

\

+ = Φ
(2.25)
The Kolmogorov, von Karman, and modified atmospheric spectrums are plotted
in Fig. 2.2 for scale sizes of L
o
= 10 m and l
o
= 5 mm. The high wavenumber “bump”
occurs in the transition between the inertial subrange and dissipation range. The bump
occurs in the plot of the modified atmospheric spectrum at the boundary between the
energy dissipation range and the inertial subrange in Fig. 2.2, and is especially apparent
in the scaled spectral plot in Fig. 2.3.
Fig. 2.2: Three wavenumber spectrum models for refractive index fluctuations. The
three energy ranges relevant to turbulence statistics are indicated.
19
Fig. 2.3: Scaled spectral models for the von Karman spectrum, Eq. (2.24), and modified
atmospheric spectrum, Eq. (2.25), plotted against the wavenumber scaled by the inner
scale.
2.5 Rytov Approximation
The Rytov approximation is a method to solve Maxwell’s equations for
electromagnetic wave propagation while accounting for diffraction effects [1,21,22]. A
derivation of optical wave propagation is begun from Maxwell’s equations, assuming
no free charges, µ
0
permeability, and an electromagnetic field with a harmonic time
dependence of e
jωt
[1,22]:
( )
( ) ( )  
( ) ( ) ( )
( ) ( ) t r H jk t r E
t r E t r n jk t r H
t r E t r n
t r H
, ,
, , ,
0 , ,
0 ,
0
2
0
2
r
r
r
r
r
r
r r
r
r
r
r
r
r
= × ∇
− = × ∇
= ⋅ ∇
= ⋅ ∇
(2.26)
20
where c k ω =
0
is the free space wavenumber, c is the speed of light in vacuum, n =
n(r) is the refractive index of the atmosphere, and 1 − = j . It is also assumed that the
variation of the field with time is much slower than e
jωt
, which justifies a quasisteady
state approach. After some manipulation with vector identities and substitution [1,22],
a wave equation is found:
( ) ( ) 0 ln 2
2 2 2
= ∇ ⋅ ∇ − + ∇ n E E n k E
r r r
(2.27)
where the third term on the left hand side is a depolarization term. Since it has been
shown that the change in polarization of a propagating wave is negligible for both cases
of λ << l
o
and λ > l
o
[1], this term may be dismissed. Eq. (2.28) becomes:
0
2 2 2
= + ∇ E n k E
r r
(2.28)
Eq. (2.28) may be transformed into a scalar equation for each of the three electric field
components. Given a mean air refractive index of 1, the total refractive index is:
( ) r n r n
r r
δ + =1 ) ( (2.29)
where δn << 1. The Rytov method considers a solution of the form:
( ) ( ) ( ) r r E
r r
Ψ = exp (2.30)
Substitution of Eq. (2.30) into Eq. (2.29) yields the nonlinear Ricatti equation [22,23]:
( ) ( ) ( ) ( ) r n k r r r
r r r r
2 2
0
2
δ − = Ψ ∇ ⋅ Ψ ∇ + Ψ ∇ (2.31)
The Ricatti equation may be solved by a multiplicative perturbation method with a
solution [22,23,45]:
( ) ( ) ( ) ( ) ( ) ...
3 2 1 0
+ Ψ + Ψ + Ψ + Ψ = Ψ r r r r r
r r r r r
(2.32)
21
The field equation for the basic Rytov solution is a result of keeping the first two terms
in the expansion [22,23,45]:
( ) ( ) ( ) ( ) r n k r r r
r r r r
δ
2
0 1 0 1
2
2 2 − = Ψ ∇ ⋅ Ψ ∇ + Ψ ∇ (2.33)
Amplitude and phase information for the propagating wave falls out of the Rytov
approximation. The first iterative solution to Eq. (2.33) may be written as the perturbed
field ( ) ( ) ( ) r r U r U
r r r
1 0
exp ) ( Ψ = , where ( ) r U
r
0
is the unperturbed field [45]. Following
Refs. [20] and [45], the complex phase perturbation is written as ( )
1 1
iS r + = Ψ χ
r
) ( ) / ln(
0 0
S S i A A − + = where χ is the logarithm of the amplitude A, and S is the phase
of the field ( ) r U
r
. The unperturbed amplitude and unperturbed phase are denoted by A
0
and S
0
, respectively. This analysis shows that under the Rytov approximation,
irradiance fluctuations obey a lognormal distribution, since   2 / ) / ln(
2
A I = χ [20,45].
The Rytov approximation yields a log amplitude variance
2
χ for infinite plane wave
propagation [22,45]:
6 / 7 6 / 11 2 2 2
307 . 0 k L C
n
= =
χ
σ χ (2.34)
It has been shown that Eq. (2.34) is a good approximation to the log amplitude variance
in the range
2
χ
σ < 1. The log intensity is related to the log amplitude variance by [1,45]:
2
2
2
log
4 log log
χ
σ σ = − = I I
I
(2.35)
and
2 6 / 7 6 / 11 2 2
log
23 . 1
R n I
k L C σ σ = = (2.36)
22
where
2
R
σ is the Rytov variance. The Rytov variance of an infinite plane wave helps to
define the strength of irradiance fluctuations, although it is not an absolute measure of
turbulence strength, as shown in Table 2.1.
Strength of Fluctuations Rytov variance
Weak
3 . 0
2
<
R
σ
Intermediate
1 ~
2
R
σ
Strong
1
2
>>
R
σ
Table 2.1: Typical Rytov variance ranges corresponding to weak, intermediate, and
strong turbulence levels.
A better comparison of the strength of irradiance fluctuations is provided by the
transverse coherence length for a optical wave [13,24]. The coherence length for a
plane wave is:
( )
5 / 3
2 2
0
46 . 1
−
=
n
LC k ρ (2.37)
The coherence length for a spherical wave is:
( )
5 / 3
2 2
0
546 . 0
−
=
n
C Lk ρ (2.38)
Fried defined a coherence radius
0 0
099 . 2 ρ = r where ρ
0
is defined in Eq. (2.38) [25,1].
A better way to determine the region of turbulence experienced by a propagating
optical wavefront is to plot the three relevant scale sizes: the transverse coherence
length, ρ
0
, the Fresnel zone size, z λ , and the scattering disk size,
0
/ ρ k L . These are
plotted in Fig. 2.3 for the plane wave case with C
n
2
= 5×10
14
m
2/3
, Fig. 2.4 for the
spherical wave case with C
n
2
= 5×10
15
m
2/3
, and Fig. 2.5 for the spherical wave case
with C
n
2
= 5×10
13
m
2/3
. The intersection of the three scale sizes denotes the onset of
strong scintillation [19].
23
Eq. (2.37) is bounded by inner scale and outer scale scattering parameters, since
optical propagation in the visible and nearinfrared wavelengths near the Earth’s surface
is dominated by Fresnel scattering [21]. The large eddies in the spectrum set the outer
scale scattering parameter:
2
0 0
2 L Lλ π ζ = (2.39)
while the small eddies define the inner scale scattering parameter:
2
0
2
2
56 . 5
l
λ κ
ζ
L
k
L
m
m
= = (2.40)
For a test range length of L = 863 m and a wavelength λ = 0.6328 µm, the Fresnel
length is 2.47 cm, which is well away from the scale of either scattering parameter.
Fig. 2.4: The Fresnel zone size, coherence length, and scattering disk size plotted for a
plane wave against propagation distance. For L = 863 m, there are scale sizes on the
link contributing to strong fluctuations. The shaded area shows scale sizes that do not
contribute to strong fluctuations.
24
Fig. 2.5: Three scale sizes plotted against the propagation distance for the spherical
wave case with L = 863 m. When C
n
2
= 5×10
15
m
2/3
, the wavefront sees weak
scintillations with contributing scale sizes on the order of the Fresnel zone size.
Fig. 2.6: Three scale sizes plotted against the propagation distance for the spherical
wave case with L = 863 m. From the graph, the wavefront experiences strong
fluctuations when C
n
2
= 5×10
13
m
2/3
, with contributing scale sizes beyond the shaded
area.
25
2.6 Strong Turbulence Theory
Multiple scattering of an optical wave by refractive index inhomogeneities
results in strong fluctuations of the irradiance. As the amount of multiple scattering
increases with increasing propagation path length, irradiance fluctuations saturate and
approach a value of one. This phenomenon was first experimentally reported in a
Russian journal by Gracheva and Gurvich [26]. Although Rytov and other Markov
based solutions to the stochastic wave equation were investigated [27], no existing
analysis method was successful in describing the saturation of scintillation. New
research in perturbation theory led to the development of the asymptotic theory for
strong scintillations [27,28,29,30]. Other scientists used heuristic methods to predict
the saturation of amplitude fluctuations [31,32]. Although the heuristic methods show
good results, there is no welldefined relationship between amplitude and irradiance
fluctuations in strong turbulence. Heuristic theory will not be used in the analysis of
aperture averaging, since the analysis requires information of the value of the logarithm
of the amplitude.
2.6.1 Andrews – Prokhorov Asymptotic Analysis
The propagation of an optical wave is defined by different moments of the field.
The angular spread of the beam is characterized by the second order moment, which is
known as the mutual coherence function (MCF). The fourth order moment defines the
irradiance fluctuations of the propagating wave. When the propagating wavelength is
small compared to the scale size of inhomogeneities in the random medium, and the
variation of the refractive index is much less than one, the fourth order moment with E
r
in the zˆ propagation direction is [27,29]:
26
( ) ( ) ( ) ( ) ( )
2 1
* *
2 1 2 1 4
, 0 , , , , , r r z E z E r z E r z E r r z + = Γ (2.41)
The irradiance variance is related to the fourth order moment by [29]:
( ) ( ) 1 0 , 0 ,
4
2
2
− Γ = − = z I I
I
σ (2.42)
With the lack of an analytical solution to Eq. (2.42), asymptotic solutions for the fourth
order moment yield the best results for strong turbulence over long propagation paths.
Prokhorov summarized the development of asymptotic analysis using the local method
of small perturbations [27]. He constructed a set of equations for the averaged field to
which the moments of the field were solutions. The local method of small perturbations
required that the phase of the fluctuating wave experiences only small variations over a
propagating distance
0
z z − , and satisfies the limit:
1
2 2 2
<< ∆ l ε k (2.43)
where ( ) ε ε ε ε / 2 − = ∆ and ℓ is the scale of the inhomogeneities in the turbulent
medium.
2.6.1.1 Andrews Asymptotic Analysis for a Plane Wave
The equation for the fourthorder moment of the electric field of a plane wave
propagating in a statistically homogeneous random medium is [27]:
( )
2 1
4
2
4
4
r r
k
i
V
z
r r
∂ ∂
Γ ∂
− Γ − =
∂
Γ ∂
(2.44)
with
( ) ( )
( ) ( )
2 2
2 1 2 1
2 1
r r D r r D
r D r D V
r r r r
r r − ′
−
+ ′
− ′ + ′ = (2.45)
27
where Γ
4
is the fourth moment of the field and ( ) u D′ is the derivative of the phase
structure function. Prokhorov wrote the spectral function of irradiance fluctuations as
[27]:
( )   ( )
2 1 4 2
, , exp
2
1
, , r r z u q i u d r q z M
r r r r r r r
Γ − =
∫
π
(2.46)
which is the Fourier transform of the fourth order moment with q representing the scale
of frequency components in the spectrum. The full power spectrum in the strong
fluctuation regime is presented in Eq. (4.17) of Ref. [27]. In his analysis, Prokhorov
derived the scintillation index, m
2
, or irradiance variance, in strong, saturated turbulence
conditions [27]:
( ) ( ) { }
( )( )
( ) 1 / for
/ 1
2 / 2
3
2 2
>>
+ = =
− −
k z D
k z D N m
I
α α
α σ
(2.47)
where
( )
( )
( ) ( )
{ } ) 1 /( ; ; 1 , 1
4
1
1
1
4
2
1
2
sin
2
1 2
2
1
3
+ − − ×

.

\

−
− Γ + Γ

.

\

=
+
α α α α
α
α
α
α
πα
πα
α
α
F
N
(2.48)
where α is the powerlaw exponent of the phase structure function,
α
) ( ) ( ku C u D = , Γ
represents the gamma function, and
2
F
1
is the confluent hypergeometric function. The
phase structure function in the inertial subrange approximates a 5/3 power relation [13].
For α = 5/3, N
3
(α) = 1.22. A solution to the scintillation index can be found using the
phase structure function representation [27]:
( )
3 / 5 3 2
9 . 2 zr k C r D
n
= (2.49)
28
By substituting this into Eq. (2.47), we find the asymptotic relation between the
scintillation index and the Rytov variance for a plane wave in saturated strong
turbulence [20,27]:
( )
5 / 2
2
2
86 . 0
1
R
I
σ
σ + = (2.50)
For the duration of this thesis, we will refer to Eq. (2.50) as the Andrews asymptotic
model for the plane wave case.
The generalized asymptotic form for the covariance may also be written in terms
of the transverse coherence length [24,11]. The form of the covariance for a plane wave
is given by a twoscale model:
( ) ( ) ( ) ( )   ρ ρ
ρ
ρ
ρ
ρ
2 1
3 / 1
2
0
3
5
3
3 / 5
0
2
1
exp b b
L
k
N C
I
+


.

\

+


.

\

− = (2.51)
where b
1
and b
2
are scale functions that represent the influence of two different scale
sizes on the propagating wave. When ρ approaches zero, b
1
and b
2
go to unity; while as
ρ approaches infinity, b
1
and b
2
go to zero. When 0 → ρ , Eq. (2.51) becomes the
asymptotic form of the irradiance variance in Eq. (2.50), as expected.
2.6.1.2 Andrews Asymptotic Analysis for the Spherical Wave
In spherical coordinates, the fourthorder moment of the electric field for a
propagating spherical wave in a statistically homogeneous random medium is [27]:
¦
)
¦
`
¹
¦
¹
¦
´
¦
∂ ∂
Γ ∂
+
∂ ∂
Γ ∂

.

\

− Γ − =
∂
Γ ∂
φ φ θ θ
η ξ η ξ
4
2
4
2
2
4
4
kr
i
V
r
(2.52)
29
where ( ) φ θ, , r are spherical coordinates with ( )
φ θ
ξ ξ , and ( )
φ θ
η η , are constant
differences in angular coordinates due to the conversion from cartesian to spherical
coordinates. In the region of the interest of the fourth moment, the angles of interest,
( )
φ θ
ξ ξ , and ( )
φ θ
η η , , are small and near the equator of the spherical coordinate system
with θ ≈ π/2 [27]. Therefore, due to the small angles in the second derivative on the
right hand side of Eq. (2.52), there will be an additional constant factor in the term. In
saturated strong turbulence, the scintillation index is [27]:
( ) ( ) { }
( )( ) α α
α σ
− −
+ = =
2 / 2
2 2
/ 1 k r D N m
sp sp sp I
(2.53)
with the spherical wave structure function related to the plane wave structure function
in Eq. (2.49) by [27]:
( ) ( ) ( ) u D u D
sp
1
1
−
+ = α (2.54)
and
( ) ( ) ( ) ( )
( ) ( ) 2 2 1 1
4
2 / 1 2 / sin 1
2
1 2
2
1
− Γ − Γ

.

\

− Γ ×
+ Γ +
=
−
+
α α
α
α απ α
πα
α
α
sp
N
(2.55)
The spherical wave scintillation index is related to the plane wave phase structure
function by [27]:
( ) ( ) { }
( )( ) α α
α σ
− −
+ = =
2 / 2
, 3
2 2
/ 1 k r D N m
sp sp I
(2.56)
where we have defined N
3,sp
as:
( ) ( )
( )( )
( ) α α α
α α
sp sp
N N
− −
+ =
2 / 2
, 3
1 (2.57)
30
As for the plane wave, we assume the phase structure function obeys a 5/3 power law in
the inertial subrange, giving N
3,sp
(5/3) = 3.86. The asymptotic scintillation index for a
spherical wave in saturated strong turbulence conditions with α = 5/3 becomes [20,27]:
( )
5 / 2
2
2
73 . 2
1
R
I
σ
σ + = (2.58)
The asymptotic formula for the covariance of a spherical wave is of the same form as
the twoscale model given by Eq. (2.51), with the spherical values for N
3
, ρ
0
, b
1
, and b
2
used. Throughout the rest of this thesis, we will refer to Eq. (2.58) as the Andrews
asymptotic model for the spherical wave for convenience.
2.6.2 Churnside Asymptotic Analysis
After the phenomenon of saturation of scintillation was understood, Churnside
built upon Fried’s work and published the first significant application of asymptotic
theory to the study of aperture averaging. Churnside used the covariance function in
Eq. (2.48) to devise a framework to study intensity scintillations when the inner scale is
small.
2.6.2.1 Churnside Asymptotic Analysis for the Plane Wave
Churnside evaluated the twoscale model of the covariance function by
representing the two scale functions, b
1
and b
2
, as:
( ) ( )
( )


.

\

− =
=
∫
3 / 5
0
2
1
0
0 0
3 / 4
1
exp
/
3
7
ρ
ρ
ρ
ρρ ρ
b
L x k J dxx b
(2.59)
31
Both functions go to unity when ρ approaches zero, and go to zero when ρ approaches
infinity. Churnside’s scale functions are compact forms of the functions presented in
Chapter 3 of Ref. [1]. Different authors have evaluated these equations under different
constraints [1,29], and Churnside uses limits relevant to aperture averaging. Based on
Eq. (2.59), Churnside calculated the irradiance variance to be:
( )
3 / 1
2
0 2
22 . 1 1 0


.

\

+ = =
L
k
C
I I
ρ
σ (2.60)
or, in terms of the plane wave Rytov variance:
( )
5 / 2
2
2
14 . 1
1
R
I
σ
σ + = (2.61)
The Churnside asymptotic approximation is slightly different from that of Sec. 2.6.1.
2.6.2.2 Churnside Asymptotic Analysis for the Spherical Wave
The twoscale model of the covariance function in Eq. (2.51) is evaluated using
Churnside’s spherical wave representations for b
1
:
( ) ( )
( )  
3 / 5 3 / 5
1
0 0
0
0
3 / 4 2 3 / 1
1
1 exp
1 915 . 0
x r
L
k
J d x dxx b
− − ×

.

\

− =
∫ ∫
∞
−
τ ρρ
ττ ρ
(2.62)
where b
2
is the same as in the plane wave case. The Churnside asymptotic equation for
the irradiance variance of a spherical wave is:
3 / 1
2
0 2
86 . 3 1


.

\

+ =
L
k
I
ρ
σ (2.63)
or, in terms of the plane wave Rytov variance:
32
( )
5 / 2
2
2
35 . 5
1
R
I
σ
σ + = (2.64)
The multiplicative constant is slightly higher by Churnside’s calculation, compared to
the earlier theory in Section 2.6.1.
2.7 Irradiance Variance Models Valid in Both Weak and Strong Turbulence
Conditions
2.7.1 Scintillation Index Model for a Plane Wave
Andrews and Philips developed a model to describe plane wave characteristics
over the entire range of fluctuation conditions, from weak to strong turbulence [33].
They use the effective Kolmogorov spectrum, which modifies the Kolmogorov
spectrum by two filter functions that exclude midscale sizes that have an insignificant
effect on the propagating wave in the moderatetostrong turbulence region. The
effective Kolmogorov spectrum model is [20,33]:
( ) ( ) ( )   κ κ κ κ
y x n n
G G C + = Φ
− 3 / 11 2
,
033 . 0
l
(2.65)
where the largescale filter function that passes only spatial frequencies
x
κ κ < is:
( )


.

\

− =
2
2
exp
x
x
G
κ
κ
κ (2.66)
and the smallscale filter function, passing only spatial frequencies
y
κ κ > , is:
( )
( )
6 / 11
3 / 11
y
y
G
κ κ
κ
κ
+
= (2.67)
33
The wavenumber spectrum given by Eq. (2.65) is a twoscale model for generic scale
sizes x and y. Likewise, modified Rytov theory may be used to define the scintillation
index,
2
I
σ , in terms of largescale and smallscale scintillations [20,33]:
( ) 1 exp
2
ln
2
ln
2
− + =
y x I
σ σ σ (2.68)
where
2
ln x
σ and
2
ln y
σ are the largescale and smallscale log irradiance fluctuations.
These functions are evaluated in Ref. [33], to define the scintillation index for a plane
wave, excluding inner scale effects, as:
( ) ( )
1
69 . 0 1
51 . 0
11 . 1 1
49 . 0
exp
6 / 5
5 / 12
2
6 / 7
5 / 12
2
2
−
+
+
+
=
R
R
R
R
I
σ
σ
σ
σ
σ (2.69)
where
2
R
σ is the Rytov variance for a plane wave. Eq. (2.69) reduces to the Rytov
approximation in weak turbulence conditions, and the asymptotic model from Eq. (2.58)
in strong turbulence saturation conditions.
2.7.2 Scintillation Index Model for a Spherical Wave
The scintillation index (SI) model attempts to describe the behavior of irradiance
fluctuations over the entire range of turbulence strengths. The model for a spherical
wave is developed using the effective Kolmogorov spectrum, using the same twoscale
filter functions presented in Eqs. (2.66) and (2.67) for the plane wave case. The small
scale and large scale log irradiance fluctuations are evaluated to define the scintillation
index of a spherical wave, neglecting inner scale effects, by [20,33]:
( ) ( )
1
23 . 0 1
20 . 0
19 . 0 1
20 . 0
exp
6 / 5
5 / 12
2
6 / 7
5 / 12
2
2
−
+
+
+
=
R
R
R
R
I
σ
σ
σ
σ
σ (2.70)
34
where
2
R
σ is the plane wave Rytov variance. Eq. (2.70) has the same form as Eq.
(2.69), when we rewrite Eq. (2.69) using the relation between the spherical wave and
plane wave Rytov variances
2 2
,
4 . 0
R sph R
σ σ = .
2.7.3 Scintillation Index Model using the Atmospheric Spectrum
The modified atmospheric spectrum is useful when accounting for the influence
of the inner scale, l
o
, in the calculation of the irradiance variance. The spectrum is still
represented by a two scale model, with an enhanced description of the largescale filter
function integrated into the model. In order to properly account for the influence of the
inner scale, the new largescale filter function is defined as [20]:
( ) ( ) ( )


.

\
 −
×
¦
)
¦
`
¹
¦
¹
¦
´
¦


.

\

−


.

\

+


.

\
 −
=
− =
2
2
6 / 7
2
2
2 2
exp
25 . 0 80 . 1 1 exp
exp ,
x
l l l
x o o x
f G
κ
κ
κ
κ
κ
κ
κ
κ
κ κ κ κ l l
(2.71)
where the term in the {} arises out of the modified atmospheric spectrum in Eq. (2.25).
The smallscale filter function has the same form as in Eq. (2.67). Eqs. (2.71) and
(2.67) may be combined to form the Effective Atmospheric Spectrum, which is an
extension of the modified atmospheric spectrum [20]:
( ) ( ) ( )   κ κ κ κ
y o x n n
G G C + = Φ
−
l , 033 . 0
3 / 11 2
(2.72)
The irradiance variance of a spherical wave using the spectrum in Eq. (2.72) may be
broken down into components attributed to the modified atmospheric spectrum and
effective atmospheric spectrum [20]:
35
( ) ( ) 
( )
1
62 . 0 9 . 0 1
69 . 0 1 51 . 0
, exp ,
5 / 12
,
2
5 / 6
2
2
, 2
6 / 5
5 / 12 2
2
ln
2
−
+


.

\

+
+
+
=
−
sp R
mas
sp R
mas mas
o x o I
d d
D D
σ
σ
σ
σ σ
σ σ l l
(2.73)
where
L
kD
d
4
2
= . The components of Eq. (2.68) are defined in Eqs. (2.71) through
(2.74). The spherical wave Rytov variance is given by:
6 / 11 6 / 7 2 2
,
497 . 0 L k C
n sp R
= σ (2.74)
The scintillation index using the modified atmospheric spectrum in weak turbulence is
[20]:
( )
( )
( )
1 for
50 . 3
3
tan
4
5
sin
9
52 . 0
3
tan
3
4
sin
9
61 . 2
3
tan
6
11
sin / 9 1 4 . 0 65 . 9
2
,
6 / 5
1
24 / 7
2
1
4 / 1
2
1
12 / 11
2 2
,
2
<
¦
)
¦
`
¹
−

.

\

+
−

.

\

+
+
¹
´
¦

.

\

+ ≅
−
−
−
sp R
l
l
l
l
l
l
l sp R mas
Q
Q
Q
Q
Q
Q
Q
σ
σ σ
(2.75)
where
2
/ 89 . 10
o l
k L Q l = . In saturated strong turbulence, the scintillation index becomes
[20]:
( )
6 / 7
2
, 6 / 1
6 / 7 2
,
2
100
for ,
57 . 6
1
l
sp R
l sp R
mas
Q
Q
>> + ≅ σ
σ
σ (2.76)
Either Eq. (2.75) or Eq. (2.76) may be used in Eq. (2.73) depending on the strength of
turbulence experienced on the link. Finally, the part of the scintillation index resulting
from the effective atmospheric spectrum, ( )
o x
D l ,
2
ln
σ , in Eq. (2.73) is defined by [20]:
36
( )


.

\

+
−


.

\

+
+ ×


.

\

+
=
12 / 7
2 / 1
6 / 7
2
,
2
ln
25 . 0
75 . 1 1
04 . 0 ,
l xd
xd
l xd
xd
l xd
l xd
sp R o x
Q
Q
Q
Q
D
η
η
η
η
η
η
σ σ l
(2.77)
where
6 / 1 2
,
2
20 . 0 18 . 0 1
56 . 8
l sp R
xd
Q d σ
η
+ +
= (2.78)
Although Eq. (2.73) may be used to quantify the effects of the inner scale l
o
on the
irradiance variance, it is generally accepted that l
o
has little effect on the strong
turbulence irradiance variance [20]. As such, the scintillation index model using the
atmospheric spectrum presented in Eq. (2.73) will only be used in conjunction with
weak turbulence analysis.
37
Chapter 3
Aperture Averaging in Weak Turbulence
3.1 Introduction
As electromagnetic waves propagate through the atmosphere, they undergo
refraction by turbulent eddies. They eddies are characterized by localized variations in
refractive index from the mean. As a propagating wave moves through these eddies,
spatially and temporally random irradiance patterns are formed. When a portion of the
wave reaches a receiver, the receiver aperture will average across all of these random
spatial fluctuations unless the aperture is smaller than the spatial scale of the
fluctuations. This phenomenon is called aperture averaging. The aperture averaging
factor of the receiver is defined as the ratio of the irradiance fluctuations seen at a
receiver with diameter D, to those fluctuations seen by a point receiver (or a receiver
that is small enough to approximate a point receiver).
Aperture averaging theory has been extensively developed for plane and
spherical waves in weak turbulence conditions [1,10,11,16,21,24,33]. Minimal theory
is available for the strong turbulence regime [20,24]. There has been some previous
experimental work [10,11], but early experiments did not account for scintillation
saturation, and resulted in data markedly different from that predicted by theory. Later
experiments did not sufficiently account for background light, and were limited by the
short path lengths under investigation[11].
38
Aperture averaging of optical scintillations is an essential consideration in any
receiver telescope design. Incorporating aperture averaging techniques helps optimize
optical receiver designs; in that a larger aperture will collect more incoming radiation
and reduce the likelihood of fading on an optical communications link. In this chapter,
an experiment to quantify the system enhancement of an aperture averaged receiver is
conducted. This experiment is built upon the theoretical foundation presented in
Chapter 2. A description of the methodology of this experiment is followed by a
presentation and discussion of experimental results.
3.2 Data Analysis for Weak Turbulence
3.2.1 General Aperture Averaging Form
The generic form of the aperture averaging factor for a circular aperture with
diameter D is [10,11]:
( )
( )
( )
( )
( ) ( ) ( )
∫
− − = =
−
D
D
r
D
r
D
r
I
I
I
I
rdr
D C
r C D
A
0
2 / 1
1
2 2
2
2
2
1 cos
16
0 0 π σ
σ
(3.1)
where r is the distance between two points on the aperture surface, C
I
(r) is the spatial
covariance of the irradiance, C
I
(0) is the variance of the irradiance, and the term in the
square brackets is the modulation transfer function (MTF) of the circular aperture.
3.2.2 Weak Turbulence
Weak turbulence was defined in Fig. 2.4 as the case when the transverse
coherence length of the received wave is much larger than the Fresnel length. The
covariance of the irradiance is [11]:
39
( ) ( ) ( )
( )
∫ ∫
∞
−
Φ =
0 0
2
2
0
2 2
2
sin 16
L
n I
k
s z L
rs dzJ d k r C
κ
κ κ κκ π (3.2)
where k = 2π/λ, ( ) κ
n
Φ is the wavenumber spectrum, z is the distance traveled over path
length L, J
0
is a Bessel function of the first kind, and s is a scaling factor that is 1 for a
plane wave and z/L for a spherical wave.
3.2.2.1 Plane Wave, Small l
o
To calculate the covariance values necessary to input into the aperture averaging
factor equation, the Kolmogorov spectrum in Eq. (2.22) may be used since the inner
scale is much smaller than the Fresnel length, l
o
<< (L/k)
1/2
. After the Kolmogorov
spectrum is substituted into Eq. (3.2), and we assume a point receiver with r = 0, the
variance of the irradiance is found as [11]:
( )
2 2 6 / 11 6 / 7 2
23 . 1 0
R I n I
L k C r C σ σ ≡ = = = (3.3)
The exact aperture averaging factor for a plane wave is [11]:
( )
( ) ( )  
∫
∫ ∫
− −
×
−
=
−
∞
− −
1
0
2 / 1
2 1
0
2
2
0
3 / 8 6 / 11 6 / 5
1 cos
2
sin 6 . 21
ydy y y y Dy J
k
z L
dz d L k A
κ
κ
κκ
(3.4)
The inner integral of the Bessel function is given by ( )
2 2 2
1
2 / D D J κ κ π . After a change
of variables to let u = κD/2, A becomes solely a function of (kD
2
/4L)
1/2
[11]:
( )
∫


.

\

−


.

\

=
−
du
kD
Lu
Lu
kD
u J u
L
kD
A
2
2
2
2
2
1
3 / 14
6 / 5
2
4
sin
4
1
4
47 . 8 (3.5)
40
It is seen that for a small receiver aperture, when kD
2
/4L << 1, A approaches one as
expected. For large apertures, when kD
2
/4L >> 1, the sine term is small and may be
replaced by the expansion
6
3
sin
γ
γ γ − ≈ , so that [11]:
( )
∫
∞
−
−
−


.

\

=


.

\

≈
0
6 / 7
2
2
1
3 / 2
6 / 7
2
4
932 . 0
4
41 . 1
L
kD
du u J u
L
kD
A (3.6)
The D
7/3
dependence indicates that the scattering due to turbulence is over a small
angle. The light will not be scattered wide enough to miss the receiver aperture
entirely. The total aperture averaging factor in the weak turbulence plane wave region
has been approximated by Churnside as [11,24,33]:
1
6 / 7
2
4
07 . 1 1
−


.

\

+ =
L
kD
A (3.7)
Eq. (3.7) is plotted in Fig. 3.1 along with the exact theory formula, Eq. (3.4), for the
plane wave. There is a 17% difference in the approximate value of A from the exact
theory when 1
4
2
=
L
kD
[11]. Other approximations have been made to the
theoretical aperture averaging factor, and show slightly better results. Andrews
reported the approximate form for the aperture averaging factor as [34]:
6 / 7
2
4
062 . 1 1
−


.

\

+ =
L
kD
A (3.8)
At the same location, where 1
4
2
=
L
kD
, the Andrews approximation is 7% higher
than the exact theory assuming a Kolmogorov spectrum.
41
Fig. 3.1: Aperture averaging factor for a plane wave plotted against the ratio of the
aperture radius to the Fresnel zone size. The solid line is the exact theory from Eq. (3.4)
while the dashed line is Churnside’s approximation in Eq. (3.7) [11].
3.2.2.2 Plane Wave, Large l
o
Either the Tatarskii spectrum from Eq. (2.23) or the Hill spectrum [19] may be
used when the inner scale is large l
o
>> (L/k)
1/2
. The Tatarskii spectrum is analytically
simpler than the Hill spectrum, and the results are similar [24]. By substituting Eq.
(2.23) into the covariance Eq. (3.2), and letting 0 → r , the covariance becomes [11]:
( )
3 / 7 3 2
8 . 12 0
−
=
o n I
L C C l (3.9)
By following a similar method as in Section 3.2.2.1, and making the substitution u =
kD/2, the aperture averaging factor is [11]:
( )
∫
∞
−
−


.

\

−


.

\

=
0
2
2 2
2
1
3 / 2
3 / 7
1141 . 0 exp 686 . 0 du
D
u
u J u
D
A
o
o
l
l
(3.10)
42
When D >> 1, the exponential in Eq. (3.10) approaches unity, and A can be
approximated by [11]:
1
3 / 7
21 . 2 1
−


.

\

+ =
o
D
A
l
(3.11)
An improved approximation to Eq. (3.11) is [11]:
6 / 7
2
19 . 2 1
−


.

\

+ =
o
D
A
l
for 5 . 0 / 0 ≤ ≤
o
D l (3.12)
3.2.2.3 Plane Wave, l
o
on the Order of the Fresnel Length
In an intermediate regime, when the inner scale is similar to the Fresnel length, a
decision must be made as to which approximation to the aperture averaging factor to
use. By setting the two aperture averaging approximations equal to each other, Eq.
(3.8) = Eq. (3.12), they are found to be equal when l
o
= 2.73(L/k)
1/2
. Churnside has
recommended using the small inner scale approximation when l
o
< 2.73(L/k)
1/2
, and the
large inner scale approximation when l
o
> 2.73(L/k)
1/2
[11]. Andrews recommends
using Eq. (3.8) when kD
2
/4L < 1 [34].
3.2.2.4 Spherical Wave, Small l
o
The Kolmogorov spectrum may be used to evaluate the covariance of the
irradiance (Eq. (3.2)), along with s = z/L for a spherical wave, and r = 0 [11]:
( )
2
,
2 6 / 11 6 / 7 2
497 . 0 0
sp R I n I
L k C C σ σ ≡ = = (3.13)
The exact form of the aperture averaging factor in Eq. (3.1) is written using the
spherical wave covariances [11]:
43
( )
( )  
∫
∫ ∫
− −

.

\

×
−
=
−
∞
− −
1
0
2 / 1
2 1
0
0
1
0
2
2 3 / 8 6 / 11 6 / 5
1 cos
2
sin 4 . 53
y y y
L
KDzy
dyyJ
kL
z L z K
dz dKK L k A
(3.14)
After evaluating the aperture averaging factor integral as in Section 3.2.2.1, it can be
approximated as [11]:
1
6 / 7
2
4
214 . 0 1
−


.

\

+ =
L
kD
A (3.15)
The exact and approximate forms of the aperture averaging factor for a spherical wave
with small inner scale are shown in Fig. 3.2. Where 1
4
2
=
L
kD
, the approximate
version predicts the amount of aperture averaging to be 86% less than the theory
[11,24].
Fig. 3.2: Aperture averaging factor for a spherical wave with small inner scale. The
solid line is the exact theory given a Kolmogorov spectrum, while the dashed line is the
approximate formula of Eq. (3.15) [11].
44
3.2.2.5 Spherical Wave, Large l
o
By evaluating the covariance of the irradiance in Eq. (3.2) with the Tatarskii
spectrum, and r = 0 [11]:
( )
2 3 / 7 3 2
28 . 1 0
I o n I
L C C σ = =
−
l (3.16)
The aperture averaging factor is approximated by [11]:
1
3 / 7
109 . 0 1
−


.

\

+ =
o
D
A
l
(3.17)
3.2.2.6 Spherical Wave, l
o
on the Order of the Fresnel Length
By following the argument in Section 3.2.2.3, Eq. (3.15) is equal to Eq. (3.17)
when l
o
is 1.5 times the Fresnel length. Churnside recommends using the small inner
scale approximation when l
o
< 1.5(L/k)
1/2
and the large inner scale approximation when
l
o
>1.5(L/k)
1/2
[11]. Andrews recommends using the small inner scale result when
kD
2
/4L <1.
3.3 Aperture Averaging Experiment
Experimental aperture averaging data is necessary to verify previously
developed theory [1,13,16,17,20] and ensure that modern optical communication
systems are correctly optimized. After the development of plane wave theory of optical
propagation through turbulence, the advent of the laser allowed experimental studies of
aperture averaging. However, these experiments were conducted over very long paths
and were susceptible to the effects of saturation of scintillation [17]. A more recent
experiment by Churnside [11,24] used 100, 250, 500, and 1000 m paths to avoid
45
saturation effects, and produced results in reasonably good agreement with spherical
wave theory over short paths of 250 m [11,24]. Over longer paths, the experimental
data tended to diverge from theory. The scintillometer used in the experiment measured
turbulence over a 250 m path, and not the path length under test, which reduces
confidence in the precision of the C
n
2
data. Other shortcomings of this experiment
include the lack of simultaneous background light measurements and a use of only six
apertures, ranging from 1 mm to 5 cm. Many commercial free space optical
communication units use apertures with diameters larger than 5 cm, so breadth of data
collected by Churnside in Ref. [11] is insufficient for practical use.
3.3.1 Aperture Averaging Transmitter and Receiver Systems
The aperture averaging transmitter uses a 21 mW JDS Uniphase HeNe laser.
The laser operates in a single mode (TEM
00
) at 632.8 nm, with a 0.70 mm output
diameter and 1.15mrad beam divergence. The beam passes through an optical chopper
and a 30x Melles Griot beam expander. The beam expander is adjustable so that the
beam diameter at the receiver may be tuned; for this experiment, the beam diameter at
the receiver is ~1.5 m. The optical chopper serves an important purpose during the data
acquisition in that it allows us to calculate mean background light levels at the same
time that intensity data is being recorded. The optical chopper is operated with a
frequency of 3.636 ms, which is on the order of the time that the turbulence on a
specific path through the atmosphere remains constant. A photograph of the aperture
averaging transmitter is shown in Fig. 3.3.
46
Fig. 3.3: Aperture averaging transmitter on the roof of the A.V. Williams Building at
the University of Maryland, College Park.
The laser light propagates over an 863 m path between the A.V. Williams
Building and the Chesapeake Building. The transmitter on the roof of A.V. Williams is
approximately 14 meters above the ground, while the receiver in the Chesapeake
Building is approximately 12 meters above ground. An aerial photograph of the path,
with the line of sight indicated, is shown in Fig. 3.4. The terrain propagated over is
mostly asphalt parking lots, with a few trees and grassy fields. One part of the path
passes near the rotunda on the east side of the Comcast Center, which may have some
affect on the homogeneity of the turbulence measurements.
47
Figure 3.4: An aerial photograph of the propagation path (dotted line). Satellite
photograph taken in 2002 by GlobeXplorer.
The receiver system is composed of two receive apertures, a point receiver and a
variable aperture receiver, shown in Fig. 3.5. The point receiver is an effective
scintillometer and is used to calculate pathaveraged C
n
2
measurements. Light is first
filtered through a laser line filter at 632.8 nm with a 10 nm passband width. The filter
removes any stray light from interfering with the operation of the scintillometer. The
laser line filter is angle dependent, so only light parallel to the propagation path at a
wavelength near 632.8 nm will pass. The aperture diameter of the point receiver is 5
mm, which is smaller than the Fresnel zone size for the path, cm 34 . 2 = L λ . The
minimum aperture size, required to measure A = 1 in weak turbulence, is the diameter
48
where 1
4
2
=
L
kD
. For this experimental setup, D
min
= 1.86 cm. Since the
scintillometer aperture diameter is smaller than the minimum aperture diameter for A =
1, the baseline C
n
2
measurements are true pathaveraged values. After passing through
the scintillometer receive aperture, light is detected by a PerkinElmer FFD100 Si
photodetector. The FFD100 has a 5.1 mm
2
active area, with a responsivity of 0.43
A/W at 632.8 nm. The detected signal is amplified by a variable transimpedance
amplifier and recorded by LabVIEW.
The variable aperture receiver is a Rolyn Optics 20 cm planoconvex lens. An
aperture stop of sizes ranging from 1 cm to 16 cm is placed behind the plane side of the
lens, to limit the aperture diameter. To ensure that the receiver does not saturate, a New
Focus neutral density filter wheel is placed in front of the photodetector. At large
receive apertures, the filter wheel is necessary to avoid detector saturation and measure
the intensity variance accurately. Since neutral density filters are not angle dependent,
they will not affect the distribution of irradiance fluctuations. The incident beam is then
received by a Perkin Elmer FFD200 Si photodetector and amplified through a variable
transimpedance amplifier. The transimpedance gain is set when the photodetector is
illuminated through the smallest aperture diameter; thereby allowing the electronic gain
to remain constant, while the receive optical power is controlled by changing the neutral
density filter transmittance. The FFD200 has a 20 mm
2
active area with a responsivity
of 0.43 A/W at 632.8 nm. The signal is than recorded and processed in LabVIEW.
Typical oscilloscope traces of the aperture averaged and scintillometer signals are
shown in Fig. 3.6.
49
Fig. 3.5: Diagram of the aperture averaging receiver setup. The scintillometer channel
uses a 5 mm diameter receive lens.
Fig. 3.6: An optically chopped signal propagated over 863 m and viewed on the
Tektronix oscilloscope. Channel 1 is the signal detected by the scintillometer, while
Channel 2 is detected by the aperture averaged receiver. The signal detected on
Channel 2 shows smaller intensity fluctuations.
50
3.3.2 LabVIEW Data Acquisition
A customized LabVIEW program, whose front panel is shown in Fig. 3.7,
samples the optically chopped 275 Hz signal at a rate of 3000 samples/sec. Data
collection for each aperture is typically done in 5 to 15 minute intervals. The maximum
and minimum interchannel sampling delays for two channels are 62.5 µsec and 5 µsec,
respectively. This time difference is well below the reordering time of the turbulent
atmosphere, so the turbulent channel is quasistationary.
Fig. 3.7: Front panel of the LabVIEW program designed to calculate and record
irradiance statistics for the aperture averaging experiment.
The LabVIEW program calculates irradiance statistics relevant to determining
the aperture averaging factor over a 1 minute sampling interval. Fig. 3.8 is a flowchart
representing the data acquisition process for one channel. The chart explains the
methodology used to determine the normalized irradiance variance,
2
I
σ , from the
51
measured waveform data. The waveform type in LabVIEW is a new data type
introduced in LabVIEW 6.0 that stores information on sampled voltage (y), acquisition
start time (t
0
), and sampling interval (∆t). The waveform type makes signal analysis
easier because the sampled data is synchronized with the sampling time. LabVIEW
also includes an extensive functionality to take advantage of the waveform type.
From Fig. 3.8, the detected chopped signal is sorted into high and low voltage
values. The mean of the low values is the average background signal. This background
signal is than subtracted from the recorded high values, which is then averaged over a 1
minute interval to give the mean intensity. Simultaneously, the variance of the high
values for that acquisition period is calculated. The variance of the high values is then
normalized by the square of the mean intensity, which gives
2
I
σ .
Table 3.1 shows the 13 values that are output by LabVIEW into a data file. The
two channels recorded are called Ch. 0 and Ch. 1, due to the naming conventions in
LabVIEW. The one 1min path averaged C
n
2
is then calculated from the normalized
irradiance variance
2
I
σ using the equation for the plane wave Rytov variance. The
aperture averaging factor, A, is calculated each sampling interval from the ratio of the
normalized intensity variances of the two channels, as in Eq. (3.1). These values are
then written to a data file every minute. The background variance values are only
output to ensure that there is no significant baseline wander on the measured signal that
could distort the measured irradiance statistics.
The LabVIEW program requires a large storage buffer to hold 60 seconds worth
of samples, measured at 3000 samples per second. Shorter averaging times were also
used in the initial experiments, however results were much more erratic. The longer
52
averaging time ensures that there are a sufficient number of samples so that lognormal
statistics are obeyed [35]. Only half the number of samples acquired may be used in the
calculation of the irradiance variance, since the signal is optically chopped. The other
half of the samples are used to determine the mean background level.
Fig. 3.8: Flowchart of the data acquisition process in LabVIEW for one channel. The
diagram explains how intensity and background levels are determined for the sampled
signal averaged over a 1 min. interval.
53
Time measured to 1/1000
th
of a second
Signal Ratio
0 .
1 .
Ch
Ch
I
I
Aperture Averaging Factor, A
1 .
2
0 .
2
Ch I
Ch I
σ
σ
C
n
2
Ch. 0
6 / 11 6 / 7
0 .
2
23 . 1 L k
Ch I
σ
Irradiance Variance Ch. 0
0 .
2
Ch I
σ , variance of high values normalized
by
2
0 . Ch
I
Average Intensity Ch. 0
0 . Ch
I , mean of high values with
background level subtracted out
Average Background Ch. 0 mean of low values
Background Variance Ch. 0 variance of low values
C
n
2
Ch. 1
6 / 11 6 / 7
1 .
2
23 . 1 L k
Ch I
σ
Irradiance Variance Ch. 1
1 .
2
Ch I
σ , variance of high values normalized
by
2
1 . Ch
I
Average Intensity Ch. 1
1 . Ch
I , mean of high values with
background level subtracted out
Average Background Ch. 1 mean of low values
Background Variance Ch. 1 variance of low values
Table 3.1: List of 13 values output per averaging interval in LabVIEW data file. Here,
Channel 0 represents the aperture averaged signal, while Channel 1 represents the
scintillometer signal.
3.3.3 Fetch Effects
The height of the path above ground, approximately 12 m, limits the strength of
turbulence experienced on an average day. Although the terrain along the path is
mostly flat with grass of pavement, there are a few trees and two buildings that impinge
54
on the path, which could result in fetch effects. To be able to neglect fetch effects, the
height above ground must be larger than the boundary layer 100to1 rule for fetch
effects [36,37]. The rule states that there will be an internal boundary layer established
at a height that is 1/100
th
of the half width of the path. For our path of 863 m, this
height is approximately 4.3 m. Since that propagation path is higher than this internal
boundary layer, we can assume that we are nearly free of fetch effects [36,37].
3.4 Plane Wave Experimental Results
Theory and approximation of the plane wave aperture averaging factor with
small inner scale was presented in Sec. 3.2.2.1. The aperture averaging factor, A, is
typically plotted against the ratio of the aperture radius to the Fresnel zone size. Recall
that the Fresnel zone size is the dominant scale size in cases of weak turbulence. Three
accepted approximations to the aperture averaging factor are calculated and plotted in
Fig. 3.9. The scintillation index model is an attempt to define an irradiance variance
that is valid across all turbulence levels. The plane wave scintillation index diverges
from its Rytov predicted weak turbulence value when 5 . 0
2
≥
R
σ . Therefore, for
different strengths of turbulence indicated by C
n
2
, the scintillation index model does
predict different aperture averaging results from the accepted approximations to weak
turbulence theory. A C
n
2
value of 10
16
m
2/3
corresponds to
2
R
σ = 0.0043. A C
n
2
of
10
14
m
2/3
changes the Rytov variance by a factor of 100 and we find that
2
R
σ = 0.43.
55
Fig. 3.9: Plane wave aperture averaging factor plotted as the Churnside approximation
Eq. (3.7), the Andrews approximation Eq. (3.8), and the scintillation index (SI) model
for various values of C
n
2
.
Experimental data plotted against the ratio
L
kD
4
2
is shown in Fig. 3.10. The
mean data is plotted with error bars of one standard deviation. The data does follow the
trend predicted by theory. For smaller apertures, the experimental aperture averaging
data seems to retain high values, whereas the approximate models drop off. Since the
laser is diverged, we expect better agreement with the spherical wave aperture
averaging models.
The Andrews scintillation index model is the only model that takes into account
the strength of turbulence in the weak fluctuation region. As seen from Fig. 3.10, the SI
model predicts less aperture averaging than weak fluctuation theory for given values of
receiver aperture. When our experimental data was analyzed, there was no significant
56
dependence of A on C
n
2
in the weak turbulence region, and certainly not the substantial
variation predicted by the scintillation index model.
Fig. 3.10: New mean experimental data plotted with error bars, along with
approximations to the aperture averaging factor.
The mean data is plotted along with the experimental Churnside data in Fig.
3.11. The Churnside data was measured on a 500 m path, although C
n
2
was measured
over only half the path using an incoherent scintillometer [11,24]. It was assumed that
due to the flatness of the terrain that the measured C
n
2
was indicative of the path
averaged value. The laser and scintillometer were propagated over a plateau of
grassland at a height of 1.5 m above ground. The 1.2 mrad divergence of the HeNe was
used directly, so it was assumed that the data would be a better fit to the spherical wave
theory. It is apparent from Fig. 3.11 that the plane wave theory does not sufficiently
57
characterize the amount of aperture averaging taking place, especially at larger aperture
diameters.
Fig. 3.11: New mean aperture averaging data plotted with Churnside experimental data
taken over a 500 m path. Neither the plane wave Churnside approximation nor the SI
model are a good fit to the data.
3.5 Spherical Wave Experimental Results
Spherical wave weak turbulence theory and approximation was presented in
Section 3.2.2.4. Since the beam divergence in this experiment is ~ 1 mrad, we expect
the data to be in good agreement with the spherical wave analysis. Fig. 3.12 compares
the Churnside approximation to the scintillation index model for the aperture averaging
factor. The Churnside model predicts a higher knee than the scintillation index model.
As shown in Fig. 3.2, this knee overpredicts the exact theory given a Kolmogorov
spectrum, thereby underestimating the amount of aperture averaging taking place.
58
Fig. 3.12: Comparison of two different models for the spherical wave in weak
turbulence. The Churnside approximation, Eq. (3.7), predicts a higher knee in the curve
than the SI model.
Experimental mean aperture averaging data with error bars is shown in Fig.
3.13. In contrast to what is predicted by the exact theory, we see a higher knee in the
data that the theory predicts near 1
4
2
≈
L
kD
. It also shows that the Churnside
approximation is a better fit to the spherical wave data than the scintillation index
model. This supports a generally accepted notion that aperture averaging in weak
turbulence is independent of C
n
2
.
59
Fig. 3.13: Experimental mean aperture averaging data plotted against the ratio of
aperture radius to the Fresnel zone size.
The Churnside experimental data measured over a 500 m path is displayed along
with our mean data in Fig. 3.14. The Churnside data does not exhibit the same knee
that our new data shows. Churnside assumed that a high knee is a result of a violation
of propagation uniformity due to cloud cover over part of the path [11]. This seems to
be an unlikely explanation for the knee in our new data, since the data was taken under
clear weather conditions with wind speeds less than 4 m/s and low relative humidity. It
is possible that a few trees and a building near the propagation path adversely affect the
homogeneity of the atmosphere in that area. There may be additional scintillations
induced near the transmitter since the laser propagates about 1.5 m above the roof of
A.V. Williams over a length of 10 m.
60
Fig. 3.14: Comparison of our new mean data with the Churnside data.
It is not apparent that the knee in the new data is due to inner scale effects. A
large inner scale (larger than the Fresnel zone size), should reduce the aperture
averaging factor, and not increase it as we are seeing. The Fresnel zone size is on the
order of 1 cm, making it unlikely that the inner scale ℓ
o
is larger than 1 cm for this path
length and height. However, an inner scale that is smaller than the Fresnel zone size
will still impact the data in weak turbulence conditions. Therefore, irradiance variance
models including an ℓ
o
dependence should be investigated.
3.6 New Aperture Averaging Model
Data was fit to the generalized form of the Churnside model in Eq. (3.15). A
must be 1 when the aperture diameter becomes infinitesimal, so only the coefficient of
61
the
L
kD
4
2
term was varied. The genfit function in MathCad allows arbitrary
functions to be fitted to data matrices. Using this function, a fit to the measured
aperture averaging data was found to be:
1
6 / 7
2
4
196 . 0 1
−


.

\

+ =
L
kD
A (3.18)
This new recommended fit, along with the new data and the Churnside approximation,
given by Eq. (3.15), are plotted in Fig. 3.15.
Fig. 3.15: New weak turbulence fit (Eq. 3.18) shown with new data and the Churnside
model (Eq. 3.15).
The limitations of the
L
kD
4
2
term do not pick up on the nuances of the data;
more specifically, it does not predict the knee that we see in Fig. 3.14. Including a
62
square root term in the model does pick up on the knee, and we find the approximate
formula using MathCad:
1
12 / 7
2
6 / 7
2
4
162 . 0
4
266 . 0 1
−


.

\

−


.

\

+ =
L
kD
L
kD
A (3.19)
Fried had previously proposed that a fit could be improved with a
12 / 7
2
4


.

\

L
kD
term,
although Churnside discounted the term as unnecessary [16].
Fig. 3.16: New fit to weak turbulence data with an additional square root term, given by
Eq. (3.19), plotted along with the Churnside approximation in Eq. (3.15).
3.7 Determination of the Inner Scale of Turbulence, ℓ
o
Although inner scale could not be measured during the aperture averaging
experiments, ℓ
o
may be determined by fitting available aperture averaging models to the
63
collected new data. Similar to the development of a new aperture averaging model in
Section 3.6, the genfit function in MathCad is used to determine ℓ
o
.
In a first attempt to extract a value of ℓ
o
, the new data is fit to the plane wave
aperture averaging factor given a large ℓ
o
given by Eq. (3.11). The fit determined ℓ
o
=
5.27 cm, a value that is approximately an order of magnitude larger than what would be
expected in weak turbulence conditions. The new data is plotted along with Eq. (3.11)
using ℓ
o
= 5.27 cm in Fig. 3.17. In our previous analysis, spherical wave models are
expected to be in better agreement with data acquired on the ChesapeakeAVW test
range. Although the fit in Fig. 3.17 is reasonable, the assumption that the diverged
beam approximates a plane wave is not.
Fig. 3.17: Eq. (3.11) plotted using the curve fit value of ℓ
o
= 5.27 cm, against the ratio
of the aperture diameter to the inner scale. The dashed line represents Eq. (3.11).
64
The new data was next fitted to the aperture averaging factor for a spherical
wave given a large inner scale, given by Eq. (3.17). The fit resulted in a predicted ℓ
o
=
1.45 cm. This value is on the high end of the range of inner scale measurements
reported in the literature [60,61]. The new data along with Eq. (3.17) with ℓ
o
= 1.45 cm
is shown in Fig. 3.18.
Fig. 3.18: Spherical wave aperture averaging factor for ℓ
o
= 1.45 cm using Eq. (3.17)
plotted along with the new data. The dashed line represents Eq. (3.17).
Since a reasonable value of ℓ
o
has not been determined using the standard
approximation to the aperture averaging factor, a more involved model for the
dependence of the aperture averaging factor on ℓ
o
is investigated. Recall that the
modified atmospheric spectrum and effective atmospheric spectrum presented in
Section 2.7.3 compose the best model to date for determining the influence of inner
65
scale on the irradiance variance. Using these spectra and assuming spherical wave
propagation, the aperture averaging factor is calculated by taking the ratio of Eq. (2.73)
with aperture diameter D to Eq. (2.73) with D = 0:
( )
( )
o I
o I
D
A
l
l
, 0
,
2
2
σ
σ
= (3.20)
By using the genfit function in MathCad, the aperture averaging factor using the
atmospheric spectrum predicts ℓ
o
= 4.008 mm. The new data along with the
atmospheric spectrum aperture averaging factor for ℓ
o
= 4.008 mm are plotted in Fig.
3.19.
Fig. 3.19: New data is plotted using the predicted value ℓ
o
= 4.008 mm. The inner
scale was determined using the atmospheric spectrum, which is represented by the
dashed line.
66
The primary reason the Churnside plane wave and spherical wave
approximations to the aperture averaging factor, given by Eq. (3.11) and (3.17), did not
produce accurate inner scale values is that the models are meant to be used when the
inner scale is larger than the Fresnel zone size. In our case, we expect the Fresnel zone
size to be larger than ℓ
o
. The atmospheric spectrum produces reasonable results
because it is not constrained by the Fresnel zone size. The atmospheric spectrum
mandates that the inner scale always impacts the irradiance variance in weak turbulence
conditions. Although the model is much more complex than the Churnside
approximations, its usefulness in determining the inner scale, especially when ℓ
o
is not
measurable, is proven.
Referring back to the discussion in Sec. 3.5, the atmospheric spectrum aperture
averaging model with ℓ
o
= 4.008 mm in Fig. 3.19 does not follow the knee of the data,
as the Churnside spherical wave approximation in Fig. 3.18 does. It is still unclear what
physical mechanism contributes to the knee. From Fig. 3.19, we can ascertain that even
small values of ℓ
o
will reduce A. It seems likely that the terrain uniformity and slightly
offhorizontal path could act as forces counteracting the impact of ℓ
o
. Either way, it is
clear that improvements must be made to the atmospheric spectrum models of the
irradiance variance in order to capture the physical nature of the relationship between
the aperture averaging factor and ℓ
o
.
3.8 Conclusions
The experimental results presented in this chapter show the behavior of the
aperture averaging factor in weak turbulence. The data is consistent with models
developed from the Kolmogorov, Tatarskii, and atmospheric spectra for spherical wave
67
propagation. This experiment was carefully planned out, and the results are the most
accurate to date. The aperture averaging measurements presented here consider a
variety of receive aperture diameters, simultaneous measurement of background light
levels, and concurrent measurement of pathaveraged C
n
2
values. Inner scale was also
estimated from the data.
68
Chapter 4
Aperture Averaging in Strong Turbulence
4.1 Introduction
Aperture averaging in strong turbulence is extremely difficult to study because
of the lack of a comprehensive theory for strong intensity scintillations including
saturation effects [1,11,20,21,27,29,33,40,41]. In order to make useful comparisons of
the experimentally determined aperture averaging factor with previously developed
theory and numerical approximations, nonmeasurable quantities must be evaluated for
this experiment. These include the Rytov variance, transverse coherence length, Fresnel
zone, inner scale, and scattering disk size.
4.2 Data Analysis in Strong Turbulence
4.2.1 General Aperture Averaging Form
The general form of the aperture averaging factor for the spherical wave has the
same representation as in weak turbulence theory, given by Eq. (3.1). The spatial
covariance terms are represented by their strong turbulence theory counterparts.
4.2.2 Plane Wave Analysis
4.2.2.1 Churnside Asymptotic Analysis for the Plane Wave
The aperture averaging factor in strong turbulence may be evaluated using
asymptotic theory [11,27]. The theory relies on the assumption that the irradiance
69
variance is the sum of 1 and a perturbation term, as discussed in Section 2.6.1.
Asymptotic theory is the best theory to describe behavior in the saturation regime,
because it uses a twoscale model to describe behavior in the inertial subrange and the
transition into the dissipation region.
The asymptotic method uses three terms from the series expansion of the
covariance function [27]. The asymptotic form of the covariance function for a plane
wave with small inner scale is [11]:
( ) ( )   ( ) ( ) ( )   ρ ρ ρ
ρ
ρ
ρ
2 1
3 / 1
3 2
1
3 / 5
2
0
0
exp b b N C
L
k
I
+ + − = (4.1)
where N
3
is a constant equal to 1.22 for a plane wave, and b
1
and b
2
are functions that
go to unity as ρ goes to zero and go to zero when ρ goes to infinity. Specific
descriptions of the formulas and the constant are found in Sec 2.6.1 and Ref. 27.
Asymptotic theory includes the impact of two scale sizes, the scattering disk, L/kρ
0
, and
the transverse coherence length, ρ
0
. Asymptotic theory is expected to be valid only
where the irradiance variance is close to unity, indicating a region of strong, saturated
pathintegrated turbulence [11].
For zero inner scale, the variance of the irradiance is related to the transverse
coherence length by:
( ) ( )
3 / 1
2
2
0
22 . 1 1 0
L
k
I I
C
ρ
ρ σ + = = = (4.2)
By solving for ρ
0
in Eq. (4.2), we can find the predicted asymptotic value of ρ
0
by using
measured values of
2
I
σ . We can also use the transverse coherence length, ρ
0
, for plane
wave propagation, given in Eq. (4.3), to find predicted values of C
n
2
and, consequently,
2
R
σ using:
70
( )
5 / 3
2 2
0
46 . 1
−
=
n
LC k ρ (4.3)
and
6 / 11 6 / 7
2
2
23 . 1 L k
C
R
n
σ
= (4.4)
Churnside evaluated the aperture averaging factor as a sum of two terms, based on the
two scale functions b
1
and b
2
in Eq. (4.1):
( )  
∫
∫

.

\
 −
=


.

\

− − −
+
=
−
−
1
0
0 2
1
3 / 2
2
2
2
0
2 2
2
2
1
0
3 / 5
0
2 / 1
2 1
2
2
1
2 2
1
3
56
exp 1 cos
2
1 16
L
x kD
J dxx
D k
L
A
Dy
y y y dyy A
I
I
I
I
ρ
σ
σ
ρ
ρ σ
σ
π
(4.5)
By evaluating the two components of the aperture averaging factor for large and small
diameters, the approximate aperture averaging factor is given as [11]:
( )   ( )  
1
3 / 7
2 2
2
1
2 2
2
0
0
162 . 0 1
2
1
908 . 0 1
2
1
−
−
+
−
+ +
+
=
L
D k
I
I D
I
I
A
ρ
ρ
σ
σ
σ
σ
(4.6)
The approximate aperture averaging factor, given in Eq. (4.6), for a plane wave with a
small inner scale is plotted in Fig. 4.1, using AVWChesapeake test range characteristic
values.
71
Fig. 4.1: The plane wave aperture averaging factor A vs. the ratio of the aperture radius
D/2 to the transverse coherence length ρ
0
, for L = 863 m and λ = 632.8 nm. The legend
indicates different values of irradiance variance
2
I
σ , where turbulence strength
increases in the asymptotic limit of 1
2
→
I
σ .
The impact of the two scale sizes, the scattering disk and the transverse coherence
length, is especially evident in the curves with
2
I
σ = 1.25 and 1.5 in Fig. 4.1. According
to asymptotic theory, values of
2
I
σ closer to unity indicate stronger levels of turbulence.
From Eq. (4.2), the width of the plateau, which is the separation of the two scales, may
be determined. For
2
I
σ = 1.5, the width is 14.5. For
2
I
σ = 1.1, the separation is 1815,
which indicates that the drop off in Fig. 4.1 is beyond the range of the chart. The
location of the plateau is also at a lower value of A for
2
I
σ = 1.1.
The graphs for various irradiance variances in Fig. 4.1 show the trend of the
aperture averaging factor in strong turbulence. Initially, there is more aperture
72
averaging in strong turbulence conditions, because A drops off faster in strong
turbulence than in weak turbulence. At larger values of D/2ρ
0
, plateaus form in the
strong turbulence aperture averaging factor from the influence of the twoscale sizes,
the transverse coherence length and the scattering disk size. The plateaus cause the
aperture averaging factor to remain high over a wide range of aperture diameters, and
allow the weak turbulence aperture averaging factor to catch up to the strong turbulence
value. For most practical purposes, D/2ρ
0
will always be small enough so that the link
will experience more aperture averaging in strong turbulence conditions.
4.2.2.2 Andrews Asymptotic Analysis for the Plane Wave
Andrews approached the theory of aperture averaging using the effective
Kolmogorov spectrum and modified Rytov theory to determine a model for aperture
averaging valid over the range of weaktostrong scintillations. This approach will be
fully developed in the following section. In the strong fluctuation region, the
asymptotic relation of the irradiance variance for a plane wave to the Rytov variance
developed from Prokhorov is used [20,27]:
( )
5 / 2
2
2
86 . 0
1
R
I
σ
σ + = (4.7)
which is valid when
2
I
σ >> 1 and the spatial coherence radius,
0
2ρ ρ ≈ , is larger than
the inner scale, ℓ
o
. The aperture averaging factor given by Eq. (4.7) has been plotted in
Fig. 4.2 for a variety of irradiance variance values. Given a value of
2
I
σ , Eq. (4.7) may
be solved for C
n
2
, from which the transverse coherence length for a plane wave is
determined and substituted into Eq. (4.6) to calculate A. This model for A reduces the
width of the plateau on the aperture averaging curve.
73
Fig. 4.2: Aperture averaging factor using the Andrews asymptotic model vs. the ratio of
aperture radius to the transverse coherence length.
4.2.2.3 Scintillation Index (SI) Model for the Plane Wave
Andrews and Philips developed a model to describe plane wave characteristics
over the entire range of fluctuation conditions, from weak to strong turbulence [33].
They use the effective Kolmogorov spectrum, which modifies the Kolmogorov
spectrum by two filter functions to exclude midscale sizes that have an insignificant
effect on the propagating wave in the moderatetostrong turbulence region. The
effective Kolmogorov spectrum model is [20,33]:
( ) ( ) ( )   κ κ κ κ
y x n n
G G C + = Φ
− 3 / 11 2
,
033 . 0
l
(4.8)
where the largescale filter function that passes only spatial frequencies
x
κ κ < is:
( )


.

\

− =
2
2
exp
x
x
G
κ
κ
κ (4.9)
74
and the smallscale filter function, passing only spatial frequencies
y
κ κ > , is:
( )
( )
6 / 11
3 / 11
y
y
G
κ κ
κ
κ
+
= (4.10)
Likewise, modified Rytov theory may be used to define the scintillation index,
2
I
σ , in
terms of largescale and smallscale scintillations [20,33]:
( ) 1 exp
2
ln
2
ln
2
− + =
y x I
σ σ σ (4.12)
where
2
ln x
σ and
2
ln y
σ are the largescale and smallscale log irradiance fluctuations.
These functions are evaluated in Ref. [33], to define the scintillation index for a plane
wave, excluding inner scale effects, as:
( )
( ) ( )
1
69 . 0 1
51 . 0
11 . 1 1
49 . 0
exp
6 / 5
5 / 12
2
6 / 7
5 / 12
2
2
−
+
+
+
=
R
R
R
R
I
L
σ
σ
σ
σ
σ (4.13)
where
2
R
σ is the Rytov variance for a plane wave. Eq. (4.13) reduces to the Rytov
approximation in weak turbulence conditions, while it reduces to the Andrews
asymptotic model in saturated strong turbulence conditions.
4.2.2.4 Comparison of Strong Turbulence Models for the Plane Wave
The three irradiance variance models, along with weak turbulence theory
(
2 2
R I
σ σ = ), are plotted in Fig. 4.3. The Andrews and Churnside asymptotic models
follow a similar trend, with the Andrews asymptotic model matching the scintillation
index model when 36
2
>
R
σ [33]. In the weaktomoderate fluctuation regime with
2 5 . 0
2
< <
R
σ , the scintillation index model has a lower slope than what weak turbulence
theory would predict. It is well known that weak turbulence theory is most accurate
75
when 3 . 0
2 2
< =
I R
σ σ , and the scintillation index model accounts for the deviation of data
in this region from weak turbulence theory.
The coherence length is expected to follow asymptotic behavior in the saturation
region. When the irradiance variance
2
I
σ approaches 1, asymptotic theory predicts that
the transverse coherence length will become infinitely small. This would require the
presence of unreasonably high turbulence levels on the link. Fig. 4.4 shows the
behavior of the transverse coherence length based on three different models for the
irradiance variance in strong turbulence. The scintillation index model is plotted in the
onsetofstrong turbulence region, where 2 5 . 0
2
< <
R
σ . In the onset of strong
turbulence, ρ
0
is in the range of 2 mm to 6 mm for a path length of 863 m and λ = 632.8
nm.
Fig. 4.3: Irradiance variance, or scintillation index, of a plane wave versus Rytov
variance. Asymptotic models in the saturation region are shown.
76
Fig. 4.4: Transverse coherence length based on strong turbulence models plotted against
irradiance variance,
2
I
σ .
4.2.3 Spherical Wave Analysis
4.2.3.1 Churnside Asymptotic Analysis for the Spherical Wave
Aperture averaging theory for spherical wave propagation with small inner scale
is also evaluated using asymptotic theory as in Sec. 4.2.2.1. The spherical wave
analysis uses the same covariance function as presented in Eq. (4.1), although N
3
, b
1
,
and b
2
take on spherical wave representations. The transverse coherence length for the
spherical wave is defined by:
( )
5 / 3
2 2
0
545 . 0
−
=
n
LC k ρ (4.14)
77
The coefficient in Eq. (4.14) is the only difference between the plane and spherical
wave cases, otherwise the behavior of
0
ρ is the same. The constant N
3
also takes on a
new value of 3.86 for the spherical wave case. The two scale functions b
1
and b
2
are
also modified from the plane wave situation, although their behavior as ρ goes to zero
and infinity are the same. For negligible inner scale, the irradiance variance is related to
ρ
0
by [11]:
3 / 1
2
0 2
86 . 3 1


.

\

+ =
L
k
I
ρ
σ (4.15)
The irradiance variance is related to the plane wave Rytov variance
2
R
σ by using the
form presented in Eq. (2.36).
Churnside evaluated the strong turbulence spherical wave aperture averaging
factor as a sum of two terms which are strongly influenced by b
1
and b
2
in the
evaluation of the covariance equation [11]:
( )  
( ) ( )


.

\

− ×
−
−


.

\

=


.

\

− − −
+
=
∫ ∫
∫
∞
− −
−
3 / 5 3 / 5
3 / 5
0
0
1
0
3 / 1 2 2
1
3 / 2
2
2
3 / 7
0
2
1
0
3 / 5
0
2 / 1
2 1
2
2
1
2
exp
1
2
1 2
66 . 3
exp 1 cos
2
1 16
x u
kD
L
x dxx u J duu
kD
L
A
Dy
y y y dyy A
I
I
I
I
ρ
σ
σ
ρ
ρ σ
σ
π
(4.16)
As in Eq. (4.15), the term A
1
represents scale sizes smaller than ρ
0
, while the term A
2
represents scale sizes larger than the scattering disk, L/kρ
0
. The total aperture averaging
factor is A=A
1
+A
2
, and is given by [11]:
78
1
3 / 7
0
2
2
1
2
0
2
2
2
613 . 0 1
2
1
2
908 . 0 1
2
1
−
−

.

\

+
−
+


.

\

+
+
=
L
kD
D
A
I
I
I
I
ρ
σ
σ
ρ σ
σ
(4.17)
Eq. (4.17) is plotted for various values of irradiance variance in Fig. 4.5 below. The
separation of scale sizes for the spherical wave case, calculated from the width of the
plateau, is wider than in the plane wave case. For
2
I
σ = 1.1, the Fresnel zone size is
240ρ
0
, and the width of the plateau is beyond the range of the graph at 57512. When
2
I
σ = 1.25, the Fresnel zone size is smaller at 60.6ρ
0
, and the plateau width is 3681. For
2
I
σ =1.5, the Fresnel zone size is 21.4ρ
0
, and the plateau width is 460. As in the plane
wave case, larger irradiance variances show higher plateaus in A.
Fig. 4.5: Churnside approximate aperture averaging factor plotted against the ratio of
the aperture radius to the transverse coherence length, for
2
I
σ = 1.1, 1.25, and 1.5.
Curves calculated for an AVW test range path length of 863 m.
79
4.2.3.2 Andrews Asymptotic Analysis for the Spherical Wave
Andrews used the Kolmogorov spectrum to relate the irradiance variance in the
saturation region to the plane wave Rytov variance
2
R
σ [11]:
( )
5 / 2
2
2
73 . 2
1
R
I
σ
σ + = (4.19)
where
2
I
σ >> 1. This asymptotic model predicts a transverse coherence length that is
different from the Churnside asymptotic model by a factor of 2.74. This coherence
length is then used in the aperture averaging factor model given by Eq. (4.17). The shift
in the aperture averaging factor from its value given by the Churnside model is visible
in Fig. 4.6. The plateaus are also narrower than those of the Churnside model. The
Andrews asymptotic analysis is a good fit to the scintillation index model for a spherical
wave, which will be presented in Sec. 4.2.3.3.
Fig. 4.6: Andrews asymptotic model for the spherical wave plotted against the ratio of
the aperture radius to the transverse coherence length.
80
4.2.3.3 Scintillation Index (SI) Model for the Spherical Wave
The scintillation index model attempts to describe the behavior of irradiance
fluctuations over the entire range of turbulence strengths. The model for a spherical
wave is developed using the effective Kolmogorov spectrum, using the same filter
functions as presented in Eqs. (4.9) and (4.10) for the plane wave case. The small scale
and large scale log irradiance fluctuations are evaluated to define the scintillation index
of a spherical wave, neglecting inner scale effects, by [20]:
( )
( ) ( )
1
23 . 0 1
20 . 0
19 . 0 1
20 . 0
exp
6 / 5
5 / 12
2
6 / 7
5 / 12
2
2
−
+
+
+
=
R
R
R
R
I
L
σ
σ
σ
σ
σ (4.20)
where
2
R
σ is the plane wave Rytov variance.
4.2.3.4 Comparison of Strong Turbulence Models for the Spherical Wave
Both asymptotic models, along with the scintillation index model and weak
turbulence theory, are plotted in Fig. 4.7. As in the plane wave case, the asymptotic
models are similar, with the Andrews model matching the scintillation index model
when
2
R
σ > 64. In the onset of strong scintillation region, where 0.5 <
2
R
σ < 2, the
scintillation index model has a lower slope than that predicted by weak turbulence
theory.
81
Fig. 4.7: Spherical wave irradiance variance plotted as a function of plane wave Rytov
variance. Asymptotic and weak turbulence relations are indicated.
The transverse coherence length also behaves asymptotically in the saturation
region. Figure 4.8 shows that ρ
0
becomes infinitely small as 1
2
→
I
σ . The scintillation
index model predicts values of ρ
0
in the range of 1 mm to 1 cm. In the saturated region,
the scintillation index model follows the Andrews asymptotic model. The asymptotic
models predict that for strong saturation conditions, the transverse coherence length
becomes very small (0.1 mm to 0.001 mm) for path lengths on the order of the range
used in this experiment.
82
Fig. 4.8: Behavior of the transverse coherence length in the strong fluctuation region,
plotted as a function of the plane wave Rytov variance.
4.3 Aperture Averaging Experiment
4.3.1 Experimental setup
Strong turbulence aperture averaging data was collected over multiple days
along the 863 m AVWChesapeake test range. The only valid experimental method for
determining the aperture averaging factor in strong turbulence is to directly measure the
irradiance variance of a propagating wave [40]. Unlike the situation in weak
turbulence, there is no way of equating the irradiance variance to the logirradiance
variance in strong turbulence. On this test range, measured intensity variance values
ranged from 1.01 to 1.35; however, a sufficiently large data set is available only at
2
I
σ =
1.1.
83
The transmitter and receiver systems are the same as those used in the weak
turbulence aperture averaging experiment, presented in Section 3.3.1. Aperture
diameters ranging from 1 cm to 16 cm were used; however, due to atmospheric
variations during the data collection period, not all available aperture diameters
produced data in strong turbulence conditions.
Data was collected during clear weather conditions. Winds ranged from 1 m/s
to 4.5 m/s, with low relative humidity (less than 50%). Temperatures ranged from 60°F
to 83°F. Strong turbulence values were always measured in daylight.
4.3.2 LabVIEW Data Acquisition
Data acquisition for the strong turbulence case proceeded in the same manner as
for the weak turbulence case, which was presented in Section 3.3.2. The LabVIEW
program was shown in Fig. 3.7, and outputs the same 13 data values presented in Table
3.1. Received intensity was sampled in LabVIEW at 3000 samples/sec, with intensity
variance values averaged over 1 minute intervals. The 1 minute averaging time
allowed for more accurate measurements. Data is sorted according to the measured
irradiance variance of the scintillometer channel. Data is then grouped by irradiance
variance value and separated into two regions: weak irradiance fluctuations, 0 . 1
2
<
I
σ ,
and strong irradiance fluctuations, 0 . 1
2
>
I
σ .
4.4 Experimental Results for the Plane Wave Case
Data was analyzed by the process described in Section 4.2.1. The Fresnel zone
size is 9.32 mm, and the scattering disk is 8.69·10
5
/ρ
0
. To neglect inner scale effects,
0
ρ ≤
o
l and ρ
0
< (L/k)
1/2
. As in the weak turbulence case, we expect the spherical wave
84
analysis to be in better agreement with the theory, due to the divergence of the laser
beam.
New data is plotted in Fig. 4.9 using the Churnside asymptotic value for ρ
0
. The
data plotted has an average
2
I
σ ranging from 1.07 to 1.12. Therefore, the mean data
should follow the solid line for
2
I
σ = 1.1. Error bars represent one standard deviation of
the measured values. Since ρ
0
is calculated from measured values of irradiance
variance, it also has a range of values within one standard deviation of the mean. The
asymptotic theory curve seems to underpredict the aperture averaging factor, thereby
overestimating the amount of aperture averaging that is taking place in strong
turbulence. In the plateau region, near D/2ρ
0
= 350, the experimental aperture
averaging factor is 43% higher than the value predicted by the model.
Figure 4.9: Average aperture averaged data in strong turbulence using Churnside
asymptotic analysis plotted against the ratio of the aperture radius to the transverse
coherence length. Churnside asymptotic analysis predicts C
n
2
= 9.3×10
12
m
2/3
.
85
As mentioned previously, the direct measurement of the intensity to determine
2
I
σ is the only valid method for determining aperture averaging in strong turbulence
[40]. In the absence of independent measurements of C
n
2
by an nonsaturating
incoherent scintillometer [11,24], the asymptotic model presented in Sec. 4.2.2.1 was
used to extract values for
2
R
σ and ρ
0
. Typical values of ρ
0
in weak to nonsaturated
strong turbulence are on the order of 1 mm. Using measured data values the asymptotic
model predicts ρ
0
in the order of 0.01 mm to 0.1 mm for saturated strong turbulence,
along with values of C
n
2
averaging 9.3×10
12
m
2/3
. These values are consistent with
those expected experimentally, since very long propagation paths have very small
transverse coherence lengths [42].
New data is plotted in Fig. 4.10 using the Andrews asymptotic relation of Eq.
(4.7). The only difference between this data and the data plotted using the Churnside
asymptotic representation in Fig. 4.9 is that the data is leftshifted on the xaxis by a
factor of 1.52. There is a small constant offset between Eqs. (4.2) and (4.7), leading to
this shift. Once again, the data clustered near A = 0.06 appears to exhibit a plateau
effect. The Andrews asymptotic model gives ρ
0
= 0.39 mm and C
n
2
= 5.7×10
12
m
2/3
for this data.
86
Fig. 4.10: New data using the Andrews asymptotic theory plotted against the ratio of
the aperture radius to the transverse coherence length. The solid line represents the
aperture averaging factor given by the Andrews asymptotic theory for
2
I
σ = 1.1.
We previously introduced an expression for irradiance variance
2
I
σ based on
2
R
σ
that is valid over all turbulence conditions, with inner scale effects neglected. This is
the scintillation index (SI) model. In order to determine the transverse coherence length
from Eq. (4.3), we have interpolated values of
2
I
σ based on
2
R
σ for the rising portion of
the scintillation index curve shown in Fig. 4.3. Therefore, the data is analyzed
assuming strong turbulence without saturation, giving a value for ρ
0
that is consistent
with the conditions on the experimental test range. This new data is presented in Fig.
4.11, along with scintillation index model curves using saturated and nonsaturated
values of the Rytov variance for
2
I
σ = 1.1.
87
Fig. 4.11: New aperture averaging data calculated using the SI model plotted against
the ratio of aperture radius D/2 to the transverse coherence length ρ
0
for a plane wave
propagating in strong turbulence. The curves represent the aperture averaging factor
given saturated and nonsaturated strong turbulence values for
2
R
σ .
The new data in Fig. 4.11 is in better agreement with theory, and visibly follows
the predicted trend for 1 . 1
2
=
I
σ . Using the scintillation index model to find a Rytov
variance of 9 . 2
2
≈
R
σ when 1 . 1
2
=
I
σ , the transverse coherence length is found to be ρ
0
≈ 4.4 mm. The turbulence strength average is C
n
2
= 8.4×10
14
m
2/3
. Since our data was
not acquired under strong saturation conditions [42], our receive aperture on the
scintillometer channel with D = 5 mm is sufficient to avoid aperture averaging effects.
88
4.5 Experimental Results for the Spherical Wave Case
Strong turbulence data for the spherical wave case was acquired through the
process presented in Section 4.3. Using the models presented in Section 4.2.3, we can
compare measured aperture averaging data with that of the theoretical approximations
for the spherical wave case. The Fresnel zone size is 9.32 mm, while the typical
transverse coherence length was near 7 mm, indicating that we are within the strong
turbulence region.
New data with 1 . 1
2
≈
I
σ is shown in Fig. 4.12 using the Churnside asymptotic
model, presented in Section 4.2.3.1, to determine ρ
0
. The approximate formula from
Eq. (4.17) is also plotted in Fig. 4.11, for
2
I
σ = 1.1. The data is slightly higher than
what the approximate formula for 1 . 1
2
=
I
σ predicts. The Churnside asymptotic model
predicts ρ
0
≈ 0.04 mm and C
n
2
= 7.2 ×10
10
m
2/3
when 1 . 1
2
=
I
σ . The predicted C
n
2
indicates very strong turbulence because the asymptotic model assumes strong
saturation conditions that force the transverse coherence length to be a very small value,
on the order of 0.01 mm. The drawbacks of the asymptotic model are explained in
more detail later in this chapter.
The Andrews asymptotic relationship for the transverse coherence length was
used to plot the aperture averaging factor in Fig. 4.13 for 1 . 1
2
≈
I
σ . Eqs. (4.14) and
(4.19) were used to determine ρ
0
for this set of data. As with the Churnside asymptotic
analysis, the data is higher than what the asymptotic approximations predict. The
Andrews model predicts the transverse coherence length for this path to be on the order
of 0.1 mm. Andrews asymptotic analysis estimates the level of turbulence at C
n
2
= 1.2
×10
10
m
2/3
.
89
In the case of both asymptotic relations, the Churnside and Andrews asymptotic
models predict that A will always be much less than 1 for an aperture diameter of any
reasonable size. In order to measure A near 1, the receiver aperture diameter must be on
the order of the saturated value of ρ
0
, which ranges from 0.01 mm to 0.1 mm. The
residual value in Figs. 4.12 and 4.13 is that they clearly show that this data was not
taken in saturated strong turbulence conditions. The two data points located at A = 0.78
and A = 0.18 would not have been measured if the turbulence was saturated. The rest of
the data points indicate a plateau effect, although the mean of those five points is about
36% higher than the asymptotic model values at D/2ρ
0
= 2000. Recall that the
discrepancy between the Churnside and Andrews asymptotic models is that ρ
0
in the
Andrews model is a factor of 2.74 larger than that of the Churnside asymptotic model.
Fig. 4.12: New strong turbulence data analyzed using the Churnside asymptotic method
for a spherical wave, with irradiance variances near 1.1.
90
Fig. 4.13: New strong turbulence data plotted using the Andrews asymptotic model for
the spherical wave from Eq. (4.19). The curve represents the Andrews asymptotic
model aperture averaging factor for 1 . 1
2
=
I
σ .
To address the shortcomings of the asymptotic theories, the scintillation index
model for the irradiance variance in Eq. (4.20) was used to analyze the experimental
data in a nonsaturation strong turbulence region. The ρ
0
for 1 . 1
2
=
I
σ was near 7 mm,
putting us well within the strong turbulence region on this path. This corresponds to an
average C
n
2
= 8.4×10
14
m
2/3
. Fig. 4.14 shows the mean data plotted with the
approximation to the aperture averaging theory when 1 . 1
2
=
I
σ . The data appears to
show a tapering off and potentially a plateau area near D/2ρ
0
= 10, similar to that
predicted by the approximate curve. This is consistent with the turbulence working to
render scale sizes larger than ρ
0
ineffective along the path.
91
Fig. 4.14: Mean data with error bars plotted against the ratio of the aperture radius to
the transverse coherence length. The solid line is the aperture averaging approximation
in Eq. (4.17) for 1 . 1
2
=
I
σ .
The new data using the asymptotic and SI models have been plotted along with
the Churnside experimental data in Figs. 4.15 and 4.16. The Churnside data is the only
wellaccepted set of aperture averaging data in strong turbulence. The Churnside data
was taken by propagating a HeNe laser over a 1000 m path, with an average irradiance
variance of 3.08 ± 0.38. The Churnside asymptotic curve for 08 . 3
2
=
I
σ is shown in
Fig. 4.15 to contrast with the Churnside data. Data for six apertures was collected
simultaneously; and C
n
2
was measured over a 250 m path, adjacent to the 1000 m
aperture averaging path, using an incoherent scintillometer. The inability to measure
C
n
2
over the same path as A calls into question the true pathaveraged nature of the
turbulence data. The average C
n
2
measured was 1.29 ± 0.39 × 10
12
m
2/3
, and ρ
0
= 1.74
92
mm ± 0.24 mm. The Churnside data is limited in breadth, since Churnside did not have
large enough apertures available to measure very small aperture averaging factors. This
also limited their ability to see any scale size effects due to the size of turbulent eddies
relative to ρ
0
and the scattering disk.
In a condition of strong saturated turbulence, the new data analysis using
asymptotic theory in Fig. 4.15 results in data that is beyond the range of the Churnside
data. Although the analyzed data demonstrates a plateau in reasonable agreement with
approximate theory, the data should only be treated as an estimate of what could occur
if the link experienced strong saturated turbulence, following the discussion presented
earlier regarding Figs. 4.12 and 4.13. The new data reinforces the idea that it will be
difficult to surpass the plateau and reach the scattering disk scale in saturated
conditions; this would require an extremely large diameter receiver. By plotting the
aperture averaging factor on the plateau, an aperture size near the ρ
0
scale may be
chosen for a specific link, if it is required to work in saturated scintillation conditions.
As mentioned previously, the ρ
0
and scattering disk scale sizes result in a plateau
region in the aperture averaging factor for strong turbulence. The Churnside data does
not show any scale effects, while the new aperture averaging data using the SI model
(Fig. 4.16) begins to exhibit a tapering around D/2ρ
0
= 10, likely leading to a plateau.
The new data is in good agreement with the predicted aperture averaging curve, even at
higher values of A. The spherical wave treatment is also much improved over the plane
wave analysis of Section 4.4, which is due to the beam divergence of the transmitter.
Only the point at A = 0.188 and D/2ρ
0
= 4.32 in Fig. 4.16 does not include the
93
approximate theory for the aperture averaging factor within the error bars. This is due
to a limited number of data points at that aperture size.
Fig. 4.15: Mean data analyzed with the Churnside asymptotic model, plotted along with
the strong turbulence Churnside data taken over a 1000 m path.
The new data in Fig. 4.16 is shown with both the SI model curves for saturated
and nonsaturated strong turbulence at 1 . 1
2
=
I
σ . The new data is in good agreement
with both curves. Although Fig. 4.16 is beginning to show scale size effects, data for
larger apertures is needed to discern the validity of these results. Given that the amount
of strong turbulence data over this test range was limited, these results should encourage
further experimental studies of aperture averaging in strong turbulence in order to
improve the quality of theoretical models available in saturated and nonsaturated
strong atmospheric turbulence.
94
Fig. 4.16: Mean data analyzed using the scintillation index model in the nonsaturated
strong turbulence region, plotted along with the Churnside data for a 1000 m path.
Saturated and nonsaturated SI model curves for 1 . 1
2
=
I
σ are also shown.
4.5 Conclusions
Strong turbulence aperture averaging data, measured on an 863 m test range at
the University of Maryland, College Park, has been analyzed using the Churnside
Asymptotic, Andrews Asymptotic, and Scintillation Index models. The models are
used to determine values for ρ
0
and
2
R
σ , which otherwise could only be measured by
using an incoherent light scintillometer. The SI model shows the best aperture
averaging results for the spherical wave in strong turbulence. Due to the height of the
propagation path above ground, it was difficult to acquire enough data points in strong
turbulence conditions for analysis. The only large data set was available for 1 . 1
2
=
I
σ .
In spite of these limitations, this is the first set of strong turbulence aperture averaging
data to be acquired in nonsaturation conditions.
95
Chapter 5
Free Space Optical Communication using Nonimaging Optics
5.1 Introduction
Atmospheric turbulence adversely affects the propagation of optical waves. As
we have investigated in the previous chapters, the atmosphere distorts the wavefront
and produces intensity scintillations. Another manifestation of the interaction of
atmospheric turbulence with an optical wave is the receiver spot wander in the focal
plane. High speed optical receivers use photodetectors with diameters on the order of
10s of microns. In an onoff keyed transmission system, achieving a good bit error ratio
(BER) requires that a maximum amount of received light is received by the
photodetector. When turbulence induced beam wander moves the focal spot, there is a
definite increase in the BER. Commercial systems for data transmission typically
require a BER of 10
9
or better [43]. Minimizing beam wander in the focal plane of the
receiver is an important and relatively low cost step in maximizing link reliability.
5.2 TurbulenceInduced Beam Motion
5.2.1 Beam Wander
After a beam propagates through a length equal to a few transmitter diameters of
turbulent atmosphere, the beam begins to wander randomly in the plane transverse to
the propagation direction [44]. Beam wander is a result of the electromagnetic wave
interacting with turbulent eddies of sizes on the order of the transmit aperture diameter
96
[44]. Typically, the influential scale sizes are larger than the inner scale, ℓ
o
. Although
the beam wanders, it tends to maintain the shape it would have if the beam was
propagating through free space [45]. It has been shown experimentally that the time
constant of the beam wander is on the order of the ratio of the beam size to the wind
velocity. Previous work has proposed that a fast tracking transmitter be used to
minimize beam wander [20,45].
The motion of the centroid of the beam, r
c
, is defined by [20]:
2 2
2 / 1
2
ST LT c
W W r − = (5.1)
where W
ST
is the short term beam width, and W
LT
is the long term beam width, shown in
Fig. 5.1. Beam wander may be observed while propagating a visible laser (such as a
HeNe) at night.
Fig. 5.1: Beam wander at the receiver plane is characterized by the short term and long
term beam widths [adapted from Ref. 20].
5.2.2 Angle of Arrival and Image Dancing
Imaging dancing, also called beam wander in the focal plane, is related to
turbulenceinduced phase fluctuations on the propagating optical wave. The angleof
97
arrival is a result of a phase shift along the receiver lens diameter that is manifested as
an optical path difference. The angleofarrival, α, is defined by [20]:
D
S
D
l ∆
=
∆
= α (in radians) (5.2)
where ∆l is the optical path difference and ∆S is the phase shift across a receiver
aperture with diameter D. Using Eq. (5.2), the RMS image displacement, ∆s
RMS
, is [20]:
( ) f LD C f s
n RMS
2 / 1
3 / 1 2
2 / 1
2
91 . 2
−
= = ∆ α for
o
D l >> (5.3)
where f is the focal length of the optical receiver in meters. Using the specifications of
link under test and typical weak turbulence conditions, C
n
2
= 5×10
15
m
2/3
, L = 863 m, D
= 16 in or 40.6 cm, and f/10, the RMS image displacement is 16.47 µm. In testing high
speed (> 1 Gbps) transmission systems, lowcapacitance, small aperture photodetectors
with diameters of less than 100µm are used. Any image motion will result in the loss of
photons and will increase the BER.
5.3 BER in OnOff Keyed Systems
An onoff keyed (OOK) system is binary transmission protocol where 1s and 0s
are represented by the transmission of a large number of photons or the lack of photons
in a bit slot. The optical signal is created by either the direct modulation of a laser, or
the use of an external modulator to control the emission from a continuouswave (CW)
laser source. Depending on the power and subsequent amplification of the laser source,
the number of photons transmitted for a “1” may be upwards of 1,000,000,000 photons.
On an atmospheric link with a fade on the order of 40 dB during a “1” bit period, and a
transmitter power of 0 dBm = 1 mW, only 7800 photons may make it to the receiver.
98
For a pin photodetector operating at 100 Mbps with a minimum receive power of 42
dBm, a minimum of 4925 photons must make it to the photodetector surface. Loosing
photons due to beam wander in the focal plane is an unnecessary liability in an optical
wireless link.
Measuring the BER is the best way to characterize the performance of an optical
wireless communication system. The BER also accounts for the modulation format of
the transmission, which in this case is the OOK format. In an OOK system, the
probability of an error may be written as [20]:
) 1  0 Pr( ) 0  1 Pr( ) Pr(
1 0
p p error + = (5.4)
where p
0
is the probability that a “0” is transmitted, p
1
is the probability that a “1” is
transmitted, and p
0
+ p
1
= 1. When a pseudorandom sequence is transmitted, an equal
number of “1” and “0” bits are transmitted on average, implying that p
0
= p
1
= 0.5. The
BER of an OOK system is derived from the theoretical probability of a bit error
assuming a Gaussian noise distribution [20]:


.

\

=
N
S
i
erfc error
σ 2 2 2
1
) Pr( (5.5)
where i
S
is the signal current of one bit pulse, and σ
N
is the standard deviation of the
Gaussian noise distribution. The BER for an OOK system is [20,46]:
( ) dS
i
S SNR
S p OOK BER
S
I


.

\

=
∫
∞
2 2
erfc
2
1
) (
0
(5.6)
where SNR is the signaltonoise ratio of the transmission, and p
I
(S) is the Nakagami
gammagamma distribution with variables representing the contribution of smallscale,
γ, and largescale, ψ, components to the irradiance of the transmission, I = γ ψ [20,47]:
99
( )
( )
( )
( ) ( )


.

\



.

\

Γ Γ
=
−
− 
.

\
 +
+
S S S
I
i
S
K
i
S
i
S p
ψγ
γ ψ
ψγ
γ ψ
γ ψ
γ ψ
2
2
1
2 2 /
(5.7)
where K
ψγ
() is the modified Bessel function of the second kind. The gammagamma
distribution is the representative probability distribution function for a range of physical
phenomena, including the propagation of radio waves in the troposphere or ionosphere
[20]. The components of the irradiance in the presence of atmospheric turbulence may
be defined by [20,47]:
( )
( )
2 2
,
2
1
6 / 7
5 / 12
,
2
2
,
1
5 / 12
,
2 2
6 / 5
5 / 12
,
2
,
4065 . 0
4
1
56 . 0 18 . 0 1
49 . 0
exp
1
62 . 0 90 . 0 1
69 . 0 1 51 . 0
exp
R sph R
sph R
sph R
sph R
sph R sph R
L
kD
d
d
d d
σ σ
σ
σ
ψ
σ
σ σ
γ
=
=


.

\

−
+ +
=


.

\

−
+ +
+
=
−
−
−
(5.8)
By defining a new random variable
S
i
S
x = , and its derivative dx i dS
S
= , Eq. (5.6)
can be transformed into the BER of an OOK system in the presence of atmospheric
turbulence [47]:
( )
( ) ( )
( ) ( )dx x K x
x SNR
BER
ψγ
γ ψ
ψγ
γ ψ
γ ψ
γ ψ
2
2 2
erfc
1
2
0
2 / ) (
−
− 
.

\
 +
∞
+
∫


.

\

×
Γ Γ
=
(5.9)
Eq. (5.9) is plotted in Fig. 5.2 to demonstrate the relationship between the BER and
SNR for different values of irradiance variance. The three values of irradiance
fluctuation were chosen to represent weak 3 . 0
2
=
R
σ , intermediate 6 . 0
2
=
R
σ , and non
100
saturated strong turbulence 0 . 2
2
=
R
σ . Three more curves in Fig. 5.2 are plotted to show
the BER improvement for a 4 inch aperture along a 1.726 km link. For a BER of 10
9
,
there is a between a 4 dB and 15 dB improvement with aperture averaging across each
level of irradiance fluctuations.
Fig. 5.2: The BER of a communications link with a gammagamma probability
distribution function (PDF) as a function of signaltonoise ratio. The Rytov variances
plotted are 0.3, 0.6, and 2.0. When 1
4
2
=
L
kD
, there is no aperture averaging present
because the aperture radius is equal to the Fresnel zone size (FZ). Three more curves
are plotted for an aperture averaged case using a 4 inch diameter aperture. The
irradiance variances used in the calculation are the same in both cases.
5.4 Nonimaging Optics
To address the problem of beam wander in the focal plane, presented in Sec. 5.2,
we have chosen to integrate a nonimaging optical element into the optical receiver of an
optical wireless link.
101
5.4.1 Theory of the Compound Parabolic Concentrator
Roland Winston and W.T. Welford published the first book on nonimaging
optics in 1978 [48]. Initially, nonimaging optical elements were investigated for
application to the concentration of solar energy. In 1976, the design of the dielectric
filled Compound Parabolic Concentrator (CPC) for energy concentration was first
published [49]. The CPC is an extension of the basic cone concentrator. The cone
concentrator was limited in its performance, because some rays would be reflected back
through the entrance aperture of the cone. This behavior is indicated by the double
arrows in Fig. 5.3.
Fig. 5.3: The cone concentrator with cone angle γ and maximum entrance angle θ
i
.
(adapted from Ref. 48). The cone is hollow with reflective (metallic) edges.
The CPC uses surfaces traced out by an offaxis parabola to improve the
concentration ratio over that of the cone concentrator. Figure 5.4 shows a diagram of
the CPC. As in the case of the cone concentrator, θ
i
is the maximum entrance angle of
the CPC. This angle is related to the entrance and exit apertures by:
102
( )
in
out
i
a
a
= θ sin (5.10)
where a
out
is the radius of the exit (smaller) aperture, and a
in
is the radius of the entrance
(larger) aperture. The parabola has a focal length of:
( ) ( )
i out
a f θ sin 1+ = (5.11)
The design equations of the CPC in Cartesian coordinates are easily determined from
Fig. 5.4:
( )
( )
( )
( )
( )
( ) φ
θ φ
θ φ
φ
θ φ
θ φ
cos 1
cos 2
cos
cos 1
sin 2
sin
−
−
= − =
−
−
−
= − − =
i
i
out
i
out i
f
r z
a
f
a r y
(5.12)
where the parametric angle φ is the angle between the axis of the parabola and the
vector r. The parametric angle has minimum and maximum angles of
i
θ φ 2
min
= and
i
θ
π
φ + =
2
max
. The length, L, of the concentrator is found by calculating z for φ = φ
min
:
( )
( )
( ) ( ) ( )
( )
( ) ( )
i out in
i
i i out
i
i
a a
a f
L
θ
θ
θ θ
θ
θ
cot
sin
cos sin 1
2 cos 1
cos 2
2
+ =
+
=
−
=
(5.13)
The theoretical maximum concentration ratio for a 3dimensional CPC is:
( )
i
out
in
a
a
C θ
2
2
max
sin
−
=


.

\

= (5.14)
In fact, the maximum concentration ratio will depend on the reflectance of the
concentrator surface.
103
CPCs are especially useful for integration into FSO communications receivers;
they are easily designed, highly efficient, add minimal size and weight to the system,
and are self baffling. The self baffling nature of the CPC limits the amount of stray
light incident on the photodiode.
Fig. 5.4: The design of the CPC surface. The axis of the parabola and the CPC axis are
different. The surface of the CPC is traced out by moving the vector r through the
parametric angle φ.
5.4.2 The DielectricFilled CPC
Dielectricfilled CPCs take advantage of the principle of total internal reflection
to maximize the concentration ratio. Since total internal reflection is nearly 100%
efficient, the theoretical maximum concentration ratio becomes [48]:
104
( )
e out
in
dielectric
n
a
a
C
θ
2
2
2
max,
sin
=


.

\

= (5.15)
where n is the refractive index of the dielectric material, and θ
e
is the maximum angle of
incidence of an entrance ray on the dielectric surface. Figure 5.5 is a plot of θ
e
and the
internal maximum transmission angle, θ
n
. The angles θ
e
and θ
n
are related to n by:
( )
( )
2
/ 2 1 sin
/ 2 sin
n
n n
n
e
− ≤
− ≤
θ
θ
(5.16)
If n > 2, the angle θ
e
becomes imaginary. Since the CPC cannot accept an angle
larger than 90°, the maximum angle that is refracted into the dielectric is given by
Snell’s Law as ( ) n / 1 sin
1 −
.
Fig. 5.5: Plot of the maximum entrance angle in air, θ
e
, and the maximum internal ray
angle, θ
n
for a range of refractive indices.
105
The parametric equations for the dielectric CPC are the same as those used to
describe the hollow CPC, given by Eq. (5.11) through (5.13), with θ
i
replaced by the
maximum internal transmission angle of the CPC.
5.4.3 The CPCPhotodetector Combination
The coupling of a CPC to a light detection device is logical method of increasing
the probability that an optical signal is accurately detected. CPCs must obey the
brightness principle, which states that an image cannot be brighter than the incident
object [48,50,51]. When considering a CPC, the brightness theorem dictates that
radiation incident on the CPC over an angle of incidence θ
i
will exit the CPC
maintaining the same flux per unit area; therefore, the radiation will exit the CPC over a
range of angles larger than θ
i
.
Using the brightness principle, a relation is found between the entrance aperture
of the CPC and the photodetector area required to achieve maximum efficiency [50,51]:
( )
2
2
sin
n
A A
i
i PD
θ
= (5.17)
where A
PD
is the area of a photodetector and A
i
is the area of the entrance aperture of the
CPC. In order to minimize the detection noise and maximize the bit rate of the system,
the smallest photodetector allowed by Eq. (5.17) should be used. The maximum
collection efficiency expected by Eq. (5.17) will be less than 100%, since Fresnel losses
may exist at the entrance and exit surfaces of the CPC.
Alternative designs of nonimaging concentrators have been proposed, including
the θ
i
/θ
o
concentrator [48], and RX and RXI dielectric concentrators [50,51]. The
106
former design restricts the exit angle by using a cone section at the exit aperture. The
latter designs enhance the uniformity of the output irradiance distribution.
5.5 1.7 km CPC Link Experiment
The hypothesis of a BER improvement when using a CPC in the optical receiver
of an optical wireless link was tested on a 1.7 km range at the University of Maryland,
College Park.
5.5.1 Experimental Setup and Test Range
The test range used for the CPC link experiment is the same range that was used
for aperture averaging experiments. The addition of a hollow, cornercube
retroreflector (CCR) allowed the link range to be doubled to 1.726 km. A schematic of
the link experiment is shown in Fig. 5.6. The bit error ratio tester (BERT) generates a
pseudorandom binary sequence (2
7
1), which modulates the laser diode. The output of
the fibercoupled laser diode is transmitted to the erbium doped fiber amplifier (EDFA),
which amplifies the signal to approximately 158 mW. After the signal exits the optical
fiber, the optical beam is transmitted by a beam expansion system to the CCR
downrange. The beam expander contains a 10 x magnifying lens and a 2.25 in.
biconvex lens. The 5 in. cornercube retroreflector is goldplated for maximum infrared
reflectance. Since the retroreflector has a narrow divergence angle of 1 mrad, the
transmitting lens system must be placed in front of the center of the 16 in. Meade
SchmidtCassegrain telescope. Almost fortunately, part of the primary aperture on the
telescope is obstructed by the secondary mirror, so the amount of obstruction due to the
transmitter lens system is minimized.
107
After a portion of the beam is reflected by the CCR, the reflected wave expands
and is incident on the 16 in. telescope. The Meade is an f/10 telescope, although the
focus is adjustable. A ZnSe dielectricfilled CPC with a 1 cm entrance aperture is
placed in front of the focal plane of the telescope to maximize collection efficiency.
Details of the ZnSe CPC are discussed in the following section. The exit aperture (1
mm diameter) of the CPC is buttcoupled to a New Focus DC  125 Mhz InGaAs
photodiode (PD), Model 1611. The photodiode signal is amplified through two RF
amplifiers and subsequently read by the BERT receiver. The BERT is an Anritsu
ME522, which can transmit data at rates from 50 Mbps to 700 Mbps. An external
multiplexer/demultiplexer will increase the limit to 1.4 Gbps. The BERT measurement
period is variable from 1 second to hours.
EDFA TX
Laser,
Power Supply
CPC
CLOCK DATA
IR Camera
PD
Errors
To CCR
CPC
BERT
16″ Meade
TX Lens
Amplifier
EDFA TX
Laser,
Power Supply
CPC
CLOCK DATA
IR Camera
PD
Errors
To CCR
CPC
BERT
16″ Meade
TX Lens
Amplifier
Fig. 5.6: Setup of the CPC link experiment. Black lines are electrical connections;
orange lines are fiber optic connections. The IR camera is used for alignment purposes
only, and is lowered below the field of view of the 16 in. Meade telescope when not in
use.
108
5.5.2 ZnSe CPC
The CPC used in this experiment was made of ZnSe, due to its high
transmission at 1.55 µm. The refractive index of ZnSe at 1.55 µm is 2.48. The CPC is
approximately 2.25 cm long, with an entrance diameter of 1 cm and an exit diameter of
1 mm. A profile of the CPC is shown in Fig. 5.7, with a canned photodiode centered at
the exit aperture of the CPC. The CPC has a maximum entrance angle of ~5°. At
normal incidence, 18% of the light is reflected at each aperture.
For the 1.7 km experiment, the photodetector integrated into the package was
removed. The New Focus InGaAs photodetector was used in its place.
ZnSe CPC
InGaAs Photodiode
Al mounting tube
ZnSe CPC
InGaAs Photodiode
Al mounting tube
ZnSe CPC
InGaAs Photodiode
Al mounting tube
Fig. 5.7: Profile of the mounted ZnSe CPC directly coupled to an InGaAs PD.
5.5.3 Link Budget
The link budget is a design method to ensure that enough power will reach the
optical receiver given that there will be some losses on the optical link. It may be
written in the form [52]:
s L RX TX
M C P P + + = (5.18)
109
where P
TX
is the average transmitter power, P
RX
is the receiver sensitivity in dBm, C
L
represents the channel losses, and M
s
is the system margin, which represents the
maximum additional loss the system can handle. The link budget is normally written
with the optical power in dBm, where ( ) mW 1 / 10log dBm 1
10
P = , and the electrical
power in dB. The link budget of the 1.7 km link, without the CPC, is articulated in
Table 5.1.
The system margin is given by:
dB 2 . 28
dB 5 . 38 dBm 3 . 44 dBm 22
=
− + =
− − =
L RX TX s
C P P M
(5.19)
Although this system margin is sufficient to maintain a BER of 10
9
in a digital
communications system, where an electronic noise margin of 21 dB is required [20,
46,53], it does not account for fading on the optical wireless link. A fade is a sudden
drop in the signal strength that may result in the erroneous detection of a bit. Fading on
optical wireless links is induced by the motion of turbulent eddies in the atmosphere.
Fades of 10dB are common, although fades of over 40 dB have been reported [54]. Fig.
5.2 shows that for a BER of 10
9
in weak turbulence ( 3 . 0
2
=
R
σ ), a SNR of 17 dB is
needed. This is a clear example of how receivers integrated with nonimaging
concentrators should improve the link margin without placing any additional size,
weight, or power burdens on the system.
110
Link component Component specifications Power (in dB or dBm)
1.55 µm laser:
Force, Inc.
model 2666A
Max. launch power 0.15 mW,
extinction ratio 10:1
8 dBm
Optigain EDFA
model 2000
Max. gain 30 dB
Max. output 23 dBm
30 dB
P
TX
Transmitter power = 158 mW 22 dB
Fiber splice 2 splices @ 0.15 dB 0.3 dB
Fiber connectors 2 connectors @ 0.5 dB 1 dB
Viracon thermal pane
window in
Chesapeake Bldg.
Average 25% transmission for
infrared wavelengths
6 dB
Atmospheric
transmission
Assume an attenuation of 0.2
dB/km for clear air [55]
0.75 dB
Retroreflector Loss due to mismatch of beam
width and CCR diameter (129mm)
11.7 dB
Surface reflectivity 0.175 dB
Atmospheric
transmission
0.75 dB
Window loss 6 dB
Meade LX 200 EMC Loss due to mismatch of beam
width and 16 in. aperture
6.54 dB
Loss due to secondary mirror
obstruction of SchmidtCassegrain
design telescope
0.446 dB
Telescope transmission at 1.55µm
with EMC coatings
1.25 dB
Coupling loss Coupling of focal spot to
photodetector
0.5 dB
Total channel loss, C
L
38.5 dB
P
RX
New Focus 1611
InGaAs photodetector
Receiver sensitivity
NEP = 2.5pW/(Hz)
1/2
44.3 dBm
Table 5.1: Link budget analysis for the 1.7 km link, neglecting atmospheric turbulence.
5.5.4 C
n
2
Measurements
The strength of turbulence, C
n
2
, is measured over the same time interval as the
BER. Measurements over 1 minute intervals gave the most reliable results. The C
n
2
111
was measured over the scintillometer channel used in the aperture averaging
experiment.
5.5.5 Experimental Results and Analysis
A 100 Mbps pseudorandom binary sequence with a pattern length of 2
7
1 bits
was transmitted downrange and reflected by the corner cube retroreflector. Test runs
were done with and without a ZnSe CPC placed directly in front of the New Focus 1611
InGaAs photodetector. The exit aperture of the CPC was in direct contact with the
photodetector window in order to achieve maximum optical coupling. This coupling is
especially important since there is a large area mismatch between the CPC exit aperture
(1 mm diameter) and PD (0.3 mm diameter). The coupling of the CPC and
photodetector is very angle sensitive, and the CPC was placed on a 6 degreeoffreedom
stage to attempt to maximize light output. The PD was mounted on an xyz stage. The
actual PD surface is set back 0.5 mm from the lens case. A significant loss of coupling
was observed when there was more than 2 mm of air space between the PD and the
CPC.
Data was taken over two consecutive days in weak turbulence conditions
( ) 3 . 0
2
<
I
σ . BER for 100 Mbps PRBS transmission with and without the CPC is plotted
against the measured C
n
2
in Fig. 5.8, with error bars representing the fluctuation of C
n
2
over a 1 min. acquisition period. Fig. 5.9 plots a first order linear fit along with the
mean data. The curve fit equation is also shown in the plot.
From Figs. 5.8 and 5.9, the measured BERs were found to be on the order of
10
3
to 10
4
. The range of BER measurements indicates that the beam may not have
been as collimated as was thought. Since the system operates at 1.55 µm, the width of
112
the beam when it reaches the 5 inch retroreflector is extremely difficult to measure.
Though the BERs on the FSO link are high, the important result is the relative
enhancement of the BER when the optical communications receiver includes a CPC.
Fig. 5.8: BER plotted as a function of C
n
2
for the 1.7 km link with and without the CPC
integrated into the optical receiver. The error bars in C
n
2
represent its fluctuation over
the 1 min averaging interval.
113
Fig. 5.9: Same data as in Fig. 5.8, plotted along with a first order linear fit to the mean
data.
There is approximately a factor of 5 improvement in the BER when using the
CPC in the optical receiver. This enhancement is particularly encouraging, considering
that only about 11% of the light that exits the CPC is collected by the PD, due to the
area mismatch. A CPCPD combination where the PD area is either the same or
slightly larger than the exit aperture of the CPC should yield a better BER enhancement.
Also, the CPC and PD were mounted on a tripod, which caused the signal to be
sensitive to vibrations of the floor. Since the Meade telescope is in its own, very large,
heavy tripod, it does not feel the vibrations. A telescope with the CPC and PD using the
same optomechanical mount would also improve the performance by making the
system immune to vibrations.
114
Fig. 5.10 is an oscilloscope trace of the amplified received data, before and after
passing the signal through a limiting amplifier. The second trace in Fig. 5.10 is actually
the inverse of the data, which is connected to the oscilloscope so that each of the
limiting amplifier terminals are properly terminated. Nonetheless, it shows how the
limiting amplifier attempts to clean up the turbulence induced fluctuations of the data,
but still leaves residual noise on the data that it cannot clean up. In using the BERT, the
user must be careful that the terminals of the BERT are not overloaded; otherwise there
could be significant damage to the system. The turbulenceinduced optical fluctuations
of the receive signal affect the quality of the electrical signal output of the RF
amplifiers. The amplified signal even showed some baseline wander due to the
fluctuations in the intensity of the received signal. The amplified signal was attenuated
to between 30 mV
pp
and 500 mV
pp
so that any sudden spikes in intensity would not
overload the limiting amplifier. The limiting amplifier then outputs a 1.2 V
pp
signal to
the BERT.
115
Fig. 5.10: Trace 1 shows the data stream after passing through two 20 dB RF amplifiers.
Trace 2 shows the inverted data after passing through a limiting amplifier. The signal
after the limiting amplifier is much cleaner; however there is still enough turbulence
induced noise to degrade the BER.
5.6 NRL Test Range Experiment over the Chesapeake Bay
An experiment to demonstrate the BER enhancement due to the use of a CPC
PD combination in an optical receiver was conducted at the Naval Research
Laboratory’s Chesapeake Bay Detachment, in Maryland (NRLCBD).
5.6.1 Experimental Setup and Test Range
The Naval Research Laboratory has a facility located on the Chesapeake Bay in
Maryland that they have set up for use as a laser communications test site. They have
previously used the facility to test their modulating retroreflector (MRR) array
technology [56,57]. To test high speed laser communications, a solid retroreflector
array is located on a tower at Tilghman, MD, and can be interchanged with an MRR
116
array. The Tilghman site is approximately 16.2 km across the bay from NRLCBD.
Fig. 5.11 is a diagram of the test range.
Fig. 5.11: The NRLCBD laser communication test range [56].
The retroreflector array at the Tilghman tower is composed of 12 solid
retroreflectors, aligned three wide by four tall, spaced on 5 inch centers. Each
retroreflector is 2 inches in diameter. The optical transmitter uses a fibercoupled diode
laser at 1.55 µm that is subsequently amplified to 2 W. The signal exiting the fiber is
expanded and collimated, and sent downrange to the retroreflector array. Fig. 5.12 is a
diagram of the optical transmitter. A BERT pattern generator (Agilent model 86130A)
modulates the fibercoupled diode with a 2
7
1 PRBS pattern.
Fig. 5.12: Optical transmitter at NRLCBD. The collimated optical output is
approximately 2 W [56].
When the retroreflected signal returns, it is collected by a 16 in. Meade LX200
SchmidtCassegrain telescope. This is the same telescope used in the 1.7 km link at the
University of Maryland. It is estimated that 1.84 mW of the 2 W launch power is
117
collected by the 16 in. receiver telescope [56]. Since the signal strength is very low
after traversing a 32 km roundtrip, the signal is amplified with a Femto variable gain
voltage amplifier that can amplify a signal by up to 60 dB (model DHPVA100). The
signal is then detected by the BERT receiver. A description of the optical receiver is
shown in Fig. 5.13.
Fig. 5.13: The optical receiver at NRLCBD [56].
5.6.2 CPCPD Combination System
The integrated CPCphotodiode optical receiver, shown in Fig. 5.7, is used as
the detection system in the NRLCBD experiment. The detector integrated into the
package is a Telecom Devices 35PD500TO InGaAs photodiode with a 0.5 mm
diameter. The distance between the TO46 package window and the photodiode surface
is 0.38 mm. At 1.5 µm, the photodiode has a responsivity of 1.0 A/W. The back end of
the aluminum mounting tube has bond pads and short leads that connect the photodiode
signal to a Maxim 3266 low noise transimpedance preamplifier chip.
The input sensitivity of the preamplifier for a BER of 10
9
is:
( )
( )
( )
( )
dBm 38 . 28
1000
1  10 1.0A/W 2
1 10 nA 200 89 . 11
log 10
1000
1 2
1
log 10
− =


.

\

×
×
+ × ×
=


.

\

×
− ℜ
+ ×
=
ex
ex N
r
r I OSNR
y Sensitivit
(19)
118
where OSNR is the optical SNR required to achieve a BER of 10
9
, I
N
is the input
referred RMS noise current of the Maxim 3266, r
ex
is the extinction ratio of the laser
transmitter (defined here by the ratio P
1
/P
0
) [58,59], and ℜis the responsivity of the
photodiode in A/W. The optical power in a “1” bit and “0” bit may also be calculated.
When using a PRBS transmission pattern, the optical power in a “1” bit is approximated
by twice the average power [43]. Given a 10:1 extinction ratio and assuming a 2W (33
dBm) average power, the optical power in a “0” bit is 10 dB lower than that of a “1” bit:
mW 00 4 W 4 1 . 0 P
W 4
0
1
= × =
= P
(20)
where P
0
is the power in a “0” bit, and P
1
is the power in a “1” bit.
5.6.3 Experimental Results and Analysis
BER measurements were taken over the 32.4 km test range at NRLCBD. The
CPCPD receiver was used in place of a 62.5 µm optical fiber coupled receiver.
Measurements were made at standard telecommunication data transmission rates of OC
1 (51.84 Mbps), OC2 (103.68 Mbps), and OC3 (155.52 Mbps). Data was taken over a
4 hour window with an atmospheric temperature range of 2.4° to 3.5°, a water
temperature range of 5.8° to 5.9°, and an average windspeed ranging from 1.0 m/s to
2.6 m/s. These temperature and windspeed measurements were taken from a NOAA
buoy in proximity to the NRLCBD test range at Thomas Point, MD. The experiment
was attempted twice prior to the successful test, but had to be cancelled because of the
amount of turbulence present on the link. The turbulence made it impossible to align
the test link using LED beacons on the tower at Tilghman and a position sensitive
detector located at NRLCBD.
119
From Eq. (5.3), the beam wander in the focal plane is estimated at ∆s
RMS
= 339.6
µm for a C
n
2
= 1×10
13
m
2/3
. A high value of C
n
2
is assumed because link must be in
the saturated strong turbulence range, due to the extremely long propagation length.
Data collected at OC1 is shown in Fig. 5.14. The overall BER is plotted in 1
min intervals, along with the “0” error rate and “1” error rate. The Agilent BERT has
the ability to count “0” errors and “1” errors. Typically, we expect that “0” errors are
much lower than “1” errors because the power in a “0” bit should be below the
threshold of the BERT. For the OC1 test, the “0” error rate is lower than the “1” error
rate, although not substantially lower.
Fig. 5.14: Measurements of BER using the CPCPD receiver when an OC1 PRBS
transmission pattern was tested during a 35 minutes test period. The BER of 1 bit in a
51.84 Mbps transmission is 3.215×10
10
, as indicated by the dark horizontal line. There
were no errors between minutes 1 and 19. The only errors were detected in the 31
st
minute.
120
BER measurements taken at OC2 are shown in Fig. 5.15. In the 11
th
minute,
the “1” error rate is much higher than the “0” error rate. Fig. 5.16 shows OC2 BER
data published in an NRL paper, with a 62.5 µm optical fiber coupled receiver in place
of the CPCPD receiver. The signal is measured by an OC12 receiver, regenerated and
collected by an OC48 receiver, and finally received by the Agilent BERT.
Fig. 5.15: BER at OC2 over a 12 minute acquisition period using the CPCPD receiver.
A single bit error in a 103.68 Mbps transmission at OC2 gives a BER of 1.607×10
10
.
121
Fig. 5.16: BER at 100Mbps measured over the NRLCBD test range with an optical
fiber coupled receiver, from Ref. 46. Note that data is plotted for 5 sec intervals. The
minimum BER for a 5 sec interval is 2×10
9
, which is beyond the range of the graph.
A comparison of the data from the CPCPD receiver (Fig. 5.15) and the NRL
CBD optical fiber coupled receiver (Fig. 5.16) shows a significant improvement in BER
when using the CPC. The long link length contributes to the amount of beam wander in
the focal plane. In Fig. 5.16, there is not one acquisition interval that is errorfree. The
problem of beam wander in the focal plane is visibly reduced, and manifested in the
BER improvement.
BER results from tests with the highest transmission rate, OC3, are shown in
Fig. 5.17. The first 4 minutes of acquisition are error free. The following 11 minutes
show significant errors. This behavior is indicative of transmitter alignment problems
and not problems with the CPCPD receiver. There is no autotracking between the
122
transmitter and receiver on the link. There was also an evening haze rolling in during
this acquisition time, which could affect alignment over such a long range.
The first 4 minutes of error free acquisition are extremely encouraging. Fig.
5.18 shows published results from the link with an optical fiber coupled receiver for
data transmitted at 200 Mbps. A direct comparison with OC3 data is not available.
Fig. 5.18 shows a consistent and significant number of errors throughout the acquisition
period. Compared with the new data in Fig. 5.17, it appears that the addition of a CPC
into the receiver system would improve the BER. New data could not be acquired at
200 Mbps, due to the bandwidth range of the Femto variable gain amplifier. The device
had a bandwidth of 100 MHz, and the receive signal intensity at 200 Mbps fell off
significantly.
Overall, each of the three data rates tested showed a significant amount of error
free time. Although this CPCPD receiver was not optimized for the link, nor was the
optical coupling between the CPC exit aperture and PD surface optimized, the device
performed extremely well.
123
Fig. 5.17: BER at OC3 measured over a 15 minute period using the CPCPD receiver.
The single error BER for a 1 min acquisition period is 1.072×10
10
.
Fig. 5.18: BER at 200Mbps measured over the NRLCBD test range with an optical
fiber coupled receiver, from Ref. 46. Note that data is plotted for 5 sec intervals. The
minimum BER for a 5 sec interval is 1×10
9
, which is beyond the range of the graph.
124
5.7 Conclusions
The experimental results presented in this chapter show the performance
enhancement in BER due to the integration of nonimaging optical elements into a FSO
optical receiver. The data shows a factor of 5 improvement in BER when a compound
parabolic concentrator was integrated into the optical receiver system of a 100 Mbps
link over a 1.7 km retroreflected test range at the University of Maryland, College Park.
An integrated CPCphotodiode device was tested at the Naval Research Laboratory’s
32.4 km Chesapeake Bay test range. The link showed only rare periods of burst errors
during this experiment, as compared to significant, sustained errors without the use of
the device. Both experiments validate the ability of the CPC to significantly reduce the
number of data transmission errors due to beam wander in the focal plane of the
receiver. Optimized CPCphotodiode combinations will improve on the results
presented here.
125
Chapter 6
Conclusion
6.1 Summary of Contributions
This dissertation describes two techniques that will enhance the performance of
free space optical communication systems, while minimizing the size, weight, and
power of the system. Through experimental studies and comparison with available
models and theory, onoff keyed FSO communication systems will show a significant
performance improvement in data transmission when aperture averaging methods and
nonimaging optical elements are incorporated into the system design. Both studies
show the ability to mitigate the effects of atmospheric turbulence on the propagating
coherent optical wave.
Aperture averaging techniques specifically aim to reduce the atmospheric
turbulenceinduced intensity fluctuations that appear on a propagating optical
wavefront. Experimental studies of aperture averaging were conducted in both weak
and strong fluctuation conditions over a test range at the University of Maryland,
College Park. Due to the technical complexities involved in designing an aperture
averaging experiment, only one set of published data to this point has been available to
compare with analytical models. The experiment presented in this dissertation exceeds
the scope of the previously published data. The new data presented here represents
significantly improved measurements in both weak turbulence and strong turbulence
conditions due to: the examination of a wider range of aperture diameters; the use of
126
better quality electronics, computers, and data acquisition equipment; the measurement
of C
n
2
over the entire length of the aperture averaging test range; the simultaneous
measurement of background light levels with aperture averaging data; and the use of an
elevated test range allowed for minimal fetch influences. This is also the first report of
aperture averaging data collected for nonsaturated strong turbulence conditions. Since
the behavior of the refractive index spectrum in the energy dissipation range is still in
contention, strong turbulence theory is still questionable. This data will be useful in
improving the quality of strong turbulence theory and models available for the study of
aperture averaging in strong fluctuation conditions.
Compound parabolic concentrators integrated into optical wireless receivers are
shown to counteract the bit error ratio degradation attributed to atmospheric turbulence
induced beam wander in the focal plane of the receiver. This dissertation is the first
report of an experimental characterization of the performance enhancement due to the
integration of the CPC into the receiver. Promising BER improvements by a factor of 5
resulted from the testing of 100 Mbps data in weak turbulence conditions on the 1.7 km
link at the University of Maryland, College Park. Due to the small link margin and
difficulty in link alignment, PRBS bit error ratio tests were not achieved at rates beyond
100 Mbps. This experiment is supplemented by BER testing using a CPCphotodiode
combination receiver at the Naval Research Laboratory’s Chesapeake Bay Detachment.
The 32.4 km retroreflected link always experiences saturated strong turbulence
conditions. By comparison with previously published NRL data, the CPCPD receiver
showed only minimal burst errors over the testing time, at optical communication
standard rates of OC1 through OC3. The performance enhancements demonstrated in
127
this dissertation, along with the compact size of the device, should make the CPC a
mandatory part of any OOK FSO communication receiver. By designing the CPC in
such a way that its exit aperture maximizes concentration of optical energy onto the
surface of a photodiode, data transmission errors due to beam wander in the focal plane
could be absolutely canceled.
6.2 Future Work
This dissertation has begun to address perhaps the most significant problem
plaguing the design of FSO systems: the lack of experimental data pertaining to the
effects of atmospheric turbulence on the propagation of coherent optical waves. The
experimental data presented here provides valuable insights into how FSO receivers
may be optimized to improve the transmission of OOK systems. Although the data
spans both weak and strong turbulence conditions, more data in strong turbulence
conditions should be gathered to complement the data presented here.
Due to the height of the optical test range at the University of Maryland, only
limited data was collected for aperture averaging in strong turbulence. The testing of
smaller and larger receiver apertures on a new range that consistently experiences
saturated strong turbulence conditions would also be beneficial.
Regarding the CPC experiments, an optimized CPCPD device should allow the
experimental determination of the extent of a BER performance enhancement. A new
electronic preamplifier and receiver would have to be designed to test the device at data
rates beyond 1 Gbps. It would certainly be interesting to characterize the maximum
OOK transmission rate the CPC could handle.
128
Finally, the data presented here should be studied to improve the quality of the
refractive index spectrums designed for strong turbulence. Although the Hill spectrum
has been shown to be the most accurate in strong turbulence conditions, the fact that it
needs to be solved numerically makes it difficult to use in practical situations.
Approximations to the Hill spectrum, like the modified and effective atmospheric
spectrums, still leave room for improvement. The new aperture averaging data
presented here is valuable because it directly measured the irradiance variances in
strong turbulence. In the past, spectrums have been derived from logirradiance
variance statistics, which do not address the practical measurements of the irradiance
variances in strong turbulence conditions. Theoretical researchers need to address the
refractive index spectrum in strong turbulence by studying experimental data collected
for the strong turbulence irradiance variance.
129
References
[1] Strobehn, J.W. (ed.). Topics in Applied Physics, Vol. 25, Laser Beam
Propagation in the Atmosphere. SpringerVerlag: New York, 1978.
[2] Lucy, R.F. and K. Lang. “Optical communications experiments at 6328 Å and
10.6 µm.” Appl. Opt., Vol. 7, No. 10, Oct. 1968, p. 196570.
[3] Lipsett, M.S. et al. “Space instrumentation for laser communications.” IEEE J.
Quant. Elec., Vol. 5, No. 6, June 1969, p. 3489.
[4] Solimeno, S. et al. “Optical communication in turbulent media.” Proceedings of
the 1970 International Symposium on Information Theory. IEEE: New York,
1970.
[5] Masud, Sam et al. “Ten Hottest Technologies.” Telecommunications Magazine,
May 2001.
[6] Li, Tiffany Jing and Murat Uysal. “Achievable Information Rate for Outdoor
Free Space Optical Communication with Intensity Modulation and Direct
Detection.” Proc. GLOBECOMM, Vol. 2, 2003, p. 26548.
[7] PolakDingles, P. et al. “Performance update of a terrestrial laser
communications link with quadrature amplitude modulation.” Proc. SPIE, Vol.
2990, Free Space Laser Communication Technologies IX, April 1997, p. 8293.
[8] Ohtsuki, Tomoaki. “Performance Analysis of Atmospheric Optical PPM
CDMA Systems.” J. Lightwave Technol., Vol. 21, No. 2, February 2003, p. 406
11.
130
[9] Zhu, Xiaomihng and J. M. Kahn. “Performance Bounds for Coded FreeSpace
Optical Communications Through Atmospheric Turbulence Channels.” IEEE
Trans. Commun., Vol. 51, No. 8, August 2003, p. 12339.
[10] Fried, D.L. “Aperture Averaging of Scintillation.” J. Opt. Soc. Am., Vol. 57, No.
2, February 1967, p. 16975.
[11] Churnside, James H. “Aperture averaging of optical scintillations in the
turbulent atmosphere.” Appl. Opt., Vol. 30, No. 15, 20 May 1991, p. 198294.
[12] Vinnichenko, N.K. et al. Turbulence in the Free Atmosphere. Consultants
Bureau: New York, 1980.
[13] Wheelon, Albert D. Electromagnetic Scintillation, Vol.1, Geometrical Optics.
Cambridge University Press: Cambridge, 2001.
[14] Andrews, David G. An Introduction to Atmospheric Physics. Cambridge
University Press: Cambridge, 2000.
[15] Panofsky, Hans A. and John A. Dutton. Atmospheric Turbulence: Models and
Methods for Engineering Applications. John Wiley & Sons: New York, 1984.
[16] Andrews, Larry C. and Ronald L. Philips. Laser Beam Propagation through
Random Media. SPIE Optical Engineering Press: Bellingham, Washington,
1998.
[17] Lawrence, Robert S. and John W. Strobehn. “A Survey of ClearAir Propagation
Effects Relevant to Optical Communications.” Proc. IEEE, Vol. 58, No. 10,
October 1970, p. 152344.
[18] von Karman, Theodore. “Progress in the Statistical Theory of Turbulence.”
Proc. NAS, Vol. 34, November 1948, p. 5309.
131
[19] Hill, R.J. and S.F. Clifford. “Modified spectrum of atmospheric temperature
fluctuations and its application to optical propagation.” J. Opt. Soc. Am., Vol.
68, No. 7, 1978, p. 8929.
[20] Andrews, Larry C., et al. Laser Beam Scintillation with Applications. SPIE
Press: Bellingham, Washington, 2001.
[21] Wheelon, Albert D. Electromagnetic Scintillation, Vol. 2, Weak Scattering.
Cambridge University Press: Cambridge, 2003.
[22] Sasiela, Richard J. (ed.). Electromagnetic Wave Propagation in Turbulence.
SpringerVerlag: Berlin, 1994.
[23] Goodman, J.W. Statistical Optics. Wiley: New York, 2000.
[24] Churnside, James H. “ApertureAveraging Factor for Optical Propagation
Through the Turbulent Atmosphere.” NOAA Technical Memorandum ERL
WPL188, November 1990. (available from the National Technical Information
Service <http://www.ntis.gov>).
[25] Fried, D.L. “Optical Resolution Through a Randomly Inhomogeneous Medium
for Very Long and Very Short Exposures.” J. Opt. Soc. Am., Vol. 56, No. 10,
October 1966, p. 13729.
[26] Gracheva, M.E. and A.S. Gurvich. “Strong fluctuations in the intensity of light
propagated through the atmosphere close to the Earth.” Sov. Radiophys., Vol. 8,
No. 4, 1965, p. 5115.
[27] Prokhorov, A.M. et al. “Laser Irradiance Propagation in Turbulent Media.”
Proc. IEEE, Vol. 63, No. 5, May 1975, p. 790811.
132
[28] Coles, W.A. and R.G. Frehlich. “Simultaneous measurements of angular
scattering and intensity scintillation in the atmosphere.” J. Opt. Soc. Am., Vol.
72, No. 8, August 1982, p. 10428.
[29] Fante, Ronald L. “Innerscale size effect on the scintillations of light in the
turbulent atmosphere.” J. Opt. Soc. Am., Vol. 73, No. 3, March 1983, p. 277
281.
[30] Frehlich, R.G. “Intensity covariance of a point source in a random medium with
a Kolmogorov spectrum and an inner scale of turbulence.” J. Opt. Soc. Am., Vol.
4, No. 2, February 1987, p. 3606.
[31] Hill, R.J. and S.F. Clifford. “Theory of saturation of optical scintillation by
strong turbulence for arbitrary refractiveindex spectra.” J. Opt. Soc. Am., Vol.
71, No. 6, June 1981, p. 67586.
[32] Clifford, S.F. and H.T. Yura. “Equivalence of two theories of strong optical
scintillation.” J. Opt. Soc. Am., Vol. 64, No. 12, December 1974, p. 16414.
[33] Andrews, L.C. et al. “Aperture averaging of optical scintillations: power
fluctuations and the temporal spectrum.” Waves Random Media, Vol. 10, 2000,
p. 5370.
[34] Andrews, L.C. “Apertureaveraging factor for optical scintillations of plane and
spherical waves in the atmosphere.” J. Opt. Soc. Am., Vol. 9, No. 4, April 1992,
p. 597600.
[35] Churnside, James H. “Scintillation from a Diverging Laser Beam.” NOAA
Technical Memorandum ERL ETL240, September 1994. (available from the
National Technical Information Service <http://www.ntis.gov>).
133
[36] Hutt, D.L. “Modeling and measurements of atmospheric optical turbulence over
land.” Opt. Eng., Vol. 38, No. 8, August 1999, p. 128895.
[37] Oke, T.R. Boundary Layer Climates. Methuen & Co.: London, 1978.
[38] Ochs, Gerard R. and Reginald J. Hill. “Opticalscintillation method of
measuring turbulence inner scale.” Appl. Opt., Vol. 24, No. 15, 1 August 1985,
p. 24302.
[39] Frehlich, Rod G. “Estimation of the parameters of the atmospheric turbulence
spectrum using measurements of the spatial intensity covariance.” Appl. Opt.,
Vol. 27, No. 11, June 1988, p. 21948.
[40] Hill, R.J. “Saturation resistance and innerscale resistance of a largeaperture
scintillometer: a case study.” Appl. Opt., Vol. 20, No. 22, 15 November 1981, p.
38224.
[41] Hill, Reginald J. and Rod. G. Frehlich. “Onset of strong scintillation with
application to remote sensing of turbulence inner scale.” Appl. Opt., Vol. 35,
No. 6, 20 February 1996, p. 98697.
[42] Churnside, James H. et al. “Aperture size and bandwidth requirements form
measuring strong scintillation in the atmosphere.” Appl. Opt., Vol. 28, No. 19, 1
October 1989, p. 412632.
[43] Derickson, Dennis (ed.). Fiber Optic Test and Measurement. Prentice Hall PTR:
New Jersey, 1998.
[44] Churnside, James H. and Richard J. Lataitis. “Wander of an optical beam in the
turbulent atmosphere.” Appl. Opt., Vol. 29, No. 7, 1 March 1990, p. 92630.
134
[45] Ishimaru, Akira. Wave Propagation and Scattering in Random Media. IEEE
Press and Oxford University Press: New York, 1997.
[46] Tyson, Robert K. and Douglas E. Canning. “Indirect measurement of a laser
communication biterrorrate reduction with loworder adaptive optics.” Appl.
Opt., Vol. 42, No. 21, 20 July 2003, p. 423943.
[47] Korotkova, Olga and Larry C. Andrews. “Model for a partially coherent
Gaussian beam in atmospheric turbulence with application in Lasercom.” Opt.
Eng., Vol. 43, No. 2, February 2004, p. 33041.
[48] Welford, W.T. and R. Winston. High Collection Nonimaging Optics. Academic
Press: San Diego, 1989.
[49] Winston, R. “Dielectric compound parabolic concentrators.” Appl. Opt., Vol. 15,
No. 2, February 1976, p. 2912.
[50] Miñano, Juan C. et al. “Ultra compact optics for optical wireless
communications.” Proc. SPIE, Vol. 3850, Optical Wireless Communications II,
September 1999, p. 8090.
[51] Mohedano, Rubén et al. “Ultracompact nonimaging devices for optical wireless
communications.” Opt. Eng., Vol. 39, No. 10, October 2000, p. 27407.
[52] Agrawal, Govind P. FiberOptic Communication Systems. John Wiley & Sons:
New York, 1997.
[53] Davis, Christopher C. and Igor I. Smolyaninov. “The Effect of Atmospheric
Turbulence on BitErrorRate in an OnOffKeyed Optical Wireless System.”
Proc. SPIE, Vol. 4489, FreeSpace Laser Communication and Laser Imaging,
July 2002, p. 126137.
135
[54] Biswas, A. and M.W. Wright. “MountainToptoMountainTop Optical Link
Demonstration: Part I.” IPN Progress Report 42149, Jet Propulsion Laboratory,
Pasedena, CA, 15 May 2002.
[55] Kim, Isaac I. et al. “Comparison of laser beam propagation at 785 nm and 1550
nm in fog and haze for optical wireless communications.” Proc. SPIE, Vol.
4214, Optical Wireless Communications III, February 2001, p. 2537.
[56] Moore, C.I. et al. “Freespace highspeed laser communication link across the
Chesapeake Bay.” Proc. SPIE, Vol. 4821, Free Space Laser Communication
and Laser Imaging II, December 2002, p. 474485.
[57] Gilbreath, C.G. and W.S. Rabinovich. “Research in Freespace Optical Data
Transfer at the U.S. Naval Research Laboratory.” Proc. SPIE, Vol. 5160, Free
Space Laser Communication and Active Laser Illumination III, January 2003, p.
22533.
[58] “Maxim 1.25Gbps/2.5Gbps, +3V to +5.5V, LowNoise Transimpedance
Preamplifiers for LANs.” Rev. 1, June 2000, Maxim Integrated Products.
[59] “Accurately Estimating Optical Receiver Sensitivity.” Application Note: HFAN
3.0.0, October 2001, Maxim Integrated Products.
propagation. New expressions for the aperture averaging factor in weak turbulence are given. In strong turbulence conditions, aperture averaging is analyzed with special attention to the various wavenumber spectrum models. This is the first report of experimental strong fluctuation aperture averaging data acquired in nonsaturated conditions. Nonimaging optical elements are particularly useful for the mitigation of atmospheric turbulenceinduced beam wander in the focal plane of a free space optical communication receiver. Experimental results of the bit error ratio enhancement due to the incorporation of a nonimaging optical element, specifically a compound parabolic concentrator, are presented. Two link ranges were tested, a 1.7 km link at the University of Maryland experiencing weak turbulence, and a 32.4 km link at the Naval Research Laboratory’s Chesapeake Bay Detachment experiencing saturated, strong turbulence. These results are the first reported experimental test of a nonimaging optical element integrated into an outdoor free space optical communications system.
TECHNIQUES TO MITIGATE THE EFFECTS OF ATMOSPHERIC TURBULENCE ON FREE SPACE OPTICAL COMMUNICATION LINKS
by Linda Marie Wasiczko
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2004
Advisory Committee: Professor Christopher C. Davis, Chair Professor Pamela Abshire Professor Agisilaos A. Iliadis Professor Thomas E. Murphy Professor Owen E. Thompson
©Copyright by Linda Marie Wasiczko 2004 .
humility. Bernadine & Dennis. and faith Keep smiling ii .Dedication To my Parents. for teaching me perseverance.
Ray Burris. Finally. Dr. Igor Smolyaninov. and Dr. I also want to thank the members of my dissertation committee for their service and guidance. Pamela. Charles Thompson of the Lawrence Livermore National Laboratory. my sister. Kyuman Cho for introducing me to this project. Jennifer Ricklin and Dr.. Chris Moore. and my friends who have supported (and antagonized) me throughout my many years in graduate school. I want to thank my parents. Prof. iii . of the Naval Research Laboratory.Acknowledgements I want to express my appreciation to my advisor. Mikhail Vorontsov. and Dr. especially: Dr. and Prof. especially: Dr. I am thankful for the technical and nontechnical discussions with my colleagues in the Maryland Optics Group at the University of Maryland. Dr. H. Dr. Craig Casey. J. Vildana Hodzic. Patricia Mead and Dr. I want to thank two dear colleagues and friends for their continuing encouragement: Dr. and support during this experience. for which I am extremely grateful. and Mr. Christopher Davis. Mario Dagenais. Dr. I appreciate the equipment loans to this dissertation research from Prof. Dr. Quirino Balzano. The American Association of University Women sponsored my dissertation year work through a Selected Professions Fellowship. I also want to thank Dr. Michael Vilcheck. Stuart Milner. friendship. I have enjoyed fruitful conversations and collaboration with members of the free space optics community. of the Army Research Laboratory. for his wisdom. Agis Iliadis. Inc. “Buzz” Graves at AOptix.
.2 Nonimaging Optics . . .3. . . . . . . . . 3. .4 2. . . . . . . . 2.1 Churnside Asymptotic Analysis for the Plane Wave . . . . . . . . . . . . . . . .6 . . . . . . . . . . . 2. . 2. . . . . List of Figures . . . . . . . . . . .TABLE OF CONTENTS List of Tables. . . . . . . . .2. . . . . . . . . . . . . . .1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . . . . . . . . . . . . . . . . . . . . . . vii viii 1 1 2 4 4 5 6 7 9 9 10 12 15 19 25 25 26 28 30 30 31 32 32 33 34 37 37 38 38 38 Optical Wave Propagation through Atmospheric Turbulence . . . . . . . . . . . . .3 2. . . . . . . . . . 2. 2. . . . . . . . . . . .1 History of Free Space Optical Communication . . . . . . . . . . 2. . . . . . . . . . . . 1. . . . . . . . .7. . . . . . . . . . . . . . . .2 Churnside Asymptotic Analysis. . . . . . . . . . .1 Scintillation Index Model for a Plane Wave. . . . . . . . . . . . . . . . . . . .2. . 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chapter 2: 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . .2 Weak Turbulence. . . . . . . . . . .3. . 2. . . . . .6. . . . .1 General Aperture Averaging Form. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Atmosphere .7 Irradiance Variance Models Valid in Both Weak and Strong Turbulence Conditions . . 1. . . . . .2 2. Kolmogorov Turbulence .1 Aperture Averaging . . . .6. . . . . . .1 2. . Strong Turbulence Theory . . . .5 2. . .6. . . . . . Turbulent Energy Flow . . . . . . . . . . . . .2 Churnside Asymptotic Analysis for the Spherical Wave . . . . . . . . . . . . . . . . .5 Organization . . . . . . . . .1. . .7. . . . . . . . . .7. . .1 Andrews Asymptotic Analysis for the Plane Wave 2. iv . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . .1. . Chapter 3: Aperture Averaging in Weak Turbulence 3. . . . 1. . . . . . . . . . . 1. . . . . . . Wavenumber Spectrum Models . . . . . . . . Chapter 1: Introduction.2. 3. 1. . . . . .2 Andrews Asymptotic Analysis for the Spherical Wave . .6. . . . . . . . . . . . . . . . . .2 Data Analysis for Weak Turbulence . . . Rytov Approximation . .3 Scintillation Index Model using the Atmospheric Spectrum. .6. . . . . . . . . . . . . . . . . 3. . .2 Free Space Optical Communication Techniques . . . . . . . . . . . . . . . .3 LowOverhead Techniques to Mitigate the Effects of Atmospheric Turbulence . . . . . . . . . . . . . .6. . . . . . . . . . . . . . .1 Andrews – Prokhorov Asymptotic Analysis . . . . . . . . 2. . . . . . . . .4 Useful Terminology . . . . . . . . . . . .2 Scintillation Index Model for a Spherical Wave. . . . .
. . . . . . . . . . . . . .3. . . . . 3. .2.2. . . . . . .7 3. . . . .3 Spherical Wave Analysis. .4 Comparison of Strong Turbulence Models for the Plane Wave . . . . . . . 4. . . . . . . . . . . 4. . .2. . . . . . . . . .1 Introduction .5 Spherical Wave. . . . . . . . . . . . .2 Data Analysis in Strong Turbulence . . . . . . . .2. . .2. . . . . . . . . . . . . .3 3. . . .2. .3. .2. . . . . .3. . . . . . . . . . ℓo on the Order of the Fresnel Length 3. . . . . . . . . . . . . . . . . . .2 Plane Wave. . . . . . . . . . .6 3. . . .3 Aperture Averaging Experiment .3. . 3. . . . . . .2. . . . . 4. . . . . . . . . . . . . 4. .2. . . . .1 Introduction . . . . . . . . .1 General Aperture Averaging Form. . . 5. . . . .2. 5. . . Conclusions. ℓo on the Order of the Fresnel Length.1 Experimental Setup . . .2. . . . . . . . . . . . . . . Aperture Averaging Experiment. . . . . . 4. . . 4. . . . 3. . . . . . . . . . . Determination of the Inner Scale of Turbulence. .2. . . . . . . . . . . . . . . . . . . . . . 3. . . . . . 4. .2 Andrews Asymptotic Analysis for the Spherical Wave . . . . . . .6 Conclusions . . . . . . . . . Small ℓo . . . . . . . . . . 3. New Aperture Averaging Model. . . . . .3. . . . . . . . . . . . . . . .3 Scintillation Index Model for the Plane Wave . . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . 39 41 42 42 44 44 44 45 50 53 54 57 60 62 66 68 68 68 68 68 68 72 73 74 76 76 79 80 80 82 82 83 83 88 94 95 95 95 Chapter 4: Aperture Averaging in Strong Turbulence . . . . . . . . .1 Churnside Asymptotic Analysis for the Spherical Wave . . . . . . . . .3. . . . .4 Experimental Results for the Plane Wave Case . . . . . . . . . . .2 Plane Wave Analysis. . . . . . .2 LabVIEW Data Acquisition . . . . . . 4. . . . . . . . . . . . . . . . . . . . .2 Andrews Asymptotic Analysis for the Plane Wave 4. . . .3.2 TurbulenceInduced Beam Motion. . . . . . . .2. . . .1 Churnside Asymptotic Analysis for the Plane Wave . . . Large ℓo . 4. . . . . .2. . . . v . . . . . .3. . . . . . . . . .2. Large ℓo . . . . . . . .3. . . . . . . . . . . . .3. . . . . 4. . . . 4. . . . . . . . . . . . . . . . 4. . . .1 Aperture Averaging Transmitter and Receiver Systems.2. . . . .2. .2.3 Scintillation Index Model for the Spherical Wave.4 Comparison of Strong Turbulence Models for the Spherical Wave . . . . . 4. 4. .2. . . . . . . . . .6 Spherical Wave. . . . . . . . . . . . .2. . . 4. . . . . . . .3 Plane Wave.4 Spherical Wave. . . . .2. .2. . .2. . . . . . .2. .2. . .3 Fetch Effects . . . . . Spherical Wave Experimental Results . . . . . . . 3. Small ℓo . . . . . .5 Experimental Results for the Spherical Wave Case .2. . ℓo . . 3. . . . Plane Wave Experimental Results . . . . . . . . . . . . . . . . . . . Chapter 5: Free Space Optical Communication using Nonimaging Optics . . 4. . . . . . . . .4 3. . . . . .8 3.1 Plane Wave. . . . . 4. . . . . . .2 LabVIEW Data Acquisition . . . . . .5 3. . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . .
. . 115 5. . . . . . 108 5. . . . . . . . . . . . . .2 ZnSe CPC . . . . .5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 Conclusions . . . . . .5. . . . .4. . . . . . . . . . . . . 105 1. . .1 Summary of Contributions . . . . . . .5.1 Theory of the Compound Parabolic Concentrator . . . .4 Cn2 Measurements . . . .3 Link Budget . . . . . . . . . . . . . . . . . . . . . . . . .2 Future Work. . . . . . . . . . 106 5. . . . 103 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 AngleofArrival and Image Dancing . . . . 110 5. . . . . . 125 6. . .3 The CPCPhotodetector Combination . . . . . . . . . . . . . .4. . . . . . . . . .2 CPCPD Combination . . . . . . . . . . . . . . . . . . 129 vi . . . . .6. . . . . . .6. .3 5. . . . . . . . . .5 5. . .1 Beam Wander .1 Experimental Setup and Test Range . . . . . . . . . . . . 101 5. . . . . . . . . . . 117 5. . . . . . . . . . . . . . . . 108 5. . . . . .4 5. . . . . . . .5. . . . . 115 5. . . . . . . . 125 6. . . . . . . . . . . . . .7 5. . . . . . . . . . . . . . . . . . . . . . . 95 5. 96 BER in OnOff Keyed Systems . . . . . . . . . . . .5 Experimental Results and Analysis .1 Experimental Setup and Test Range . . . 127 References . . . . .6 5. . . . . . . . . . . 111 NRL Link Experiment over the Chesapeake Bay . . . . . . . 97 Nonimaging Optics .2. .2 The DielectricFilled CPC . . . 106 5.7 km Link Experiment . . . . . .5. .5. . . . . . . . .3 Experimental Results and Analysis . . . . .6. . . . . . . 100 5. . . . . . . . 124 Chapter 6: Conclusion . . . . . .
. . . Link budget analysis for the 1. . . . . . . . . . . . . . intermediate. .1: Typical Rytov variance ranges corresponding to weak. . .7 km link. . . . .. . . . . . . . Channel 0 represents the aperture averaged signal. neglecting atmospheric turbulence. . 22 53 5. . .LIST OF TABLES 2. List of 13 values output per averaging interval in LabVIEW data file. and strong turbulence levels. . . Here. . . . . . . . . . . . . . . .1: 3. . . . . . . . . .1: 110 vii . . . . . . . . . . while Channel 1 represents the scintillometer signal. .
Eq. . . . . . . . . . . Satellite photograph taken in 2002 by GlobeXplorer. . . . . . showing the energy cascade and subsequent division of turbulent eddies in the atmosphere [adapted from Ref.4) while the dashed line is Churnside’s approximation in Eq. . Eq. coherence length. . . . . . . .25). . The solid line is the exact theory given a Kolmogorov spectrum. . . while the dashed line is the approximate formula of Eq. . . . .4: 23 2. . The solid line is the exact theory from Eq. . . . . The shaded area shows scale sizes that do not contribute to strong fluctuations. . the wavefront experiences strong fluctuations when Cn2 = 5×1013 m2/3. . . 11 18 2. .15) [11]. Scaled spectral models for the von Karman spectrum. . . . . . Three wavenumber spectrum models for refractive index fluctuations. . . . . Williams Building at the University of Maryland. . . . . Aperture averaging transmitter on the roof of the A. .3: 3. . . . . The Fresnel zone size. . . (3. . . . . . .3: 19 2. . . . .7) [11]. . . . . . . . . and modified atmospheric spectrum. . .1: Depiction of the process of turbulent decay.1: 41 3. . . . (2. . . . . Three scale sizes plotted against the propagation distance for the spherical wave case with L = 863 m. . . . . Aperture averaging factor for a plane wave plotted against the ratio of the aperture radius to the Fresnel zone size. . .. . .24). . . . . . . .V. . . 13]. . . . . . (3. .6: 24 3. . From the graph. . . . An aerial photograph of the propagation path (dotted line). . . plotted against the wavenumber scaled by the inner scale.2: 43 46 47 3. and scattering disk size plotted for a plane wave against propagation distance. (3. . . . . . . . . . . . . where the contributing scale sizes are outside the shaded area. . . . . . . When Cn2 = 5×1015 m2/3 the wavefront sees weak scintillations with contributing scale sizes on the order of the Fresnel zone size. .2: 2. . .4: viii . . For L = 863 m. College Park. . . . . . . . . .5: 24 2. Aperture averaging factor for a spherical wave with small inner scale. . . .LIST OF FIGURES 2. Three scale sizes plotted against the propagation distance for the spherical wave case with L = 863 m. . . . . . The three energy ranges relevant to turbulence statistics are indicated. . there are scale sizes on the link contributing to strong fluctuations. . . . . (2. . . . .
Flowchart of the data acquisition process in LabVIEW for one channel. 3. . . .. . . . (3. . 3. . .8). .11) plotted using the curve fit value of ℓo = 5.6: Diagram of the aperture averaging receiver setup. Front panel of the LabVIEW program designed to calculate and record irradiance statistics for the aperture averaging experiment. interval. . . . . . . . . .10: New mean experimental data plotted with error bars. . 3. . . . .15: New weak turbulence fit (Eq. . . . . . . . . . .7: 3. .8: 52 3. . . . . .7). . .14: Comparison of our new mean data with the Churnside data. 3.11). . . . . Neither the plane wave Churnside approximation nor the SI model are a good fit to the data. . . . . . . . . . . . . . . . (3. . . . . . . . . . . . predicts a higher knee in the curve than the SI model. . Plane wave aperture averaging factor plotted as the Churnside approximation Eq. . (3. 3. . . and the scintillation index (SI) model for various values of Cn2. . . . . . (3. 3. Eq. . 3. . . . . .16: New fit to weak turbulence data with an additional square root term. . . . . . . . . .3. The signal detected on Channel 2 shows smaller intensity fluctuations. . . . . . 3. . .9: 55 3. . . . . . .27 cm. . . . . . . . . given by Eq. . . . . . . . The dashed line represents Eq. . . . . plotted along with the Churnside approximation in Eq. . . . . . . 3. . An optically chopped signal propagated over 863 m and viewed on the Tektronix oscilloscope. .15). . (3. . . . . .12: Comparison of two different models for the spherical wave in weak turbulence. The scintillometer channel uses a 5 mm diameter receive lens. .18) shown with new data and the Churnside model (Eq. . . 49 49 50 3. . .13: Experimental mean aperture averaging data plotted against the ratio of aperture radius to the Fresnel zone size. . . . . . . . . . . . . . . . . while Channel 2 is detected by the aperture averaged receiver. . . . . . . . . . The Churnside approximation. the Andrews approximation Eq. . . . . . . (3. .7). ix 56 57 58 59 60 61 62 63 . . . . . .15).17: Eq. . . . . . . . . . . . . . against the ratio of the aperture diameter to the inner scale. . . . . along with approximations to the aperture averaging factor. . .5: 3. . . . . . . . . . Channel 1 is the signal detected by the scintillometer. .11: New mean aperture averaging data plotted with Churnside experimental data taken over a 500 m path. . . The diagram explains how intensity and background levels are determined for the sampled signal averaged over a 1 min. . . (3. .19). . . .
. . . (3.19: New data is plotted using the predicted value ℓo = 4. . the ratio of aperture radius to the transverse coherence length. . . . .8 nm. . . . .18: Spherical wave aperture averaging factor for ℓo = 1. . . . . . . .45 cm using Eq. where turbulence strength increases in the asymptotic limit of σ I2 → 1 .3: 4. .2: 4. . Churnside approximate aperture averaging factor plotted against the ratio of the aperture radius to the transverse coherence length. . . . . . . . . . . . Asymptotic models in the saturation region are shown. .5: 78 4. . . . . Aperture averaging factor using the Andrews asymptotic model vs. Behavior of the transverse coherence length in the strong fluctuation region. . . . . . . . .17). . . . . . . . . . . . . . . . . . . . . for L = 863 m and λ = 632. .1: The plane wave aperture averaging factor A vs. . . . . .6: 4. Average aperture averaged data in strong turbulence using Churnside asymptotic analysis plotted against the ratio of the aperture radius to the transverse coherence length. The dashed line represents Eq. . . . . . . . Asymptotic and weak turbulence relations are indicated. . and 1. . . the ratio of the aperture radius D/2 to the transverse coherence length ρ0. . . . . . . . . . .9: 79 81 82 84 x . . . . . . .1. . .008 mm. . . . . . . . . 3. for σ I2 = 1. . . .17) plotted along with the new data. . . . . . . . . . Curves calculated for an AVW test range path length of 863 m. . . Andrews asymptotic model for the spherical wave plotted against the ratio of the aperture radius to the transverse coherence length. . . . . . . . . . . . . The inner scale was determined using the atmospheric spectrum. . or scintillation index. . . . . . .5. Irradiance variance. . . . . . . of a plane wave versus Rytov variance. . .4: 73 75 76 4. . . 64 65 71 4. . . . . . . . . . . . . . . . .25. .3. 1. . . σ I2 . 4. . . . . . Transverse coherence length based on strong turbulence models plotted against irradiance variance. . plotted as a function of the plane wave Rytov variance. . . . . . . Churnside asymptotic analysis predicts Cn2 = 9. .3×1012 m2/3. . . . . . . . Spherical wave irradiance variance plotted as a function of plane wave Rytov variance. . (3. . . . . . The legend indicates different values of irradiance variance σ I2 . . . . . . . . .8: 4.7: 4. which is represented by the dashed line.. . .
there 4L is no aperture averaging present because the aperture radius is equal to the Fresnel zone size (FZ). . . . . . . . . . . . . . . . . . . . . . . . 4. . . .13: New strong turbulence data plotted using the Andrews asymptotic model for the spherical wave from Eq. . plotted along with the Churnside data for a 1000 m path. . . The curves represent the aperture averaging factor given saturated and nonsaturated 2 strong turbulence values for σ R . . . .1. . . . .1. . . . The BER of a communications link with a gammagamma probability distribution function (PDF) as a function of signaltonoise ratio. . . The curve represents the Andrews asymptotic model aperture averaging factor for σ I2 = 1. . . . . 4. Saturated and nonsaturated SI model curves for σ I2 = 1. 20]. . . . . with irradiance variances near 1. . . .0. . . . . . . . 48). . .1 . .3: The cone concentrator with cone angle γ and maximum entrance angle θi. The = 1 . . . . . . .3. . . . . . .10: New data using the Andrews asymptotic theory plotted against the ratio of the aperture radius to the transverse coherence length. . . 0. .15: Mean data analyzed with the Churnside asymptotic model. . . . . . . . . . . When kD 2 100 5. . . 4. .1: 5. . . . 86 87 89 90 91 93 94 96 Rytov variances plotted are 0. . . . . . The cone is hollow with reflective (metallic) edges. The irradiance variances used in the calculation are the same in both cases. . . . . (4. . . . . . . . . . . . . and 2. (4. . 4. . 4. . 4. . . . .19). . . .14: Mean data with error bars plotted against the ratio of the aperture radius to the transverse coherence length.1 .2: Beam wander at the receiver plane is characterized by the short term and long term beam widths [adapted from Ref. .11: New aperture averaging data calculated using the SI model plotted against the ratio of aperture radius D/2 to the transverse coherence length ρ0 for a plane wave propagating in strong turbulence.1 are also shown. . . . . The solid line is the aperture averaging approximation in Eq. . . . Three more curves are plotted for an aperture averaged case using a 4 inch diameter aperture. . (adapted from Ref. 101 xi . . plotted along with the strong turbulence Churnside data taken over a 1000 m path.4. . . . . . . . . .16: Mean data analyzed using the scintillation index model in the nonsaturated strong turbulence region. . .6. . . . . .12: New strong turbulence data analyzed using the Churnside asymptotic method for a spherical wave. . The solid line represents the aperture averaging factor given by the Andrews asymptotic theory for σ I2 = 1.17) for σ I2 = 1. 5. . . . . .
. . . .9: 5. . 5. . The axis of the parabola and the CPC axis are different. . 5. There were no errors between minutes 1 and 19. The BER of 1 bit in a 51. and is lowered below the field of view of the 16 in. . . . . . . .10: Trace 1 shows the data stream after passing through two 20 dB RF amplifiers. . . . . BER plotted as a function of Cn2 for the 1. .12: Optical transmitter at NRLCBD. . . . . . . . . .6: 107 108 5. . . . . . . . . . . . 115 116 116 117 119 120 xii . . The only errors were detected in the 31st min. . . . . . . The collimated optical output is approximately 2 W [56]. 5. . . . . . . . The error bars in Cn2 represent its fluctuation over the 1 min averaging interval. . . . . . .7: 5. . . . θe. . . . . . . . . . . . . . . . . . Profile of the mounted ZnSe CPC directly coupled to an InGaAs PD. .11: The NRLCBD laser communication test range [56]. Same data as in Fig. . . . . . . . . . . . . . . . . .68 Mbps transmission at OC2 gives a BER of 1. . . . . . . .8: 112 113 5. . Trace 2 shows the inverted data after passing through a limiting amplifier. . 103 104 5. . . . . . . . . . . .5. . . . . . . as indicated by the dark horizontal line. . . . orange lines are fiber optic connections. . . . . 5. . . . 5. The signal after the limiting amplifier is much cleaner. .15: BER at OC2 over a 12 minute acquisition period using the CPCPD receiver. . . . . . . . . however there is still enough turbulenceinduced noise to degrade the BER. . Meade telescope when not in use. . .4: The design of the CPC surface. . . . . . . . . . . . . . . .14: Measurements of BER using the CPCPD receiver when an OC1 PRBS transmission pattern was tested during a 35 minutes test period. . . Black lines are electrical connections. . . .84 Mbps transmission is 3. . . . . . .607×1010. . θn for a range of refractive indices. . . . . The surface of the CPC is traced out by moving the vector r through the parametric angle φ.5: 5. . . . . . . plotted along with a first order linear fit to the mean data. . . . . . . 5. . . . . . . . . The IR camera is used for alignment purposes only. .7 km link with and without the CPC integrated into the optical receiver. . . . and the maximum internal ray angle. . . . .215×1010. . . . . . . . . . . . . Plot of the maximum entrance angle in air. . . . .8. A single bit error in a 103. . . . .13: The optical receiver at NRLCBD [56]. . Setup of the CPC link experiment. . . . .
. . . Note that data is plotted for 5 sec intervals. . . . . . . . .5. . . . . . The single error BER for a 1 min acquisition period is 1. . . . . . . from Ref. . . . 46. .17: BER at OC3 measured over a 15 minute period using the CPCPD receiver. . . . from Ref. .072×1010. which is beyond the range of the graph. . . . Note that data is plotted for 5 sec intervals. . . . which is beyond the range of the graph. . .16: BER at 100Mbps measured over the NRLCBD test range with an optical fiber coupled receiver. . . . 5. . . . . 46. . . . . 121 123 123 xiii . . . . . . . . . . . . . . . .18: BER at 200Mbps measured over the NRLCBD test range with an optical fiber coupled receiver. . . . . . . 5. . . . . . . . . . The minimum BER for a 5 sec interval is 1×109. The minimum BER for a 5 sec interval is 2×109.
The advantages of FSO were trumpeted throughout the 1 . optical fiber. By this time.3. FSO was voted one of the ten hottest technologies in 2001 [5].4]. After September 11. scientists have attempted to establish reliable communications channels by modulating and propagating optical signals over a lineofsight path [1]. costeffective components in FSO systems.to late1990s.1 History of Free Space Optical Communication Since the invention of the Ruby laser in 1960. free space laser communication was once again resurrected. after copper wires. and researchers had found ways to use these efficient. renewed interest in free space optics (FSO). This time. optical fiber communications was well established. research in free space optical communication began a cycle of declining interest and funding. and RF antennas were destroyed. Due to the relative immaturity of laser technology and flawed system demonstrations. an increasing demand for high bandwidth communications. Most of the early development of free space optical communication links was directed towards spacebased applications [2. FSO systems were deployed to reconnect parts of the critical New York City infrastructure. During the mid. fueled by the explosion of the internet. only to be resurrected over and over again.Chapter 1 Introduction 1. 2001. The hype surrounding the technology was partly due to the lack of optical fiber backbone connectivity in metropolitan areas and the high cost of laying optical fiber in those areas.
the ability to extend the reach of the optical fiber backbone. This is a low overhead technique. Atmospheric turbulence may easily prevent the detection of a sufficient number of photons to correctly interpret the received bits. aerosols. Even with the significant capital investment in FSO during this time. The research presented here demonstrates techniques to mitigate the distortions induced on the optical beam by atmospheric turbulence. This left FSO vendors without telecommunication and cable carriers who could purchase and deploy FSO units on a large scale. 2 . and molecular attenuation. This transmission scheme attempts to collect enough photons at the receiver to correctly detect “1” and “0” bits. A “bare bones” FSO system transmits “1” bits by turning the optical source on.telecommunications sector. 1. high security. Unfortunately. the technology bubble in the stock market burst. This method is called intensity modulation/direct detection (IM/DD).2 Free Space Optical Communication Techniques Researchers have pursued a variety of system architectures in an attempt to improve the reliability of FSO links. researchers did not solve the technology’s reliability problems due to atmospheric turbulence. support for high bandwidth transmissions. These three environmental properties destroy the coherence of a propagating optical wavefront. These advantages include: rapidly deployablility. because there is no attempt made to ensure the integrity of the signal. and use in disaster recovery situations. or onoff keying (OOK). and “0” bits by turning the optical source off.
there is an increased likelihood that the detected signal will be read correctly. researchers have tried to come up with techniques to improve the likelihood that data bits will be received correctly. Although the techniques are promising. Scientists have used adaptive optics techniques. Diversity can occur in the form of spatial diversity (requiring multiple transmitters and/or receivers). to try to restore a distorted wavefront to its original state before it was destroyed by atmospheric turbulence. separated by a time delay). originally developed to improve the quality of stellar telescopes. they do require a significant electronic overhead in the retiming and synchronization process. Diversity techniques are another method of improving the reliability of an FSO system. Additionally. to predistort the beam and hope that atmospheric turbulence. Each of these techniques requires a synchronization of the received signals. have been 3 . temporal diversity (requiring a signal to be transmitted twice. Methods requiring heterodyne detection. including phase shift keying and quadrature amplitude modulation. they require bulky and computationintensive systems to achieve wavefront correction. systems have not been able to compute the wavefront correction fast enough to entirely counteract the affects of atmospheric turbulence. Some used phase conjugation techniques. by distorting the wavefront. will actually work to render the wavefront to its unperturbed form. or wavelength diversity (requiring the transmission of data on at least two distinct wavelengths). Although these techniques have shown limited success. Finally.Beyond the basic IM/DD scheme. In effect. if the optical wavefront propagates in at least two distinct ways. coding schemes used in RF and wired communications systems have been adapted for FSO communication.
1. a lens of any diameter will collect more photons than an ideal point receiver.7]. Aperture averaging was experimentally investigated for both horizontal and spacetoground paths under weak 4 . Forward errorcorrection (FEC) codes insert check bits into the data stream. 1.experimentally demonstrated but do not guarantee enough improvement to be viable commercially [6.8.9]. Although coding provides an additional layer of information security. bandwidth. In effect. size. aperture averaging will increase the likelihood of correctly detecting the transmitted data stream. Fried developed a theoretical expression characterizing the amount of aperture averaging given an incident infinite plane wave. such as OOK. which states that the amount of measured radiation may be increased by increasing the size of the receiver collecting lens aperture.3 LowOverhead Techniques to Mitigate the Effects of Atmospheric Turbulence The research presented here demonstrates how highperformance FSO communication systems may be designed using techniques that add nearly no power.3. studies have shown that even the best FEC codes cannot negate the affects of atmospheric turbulence alone [6. any irradiance fluctuations across the collecting lens are “averaged” by the size of the lens.1 Aperture Averaging Aperture averaging is a well known concept. Therefore. In a binary communications transmission. or weight overhead to the overall design. Aperture averaging of irradiance fluctuations was first studied by Fried in 1967 [10]. which contribute an additional power and bandwidth overhead on the system.
As such. we are only concerned with collecting a 5 . In the case of FSO with IM/DD. although its size may be magnified or reduced. in 1991. If a converging lens is used. Later in 1973. Fried came to an erroneous conclusion that the amount of aperture averaging is proportional to the inverse square of the aperture diameter.2 Nonimaging Optics On a basic level. 1. so scientists could not truly differentiate between weak and strong fluctuations. A diverging lens will form a virtual image on the same side of the lens as the object. imaging optics form an image based on the transformation of an object by an imaging element. the image or virtual image is a perfect replica of the object. A lens or compound lens is a typical imaging element.levels of atmospheric turbulence. although new theories and models have been developed recently. when the object is placed at or beyond the focal length of the lens.3. Fried came to the correct conclusion that there is a − 7 / 3 power dependence. there have been no consistent experimental results for aperture averaging in strong turbulence. then the object and image are located on opposite sides of the lens. Ideally. The concept that the amount of irradiance fluctuations saturates after a certain propagation distance was not understood. Subsequent experiments on aperture averaging in weak atmospheric turbulence found very poor agreement with theory. Churnside published experimental results that have shown the best agreement with theory in weak turbulence conditions until now [11]. Finally. The behavior of optical waves propagating in strong turbulence is not well understood. Imaging optics are useful when the quality of an optical wavefront must be preserved. However.
a nonimaging concentrator is used to mitigate the affects of atmospheric turbulence on the focal plane of an imaging optical receiver. Projected area – the area projected onto a plane whose normal is the line of sight. 6 . Only the conservation of energy is required. A sphere has a 4π solid angle. Instead. the area is multiplied by the cosine of the angle between the normal to the area and the line of sight. and dispense radiation over a wider angular range. In this research. the projection of an area onto a unit sphere. These elements collect radiation over a restricted angular range.maximum number of photons.4 Useful Terminology Radiometric terminology has been loosely used in scientific research. but will improve the detection efficiency of the receiver. 1. what would be the image may be partially or completely randomized. measured in steradians (sr). Atmospheric turbulence will destroy the coherence of a propagating optical wave. we specifically define radiometric terms used throughout this dissertation. Solid angle – Ω. Any information about the phase of the wavefront is ignored. and induce phenomena that will affect the collection of a maximum number of photons. Nonimaging optics are not required to preserve the integrity of the object. A special type of nonimaging optical element is called a concentrator. The nonimaging concentrator will not preserve the coherence of the wavefront. Here. The concentrator operates on the principle of conservation of brightness. preserving the relationship between the irradiance and its angular distribution.
1. the number of cycles per second. A framework for studying optical wave propagation through atmosphere turbulence. expressed in W/m2. also known in this work as intensity. 7 . Chapter 2 reviews both the physical nature of atmospheric turbulence and theories developed to characterize atmospheric turbulence. where λ is the wavelength of the optical wave. is presented. the distribution of radiance per unit wavelength. Irradiance – I = dP dΩ . Spectral radiance – Lλ = dL dλ . Radiance – L.5 Organization This dissertation is organized in four chapters detailing the theory. experimental methodology. the flow of energy per unit area per unit time. the inverse of a spatial scale length. Aperture averaging models. The spectral radiance is meaningful by integrating it over a wavelength range. The irradiance is proportional to the square of the amplitude of the electric field of an optical wave.Optical power – P. expressed in W/m2srHz. Chapter 3 involves an in depth study of aperture averaging in weak turbulence conditions. Optical wavenumber – k = 2π λ . the optical power (or radiant flux) per unit projected area per unit solid angle. Frequency – ν . expressed in W/m2. and analysis used to study techniques to mitigate the effects of atmospheric turbulence. in strong and weak turbulence conditions. expressed in W/m2sr. the flow of radiant energy per unit time. also known as radiant flux. Spatial wavenumber – κ. methodology. measured in Hz or sec1.
but instead addresses aperture averaging in strong turbulence conditions. 8 . Chapter 5 studies and characterizes the performance enhancement of free space optical communication systems when nonimaging optics are integrated into the optical receiver. Chapter 4 follows the organization of Chapter 3. New experimental data in the strong turbulence region are analyzed and compared with available models. Chapter 6 summarizes the contributions of this work and addresses areas for future research.and analysis of new experimental data are explained.
1) where n is the total refractive index. At sea level. and T is the temperature in K. humidity. 9 . The refractive index of air at optical frequencies is. Atmospheric turbulence is a result of localized variations of temperature.Chapter 2 Optical Wave Propagation through Atmospheric Turbulence 2. λ is the wavelength in µm. p is the pressure in mb. n − 1 is typically 3×104 [1]. since humidity affects the value of the refractive index by less than 1%. but the cumulative effect of eddies over a 1 km path is great. scattering off aerosols. where each localized area of lower or higher refractive index is known as a turbulent eddy [1]. This thesis addresses the predominant cause of distortion of optical waves in the atmosphere. 7. which is atmospheric turbulence. and atmospheric absorption. Humidity effects are typically neglected over land.1 The Atmosphere Propagation of optical waves through the atmosphere is affected by atmospheric turbulence. and pressure in the atmosphere.52 ⋅ 10 −3 p − 6 ⋅ 10 n − 1 = 77.61 + T λ2 (2. These variations result in localized refractive index fluctuations. The refractive index of each individual eddy is not much greater than unity.
inertial subrange.2) where U is the characteristic flow velocity. L is the characteristic dimension of viscous flow. with an energy input region. and eventually give rise to eddies of smaller scale sizes and lower velocities. Re ( j ) = U ( j ) L( j ) υ ( j ) where j is the order of the eddy.3) (2.2.2 Turbulent Energy Flow Turbulent flows result when large inertial forces draw together fluid volumes with very different velocities. and energy dissipation region. Since Reynolds numbers in the atmosphere are large. The process is shown pictorially in Figure 2. Turbulent flows exist when the Reynolds number is greater than 2500 to 5000 [12]. and υ is the kinematic viscosity of the fluid.1. Eddies of scale sizes on the order of flow dimensions move randomly. a portion of kinetic energy in the atmosphere is converted into turbulent energy [13]. When the characteristic length reaches a specified 10 .4) At large characteristic lengths ℓ. where the Reynolds number is defined by: Re = UL / υ (2. eddies become small enough that viscosity forces overcome inertial forces. Richardson first developed a picture of the turbulent energy redistribution in the atmosphere. and the eddies can no longer decay. The breakdown in eddies is dictated by [12]. Eventually. U l ≈ (εl ) 1/ 3 (2. the associated fluid flows are also highly unstable. and irregular velocity fluctuations are apparent. Higher orders denote smaller eddies. The dissipation rate ε is related to the velocity Uℓ of an eddy with characteristic length ℓ by [12].
ℓo is typically on the order of 4 mm [12].outer scale length.1: Depiction of the process of turbulent decay. energy begins to cascade. showing the energy cascade and subsequent division of turbulent eddies in the atmosphere [adapted from Ref. l o = 7.148 cm2s1 [13]. 11 . The inner scale length.4 υ 3 ε ( ) 1/ 4 (2. Kolmogorov proposed that in the inertial subrange. Lo.5) At the surface layer of the Earth. until eddies reach a size equal to the inner scale length. and υ is typically 0. ℓo. 13]. is defined by. turbulence is isotropic and may be transferred from eddy to eddy without loss. or Kolmogorov microscale. The energy of one eddy is progressively redistributed into eddies of smaller scales. where Lo > ℓ > ℓo. Energy Input Lo Turbulence Spectrum Φ(κ) ℓo Inertial Subrange κin = 1/Lo Energy Dissipation κd= 1/ℓo Spatial Wavenumber κ = 1/l Figure 2. the energy of the eddy is dissipated as heat energy through viscosity processes [13]. When the diameter of a decaying eddy reaches ℓo.
These assumptions allow the velocity structure function to be treated as a scalar function [1]: r r r r Dij (r ) = [ D (r ) − D⊥ (r )]ni n j − D⊥ (r )δ ij (2.6) r r r v where r is the displacement between two points in space. The first is to assume local homogeneity in the atmosphere.3 Kolmogorov Turbulence Turbulence is by nature a random process. To continue this analysis. and i and j are two velocity components.2. and ni.6) to a dependence on the magnitude r r of r .0 ⇒ i ≠ j} . which restricts r the dependence of the velocity statistics in Eq. A second assumption of local isotropy limits Eq. given as r . the flow is assumed r r to be incompressible.6) to the vector displacement. r . To continue the structure function analysis. Kolmogorov first developed a universal description of atmospheric turbulence by developing a structure tensor to describe a mean square velocity difference between two points in the atmosphere. In 1941. is [1]: r r r r r r r Dij (r ) = [vi (r1 + r ) − vi (r1 )] ⋅ v j (r1 + r ) − v j (r1 ) [ ] (2. This assumption is valid when ν 2 a 2 << 1 . and as such may be described using statistical quantities. δ ij ∈ { ⇒ i = j. two approximations are made given the condition that the displacement is within the inertial subrange. The structure function. r1 + r and r1 [1]. r Dij (r ) . 12 . (2. nj are the r components of a unit vector along r [1]. giving ∇ ⋅ v = 0 .7) r r where D (r ) and D⊥ (r ) are the structure function components of the wind velocity r 1 field parallel and transverse to r . (2.
8 °/km.10) must be transformed into a refractive index fluctuation form.11) where T is the absolute temperature in Celsius. the potential temperature can be approximated by [15]: θ = T + cg h p (2. Kolmogorov determined that as long as r is within the inertial subrange. To use this relation in electromagnetic wave propagation problems.where a is the velocity of sound [12]. we can relate the parallel and transverse components of the velocity structure function: D⊥ = 1 d 2 r dr (r D ). the velocity fluctuation form of Eq. g c p = 9. For small values of r. which is the temperature of a parcel of air that is brought adiabatically (where heat is neither gained nor lost) from a state having pressure p and temperature T to a state where pressure p0 = 1000 mbar [14]. cp is the specific heat. we can describe the structure function in terms of one component: D = [v (r1 + r ) − v(r1 )] 2 (2.10) is the KolmogorovObukhov “twothirds” power law [12]. The value g c p is also the adiabatic rate of decrease of the absolute 13 . This is accomplished by using the potential temperature. Eq. g is the acceleration of gravity. 2  (2.16].9) As mentioned previously. θ. (2.8) With this relation. and h is the height of the parcel above the Earth’s surface [1. (2. we can assume localized fluctuations and the structure function becomes [1]: D = Cv2 r 2 / 3 (2. as when r is in the inertial subrange.10) where Cv2 is the velocity structure constant with units m2/3. Using the assumption of an incompressible flow.
Passive additives to do not affect turbulence statistics.11). the adiabatic rate g c p can be neglected and T is treated as a passive additive [1.16).52 ⋅ 10 −3 p 2 ∂T ⋅ 10 − 6 ∂n = 77.06 2 ∂θ ⋅ 10− 6 T (2.16) 14 .14) Using the relation of potential temperature to absolute temperature from Eq. (2. We can now infer a relationship between the structure parameter for refractive index and that of the potential temperature as: p 2 Cn = 79.15) The direct relationship between refractive index fluctuations and potential temperature fluctuations is shown in Eq. (2.06 2 Cθ2 ⋅ 10− 6 T (2.6328µm: p ∂n = 79.1) with respect to n.16. and is described by: Dθ (r ) = Cθ2 r 2 / 3 By taking the derivative of both sides of Eq. giving: 7.13) Knowing that pressure fluctuations are relatively small compared to temperature fluctuations. we find that for λ = 0.17]. At small heights above the Earth’s surface. we can approximate ∂p → 0 . The structure function for the potential temperature is well known [1].12) (2. and the relation of their derivatives ∂T = ∂θ because T is a passive additive.10).52 ⋅ 10−3 p ∂p ∂T − ∂n = 77.temperature. (2. and therefore do not alter the twothirds law in Eq. we find: 7. (2.61 + T p λ2 T (2.61 + 2 T λ (2.
4 Wavenumber Spectrum Models (2. allowing 15 . The value of Cn typically varies from 1017 m2/3 under “weak turbulence” conditions to 1013 m2/3 in “strong turbulence” conditions. we can define a Fourier transform to quantify the ability of different eddy sizes to influence the refractive index of a random medium. Cn is formally known as the refractive index structure parameter. the refractive index structure function obeys the same KolmogorovObukhov “twothirds” power law: r 2 r 2/3 Dn (r ) = C n r 2. To develop a Fourier spectrum. r1 ) = δn(r1 + r )δn(r1 ) (2.18) where δn is the fluctuating part of the refractive index. the assumption of homogeneity implies that only the separation between two points and not the location of r r r r the two points in a random medium affects the physics. Following the previous discussion of Kolmogorov turbulence (Sec.3). Near the Earth’s surface. and varies spatially and temporally. so that Bn (r1 + r . This relationship is given by a spatial covariance [1. the random medium may be treated as isotropic. We use this indicator to quantify the strength of 2 turbulence along the path of the optical propagation.Optical refractive index fluctuations near the Earth’s surface are a result of temperature 2 variations [1].13]: r r r r r r Bn (r1 + r . r1 ) → Bn (r ) . we must first determine how refractive index fluctuations vary between different points in the atmosphere.17) By using a methodology similar to that used for the characterization of the strength of various frequency components in a timevarying electrical signal. 2. Using previous information. with n = n + δn .
<< κ << 1 lo .19) = 4π ∫ dκκ 2 Φ n (κ ) where Φ n (κ ) is the isotropic and homogeneous turbulence spectrum.22) is the Kolmogorov spectrum.17) and (2. In spherical coordinates.20) which has been derived from the generic Fourier transform description of the spatial covariance and the definition of the refractive index structure function. (2. (2.033Cn κ −11 / 3 (2. The two descriptions of the refractive index 2 structure function from Eqs. r1 ) = [δn(r1 + r ) − δn(r1 )] .22) 1 Lo Eq. 16 . Dn (r1 + r . the threedimensional Fourier transform for the spatial covariance is: r r r r Bn (r ) = ∫ d 3κ exp[iκ ⋅ r ]Φ n (κ ) ∞ π 2π = −∞ 0 0 ∞ ∫ ∫ ∫ dκκ 0 2 r r r sin (θ )dθdφdκ exp[iκ ⋅ r ]Φ n (κ ) sin (κr ) κr (2. the wavenumber spectrum is [1. A Fourier integral also relates the structure function to the isotropic and homogeneous turbulence spectrum: sin (κr ) Dn (r ) = 8π ∫ dκκ 2 Φ n (κ )1 − κr 0 ∞ (2.r Bn (r ) → Bn (r ) .21) Using power law relations.20) may be compared to determine the wavenumber power spectrum: sin (κr ) = 8π ∫ dκκ 2 Φ n (κ )1 − κr 0 ∞ C r 2 2/3 n (2. It is only valid in the range Φ n (κ ) is an equivalent representation of Dn (r ) in inverse space.3]: 2 Φ n (κ ) = 0.
Andrews developed an analytical approximation to the Hill spectrum.92 . since the Hill spectrum uses a secondorder differential equation that has to be solved numerically. Although it does an excellent job of describing large eddy formation in the troposphere. Hill constructed an exact spectrum that accounts for the high wavenumber rise [19].033Cn 2 exp(− κ 2 / κ m ) (κ 2 2 + κ in ) 11 / 6 for 0 < κ < κ m (2. In the inertial subrange.92 l o and κ in = 1 Lo .24) where κ m = 5.There are other spectrum models that attempt to describe behavior beyond the inertial subrange. The von Karman spectrum integrates the small eddy dependence in the Tatarskii spectrum with a large eddy description to give: 2 Φ n (κ ) = 0. The von Karman model is the most widelyused model to describe lo characteristics in the energyinput region ( κ < 1 Lo ). The Tatarskii spectrum uses a Gaussian function to extend coverage to the dissipation range ( κ > 1 lo ). the model was originally proposed to describe fluid flow in a circular pipe [18]. which has the representation [20]: 17 .033Cn κ −11 / 3 exp − κ κm (2. the von Karman spectrum reduces to the Kolmogorov spectrum value. near κ m . Andrews calls his approximation the Modified Atmospheric Spectrum. The Kolmogorov and von Karman spectrums fail to show a “bump” at high wavenumber values in the inertial subrange. where small eddies are influential: 2 Φ n (κ ) = 0.23) where κ m = 5.
2. 2.254 κ l × exp − κ κ 2 κ 2 + κ in 2 ( ( 7/6 ) 2 l ) (2. von Karman. Fig.033C 1 + 1. and modified atmospheric spectrums are plotted in Fig. The high wavenumber “bump” occurs in the transition between the inertial subrange and dissipation range. and is especially apparent in the scaled spectral plot in Fig.25) The Kolmogorov.3. The bump occurs in the plot of the modified atmospheric spectrum at the boundary between the energy dissipation range and the inertial subrange in Fig.802 κ l 2 n κ − 0. κ Φ n (κ ) = 0. 2. 18 . 2. 2.2: Three wavenumber spectrum models for refractive index fluctuations. The three energy ranges relevant to turbulence statistics are indicated.2 for scale sizes of Lo = 10 m and lo = 5 mm.
t ) E (r . t ) [ ] (2. 2. t ) r r r r ∇ × E (r . A derivation of optical wave propagation is begun from Maxwell’s equations. Eq. Eq. assuming no free charges. t ) = jk0 H (r . and modified atmospheric spectrum.25). t ) = 0 r 2r r ∇ ⋅ n(r .21.22]. and an electromagnetic field with a harmonic time dependence of ejωt [1. (2.5 Rytov Approximation The Rytov approximation is a method to solve Maxwell’s equations for electromagnetic wave propagation while accounting for diffraction effects [1. t ) = − jk0 n(r .Fig. t ) E (r . plotted against the wavenumber scaled by the inner scale.3: Scaled spectral models for the von Karman spectrum.24). (2.22]: r r ∇ ⋅ H (r . 2. t ) = 0 r r r 2r r ∇ × H (r . µ0 permeability.26) 19 .
and j = − 1 .28) may be transformed into a scalar equation for each of the three electric field components. The Rytov method considers a solution of the form: r r E (r ) = exp(Ψ (r )) (2. c is the speed of light in vacuum. Since it has been shown that the change in polarization of a propagating wave is negligible for both cases of λ << lo and λ > lo [1]. n = n(r) is the refractive index of the atmosphere. which justifies a quasisteady state approach. (2. Eq.where k 0 = ω c is the free space wavenumber. this term may be dismissed. It is also assumed that the variation of the field with time is much slower than ejωt. the total refractive index is: r r n(r ) = 1 + δn(r ) (2. (2.28) becomes: r r ∇2 E + k 2n2 E = 0 (2.27) where the third term on the left hand side is a depolarization term.31) The Ricatti equation may be solved by a multiplicative perturbation method with a solution [22. (2.45]: r r r r r Ψ (r ) = Ψ0 (r ) + Ψ1 (r ) + Ψ2 (r ) + Ψ3 (r ) + . (2.. (2.28) Eq.30) into Eq.30) Substitution of Eq. Given a mean air refractive index of 1..32) 20 .29) where δn << 1. a wave equation is found: r r r ∇ 2 E + k 2 n 2 E − 2∇ E ⋅ ∇ ln (n ) = 0 ( ) (2.29) yields the nonlinear Ricatti equation [22. After some manipulation with vector identities and substitution [1.23.23]: r r r r ∇ 2 Ψ (r ) + ∇Ψ (r ) ⋅ ∇Ψ (r ) = −k02δn 2 (r ) (2.22].
The field equation for the basic Rytov solution is a result of keeping the first two terms in the expansion [22.36) 21 . (2.45].34) is a good approximation to the log amplitude variance 2 in the range σ χ < 1. The log intensity is related to the log amplitude variance by [1.45]: 2 χ 2 = σ χ = 0.307Cn2 L11 / 6 k 7 / 6 [ ] (2.34) It has been shown that Eq.33) Amplitude and phase information for the propagating wave falls out of the Rytov approximation. irradiance fluctuations obey a lognormal distribution. The unperturbed amplitude and unperturbed phase are denoted by A0 and S0. Following r Refs.35) and 2 2 σ log I = 1. since χ = ln( I / A 2 ) / 2 [20. where U 0 (r ) is the unperturbed field [45]. the complex phase perturbation is written as Ψ1 (r ) = χ + iS1 = ln( A / A0 ) + i ( S − S 0 ) where χ is the logarithm of the amplitude A. This analysis shows that under the Rytov approximation. respectively. (2.23. [20] and [45]. The Rytov approximation yields a log amplitude variance χ 2 for infinite plane wave propagation [22. The first iterative solution to Eq.45]: 2 σ log I = log I − log I 2 2 = 4σ χ (2.45]: r r r r ∇ 2 Ψ1 (r ) + 2∇Ψ0 (r ) ⋅ ∇Ψ1 (r ) = −2k02δn(r ) (2. and S is the phase r of the field U (r ) .33) may be written as the perturbed r r r r field U (r ) = U 0 (r ) exp(Ψ1 (r )) .23Cn2 L11/ 6 k 7 / 6 = σ R (2.
the Fresnel zone size. ρ0.38) [25.46k 2 LCn2 ) −3 / 5 (2.37) The coherence length for a spherical wave is: ρ 0 = (0.099 ρ 0 where ρ0 is defined in Eq. as shown in Table 2. (2. A better comparison of the strength of irradiance fluctuations is provided by the transverse coherence length for a optical wave [13.38) Fried defined a coherence radius r0 = 2. 2. λz .4 for the spherical wave case with Cn2 = 5×1015 m2/3.1].5 for the spherical wave case with Cn2 = 5×1013 m2/3. The Rytov variance of an infinite plane wave helps to define the strength of irradiance fluctuations. intermediate. Fig. Strength of Fluctuations Weak Intermediate Strong Rytov variance 2 σ R < 0.1.3 2 σR ~1 2 σ R >> 1 Table 2. and strong turbulence levels.1: Typical Rytov variance ranges corresponding to weak. 2. A better way to determine the region of turbulence experienced by a propagating optical wavefront is to plot the three relevant scale sizes: the transverse coherence length.24]. 2.2 where σ R is the Rytov variance. and Fig. and the scattering disk size. The coherence length for a plane wave is: ρ 0 = (1.3 for the plane wave case with Cn2 = 5×1014 m2/3. 22 . L / kρ0 . The intersection of the three scale sizes denotes the onset of strong scintillation [19]. These are plotted in Fig. although it is not an absolute measure of turbulence strength.546 Lk 2Cn2 ) −3 / 5 (2.
coherence length. and scattering disk size plotted for a plane wave against propagation distance.56 2 k l0 (2.4: The Fresnel zone size.39) (2.40) For a test range length of L = 863 m and a wavelength λ = 0. (2.37) is bounded by inner scale and outer scale scattering parameters. since optical propagation in the visible and nearinfrared wavelengths near the Earth’s surface is dominated by Fresnel scattering [21].47 cm.6328 µm. the Fresnel length is 2.Eq. 2. The shaded area shows scale sizes that do not contribute to strong fluctuations. which is well away from the scale of either scattering parameter. 23 . Fig. The large eddies in the spectrum set the outer scale scattering parameter: ζ 0 = 2π Lλ L2 0 while the small eddies define the inner scale scattering parameter: 2 ζm = 2 Lκ m Lλ = 5. there are scale sizes on the link contributing to strong fluctuations. For L = 863 m.
2.5: Three scale sizes plotted against the propagation distance for the spherical wave case with L = 863 m. From the graph.Fig. the wavefront sees weak scintillations with contributing scale sizes on the order of the Fresnel zone size. the wavefront experiences strong fluctuations when Cn2 = 5×1013 m2/3.6: Three scale sizes plotted against the propagation distance for the spherical wave case with L = 863 m. 2. 24 . with contributing scale sizes beyond the shaded area. When Cn2 = 5×1015 m2/3. Fig.
29]: 25 . The angular spread of the beam is characterized by the second order moment. When the propagating wavelength is small compared to the scale size of inhomogeneities in the random medium.29. Heuristic theory will not be used in the analysis of aperture averaging. there is no welldefined relationship between amplitude and irradiance fluctuations in strong turbulence.6 Strong Turbulence Theory Multiple scattering of an optical wave by refractive index inhomogeneities results in strong fluctuations of the irradiance. Other scientists used heuristic methods to predict the saturation of amplitude fluctuations [31. 2.2. As the amount of multiple scattering increases with increasing propagation path length. and the r variation of the refractive index is much less than one. no existing analysis method was successful in describing the saturation of scintillation. Although Rytov and other Markov based solutions to the stochastic wave equation were investigated [27]. Although the heuristic methods show good results.30]. irradiance fluctuations saturate and approach a value of one. the fourth order moment with E ˆ in the z propagation direction is [27. which is known as the mutual coherence function (MCF). The fourth order moment defines the irradiance fluctuations of the propagating wave. since the analysis requires information of the value of the logarithm of the amplitude.6.1 Andrews – Prokhorov Asymptotic Analysis The propagation of an optical wave is defined by different moments of the field. New research in perturbation theory led to the development of the asymptotic theory for strong scintillations [27. This phenomenon was first experimentally reported in a Russian journal by Gracheva and Gurvich [26].32].28.
Prokhorov summarized the development of asymptotic analysis using the local method of small perturbations [27].41) The irradiance variance is related to the fourth order moment by [29]: σ I2 = (I − I ) 2 = Γ4 ( z . and satisfies the limit: k 2 ∆ε 2 l 2 << 1 (2. (2. r1 )E ( z .0 ) − 1 (2.0.42) With the lack of an analytical solution to Eq.45) 26 .43) where 2∆ε = (ε − ε )/ ε and ℓ is the scale of the inhomogeneities in the turbulent medium.44) with r r r r D′(r1 + r2 ) D′(r1 − r2 ) r r − V = D′(r1 ) + D′(r2 ) − 2 2 (2. r2 )E * ( z . He constructed a set of equations for the averaged field to which the moments of the field were solutions. r2 ) = E ( z .6. asymptotic solutions for the fourth order moment yield the best results for strong turbulence over long propagation paths. r1 + r2 ) (2.1.1 Andrews Asymptotic Analysis for a Plane Wave The equation for the fourthorder moment of the electric field of a plane wave propagating in a statistically homogeneous random medium is [27]: ∂Γ4 ∂ 2Γ4 = −VΓ4 − i r r k ∂r ∂r ∂z 1 2 ( ) (2.Γ4 (z . The local method of small perturbations required that the phase of the fluctuating wave experiences only small variations over a propagating distance z − z0 . r1 . 2.42).0)E * ( z .
r2 ) = ∫ du exp[− iqu ]Γ4 (z. A solution to the scintillation index can be found using the phase structure function representation [27]: 2 D(r ) = 2.9Cn k 3 zr 5 / 3 (2. m2. (4. Prokhorov wrote the spectral function of irradiance fluctuations as [27]: r r r rr r r 1 M ( z. saturated turbulence conditions [27]: σ I2 = m 2 = 1 + N 3 (α ) D z / k {( for D z / k >> 1 where N 3 (α ) = 2(α +1) 1 πα 2 sin Γ 1 + α 2 Γ 4 α − 1 πα α −1 2 ×2 F1 4 − 1. Γ represents the gamma function. Prokhorov derived the scintillation index. The phase structure function in the inertial subrange approximates a 5/3 power relation [13]. D(u ) = C (ku )α .22.49) 27 . For α = 5/3. in strong. or irradiance variance.17) of Ref. [27].48) where α is the powerlaw exponent of the phase structure function.46) which is the Fourier transform of the fourth order moment with q representing the scale of frequency components in the spectrum. r1. q .α /(α + 1) ( )} ( − 2 / α )( 2 −α ) ) (2. In his analysis. r2 ) 2π (2. N3(α) = 1.α . The full power spectrum in the strong fluctuation regime is presented in Eq.where Γ4 is the fourth moment of the field and D′(u ) is the derivative of the phase structure function.47) ( )( ) {α } (2.α − 1. and 2F1 is the confluent hypergeometric function.
When ρ approaches zero. (2.11]. (2.1. b1 and b2 go to unity.50). When ρ → 0 .51) ρ0 2 where b1 and b2 are scale functions that represent the influence of two different scale sizes on the propagating wave. b1 and b2 go to zero.6. we will refer to Eq. The form of the covariance for a plane wave is given by a twoscale model: 2 1/ 3 ρ 5 / 3 1 kρ 0 5 CI (ρ ) = exp − + N 3 ( 3 ) L [b1 (ρ ) + b2 (ρ )] (2. The generalized asymptotic form for the covariance may also be written in terms of the transverse coherence length [24. 2.27]: σ I2 = 1 + (σ ) 0. as expected.50) For the duration of this thesis. (2.47). Eq.52) 28 . (2.2 Andrews Asymptotic Analysis for the Spherical Wave In spherical coordinates.50) as the Andrews asymptotic model for the plane wave case. the fourthorder moment of the electric field for a propagating spherical wave in a statistically homogeneous random medium is [27]: 2 ∂ 2 Γ4 ∂Γ4 i ∂ Γ4 + = −VΓ4 − 2 ∂r kr ∂ξθ ∂ηθ ∂ξ φ ∂ηφ (2. while as ρ approaches infinity. we find the asymptotic relation between the scintillation index and the Rytov variance for a plane wave in saturated strong turbulence [20.By substituting this into Eq.51) becomes the asymptotic form of the irradiance variance in Eq.86 2 2/5 R (2.
are small and near the equator of the spherical coordinate system with θ ≈ π/2 [27]. there will be an additional constant factor in the term. In saturated strong turbulence.θ .56) where we have defined N3. (2. due to the small angles in the second derivative on the right hand side of Eq. the scintillation index is [27]: 2 σ I2 = msp = 1 + N sp (α ) Dsp r / k { ( )} ( − 2 / α )( 2 −α ) (2. the angles of interest.54) and 2α +1 (α + 1)sin (απ / 2)Γ 2 (1 + α / 2) N sp (α ) = πα 4 × Γ − 1Γ 2 (α − 1)Γ −1 (2α − 2 ) α (2. In the region of the interest of the fourth moment. sp (α ) = (α + 1) − ( 2 / α )( 2 −α ) N sp (α ) (2.ηφ ) are constant differences in angular coordinates due to the conversion from cartesian to spherical coordinates.ηφ ) .53) with the spherical wave structure function related to the plane wave structure function in Eq. φ ) are spherical coordinates with (ξθ .sp as: N 3.52). sp (α ) D r / k {( )} ( − 2 / α )( 2 −α ) (2. ξφ ) and (ηθ .49) by [27]: Dsp (u ) = (α + 1) D(u ) −1 (2.55) The spherical wave scintillation index is related to the plane wave phase structure function by [27]: 2 σ I2 = msp = 1 + N 3. ξφ ) and (ηθ .57) 29 . (2. Therefore.where (r . (ξ θ .
Throughout the rest of this thesis. Churnside built upon Fried’s work and published the first significant application of asymptotic theory to the study of aperture averaging. (2. 2. b1 and b2. ρ0. we assume the phase structure function obeys a 5/3 power law in the inertial subrange.2 Churnside Asymptotic Analysis After the phenomenon of saturation of scintillation was understood. giving N3.As for the plane wave.73 2 2/5 R (2. (2.59) 30 .51). (2. Churnside used the covariance function in Eq. b1. we will refer to Eq. 2.48) to devise a framework to study intensity scintillations when the inner scale is small.6. as: b1 (ρ ) = 7 dxx 4 / 3 J 0 (kρρ0 x / L ) 3∫ 0 1 ρ 5 / 3 b2 (ρ ) = exp − ρ0 (2. The asymptotic scintillation index for a spherical wave in saturated strong turbulence conditions with α = 5/3 becomes [20.1 Churnside Asymptotic Analysis for the Plane Wave Churnside evaluated the twoscale model of the covariance function by representing the two scale functions.2. and b2 used.58) as the Andrews asymptotic model for the spherical wave for convenience.86.27]: σ I2 = 1 + (σ ) 2.6. with the spherical values for N3.58) The asymptotic formula for the covariance of a spherical wave is of the same form as the twoscale model given by Eq.sp(5/3) = 3.
6. (2.51) is evaluated using Churnside’s spherical wave representations for b1: kρρ 0τ 2 b1 (ρ ) = 0. Churnside’s scale functions are compact forms of the functions presented in Chapter 3 of Ref. and Churnside uses limits relevant to aperture averaging.86 0 L 2 I 1/ 3 (2.22 0 L 2 I 1/ 3 (2.61) The Churnside asymptotic approximation is slightly different from that of Sec.60) or.62) where b2 is the same as in the plane wave case. Based on Eq.2 Churnside Asymptotic Analysis for the Spherical Wave The twoscale model of the covariance function in Eq. [1]. Churnside calculated the irradiance variance to be: kρ 2 σ = CI (0) = 1 + 1.1. (2.Both functions go to unity when ρ approaches zero. Different authors have evaluated these equations under different constraints [1. 2.14 2 2/5 R (2. 2.915∫ dxx −1 / 3 (1 − x ) ∫ dττ 4 / 3 J 0 L 0 0 1 ∞ × exp − r [ 5/3 (1 − x ) 5/3 ] (2. and go to zero when ρ approaches infinity. in terms of the plane wave Rytov variance: σ I2 = 1 + (σ ) 1.63) or.6. in terms of the plane wave Rytov variance: 31 .59). The Churnside asymptotic equation for the irradiance variance of a spherical wave is: kρ 2 σ = 1 + 3.29].2.
σ I2 = 1 +
(σ )
5.35
2 2/5 R
(2.64)
The multiplicative constant is slightly higher by Churnside’s calculation, compared to the earlier theory in Section 2.6.1. 2.7 Irradiance Variance Models Valid in Both Weak and Strong Turbulence Conditions 2.7.1 Scintillation Index Model for a Plane Wave Andrews and Philips developed a model to describe plane wave characteristics over the entire range of fluctuation conditions, from weak to strong turbulence [33]. They use the effective Kolmogorov spectrum, which modifies the Kolmogorov spectrum by two filter functions that exclude midscale sizes that have an insignificant effect on the propagating wave in the moderatetostrong turbulence region. The effective Kolmogorov spectrum model is [20,33]:
2 Φ n , l (κ ) = 0.033Cn κ −11 / 3 Gx (κ ) + G y (κ )
[
]
(2.65)
where the largescale filter function that passes only spatial frequencies κ < κ x is:
κ Gx (κ ) = exp − 2
2
κx
(2.66)
and the smallscale filter function, passing only spatial frequencies κ > κ y , is:
G y (κ ) =
κ 11 / 3 (κ + κ y )11 / 6
(2.67)
32
The wavenumber spectrum given by Eq. (2.65) is a twoscale model for generic scale sizes x and y. Likewise, modified Rytov theory may be used to define the scintillation index, σ I2 , in terms of largescale and smallscale scintillations [20,33]:
2 2 σ I2 = exp(σ ln x + σ ln y ) − 1
(2.68)
2 2 where σ ln x and σ ln y are the largescale and smallscale log irradiance fluctuations.
These functions are evaluated in Ref. [33], to define the scintillation index for a plane wave, excluding inner scale effects, as:
2 0.49σ R 2 0.51σ R
σ I2 = exp
1 + 1.11σ R
(
12 / 5 7 / 6
)
+
(1 + 0.69σ )
12 / 5 R
−1 5/6
(2.69)
2 where σ R is the Rytov variance for a plane wave. Eq. (2.69) reduces to the Rytov
approximation in weak turbulence conditions, and the asymptotic model from Eq. (2.58) in strong turbulence saturation conditions. 2.7.2 Scintillation Index Model for a Spherical Wave The scintillation index (SI) model attempts to describe the behavior of irradiance fluctuations over the entire range of turbulence strengths. The model for a spherical wave is developed using the effective Kolmogorov spectrum, using the same twoscale filter functions presented in Eqs. (2.66) and (2.67) for the plane wave case. The small scale and large scale log irradiance fluctuations are evaluated to define the scintillation index of a spherical wave, neglecting inner scale effects, by [20,33]:
2 0.20σ R 2 0.20σ R
σ I2 = exp
1 + 0.19σ
(
12 / 5 7 / 6 R
)
+
(1 + 0.23σ )
12 / 5 5 / 6 R
−1
(2.70)
33
2 where σ R is the plane wave Rytov variance. Eq. (2.70) has the same form as Eq.
(2.69), when we rewrite Eq. (2.69) using the relation between the spherical wave and
2 2 plane wave Rytov variances σ R , sph = 0.4σ R .
2.7.3 Scintillation Index Model using the Atmospheric Spectrum The modified atmospheric spectrum is useful when accounting for the influence of the inner scale, lo, in the calculation of the irradiance variance. The spectrum is still represented by a two scale model, with an enhanced description of the largescale filter function integrated into the model. In order to properly account for the influence of the inner scale, the new largescale filter function is defined as [20]:
2 Gx (κ , l o ) = f (κl o )exp − κ 2 κ x
(
)
κ − 0.25 κ l
7/6
− κ 2 κ = exp 2 1 + 1.80 κ κ l l −κ2 × exp 2 κ x
(2.71)
where the term in the {} arises out of the modified atmospheric spectrum in Eq. (2.25). The smallscale filter function has the same form as in Eq. (2.67). Eqs. (2.71) and (2.67) may be combined to form the Effective Atmospheric Spectrum, which is an extension of the modified atmospheric spectrum [20]:
2 Φ n (κ ) = 0.033Cn κ −11 / 3 Gx (κ , l o ) + G y (κ )
[
]
(2.72)
The irradiance variance of a spherical wave using the spectrum in Eq. (2.72) may be broken down into components attributed to the modified atmospheric spectrum and effective atmospheric spectrum [20]:
34
l o ) 2 12 / − 5 / 6 0. in Eq. l o ) = exp[σ ln x (D. (2.497Cn2 k 7 / 6 L11 / 6 (2.73) 2 where d = kD 4L .69σ mas5 + −1 6/5 2 σ R .62d 2σ 12.73) is defined by [20]: 35 .89 L / kl 2 . the scintillation index becomes o [20]: 2 σ mas ≅ 1 + (σ 6. The spherical wave Rytov variance is given by: 2 σ R .50 − 5 / 6 Ql 2 for σ R . (2. for σ R . sp / 1 + 0. sp 0.51σ mas 1 + 0.76) may be used in Eq. l o ) . The components of Eq.73) depending on the strength of turbulence experienced on the link. (2.2 σ I2 (D.4(1 + 9 / Ql2 ) 11 / 12 11 −1 Ql sin 6 tan 3 Q 4 sin tan −1 l 3 3 3.9d 2 2 + 0.75) ( Q 5 sin tan −1 l 7 / 24 3 4 9 + Ql2 ) where Ql = 10. (2. In saturated strong turbulence.sp5 R σ mas ( ) (2.74) The scintillation index using the modified atmospheric spectrum in weak turbulence is [20]: 2 2 σ mas ≅ 9. (2.75) or Eq. (2.65σ R .76) Either Eq. the part of the scintillation index resulting 2 from the effective atmospheric spectrum.74).57 2 R . Finally. σ ln x (D. sp = 0. sp < 1 + − 0.61 2 1/ 4 l (2.68) are defined in Eqs. sp >> 100 Ql7 / 6 (2. sp Q 7 / 6 1/ 6 l ) 2 .71) through (2.52 (9 + Q ) 2.
the scintillation index model using the atmospheric spectrum presented in Eq.78) Although Eq.75 η + Q l xd η xd − 0.73) may be used to quantify the effects of the inner scale lo on the irradiance variance. (2.20σ R . (2.18d + 0.25 η + Q l xd where 1/ 2 (2.73) will only be used in conjunction with weak turbulence analysis.56 2 1 + 0. sp η xd Ql η + Q l xd 7/6 η xd × 1 + 1. As such. 36 .04σ 2 R .σ 2 ln x (D.77) 7 / 12 η xd = 8. spQl1 / 6 2 (2. l o ) = 0. it is generally accepted that lo has little effect on the strong turbulence irradiance variance [20].
but early experiments did not account for scintillation saturation.21.11]. There has been some previous experimental work [10.16.24]. They eddies are characterized by localized variations in refractive index from the mean. Later experiments did not sufficiently account for background light. This phenomenon is called aperture averaging. Minimal theory is available for the strong turbulence regime [20. and resulted in data markedly different from that predicted by theory.Chapter 3 Aperture Averaging in Weak Turbulence 3. The aperture averaging factor of the receiver is defined as the ratio of the irradiance fluctuations seen at a receiver with diameter D.24.11. Aperture averaging theory has been extensively developed for plane and spherical waves in weak turbulence conditions [1.33]. to those fluctuations seen by a point receiver (or a receiver that is small enough to approximate a point receiver). 37 . the receiver aperture will average across all of these random spatial fluctuations unless the aperture is smaller than the spatial scale of the fluctuations. As a propagating wave moves through these eddies.10. and were limited by the short path lengths under investigation[11]. When a portion of the wave reaches a receiver. they undergo refraction by turbulent eddies. spatially and temporally random irradiance patterns are formed.1 Introduction As electromagnetic waves propagate through the atmosphere.
CI(r) is the spatial covariance of the irradiance. in that a larger aperture will collect more incoming radiation and reduce the likelihood of fading on an optical communications link.2 3.Aperture averaging of optical scintillations is an essential consideration in any receiver telescope design. 3. and the term in the square brackets is the modulation transfer function (MTF) of the circular aperture.2 Weak Turbulence Weak turbulence was defined in Fig. Incorporating aperture averaging techniques helps optimize optical receiver designs. The covariance of the irradiance is [11]: 38 . CI(0) is the variance of the irradiance.4 as the case when the transverse coherence length of the received wave is much larger than the Fresnel length.11]: A= σ I2 (D ) D C I (r ) 16 r r r cos −1 ( D ) − D (1 − D = σ I2 (0) ∫ C I (0 ) πD 2 0 ( 2 2 ) 1/ 2 )rdr (3. 2.1 Data Analysis for Weak Turbulence General Aperture Averaging Form The generic form of the aperture averaging factor for a circular aperture with diameter D is [10. 3.2.1) where r is the distance between two points on the aperture surface. This experiment is built upon the theoretical foundation presented in Chapter 2. A description of the methodology of this experiment is followed by a presentation and discussion of experimental results. In this chapter. an experiment to quantify the system enhancement of an aperture averaged receiver is conducted.2.
2).2) where k = 2π/λ. After the Kolmogorov spectrum is substituted into Eq. and s is a scaling factor that is 1 for a plane wave and z/L for a spherical wave. lo << (L/k)1/2.∞ L κ 2 (L − z )s C I (r ) = 16π 2 k 2 ∫ dκκΦ n (κ )∫ dzJ 0 (κrs ) sin 2 2k 0 0 (3. (3.3) The exact aperture averaging factor for a plane wave is [11]: ∞ A = 21. the Kolmogorov spectrum in Eq. Small lo To calculate the covariance values necessary to input into the aperture averaging factor equation. Φ n (κ ) is the wavenumber spectrum. 3.22) may be used since the inner scale is much smaller than the Fresnel length.23C n k 7 / 6 L11 / 6 = σ I2 ≡ σ R (3. the variance of the irradiance is found as [11]: 2 2 C I (r = 0) = 1.2.6k 1 5/6 L −11 / 6 ∫ J (κDy )[cos 0 0 κ 2 (L − z ) dκκ −8 / 3 ∫ dz sin 2 × ∫ 2k 0 −1 y − y 1− y ( 2 1/ 2 ) ]ydy (3. J0 is a Bessel function of the first kind.4) The inner integral of the Bessel function is given by πJ12 (κD / 2) κ 2 D 2 . A becomes solely a function of (kD2/4L)1/2 [11]: kD 2 A = 8. and we assume a point receiver with r = 0. z is the distance traveled over path length L. (2.2.5) 39 .47 4L 5/ 6 ∫u −14 / 3 4 Lu 2 kD 2 du sin J (u )1 − 4 Lu 2 kD 2 2 1 (3.1 Plane Wave. After a change of variables to let u = κD/2.
the sine term is small and may be replaced by the expansion sin γ ≈ γ − γ 3 6 .6) The D7/3 dependence indicates that the scattering due to turbulence is over a small angle. where kD 2 4L −7 / 6 (3.24. 40 .It is seen that for a small receiver aperture.4). when kD2/4L >> 1. Andrews reported the approximate form for the aperture averaging factor as [34]: kD 2 A = 1 + 1. for the plane wave.932 4L 2 1 −7 / 6 (3. (3. (3. For large apertures.8) = 1 . The total aperture averaging factor in the weak turbulence plane wave region has been approximated by Churnside as [11.33]: 7/6 kD 2 A = 1 + 1. so that [11]: kD 2 A ≈ 1.7) Eq. when kD2/4L << 1.062 4 L At the same location. The light will not be scattered wide enough to miss the receiver aperture entirely. the Andrews approximation is 7% higher than the exact theory assuming a Kolmogorov spectrum. 3.7) is plotted in Fig. Eq. There is a 17% difference in the approximate value of A from the exact theory when kD 2 4L = 1 [11].41 4L −7 / 6 ∞ ∫u 0 −2 / 3 kD 2 J (u )du = 0. A approaches one as expected.1 along with the exact theory formula. Other approximations have been made to the theoretical aperture averaging factor. and show slightly better results.07 4L −1 (3.
3.8Cn L3 l o 7 / 3 (3.10) 41 .1: Aperture averaging factor for a plane wave plotted against the ratio of the aperture radius to the Fresnel zone size.2). (2.23) into the covariance Eq.Fig.23) or the Hill spectrum [19] may be used when the inner scale is large lo >> (L/k)1/2. Large lo Either the Tatarskii spectrum from Eq.9) By following a similar method as in Section 3.2 Plane Wave.2.1. and the results are similar [24].2.686 l o −7 / 3 ∞ l 2u 2 u −2 / 3 J12 (u ) exp − 0. By substituting Eq. (3. (3.7) [11]. (2. The solid line is the exact theory from Eq.2. the aperture averaging factor is [11]: D A = 0. the covariance becomes [11]: 2 − C I (0) = 12. 3. The Tatarskii spectrum is analytically simpler than the Hill spectrum.4) while the dashed line is Churnside’s approximation in Eq. (3.2. and letting r → 0 . and making the substitution u = kD/2.1141 o 2 du ∫ D 0 (3.
10) approaches unity. Small lo The Kolmogorov spectrum may be used to evaluate the covariance of the irradiance (Eq. (3.2. (3.2.When D >> 1.1) is written using the spherical wave covariances [11]: 42 . By setting the two aperture averaging approximations equal to each other.3 Plane Wave.8) = Eq. (3.4 Spherical Wave. (3.2.2. a decision must be made as to which approximation to the aperture averaging factor to use. sp (3. Eq.21 l o −1 (3. (3. and the large inner scale approximation when lo > 2.19 lo −7 / 6 for 0 ≤ D / l o ≤ 0. (3.8) when kD2/4L < 1 [34]. the exponential in Eq.5 (3.11) is [11]: 2 D A = 1 + 2.12).497C n k 7 / 6 L11 / 6 = σ I2 ≡ σ R . lo on the Order of the Fresnel Length In an intermediate regime.2)).73(L/k)1/2. Churnside has recommended using the small inner scale approximation when lo < 2.73(L/k)1/2. (3. they are found to be equal when lo = 2.12) 3.13) The exact form of the aperture averaging factor in Eq. and A can be approximated by [11]: 7/3 D A = 1 + 2. along with s = z/L for a spherical wave. and r = 0 [11]: 2 2 C I (0) = 0. when the inner scale is similar to the Fresnel length.73(L/k)1/2 [11].11) An improved approximation to Eq. Andrews recommends using Eq. 3.
2.214 4L −1 (3. The solid line is the exact theory given a Kolmogorov spectrum. Where kD 2 4L = 1 .1. 3. (3.24].4k 5 / 6 L−11 / 6 ∫ dKK − 8 / 3 ∫ dz sin 2 2kL 0 0 KDzy 2 −1 × ∫ dyyJ 0 cos y − y 1 − y L 0 1 [ ( ) 1/ 2 ] (3. Fig. it can be approximated as [11]: 7/6 kD 2 A = 1 + 0.1 ∞ K 2 z (L − z ) A = 53.2: Aperture averaging factor for a spherical wave with small inner scale. 3.15) The exact and approximate forms of the aperture averaging factor for a spherical wave with small inner scale are shown in Fig.15) [11].14) After evaluating the aperture averaging factor integral as in Section 3.2. while the dashed line is the approximate formula of Eq. the approximate version predicts the amount of aperture averaging to be 86% less than the theory [11. 43 .2.
5 times the Fresnel length.3. (3. 250.28Cn L3 l o 7 / 3 = σ I2 (3.17. After the development of plane wave theory of optical propagation through turbulence.15) is equal to Eq. Andrews recommends using the small inner scale result when kD2/4L <1.17) when lo is 1. the advent of the laser allowed experimental studies of aperture averaging.5 Spherical Wave. and 1000 m paths to avoid 44 .6 Spherical Wave. Churnside recommends using the small inner scale approximation when lo < 1.3 Aperture Averaging Experiment Experimental aperture averaging data is necessary to verify previously developed theory [1.2. 3.13. Eq.5(L/k)1/2 [11].20] and ensure that modern optical communication systems are correctly optimized.16.16) The aperture averaging factor is approximated by [11]: 7/3 D A = 1 + 0.17) 3. and r = 0 [11]: 2 − C I (0) = 1.3. A more recent experiment by Churnside [11.2.2) with the Tatarskii spectrum. (3. Large lo By evaluating the covariance of the irradiance in Eq.24] used 100.2.109 l o −1 (3. these experiments were conducted over very long paths and were susceptible to the effects of saturation of scintillation [17].2.5(L/k)1/2 and the large inner scale approximation when lo >1. 500.2. lo on the Order of the Fresnel Length By following the argument in Section 3. However.2. (3.
Over longer paths. with a 0. 3. The scintillometer used in the experiment measured turbulence over a 250 m path.8 nm. Other shortcomings of this experiment include the lack of simultaneous background light measurements and a use of only six apertures.70 mm output diameter and 1. The beam passes through an optical chopper and a 30x Melles Griot beam expander. Many commercial free space optical communication units use apertures with diameters larger than 5 cm. so breadth of data collected by Churnside in Ref.15mrad beam divergence. which is on the order of the time that the turbulence on a specific path through the atmosphere remains constant. for this experiment.3. [11] is insufficient for practical use. the beam diameter at the receiver is ~1. The optical chopper serves an important purpose during the data acquisition in that it allows us to calculate mean background light levels at the same time that intensity data is being recorded.24].1 Aperture Averaging Transmitter and Receiver Systems The aperture averaging transmitter uses a 21 mW JDS Uniphase HeNe laser.saturation effects. ranging from 1 mm to 5 cm. A photograph of the aperture averaging transmitter is shown in Fig. and produced results in reasonably good agreement with sphericalwave theory over short paths of 250 m [11. 45 .3.5 m. the experimental data tended to diverge from theory. The optical chopper is operated with a frequency of 3. The beam expander is adjustable so that the beam diameter at the receiver may be tuned. 3. which reduces confidence in the precision of the Cn2 data. The laser operates in a single mode (TEM00) at 632.636 ms. and not the path length under test.
which may have some affect on the homogeneity of the turbulence measurements.V.3: Aperture averaging transmitter on the roof of the A. College Park. Williams is approximately 14 meters above the ground. with a few trees and grassy fields. An aerial photograph of the path. with the line of sight indicated. The laser light propagates over an 863 m path between the A.Fig. Williams Building and the Chesapeake Building.4. 3.V. The transmitter on the roof of A. while the receiver in the Chesapeake Building is approximately 12 meters above ground. The terrain propagated over is mostly asphalt parking lots. Williams Building at the University of Maryland. 46 .V. is shown in Fig. 3. One part of the path passes near the rotunda on the east side of the Comcast Center.
which is smaller than the Fresnel zone size for the path.Figure 3.4: An aerial photograph of the propagation path (dotted line). 3. so only light parallel to the propagation path at a wavelength near 632. a point receiver and a variable aperture receiver. The minimum aperture size.8 nm will pass. Satellite photograph taken in 2002 by GlobeXplorer.8 nm with a 10 nm passband width.5. required to measure A = 1 in weak turbulence. The point receiver is an effective scintillometer and is used to calculate pathaveraged Cn2 measurements. The aperture diameter of the point receiver is 5 mm. The filter removes any stray light from interfering with the operation of the scintillometer. is the diameter 47 . The receiver system is composed of two receive apertures. λL = 2. Light is first filtered through a laser line filter at 632.34 cm . The laser line filter is angle dependent. shown in Fig.
The transimpedance gain is set when the photodetector is illuminated through the smallest aperture diameter. An aperture stop of sizes ranging from 1 cm to 16 cm is placed behind the plane side of the lens. with a responsivity of 0. to limit the aperture diameter. Since neutral density filters are not angle dependent. The FFD200 has a 20 mm2 active area with a responsivity of 0. After passing through the scintillometer receive aperture. Dmin = 1. they will not affect the distribution of irradiance fluctuations.43 A/W at 632. 48 . while the receive optical power is controlled by changing the neutral density filter transmittance.2 where kD 4L = 1 . light is detected by a PerkinElmer FFD100 Si photodetector. a New Focus neutral density filter wheel is placed in front of the photodetector. The variable aperture receiver is a Rolyn Optics 20 cm planoconvex lens. To ensure that the receiver does not saturate. Typical oscilloscope traces of the aperture averaged and scintillometer signals are shown in Fig.43 A/W at 632. thereby allowing the electronic gain to remain constant. The incident beam is then received by a Perkin Elmer FFD200 Si photodetector and amplified through a variable transimpedance amplifier. Since the scintillometer aperture diameter is smaller than the minimum aperture diameter for A = 1. The FFD100 has a 5.6.8 nm.8 nm. The signal is than recorded and processed in LabVIEW. 3. For this experimental setup. the baseline Cn2 measurements are true pathaveraged values.86 cm.1 mm2 active area. the filter wheel is necessary to avoid detector saturation and measure the intensity variance accurately. At large receive apertures. The detected signal is amplified by a variable transimpedance amplifier and recorded by LabVIEW.
Fig.5: Diagram of the aperture averaging receiver setup. 3. Fig. 49 . The scintillometer channel uses a 5 mm diameter receive lens. The signal detected on Channel 2 shows smaller intensity fluctuations. while Channel 2 is detected by the aperture averaged receiver.6: An optically chopped signal propagated over 863 m and viewed on the Tektronix oscilloscope. Channel 1 is the signal detected by the scintillometer. 3.
3. 3. from the 50 . respectively. Fig. The LabVIEW program calculates irradiance statistics relevant to determining the aperture averaging factor over a 1 minute sampling interval. so the turbulent channel is quasistationary.2 LabVIEW Data Acquisition A customized LabVIEW program. This time difference is well below the reordering time of the turbulent atmosphere.3. 3. σ I2 . samples the optically chopped 275 Hz signal at a rate of 3000 samples/sec.5 µsec and 5 µsec.8 is a flowchart representing the data acquisition process for one channel. The chart explains the methodology used to determine the normalized irradiance variance. Data collection for each aperture is typically done in 5 to 15 minute intervals. 3.7. whose front panel is shown in Fig. Fig. The maximum and minimum interchannel sampling delays for two channels are 62.7: Front panel of the LabVIEW program designed to calculate and record irradiance statistics for the aperture averaging experiment.
(3. the variance of the high values for that acquisition period is calculated. 3.measured waveform data.8. however results were much more erratic. The waveform type makes signal analysis easier because the sampled data is synchronized with the sampling time. The one 1min path averaged Cn2 is then calculated from the normalized irradiance variance σ I2 using the equation for the plane wave Rytov variance. and sampling interval (∆t). as in Eq. These values are then written to a data file every minute. LabVIEW also includes an extensive functionality to take advantage of the waveform type. The longer 51 . Table 3. The waveform type in LabVIEW is a new data type introduced in LabVIEW 6. which gives σ I2 . due to the naming conventions in LabVIEW. Simultaneously. The variance of the high values is then normalized by the square of the mean intensity. The mean of the low values is the average background signal. The background variance values are only output to ensure that there is no significant baseline wander on the measured signal that could distort the measured irradiance statistics.1 shows the 13 values that are output by LabVIEW into a data file.0 that stores information on sampled voltage (y). which is then averaged over a 1 minute interval to give the mean intensity. This background signal is than subtracted from the recorded high values. 1. The LabVIEW program requires a large storage buffer to hold 60 seconds worth of samples.1). the detected chopped signal is sorted into high and low voltage values. Shorter averaging times were also used in the initial experiments. From Fig. measured at 3000 samples per second. acquisition start time (t0). 0 and Ch. The two channels recorded are called Ch. The aperture averaging factor. A. is calculated each sampling interval from the ratio of the normalized intensity variances of the two channels.
interval. Fig. 3. The other half of the samples are used to determine the mean background level. 52 .averaging time ensures that there are a sufficient number of samples so that lognormal statistics are obeyed [35]. Only half the number of samples acquired may be used in the calculation of the irradiance variance. The diagram explains how intensity and background levels are determined for the sampled signal averaged over a 1 min.8: Flowchart of the data acquisition process in LabVIEW for one channel. since the signal is optically chopped.
0 1.1 1. Although the terrain along the path is mostly flat with grass of pavement. while Channel 1 represents the scintillometer signal.0 Aperture Averaging Factor. 1 . mean of high values with background level subtracted out mean of low values variance of low values Cn2 Ch. 0 . variance of high values normalized by I I Ch . Channel 0 represents the aperture averaged signal. 0 σ I2 Ch. mean of high values with background level subtracted out mean of low values variance of low values Table 3. 1 Average Background Ch.1 σ I2Ch. there are a few trees and two buildings that impinge 53 . 0 Background Variance Ch.23k 7 / 6 L11 / 6 Irradiance Variance Ch. 1 Background Variance Ch. variance of high values normalized by I I Ch.0 σ I2 Ch. 1 σ I2Ch.0 .1 2 Ch.1 . approximately 12 m. 1 σ I2Ch.Time Signal Ratio measured to 1/1000th of a second I I Ch. Here.0 Average Intensity Ch. 0 σ I2 Ch.23k 7 / 6 L11 / 6 Irradiance Variance Ch. A Cn2 Ch.3 Fetch Effects The height of the path above ground.3. 0 Average Background Ch.1: List of 13 values output per averaging interval in LabVIEW data file.1 Average Intensity Ch. limits the strength of turbulence experienced on an average day.0 2 Ch. 3.1 Ch.
Therefore. The plane wave scintillation index diverges 2 from its Rytov predicted weak turbulence value when σ R ≥ 0. To be able to neglect fetch effects. A Cn2 of 2 1014 m2/3 changes the Rytov variance by a factor of 100 and we find that σ R = 0.1. 3. The aperture averaging factor. Three accepted approximations to the aperture averaging factor are calculated and plotted in Fig. A Cn2 value of 1016 m2/3 corresponds to σ R = 0. For our path of 863 m.37]. which could result in fetch effects. for different strengths of turbulence indicated by Cn2. The rule states that there will be an internal boundary layer established at a height that is 1/100th of the half width of the path.0043.37]. Recall that the Fresnel zone size is the dominant scale size in cases of weak turbulence. 3. Since that propagation path is higher than this internal boundary layer.3 m.4 Plane Wave Experimental Results Theory and approximation of the plane wave aperture averaging factor with small inner scale was presented in Sec. 54 . the height above ground must be larger than the boundary layer 100to1 rule for fetch effects [36. 3. we can assume that we are nearly free of fetch effects [36.9. A.on the path. this height is approximately 4. is typically plotted against the ratio of the aperture radius to the Fresnel zone size.2. the scintillation index model does predict different aperture averaging results from the accepted approximations to weak 2 turbulence theory.43.5 .2. The scintillation index model is an attempt to define an irradiance variance that is valid across all turbulence levels.
For smaller apertures.7). (3. we expect better agreement with the spherical wave aperture averaging models.Fig.10. Since the laser is diverged.8). the experimental aperture averaging data seems to retain high values. the SI model predicts less aperture averaging than weak fluctuation theory for given values of receiver aperture. As seen from Fig. 3.10. there was no significant 55 . The Andrews scintillation index model is the only model that takes into account the strength of turbulence in the weak fluctuation region. Experimental data plotted against the ratio kD 2 4L is shown in Fig. The mean data is plotted with error bars of one standard deviation. and the scintillation index (SI) model for various values of Cn2. whereas the approximate models drop off. the Andrews approximation Eq. (3. 3. When our experimental data was analyzed.9: Plane wave aperture averaging factor plotted as the Churnside approximation Eq. The data does follow the trend predicted by theory. 3.
so it was assumed that the data would be a better fit to the spherical wave theory. and certainly not the substantial variation predicted by the scintillation index model.2 mrad divergence of the HeNe was used directly. It is apparent from Fig. The Churnside data was measured on a 500 m path. although Cn2 was measured over only half the path using an incoherent scintillometer [11.5 m above ground.11.10: New mean experimental data plotted with error bars. along with approximations to the aperture averaging factor. 3.24].dependence of A on Cn2 in the weak turbulence region. It was assumed that due to the flatness of the terrain that the measured Cn2 was indicative of the path averaged value. The mean data is plotted along with the experimental Churnside data in Fig. Fig. The laser and scintillometer were propagated over a plateau of grassland at a height of 1. 3.11 that the plane wave theory does not sufficiently 56 . 3. The 1.
3.2.characterize the amount of aperture averaging taking place. 3.4. especially at larger aperture diameters. we expect the data to be in good agreement with the spherical wave analysis. Since the beam divergence in this experiment is ~ 1 mrad. Fig.11: New mean aperture averaging data plotted with Churnside experimental data taken over a 500 m path. 3. Neither the plane wave Churnside approximation nor the SI model are a good fit to the data. thereby underestimating the amount of aperture averaging taking place. this knee overpredicts the exact theory given a Kolmogorov spectrum.2.12 compares the Churnside approximation to the scintillation index model for the aperture averaging factor.5 Spherical Wave Experimental Results Spherical wave weak turbulence theory and approximation was presented in Section 3. 3. 57 .2. The Churnside model predicts a higher knee than the scintillation index model. Fig. As shown in Fig.
(3. It also shows that the Churnside approximation is a better fit to the spherical wave data than the scintillation index model.13. In contrast to what is predicted by the exact theory. we see a higher knee in the data that the theory predicts near kD 2 4L ≈ 1 .7).Fig. 3. 3. The Churnside approximation. Experimental mean aperture averaging data with error bars is shown in Fig. predicts a higher knee in the curve than the SI model.12: Comparison of two different models for the spherical wave in weak turbulence. 58 . Eq. This supports a generally accepted notion that aperture averaging in weak turbulence is independent of Cn2.
Fig. 3.13: Experimental mean aperture averaging data plotted against the ratio of aperture radius to the Fresnel zone size. The Churnside experimental data measured over a 500 m path is displayed along with our mean data in Fig. 3.14. The Churnside data does not exhibit the same knee that our new data shows. Churnside assumed that a high knee is a result of a violation of propagation uniformity due to cloud cover over part of the path [11]. This seems to be an unlikely explanation for the knee in our new data, since the data was taken under clear weather conditions with wind speeds less than 4 m/s and low relative humidity. It is possible that a few trees and a building near the propagation path adversely affect the homogeneity of the atmosphere in that area. There may be additional scintillations induced near the transmitter since the laser propagates about 1.5 m above the roof of A.V. Williams over a length of 10 m.
59
Fig. 3.14: Comparison of our new mean data with the Churnside data. It is not apparent that the knee in the new data is due to inner scale effects. A large inner scale (larger than the Fresnel zone size), should reduce the aperture averaging factor, and not increase it as we are seeing. The Fresnel zone size is on the order of 1 cm, making it unlikely that the inner scale ℓo is larger than 1 cm for this path length and height. However, an inner scale that is smaller than the Fresnel zone size will still impact the data in weak turbulence conditions. Therefore, irradiance variance models including an ℓo dependence should be investigated. 3.6 New Aperture Averaging Model Data was fit to the generalized form of the Churnside model in Eq. (3.15). A must be 1 when the aperture diameter becomes infinitesimal, so only the coefficient of
60
the
kD 2
4L
term was varied. The genfit function in MathCad allows arbitrary
functions to be fitted to data matrices. Using this function, a fit to the measured aperture averaging data was found to be:
7/6 kD 2 A = 1 + 0.196 4L
−1
(3.18)
This new recommended fit, along with the new data and the Churnside approximation, given by Eq. (3.15), are plotted in Fig. 3.15.
Fig. 3.15: New weak turbulence fit (Eq. 3.18) shown with new data and the Churnside model (Eq. 3.15). The limitations of the
kD 2
4L
term do not pick up on the nuances of the data;
more specifically, it does not predict the knee that we see in Fig. 3.14. Including a
61
plotted along with the Churnside approximation in Eq.16: New fit to weak turbulence data with an additional square root term. Fig. 3. ℓo Although inner scale could not be measured during the aperture averaging experiments. and we find the approximate formula using MathCad: 7/6 7 / 12 kD 2 kD 2 A = 1 + 0.162 4 L −1 (3. 3.square root term in the model does pick up on the knee.7 Determination of the Inner Scale of Turbulence.19). given by Eq. (3.15).19) kD 2 Fried had previously proposed that a fit could be improved with a 4L although Churnside discounted the term as unnecessary [16]. (3.266 4 L − 0. ℓo may be determined by fitting available aperture averaging models to the 62 . 7 / 12 term.
27 cm. spherical wave models are expected to be in better agreement with data acquired on the ChesapeakeAVW test range. the new data is fit to the plane wave aperture averaging factor given a large ℓo given by Eq. In our previous analysis. against the ratio of the aperture diameter to the inner scale. The fit determined ℓo = 5.11) using ℓo = 5.27 cm.27 cm in Fig.17: Eq. Similar to the development of a new aperture averaging model in Section 3.11). (3. 3.6. 63 . The new data is plotted along with Eq. 3. The dashed line represents Eq.17. In a first attempt to extract a value of ℓo. 3.11) plotted using the curve fit value of ℓo = 5. a value that is approximately an order of magnitude larger than what would be expected in weak turbulence conditions. the genfit function in MathCad is used to determine ℓo. Fig.collected new data. (3. (3.11). the assumption that the diverged beam approximates a plane wave is not.17 is reasonable. Although the fit in Fig. (3.
45 cm using Eq. The fit resulted in a predicted ℓo = 1. 3.17) plotted along with the new data. Since a reasonable value of ℓo has not been determined using the standard approximation to the aperture averaging factor.The new data was next fitted to the aperture averaging factor for a spherical wave given a large inner scale.45 cm.17) with ℓo = 1. Recall that the modified atmospheric spectrum and effective atmospheric spectrum presented in Section 2.17). (3.18: Spherical wave aperture averaging factor for ℓo = 1.61]. given by Eq.3 compose the best model to date for determining the influence of inner 64 .17). The new data along with Eq.7. (3. 3. (3. Fig.18. a more involved model for the dependence of the aperture averaging factor on ℓo is investigated.45 cm is shown in Fig. The dashed line represents Eq. This value is on the high end of the range of inner scale measurements reported in the literature [60. (3.
73) with aperture diameter D to Eq. 65 .73) with D = 0: A= σ I2 (D. l o ) (3.19: New data is plotted using the predicted value ℓo = 4. Using these spectra and assuming spherical wave propagation.scale on the irradiance variance.008 mm are plotted in Fig. (2.008 mm. which is represented by the dashed line.19. l o ) σ I2 (0. 3. The new data along with the atmospheric spectrum aperture averaging factor for ℓo = 4. the aperture averaging factor is calculated by taking the ratio of Eq. Fig. the aperture averaging factor using the atmospheric spectrum predicts ℓo = 4. (2. The inner scale was determined using the atmospheric spectrum.20) By using the genfit function in MathCad.008 mm. 3.
8 Conclusions The experimental results presented in this chapter show the behavior of the aperture averaging factor in weak turbulence. 3. (3. we expect the Fresnel zone size to be larger than ℓo.11) and (3.19 does not follow the knee of the data. 3. The data is consistent with models developed from the Kolmogorov. The atmospheric spectrum mandates that the inner scale always impacts the irradiance variance in weak turbulence conditions. From Fig. Referring back to the discussion in Sec. 3. Either way. especially when ℓo is not measurable.18 does. is proven. we can ascertain that even small values of ℓo will reduce A. the atmospheric spectrum aperture averaging model with ℓo = 4. In our case. Although the model is much more complex than the Churnside approximations. The atmospheric spectrum produces reasonable results because it is not constrained by the Fresnel zone size. its usefulness in determining the inner scale.The primary reason the Churnside plane wave and spherical wave approximations to the aperture averaging factor.5.19. It is still unclear what physical mechanism contributes to the knee. as the Churnside spherical wave approximation in Fig. 3. 3. Tatarskii.17).008 mm in Fig. did not produce accurate inner scale values is that the models are meant to be used when the inner scale is larger than the Fresnel zone size. and atmospheric spectra for spherical wave 66 . It seems likely that the terrain uniformity and slightly offhorizontal path could act as forces counteracting the impact of ℓo. given by Eq. it is clear that improvements must be made to the atmospheric spectrum models of the irradiance variance in order to capture the physical nature of the relationship between the aperture averaging factor and ℓo.
The aperture averaging measurements presented here consider a variety of receive aperture diameters.propagation. simultaneous measurement of background light levels. Inner scale was also estimated from the data. and concurrent measurement of pathaveraged Cn2 values. This experiment was carefully planned out. 67 . and the results are the most accurate to date.
27.1).21. given by Eq.2.40. These include the Rytov variance. 4. nonmeasurable quantities must be evaluated for this experiment. The theory relies on the assumption that the irradiance 68 .33.2.2 Plane Wave Analysis 4. and scattering disk size.2.Chapter 4 Aperture Averaging in Strong Turbulence 4. inner scale.27].29.1 Introduction Aperture averaging in strong turbulence is extremely difficult to study because of the lack of a comprehensive theory for strong intensity scintillations including saturation effects [1.11. (3. In order to make useful comparisons of the experimentally determined aperture averaging factor with previously developed theory and numerical approximations.1 Data Analysis in Strong Turbulence General Aperture Averaging Form The general form of the aperture averaging factor for the spherical wave has the same representation as in weak turbulence theory. transverse coherence length.2 4. The spatial covariance terms are represented by their strong turbulence theory counterparts.41]. Fresnel zone.2.20.1 Churnside Asymptotic Analysis for the Plane Wave The aperture averaging factor in strong turbulence may be evaluated using asymptotic theory [11. 4.
1. and b1 and b2 are functions that go to unity as ρ goes to zero and go to zero when ρ goes to infinity. 2 σ R using: 69 . ρ0.2).2) By solving for ρ0 in Eq. consequently. as discussed in Section 2. the variance of the irradiance is related to the transverse coherence length by: σ I2 = C I (ρ = 0 ) = 1 + 1.6. and the transverse coherence length. to find predicted values of Cn2 and. (4.22 ( ) 2 kρ 0 L 1/ 3 (4. we can find the predicted asymptotic value of ρ0 by using measured values of σ I2 . given in Eq. because it uses a twoscale model to describe behavior in the inertial subrange and the transition into the dissipation region. for plane wave propagation. the scattering disk. For zero inner scale.3).1) where N3 is a constant equal to 1. Asymptotic theory is the best theory to describe behavior in the saturation regime. The asymptotic method uses three terms from the series expansion of the covariance function [27]. Specific descriptions of the formulas and the constant are found in Sec 2. Asymptotic theory includes the impact of two scale sizes. ρ0.6. (4. Asymptotic theory is expected to be valid only where the irradiance variance is close to unity.22 for a plane wave. 27.variance is the sum of 1 and a perturbation term. We can also use the transverse coherence length. The asymptotic form of the covariance function for a plane wave with small inner scale is [11]: C I (ρ ) = exp − [ ( ) ]+ N ( ) ρ 5/3 ρ0 1 2 3 2 kρ0 L 1/ 3 [b1 (ρ ) + b2 (ρ )] (4. L/kρ0.1 and Ref. saturated pathintegrated turbulence [11]. indicating a region of strong.
3) 1. (4.5) 1 2 2 56 L σ − 1 kDρ 0 x A2 = 2 2 2 I 2 ∫ dxx − 2 / 3 J12 3k D ρ 0 2σ I 0 2L 16 σ I2 + 1 A1 = dyy cos −1 y − y 1 − y 2 π 2σ I2 ∫ 0 1 [ ( ) 1/ 2 ] By evaluating the two components of the aperture averaging factor for large and small diameters.4) Churnside evaluated the aperture averaging factor as a sum of two terms.6) The approximate aperture averaging factor. (4.23k 7 / 6 L11/ 6 (4.1): Dy 5 / 3 exp − ρ0 (4.908(2D ρ 2 2σ I [ 0 )] −1 + 7 / 3 −1 σ I2 − 1 1 + 0.6). given in Eq.162(kρ LD ) 2 2 2σ I 0 [ ] (4. for a plane wave with a small inner scale is plotted in Fig.1.46k 2 LCn2 ) and 2 Cn = 2 σR −3 / 5 (4. 70 . 4. based on the two scale functions b1 and b2 in Eq. using AVWChesapeake test range characteristic values. the approximate aperture averaging factor is given as [11]: A= σ I2 + 1 1 + 0.ρ 0 = (1.
1: The plane wave aperture averaging factor A vs.Fig.1 is beyond the range of the chart. The legend indicates different values of irradiance variance σ I2 .1.25 and 1. values of σ I2 closer to unity indicate stronger levels of turbulence. the ratio of the aperture radius D/2 to the transverse coherence length ρ0. The graphs for various irradiance variances in Fig. is especially evident in the curves with σ I2 = 1.5. the separation is 1815. the width is 14. 4. may be determined.1. 4. For σ I2 = 1. The impact of the two scale sizes. the scattering disk and the transverse coherence length. which is the separation of the two scales. there is more aperture 71 . According to asymptotic theory.5 in Fig.1 show the trend of the aperture averaging factor in strong turbulence. 4. For σ I2 = 1. Initially.1. which indicates that the drop off in Fig.5. the width of the plateau. 4.2). From Eq. The location of the plateau is also at a lower value of A for σ I2 = 1. for L = 863 m and λ = 632. where turbulence strength increases in the asymptotic limit of σ I2 → 1 .8 nm. (4.
2 for a variety of irradiance variance values.86 2 2/5 R (4. and allow the weak turbulence aperture averaging factor to catch up to the strong turbulence value. This model for A reduces the width of the plateau on the aperture averaging curve.2.2.27]: σ I2 = 1 + (σ ) 0. 72 . because A drops off faster in strong turbulence than in weak turbulence.2 Andrews Asymptotic Analysis for the Plane Wave Andrews approached the theory of aperture averaging using the effective Kolmogorov spectrum and modified Rytov theory to determine a model for aperture averaging valid over the range of weaktostrong scintillations. Eq. This approach will be fully developed in the following section. (4. 4. (4. from which the transverse coherence length for a plane wave is determined and substituted into Eq.averaging in strong turbulence conditions. 4. The plateaus cause the aperture averaging factor to remain high over a wide range of aperture diameters. D/2ρ0 will always be small enough so that the link will experience more aperture averaging in strong turbulence conditions.7) has been plotted in Fig. is larger than the inner scale. Given a value of σ I2 .7) which is valid when σ I2 >> 1 and the spatial coherence radius.7) may be solved for Cn2. the transverse coherence length and the scattering disk size. For most practical purposes. the asymptotic relation of the irradiance variance for a plane wave to the Rytov variance developed from Prokhorov is used [20. ℓo. In the strong fluctuation region. plateaus form in the strong turbulence aperture averaging factor from the influence of the twoscale sizes. At larger values of D/2ρ0. (4. ρ ≈ 2 ρ 0 . The aperture averaging factor given by Eq.6) to calculate A.
Fig. 4.2: Aperture averaging factor using the Andrews asymptotic model vs. the ratio of aperture radius to the transverse coherence length. 4.2.2.3 Scintillation Index (SI) Model for the Plane Wave Andrews and Philips developed a model to describe plane wave characteristics over the entire range of fluctuation conditions, from weak to strong turbulence [33]. They use the effective Kolmogorov spectrum, which modifies the Kolmogorov spectrum by two filter functions to exclude midscale sizes that have an insignificant effect on the propagating wave in the moderatetostrong turbulence region. The effective Kolmogorov spectrum model is [20,33]:
2 Φ n , l (κ ) = 0.033Cn κ −11 / 3 Gx (κ ) + G y (κ )
[
]
(4.8)
where the largescale filter function that passes only spatial frequencies κ < κ x is:
κ2 Gx (κ ) = exp − 2 κ x
(4.9)
73
and the smallscale filter function, passing only spatial frequencies κ > κ y , is:
G y (κ ) =
κ 11 / 3 (κ + κ y )11 / 6
(4.10)
Likewise, modified Rytov theory may be used to define the scintillation index, σ I2 , in terms of largescale and smallscale scintillations [20,33]:
2 2 σ I2 = exp(σ ln x + σ ln y ) − 1
(4.12)
2 2 where σ ln x and σ ln y are the largescale and smallscale log irradiance fluctuations.
These functions are evaluated in Ref. [33], to define the scintillation index for a plane wave, excluding inner scale effects, as:
2 0.49σ R σ (L ) = exp 1 + 1.11σ 12 / 5 R 2 I 2 0.51σ R
(
)
7/6
+
(1 + 0.69σ )
12 / 5 R
−1 5/6
(4.13)
2 where σ R is the Rytov variance for a plane wave. Eq. (4.13) reduces to the Rytov
approximation in weak turbulence conditions, while it reduces to the Andrews asymptotic model in saturated strong turbulence conditions. 4.2.2.4 Comparison of Strong Turbulence Models for the Plane Wave The three irradiance variance models, along with weak turbulence theory
2 ( σ I2 = σ R ), are plotted in Fig. 4.3. The Andrews and Churnside asymptotic models
follow a similar trend, with the Andrews asymptotic model matching the scintillation
2 index model when σ R > 36 [33]. In the weaktomoderate fluctuation regime with 2 0.5 < σ R < 2 , the scintillation index model has a lower slope than what weak turbulence
theory would predict. It is well known that weak turbulence theory is most accurate
74
2 when σ R = σ I2 < 0.3 , and the scintillation index model accounts for the deviation of data
in this region from weak turbulence theory. The coherence length is expected to follow asymptotic behavior in the saturation region. When the irradiance variance σ I2 approaches 1, asymptotic theory predicts that the transverse coherence length will become infinitely small. This would require the presence of unreasonably high turbulence levels on the link. Fig. 4.4 shows the behavior of the transverse coherence length based on three different models for the irradiance variance in strong turbulence. The scintillation index model is plotted in the
2 onsetofstrong turbulence region, where 0.5 < σ R < 2 . In the onset of strong
turbulence, ρ0 is in the range of 2 mm to 6 mm for a path length of 863 m and λ = 632.8 nm.
Fig. 4.3: Irradiance variance, or scintillation index, of a plane wave versus Rytov variance. Asymptotic models in the saturation region are shown. 75
although N3. The transverse coherence length for the spherical wave is defined by: ρ 0 = (0.545k 2 LCn2 ) −3 / 5 (4. 4.2.1 Churnside Asymptotic Analysis for the Spherical Wave Aperture averaging theory for spherical wave propagation with small inner scale is also evaluated using asymptotic theory as in Sec. σ I2 . b1.2.3.2. (4.1).Fig.14) 76 . 4. 4.4: Transverse coherence length based on strong turbulence models plotted against irradiance variance. and b2 take on spherical wave representations.2.1. The spherical wave analysis uses the same covariance function as presented in Eq.3 Spherical Wave Analysis 4.
16) duu − 2 / 3 J12 (u )∫ dxx 2 (1 − x ) 2 ∫ 2σ I 0 0 2 L 5 / 3 u 5 / 3 x5 / 3 × exp − kDρ 0 ∞ As in Eq. The two scale functions b1 and b2 are also modified from the plane wave situation.36).86 0 L 2 I 1/ 3 (4. while the term A2 represents scale sizes larger than the scattering disk.86 for the spherical wave case. (4. the term A1 represents scale sizes smaller than ρ0. although their behavior as ρ goes to zero and infinity are the same. and is given by [11]: 77 . L/kρ0.15).The coefficient in Eq. Churnside evaluated the strong turbulence spherical wave aperture averaging factor as a sum of two terms which are strongly influenced by b1 and b2 in the evaluation of the covariance equation [11]: 16 σ I2 + 1 A1 = dyy cos −1 y − y 1 − y 2 2 ∫ π 2σ I 0 1 [ ( ) 1/ 2 ] Dy 5 / 3 exp − ρ0 1 2L A2 = 3. The total aperture averaging factor is A=A1+A2. (2.15) 2 The irradiance variance is related to the plane wave Rytov variance σ R by using the form presented in Eq.14) is the only difference between the plane and spherical wave cases. The constant N3 also takes on a new value of 3. otherwise the behavior of ρ 0 is the same.66 kDρ 0 7/3 σ I2 − 1 −1 / 3 (4. (4. the irradiance variance is related to ρ0 by [11]: kρ 2 σ = 1 + 3. For negligible inner scale.
the Fresnel zone size is 240ρ0.17) is plotted for various values of irradiance variance in Fig.5. When σ I2 = 1. 4.908 A= 2ρ 2σ I2 0 −1 7 / 3 −1 (4. is wider than in the plane wave case. For σ I2 =1. and the plateau width is 3681.613 2 2σ I 2L 2 I Eq.1.2 D σ I2 + 1 1 + 0. For σ I2 = 1. The separation of scale sizes for the spherical wave case. Fig.4ρ0. 1. and 1.17) + σ −1 kDρ 0 1 + 0. calculated from the width of the plateau. and the width of the plateau is beyond the range of the graph at 57512.25.25. the Fresnel zone size is smaller at 60.5 below.5. Curves calculated for an AVW test range path length of 863 m. for σ I2 = 1. and the plateau width is 460.5: Churnside approximate aperture averaging factor plotted against the ratio of the aperture radius to the transverse coherence length.6ρ0.1. the Fresnel zone size is 21. As in the plane wave case. 78 . (4. larger irradiance variances show higher plateaus in A. 4.
2. 4. This coherence length is then used in the aperture averaging factor model given by Eq. Fig.6. which will be presented in Sec. (4. The shift in the aperture averaging factor from its value given by the Churnside model is visible in Fig.2 Andrews Asymptotic Analysis for the Spherical Wave Andrews used the Kolmogorov spectrum to relate the irradiance variance in the 2 saturation region to the plane wave Rytov variance σ R [11]: σ I2 = 1 + (σ ) 2. 79 . 4.74.3.2. This asymptotic model predicts a transverse coherence length that is different from the Churnside asymptotic model by a factor of 2.4. The Andrews asymptotic analysis is a good fit to the scintillation index model for a spherical wave. The plateaus are also narrower than those of the Churnside model.17).19) where σ I2 >> 1.73 2 2/5 R (4.3. 4.6: Andrews asymptotic model for the spherical wave plotted against the ratio of the aperture radius to the transverse coherence length.3.
3 Scintillation Index (SI) Model for the Spherical Wave The scintillation index model attempts to describe the behavior of irradiance fluctuations over the entire range of turbulence strengths. 4. The small scale and large scale log irradiance fluctuations are evaluated to define the scintillation index of a spherical wave. are plotted in Fig.20σ R 12 / 5 7 / 6 ) + (1 + 0.3. As in the plane wave case. The model for a spherical wave is developed using the effective Kolmogorov spectrum. using the same filter functions as presented in Eqs.5 < σ R < 2. by [20]: σ I2 (L ) = exp 1 + 0.9) and (4. along with the scintillation index model and weak turbulence theory.20) 2 where σ R is the plane wave Rytov variance.3. 4.4. the asymptotic models are similar. where 0. the scintillation index model has a lower slope than that predicted by weak turbulence theory.10) for the plane wave case. 80 .19σ R ( 2 0. In the onset of strong scintillation region.2.20σ R −1 5/6 (4.4 Comparison of Strong Turbulence Models for the Spherical Wave Both asymptotic models.23σ ) 12 / 5 R 2 0.7.2. (4. with the Andrews model matching the scintillation index model 2 2 when σ R > 64. neglecting inner scale effects.
8 shows that ρ0 becomes infinitely small as σ I2 → 1 . the scintillation index model follows the Andrews asymptotic model. In the saturated region. the transverse coherence length becomes very small (0. The asymptotic models predict that for strong saturation conditions.1 mm to 0.001 mm) for path lengths on the order of the range used in this experiment. 4. Figure 4. The scintillation index model predicts values of ρ0 in the range of 1 mm to 1 cm.Fig. The transverse coherence length also behaves asymptotically in the saturation region.7: Spherical wave irradiance variance plotted as a function of plane wave Rytov variance. Asymptotic and weak turbulence relations are indicated. 81 .
4.1 Experimental setup Strong turbulence aperture averaging data was collected over multiple days along the 863 m AVWChesapeake test range.Fig. On this test range.8: Behavior of the transverse coherence length in the strong fluctuation region.3 Aperture Averaging Experiment 4.35. 4. plotted as a function of the plane wave Rytov variance. however.3. a sufficiently large data set is available only at σ I2 = 1. The only valid experimental method for determining the aperture averaging factor in strong turbulence is to directly measure the irradiance variance of a propagating wave [40]. Unlike the situation in weak turbulence. measured intensity variance values ranged from 1. 82 . there is no way of equating the irradiance variance to the logirradiance variance in strong turbulence.1.01 to 1.
Data was collected during clear weather conditions.0 . and strong irradiance fluctuations. and the scattering disk is 8. 3. The LabVIEW program was shown in Fig. we expect the spherical wave 83 .1. and outputs the same 13 data values presented in Table 3.3. Winds ranged from 1 m/s to 4.3. presented in Section 3.The transmitter and receiver systems are the same as those used in the weak turbulence aperture averaging experiment.1.2 LabVIEW Data Acquisition Data acquisition for the strong turbulence case proceeded in the same manner as for the weak turbulence case.0 . The 1 minute averaging time allowed for more accurate measurements. l o ≤ ρ 0 and ρ0 < (L/k)1/2. σ I2 < 1. Strong turbulence values were always measured in daylight.4 Experimental Results for the Plane Wave Case Data was analyzed by the process described in Section 4. Temperatures ranged from 60°F to 83°F. 4. As in the weak turbulence case.3.1.2.32 mm. however. σ I2 > 1. with low relative humidity (less than 50%). due to atmospheric variations during the data collection period. The Fresnel zone size is 9. To neglect inner scale effects.7. 4.5 m/s. Data is sorted according to the measured irradiance variance of the scintillometer channel. not all available aperture diameters produced data in strong turbulence conditions. Data is then grouped by irradiance variance value and separated into two regions: weak irradiance fluctuations. Aperture diameters ranging from 1 cm to 16 cm were used. Received intensity was sampled in LabVIEW at 3000 samples/sec. with intensity variance values averaged over 1 minute intervals. which was presented in Section 3.69·105/ρ0.2.
Figure 4. Churnside asymptotic analysis predicts Cn2 = 9. near D/2ρ0 = 350.1. 84 .analysis to be in better agreement with the theory. New data is plotted in Fig.07 to 1. due to the divergence of the laser beam. it also has a range of values within one standard deviation of the mean.3×1012 m2/3. 4. In the plateau region. the mean data should follow the solid line for σ I2 = 1. the experimental aperture averaging factor is 43% higher than the value predicted by the model. The asymptotic theory curve seems to underpredict the aperture averaging factor.9 using the Churnside asymptotic value for ρ0. The data plotted has an average σ I2 ranging from 1.9: Average aperture averaged data in strong turbulence using Churnside asymptotic analysis plotted against the ratio of the aperture radius to the transverse coherence length. thereby overestimating the amount of aperture averaging that is taking place in strong turbulence.12. Error bars represent one standard deviation of the measured values. Therefore. Since ρ0 is calculated from measured values of irradiance variance.
The only difference between this data and the data plotted using the Churnside asymptotic representation in Fig. (4.1 mm for saturated strong turbulence.39 mm and Cn2 = 5. along with values of Cn2 averaging 9. Once again. Typical values of ρ0 in weak to nonsaturated strong turbulence are on the order of 1 mm. the direct measurement of the intensity to determine σ I2 is the only valid method for determining aperture averaging in strong turbulence [40]. 4.7×1012 m2/3 for this data. In the absence of independent measurements of Cn2 by an nonsaturating incoherent scintillometer [11. leading to this shift.2. The Andrews asymptotic model gives ρ0 = 0. 4.2) and (4. New data is plotted in Fig.01 mm to 0.06 appears to exhibit a plateau effect.7). There is a small constant offset between Eqs. the asymptotic model presented in Sec. 85 . since very long propagation paths have very small transverse coherence lengths [42]. (4. These values are consistent with those expected experimentally.As mentioned previously.1 was 2 used to extract values for σ R and ρ0.9 is that the data is leftshifted on the xaxis by a factor of 1. Using measured data values the asymptotic model predicts ρ0 in the order of 0.2.10 using the Andrews asymptotic relation of Eq.52.3×1012 m2/3. 4.24].7). the data clustered near A = 0.
86 . we have interpolated values of σ I2 based on σ R for the rising portion of the scintillation index curve shown in Fig. 4. The solid line represents the aperture averaging factor given by the Andrews asymptotic theory for σ I2 = 1.3.Fig. This is the scintillation index (SI) model. (4. 2 We previously introduced an expression for irradiance variance σ I2 based on σ R that is valid over all turbulence conditions. This new data is presented in Fig. giving a value for ρ0 that is consistent with the conditions on the experimental test range.1.3).10: New data using the Andrews asymptotic theory plotted against the ratio of the aperture radius to the transverse coherence length.11.1. along with scintillation index model curves using saturated and nonsaturated values of the Rytov variance for σ I2 = 1. 4. the data is analyzed assuming strong turbulence without saturation. 4. with inner scale effects neglected. Therefore. In order to determine the transverse coherence length 2 from Eq.
4 mm. the transverse coherence length is found to be ρ0 ≈ 4. 4. 4. The curves represent the aperture averaging factor 2 given saturated and nonsaturated strong turbulence values for σ R . Using the scintillation index model to find a Rytov 2 variance of σ R ≈ 2. The turbulence strength average is Cn2 = 8.11 is in better agreement with theory.11: New aperture averaging data calculated using the SI model plotted against the ratio of aperture radius D/2 to the transverse coherence length ρ0 for a plane wave propagating in strong turbulence. Since our data was not acquired under strong saturation conditions [42].1 . The new data in Fig. and visibly follows the predicted trend for σ I2 = 1. our receive aperture on the scintillometer channel with D = 5 mm is sufficient to avoid aperture averaging effects.4×1014 m2/3. 87 .Fig.1 .9 when σ I2 = 1.
for σ I2 = 1. on the order of 0.11. New data with σ I2 ≈ 1. 4.04 mm and Cn2 = 7. The Fresnel zone size is 9.3. (4.2.2 ×1010 m2/3 when σ I2 = 1. we can compare measured aperture averaging data with that of the theoretical approximations for the spherical wave case.2.01 mm.12 using the Churnside asymptotic model. Using the models presented in Section 4. The Andrews asymptotic relationship for the transverse coherence length was used to plot the aperture averaging factor in Fig. The data is slightly higher than what the approximate formula for σ I2 = 1.1 predicts. The drawbacks of the asymptotic model are explained in more detail later in this chapter. (4.1 mm. to determine ρ0. the data is higher than what the asymptotic approximations predict.1 .5 Experimental Results for the Spherical Wave Case Strong turbulence data for the spherical wave case was acquired through the process presented in Section 4.4. Andrews asymptotic analysis estimates the level of turbulence at Cn2 = 1. The predicted Cn2 indicates very strong turbulence because the asymptotic model assumes strong saturation conditions that force the transverse coherence length to be a very small value.14) and (4. Eqs. 88 .19) were used to determine ρ0 for this set of data.32 mm. presented in Section 4.3.13 for σ I2 ≈ 1. As with the Churnside asymptotic analysis.1 is shown in Fig.2 ×1010 m2/3. The Andrews model predicts the transverse coherence length for this path to be on the order of 0.1 .3. 4.17) is also plotted in Fig. while the typical transverse coherence length was near 7 mm.1. 4. The Churnside asymptotic model predicts ρ0 ≈ 0. The approximate formula from Eq. indicating that we are within the strong turbulence region.1.
which ranges from 0.74 larger than that of the Churnside asymptotic model. Recall that the discrepancy between the Churnside and Andrews asymptotic models is that ρ0 in the Andrews model is a factor of 2. The two data points located at A = 0. the Churnside and Andrews asymptotic models predict that A will always be much less than 1 for an aperture diameter of any reasonable size. 4.12: New strong turbulence data analyzed using the Churnside asymptotic method for a spherical wave. 4. Fig. The residual value in Figs.1. In order to measure A near 1. the receiver aperture diameter must be on the order of the saturated value of ρ0. 89 .1 mm. although the mean of those five points is about 36% higher than the asymptotic model values at D/2ρ0 = 2000.13 is that they clearly show that this data was not taken in saturated strong turbulence conditions. The rest of the data points indicate a plateau effect.78 and A = 0.01 mm to 0.18 would not have been measured if the turbulence was saturated.In the case of both asymptotic relations.12 and 4. with irradiance variances near 1.
the scintillation index model for the irradiance variance in Eq.Fig. putting us well within the strong turbulence region on this path. (4.1 .20) was used to analyze the experimental data in a nonsaturation strong turbulence region.1 was near 7 mm. This corresponds to an average Cn2 = 8. Fig. 90 . 4. This is consistent with the turbulence working to render scale sizes larger than ρ0 ineffective along the path. To address the shortcomings of the asymptotic theories. The ρ0 for σ I2 = 1.4×1014 m2/3.14 shows the mean data plotted with the approximation to the aperture averaging theory when σ I2 = 1. The data appears to show a tapering off and potentially a plateau area near D/2ρ0 = 10. (4.1 . The curve represents the Andrews asymptotic model aperture averaging factor for σ I2 = 1.19).13: New strong turbulence data plotted using the Andrews asymptotic model for the spherical wave from Eq. 4. similar to that predicted by the approximate curve.
The new data using the asymptotic and SI models have been plotted along with the Churnside experimental data in Figs. The inability to measure Cn2 over the same path as A calls into question the true pathaveraged nature of the turbulence data. and ρ0 = 1. 4. The solid line is the aperture averaging approximation in Eq. adjacent to the 1000 m aperture averaging path.14: Mean data with error bars plotted against the ratio of the aperture radius to the transverse coherence length.15 and 4. The Churnside data is the only wellaccepted set of aperture averaging data in strong turbulence. and Cn2 was measured over a 250 m path. (4.16. 4.Fig. The Churnside data was taken by propagating a HeNe laser over a 1000 m path. with an average irradiance variance of 3. using an incoherent scintillometer.08 is shown in Fig. The average Cn2 measured was 1.15 to contrast with the Churnside data.74 91 .38.08 ± 0. The Churnside asymptotic curve for σ I2 = 3.29 ± 0.1 .39 × 1012 m2/3.17) for σ I2 = 1. 4. Data for six apertures was collected simultaneously.
mm ± 0. 4. an aperture size near the ρ0 scale may be chosen for a specific link.13. which is due to the beam divergence of the transmitter. The new data reinforces the idea that it will be difficult to surpass the plateau and reach the scattering disk scale in saturated conditions. 4.4. Although the analyzed data demonstrates a plateau in reasonable agreement with approximate theory. likely leading to a plateau. The Churnside data is limited in breadth. since Churnside did not have large enough apertures available to measure very small aperture averaging factors. The Churnside data does not show any scale effects. if it is required to work in saturated scintillation conditions. while the new aperture averaging data using the SI model (Fig.16 does not include the 92 .12 and 4.24 mm. the ρ0 and scattering disk scale sizes result in a plateau region in the aperture averaging factor for strong turbulence.32 in Fig. 4. following the discussion presented earlier regarding Figs. this would require an extremely large diameter receiver. This also limited their ability to see any scale size effects due to the size of turbulent eddies relative to ρ0 and the scattering disk. By plotting the aperture averaging factor on the plateau. The new data is in good agreement with the predicted aperture averaging curve.16) begins to exhibit a tapering around D/2ρ0 = 10.188 and D/2ρ0 = 4. even at higher values of A. the data should only be treated as an estimate of what could occur if the link experienced strong saturated turbulence. As mentioned previously.15 results in data that is beyond the range of the Churnside data. 4. The spherical wave treatment is also much improved over the plane wave analysis of Section 4. Only the point at A = 0. the new data analysis using asymptotic theory in Fig. In a condition of strong saturated turbulence.
plotted along with the strong turbulence Churnside data taken over a 1000 m path. these results should encourage further experimental studies of aperture averaging in strong turbulence in order to improve the quality of theoretical models available in saturated and nonsaturated strong atmospheric turbulence. Fig. The new data is in good agreement with both curves.1 . 93 . Although Fig. data for larger apertures is needed to discern the validity of these results.approximate theory for the aperture averaging factor within the error bars.15: Mean data analyzed with the Churnside asymptotic model. 4. This is due to a limited number of data points at that aperture size.16 is beginning to show scale size effects. Given that the amount of strong turbulence data over this test range was limited.16 is shown with both the SI model curves for saturated and nonsaturated strong turbulence at σ I2 = 1. 4. The new data in Fig. 4.
1 are also shown. The only large data set was available for σ I2 = 1. Due to the height of the propagation path above ground. The SI model shows the best aperture averaging results for the spherical wave in strong turbulence.16: Mean data analyzed using the scintillation index model in the nonsaturated strong turbulence region. College Park. Andrews Asymptotic. The models are 2 used to determine values for ρ0 and σ R . measured on an 863 m test range at the University of Maryland. which otherwise could only be measured by using an incoherent light scintillometer. it was difficult to acquire enough data points in strong turbulence conditions for analysis. 4.Fig. and Scintillation Index models. plotted along with the Churnside data for a 1000 m path.5 Conclusions Strong turbulence aperture averaging data. Saturated and nonsaturated SI model curves for σ I2 = 1. has been analyzed using the Churnside Asymptotic. this is the first set of strong turbulence aperture averaging data to be acquired in nonsaturation conditions. 4.1 . 94 . In spite of these limitations.
there is a definite increase in the BER.2. High speed optical receivers use photodetectors with diameters on the order of 10s of microns.1 TurbulenceInduced Beam Motion Beam Wander After a beam propagates through a length equal to a few transmitter diameters of turbulent atmosphere.2 5. the atmosphere distorts the wavefront and produces intensity scintillations. 5.1 Introduction Atmospheric turbulence adversely affects the propagation of optical waves. achieving a good bit error ratio (BER) requires that a maximum amount of received light is received by the photodetector. Beam wander is a result of the electromagnetic wave interacting with turbulent eddies of sizes on the order of the transmit aperture diameter 95 . Commercial systems for data transmission typically require a BER of 109 or better [43]. When turbulence induced beam wander moves the focal spot. In an onoff keyed transmission system. the beam begins to wander randomly in the plane transverse to the propagation direction [44].Chapter 5 Free Space Optical Communication using Nonimaging Optics 5. Another manifestation of the interaction of atmospheric turbulence with an optical wave is the receiver spot wander in the focal plane. Minimizing beam wander in the focal plane of the receiver is an important and relatively low cost step in maximizing link reliability. As we have investigated in the previous chapters.
5. Although the beam wanders.2. is defined by [20]: rc2 1/ 2 2 2 = WLT − WST (5. The angleof 96 .[44]. rc. also called beam wander in the focal plane. the influential scale sizes are larger than the inner scale. Typically. ℓo. 5. and WLT is the long term beam width. 5. shown in Fig.1.1: Beam wander at the receiver plane is characterized by the short term and long term beam widths [adapted from Ref. Fig. The motion of the centroid of the beam. It has been shown experimentally that the time constant of the beam wander is on the order of the ratio of the beam size to the wind velocity. 20]. it tends to maintain the shape it would have if the beam was propagating through free space [45]. is related to turbulenceinduced phase fluctuations on the propagating optical wave.45]. Beam wander may be observed while propagating a visible laser (such as a HeNe) at night.1) where WST is the short term beam width.2 Angle of Arrival and Image Dancing Imaging dancing. Previous work has proposed that a fast tracking transmitter be used to minimize beam wander [20.
L = 863 m. 5. 97 . The optical signal is created by either the direct modulation of a laser.arrival is a result of a phase shift along the receiver lens diameter that is manifested as an optical path difference.91Cn LD −1 / 3 ( ) 1/ 2 f for D >> l o (5. the RMS image displacement.3) where f is the focal length of the optical receiver in meters.2) where ∆l is the optical path difference and ∆S is the phase shift across a receiver aperture with diameter D. ∆sRMS. small aperture photodetectors with diameters of less than 100µm are used.47 µm. or the use of an external modulator to control the emission from a continuouswave (CW) laser source.000. lowcapacitance. In testing high speed (> 1 Gbps) transmission systems. and f/10. Depending on the power and subsequent amplification of the laser source. α. Any image motion will result in the loss of photons and will increase the BER.000 photons.000. the number of photons transmitted for a “1” may be upwards of 1. the RMS image displacement is 16. The angleofarrival.3 BER in OnOff Keyed Systems An onoff keyed (OOK) system is binary transmission protocol where 1s and 0s are represented by the transmission of a large number of photons or the lack of photons in a bit slot. is defined by [20]: α= ∆l ∆S = (in radians) D D (5. D = 16 in or 40. only 7800 photons may make it to the receiver.6 cm. Using Eq. On an atmospheric link with a fade on the order of 40 dB during a “1” bit period.2). Cn2 = 5×1015 m2/3. Using the specifications of link under test and typical weak turbulence conditions. and a transmitter power of 0 dBm = 1 mW. is [20]: ∆sRMS = α 2 1/ 2 2 f = 2. (5.
For a pin photodetector operating at 100 Mbps with a minimum receive power of 42 dBm.46]: ∞ SNR S 1 dS BER(OOK ) = ∫ pI (S )erfc 2 2 i 20 S (5. and σN is the standard deviation of the Gaussian noise distribution. The BER of an OOK system is derived from the theoretical probability of a bit error assuming a Gaussian noise distribution [20]: iS 1 Pr(error ) = erfc 2 2σ 2 N (5. implying that p0 = p1 = 0. an equal number of “1” and “0” bits are transmitted on average. the probability of an error may be written as [20]: Pr(error ) = p0 Pr(1  0) + p1 Pr(0  1) (5. ψ.5.6) where SNR is the signaltonoise ratio of the transmission. and largescale. which in this case is the OOK format. Measuring the BER is the best way to characterize the performance of an optical wireless communication system. The BER for an OOK system is [20. and p0 + p1 = 1.47]: 98 . γ. In an OOK system. a minimum of 4925 photons must make it to the photodetector surface. p1 is the probability that a “1” is transmitted. and pI(S) is the Nakagami gammagamma distribution with variables representing the contribution of smallscale. components to the irradiance of the transmission. I = γ ψ [20.4) where p0 is the probability that a “0” is transmitted.5) where iS is the signal current of one bit pulse. When a pseudorandom sequence is transmitted. The BER also accounts for the modulation format of the transmission. Loosing photons due to beam wander in the focal plane is an unnecessary liability in an optical wireless link.
9) is plotted in Fig. Eq.90d + 0.9) Eq.sph R ( ) −1 ( ) − 1 7/6 −1 (5.62d σ R .49σ R .56σ 12. (5. The components of the irradiance in the presence of atmospheric turbulence may be defined by [20.6 . The gammagamma distribution is the representative probability distribution function for a range of physical phenomena.8) d= 2 σ R . sph 2 0. sph kD 2 4L 2 = 0. and its derivative dS = iS dx . The three values of irradiance 2 2 fluctuation were chosen to represent weak σ R = 0. sph 1 + 0.3 .18d 2 + 0. and non 99 . including the propagation of radio waves in the troposphere or ionosphere [20].51σ R . intermediate σ R = 0.4065σ R iS By defining a new random variable x = S . (5.2 to demonstrate the relationship between the BER and SNR for different values of irradiance variance.69σ R .47]: 12 / 5 − 5 / 6 2 exp 0. sph ψ = exp /5 1 + 0. 5.7) where Kψγ() is the modified Bessel function of the second kind.6) can be transformed into the BER of an OOK system in the presence of atmospheric turbulence [47]: (ψγ )(ψ +γ ) / 2 BER = Γ(ψ )Γ(γ ) SNR x ψ +γ −1 2 Kψ −γ 2 xψγ dx × ∫ erfc 2 2 ( x ) 0 ∞ ( ) (5. sph − 1 γ= 2 2 12 / 5 1 + 0.2(ψγ ) pI (S ) = Γ(ψ )Γ(γ ) iS (ψ + γ / 2 ) S i S ψ + γ −1 2 ψγS Kψ − γ 2 iS (5.
Three more curves in Fig. 5. we have chosen to integrate a nonimaging optical element into the optical receiver of an optical wireless link. presented in Sec.0 . The irradiance variances used in the calculation are the same in both cases.0. kD 2 5. 5. there is no aperture averaging present 4L because the aperture radius is equal to the Fresnel zone size (FZ). 100 . 0.2 saturated strong turbulence σ R = 2. Fig. there is a between a 4 dB and 15 dB improvement with aperture averaging across each level of irradiance fluctuations.6. Three more curves are plotted for an aperture averaged case using a 4 inch diameter aperture.2 are plotted to show the BER improvement for a 4 inch aperture along a 1.4 Nonimaging Optics To address the problem of beam wander in the focal plane. 5. For a BER of 109.2. The Rytov variances plotted are 0. and 2.3.2: The BER of a communications link with a gammagamma probability distribution function (PDF) as a function of signaltonoise ratio.726 km link. When = 1 .
This behavior is indicated by the double arrows in Fig. Welford published the first book on nonimaging optics in 1978 [48]. 48). Fig. The cone concentrator was limited in its performance.3.3: The cone concentrator with cone angle γ and maximum entrance angle θi. The cone is hollow with reflective (metallic) edges. 5. Figure 5.T.4. because some rays would be reflected back through the entrance aperture of the cone. the design of the dielectricfilled Compound Parabolic Concentrator (CPC) for energy concentration was first published [49]. The CPC is an extension of the basic cone concentrator.4 shows a diagram of the CPC. As in the case of the cone concentrator.1 Theory of the Compound Parabolic Concentrator Roland Winston and W. nonimaging optical elements were investigated for application to the concentration of solar energy. This angle is related to the entrance and exit apertures by: 101 . In 1976. (adapted from Ref. Initially.5. The CPC uses surfaces traced out by an offaxis parabola to improve the concentration ratio over that of the cone concentrator. θi is the maximum entrance angle of the CPC. 5.
14) In fact. 102 . the maximum concentration ratio will depend on the reflectance of the concentrator surface. The length.11) The design equations of the CPC in Cartesian coordinates are easily determined from Fig. 5.12) where the parametric angle φ is the angle between the axis of the parabola and the vector r.10) where aout is the radius of the exit (smaller) aperture.sin (θi ) = aout ain (5. and ain is the radius of the entrance (larger) aperture.4: 2 f sin (φ − θ i ) − aout 1 − cos(φ ) 2 f cos(φ − θ i ) z = r cos(φ − θ i ) = 1 − cos(φ ) y = r sin (φ − θ i ) − aout = (5. The parametric angle has minimum and maximum angles of φmin = 2θ i and φmax = π 2 + θi .13) The theoretical maximum concentration ratio for a 3dimensional CPC is: Cmax a = in a out = sin −2 (θ i ) 2 (5. The parabola has a focal length of: f = aout (1 + sin (θi )) (5. L. of the concentrator is found by calculating z for φ = φmin: 2 f cos(θi ) aout (1 + sin (θ i ))cos(θi ) = 1 − cos(2θi ) sin 2 (θi ) L= = (ain + aout )cot (θi ) (5.
and are self baffling. add minimal size and weight to the system.4. they are easily designed. The axis of the parabola and the CPC axis are different. highly efficient. The self baffling nature of the CPC limits the amount of stray light incident on the photodiode. 5.2 The DielectricFilled CPC Dielectricfilled CPCs take advantage of the principle of total internal reflection to maximize the concentration ratio. Fig.CPCs are especially useful for integration into FSO communications receivers. The surface of the CPC is traced out by moving the vector r through the parametric angle φ. Since total internal reflection is nearly 100% efficient.4: The design of the CPC surface. 5. the theoretical maximum concentration ratio becomes [48]: 103 .
15) where n is the refractive index of the dielectric material. and θe is the maximum angle of incidence of an entrance ray on the dielectric surface. 5. Fig. θe.5 is a plot of θe and the internal maximum transmission angle. θn.Cmax. the maximum angle that is refracted into the dielectric is given by Snell’s Law as sin −1 (1 / n ) . The angles θe and θn are related to n by: sin θ e ≤ n − (2 / n ) sin θ n ≤ 1 − 2 / n 2 ( ) (5. 104 . the angle θe becomes imaginary. and the maximum internal ray angle. Since the CPC cannot accept an angle larger than 90°. dielectric a n2 = in = a sin 2 (θ ) e out 2 (5.16) If n > 2. Figure 5. θn for a range of refractive indices.5: Plot of the maximum entrance angle in air.
The parametric equations for the dielectric CPC are the same as those used to describe the hollow CPC, given by Eq. (5.11) through (5.13), with θi replaced by the maximum internal transmission angle of the CPC. 5.4.3 The CPCPhotodetector Combination The coupling of a CPC to a light detection device is logical method of increasing the probability that an optical signal is accurately detected. CPCs must obey the brightness principle, which states that an image cannot be brighter than the incident object [48,50,51]. When considering a CPC, the brightness theorem dictates that radiation incident on the CPC over an angle of incidence θi will exit the CPC maintaining the same flux per unit area; therefore, the radiation will exit the CPC over a range of angles larger than θi. Using the brightness principle, a relation is found between the entrance aperture of the CPC and the photodetector area required to achieve maximum efficiency [50,51]: sin 2 (θi ) = Ai n2
APD
(5.17)
where APD is the area of a photodetector and Ai is the area of the entrance aperture of the CPC. In order to minimize the detection noise and maximize the bit rate of the system, the smallest photodetector allowed by Eq. (5.17) should be used. The maximum collection efficiency expected by Eq. (5.17) will be less than 100%, since Fresnel losses may exist at the entrance and exit surfaces of the CPC. Alternative designs of nonimaging concentrators have been proposed, including the θi/θo concentrator [48], and RX and RXI dielectric concentrators [50,51]. The
105
former design restricts the exit angle by using a cone section at the exit aperture. The latter designs enhance the uniformity of the output irradiance distribution. 5.5 1.7 km CPC Link Experiment The hypothesis of a BER improvement when using a CPC in the optical receiver of an optical wireless link was tested on a 1.7 km range at the University of Maryland, College Park. 5.5.1 Experimental Setup and Test Range The test range used for the CPC link experiment is the same range that was used for aperture averaging experiments. The addition of a hollow, cornercube retroreflector (CCR) allowed the link range to be doubled to 1.726 km. A schematic of the link experiment is shown in Fig. 5.6. The bit error ratio tester (BERT) generates a pseudorandom binary sequence (271), which modulates the laser diode. The output of the fibercoupled laser diode is transmitted to the erbium doped fiber amplifier (EDFA), which amplifies the signal to approximately 158 mW. After the signal exits the optical fiber, the optical beam is transmitted by a beam expansion system to the CCR downrange. The beam expander contains a 10 x magnifying lens and a 2.25 in. biconvex lens. The 5 in. cornercube retroreflector is goldplated for maximum infrared reflectance. Since the retroreflector has a narrow divergence angle of 1 mrad, the transmitting lens system must be placed in front of the center of the 16 in. Meade SchmidtCassegrain telescope. Almost fortunately, part of the primary aperture on the telescope is obstructed by the secondary mirror, so the amount of obstruction due to the transmitter lens system is minimized.
106
After a portion of the beam is reflected by the CCR, the reflected wave expands and is incident on the 16 in. telescope. The Meade is an f/10 telescope, although the focus is adjustable. A ZnSe dielectricfilled CPC with a 1 cm entrance aperture is placed in front of the focal plane of the telescope to maximize collection efficiency. Details of the ZnSe CPC are discussed in the following section. The exit aperture (1 mm diameter) of the CPC is buttcoupled to a New Focus DC  125 Mhz InGaAs photodiode (PD), Model 1611. The photodiode signal is amplified through two RF amplifiers and subsequently read by the BERT receiver. The BERT is an Anritsu ME522, which can transmit data at rates from 50 Mbps to 700 Mbps. An external multiplexer/demultiplexer will increase the limit to 1.4 Gbps. The BERT measurement period is variable from 1 second to hours.
TX Lens
16″ Meade
To CCR
CPC CPC PD Amplifier Errors
BERT
IR Camera CLOCK EDFA TX Laser, Power Supply DATA
Fig. 5.6: Setup of the CPC link experiment. Black lines are electrical connections; orange lines are fiber optic connections. The IR camera is used for alignment purposes only, and is lowered below the field of view of the 16 in. Meade telescope when not in use.
107
The refractive index of ZnSe at 1.5.7 km experiment.18) 108 . A profile of the CPC is shown in Fig. For the 1.48. The CPC has a maximum entrance angle of ~5°.25 cm long. the photodetector integrated into the package was removed. with an entrance diameter of 1 cm and an exit diameter of 1 mm. due to its high transmission at 1. Al mounting tube ZnSe CPC InGaAs Photodiode Fig. The CPC is approximately 2.2 ZnSe CPC The CPC used in this experiment was made of ZnSe. 5. with a canned photodiode centered at the exit aperture of the CPC.55 µm.5.5. At normal incidence. It may be written in the form [52]: PTX = PRX + CL + M s (5. 5. 18% of the light is reflected at each aperture. 5.7: Profile of the mounted ZnSe CPC directly coupled to an InGaAs PD.55 µm is 2. The New Focus InGaAs photodetector was used in its place.7.3 Link Budget The link budget is a design method to ensure that enough power will reach the optical receiver given that there will be some losses on the optical link.
A fade is a sudden drop in the signal strength that may result in the erroneous detection of a bit. and the electrical power in dB.1.2dB Although this system margin is sufficient to maintain a BER of 109 in a digital communications system. a SNR of 17 dB is (5.where PTX is the average transmitter power. The system margin is given by: M s = PTX − PRX − CL = 22dBm + 44. which represents the maximum additional loss the system can handle.19) needed. or power burdens on the system. Fades of 10dB are common. Fading on optical wireless links is induced by the motion of turbulent eddies in the atmosphere. The link budget is normally written with the optical power in dBm. where 1 dBm = 10log10 (P / 1 mW ) . PRX is the receiver sensitivity in dBm. 2 5. This is a clear example of how receivers integrated with nonimaging concentrators should improve the link margin without placing any additional size. where an electronic noise margin of 21 dB is required [20.3 ).7 km link. and Ms is the system margin. without the CPC. weight. Fig. 46. it does not account for fading on the optical wireless link. The link budget of the 1.53]. although fades of over 40 dB have been reported [54]. is articulated in Table 5. CL represents the channel losses.5dB = 28. 109 .2 shows that for a BER of 109 in weak turbulence ( σ R = 0.3dBm − 38.
Inc. CL Loss due to mismatch of beam width and 16 in.Link component 1.7 km link. model 2666A Optigain EDFA model 2000 PTX Fiber splice Fiber connectors Viracon thermal pane window in Chesapeake Bldg.7 dB 0.75 dB 11. Atmospheric transmission Retroreflector Atmospheric transmission Window loss Meade LX 200 EMC Component specifications Max. neglecting atmospheric turbulence. 5.5pW/(Hz) InGaAs photodetector Table 5. The Cn2 110 . aperture Loss due to secondary mirror obstruction of SchmidtCassegrain design telescope Telescope transmission at 1.25 dB 0. Cn2.446 dB 1.2 dB/km for clear air [55] Loss due to mismatch of beam width and CCR diameter (129mm) Surface reflectivity Power (in dB or dBm) 8 dBm 30 dB 22 dB 0.15 mW.3 dB 1 dB 6 dB 0.54 dB 0. output 23 dBm Transmitter power = 158 mW 2 splices @ 0. extinction ratio 10:1 Max.55 µm laser: Force.1: Link budget analysis for the 1.5 dB Coupling loss Total channel loss.75 dB 6 dB 6.175 dB 0.15 dB 2 connectors @ 0. is measured over the same time interval as the BER. Measurements over 1 minute intervals gave the most reliable results. launch power 0.5 dB Average 25% transmission for infrared wavelengths Assume an attenuation of 0.4 Cn2 Measurements The strength of turbulence.5 dB 38.3 dBm PRX 1/2 New Focus 1611 NEP = 2. gain 30 dB Max.55µm with EMC coatings Coupling of focal spot to photodetector Receiver sensitivity 44.5.
Fig.5. Data was taken over two consecutive days in weak turbulence conditions ( σ I2 < 0. The range of BER measurements indicates that the beam may not have been as collimated as was thought. BER for 100 Mbps PRBS transmission with and without the CPC is plotted against the measured Cn2 in Fig. The exit aperture of the CPC was in direct contact with the photodetector window in order to achieve maximum optical coupling. the width of 111 . The curve fit equation is also shown in the plot. the measured BERs were found to be on the order of 103 to 104.was measured over the scintillometer channel used in the aperture averaging experiment.5 Experimental Results and Analysis A 100 Mbps pseudorandom binary sequence with a pattern length of 271 bits was transmitted downrange and reflected by the corner cube retroreflector. Since the system operates at 1. The actual PD surface is set back 0.8 and 5. The PD was mounted on an xyz stage.8. and the CPC was placed on a 6 degreeoffreedom stage to attempt to maximize light output.3) .9.5 mm from the lens case.9 plots a first order linear fit along with the mean data. This coupling is especially important since there is a large area mismatch between the CPC exit aperture (1 mm diameter) and PD (0. A significant loss of coupling was observed when there was more than 2 mm of air space between the PD and the CPC. 5. The coupling of the CPC and photodetector is very angle sensitive. 5.3 mm diameter). From Figs. Test runs were done with and without a ZnSe CPC placed directly in front of the New Focus 1611 InGaAs photodetector. acquisition period. with error bars representing the fluctuation of Cn2 over a 1 min.55 µm. 5. 5.
The error bars in Cn2 represent its fluctuation over the 1 min averaging interval. 112 . Fig.the beam when it reaches the 5 inch retroreflector is extremely difficult to measure.8: BER plotted as a function of Cn2 for the 1. 5. the important result is the relative enhancement of the BER when the optical communications receiver includes a CPC.7 km link with and without the CPC integrated into the optical receiver. Though the BERs on the FSO link are high.
5. which caused the signal to be sensitive to vibrations of the floor. There is approximately a factor of 5 improvement in the BER when using the CPC in the optical receiver. considering that only about 11% of the light that exits the CPC is collected by the PD. 113 . 5. it does not feel the vibrations. heavy tripod. plotted along with a first order linear fit to the mean data. Since the Meade telescope is in its own. very large. This enhancement is particularly encouraging.8. due to the area mismatch. A telescope with the CPC and PD using the same optomechanical mount would also improve the performance by making the system immune to vibrations. A CPCPD combination where the PD area is either the same or slightly larger than the exit aperture of the CPC should yield a better BER enhancement.Fig. Also.9: Same data as in Fig. the CPC and PD were mounted on a tripod.
otherwise there could be significant damage to the system.2 Vpp signal to the BERT.Fig. The amplified signal was attenuated to between 30 mVpp and 500 mVpp so that any sudden spikes in intensity would not overload the limiting amplifier. In using the BERT. but still leaves residual noise on the data that it cannot clean up. before and after passing the signal through a limiting amplifier.10 is an oscilloscope trace of the amplified received data. Nonetheless. 5. The limiting amplifier then outputs a 1.10 is actually the inverse of the data. which is connected to the oscilloscope so that each of the limiting amplifier terminals are properly terminated. 114 . The second trace in Fig. The amplified signal even showed some baseline wander due to the fluctuations in the intensity of the received signal. it shows how the limiting amplifier attempts to clean up the turbulence induced fluctuations of the data. The turbulenceinduced optical fluctuations of the receive signal affect the quality of the electrical signal output of the RF amplifiers. 5. the user must be careful that the terminals of the BERT are not overloaded.
6 NRL Test Range Experiment over the Chesapeake Bay An experiment to demonstrate the BER enhancement due to the use of a CPC PD combination in an optical receiver was conducted at the Naval Research Laboratory’s Chesapeake Bay Detachment.1 Experimental Setup and Test Range The Naval Research Laboratory has a facility located on the Chesapeake Bay in Maryland that they have set up for use as a laser communications test site. To test high speed laser communications. a solid retroreflector array is located on a tower at Tilghman.Fig. Trace 2 shows the inverted data after passing through a limiting amplifier. 5.57]. however there is still enough turbulenceinduced noise to degrade the BER. and can be interchanged with an MRR 115 .6. 5. They have previously used the facility to test their modulating retroreflector (MRR) array technology [56.10: Trace 1 shows the data stream after passing through two 20 dB RF amplifiers. in Maryland (NRLCBD). MD. 5. The signal after the limiting amplifier is much cleaner.
5. spaced on 5 inch centers.12 is a diagram of the optical transmitter. 5. Fig. 5. aligned three wide by four tall.2 km across the bay from NRLCBD. The optical transmitter uses a fibercoupled diode laser at 1. The Tilghman site is approximately 16. The collimated optical output is approximately 2 W [56]. Fig. A BERT pattern generator (Agilent model 86130A) modulates the fibercoupled diode with a 271 PRBS pattern.12: Optical transmitter at NRLCBD. When the retroreflected signal returns.7 km link at the University of Maryland. Fig.55 µm that is subsequently amplified to 2 W.11: The NRLCBD laser communication test range [56]. and sent downrange to the retroreflector array. It is estimated that 1. Fig. 5.array.11 is a diagram of the test range. Meade LX200 SchmidtCassegrain telescope.84 mW of the 2 W launch power is 116 . Each retroreflector is 2 inches in diameter. The signal exiting the fiber is expanded and collimated. This is the same telescope used in the 1. it is collected by a 16 in. The retroreflector array at the Tilghman tower is composed of 12 solid retroreflectors.
Fig. receiver telescope [56].1) × 1000 = −28. At 1.38dBm (19) 117 .6. 5.13: The optical receiver at NRLCBD [56]. 5.5 µm.5 mm diameter.13.38 mm. shown in Fig. The back end of the aluminum mounting tube has bond pads and short leads that connect the photodiode signal to a Maxim 3266 low noise transimpedance preamplifier chip. The signal is then detected by the BERT receiver. the photodiode has a responsivity of 1. A description of the optical receiver is shown in Fig. is used as the detection system in the NRLCBD experiment.collected by the 16 in.0A/W(10 . Since the signal strength is very low after traversing a 32 km roundtrip. The distance between the TO46 package window and the photodiode surface is 0. The input sensitivity of the preamplifier for a BER of 109 is: OSNR × I N (rex + 1) Sensitivity = 10 log × 1000 2ℜ(rex − 1) 11.2 CPCPD Combination System The integrated CPCphotodiode optical receiver. the signal is amplified with a Femto variable gain voltage amplifier that can amplify a signal by up to 60 dB (model DHPVA100).7. 5.89 × 200nA × (10 + 1) = 10 log 2 × 1. The detector integrated into the package is a Telecom Devices 35PD500TO InGaAs photodiode with a 0. 5.0 A/W.
6 m/s. but had to be cancelled because of the amount of turbulence present on the link.6. The experiment was attempted twice prior to the successful test.where OSNR is the optical SNR required to achieve a BER of 109.4° to 3.84 Mbps).3 Experimental Results and Analysis BER measurements were taken over the 32. 5. rex is the extinction ratio of the laser transmitter (defined here by the ratio P1/P0) [58. a water temperature range of 5. The CPCPD receiver was used in place of a 62. the optical power in a “0” bit is 10 dB lower than that of a “1” bit: P = 4W 1 P0 = 0. IN is the inputreferred RMS noise current of the Maxim 3266. the optical power in a “1” bit is approximated by twice the average power [43]. Given a 10:1 extinction ratio and assuming a 2W (33 dBm) average power. OC2 (103. Measurements were made at standard telecommunication data transmission rates of OC1 (51. and an average windspeed ranging from 1.5°.68 Mbps).0 m/s to 2. MD.5 µm optical fiber coupled receiver. When using a PRBS transmission pattern. The turbulence made it impossible to align the test link using LED beacons on the tower at Tilghman and a position sensitive detector located at NRLCBD.59]. Data was taken over a 4 hour window with an atmospheric temperature range of 2. These temperature and windspeed measurements were taken from a NOAA buoy in proximity to the NRLCBD test range at Thomas Point. (20) 118 . and OC3 (155.9°.8° to 5.1 × 4W = 400mW where P0 is the power in a “0” bit. The optical power in a “1” bit and “0” bit may also be calculated.4 km test range at NRLCBD. and ℜ is the responsivity of the photodiode in A/W. and P1 is the power in a “1” bit.52 Mbps).
A high value of Cn2 is assumed because link must be in the saturated strong turbulence range.14. 119 .84 Mbps transmission is 3. The Agilent BERT has the ability to count “0” errors and “1” errors. The only errors were detected in the 31st minute. 5. as indicated by the dark horizontal line. the beam wander in the focal plane is estimated at ∆sRMS = 339. The overall BER is plotted in 1 min intervals. There were no errors between minutes 1 and 19. Typically. Data collected at OC1 is shown in Fig.3).14: Measurements of BER using the CPCPD receiver when an OC1 PRBS transmission pattern was tested during a 35 minutes test period. along with the “0” error rate and “1” error rate.From Eq. 5. The BER of 1 bit in a 51. (5. we expect that “0” errors are much lower than “1” errors because the power in a “0” bit should be below the threshold of the BERT. Fig. the “0” error rate is lower than the “1” error rate.6 µm for a Cn2 = 1×1013 m2/3.215×1010. due to the extremely long propagation length. although not substantially lower. For the OC1 test.
5.5 µm optical fiber coupled receiver in place of the CPCPD receiver. A single bit error in a 103.BER measurements taken at OC2 are shown in Fig. The signal is measured by an OC12 receiver. 5. 5. In the 11th minute. and finally received by the Agilent BERT. the “1” error rate is much higher than the “0” error rate. Fig.68 Mbps transmission at OC2 gives a BER of 1.16 shows OC2 BER data published in an NRL paper. 120 .15: BER at OC2 over a 12 minute acquisition period using the CPCPD receiver. regenerated and collected by an OC48 receiver.607×1010.15. with a 62. Fig.
A comparison of the data from the CPCPD receiver (Fig.16. and manifested in the BER improvement.16) shows a significant improvement in BER when using the CPC. Note that data is plotted for 5 sec intervals.16: BER at 100Mbps measured over the NRLCBD test range with an optical fiber coupled receiver. The long link length contributes to the amount of beam wander in the focal plane. 46. The first 4 minutes of acquisition are error free. 5. from Ref. there is not one acquisition interval that is errorfree. BER results from tests with the highest transmission rate. The minimum BER for a 5 sec interval is 2×109. 5. There is no autotracking between the 121 .15) and the NRLCBD optical fiber coupled receiver (Fig. 5. which is beyond the range of the graph. The problem of beam wander in the focal plane is visibly reduced. 5. This behavior is indicative of transmitter alignment problems and not problems with the CPCPD receiver. 5. are shown in Fig. The following 11 minutes show significant errors.Fig. In Fig.17. OC3.
Fig. 5. the device performed extremely well. New data could not be acquired at 200 Mbps. and the receive signal intensity at 200 Mbps fell off significantly. Although this CPCPD receiver was not optimized for the link. Fig. 5. 122 .18 shows a consistent and significant number of errors throughout the acquisition period. it appears that the addition of a CPC into the receiver system would improve the BER. 5. each of the three data rates tested showed a significant amount of errorfree time.18 shows published results from the link with an optical fiber coupled receiver for data transmitted at 200 Mbps. The device had a bandwidth of 100 MHz.transmitter and receiver on the link. nor was the optical coupling between the CPC exit aperture and PD surface optimized. Overall. The first 4 minutes of error free acquisition are extremely encouraging. due to the bandwidth range of the Femto variable gain amplifier. There was also an evening haze rolling in during this acquisition time.17. A direct comparison with OC3 data is not available. which could affect alignment over such a long range. Compared with the new data in Fig.
18: BER at 200Mbps measured over the NRLCBD test range with an optical fiber coupled receiver. Note that data is plotted for 5 sec intervals. from Ref. 5. 46. The minimum BER for a 5 sec interval is 1×109. 5. The single error BER for a 1 min acquisition period is 1. Fig. 123 . which is beyond the range of the graph.17: BER at OC3 measured over a 15 minute period using the CPCPD receiver.072×1010.Fig.
7 Conclusions The experimental results presented in this chapter show the performance enhancement in BER due to the integration of nonimaging optical elements into a FSO optical receiver. sustained errors without the use of the device.7 km retroreflected test range at the University of Maryland. The data shows a factor of 5 improvement in BER when a compound parabolic concentrator was integrated into the optical receiver system of a 100 Mbps link over a 1. The link showed only rare periods of burst errors during this experiment. College Park. as compared to significant. 124 . Both experiments validate the ability of the CPC to significantly reduce the number of data transmission errors due to beam wander in the focal plane of the receiver.4 km Chesapeake Bay test range. Optimized CPCphotodiode combinations will improve on the results presented here.5. An integrated CPCphotodiode device was tested at the Naval Research Laboratory’s 32.
1 Summary of Contributions This dissertation describes two techniques that will enhance the performance of free space optical communication systems. The new data presented here represents significantly improved measurements in both weak turbulence and strong turbulence conditions due to: the examination of a wider range of aperture diameters. the use of 125 . College Park. The experiment presented in this dissertation exceeds the scope of the previously published data. Through experimental studies and comparison with available models and theory. only one set of published data to this point has been available to compare with analytical models. Experimental studies of aperture averaging were conducted in both weak and strong fluctuation conditions over a test range at the University of Maryland. Aperture averaging techniques specifically aim to reduce the atmospheric turbulenceinduced intensity fluctuations that appear on a propagating optical wavefront. Both studies show the ability to mitigate the effects of atmospheric turbulence on the propagating coherent optical wave.Chapter 6 Conclusion 6. while minimizing the size. and power of the system. Due to the technical complexities involved in designing an aperture averaging experiment. onoff keyed FSO communication systems will show a significant performance improvement in data transmission when aperture averaging methods and nonimaging optical elements are incorporated into the system design. weight.
and data acquisition equipment.better quality electronics.7 km link at the University of Maryland. strong turbulence theory is still questionable. The performance enhancements demonstrated in 126 . computers. the CPCPD receiver showed only minimal burst errors over the testing time. the simultaneous measurement of background light levels with aperture averaging data. Promising BER improvements by a factor of 5 resulted from the testing of 100 Mbps data in weak turbulence conditions on the 1.4 km retroreflected link always experiences saturated strong turbulence conditions. Compound parabolic concentrators integrated into optical wireless receivers are shown to counteract the bit error ratio degradation attributed to atmospheric turbulenceinduced beam wander in the focal plane of the receiver. the measurement of Cn2 over the entire length of the aperture averaging test range. and the use of an elevated test range allowed for minimal fetch influences. The 32. This is also the first report of aperture averaging data collected for nonsaturated strong turbulence conditions. Since the behavior of the refractive index spectrum in the energy dissipation range is still in contention. Due to the small link margin and difficulty in link alignment. at optical communication standard rates of OC1 through OC3. This experiment is supplemented by BER testing using a CPCphotodiode combination receiver at the Naval Research Laboratory’s Chesapeake Bay Detachment. This data will be useful in improving the quality of strong turbulence theory and models available for the study of aperture averaging in strong fluctuation conditions. By comparison with previously published NRL data. PRBS bit error ratio tests were not achieved at rates beyond 100 Mbps. College Park. This dissertation is the first report of an experimental characterization of the performance enhancement due to the integration of the CPC into the receiver.
127 . Due to the height of the optical test range at the University of Maryland. more data in strong turbulence conditions should be gathered to complement the data presented here. A new electronic preamplifier and receiver would have to be designed to test the device at data rates beyond 1 Gbps. should make the CPC a mandatory part of any OOK FSO communication receiver. only limited data was collected for aperture averaging in strong turbulence. along with the compact size of the device. It would certainly be interesting to characterize the maximum OOK transmission rate the CPC could handle. 6. an optimized CPCPD device should allow the experimental determination of the extent of a BER performance enhancement. By designing the CPC in such a way that its exit aperture maximizes concentration of optical energy onto the surface of a photodiode. The experimental data presented here provides valuable insights into how FSO receivers may be optimized to improve the transmission of OOK systems. The testing of smaller and larger receiver apertures on a new range that consistently experiences saturated strong turbulence conditions would also be beneficial. data transmission errors due to beam wander in the focal plane could be absolutely canceled. Regarding the CPC experiments.this dissertation. Although the data spans both weak and strong turbulence conditions.2 Future Work This dissertation has begun to address perhaps the most significant problem plaguing the design of FSO systems: the lack of experimental data pertaining to the effects of atmospheric turbulence on the propagation of coherent optical waves.
Although the Hill spectrum has been shown to be the most accurate in strong turbulence conditions. the fact that it needs to be solved numerically makes it difficult to use in practical situations.Finally. still leave room for improvement. which do not address the practical measurements of the irradiance variances in strong turbulence conditions. like the modified and effective atmospheric spectrums. the data presented here should be studied to improve the quality of the refractive index spectrums designed for strong turbulence. The new aperture averaging data presented here is valuable because it directly measured the irradiance variances in strong turbulence. spectrums have been derived from logirradiance variance statistics. Theoretical researchers need to address the refractive index spectrum in strong turbulence by studying experimental data collected for the strong turbulence irradiance variance. 128 . Approximations to the Hill spectrum. In the past.
. 1970.S. [4] Solimeno. M.6 µm. 7.” Proceedings of the 1970 International Symposium on Information Theory. SPIE. IEEE: New York. Free Space Laser Communication Technologies IX.). Vol. Vol. [8] Ohtsuki. 8293. (ed. Vol. P. Lightwave Technol. [7] PolakDingles.” J. No. Vol. “Optical communication in turbulent media. 26548. “Performance update of a terrestrial lasercommunications link with quadrature amplitude modulation. Topics in Applied Physics. No. 3489. May 2001. Lang. 40611. “Performance Analysis of Atmospheric Optical PPM CDMA Systems. 5. 2.. et al. 196570. and K. Sam et al. p.References [1] Strobehn. et al. April 1997. Tomoaki.W. R. Vol. No. “Ten Hottest Technologies. p. S. Vol. Opt. SpringerVerlag: New York. [5] Masud.” Proc. “Space instrumentation for laser communications. 2990. Tiffany Jing and Murat Uysal. 10. [6] Li. 1978. 6. June 1969. 21.” Telecommunications Magazine. et al. [2] Lucy. p. 25. [3] Lipsett. J. 2. p. Oct. 1968.F. February 2003. “Optical communications experiments at 6328 Å and 10. 2003. Laser Beam Propagation in the Atmosphere.” IEEE J. GLOBECOMM.. 129 . Elec. Quant. “Achievable Information Rate for Outdoor Free Space Optical Communication with Intensity Modulation and Direct Detection.” Appl.” Proc. p.
“Aperture Averaging of Scintillation. 2. Philips. [18] von Karman. No. [12] Vinnichenko. 5309. p.” Proc. John Wiley & Sons: New York. [13] Wheelon. Strobehn.” IEEE Trans. [10] Fried. 1984. An Introduction to Atmospheric Physics. Vol.” Appl. 51. 12339. 2001. p. Washington. Soc. “Progress in the Statistical Theory of Turbulence. No. “Aperture averaging of optical scintillations in the turbulent atmosphere. “A Survey of ClearAir Propagation Effects Relevant to Optical Communications. Cambridge University Press: Cambridge. [17] Lawrence. M. and John W. Xiaomihng and J. Electromagnetic Scintillation.1. 198294.L. 1980. 20 May 1991. p. November 1948. [14] Andrews. David G. [15] Panofsky. August 2003. October 1970.K. 8. p. 34. Larry C. Vol. D. 152344. and Ronald L.” J. 10. “Performance Bounds for Coded FreeSpace Optical Communications Through Atmospheric Turbulence Channels. 16975.. Opt. No. Commun. N. 30. [11] Churnside. IEEE. 15. No. 2000. Cambridge University Press: Cambridge. [16] Andrews. Vol. et al. Geometrical Optics. Vol.” Proc. Laser Beam Propagation through Random Media.. SPIE Optical Engineering Press: Bellingham. Robert S. Theodore. and John A.[9] Zhu. 57. Vol. NAS.. 58. Opt. p. Atmospheric Turbulence: Models and Methods for Engineering Applications. February 1967. Hans A. 130 . Albert D. James H. Turbulence in the Free Atmosphere. Am. 1998. Kahn. Consultants Bureau: New York. Dutton. Vol.
[23] [24] Goodman. Am. No. 63. “Laser Irradiance Propagation in Turbulent Media.” Proc. Soc. 2003. 8929. “Optical Resolution Through a Randomly Inhomogeneous Medium for Very Long and Very Short Exposures. “Strong fluctuations in the intensity of light propagated through the atmosphere close to the Earth. Vol. Vol. IEEE.E.M. and S. Clifford. Larry C. 7. Statistical Optics. et al. 1965. Soc. No. 68. James H. 8. and A.L.gov>). Electromagnetic Scintillation. p. Gurvich. “Modified spectrum of atmospheric temperature fluctuations and its application to optical propagation. Vol. No. Albert D. 2. 10.” J.W. [21] Wheelon. 5115. et al..” J. [27] Prokhorov. Am. p. “ApertureAveraging Factor for Optical Propagation Through the Turbulent Atmosphere. 5. [26] Gracheva. (available from the National Technical Information Service <http://www. 1978. Weak Scattering.[19] Hill. J. p. Washington. SpringerVerlag: Berlin. Radiophys.” NOAA Technical Memorandum ERL WPL188. 2001.). Wiley: New York. Electromagnetic Wave Propagation in Turbulence.. 131 . Opt. Vol.. [20] Andrews. Laser Beam Scintillation with Applications. [25] Fried. May 1975. 1994.ntis. [22] Sasiela.J. D. November 1990. A. 2000. 13729.” Sov.F. R. Churnside.. Opt. (ed. 56. 4. M. Cambridge University Press: Cambridge. No. 790811.S. SPIE Press: Bellingham. Vol. Richard J. p. October 1966.
” Waves Random Media..C.. p. W. “Intensity covariance of a point source in a random medium with a Kolmogorov spectrum and an inner scale of turbulence. and H. No. 2000. “Simultaneous measurements of angular scattering and intensity scintillation in the atmosphere.” J. p. [30] Frehlich.gov>). Vol. Opt. 73. Vol. Opt. No.ntis. 10428. 4. [29] Fante. Am. Soc. 3. [31] Hill. Vol. Vol. p. S.T. 72.G.” J. James H.A. Opt. p. [34] Andrews. Am. “Innerscale size effect on the scintillations of light in the turbulent atmosphere.F. Am. Soc.” J. Soc. 4. and R. 64. “Aperture averaging of optical scintillations: power fluctuations and the temporal spectrum. Yura. March 1983. Opt. Soc. 2.. Vol. No. February 1987. 12. R.” J.C.. 8. p. p. L. No.J. No. Opt. Soc. 6. “Scintillation from a Diverging Laser Beam. 9. June 1981. Am. December 1974. 71.” NOAA Technical Memorandum ERL ETL240. p. Clifford. 5370. “Theory of saturation of optical scintillation by strong turbulence for arbitrary refractiveindex spectra. 277281. Am. “Apertureaveraging factor for optical scintillations of plane and spherical waves in the atmosphere.[28] Coles. No.. R. et al. Vol.F. L. September 1994. 3606. 10. “Equivalence of two theories of strong optical scintillation. Ronald L. Am.” J. Soc. Vol. Frehlich. 132 . Opt. (available from the National Technical Information Service <http://www. 67586. 597600. 16414.G.. [33] Andrews. [32] Clifford. [35] Churnside. April 1992. and S. August 1982.” J.
R. and Richard J.” Appl. 92630.. [43] Derickson. “Wander of an optical beam in the turbulent atmosphere. James H. Vol. Vol. Rod G.” Opt. 28. 20 February 1996. 24302. 24. Frehlich. et al. “Opticalscintillation method of measuring turbulence inner scale. “Estimation of the parameters of the atmospheric turbulence spectrum using measurements of the spatial intensity covariance.L. 35. [42] Churnside. James H. 38. 133 . Opt.[36] Hutt. p. 29. p. No. Opt. June 1988. August 1999. 7. 15. No. Opt. T.: London. Opt. Opt.J.” Appl. Vol. [41] Hill. 412632. p. [37] [38] Oke.. p.. “Aperture size and bandwidth requirements form measuring strong scintillation in the atmosphere. Prentice Hall PTR: New Jersey.. 20. Hill. No.” Appl. 1 August 1985. Vol.” Appl. “Onset of strong scintillation with application to remote sensing of turbulence inner scale. p. Methuen & Co.. 15 November 1981. p. G.). 21948... R. 38224. [40] Hill. Ochs. [44] Churnside. Vol. and Rod. 27. No. 1 October 1989. 19. 128895. Vol. Lataitis. 1998. 1 March 1990. and Reginald J. [39] Frehlich. 22. Dennis (ed. 1978. D.” Appl. “Saturation resistance and innerscale resistance of a largeaperture scintillometer: a case study. 98697. No. No. 8. 11. No. Eng. Fiber Optic Test and Measurement.” Appl. p. Boundary Layer Climates. Reginald J. Gerard R. “Modeling and measurements of atmospheric optical turbulence over land. Vol. 6. Opt.
Opt. Vol. SPIE. [52] Agrawal. Rubén et al. 39. “The Effect of Atmospheric Turbulence on BitErrorRate in an OnOffKeyed Optical Wireless System. John Wiley & Sons: New York. 423943. Robert K. Olga and Larry C. 2. Vol.” Opt. No. IEEE Press and Oxford University Press: New York. 4489. No. 21. et al. 15. Optical Wireless Communications II. 134 . p. p. 33041. “Model for a partially coherent Gaussian beam in atmospheric turbulence with application in Lasercom. [50] Miñano. R.” Proc. [46] Tyson. W. 3850. p.” Appl. Wave Propagation and Scattering in Random Media. 2. and Douglas E.” Proc.. and Igor I. 1997.T. Eng. “Dielectric compound parabolic concentrators. February 1976. 27407. Canning.. No. Winston. Akira. [51] Mohedano. [48] Welford. October 2000. Academic Press: San Diego. 20 July 2003. “Ultracompact nonimaging devices for optical wireless communications. No.. FiberOptic Communication Systems. 126137. High Collection Nonimaging Optics.[45] Ishimaru.. [49] Winston. Vol. Eng. Vol.” Appl. February 2004. Vol. “Ultra compact optics for optical wireless communications. Andrews. Smolyaninov. FreeSpace Laser Communication and Laser Imaging. Opt. Juan C. p. Govind P. 10. 8090. p. SPIE. [47] Korotkova. “Indirect measurement of a laser communication biterrorrate reduction with loworder adaptive optics. 1997. July 2002. September 1999. and R. Christopher C. 42.” Opt. Vol. [53] Davis. 1989. 2912. p. 43.
Isaac I. SPIE. Wright.W.0. “Comparison of laser beam propagation at 785 nm and 1550 nm in fog and haze for optical wireless communications.” IPN Progress Report 42149. 2537. 15 May 2002.” Proc. October 2001.5Gbps. Naval Research Laboratory. 474485. Free Space Laser Communication and Laser Imaging II.S. [59] “Accurately Estimating Optical Receiver Sensitivity. CA. 4821. 5160. SPIE. Maxim Integrated Products. December 2002. “Freespace highspeed laser communication link across the Chesapeake Bay. [57] Gilbreath.25Gbps/2. January 2003. [58] “Maxim 1. C. 1. [56] Moore. Pasedena. February 2001. A.5V. Vol. and M. p.[54] Biswas. “Research in Freespace Optical Data Transfer at the U. “MountainToptoMountainTop Optical Link Demonstration: Part I. p. +3V to +5. Maxim Integrated Products. FreeSpace Laser Communication and Active Laser Illumination III. Optical Wireless Communications III. SPIE. 22533. p. June 2000.” Rev. 135 . [55] Kim. C.” Proc.” Proc. LowNoise Transimpedance Preamplifiers for LANs.0. et al.S.” Application Note: HFAN3. Vol. Jet Propulsion Laboratory. 4214. et al. and W. Vol.G.I. Rabinovich.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.