Thesis PDF

Ultra Wideband Ranging and Link Budget Design for Naval Crane
Applications
Haris I. Volos
Thesis submitted to the Faculty of the

Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Master of Science
in
Electrical Engineering
R. Michael Buehrer, Chair

Jeffrey H. Reed
Steven W. Ellingson
June 1, 2006
Blacksburg, Virginia
Keywords: UWB, Ranging, Link Budget

Copyright 2006, Haris I. Volos
Ultra Wideband Ranging and Link Budget Design for Naval Crane
Applications
Haris I. Volos
(ABSTRACT)
In this thesis a UWB-based ranging scheme is designed, simulated, implemented and tested.
This system was designed to address the problem of safely unloading cargo crates to ships in
the open seas. UWB antennas are placed on the four corners of the cargo crate, providing the
information needed to a ranging/positioning algorithm that estimates the orientation and
distance of the ship’s deck from the crate. Furthermore, the system is successfully tested
in a 1/24 scale demonstration. In addition to the UWB ranging application, this thesis
evaluates an already proposed modification to the traditional narrowband link budget based
on the Friis transmission formula. The proposed modification replaces frequency-domain
parameters with time-domain values to handle the wide bandwidth of UWB systems. The
proposed approach is shown via measurements to be much more accurate than the traditional
technique.
Acknowledgments
First of all, I would like to express my gratitude to my advisor, Dr. R. Michael Buehrer, for
his constant support and encouragement which have inspired and motivated me throughout
the course of the writing of this thesis. I would also like to thank the other members of my
thesis committee, Dr. Jeffrey Reed and Dr. Steven Ellingson, for their willing participation
in the process.
I must also thank the Fulbright Commission of Cyprus for selecting me to be one of the recip-
ients of the Cyprus America Scholarship Program (CASP). Further, I greatly appreciate the
people of AMIDEAST who have been very helpful and understanding in their administering
of my scholarship, especially Quincy Demordy, my scholarship advisor. This scholarship has
made my studies at Virginia Tech possible.
I wish to acknowledge Joseph Gaeddert for sharing his knowledge with me on all the white
boards we filled up during our fruitful discussions. I can’t forget to thank my office mate,
Haesoo Kim, for his help as well in many ways. I would also like to acknowledge the
contributions and support of Chris Anderson, Swaroop Venkatesh, Natalia Rivera, and Rekka
Mennon. The staff of MPRG has also provided timely help and support. Furthermore,
I would like to thank Randall Nealy, staff engineer of VTAG, for all his help with the
UWB antennas and crate. In addition, I thank Nader A. Nayfeh for his cooperation for the
development of the UWB ranging demo.
Among my many friends I want especially to thank Alexandros Arsalis, Vasilis Vlachakis,
Martin Holzer, and Scott Russell for being there for me. Of course I must thank my parents
and my sisters for all their love and support. Finally I thank God for being by my side.
iii
Contents
1 Introduction 1
I Ultra Wideband Ranging 3
2 UWB Ranging Theoretical Development 4

2.1 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Using UWB Pulses for ranging . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Estimating Crate Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3.1 Estimating Height above a Level Deck . . . . . . . . . . . . . . . . . 5
2.3.2 Estimating Height Above Deck with slope ϕ . . . . . . . . . . . . . . 7
2.3.3 Determining Height and Orientation Relative to Deck . . . . . . . . . 9
2.4 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Arrival Time Estimation Accuracy . . . . . . . . . . . . . . . . . . . . . . . 19
2.6 Sample Link Budget using Radar Equation . . . . . . . . . . . . . . . . . . . 19
2.7 Another Crane Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.7.1 Initial Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7.2 Future Development . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3 UWB-based Ranging:Hardware Demonstration 26

3.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Setup and Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Equipment List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
iv
3.2.2 Transmitter and Receiver . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2.3 Crate & Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2.4 Supporting Circuitry . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3 Software Application (LabVIEW) . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3.1 Single-Side Ranging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.2 Four-Side Ranging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.3 Sub-blocks Used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.4 Other Techniques Used . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4 Demo Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.1 Single-Side Ranging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.2 Four-Side Ranging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4 UWB Ranging Conclusions 46
II Ultra Wideband Link Budget Design 48
5 Ultra Wideband Link Budget Design 49

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2 The Traditional Narrowband Link Budget . . . . . . . . . . . . . . . . . . . 50
5.2.1 The Friis Transmission Formula . . . . . . . . . . . . . . . . . . . . . 50
5.2.2 Link Budget . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.3 The Proposed I-UWB Link Budget . . . . . . . . . . . . . . . . . . . . . . . 53
5.3.1 Receiver Structure Dependence . . . . . . . . . . . . . . . . . . . . . 56
5.4 Link Budgets for Systems Other than Free-Space . . . . . . . . . . . . . . . 58
5.5 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.5.1 Vector Network Analyzer (VNA) Measurements . . . . . . . . . . . . 59
5.5.2 Time Domain Digital Sampling Oscilloscope Measurements . . . . . 73
5.5.3 Link Budget Calculations . . . . . . . . . . . . . . . . . . . . . . . . 85
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
v
List of Figures
2.1 Problem Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Ranging Sensors’ Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Reflected Path Distance Estimation . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 n = a + b Minimum Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 Reflection With a Slope ϕ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.6 Crate’s Dimensions, Heights, and Slopes . . . . . . . . . . . . . . . . . . . . 10
2.7 Sample Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.8 Crate Dimensions Used for the Simulation . . . . . . . . . . . . . . . . . . . 13
2.9 Estimation Variance Vs Assumed Height (Crate Dimensions dw = 5m, dl = 5m) 15
2.10 Estimation Variance Vs Input Variance (Crate Dimensions dw = 5m, dl = 5m) 16
2.11 Estimation Variance Vs Assumed Height (Crate Dimensions dw = 5m, dl =
10m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.12 Estimation Variance Vs Input Variance (Crate Dimensions dw = 5m, dl = 10m) 18
2.13 Acquisition Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.14 Estimation σ CRLB and Simulation Results Using Gaussian Pulses . . . . . 20
2.15 Crane Illustration (not to scale) . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.16 Required Crane Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.17 Tx Location Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1 Demonstration General Layout (not to scale) . . . . . . . . . . . . . . . . . . 26

3.2 Demonstration Setup Block Diagram . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Demonstration Setup Photos . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
vi
3.4 Crate Photo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.5 Crate Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.6 Received Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.7 Crate Inside Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.8 Switch Driver Circuit Photo . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.9 Switch Driver Circuit Schematic . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.10 LabVIEW Application Screen Shots . . . . . . . . . . . . . . . . . . . . . . . 35
3.11 Actual LabVIEW Block Diagram Sample Screen Shot . . . . . . . . . . . . . 36
3.12 Single-Side Ranging Application Block Diagram . . . . . . . . . . . . . . . . 37
3.13 Four-Side Ranging Application Block Diagram . . . . . . . . . . . . . . . . . 38
3.14 Received Window Sample Screen Shot . . . . . . . . . . . . . . . . . . . . . 40
3.15 Find Pulse’s Location Sub-block . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.16 Extreme Value Rejection Sub-block . . . . . . . . . . . . . . . . . . . . . . . 41
3.17 Sample Ranging Output: Motion all the way up and then down . . . . . . . 43
3.18 Automatic Landing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.1 VNA Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

5.2 Ridged TEM Horn Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.3 Bicone Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.4 Vivaldi Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.5 TimeDomain Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.6 The Generated Pulse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.7 Calcuated 1m Received Pulses . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.8 The Standard Gain Horn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.9 Antenna Gain Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.10 Gaussian & Gaussian Monocycle Pulses . . . . . . . . . . . . . . . . . . . . . 71
5.11 GAP Vs Center Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.12 Time Domain Measurements Equipment Setup . . . . . . . . . . . . . . . . . 74
5.13 LOS Measurement Location . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
vii
5.14 Ridged TEM Antenna Results . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.15 Bicone Antenna Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.16 Vivaldi Antenna Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.17 TimeDomain Antenna Results . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.18 NLOS Measurement Locations . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.19 NLOS Bicone Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.20 Rake Receiver Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.21 Ridged TEM Antenna Received Energy and Link Budget . . . . . . . . . . . 89
5.22 Ridged TEM Antenna Received Peak Power and Link Budget . . . . . . . . 90
5.23 Bicone Antenna Received Energy and Link Budget . . . . . . . . . . . . . . 90
5.24 Bicone Antenna Received Peak Power and Link Budget . . . . . . . . . . . . 91
5.25 Vivaldi Antenna Received Energy and Link Budget . . . . . . . . . . . . . . 91
5.26 Vivaldi Antenna Received Peak Power and Link Budget . . . . . . . . . . . . 92
5.27 TimeDomain Antenna Received Energy and Link Budget . . . . . . . . . . . 92
5.28 TimeDomain Antenna Received Peak Power and Link Budget . . . . . . . . 93
viii
List of Tables
3.1 Single-Side Sample Measurements (mm) . . . . . . . . . . . . . . . . . . . . 44

3.2 Four-Side Sample Measurements . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.1 Conventional Free Space Link Budget . . . . . . . . . . . . . . . . . . . . . . 53

5.2 Proposed UWB Free Space Link Budget — Correlator Detector . . . . . . . 57
5.3 Proposed UWB Free Space Link Budget — Peak Detector . . . . . . . . . . 58
5.4 GAP for Energy (dBm2 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.5 GAP for Peak Power (dBm2 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.6 Estimated Antenna Gains . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.7 Calculated GF riis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.8 Sample GAP for different pulses . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.9 Received Antenna Noise (500 waveform average) . . . . . . . . . . . . . . . . 74
5.10 Average NLOS Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.11 TimeDomain Antenna Example Link Budget Calculations at r = 15.68m—
Correlator Detector (LOS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.12 TimeDomain Antenna Example Link Budget Calculations at r = 15.68m—
Peak Detector (LOS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.13 Bicone Antenna Example Link Budget Calculations at r = 7.43m—Correlator
Detector (NLOS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
ix
Chapter 1
Introduction
Ultra Wideband is a technology that until February 2002 was only used for radar, sensing,
military communications and other niche applications. In February 2002, the FCC allowed
UWB to be used for commercial data communications, radar and safety applications [1].
The FCC’s ruling to allow UWB usage created a wide unlicensed band of spectrum (3.1-10.6
GHz [5]) with the potential for extremely high data rate (Gbps). However, because the
allocated bandwidth overlaps with frequencies that are already used by other licensed sys-
tems, the FCC limited the maximum power transmitted in the whole band to approximately
0.5 mW, therefore limiting UWB for short range communications. However, data rate can
always be traded for distance [1].
The FCC defines UWB as a signal with either a fractional bandwidth of 20% of the center
frequency or an absolute bandwidth of 500 MHz when the center frequency is above 6 GHz [5].
This requirement can be satisfied using techniques like direct sequence (DS), spread spectrum
(SS), multicarrier (MC), and impulse radio signals. The latter is the most promising method
because systems utilizing it can have potentially low complexity and low cost, a noise-like
signal, resistant to severe multipath fading and jamming, and have very good time domain
resolution allowing for location and tracking applications [1].
This thesis will only consider impulse based UWB (I-UWB) and will examine two aspects of
UWB, each in its own part: First UWB based Ranging for naval crane applications including
theoretical development and demonstration, and second, UWB Link Budget design proposing
a modification to the traditional narrow band link budget to make it more suitable for UWB
systems.
Part I: Ultra Wideband Ranging An UWB pulse is a good candidate for ranging
applications because of its short duration which can provide a resolution of millimeters.
UWB is already being used in ranging applications for determining the location of objects
[2, 3]. In a ranging application, a UWB pulse is transmitted, reflected, and received. By
1
CHAPTER 1. INTRODUCTION 2
knowing the pulse’s round-trip travel time and propagation velocity, the distance between
the UWB ranging device and the object that reflects the pulse back can be calculated.
An application that addresses the problem of safely and smoothly unloading crate cargo to
ships that fluctuate because of sea waves will be examined both in theory and with a 1/24
scale demonstration.
Part II: Ultra Wideband Link Budget Design With the introduction of UWB systems
for communications, engineers need to be able to estimate the performance of a communica-
tions system when designed. The performance of a communications link is directly related
to the ratio of the received signal power to the system noise power. The procedure of es-
timating the signal-to-noise ratio of the link is often termed calculating a link budget. In
order to calculate the signal power to noise power ratio the received power has to estimated.
For traditional narrow band systems the Friis transmission formula is used. However, as will
be demonstrated in this part of the thesis the Friis formula is not the optimal choice for an
I-UWB system. This part of the thesis presents a suggested modification to the Friis formula
which was first presented in [5] followed by a validation section evaluating the proposed link
budget by taking necessary frequency and time domain measurements.
Part I
Ultra Wideband Ranging
3
Chapter 2
UWB Ranging Theoretical

Development
2.1 Objective
In this part of this thesis an application of UWB ranging will be examined. The scenario, as
is shown in Figure 2.1 includes a crane located on land or on a ship that is trying to unload
a crate to a ship as smoothly as possible. The ship is expected to move up and down and
to have a slope as well. In order for the crane to accommodate for the ship’s fluctuations
the crate is to be equipped with UWB receivers and transmitters in order to determine the
relative ship’s distance and orientation to the crate.
CRATE
SHIP
CRANE
Figure 2.1: Problem Scenario
4
CHAPTER 2. UWB RANGING THEORETICAL DEVELOPMENT 5
2.2 Using UWB Pulses for ranging

UWB pulses can be used in ranging applications. A UWB pulse is sent, reflected, and
received back, and by knowing the pulse’s travel time (assuming that the pulse travels at the
speed of light) the distance between the UWB ranging device and the object that reflects
the pulse back can be calculated. However, this task is not always easy because the pulse
might not take the expected path, and/or strong reflections of other paths might confuse
the ranging device. Therefore, a careful design of a ranging device that will be as immune
to unwanted conditions as possible is desired.
The proposed design suggests that the crate has four ranging devices at its four corners that
aim to find the distance of each side of the crate to the ship’s deck. This design is illustrated
in Figure 2.2.
CRATE
RX TX
TX RX
Figure 2.2: Ranging Sensors’ Location
2.3 Estimating Crate Orientation
2.3.1 Estimating Height above a Level Deck
For this problem the path traveled by the pulse must be predicted, in order to estimate its
traveled distance. When a pulse is sent, usually the receiver also observes copies of the same
pulse shifted in time (due to multiple reflections). Because of this, it is of interest to predict
the path of the first arriving pulse, which is also the shortest path. The predicted travel
path for this application is a simple reflection from the ship’s deck. Furthermore, it is known
from physics that the angle of incidence of wave is equal to the angle of reflection. The travel
path of a reflected pulse is a triangle with sides d (the base of the triangle which is equal
to the separation distance between the transmitter and the receiver), and a & b as the other
two sides as shown in Figure 2.3. This path of the pulse will be useful if it is proven that
the distance a + b is minimum when a = b.
d
d −x x
a b
ϕθ
Figure 2.3: Reflected Path Distance Estimation
n Vs x, d=10, h=10 a and b Vs x, d=10, h=10
a
23.4 b
13
23.2 12.5
23 12
n=a+b
11.5
22.8
11
22.6
10.5
22.4
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
x x
(a) n = a + b Vs x (b) a and b Vs x
Figure 2.4: n = a + b Minimum Point
Referring to Figure 2.3, from trigonometry:

d−x
sinϕ = (2.1)
a
x
sinθ = (2.2)
b
n is defined as:
n=a+b (2.3)
Substituting a and b:
d−x x
n= + (2.4)
sinϕ sinθ
Taking the first derivative with respect to x:

dn 1 1
=− + (2.5)
dx sinϕ sinθ
Setting the first derivative to zero:
1 1
= ⇔ sinϕ = sinθ ⇔ ϕ = θ (2.6)
sinθ sinϕ
d2 n
dx2
= 0, therefore, a plot (Figure 2.4) is required that there is a minimum point at ϕ = θ.
Also it’s known that:
h
a= (2.7)
cosϕ
h
b= (2.8)
cosθ
Therefore, if ϕ = θ, then a = b as well as shown in Figure 2.4(b).
2.3.2 Estimating Height Above Deck with slope ϕ
The ship’s deck is assumed to not only vary its distance from the crate, but also to have a
slope ϕ (illustrated in Figure 2.5). Essentially what is needed is a procedure to find h given
d, n = a + b and ϕ. That procedure will be developed below.
Referring to Figure 2.5, ϕ is the slope of the ship’s deck, the top line is the crate (d), and
n = a + b is the distance traveled by the pulse. Let
a = a1 + a2 (2.9)
as shown in Figure 2.5, using Law of Sines (On the triangle with sides a, b and d)
sin(ϕ + θ) sin(180 − 2θ)

=
a d
dsin(ϕ + θ)
a= (2.10)
sin(2θ)
sin(θ − ϕ) sin(180 − 2θ)

=
b d
d d
2 2
θ −ϕ ϕ +θ
a1 h1
b
h
a2
h2
θ θ
90 − ϕ
ϕ
Figure 2.5: Reflection With a Slope ϕ
dsin(θ − ϕ)
b= (2.11)
sin(2θ)
Substituting a and b:
n = a+b
d
= [sin(ϕ − θ) + sin(ϕ + θ)]
sin(2θ)
d
= [sinθcosϕ − cosθsinϕ + sinθcosϕ + cosθsinϕ]
2sinθcosθ
2dsinθcosϕ
=
2sinθcosϕ
dcosϕ
= (2.12)
cosθ
d
cosθ =cosϕ ⇔
n
d
θ = arccos( cosϕ) (2.13)
n
Using the law of sines (On the triangle with sides a1 , h1 & d/2):
sin (90 − (−ϕ + θ)) sin(90)

= ⇔
d/ a1
2
d
a1 = (2.14)
2 cos(θ − ϕ)
From Pythagorean Theorem (On the same triangle):

d2
h1 = a21 − (2.15)
4
a2 = a − a1 (2.16)
Using the law of sines (On the triangle with sides a2 , h2 & labeled deck side):
sin (90 − ϕ) sin(θ)

= ⇔
a2 h2
a2 sin(θ)
h2 = (2.17)
cos(ϕ)

d2 a2 sin(θ)
h = h1 + h2 = a21 − + (2.18)
4 cos(ϕ)
An expression for h given a+ b, d and ϕ was just found. However, a mathematical expression
to give a + b given h, d, and ϕ was not found. A MATLAB function that can find the exact
a + b that gives a specific h was written instead. This is useful for creating data for testing
the algorithm by means of simulation.
2.3.3 Determining Height and Orientation Relative to Deck
In the actual application the only available information is the separation distance between
the transmitter and the receiver (d), and the traveled distance (n = a + b), but there is no
information about the slope (ϕ). However, the d’s and traveled distances a + b from the
ranges taken on each of the four sides are available which can be used in order to estimate
the actual slope ϕ. Below is a procedure to find h, given d and n = a + b for each of the four
sides but not ϕ.
Referring to the Figure 2.6, the sides are numbered 1 to 4, and the corners are numbered c1
to c4 . Further,
c3
ϕl
c4 4 c3
c4
c4 c3
ϕw ϕw
1 3 dw
c1 c1 c2
2
c2
xw c2
dl
ϕl
xl
c1
(a) Ship’s Deck Top View
c4 dw c1 c1 dl c2 c2 dw c3 c3 dl c4
hc4 h1 hc1 hc1 h2 hc2 hc2 h3 hc3 hc3 h4 hc4
ϕ w xw xl ϕw xw xl
ϕl ϕl
Side 1 Side 2 Side 3 Side 4
(b) Crate’s Sides
Figure 2.6: Crate’s Dimensions, Heights, and Slopes
h1 − h4 are the heights of the center of the sides 1 to 4 to the deck respectively, and
hc1 − hc4 are the heights of the corners 1 to 4 to the deck respectively.
dw and dl are the distances between the transmitter and receiver for the width and length
side respectively.
Finally, ϕw and ϕl are the slope of the ship’s deck at the width’s side and length’s side
respectively.
Assuming that the heights h1 − h4 and the slopes ϕw & ϕl are known, the difference x
between the corner heights and the center heights can be found as follows:
dw
xw = tan(ϕw ) (2.19)
2
dl
xl = tan(ϕl ) (2.20)
2
If h1 > h3
hc1 = h2 + xl
hc2 = h2 − xl
hc3 = h4 − xl
hc4 = h4 + xl
If h1 < h3 the sign of xl is reversed and the above operations are performed. This is because
the ϕ’s are assumed to be always positive and the only way to tell the exact orientation is
the difference between the heights of the different sides.
If h2 > h4
hc1 = h1 + xw
hc2 = h3 + xw
hc3 = h3 − xw
hc4 = h1 − xw
If h2 < h4 the sign of xw is reversed and the above operations are performed.

|hc1 − hc4 | |hc2 − hc3 |
ϕw = arctan = arctan (2.21)
dw dw

|hc1 − hc2 | |hc3 − hc4 |
ϕl = arctan = arctan (2.22)
dl dl
To find the heights of the corners the following procedure is performed:
1. Assume ϕl = ϕw = 0
2. Find the center heights of sides 1 and 3 using Equation (2.18).

3. Find the center heights of sides 2 and 4 using Equation (2.18).
4. Find the height to corner 1 coming from the center of side 1 and coming from the
center of side 2. Do the same for the other corners. [Equations (2.19) and (2.20)]
5. Average the two heights for each corner.
6. Find two new vlaues of ϕl and ϕw using the four corners.

7. Average the new ϕ angles.
8. Repeat from 2, until the calculated angle ϕ values don’t change.
On more than 10000 tests with h in the range 0.5 to 10.5, dw = 20, dl = 20 and angles from
0 to 20, the maximum % error found was 0.002391 which is attributed to rounding errors.
This indicates that algorithm given a set of valid data, assuming specific heights and slopes,
converges to the assumed heights and slopes. Additional tests will follow in the next section.
Using the corner heights, the ship’s deck can be plotted as shown at Figure 2.7. The top
rectangle represents the crate and the rectangle below represents the ship’s deck.
0
0
-5
-5 -10
-15
-10
-20
-15 40
35 -25
30 40
20
25 -30
15 30
20 20
10 15 15 20
10 10
5 10
5 5
0 0 0 0
(a) ϕl = 8.38o, ϕw = 10.64o (b) ϕl = 14.09o, ϕw = 0o
Figure 2.7: Sample Plots
2.4 Simulation Tests

The proposed algorithm for estimating the height and orientation of the crate is now evalu-
ated by means of computer simulation.
The simulation was done for two arbitrarily chosen scenarios that roughly represent po-
tentially actual dimensions. First, the dimensions of the crate are assumed 5x5x5 m (Fig-
ure 2.8(a)) and the maximum distance from the crate to the deck is 5 m (approximately two
floors high); therefore, the maximum distance from the deck to the tx/rx antennas is 10 m.
Thus the working range for the range estimator is 5 to 10 m. For the second scenario the
crate was no longer assumed to be a cube, but to have the length side (dl ) equal to 10 m,
instead of 5 m (Figure 2.8.)
For each case the height of one of the sides was assumed, and based on that height and the
assumed slopes, the rest of the heights were calculated. Based on the assumed heights the
5m 10m
Tx Rx Tx Rx
CRATE CRATE
10m 10m
5m 5m
DECK DECK
(a) 5m Side (b) 10m Side
Figure 2.8: Crate Dimensions Used for the Simulation
respective traveled distances were also calculated. Finally, a zero mean Gaussian noise was
added to the assumed traveled distances before passing them to the estimation algorithm.
The simulation cases for each scenario were:
• 15 different input variances from 5 dBm2 to -30 dBm2 (i.e. SNR values)
• 11 different heights from 5m to 10m.
• 6 different slopes for each side from 0 to 20degrees.
• 4 different cases for each slope pair [(ϕw , ϕl ),(−ϕw , ϕl ),(ϕw , −ϕl ),(−ϕw , −ϕl ).]
• All the above were repeated for each of the four different sides.
The total was 95,040 different cases for each scenario, repeated twice for improving the
results.
From the results two plots were created for each side. The first plot is the estimation variance
(output of the algorithm) vs the assumed height (Figures 2.9 and 2.11), and the second set
of plots show the estimation variance against the input noise’s variance (Figure 2.10 and
2.12.) For the variance calculations the results from all ϕ were included.
The estimation variance against the height for the the first dimension scenario (all sides
equal to 5 m) is shown in Figure 2.9. It can be observed that the estimation variance does
not vary significantly with the assumed height. Furthermore, it can be seen that for each
input variance case the estimation variance is reduced. Additionally, the results for each
side are almost identical. Moreover, Figure 2.10 shows the estimation variance against the
input variance for the 5 m and 10 m height cases which are found to be almost identical.
The estimation variance was found to have a linear relationship with the input variance to
the algorithm. The estimation variance was found to be approximately 2.5 to 2.8 times less
than the input variance.
For the second dimension scenario (the length side equal to 10 m) the results are shown in
Figure 2.11. It can be observed that the estimation variance does not vary significantly with
the assumed height for the sides 1 and 3, which are the 5 m sides. Furthermore, the results
are very close to the results from the previous scenario. On the other hand, for sides 1 and
4, the estimation variance decreases as the assumed height increases. The reason for this
difference is because the base of the formed triangle is now twice as big as the other cases,
causing the height estimation to be more sensitive to changes in the pulse’s traveled distance
because of the different assumed slopes of the reflection surface. Those fluctuations have
less effect as the height of the formed triangle increases. In short, the estimation variance
depends on the base to height ratio of the formed triangle.
Moreover, Figure 2.12 shows the estimation variance vs the input variance for the 5 m and
10 m height cases which, for sides 1 and 3, are found to be almost identical between each
other and to the graphs shown in Figure 2.10. On the other hand, the 5 m height case, on
the 10 m length sides (2,4), was found to have its estimation variance to be approximately
1.58 to 1.66 times less than the input variance. On the contrary, the 10 m height case was
found to be almost identical to cases in the 5 m length sides. This observation supports the
conclusion that the estimation variance depends on the base to height ratio of the formed
triangle. Furthermore, as the ratio decreases the effect of the changes is reduced as shown
in the 5m length sides in which the 5 m and 10 m height cases are almost identical.
From the simulation results it can be concluded that the estimation algorithm can effectively
use four ranging measurements (one on each side of the crate) to estimate the height and
orientation of a crate relative to a ship’s deck. Furthermore, it was found that when the
height is short compared to the base of the formed triangle the estimation variance is in-
creased. However, in an actual application the reduced height will probably cause a reduced
input variance because the received UWB pulse is likely to be stronger, thus providing more
accurate acquisition.
h1 Estimation Variance Vs Assumed h for various Input Variances h2 Estimation Variance Vs Assumed h for various Input Variances
1 1
10 10
5 dBm2 5 dBm2
0 0
10 10
2.5 dBm2 2.5 dBm2
0 dBm2 0 dBm2
2
−2.5 dBm −2.5 dBm2
Estimated h1 variance (m2)

−1
−5 dBm2 −1
−5 dBm2
10 2 10
−7.5 dBm −7.5 dBm2
−10 dBm2 −10 dBm2
2
−12.5 dBm −12.5 dBm2
−15 dBm2 −15 dBm2
2
−2
10 −17.5 dBm −2
10 −17.5 dBm2
2 2
−20 dBm −20 dBm
2
−22.5 dBm −22.5 dBm2
−25 dBm2 −25 dBm2
−27.5 dBm2 −27.5 dBm2
−3 −3
10 −30 dBm2 10 −30 dBm2
−4 −4
10 10
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10
Assumed h (m) Assumed h (m)
(a) Height of Side 1 (b) Height of Side 2

1 1
10 10
5 dBm2 5 dBm2
0 0
10 10
2.5 dBm2 2.5 dBm2
2
0 dBm 0 dBm2
−2.5 dBm2 −2.5 dBm2
2
−1 −5 dBm −1 −5 dBm2
10 10
−7.5 dBm2 −7.5 dBm2
2
−10 dBm −10 dBm2
−12.5 dBm2 −12.5 dBm2
2
−15 dBm −15 dBm2
−2
10 −17.5 dBm2 −2
10 −17.5 dBm2
2
−20 dBm −20 dBm2
−22.5 dBm2 −22.5 dBm2
2
−25 dBm −25 dBm2
−27.5 dBm2 −27.5 dBm2
−3 −3
2
10 −30 dBm 10 −30 dBm2
−4 −4
10 10
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10
(c) Height of Side 3 (d) Height of Side 4
Figure 2.9: Estimation Variance Vs Assumed Height (Crate Dimensions dw = 5m, dl = 5m)
h Estimation Variance Vs Input Variance for heights 5m & 10m h Estimation Variance Vs Input Variance for heights 5m & 10m
1 2
1 1
10 10
0 0
10 10

−1 −1
10 10
5m 5m
10m 10m
−2 −2
10 10
−3 −3
10 10
−4 −4
10 10
5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30 5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30
Input Variance (dBm2) Input Variance (dBm2)

h3 Estimation Variance Vs Input Variance for heights 5m & 10m h4 Estimation Variance Vs Input Variance for heights 5m & 10m
1 1
10 10
0 0
10 10
−1 −1
10 10
5m 5m
10m 10m
−2 −2
10 10
−3 −3
10 10
−4 −4
10 10
5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30 5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30
Figure 2.10: Estimation Variance Vs Input Variance (Crate Dimensions dw = 5m, dl = 5m)
1 1
10 10
5 dBm2 5 dBm2
0 0
10 10
2.5 dBm2 2.5 dBm2
0 dBm2 0 dBm2
2
−2.5 dBm −2.5 dBm2

−1 −5 dBm2 −1 −5 dBm2
10 2 10
−7.5 dBm −7.5 dBm2
−10 dBm2 −10 dBm2
2
−12.5 dBm −12.5 dBm2
2 2
−15 dBm −15 dBm
2
−2
10 −17.5 dBm −2
10 −17.5 dBm2
2 2
−20 dBm −20 dBm
−22.5 dBm2 −22.5 dBm2
2
−25 dBm −25 dBm2
−27.5 dBm2 −27.5 dBm2
−3 −3
10 2 10 2
−30 dBm −30 dBm
−4 −4
10 10
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10

1 1
10 10
5 dBm2 5 dBm2
0 0
10 10
2.5 dBm2 2.5 dBm2
2
0 dBm 0 dBm2
−2.5 dBm2 −2.5 dBm2
2
−1 −5 dBm −1 −5 dBm2
10 10
−7.5 dBm2 −7.5 dBm2
2
−10 dBm −10 dBm2
−12.5 dBm2 −12.5 dBm2
2
−15 dBm −15 dBm2
−2
10 −17.5 dBm2 −2
10 −17.5 dBm2
−20 dBm2 −20 dBm2
2
−22.5 dBm −22.5 dBm2
−25 dBm2 −25 dBm2
2
−27.5 dBm −27.5 dBm2
−3 −3
10 −30 dBm2 10 −30 dBm2
−4 −4
10 10
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10
Figure 2.11: Estimation Variance Vs Assumed Height (Crate Dimensions dw = 5m, dl =

10m)
h Estimation Variance Vs Input Variance for heights 5m & 10m h Estimation Variance Vs Input Variance for heights 5m & 10m
1 2
1 1
10 10
0 0
10 10

−1 −1
10 10
5m 5m
10m 10m
−2 −2
10 10
−3 −3
10 10
−4 −4
10 10
5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30 5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30

h3 Estimation Variance Vs Input Variance for heights 5m & 10m h4 Estimation Variance Vs Input Variance for heights 5m & 10m
1 1
10 10
0 0
10 10
−1 −1
10 10
5m 5m
10m 10m
−2 −2
10 10
−3 −3
10 10
−4 −4
10 10
5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30 5 2.5 0 −2.5 −5 −7.5 −10 −12.5 −15 −17.5 −20 −22.5 −25 −27.5 −30
Figure 2.12: Estimation Variance Vs Input Variance (Crate Dimensions dw = 5m, dl = 10m)
2.5 Arrival Time Estimation Accuracy

The Cramer-Rao Lower Bound (CRLB) sets a lower bound on the variance of any unbiased
estimator. This can be useful on evaluating the performance and feasibility1 of a possible
estimator.
The estimator we consider here (Figure 2.13) observes the matched filter output of the
received signal for a time interval and then it estimates the time of the maximum peak
(assuming LOS and no significant unwanted reflections).
The CR lower bound of the variance for this estimator was derived in [8] and is given by:
1
στ̂2 ≥ (2.23)
8π 2 SNRβf2
where στ̂2 in seconds squared is the lower bound of the estimation variance of time delay τ ,
SNR is the signal-to-noise ratio, and βf the bandwidth of the signal in Hz.
To express the estimation variance in meters squared Equation (2.23) only needs to be
multiplied by the speed of light.
Received Select time of the

Window with Matched Filter Estimated
Noise maximum arrival time
Figure 2.13: Acquisition Process
Using the proposed acquisition procedure the CRLB was calculated assuming a Gaussian
pulse with pulse width τp = 2πσ [5] equal to 100 ps, 500 ps, 1 ns, and 2ns. The same
pulses were used in a simulation of the proposed scheme. Figure 2.14 shows the result of the
evaluation of the CRLB and the simulation. It can be observed that the proposed estimation
scheme needs 10 dB higher SNR in order to perform as well as the theoretical lower bound
suggested by the CRLB.
2.6 Sample Link Budget using Radar Equation

For further insight, it was decided to explore the requirements for archiving a certain accuracy
from the ranging scheme. We assume a desired estimation variance of 6.25×10−4 m2 (σ=0.025
m=1 inch), for which the input variance must be less than -29 dBm2 (rounded up from
Figure 2.9), therefore the input variance to the estimation algorithm must be less than
1
if it is better than the CLRB is not feasible
Estimation σ CLRB and Simulation Results using Gaussian Pulses

0
10
−1
10
Estimation Standard Deviation (m) −2

10
−3
10 100ps CRLB
500ps CRLB
1ns CRLB
−4
2ns CRLB
10 100ps Simulation
500ps Simulation
1ns Simulation
2ns Simulation
−5
10
−10 0 10 20 30
SNR (dB)
Figure 2.14: Estimation σ CRLB and Simulation Results Using Gaussian Pulses
< 1.25 × 10−3 m2 . From Figure 2.14 the SNR required for this accurancy is at least 26 dB
using a 500 ps pulse.
Assuming the crate dimensions of Figure 2.8(a) the transmitter was assumed to send a pulse
every 150 ns both to the ground and to receiver. As soon as the receiver receives the first
pulse (in 16.67 ns, 5 m distance), it starts looking for the ground reflected pulse over a
150ns window. The pulse for the maximum likehood range estimation (for a height 10 m,
the traveled distance is 21 m) is expected to arrive in 53.3 ns, in addition to the 16.67 ns
that have already elapsed. The additional 80 ns in the window is to avoid InterSymbol
Interference (ISI) due to late arriving paths.
At this point we have an estimate for the SNR needed for the required estimation accuracy
(σ=0.025 m=1 inch.) The last part of this section is to calculate a link budget in order to
get a feel for what system parameters are needed, such as antenna gains, interleaving factor
(how many pulses are needed for a complete waveform), and update rate that will archive
the required SNR. The link budget is evaluated for two cases, case 1 when the crate is at the
maximum ranging height (5 m above the ship deck, 10 m for the tx/rx), and case 2 when the
crate is half way to the ship’s deck (2.5 m.) The link budget is evaluated by estimating the
received power using the radar equation [4] (Equation (2.24).) The radar equation is used
even though might not be the optimal method for a UWB system as explained in Part II of
this thesis. However, it provides an estimate of the needed parameters.
The radar equation is
PT GT GR αλ2
PR = (2.24)
(4π)3 R4
where PR is the received power reflected from a target a distance R meters, PT is the
transmitted power, GT and GR are the gain of the transmit and receive antennas respectively.
α is the target (floor) reflectivity, and λ is the transmission wavelength. The received SNR
is given by:
PR Na Tw
SNRRx = (2.25)
N0 F
where Na is the number of waveforms averaged, Tw is the length of the received window, N0
is the Noise Power Spectral Density, and F is the noise figure of the system.
Combining and rearranging Equations (2.24) and (2.25):
1/4
Na PT GT GR αλ2
Rmax = (2.26)
(4π)3 N0 SNRreq
For the link budget calculations some parameters had to be assumed, such as the interleav-
ing factor, which is how many pulses are required for a practical receiver to construct a full
waveform. The maximum number of pulses sent is 1/Tw , this number divided by the desired
update rate yields the maximum available pulses per update. Finally, dividing the maxi-
mum available pulses per update with the interleaving factor the total number of waveforms
available for averaging is estimated. The two cases to follow trade update rate for more
waveforms to average at the higher height case for improved SNR.
Link budget parameters for case 1 and 2
2
Detection Algorithm required SNR (for an Est < 1.25 × 10−3 m2 ): 27 dB
Transmit Power (PT ): 1 mW
Tx,Rx Antenna Gain: 4 dBi
Floor Reflectivity(α): 0.5
Noise Figure (F): 1.25
Noise PSD (N0 )=4 × 10−21 W/Hz
Pulse Width=500 ps
Detection Window(Tw ): 150 ns
Interleaving factor: 250
Center frequency (fc ): 4GHz
Using the above parameters, the following cases are calculated:
Case 1
Max Traveled Distance (R): 21/2=10.5 m (Height= 10 m)
Update Rate:2.2 Hz
Max Allowed Pulses (Na )= 3 × 106

Waveform Averaging (Nw ):12 × 103
Evaluating Eequations (2.24), (2.25) and (2.26):
SNRRx =27.23 dB Rmax = 10.64 m
The values calculated by the link budget suggest the received SNR is 27.2 dB and the
maximum working range is 21.24 m. Those values agree with the desired SNR and working
range which are 27 dB and 21 m, respectively.
Case 2
Max Traveled Distance (R): 16/2= 8 m (Height=7.5 m)
Update Rate:6.67 Hz
Max Allowed Pulses (Na )= 1 × 106
Waveform Averaging (Nw ):4 × 103
SNRRx =27.19dB Rmax =8.09 m
For this system to work, the link budget suggests high antenna gains, and a low interleaving
factor which was assumed to be 250, the Tektronix CSA8000B can interleave up to 4000
waveforms. However, these values might not reflect typical equipment parameters, there-
fore, further study of actual equipment parameters is required, and an improved acquisition
scheme that will perform with as low SNR as possible is needed. Furthermore, it can be seen
that we can increase the update rate as the crate approaches the deck (SNR increases). The
latter observation suggests that we can trade range for update rate which is very useful for a
practical system that the crane controller when the crate is too high does not need accurate
and fast updates of the range. The need for more accurate and faster range updates increases
as the crate approaches the ship’s deck were care must be taken for its safe landing. Again,
these calculations were done to determine the factors that affect the system’s performance.
2.7 Another Crane Application

Besides the problem presented in this chapter, there is another problem involving a horizon-
tally moving crane on parallel tracks as shown in Figure 2.15 that is of interest.
As the crane moves the crate which is hanging below sways back and forth. The crate’s
oscillation is unwanted because of the danger that it might hit something, and because the
operator can not unload the crate unless it is fully still. An automatic crane controller
can eliminate the sway if the exact position of the crate is known. Therefore, an electronic
system needs to be designed to estimate the location of the crate. An initial idea is to have
transmitters on the top of the crate and receivers on the bottom of the crane. The receivers
will obtain ranging information from the known transmitting points, and by mathematically
manipulating the ranging information, obtain the location of the crate (Figure 2.16). The
transmitters can be replaced by transceivers that will transmit after receiving a trigger signal
from a transmitter on the crane system to avoid the need of RF cables running all the way
Crane
Crane
Tracks
Unwanted
Swing
Unwanted
Crate Swing
Figure 2.15: Crane Illustration (not to scale)
down to the crate.
y-axis
crane
crate x,y?
x-axis
Figure 2.16: Required Crane Location
The major swaying occurs along the direction of the track, and any sway that happens in the
other direction (perpendicular to the tracks) for the time being is assumed to be negligible.
2.7.1 Initial Development
As a starting point, the system can be examined in its simplest form, which assumes two
receiving points and one transmitting point. This will also allow, if so desired, an initial
demonstration for a proof of concept with the equipment that is currently available at MPRG,
which include a two channel Digital Sampling Oscilloscope (Tektronix CSA8000B) and a 30
ps pulser made by Geozandas. The range of movements that can be examined will be
determined by the antennas used for this demo. If common available large UWB antennas
are used the ability of the system to estimate the position can be demonstrated but the
range of motion might be limited.
Operation Description
Again, it is assumed that the crate moves only in one plane. A transmitter is placed on
the crate that will send a UWB pulse. There will be two receiving points on the crane.
From the time needed for the signal to travel from the transmitting position to the receiving
positions the distance between each of the receiving positions to the transmitting point will
be estimated. The exact position of the transmitting point will be estimated by simple
trigonometry. In an actual system the transmitter can be replaced with a transceiver that
will send the pulse when signaled. More transceivers and receivers can be added for additional
accuracy and 3D information.
Transmit Location Estimation
The transmit location can be estimated by knowing the location of the two receive points
and their distances from the transmit location as illustrated in Figure 2.17.
Rx1 Rx2
θ
r1 r2
Tx
Figure 2.17: Tx Location Estimation
Let
r1 : The distance from Rx1 to Tx
r2 : The distance from Rx2 to Tx
d: The distance between Rx1 and Rx2
θ: The angle between the line Rx1-Rx2 and the line Rx1-Tx
Then, it can be shown that the Tx coordinates can be given by the following equations:
r12 − r22 + d2
a= (2.27)
2d

a
θ = arccos (2.28)
r1
xtx = xRx1 + r1 cos(θ) (2.29)
ytx = yRx1 − r1 sin(θ) (2.30)
2.7.2 Future Development
Future development will include simulations of the scheme as initially proposed, as well as
more scenarios in which additional antennas and transmitting points can be evaluated.
Chapter 3
UWB-based Ranging:Hardware
Demonstration
3.1 Description
After investigating the proposed ranging scheme in theory and via simulation, it was desired
to put it to a real test. Therefore, a demonstration was planned for December 2005. The
demonstration’s purpose was to show that using UWB the range below a cargo crate can
be estimated and an automatic crane controller can make use of the ranging information
to land the crate on the platform with minimal impact. Furthermore, a real demonstration
might give insight to conditions not considered during the theoretical development.
≈5feet
Rx Tx
Tx Rx
Crate
Crane
Platform
Figure 3.1: Demonstration General Layout (not to scale)
The demonstration called for a 1/24 scale demonstration using a 5 foot boom crane (Figure
26
CHAPTER 3. UWB-BASED RANGING:HARDWARE DEMONSTRATION 27
3.1) in the ESM’s Non-linear Dynamics Lab. Originally, the system was supposed to give 10
updates/second for all sides to the crane’s control system. However, as it will be explained
later, that requirement could not be met with the available equipment.
3.2 Setup and Specifications

The demo was based around the major equipment comprising the demo, the transmitter, and
the receiver that were already available in MPRG’s radio lab. Only the minimal equipment
had to be purchased and constructed such as the crate with the antennas, an antenna switch,
the four antenna dividers/combiners, a switch driving circuit, a USB Digital I/O interface,
and cables.
3.2.1 Equipment List
The following equipment was used to make the demonstration possible:
• Geozondas Pulser T0.5 < 30ps, Vout > 30V /50Ω (GZ1106DL1 & GZ1117DN-25)
• HP33120A Function Generator
• Tektronix CSA8000B Digital Sampling Oscilloscope (DSO)
• Tektronix PS280 DC Power Supply
• National Instruments USB 6501 Digital I/O interface
• Custom made (by the author) Switch Driving Circuit
• Gateway M460 Laptop computer with LabVIEW (2.0GHz Pentium-M, 1GB RAM)
• Crate with UWB antennas made by Randall Nealy (Virginia Tech Antenna Group)
• United Microwave cables Micropore 190 (3x8ft)
• United Microwave cables Microflex 150 (3x10ft)
• Crossover Ethernet cable for direct connection between the laptop computer and the
DSO
• 9dB attenuator
• Generic cables for trigger connections

Figure 3.2 shows how the equipment was connected, and Figure 3.3 presents an actual photo
of the equipment. The Geozondas pulser was used as the transmitter generating a -30V 30ps
pulse (time width measured at half-amplitude.) The pulser was triggered by the the HP
function generator using a TTL signal of 200KHz. From the pulse the TTL triger signal was
attenuated by a 9dB attenuator in order to safely trigger the Tektronix CSA8000B DSO that
was used as the receiver. The DSO and the Switch Driving Circuit (for the antenna switch in
the crate) were controlled by the laptop. The laptop communicated with the switch driving
circuit and the crane control by the NI USB Digital I/O interface. The only information
sent to the crane controller was the current range information. The crane control software
was developed by Nader A. Nayfeh.
Tektronix PS280
RX & TX Cables:
18ft each: Crate & DC Power Crane Controller
8ft Micropore 190 + Antennas Supply
10ft Microflex 150
8bit wire
ribbon cable
Geozondas Pulse Output Switch Driving NI USB 6501
GZ1106DL1 & U>30V/50Ω
T <30ps Circuit Digital I/O Board
GZ1117DN-25 0.5
Trigger USB Cable

Trigger Input Channel 1 Input
5V Output Tektronix
0V Gateway M460
5μs t Channel 2 Input
CSA8000B
Laptop with
Digital Sampling Ethernet
LabVIEW 7.1
HP33120A 9dB Oscilloscope Cable
Function Attenuator
Trigger Input
Generator
Figure 3.2: Demonstration Setup Block Diagram

The Crate
The Shaker
(a) Crate and Shaker

Laptop with
LabVIEW
Pulse Generator
The Receiver (DSO)
Trigger Generator
200KHz
USB I/O &

Switch Driver
Circuit
Power Supply
(b) Equipment
Figure 3.3: Demonstration Setup Photos

3.2.2 Transmitter and Receiver
As was previously mentioned the transmitter was a simple pulser that produces a -30V 30ps
Gaussian pulse. Because of the availability of only one transmitter and the fact that the
crate had two transmitting sides on the crate, an antenna switch was installed inside the
crate in order to switch between transmitting sides.
The receiver was a Tektronix CSA8000B Digital Sampling Oscilloscope, capable of receiving
50, 4000 sample, waveforms per second when triggered with a 200KHz signal. Unfortunately,
the DSO was not designed for real time transfer of the acquired data to an external device.
The effective rate was at least 16 waveforms when it was continuously receiving the same
signal without interruptions or 3 waveforms per second when it was forced to clear the
acquired data and start receiving anew (which was required every time the transmission
side was changed). This limitation on update rate caused the demonstration to be done in
two ways, one with fast update rate using only one side for ranging, and one with the slow
update rate fully utilizing the four sides and the orientation estimation algorithm which was
developed in the previous chapter.
3.2.3 Crate & Antennas
(a) The Crate (b) Antennas
Figure 3.4: Crate Photo
For the demonstration one of the major components was the crate with the UWB antennas.
Both, the crate and the UWB antennas (Figure 3.4) were designed and built by Randall
Nealy, staff engineer of VTAG. Furthermore, inside the crate additional components (antenna
switch and power dividers/combines) had to be installed, those were installed by the author
with the help of Randall Nealy.
113mm
(4.45")
103mm
(4.06")
258mm (10.16")
(a) Diagram (b) Photo
Figure 3.5: Crate Dimensions
The crate as shown in Figure 3.5 had dimensions of 103 mm (4.06”) wide, 258 mm (10.16”)
long, and 113 mm (4.45”) high. The dimensions are approximately 1/24 scale of a real
container.
Regarding the antennas, an antenna design that will be as directive as possible was desirable.
However, due to the size limitations, the actual antennas were most directive and effective
in frequencies above 10 GHz. This caused reduced received signal levels because most of
the pulse’s energy was below 10 GHz. Also, because of the reduced directivity in the low
frequencies in some cases unwanted reflections were present. However, with the choice of
good quality cables and DSP the only direct effect to the demonstration was to reduce the
range from 5ft to 2ft.1 The sample pulse used for the matched filter detection process is
shown in Figure 3.6(c) and was obtained after applying a bandpass filter of 8 to 18 GHz to
the received signal.
Antenna Coupling
Due to the small dimensions of the crate, the transmitting and receiving antennas were
very close to each other, causing a signal to be directly received from the transmit antenna
(coupling). This coupling was very strong compared with the desired reflected signal. The
coupling was mitigated by recording it during calibration and subtracting it from the received
signal. The received signal and coupling are shown in Figures 3.6(a) & 3.6(b).
1
The range was also limited to 255mm (10”) because only 256 different values could be communicated to
the crate controller and a resolution of 1mm was desired.
Received Signal and Coupling Received Signal (after Coupling Subtraction and Filtering) & Matched Filter Output
0.19 0.03
Received Signal Received Signal
Estimated Coupling Matched Filter Output
0.18 0.02
0.17 0.01
Amplitude (V)
Amplitude (V)
0.16 0
0.15 −0.01
0.14 −0.02
0.13 −0.03
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
time (s) x 10
−9 time (s) x 10
−9
(a) Coupling (b) Received signal with coupling subtracted and

matched filter output
The pulse used for the Matched Filter
0.1
0.05
Amplitude (V)
−0.05
−0.1
−0.15
−0.2
0 1 2 3 4 5
time (s) −10
x 10
(c) The pulse used for matched filtering
Figure 3.6: Received Signals
Crate’s Inside Configuration
The crate had four antenna pairs, one pair at each corner. Two of the pairs were designated
transmitting pairs and the other two pairs were designated receiving pairs. The transmit
and receive pairs were located on alternate corners as shown in Figure 3.7. The signals from
the receiving antennas on each corner were added via RF combiners and the signals to the
transmitting corner antennas were distributed via RF dividers as shown in Figure 3.7.
The parts used insise the crate included:
• 4 × Narda 4456-2 2-Way Power Dividers/Combiners 2-18GHz.

• Microwave Communications Laboratories K2-8-LIS/12P Latching Electromechanical
Switch, SPDT, DC to 26.5GHz.
• 4 × 8.5” United Mircowave Microform 139 Cables.
• 4 × 9.5” United Mircowave Microform 139 Cables.
TX UWB Antenna RX
RF Divider RF Combiner
To RX
Channel 2
From
Pulser
(TX)
RF Combiner Switch Control RF Divider
To RX
Channel 1
RX TX
(a) Diagram (b) Photo
Figure 3.7: Crate Inside Configuration
Figure 3.8: Switch Driver Circuit Photo
3.2.4 Supporting Circuitry
The antenna switch used inside the crate required a driving voltage of 12 V and 100 mA peak
current. The USB I/O card used was not able to meet these requirements. For that reason,
a simple driving circuit was designed. The circuit shown in Figure 3.9 consists of three parts.
First, the power supply’s input voltage is regulated to 12 V using a standard regulator IC
providing power to the 12 V part of the circuit, and then it is further regulated to 5 V for
the 5 V components of the circuit. Second, a simple logic circuit was designed to prevent
the combination “1 1” which means both sides are transmitting. That state is undesired and
must be prevented to avoid potential damage to the driving transistors and to the switch
itself. Third, two high speed switching transistors were used to drive the switch which only
draws current when it is changing state (latching switch.) Furthermore, for additional safety
the current supplied to the circuit from the external Tektronix power supply was limited to
the minimum required for the proper operation of the circuit.
+12V Out
V1 12V V2 5V
COM
COM
+14-16V VIN VOUT VIN VOUT +5V Out
R1 470 R2
C1 100µF C2 C3
100µF 100µF
D2 D1
0V
0V
(a) Power Supply
+12V Supply
5V Area
Q1 NTE2361
Digital Logic U2a U1a 1
Position 1 1 3 Antenna
3 2 2 C Switch
C4081 Terminals
R6 C4049 R4 10K 2
10K
Q2 NTE2361
Digital Logic U2b U1b
Position 2 5 4 D3 D4
5 4 6
C4081
R7 10K C4049 R3 10K
Digital Logic
Ground or 0V
R5 470
0V
(b) Logic and Driver Part (c) Switch
Figure 3.9: Switch Driver Circuit Schematic
3.3 Software Application (LabVIEW)

The application that will read the waveforms, control the switch, process and display the
required information was developed in LabVIEW because it provided all the necessary DSP
functionality and the ability to connect with external peripherals such as the DSO and the
USB I/O interface. Two applications were developed: first, an application utilizing the one
side for ranging; and second, after the one side was perfected, the application was expanded
to four sides. Figure 3.10 shows screenshots from the two applications, one for single-side
ranging, and three samples from four-side ranging. As can be seen from the screen shots the
results were not limited to numbers, a 3D visual representation was also developed. Figure
3.11 shows a sample screen shot of an actual LabVIEW block diagram.
(a) One Side Ranging (b) Four Sides Ranging
(c) Four Sides Ranging (d) Four Sides Ranging
Figure 3.10: LabVIEW Application Screen Shots
3.3.1 Single-Side Ranging
This is a description of the single-side ranging application referring to the application’s block
diagram shown in Figure 3.12.
The Initialize sub-block establishes connections with the external devices (DSO, USB Dig-
ital I/O), reads the reference pulse to be used for the matched filter process, and initializes
all the necessary variables. The main program starts with Get waveform that communicates
with the DSO and gets the latest acquired waveform. Then, Subtract coupling subtracts
the current estimate of the coupling from the received waveform. Furthermore, the Filter
waveform sub-block applies a 0.8-20 GHz bandpass filter to remove out-of-band noise in
the received signal. Additionally, Find pulse’s location and traveled distance ap-
Figure 3.11: Actual LabVIEW Block Diagram Sample Screen Shot
plies the acquisition method in order to find the pulse’s location and estimate the distance
traveled by the pulse.—This will be discussed in more detail in its own section.—Moreover,
Estimate distance applies the Pythagorean theorem to find the height based on the trav-
eled distance assuming a right triangle with a base equal to the half antenna separation
and a hypotenuse half the traveled distance. Finally, Non-linear filtering smooths the
output by eliminating sudden output changes that might occur due to incorrect estimation.
The operation of this block will be also discussed in more detail in its own section.
The Add waveform to coupling database function helps update the coupling database
which is being maintained for the coupling estimation. Initially, the database is filled during
the calibration process, but later it is continuously updated with new waveforms. When
a new waveform is received and the pulse’s location is estimated, a half window waveform
centered in the pulse’s location is replaced with the current coupling estimation, and the rest
of the waveform is included in the database. This allows active coupling estimation without
recording the actual pulse.
3.3.2 Four-Side Ranging
The block diagram of the four-side ranging application is shown in Figure 3.13. The basic
components are the same as the single-side application repeated four times, however, there
Establish connections with the

oscilloscope and the digital I/O
board, and initialize waveforms.
The current The distance traveled by the
estimated coupling Apply filter a pulse is returned.
Initialize The is subtracted from 0.8-20GHz filter Otherwise, if the pulse is not
received the received to remove out of found a default value is
waveform waveform band noise return to indicate this.
Find pulse’s The distance is

Get waveform Subtract Filter location and the Estimate distance estimated using
coupling waveform travelled distance Pythagorean
therorem.
If the current value is

Non-Linear found too off the
Reads the Filtering expected value then it
waveform is replaced with an
from the Add Waveform to expected value
oscilloscope coupling database
Display reading Send to crane

The current waveform is added to the coupling database after the controller
region near the observed pulse’s location is replaced with the
current estimated coupling waveform. This is done in order to
avoid the reflected pulse to be included in the coupling estimation
Repeat until stoped
Figure 3.12: Single-Side Ranging Application Block Diagram
are some differences. First, at each iteration the transmit side alternates. Second, because of
the transmit side alternation at each iteration only two waveforms are processed, therefore,
only two sides have updated values. For the other two sides the last known values are used.
Third, in this application no data are sent to the crane controller. And fourth, the estimation
algorithm for full orientation estimation (as described earlier in this thesis) is utilized.
Establish connections with the

oscilloscope and the digital I/O
board, and initialize waveforms.
Initialize
Tx 1
Get waveforms Tx Side?
Tx 2
The blocks used here are the same as in
the one side ranging application.
Therefore, the same comments apply.
The process is simply repeated four
times for each side.
Ch 1
Subtract Filter Find pulse’s Non-Linear Tx 1, Ch1 Side
coupling waveform location and the Filtering

travelled distance
Add Waveform to
coupling database
Ch 2
Subtract Filter Find pulse’s Non-Linear Tx 1, Ch2 Side

travelled distance
Add Waveform to
coupling database
In each iteration
Ch 1
only two new values
are available, for the
other two, the last
Subtract Filter Find pulse’s Non-Linear Tx 2, Ch1 Side known values are
coupling waveform location and the Filtering used. The algorithm
travelled distance used is the one
developed earlier in
this thesis.
Add Waveform to
coupling database
Ch 2 Tx 2, Ch2 Side
Subtract Filter Find pulse’s Non-Linear

travelled distance Estimate corner
heights and
orientation
Add Waveform to
coupling database
Display readings
Repeat until stoped
Figure 3.13: Four-Side Ranging Application Block Diagram

3.3.3 Sub-blocks Used
Record Coupling & Calibrate Procedure
Find Pulse’s Location
This sub-block gets the input waveform from the main program, then it applies an expo-
nential threshold to received signal. The exponential threshold is applied knowing that the
signal’s level decreases exponentially. This was found to eliminate wrong detections due to
imperfect coupling cancellation. Furthermore, in parallel a match filter with a threshold is
also applied to the received waveform, then both results are combined to get a waveform
that will meet both the exponential threshold and the matched filter threshold.
After the above criteria are applied to the received waveform, if the last location of the
pulse is known, the detection is limited around the last observed location. This provides
a simple tracking knowing the pulse should be near the detected location. However, if no
pulse is found satisfying detection criteria, then the whole received window is considered for
detection. Finally, if no pulse is found in the whole window then the sub-block returns some
default values that indicate that the pulse was not found, and probably was out of range.
If a pulse is found, the second strongest peak is also found and if it is at least 90% as strong
as the first peak then it is also considered. It was observed for some cases that near the
desired pulse will be another pulse consistently before or after the pulse, depending on the
side. For that reason the program was instructed to give priority to one of them. The second
pulse was given priority when the pulse was received on the short sides.
The two separate arriving pulses was due different cables inside the crate. The cables inside
the crate where calculated thus the pulses will start arriving at the same time on the receiver,
therefore, will use the most of the received window. The cables of the short side of the crate
were in total two inches longer that the cables of the long side (one inch for the receiving
antenna, and one inch for the transmitting antenna). This caused the pulse for the long side
to be sent earlier and to be received before the intended pulse on the short side. Usually, this
pulse was attenuated, put there were cases that caused some instability on the detection,
and selecting the second strongest pulse mitigate this problem.
Figure 3.14 shows a screen shot of the received window that was used for troubleshooting
purposes. On the screen shot some of the detection features discussed are visualized: the
exponential threshold (blue), the matched filter output (yellow), and the matched filter
region that was considered for detection (red).
Figure 3.14: Received Window Sample Screen Shot
Find Pulse’s Location and Traveled Distance Sub-block

This allows a simple tracking that
makes detection more reliable.
Apply exponential
threshold Select the waveform’s Select the region around
parts that satisfy both the last observed Find the maximum Hence, the
Input the exponential location (if any, peak and the second whole
waveform threshold and the match otherwise consider the peak within 90% of window was
filter threshold whole window) the first peak considered
Apply match
filtering already.
A pulse was A pulse was not found in
Now consider the whole Find the maximum found in the the last iteration
window. That means peak and the second previous iteration
peak within 90% of No peak found
either the pulse is out of
range or moved too fast. the first peak
This is required because it
was observed that for some Either select the first Assume default values for
cases two pulses back to peak or second peak pulse’s location and
back were received. as expected for the traveled distance,
Depending on the side, the specific crate site meaning no pulse was
desired pulse was either the Return found
first or the second. location and
traveled
distance
Figure 3.15: Find Pulse’s Location Sub-block

Non-linear Filtering
Occasionally, the detector momentarily might assume detection far from the actual pulse’s
location and then return back to the correct reading. This can cause unwanted spikes
on the output of the program. For this reason the program keeps history of the last two
values. Before a value is given for an output, first, the values before and after that value are
compared, and if they are found to be within a percentage of each other, then their average
is computed. Furthermore, the percentage of the current value is compared to that average
value and if it is found to have more than a 40% diffence then it is replaced with the average
of its surounding values.
In addition, to the the procedure above, a moving average filter further smooths the output
based on a longer history record that is also kept for that reason.
Non-Linear Filtering Sub-block
Find the average Subtract the last value from
Input of this value and this average and divide the
Value the value two difference by the average
values ago
This block There is a two value delay.

smoothes the Also, a history of values is
output by rejecting >|40%| Yes Replace the last
value with the kept in order for the moving
extreme values Difference? average average function to work.
(spikes) and by No
applying a moving
average filter. Apply LabVIEW
moving average Output
filter Value
Figure 3.16: Extreme Value Rejection Sub-block
3.3.4 Other Techniques Used
Keeping a waveform database with minimum processing
A waveform database had to be maintained for the coupling estimation. Keeping a number
of waveforms requires memory and processor time. The former was not an issue for this
application: however, the latter was a problem because whenever a new waveform was
added (to the database) and one was removed at the same time, all the waveforms had to
be averaged. That caused significant delays. However, a technique was employed that used
the minimum processing time: in addition to the waveforms, their sum was maintained as
well, and when a waveform was added to the database it was also added to the total sum.
Furthermore, the waveform that was removed was subtracted from the sum. This procedure,
for the averaging, only required one addition, one subtraction and one division, compared to
N additions and one division, where N is the number of the waveforms kept in the database.
Ethernet: A GPIB replacement
GPIB (General Purpose Interface Bus) is a standard interface for communication between
instruments from different sources. The Tektronix CSA8000B DSO is also equipped with a
GPIB interface. The problem was that the GPIB’s version used has limited data transfer
capabilities. The GPIB was successfully replaced by an Ethernet connection between the
DSO and the laptop used for the DSP.
The replacement was made possible by installing the updated National Instruments VISA2
driver (replacing the Tektronix VISA driver) to the DSO. The updated driver provides the
NI-VISA server which allowed the laptop to connect using a TCP/IP network connection to
the DSO. This allowed faster data transfer and network access to the DSO folders.
Custom GetWaveform Routine Implementation
The DSO had a LabVIEW driver for issuing commands and transferring waveform data
from the DSO to the laptop running LabVIEW. However, the GetWaveform implementation
of the driver was found not to be optimized for transfer speed. It is speculated that it
issues commands that might be redundant and that opens and closes the connection to
the DSO each time a waveform is requested which causes some delays. For that reason a
new GetWaveform was written in LabVIEW that sends only the necessary commands for
obtaining the current waveform. Furthermore, the connection to the DSO was kept open
during the acquisition process; therefore no unnecessary time was lost for opening and closing
the connection. Finally, the transfer mode was in binary instead of ASCII which yielded fewer
bits to be transfered, hence a faster response.
Despite the customization of the GetWaveform the delay introduced by the DSO when clear-
ing the acquired data, required when changing tx sides, could’t be overcome. That slowed
the transfer rate for the Four-side application.
3.4 Demo Outcomes
3.4.1 Single-Side Ranging
The demo was able to provide 1 mm resolution and 16 Updates/second of the estimated
height of the the crate. At first the system was demonstrated with a stationary floor. The
crate was moved up and down to demonstrate the abilities of the system to accurately
2
The Virtual Instrument Software Architecture (VISA) is a standard for configuring, programming and
troubleshooting instrumentation systems comprising GPIB, VXI, PXI, Serial, Ethernet, and/or USB inter-
faces.
estimate the height between the crate and the floor. A sample of the system’s output is
shown at Figure 3.17 where the crate is raised from the floor to the maximum range of
the system and then is landed back to the floor. Table 3.1 shows a sample set of height
measurements. Because of the 1 mm resolution of the system obtaining good estimates of
the system’s accuracy was a hard task. In order to improve measurement accuracy and
minimize measurement errors, the measurements were taken by resting the crate in wooden
blocks of known height, instead of the crate being suspended by string wire. Nevertheless,
the sample measurements demonstrate that system is accurate in most cases within 1 mm,
and the maximum error recorded being 2 mm from 0-191 mm heights. The accuracy for
these measurements appears to be independent of height. This is due to the fact that for the
heights evaluated the accuracy is within the resolution of the system because of the good
SNR observed at these heights. The SNR for these measurements was 29.9 dB and 17.5 dB
for the 0 mm and 191 mm case respectively. For a 250 mm case the SNR was 16.8 dB. The
system could potentially range up to 2 ft (600 mm) but for this to happen several changes
are needed to the LabVIEW application in order to accommodate for the longer receive time
window (more points per waveform) and estimation of good threshold values, furthermore,
unlike the first foot (0 - 300 mm), the second foot will experience lower SNR’s that will
challenge the accuracy of the system. Finally, keep in mind that 1 foot translates to 24 feet
in an actual scale system.
Ranging Output Example − No floor movement
250
200
Height (mm)
150
100
50
0
0 2 4 6 8 10 12 14 16 18
time (s)
Figure 3.17: Sample Ranging Output: Motion all the way up and then down
After the ability of the system to provide fast and accurate range information was demon-
strated, the floor was set to oscillate with 1 Hz and the crate control system was demon-
strated, which was able to successfully land the crate on the oscillating floor with minimum
impact. The range output to the crane control system is shown on the Figure 3.18, from
which three main regions can be observed: the oscillating region, the landing region, and the
landed region.
Table 3.1: Single-Side Sample Measurements (mm)
Measured mean
Actual Trial σ mean %
#1 #2 #3 #4 #5 #6 #7 #8 #9 #10 error
0 0 0 0 0 0 0 0 0 0 0 0.00 0 0.00
19 20 20 20 20 20 19 19 19 19 18 0.70 19.4 2.11
38 38 38 40 39 39 38 38 38 37 38 0.82 38.3 0.79
57 57 57 58 58 57 57 56 57 57 57 0.57 57.1 0.18
76 77 77 77 77 76 77 76 76 75 76 0.70 76.4 0.53
95 97 96 96 96 95 96 95 95 95 96 0.67 95.7 0.74
115 114 115 115 115 114 115 114 115 114 114 0.53 114.5 -0.43
134 133 135 134 134 134 135 134 134 133 134 0.67 134 0.00
153 153 153 153 153 153 154 153 153 153 154 0.42 153.2 0.13
172 172 173 173 173 173 174 174 173 172 173 0.67 173 0.58
191 191 192 191 192 192 193 192 191 191 192 0.67 191.7 0.37
Ranging Estimates During Automated Landing

250
200
150
Height (mm)
Oscillating Region
100
Landing Region
50 Landed Region
0
0 2 4 6 8 10 12 14 16 18 20
Time (s)
Figure 3.18: Automatic Landing
3.4.2 Four-Side Ranging
The full implementation of the system was demonstrated in a slow motion manner because
of the low update rate of the system (3 updates/second). The orientation capabilities of the
algorithm were demonstrated by tilting the floor and by visualizing the tilt on the laptop’s
screen. Furthermore, because two sides were estimated at a time this caused the effective
update rate to be 1.5 updates/second, which further limited the speed of this part of the
demo. Also for this case a few sample measurements were taken which are shown in Table
3.2. These measurements were taken with the crate suspended using string wire, which
makes the measurements subject to measurement errors. The slopes were calculated using
Equations 2.21 and 2.22. Nevertheless, the system was found to have the same accuracy as
the single-side case.
Table 3.2: Four-Side Sample Measurements
Corner Height (mm) Slope (o )

C1 C2 C3 C4 Length Side Width Side
Actual 0 0 0 0 0 0
Meas. 0 0 0 0 0 0
Actual 79 78 74 75 0 -2
Meas. 78 78 74 75 0 -2
Actual 48 46 36 39 0 -5
Meas. 49 47 37 38 0 -8
Actual 25 53 55 28 6 -1
Meas. 24 52 54 27 6 -1
Actual 35 59 52 26 6 -4
Meas. 36 61 53 27 6 -5
Chapter 4
UWB Ranging Conclusions
The first part of the thesis addresses the problem of unloading cargo crates to ships, either
from a port’s crane or a crane on another ship. The problem in unloading is due to the
fluctuations of the ship’s deck due to sea waves. For this problem a scheme was proposed
that estimates the height and the orientation of the ship in reference to the cargo crate
being unloaded. The scheme uses UWB antennas at each corner of the crate, from each side
an UWB pulse was sent to the deck and received back; from the pulse’s traveled path the
distance from the crate was estimated. In this thesis an algorithm was developed that can
estimate both the height and the orientation of the deck just using the traveled distances of
the four sides.
The proposed algorithm was evaluated by means of simulation and was found to accurately
estimate the deck’s height and orientation given valid data. Furthermore, it was found that
the estimation variance at the output of the algorithm is less than the input variance by a
factor of 1.5 to 2.8, depending on the ratio of the crate side’s length and the height to be
estimated.
After the scheme was verified analytically and via simulation, a demo was planned to demon-
strate its accuracy. For the demo a 1/24 scale crate was used with small UWB antennas. The
demo was first performed using a single side for ranging which provided 16 updates/second
to the cranes controller to test the system with the crate landing algorithm developed by
Nader A. Nayfeh. The crate landing algorithm was able to land the crate, using the input
from the UWB system developed with minimum impact at an oscillating surface (1 Hz.)
In the second part of the demo, the orientation capabilities of the algorithm were demon-
strated at a slower rate. In the demo the algorithm only received the update information
of two sides at a time, also limited the update rate of the system by an additional factor of
two. It is suggested that the final application will have four receiving channels.
Because the antennas were very close to each other there was a direct received signal (cou-
pling) which was successfully mitigated. This is something that will be a lesser issue in a
46
CHAPTER 4. UWB RANGING CONCLUSIONS 47
real-world system in which the antennas will be separated by several meters and the coupling
will be attenuated.
For future development an acquisition technique can be employed which will allow more
robust acquisition and more sophisticated pulse tracking. Furthermore, pulse shapes might
be investigated as some pulses might be more suitable than others for this application because
of their different spectrum characteristics. Moreover, the application’s environment can be
further investigated for making potential suggestions and additions to the system. Finally,
the ranging information of the crate can be potentially used for other purposes such as
detecting nearby objects to the crate and predict possible collisions.
Part II
Ultra Wideband Link Budget Design
48
Chapter 5
Ultra Wideband Link Budget Design
5.1 Introduction
In a communications link the transmitter sends a signal and a receiver at some distance
attempts to detect that signal. Since the early days of wireless communication, engineers
have devised different signals with different properties to send the information to the receiver.
Some signals are better than others; simpler or more complex to implement. However,
regardless of the signal used, a communications engineer has to be able to estimate the
performance of the link based on many parameters that have to do with the link. The
performance of the link, regardless of the type of signal sent is directly related to the ratio of
received signal power to noise power. The procedure of estimating the signal-to-noise ratio
of the link is often referred to as calculating a link budget.
A link budget is a set of calculations aimed at estimating the power of the signal received
by the receiver and the noise power seen at the receiver. The received power is estimated
by knowing the transmit power, the separation distance, propagation conditions, frequency
of operation, and the antenna parameters. From the received power other parameters of
the communication link such as the ratio of energy per bit to noise power spectral density
(Eb /N0 ) can be estimated, which directly determines the achieved bit error rate of the link.
Furthermore, not only do path loss and noise affect the link, other parameters such channel
fading and system losses must be considered for an even more accurate prediction.
For most signals that are narrow in bandwidth, a procedure for calculating the link budget
exists which accurately predicts the receive signal’s power. That link budget is based on the
Friis transmission formula which estimates the received power by considering the transmit
power, the antenna gains, the frequency of operation, and the distance between the antennas.
The FCC’s decision in 2002, to allow UWB operation in a broad range of frequencies (3.1-10.6
GHz) stimulated research in UWB based communication systems. One way to implement
49
CHAPTER 5. ULTRA WIDEBAND LINK BUDGET DESIGN 50
a UWB system is to use very short duration pulses that are wide in bandwidth, a tech-
nique commonly termed Impulse-UWB (I-UWB). Analysis of I-UWB systems requires the
examination of the impulse response of the system as opposed to the steady state response.
[5].
An I-UWB system is very different from narrow band systems because of its bandwidth
that often occupies a few GHz. Attempting to use a link budget that is based on the Friis
transmission formula which assumes a single frequency of operation and is based on steady-
state sinusoidal signals is not the most accurate procedure. In other words, the assumption
that the link behaves the same at all frequencies of operation no longer holds because of the
complex spectrum characteristics of both the pulse and the antennas used. Furthermore,
besides the estimation of the received power, the whole approach of the link budget much
be reconsidered because of the channel and the receiver structures used.
In this chapter, the traditional narrowband link budget will be introduced. Furthermore, a
proposed I-UWB link budget that first appeared in [5] will also be introduced and explained.
The purpose of the remainder of the chapter is to demonstrate, mainly by measurements
done with four different UWB antennas, that the proposed I-UWB link budget is much more
accurate and suited for I-UWB systems than the link budget based on the Friis transmission
formula. Furthermore, the proposed approach is primarily defined in the time domain as
opposed to the frequency domain which simplifies the process and the data required for the
link budget calculations.
5.2 The Traditional Narrowband Link Budget

As already mentioned, a link budget for narrowband systems already exists that is based
on the Friis transmission formula which is going to be derived and explained. We will now
briefly derive and explain this link budget approach.
5.2.1 The Friis Transmission Formula
It is assumed that the transmit antenna of a communication link radiates isotropically in

free space a power PT Watts, such that the power flux density at distance r, PD is given by
[6]
PT
PD (r) = (W/m2 ) (5.1)
4πr 2
The term 4πr 2 is the surface area of a sphere of radius r. This isotropic antenna does not
really exist. Practical antennas do not radiate equally in all directions and thus have higher
gain in a particular direction, which is assumed to be used for the link. That gain is denoted
by GT and Equation (5.1) becomes
P T GT
PD (r) = (W/m2 ) (5.2)
4πr 2
where PD is the power received in the the direction of the peak gain GT of the transmit
antenna.
The term PT GT is called the Effective Isotropic Radiated Power (EIRP). The received power
PR can be found by multiplying Equation (5.2) with the effective aperture of the receive
antenna Ae . The effective aperture is a measure of the antenna’s ability to collect electro-
magnetic (EM) energy which is given by
GR λ2
Ae = (m2 ) (5.3)
4π
where GR is the gain of the receiving antenna, and λ is the wavelength of the transmit
frequency in meters. The multiplication of Equations (5.2) and (5.3) results in:
P T GT GR λ2
PR (r) = (W ) (5.4)
(4πr)2
where PR (r) is the received power at distance r 1 and the formula above is the standard Friis
transmission formula.
2
The factor Ls = 4πr λ
is referred to as the free-space path loss. The λ in the path loss
suggests that path loss is frequency dependent. However, it must be noted that this is an
antenna effect as shown in Equation (5.3) and not a propagation effect. Furthermore, it is
shown [5], that by using a constant aperture transmit and a constant gain receive antenna,
the effects of this term are canceled. These observations will be useful for the UWB link
budget development.
5.2.2 Link Budget
The received power of a narrowband signal can be estimated using the Friis transmission
formula. However, the received power is only a part of a link budget. Another important
term is the thermal noise at the receiver which is a function of the operating temperature.
For a signal that only experiences noise due to the receiver, its link is modeled as an Additive
White Gaussian Noise (AWGN) channel. The noise power spectral density N0 is given by
N0 = kT0 (W/Hz) (5.5)

1
Note that this assumes that the antennas are oriented such that their maximum gains are directed along
the link.
where k is Boltzmann’s constant and T0 is the nominal system noise temperature in Kelvin.
Since most receivers amplify this noise based on the quality of its components, we typically
also include a term called the noise figure NF which accounts for the increase in thermal
noise:
N0 = NF kT0 = kTs (5.6)
where Ts = NF T0 is the effective system noise temp.

Eb
The SNR per bit N 0
is given by dividing the received power with the noise spectral density
and the number of bits sent per second:

Eb PR
= (5.7)
N0 N0 Rb
where Rb is the bit rate in bits/sec

In addition, a fading margin MF or safety margin can be added to account for deviation
Eb
in the received power due to fading. This will ensure that the received N 0
will be above a
threshold some % of the time. The desired SNR per bit will now be equal to:

Eb PR
= (5.8)
N0 N0 RMF
The link budget also can be also expressed in dB terms:

Eb
= (PR )dB − (N0 )dB − (R)dB − (MF )dB (5.9)
N0
(PR )dB = (PT )dB + (GT )dB + (GR )dB − (Ls )dB (5.10)
If desired, the above can be rearranged to find the maximum achievable bit rate.
Table 5.1 provides a sample conventional free space link budget calculation.
Table 5.1: Conventional Free Space Link Budget
Term Value (example) Comment

PT -55dBW Transmit power in dB relative to a W
GT 3dBi Transmit antenna gain relative to isotropic radiator at
operating frequency.
Ls 86dB (4πr/λ)2 - Free space path loss given constant gain an-
tennas at a distance r (assume fc =4.6GHz and r=100m)
GR 3dBi Receive antenna gain relative to isotropic radiator at
operating frequency.
PR -140dBW (PT + GT − Ls + GR )
k -228.6dbW/Hz/K Boltzman’s constant
NF 7dB Noise Figure
T0 24.6dBK System Noise Temperature (290K)
N0 -197dBW/Hz Noise PSD= NF kT0
Rb 50dB Bit Rate (100kbps)
Eb -190dB PR − Rb
Eb /N0 7dB Eb − N0
MF 0 Fading Margin
Eb /N0 7dB Eb /N0 achieved
5.3 The Proposed I-UWB Link Budget

This section presents the proposed link budget for UWB mostly as it appeared in [5].
It is natural to want to apply the previous link budget to UWB systems. However, there are
several problems with that approach. The development of the previous link budget assumes
a frequency of operation and everything is calculated based on that frequency. It is assumed
that the estimated power approximately holds for all the nearby frequencies around the
selected frequency which is usually referred to as the carrier or center frequency of the link.
Using the center or carrier frequency for evaluating the link budget, as is usually the case
for narrowband systems, is a challenge in I-UWB systems since the transmit pulse usually
has a bandwidth of a few to several GHz. Some might suggest to use the center frequency of
the bandwidth spectrum of the transmit pulse. Still, using a formula that takes into account
only a single frequency out of a bandwidth of few GHz is not optimal. Furthermore, the
formula can not take into account pulse distortion due the antennas’ and channel’s frequency
response. In addition, usually a link budget is calculated in terms of average power. However,
in an I-UWB system because of the the low duty cycles one might want to define the link
budget in terms of energy or peak power (The Friis formula does not really apply to energy).
Going back to the Friis transmission formula and arranging it into parts in order to examine
potential modifications we have:
P T GT GR λ2
PR = (5.11)
(4πr)2
GT GR λ2 1
= PT
4π 4πr 2
The formula has three components: the transmit power PT , the spherical spreading factor
GR λ2
4πr 2 , and a term due to the antennas GT 4π . The first two terms can’t be changed. In
contrast, the antenna effects term, which consists of the antenna gains and the effective
aperture of the receive antenna, can be replaced by an “antenna-pulse coupling gain,” or
2
GAP . For comparison purposes in the validation section lets call the old term GF riis = GT G
4π
Rλ
with units m2 .
The new GAP term captures the gain in either energy or peak power in the received pulse
with respect to the generated pulse. Whether energy or power is examined depends on the
modulation scheme used, as will be discussed shortly. In some cases the two measures may
be equal, but not in general. It should be clear that this gain is very different from the
traditional antenna gain. It is different because it defines an antenna pair as opposed to a
single antenna, and because it is pulse specific. Just as the traditional “gain” of an antenna
is frequency specific, GAP is pulse specific. Additionally, it captures the receive antenna
aperture which is normally lumped into path loss. As a result, the “gain” actually has a
negative dB value in general and has units of meters squared (m2 ).
A second modification to the traditional link budget is that path loss now accounts only for
spreading loss, not effective aperture at the receive antenna. Thus, we can define path loss
as
Lp = 4πr 2 (5.12)
where r is the distance between the transmitter and the receiver. This simply represents the
power distributed over the area of a sphere of radius r; thus, it has units of square meters
(m2 ).
Another difference in the I-UWB link budget is that it should be in terms of either pulse
energy or peak pulse power. In traditional link budgets, we deal with average power. How-
ever, average power may not be useful in UWB because low duty cycle pulses are possible.
Also, we are interested in the impact of the antenna as well. Since peak power could be
affected by the antennas, we allow for GAP to be defined either in terms of peak pulse power
or pulse energy. The former would be more applicable to threshold detector receivers, while
the latter would be more applicable to correlator receivers.
The term GAP can either be determined directly from a time domain line-of-sight measure-
ment or derived from an antenna S21 measurement. Specifically, for a pulse energy-based
link budget, the gain should be determined as

εrx
GAP = 4πr 2
p
εtx
p∞
b2r (t)dt
= 4πr 2 −∞
∞ (5.13)
−∞
a2t (t)dt
where εp represents the pulse energy, at (t) is the generated pulse, and br (t) is the received
pulse measured at a distance r meters in free-space or calculated as
1 μ d r

br (t) = at (t) ∗ htx t − ∗ hrx (t) (5.14)
4πr 50 dt c
where the combined response of htx (t) and hrx (t) (the impulse responses of the transmit and
receive antenna respectively) can be determined for the antenna pair of interest using an S21
measurement. Additionally, note that we have neglected the vector nature of the general
problem. We have assumed that Equation (5.14) is calculated with both antennas aligned
for maximum response. This is analogous to narrowband antenna gain, which is typically
defined as the maximum gain versus azimuth and elevation angles. Additionally, traditional
link budget designs assume proper polarization alignment or include a polarization loss
factor.
Thus, GAP represents the energy gain of the pulse due to the antennas. Note that this term
is not waveform independent. That is, it is not merely a factor of the antenna, unlike
the traditional antenna gain. It is waveform specific. As such, care must be taken in its
use. Alternatively, the received pulse can be determined in the frequency domain
r
e−jω c μ
Br (jω) = (jω) · Htx (jω) · Hrx (jω) · At (jω) · (5.15)
4πr 50
where At (jω) is the Fourier Transform of the generated pulse and the combined term
Ht (jω)Hr (jω) is the transfer function of the transmit and receive antennas which can be
determined from network analyzer measurements.
Assuming that the S21 measurement is the combined frequency response of the link and the
antennas at a reference distance r0 , then the frequency response of the link is equal to:
r0
Br (jω) e−jω c μ
S21 (jω) = = (jω) · Htx (jω) · Hrx (jω) · (5.16)
At (jω) 4πr0 50
Rearranging Equation (5.16), the combined antenna frequency response Htx (jω) · Hrx (jω)
can be determined using the S21 parameter of the two antennas using:
4πr0 50 r0
Htx (jω) · Hrx (jω) = · S21 (jω) · e+jω c (5.17)
(jω) μ
Substituting Equation (5.17) into Equation (5.15) we arrive at

r0 jω
Br (jw) = S21 (jw) · At (jw) · e− c (r−r0 ) (5.18)
r
In the time domain this becomes

r0 r − r0
br (t) = s21 t − ∗ at (t) (5.19)
r c
The received pulse br (t) is the convolution of the time shifted s21 (t) parameter of the link,
adjusted for the spherical path loss according to the desired distance r with the transmit
pulse at (t).
The received pulse energy in dBJ (that is, dB relative to one J) is then calculated as
ER (dBJ) = ET (dBJ) + GAP (dBm2 ) − Lp (dBm2 ) (5.20)
The noise power spectral density N0 is calculated as before, and ER /N0 dB is simply
ER (dB) − N0 (dB). Of course, this is the energy per pulse, not necessarily per bit. In
many systems multiple pulses may be transmitted per bit. Thus, this value must be multi-
plied by the number of pulses per bit Ns (added in dB) to obtain the energy per bit. The
result then represents the available Eb /N0 or the SNR per bit of a matched filter output.
Due to pulse mismatch, which is much more likely in UWB due to both the channel and
limitations in replicating the received pulse, we must include a factor ρ (negative in dB or less
than unity in linear), which represents the fraction of the energy captured by the receiver.
This factor can either be due to per path distortion or overall temporal dispersion of the
channel. For example, consider a Rake receiver that uses only two to three correlators.
As discussed in [5] it is known that in many environments, such a receiver only captures
∼ 20% of the total energy. Thus, a factor of ρ = 0.2 (-7dB) should be included in this case,
provided the path loss value implies total energy capture. Various receiver structures will
have different values of ρ. The final result is the Eb /N0 available to the detector. Table 5.2
presents a sample set of calculations based on the trasnmitted pulse and the “TimeDomain”
antenna from the validation section.
5.3.1 Receiver Structure Dependence
This development has been specifically in terms of energy because we are typically interested
in the received energy per bit.Traditional link budgets can obtain Eb from the average re-
ceived power as Eb = Pr Tb , where Tb is the bit period. However, this is not necessarily true
for UWB; thus, we have set up the budget in terms of the energy per pulse. If the receiver
is based on threshold detection or is otherwise concerned only with the peak power of the
pulse, the link budget should be worked in terms of peak power. In that case, we should
Table 5.2: Proposed UWB Free Space Link Budget — Correlator Detector

ET -108dBJ Transmit energy in dB relative to a J
GAP −38dBm2 Energy gain of the link for the specific antenna pair and
pulse used.
Lp 31dBm2 4πr 2 - Free space spreading loss at a distance r=10m
ER -177dBJ (ET + GAP − Lp )
k -228.6dbJ/K Boltzman’s constant
N0 -197dBW/Hz Noise PSD= NF kT0
N0 -197dBJ Noise PSD= NF kT0
Ns 0dB Pulses per bit (1 pulse/bit)
Eb -177dBJ ER + Ns
Eb /N0 20dB Eb − N0
MF 0 Fading Margin
ρ -7dB Energy capture loss (20% Capture)
Eb /N0 13dB Eb /N0 achieved
define the pulse antenna coupling gain GAP as

Pprx
GAP = 4πr 2
Pptx rm
max {b2r (t)}
= 4πr 2 (5.21)
max {a2t (t)}
where again br (t) is the received signal at a distance of r meters, Pprx and Pptx are the peak
power of the received pulse measured at r meters and the generated pulse respectively. Note
that the notation []|rm signifies that the gain is measured at r meters. This should coincide
with the reference distance used for the path loss calculations.
An example set of link budget calculations for a peak detector receiver is presented in Ta-
ble 5.3.
As a final note, we should mention again that the preceding analysis is more applicable to
pulse-based (I-UWB) rather than Multi Carrier UWB (MC-UWB). With MC-UWB, we can
resort to traditional narrowband approaches for each carrier.
Table 5.3: Proposed UWB Free Space Link Budget — Peak Detector

PT -10dBW Transmit peak power in dBW
GAP −44dBm2 Peak power gain of the link for the specific antenna pair
and pulse used.
Lp 31dBm2 4πr 2 - Free space spreading loss at a distance r=10m
PR -85dBW (PT + GAP − Lp )
k -228.6dbW/Hz/K Boltzman’s constant
N0 -204dBW/Hz Noise PSD= kT0
N -104dBW Noise Power (9.2GHz BW)
SNR 19dB PR − N
MF 0 Fading Margin
ρ -10dB Capture loss (10%)
SNR 9dB SNR achieved
5.4 Link Budgets for Systems Other than Free-Space

The analysis in the previous section considered only free-space propagation. For non-free-
space environments, a traditional approach is to model the average received power using a
log-distance power decay law. We can represent the average received signal power at distance
r as r
n
0
pr (r) = p0 (r) (5.22)
r
where p0 (r) is the pulse power received at a reference distance assuming free-space. This is
the traditional log-distance path loss model [7]. Note that the average is a spatial average.
If we define our reference distance as 1m (typical for indoors), we would redefine total path
loss as
Lp = 10log(4π) + 10nlog(r) (5.23)
where n is the path loss exponent defined as the path loss relative to a 1 m free space
measurement. Additionally, we must include a fade margin in non-free space environments
because of multipath combining and other parameters which cause the received power to
deviate around the predicted power.
5.5 Validation
For validation purposes measurements were taken both in the frequency domain and the time
domain. The frequency domain measurements were taken to estimate the S21 parameters
and the nominal gains of the four antenna pairs used. From the S21 parameters the GAP
values were calculated as well.
From the time domain measurements, GAP was also calculated by using reference LOS
measurements. Furthermore, LOS and NLOS propagation measurements were taken for link
budget verification purposes.
Finally, the GAP was calculated for various pulses in order to demonstrate that it’s pulse
dependent and precisely captures the spectrum characteristics of the pulses. On the other
hand, the Friis transmission formula is not able to fully capture all the characteristics of the
pulses.
5.5.1 Vector Network Analyzer (VNA) Measurements
One of the methods of calculating GAP is by using the S21 parameter of the antenna pair
of interest and the UWB pulse to be used. However, the exact data (digital form) for the
antennas used were not available. Therefore, the author had to estimate them using a Vector
Network Analyzer (VNA).
Method
For the measurements, an HP8510C Vector Network Analyzer (VNA) was used. Ideally, the
measurements should have been placed inside an anechoic chamber that provides a reflection
free environment. However, because of the lack of easy access to such an environment the
measurements were taken in the MPRG Radio Lab at 471 Durham Hall. The measurements
were taken at a reference distance of 1 m (Figure 5.1). The antennas were facing each other
with their boresight side. Because of the non-ideal environment it is expected that a strong
reflection from the floor and probably one strong reflection from the ceiling would be present
in the measurements. The antennas were 1.2 m above the floor and about the same from
the ceiling. As it will be explained later, multipath reflections from the surrounding objects
were removed during post-processing.
The frequency range of the measurements was set from 50 MHz to 18.05GHz. 50 MHz was
chosen since it is the lowest frequency that the VNA could measure, and 18.05GHz was
chosen because the connectors were not rated to be used at higher frequencies. To increase
the frequency resolution, two measurements of 9 GHz per antenna were taken, one between
50 MHz to 9.05 GHz and one between 9.05 GHz to 18.05 GHz.
Before taking the measurements the VNA was calibrated to the cables used in order to
remove their effects. An averaging factor of 10 was used during the measurements.
Antenna
Separation 1m
Antenna
Height 1.2m
Port 1 Port 2
HP8510C
VNA
Figure 5.1: VNA Setup
From the measurements there were a total of 1601 points in the selected frequency range (50
MHz to 18.05 GHz). This resulted in a frequency resolution of 11.25 MHz. The resulting
frequency resolution could not allow exact zeropading to DC. For that reason the measure-
ments were interpolated by a factor of 9 using Matlab, then decimated by a factor of 8 which
made the frequency resolution 10 MHz. The measurements were then zero padded to DC
and converted to the time domain using the IFFT.
In the time domain the main pulse was identified, and the rest of the signal was zero padded
and converted back to the frequency domain using the FFT to get the final S21 parameters.
This allowed the removal of multipath reflections. The results were compared and found to
be in agreement with [5] that has S21 measurements of the same and similar antennas from
an anechoic chamber.
Results — Antenna Characterization
The following antennas were characterized and used for the measurements in this section:
1. Ridged TEM Horn — Commercially made.
2. Bicone Antenna — Commerciallly made, repaired/modified2 at Virginia Tech.
3. Vivaldi Antenna — Manufactured2 at Virginia Tech as part of a past project investi-

gating the Vivaldi antenna design for UWB systems.
2
The author was just a user of these antennas and had no involvement in any development or modifications
of the antennas.
4. “TimeDomain” — This antenna was made by Time Domain Corporation and was
supplied with their PulsON 200 UWB Development system.
Using the method described, the characterization results of each antenna Tx/Rx pair at a
reference distance of 1 m are shown in Figures 5.2 to 5.5.
Figure 5.2 shows a photo, impulse response, and frequency response of the the Ridged TEM
Horn antenna. From the frequency response of this antenna it can be observed that it is a
wideband antenna which is officially rated, by the manufacturer, to operate in the range of
0.7 to 18 GHz.
In Figure 5.3 the same information is shown for the Bicone Antenna which has a wide but
narrower frequency response compared to the Ridged TEM Horn antenna. In addition, the
Bicone antenna has an attenuated frequency response because it has a uniform azimuth
radiation pattern, compared to the Ridged TEM Horn which is a directional antenna.
Furthermore, the Vivaldi antenna shown in Figure 5.4 is also a directional antenna with a
similar bandwidth range as the Ridged TEM Horn. Finally, in Figure 5.5 the “TimeDomain”
antenna is shown, which has an almost uniform azimuthal radiation pattern. It is slightly
directional to the front and the back of the antenna, but for practical purposes it can be
assumed to have an omni-directional pattern. In addition, this antenna has a significantly
narrower bandwidth compared to all the antennas mentioned.
Ridged TEM Horn Antennas Impulse Response s21(t) Ridged TEM Horn Antennas Frequency Response |s21(jω)|
0.025
−10
0.02
0.015 −15
0.01
Amplitude (s−1
Amplitude(dB)
−20
0.005
−25
−0.005
−0.01
−30
−0.015
1 1.2 1.4 1.6 1.8 2 0 2 4 6 8 10 12 14 16 18

time (s) −8
x 10 Frequency (Hz) 9
x 10
(a) Photo (b) Impulse Response (c) Frequency Response
Figure 5.2: Ridged TEM Horn Antenna

−3 Bicone Antennas Impulse Response s (t) Bicone Antennas Frequency Response |s21(jω)|
x 10 21
−15
5
−20
4
3 −25
Amplitude (s−1
Amplitude(dB)
2
−30
1
−35
−40
1 1.2 1.4 1.6 1.8 2 0 2 4 6 8 10 12 14 16 18

time (s) x 10
−8 Frequency (Hz) 9
x 10
Figure 5.3: Bicone Antenna
Vivaldi Antennas Impulse Response s21(t) Vivaldi Antennas Frequency Response |s21(jω)|
0.01
−15
0.005
−20
Amplitude (s−1
Amplitude(dB)
0 −25
−30
−0.005
−35
−0.01
−40
1 1.2 1.4 1.6 1.8 2 0 2 4 6 8 10 12 14 16 18

time (s) x 10
−8 Frequency (Hz) x 10
9
Figure 5.4: Vivaldi Antenna

−3 TimeDomain Antennas Impulse Response s21(t) TimeDomain Antennas Frequency Response |s21(jω)|
x 10
−20
1
−25
0.5
0 −30
Amplitude (s−1
Amplitude(dB)
−0.5 −35
−1
−40
−1.5
−45
−2
1 1.2 1.4 1.6 1.8 2 0 2 4 6 8 10 12 14 16 18

time (s) −8
x 10 Frequency (Hz) 9
x 10
Figure 5.5: TimeDomain Antenna

Results — Calculated Pulses
Using the S21 parameters and the generated pulse (Figure 5.6) the received pulse at 1m was
calculated using Equation (5.19) for each antenna. However, the effects3 of the cables are
not included in the S21 parameters, therefore, for consistency with the measurements done
in the time domain in which the cables affect the measurements, the effects of the cables
were added after calculating the received pulse using Equation (5.19). Furthermore, GAP is
also calculated before and after adding the cable effects.
Generated Pulse Generated Pulse Spectrum

12
0
10
−2
8
−4
−6 6
Amplitude(V)
Energy (dB)
−8 4
−10
2
−12
0
−14
−16 −2
−18 −4
−20
−6
0 1 2 3 4 5 6 7 0 5 10 15 20
Time (s) x 10
−10 Frequency (GHz)
(a) Time Domain (b) Energy Spectrum
Figure 5.6: The Generated Pulse
Figure 5.7 shows the calculated pulses using the procedure described above (with the cable
effects added). It may be noticed that the pulses are very similar to the impulse response
of the antennas. This is because a very wide bandwidth pulse was used (9.2 GHz 10 dB
bandwidth), which approximates an impulse.
3
The transmit cable effects are a part of the generated pulse, therefore, only the effect of the receive end
cable had to be included. The effect of the cable per foot was a loss of 0.1 dB at DC linearly increasing to 0.45
dB at 18 GHz. The calculated pulse was converted to the frequency domain in which the cable’s frequency
dependent attenuation effect was added and then the signal was converted back to the time domain.
Ridged TEM Horns Calculated 1m Received Pulse Bicone Antennas Calculated 1m Received Pulse
0.02
0.3
0
0.2
−0.02
0.1
Amplitude (V)
Amplitude (V)
−0.04
0
−0.06
−0.1
−0.08
−0.2
−0.1
−0.3
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
time (s) −9
x 10 time (s) −9
x 10
(a) Ridged TEM Antenna (b) Bicone Antenna

Vivaldi Antennas Calculated 1m Received Pulse TimeDomain Antennas Calculated 1m Received Pulse
0.3
0.25
0.03
0.2
0.15 0.02
0.1
Amplitude (V)
Amplitude (V)
0.01
0.05
0
0
−0.05
−0.1 −0.01
−0.15
−0.02
−0.2
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
time (s) x 10
−9 time (s) x 10
−9
(c) Vivaldi Antenna (d) TimeDomain Antenna
Figure 5.7: Calcuated 1m Received Pulses
Results — GAP Calculations
GAP was calculated using the calculated pulses (Figure 5.7) from the S21 parameters, and
from time domain measurements at 1m. GAP was found using the two different methods
which were found to be very close, and the largest deviation is 0.5dB.
It must be noted that GAP is consistent with the antenna characteristics. The Ridge TEM
Horn and Vivaldi antennas have the highest GAP because these are directional antennas.
Furthermore, the Bicone has a lower GAP because it is an omni directional antenna. Finally,
the TimeDomain antenna has the lowest GAP both because its radiation pattern is omni-like
and because its bandwidth is significantly narrower than the other antennas. The narrow
antenna bandwidth negatively affects GAP because the pulse had a wide bandwidth of which
a small fraction was received, hence the ratio of the transmitted pulse energy over the received
Table 5.4: GAP for Energy (dBm2 )
Method
Antenna
S21 S21 (with cable effects) Time domain
Ridged TEM Horn -13.02 -15.42 -15.50
Bicone -26.79 -28.92 -28.57
Vivaldi -17.04 -19.42 -19.39
TimeDomain -35.38 -38.22 -37.99
Table 5.5: GAP for Peak Power (dBm2 )
Method
Antenna
S21 S21 (with cable effects) Time domain
Ridged TEM Horn -20.30 -23.32 -22.82
Bicone -31.58 -34.41 -34.66
Vivaldi -22.72 -25.51 -25.57
TimeDomain -40.49 -43.69 -43.82
energy was smaller.

Results — Antenna Gain Estimation Measurements
It is desirable to compare the narrowband approach with the proposed one. In order to be
able to calculate the traditional link budget using the Friis transmission formula the gain
values of our antennas had to be estimated.
In order to estimate a gain value for our antennas a standard gain horn was used which is
rated to have an 18dBi gain in the frequency range of 3.3-4.9 GHz. The frequency response
of a pair of these antennas is plotted in the same figure with the frequency response of one
of these antennas and one of the other unknown gain antennas, and with the frequency
response of a pair of the unknown gain antennas. The difference in dB between each plot
should match the gain difference between the unknown gain and the gain of the horn antenna.
These measurements were also taken at a separation distance of 1 m2 .
Figure 5.8: The Standard Gain Horn
Looking at the results in Table 5.6, as expected the Ridged TEM Horn and the Vilaldi have
the higher gain values. Furthermore, it may be noticed that the Bicone performs better
than a dipole4 (2.1 dBi) at most frequencies examined. Finally, the TimeDomain antenna
for some frequencies (4-4.6 GHz) performs better than a dipole, but for other frequencies its
performance is worse than a dipole.
4
The gain of a dipole is used solely for comparison purposes, it is acknowledged that a dipole can only
optimally work to the frequency that it was designed
Table 5.6: Estimated Antenna Gains
Est. Gain (dBi) at

Antenna Frequency (GHz)
1 2 3.3 4 4.6 5 9
Ridged TEM Horn 13 11 11.4 10.5 10.5 9.5 9.25
Bicone 6.75 4 5 3 3 4 2.5
Vivaldi 9.25 11.25 8.5 8 8 7.25 4.5
TimeDomain -12 -8.5 2 3 3 1 -2.5
STG Horn & RTEM Horn Frequency Response |s21(jω)| STG Horn & Bicone Frequency Response |s21(jω)|
−10 −10
−20 −20
−30 −30
−40 −40
−50 −50
Power(dB)
Power(dB)
−60 −60
Ridged TEM Horn Antennas
Standard Gain Horns
−70 STG Horn & RTEM Horn −70
−80 −80
−90 −90
Bicone Antennas
−100 −100 Standard Gain Horns
STG Horn & Bicone
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
Frequency (Hz) x 10
9 Frequency (Hz) x 10
9
(a) Ridged TEM Antenna (b) Bicone Antenna

STG Horn & Vivaldi Frequency Response |s21(jω)| STG Horn & Time Domain Frequency Response |s21(jω)|
−10 −10
−20 −20
−30 −30
−40
−40
−50
−50
Power(dB)
Power(dB)
−60
−60
−70
−70
−80
Vivaldi Antennas −80 TimeDomain Antennas
Standard Gain Horns Standard Gain Horns
−90 STG Horn & Vivaldi STG Horn & Time Domain
−90
−100
−100
−110
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
Frequency (Hz) x 10
9 Frequency (Hz) x 10
9
(c) Vivaldi Antenna (d) TimeDomain Antenna
Figure 5.9: Antenna Gain Measurements

Results — Link Budget Comparison
From the estimated antenna gains, the traditional Friis transmission formula link budget is
to be compared to the proposed link budget. The Friis transmission formula is presented
again:
P T GT GR λ2
PR = (5.24)
(4πr)2
GT GR λ2 1
= PT
4π 4πr 2
1
= PT GF riis
Lp
From the above it can be shown that the Friis formula GAP equivalent term is:
GT GR λ 2
GF riis = (m2 ) (5.25)
4π
In terms of received power or energy, the only difference in the two link budgets is the terms
GAP and GF riis . Therefore, any difference between the two will be reflected in those two
terms.
One might argue that a gain defined in terms of energy (GAP ) can not be compared with
a gain defined in terms of average power (GF riis ). Assuming that a pulse p(t) is repeated
every T seconds, that the duration of the pulse is less than T seconds, and that only a single
instance of the pulse can always be found in the time interval (0, T ), the energy of that single
instance of the pulse can be defined as:
∞ T
2
Ep = p (t)dt = p2 (t)dt (5.26)
−∞ 0
The average power when the same pulse is repeated every T seconds will be:

1 T 2 1
Pp = p (t)dt = Ep (5.27)
T 0 T
The power ratio R12 of two pulses p1 and p2 that repeat themselves every T seconds will be:
1
Pp1 E
T p1 Ep1
R12 = = 1 = (5.28)
Pp2 E
T p2
Ep2
The above result indicates that the ratio of the energy of a single instance of two pulses is
the same as the ratio of the average power of the two pulses when they are repeated every
T seconds. Therefore, GAP and GF riis should be comparable either when the UWB pulse
repeats every T seconds or when just a single pulse is being sent.
Using the gains in Table 5.6, GF riis was calculated and presented to Table 5.7, which also
presents the calculated GAP from the S21 parameters with no cable effects added.
Comparing GF riis and GAP (without the receive cable effects) it can be observed at some
frequencies, for example 4.6 GHz, which is the center frequency of the 10 dB bandwidth
of the generated pulse, the two gains are close to each other. This should be treated as
a coincidence, since GAP is pulse dependent and will be different for different pulses with
different spectrum characteristics. On the other hand, GF riis will remain the same with any
pulse used, if it is evaluated at the same frequency.
Table 5.7: Calculated GF riis
GF riis (dBm2 ) at
GAP (S21 )
Antenna Frequency (GHz)
2 3.3 4 4.6 5 9 (dBm2 )
Ridged TEM Horn -5.48 -9.03 -12.5 -13.71 -16.44 -22.04 -13.02
Bicone -19.48 -21.83 -27.5 -28.71 -27.44 -35.54 -26.79
Vivaldi -4.98 -14.83 -17.5 -18.71 -20.94 -31.54 -17.04
TimeDomain -44.48 -27.83 -27.5 -28.71 -33.45 -45.54 -35.38
Results — GAP for different pulses
To further demonstrate that GAP is pulse dependent, GAP was calculated using different
pulses. The chosen pulses are a 100 ps and a 500 ps Gaussian pulse and their first derivatives
(Gaussian Monocycle.) The pulse width, in seconds, for these pulses is defined as τp =
2πσ, where σ is the standard deviation of the pulse, also in seconds. The pulses and their
spectrum characteristics are shown in Figure 5.10. The resulting GAP ’s and the spectrum
characteristics (10 dB Bandwidth and center frequency) of the pulses are shown in Table 5.8.
If the GAP is compared between the 100 ps and 500 ps cases, for both the Gaussian and
Gaussian Monocycle, it may observed that the 500 ps cases have higher gains compared to
the 100 ps cases, this is because the 500 ps pulses have narrower bandwidth which makes
use of the higher gain portions of the antennas’ frequency response.
In addition to the above pulses, GAP was calculated for two Gaussian Pulses (1ns & 5ns)
which were used to modulate a sinusoidal wave in order to change the center frequency of
the pulses. The results are displayed in Figure 5.11. It can be seen that for a very few cases
the resulting GAP might be close to GF riis , but for most cases is not. Furthermore, it can
be noticed, especially for the TimeDomain antennas that as the bandwidth of the pulse is
Gaussian & Gaussian Monocycle Pulses (Unit Energy) Gaussian & Gaussian Monocycle Pulses (Unit Energy) Spectrum
0.2
100ps Gaussian Pulse −20
500ps Gaussian Pulse
0.15 100ps Gaussian Monocycle
500ps Gaussian Monocycle −40
0.1
−60
Amplitude (V)
Energy (dB)
0.05 100ps Gaussian Pulse
−80 500ps Gaussian Pulse
0 100ps Gaussian Monocycle
−100 500ps Gaussian Monocycle
−0.05
−120
−0.1
−140
−0.15
−160
0 100 200 300 400 500 600 0 5 10 15 20 25
time (ps) Frequency (GHz)
(a) Time domain (b) Energy Spectrum
Figure 5.10: Gaussian & Gaussian Monocycle Pulses
Table 5.8: Sample GAP for different pulses
Pulse
Property Gaussian Gaussian Monocycle
100ps 500ps 100ps 500ps
10dB BW (GHz) 15.2 3.1 20.1 4.1
10dB BW Center Freq. (GHz) 7.6 1.53 12.1 2.4
Ridged TEM Horns GAP (dBm2 ) -13.84 -11.79 -18.37 -10.60
Bicones GAP (dBm2 ) -27.81 -23.79 -32.56 -25.83
Vivaldis GAP (dBm2 ) -17.77 -15.95 -22.76 -14.59
TimeDomains GAP (dBm2 ) -35.09 -42.21 -37.99 -35.21
reduced (longer pulse duration) GAP ’s variation follows the same fluctuations as the antenna
frequency response, i.e., when the antenna has a low response at a frequency, GAP will be
low as well. This may observed by comparing Figures 5.5 and 5.11(d).
Ridged TEM Horn Antennas GAP Vs 10db Bandwidth Center Frequency Bicone Antennas GAP Vs 10db Bandwidth Center Frequency
−6 1ns Gaussian Pulse −20 1ns Gaussian Pulse

5ns Gaussian Pulse 5ns Gaussian Pulse
G at that frequency G at that frequency
Friis Friis
−8
−10
−25
−12
GAP (dBm2)
GAP (dBm2)
−14
−16
−30
−18
−20
−22 −35
2 3 4 5 6 7 8 9 10 2 3 4 5 6 7 8 9 10
10db Bandwidth Center Frequency (GHz) 10db Bandwidth Center Frequency (GHz)
(a) Ridged TEM Antennas (b) Bicone Antennas

Vivaldi Antennas GAP Vs 10db Bandwidth Center Frequency Time Domain Antennas GAP Vs 10db Bandwidth Center Frequency
−5
−28
GFriis at that frequency −30 GFriis at that frequency
−10
−32
−34
−15
−36
GAP (dBm2)
GAP (dBm2)
−38
−20
−40
−42
−25
−44
−46
−30
−48
2 3 4 5 6 7 8 9 10 2 3 4 5 6 7 8 9 10
10db Bandwidth Center Frequency (GHz) 10db Bandwidth Center Frequency (GHz)
(c) Vivaldi Antennas (d) TimeDomain Antennas
Figure 5.11: GAP Vs Center Frequency

5.5.2 Time Domain Digital Sampling Oscilloscope Measurements
In this section the time domain measurements taken using a Digital Sampling Oscilloscope
are described and presented. The measurements include 1 m reference measurement using
the four antenna pairs described earlier for calculating their GAP using the time domain
method. Furthermore, the received pulses were measured for additional distances up to 25
m. The additional measurements were used to evaluate the predictions of the proposed
method.
Method
For the time domain measurements the following equipment were used:
• Geozondas Pulser T0.5 < 30ps, Vout > 30V /50Ω (GZ1106DL1 & GZ1117DN-25)
• HP33120A Function Generator
• Tektronix CSA8000B Digital Sampling Oscilloscope
• United Microwave cables Micropore 190 (10 ft for each antenna)
• Generic cables for trigger connections.
The equipment were divided into two sides: the transmit side, and the receive side (Fig-
ure 5.12). The transmit side consisted of the transmit antenna connected with a 10 ft cable
to the pulser, and the function generator. On the other side, there was the receive antenna
connected with another 10 ft cable to the Digital Sampling Oscilloscope (DSO). The function
generator was used to trigger both the pulser and the DSO with a 200 KHz half amplitude
mean square wave. In order to ensure that the DSO will not miss the pulse sent by the pulser
because of the delayed triggering (caused by the trigger cable length) the trigger cable that
connected the function generator to the pulser was slightly longer than the trigger cable to
the DSO. That ensured that the DSO was triggered and was looking for the pulse slightly
before it was sent.
Each signal measurement was taken averaging 500 waveforms in order to minimize the effects
of noise. The received window duration was 5ns for the LOS case and 20 ns for the Non Line
of Sight (NLOS) case. The antennas for the LOS case were placed 1.67 m high, and for the
NLOS case 1.2 m high. For the LOS case a very precise measurement (in terms of separation
distance) was taken at 1 m. The rest of the distances were estimated by the additional time
delay.
Tx Antenna Rx Antenna
U>30V/50Ω Antenna Height:

T <30ps
0.5 LOS 1.67m
NLOS 1.2m
Geozondas Pulse Output Channel 2 Input Tektronix
GZ1106DL1 & CSA8000B
Trigger Input Digital Sampling
GZ1117DN-25
5V Oscilloscope
0V Trigger Input
5μs t
HP33120A 1V
Function 0V
5μs t
Generator
Figure 5.12: Time Domain Measurements Equipment Setup
Results — Antenna Noise
Table 5.9 presents the noise measurements done outdoors. The receive antenna was con-
nected to the Digital Sampling Oscilloscope and the received signal was recorded without
the transmitter on. Similar to the other measurements, 500 waveform averaging was used.
The results show that the higher the antenna bandwidth the higher the recorded noise is.
The time window for these measurements was 5 ns and the bandwidth assumed for the noise
power spectral density (N0 ) calculations was the noise equivalent bandwidth of the antennas.
Table 5.9: Received Antenna Noise (500 waveform average)
Noise
Antennas Measured Calculated
Energy (dBJ) Power (dBW) N0 (dBW/Hz)
Ridged TEM Horns -187.64 -104.63 -202.56
Bicones -191.62 -108.61 -192.56
Vivaldis -194.49 -111.48 -205.08
TimeDomain -213.18 -130.17 -206.77
Results — Analysis
The received waveforms were filtered using a 50 MHz to 18.05 GHz ideal bandpass filter
(using the FFT and IFFT) to match the bandwidth measured by the VNA so the received
waveforms can be compared to the waveforms generated with the S21 parameters. A reference
noise measurement was taken with each antenna and the same filter was applied to the
reference noise. The noise energy and power were calculated and subtracted from the signal
energy/power calculations.
Results — LOS Measurements
The LOS measurements were taken at a height of 1.67 m on the road right next to the
sidewalk on the back of Durham Hall (Figure 5.13). The transmit location and the final
receive location are shown in the location photo (Figure 5.13). The receive measurements
were done along the receive path also shown in the same photo.
Durham Hall
Rx
Rx Path
Tx
LOS Outdoor Measurements Location
Figure 5.13: LOS Measurement Location
The results of the measurements are shown in Figures 5.14 to 5.17. Each figure:
(a) shows the received energy along with the predicted energy based on the calculated GAP
and a path loss exponent of 2 [Equation (5.20)]. In addition the predicted energy is also
calculated using the actual observed path loss exponent which was found to vary slightly
from the expected path loss exponent of 2.
(b) shows the same information as (a) for peak power.
(c) shows the received pulse at reference distance of 1 m.
(d) shows all received pulses normalized by the distances (on these pulses the ground reflec-
tion can be clearly seen and is coming closer and closer to the main pulse as the distance
increases). The reflections were time gated and only the main part of the pulse (e) was
used for the calculations.
(e) shows the the same information as (d) but showing only time window that was used for
the energy and peak power calculations, and finally,
(f) shows the spectrum of the generated pulse along with the first and last received pulse
spectra.
The path loss exponent n of the results was found to be smaller than 2. Initially this was
believed to be from additional multipath (besides the main ground reflection) that could
not be seen by looking at the received signal. However, by deconvolving the received signal
using the CLEAN5 algorithm [9], and reconstructing the received waveform by just using the
main (strongest path) and calculating the received energy yielded results with no significant
difference.
Reasons that might have contributed to a path loss exponent n = 2
• The measurements used a very small portion of the receiver’s dynamic range making
it prone to quantization errors.
• A high frequency triggering jitter was also observed, causing the received signal to move
around its position slightly, which combined with the averaging might have contributed
to additional energy level reduction.
• The TimeDomain antennas were the only pair found with a path loss slightly over 2,
and were the narrowest in bandwidth of the antennas used. In a way those antennas
were filtering the environmental noise before arriving at the receiver. This observation
partially supports the possibility that the higher antenna bandwidth might contribute
to the higher energy levels due to the additional noise.
5
The CLEAN algorithm is a deconvolution method that correlates the received waveform with a reference
pulse, finds the higher correlation, assigns a weight and subtracts the pulse from the received signal. This
procedure is repeated until no correlations above specific threshold can not be found. In this report, a 15
dB threshold, referring to the stronger path, was used.
Ridged TEM Horn Antennas Pulse Actual & Predicted Received Energy (dB) Ridged TEM Horn Antennas Pulse Actual & Predicted Received Peak Power (dB)
−120 −20
Measured Energy Measured Peak Power
Predicted Energy (n=2) Predicted Peak Power (n=2)
Predicted Energy (n=1.9513) Predicted Peak Power (n=1.9127)
−125 −25
−130 −30
Energy (dB)
Power (dB)
−135 −35
−140 −40
−145 −45
−150 −50
0 1 0 1
10 10 10 10
Distance (m) Distance (m)
(a) Received Energy (b) Received Power

Ridged TEM Horn Antennas Receive Pulse at 1m Ridged TEM Horn Antennas All Received Pulses x Distance
0.5
0.4
Received Pulse at 1m
0.4
0.3
0.3
0.2
0.2
0.1
Amplitude(V)
Amplitude(V)
0.1
0
0
−0.1
−0.1
−0.2
−0.2
−0.3 −0.3
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4

Time (s) x 10
−9 Time (s) x 10
−9
(c) Received Pulse (1m) (d) All Received Pulses Normalized by the
Distance
Ridged TEM Horn Antennas Received Pulses x Distance (Main Pulse Zoom) Ridged TEM Horn Antennas Generated & Received Pulse Spectrum
0.5 20
0.4
0
0.3
−20
0.2
Amplitude(V)
−40
Energy (dB)
0.1
0 −60
−0.1
−80
−0.2
−100
−0.3 Generated Pulse
Received Pulse at 25.7644m
−120
0 1 2 3 4 5 6 7 8 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s) −10
x 10 Frequency (Hz) x 10
10
(e) All Received Pulses Normalized by the (f) Received Spectrum

Distance (Main Pulse Zoom)
Figure 5.14: Ridged TEM Antenna Results

Bicone Antennas Pulse Actual & Predicted Received Energy (dB) Bicone Antennas Pulse Actual & Predicted Received Peak Power (dB)
−130 −35
−135 −40
−140 −45
Energy (dB)
Power (dB)
−145 −50
−150 −55
−155 −60
−160 −65
0 1 0 1
10 10 10 10

Bicone Antennas Receive Pulse at 1m Bicone Antennas All Received Pulses x Distance
0.06
0.02
0.04
0 0.02
0
−0.02
Amplitude(V)
Amplitude(V)
−0.02
−0.04 −0.04
−0.06
−0.06
−0.08
−0.08
−0.1
−0.12
−0.1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4

Time (s) x 10
−9 Time (s) x 10
−9
Distance
Bicone Antennas Received Pulses x Distance (Main Pulse Zoom) Bicone Antennas Generated & Received Pulse Spectrum
20
0.04
0.02 0
0
−20
−0.02
Amplitude(V)
−40
Energy (dB)
−0.04
−0.06 −60
−0.08
−80
−0.1
−100
Generated Pulse
−0.12 Received Pulse at 1m
−120
0 1 2 3 4 5 6 7 8 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s) −10
10

Figure 5.15: Bicone Antenna Results

Vivaldi Antennas Pulse Actual & Predicted Received Energy (dB) Vivaldi Antennas Pulse Actual & Predicted Received Peak Power (dB)
−125 −25
−30
−130
−35
−135
Energy (dB)
Power (dB)
−40
−140
−45
−145
−50
−150 −55
0 1 0 1
10 10 10 10

Vivaldi Antennas Receive Pulse at 1m Vivaldi Antennas All Received Pulses x Distance
0.3
0.3
0.25
0.2
0.2
0.15
0.1
0.1
Amplitude(V)
Amplitude(V)
0.05
0 0
−0.05
−0.1
−0.1
−0.15
−0.2
−0.2
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (s) x 10
−9 Time (s) x 10
−9
Distance
Vivaldi Antennas Received Pulses x Distance (Main Pulse Zoom) Vivaldi Antennas Generated & Received Pulse Spectrum
20
0.3
0
0.2
−20
0.1
Amplitude(V)
−40
Energy (dB)
0 −60
−80
−0.1
−100
Generated Pulse
−0.2
−120
0 1 2 3 4 5 6 7 8 9 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s) −10
10

Figure 5.16: Vivaldi Antenna Results

TimeDomain Antennas Pulse Actual & Predicted Received Energy (dB) TimeDomain Antennas Pulse Actual & Predicted Received Peak Power (dB)
−140 −45
−145 −50
−150 −55
Energy (dB)
Power (dB)
−155 −60
−160 −65
−165 −70
−170 −75
0 1 0 1
10 10 10 10

TimeDomain Antennas Receive Pulse at 1m TimeDomain Antennas All Received Pulses x Distance
0.03 0.03
0.02 0.02
Amplitude(V)
Amplitude(V)
0.01 0.01
0 0
−0.01
−0.01
−0.02
−0.02
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4

Time (s) x 10
−9 Time (s) x 10
−9
Distance
TimeDomain Antennas Received Pulses x Distance (Main Pulse Zoom) TimeDomain Antennas Generated & Received Pulse Spectrum
20
0.03 0
−20
0.02
−40
Amplitude(V)
Energy (dB)
0.01
−60
0
−80
−0.01 −100
−120 Generated Pulse

−0.02
−140
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s) x 10
−9 Frequency (Hz) x 10
10

Figure 5.17: TimeDomain Antenna Results

Results — NLOS Measurements
In this section the results of a sample NLOS measurement set are presented. Twelve points
at MPRG’s radio lab were taken (Figure 5.18). The equipment used and the setup were the
same as the other time domain measurements previously presented. The same pulse was
used as well. The only difference was that there was no line of sight between the transmit
and receive antennas (The Bicones). In this case the received window was 20 ns.
Distance
from Tx (m):
1: 2.26
2: 2.47
3: 2.90
1 12 4: 3.90
2 9 5: 4.86
3 8
5
6 10 11 6: 5.09
4 7: 5.38
7
8: 6.21
Drawing in approximate scale 9: 6.65
Legend
Transmit Location 10: 7.43
Receive locations 11: 7.74
Benches, cabinets, and electronic equipment 12: 8.30
Wooden frame glass door; closed during measurements
Interior dry wall
Outside wall
Figure 5.18: NLOS Measurement Locations
Analysis Procedure For this measurement set the distance between the trasnmit and
receive antennas was not estimated from the time delay like the previous cases (LOS). In-
stead, it was estimated by measuring the exact location of the receive points (the transmitter
location was fixed.) The received signals were bandpass filtered with a bandpass filter of 50
MHz to 18.05 GHz. Then each signal’s channel impulse response (CIR) was estimated using
the CLEAN algorithm
After the CIR was estimated, the received signal was reconstructed6 by convolving the CIR
with the reference pulse (of unit energy). Then the energy in the whole window was calcu-
lated. The peak power was calculated as well by using the absolute maximum peak on the
received window.
Using the reconstructed received waveform the ability of a Rake receiver to capture the
energy on the channel was evaluated. To simulate a Rake receiver the signal was correlated
(match filtered) with a unit energy reference pulse, then using a peak detection method the
first N stronger peaks were selected, where N is the number of the assumed Rake fingers.
6
This ensured a received signal with resolvable paths and no noise present.
Furthermore, peaks closer than 1/4 of the reference’s pulse width to already selected peaks
were ignored. The energy on the selected peaks was calculated and divided by the total
energy on the channel (energy capture of a perfect Rake,) even though it is known that
such receivers do not exists. The results are presented at Figure 5.20 and are very close to
results presented in [10]. The receiver’s performance is required when doing a link budget in
order to predict how much energy will will be captured by the receiver, therefore, making the
necessary modifications to the link budget parameters for achieving the desired performance.
Observed Path loss The energy path loss exponent n observed from this set was 2.09.
However, other more extensive measurements observed a path loss exponent of 2.3 (σ = 2.4
dB) for Bicones and 2.4 (σ = 5.1 dB) for Horn Antennas [5]. In addition, [11] found a path
loss exponent of 2.95 to 3.12 using a Vector Network Analyzer in the range of 2 to 8 GHz.
Furthermore, the IEEE 802.15.3a standards task group adopted an n = 3.5 with σ = 0.97
[12]. The reason for getting a small path loss exponent from the measurements presented
here is because of the short distances that allow small amplitude multipath to be strong
enough to be detected. Furthermore, the pulse used was narrower and combined with the
higher sampling frequency (finner time resolution) these factors allow the resolution of more
multipath components which contributed to the received energy. On the other hand, the
path loss exponent of the peak power was found to be 3.37. Finally, high accuracy is not
expected from such a small set of measurement points which were taken for link budget
verification purposes and not from the channel characterization perspective. Nevertheless,
Table 5.10 presents the observed channel statistics for the above measurements.
Table 5.10: Average NLOS Statistics
n σ τ mean τ rms Np
2.09 5.30 dB 7.52 ns 4.98 ns 53
NLOS Link Budget Considerations When doing a link budget for NLOS operation,
additional factors must be taken in consideration. If the path loss exponent is defined in
terms of the total energy of the channel (including multipath), then a receiver performance
factor must be included, because a practical receiver will be able to capture only a fraction
of the total energy as demonstrated in Figure 5.20. Furthermore, the received energy in a
NLOS channel fluctuates significantly more than in the LOS case and a larger fading margin
must be considered.
NLOS Bicone Antennas Actual & Predicted Received Energy (dB) (20ns window) NLOS Bicone Antennas Pulse Actual & Predicted Received Peak Power (dB)
−140 −45
Predicted Energy (n=2) Predicted Peak Power (n=3.3658)
−142 Predicted Energy (n=2.085)
−50
−144
−146
−55
Energy (dB)
Power (dB)
−148
−60
−150
−152
−65
−154
−156 −70
0 1 0 1
10 10 10 10

Reference Pulse (Unit Energy) NLOS Bicone Antennas Received Window at 2.26m
0.01
0.008
0
0.006
0.004
−0.05
0.002
Amplitude(V)
Amplitude(V)
−0.1
−0.002
−0.004
−0.15 −0.006
−0.008
−0.2 −0.01
0 1 2 3 4 5 6 7 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

Time (s) x 10
−10 Time (s) x 10
−8
(c) Reference Pulse (d) 1st Received Window

NLOS Bicone Antennas All Received CIRs
NLOS Bicone Antennas Received CIR at 2.26m (Aligned to the 1st path & Normalized for Unit Energy)
0.05
0.5
0.04
0.4
0.03
0.3
0.02
0.2
0.01
Amplitude(V)
Amplitude(V)
0.1
0
0
−0.01
−0.02 −0.1
−0.03 −0.2
−0.04
−0.3
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 2 4 6 8 10 12 14 16 18 20

Time (s) x 10
−8 Time (s) −9
x 10
(e) 1st Received Window CIR (f) All Received CIRs
Figure 5.19: NLOS Bicone Results

Rake Receiver Cumulative Energy Per Finger

1
0.9
0.8
0.7
Cumulative Energy
0.6
0.5
0.4
0.3
0.2
0.1
0
5 10 15 20 25 30 35 40 45 50 55 60
Number of Rake fingers
Figure 5.20: Rake Receiver Performance

5.5.3 Link Budget Calculations
Using the data collected and calculated in this section (Validation) link budget was calculated
by estimating the received energy using the proposed method and the Friis transmission
formula.
Figures 5.21 to 5.28 show the results for the received energy and peak power along with
sample link budget calculations for the LOS cases. For the Friis formula calculations the 10
dB center bandwidth of the pulse was used which was 4.6 GHz and the received peak power
frequency of each antenna pair was used as well.
LOS Example Calculations
Tables 5.11 and 5.12 show a set of actual calculations done in the same way that were
performed for Figures 5.21 to 5.28.
For the TimeDomain antennas case, used in the example calculations, the most accurate way
for doing link budgets based on received energy and peak power was the proposed method
using GAP . However, as it can be seen from Figures 5.21 to 5.26 the energy link budget
(correlator) is also accurately estimated using the Friis transmission formula. On the other
hand, the peak power (detector) link budget is still not accurately calculated. The results
demonstrate that by using the Friis transmission formula there are cases that might be valid,
but not always, on the other hand, the proposed method using GAP is always very close to
the observed signal levels for both energy and peak power.
NLOS Example
On Table 5.13 a sample set of NLOS link budget calculations are presented. Some of the
parameters used were obtained from Section 5.5.2 were the NLOS case was investigated.
The NLOS link budget, compared to the LOS link budget, has to take in account the larger
path loss experienced in a NLOS scenario. Furthermore, the path loss observed at any
given point will deviate from its average value due to variations in the environment. The
path loss variation has been shown to follow a log-normal distribution [5]. From a link
budget perspective, a big enough fading margin must be used in order to ensure a certain
performance at a % of the time. For the calculations it was assumed n = 2.3 with a σ = 2.4
dB [5]. The fading margin was calculated to be 6dB for providing the target Eb /N0 at least
99.1 % of the cases. The probability that the received signal level assuming log-normal fading
(in dB power units) will exceed a certain value γ can be calculated from the normal CDF
using Equation (5.29) [7]:

γ − PR (r)
P r[PR (r) > γ] = Q (5.29)
σ
Table 5.11: TimeDomain Antenna Example Link Budget Calculations at r = 15.68m—Correlator Detector (LOS)
Proposed Friis Friis Using

Method Formula Formula Measured
Using GAP (4.6Ghz) (3Ghz) Energy
Received Energy
Transmitted Energy ET −95.47dBJ −95.47dBJ −95.47dBJ −95.47dBJ
Antenna Pair–Pulse GAP for Energy −37.99dBm2
Antenna Gains (GT + GR ) 6dBi 4dBi
1 −2
Path Loss 4πr −34.9dBm
λ2 2
Path Loss 4πr −69.61dB −65.9dB
Received Energy ER −168.36dBJ −159.08dBJ −157.37dBJ −168.65dBJ
Noise PSD
Boltzmann’s Constant (dBW/K/Hz) −228.6 −228.6 −228.6 −228.6
System Noise Temp (T0 = 290K) 24.6dBK 24.6dBK 24.6dBK 24.6dBK
Noise PSD N0 −204dBW/Hz −204dBW/Hz −204dBW/Hz −204dBW/Hz
Received Energy per Bit

CHAPTER 5. ULTRA WIDEBAND LINK BUDGET DESIGN
Pulses per bit Ns = 1 1 1 1 1
Energy/bit Eb = ER + 10log(Ns ) −168.36dBJ −159.08dBJ −157.37dBJ −168.65dBJ
Final Eb /N0
Fading Margin MF 0dB 0dB 0dB 0dB
Energy Capture Loss (20%) ρ −7dB −7dB −7dB −7dB
Eb /N0 = Eb − N0 − MF + ρ 28.64dB 37.92dB 39.63dB 28.35dB

86
Table 5.12: TimeDomain Antenna Example Link Budget Calculations at r = 15.68m—Peak Detector (LOS)
Proposed Friis Friis Using

Method Formula Formula Measured
Using GAP (4.6Ghz) (3Ghz) Power
Received Power
Transmitted Peak Power PT 9.26dBW 9.26dBW 9.26dBW 9.26dBW
Ant. Pair–Pulse GAP PPower −43.82dBm2
Antenna Gains (GT + GR ) 6dBi 4dBi
1 −2
Path Loss 4πr −34.9dBm
λ2 2
Path Loss 4πr −69.61dB −65.9dB
Received Peak Power PR −69.46dBW −54.35dBW −52.64dBW −70.18dBW
Noise Power
Boltzmann’s Constant (dBW/K/Hz) −228.6 −228.6 −228.6 −228.6
System Noise Temp (T0 = 290K) 24.6dBK 24.6dBK 24.6dBK 24.6dBK
Received 10db BW (4GHz) 96.02dBHz 96.02dBHz 96.02dBHz 96.02dBHz
CHAPTER 5. ULTRA WIDEBAND LINK BUDGET DESIGN
Noise Power N0 −107.98dBW −107.98dBW −107.98dBW −107.98dBW
Final Eb /N0
Fading Margin MF 0dB 0dB 0dB 0dB
Capture Loss (10%) ρ −10dB −10dB −10dB −10dB
Eb /N0 = PR − N0 − MF + ρ 28.52dB 43.63dB 45.34dB 27.8dB

87
Table 5.13: Bicone Antenna Example Link Budget Calculations at r = 7.43m—Correlator

Detector (NLOS)
Proposed Using
Method Measured
Using GAP Energy
Received Energy
Transmitted Energy ET −95.47dBJ −95.47dBJ
Antenna Pair–Pulse GAP for Energy −28.57dBm2
Path Loss 4πr1 n ,n = 2.3[5], r = 7.43m −28.41dBm2
Received Energy ER −152.45dBJ −155dBJ ∗
Noise PSD
Boltzmann’s Constant −228.6dBW/K/Hz −228.6dBW/K/Hz
System Noise Temp (T0 = 290K) 24.6dBK 24.6dBK
Noise PSD N0 −204dBW/Hz −204dBW/Hz
Received Energy per Bit

Pulses per bit Ns = 1 1 1
Energy per bit Eb = ER + 10log(Ns) −152.45dBJ −155dBJ
Final Eb /N0
Fading Margin MF
Eb
(P r{ N < T arget} < 1%) 6dB 6dB
0
Energy Capture Loss ρ (40%, 5
−4dB −4dB
Fingers, Figure 5.20)
Eb /N0 = Eb − N0 − MF + ρ 41.55dB 39dB
∗
Ideally this value should be the average multiple observations for the desired distance
Ridged TEM Horn Antennas Measured & Estimated Received Energy Ridged TEM Horn Antennas Sample Link Budget (Eb/N0) − Correlator Detector
95
Measured Using Measured Energy
−105 Using Proposed Method Using Proposed Method
Friis Est. at 10dB BW (4.6GHz) 90 Friis Est. at 10dB BW (4.6GHz)
Friis Est. at RX peak power freq.(1GHz) Friis Est. at RX peak power freq.(1GHz)
−110
85
−115
80
−120
Energy (dBJ)
75
Eb/N0 (dB)
−125
70
−130
65
−135
60
−140
55
−145
50
0 1 0 1
10 10 10 10
(a) Received Energy (b) Energy Link Budget
Figure 5.21: Ridged TEM Antenna Received Energy and Link Budget
Ridged TEM Horn Antennas Sample Link Budget (Eb/N0) − Peak Detector
Measured
Ridged TEM Horn Antennas Measured & Estimated Received Peak Power 95 Using Proposed Method
Friis Est. at 10dB BW (4.6GHz)
Measured Friis Est. at RX peak power freq.(1GHz)
0
Using Proposed Method 90
Friis Est. at 10dB BW (4.6GHz)
−5 Friis Est. at RX peak power freq.(1GHz)
85
−10
80
−15
Eb/N0 (dB)
75
Peak Power (dBW)
−20
70
−25
65
−30
−35 60
−40 55
−45
50
−50
0 1 0 1
10 10 10 10
(a) Received Peak Power (b) Peak Power Link Budget
Figure 5.22: Ridged TEM Antenna Received Peak Power and Link Budget
Bicone Antennas Measured & Estimated Received Energy Bicone Antennas Sample Link Budget (Eb/N0) − Correlator Detector
−115
Measured 80 Using Measured Energy
Using Proposed Method Using Proposed Method
−120 Friis Est. at 10dB BW (4.6GHz) Friis Est. at 10dB BW (4.6GHz)
75
−125
70
−130
65
Energy (dBJ)
−135
Eb/N0 (dB)
60
−140
55
−145
50
−150
45
−155
40
−160
0 1 0 1
10 10 10 10
Figure 5.23: Bicone Antenna Received Energy and Link Budget

Bicone Antennas Measured & Estimated Received Peak Power Bicone Antennas Sample Link Budget (Eb/N0) − Peak Detector
−10
Measured Measured
85
−15 Friis Est. at 10dB BW (4.6GHz) Friis Est. at 10dB BW (4.6GHz)
Friis Est. at RX peak power freq.(1GHz) 80 Friis Est. at RX peak power freq.(1GHz)
−20
75
−25
70
−30
Peak Power (dBW)
65
Eb/N0 (dB)
−35
60
−40
55
−45
50
−50
45
−55
40
−60
0 1 0 1
10 10 10 10
Figure 5.24: Bicone Antenna Received Peak Power and Link Budget
Vivaldi Antennas Measured & Estimated Received Energy Vivaldi Antennas Sample Link Budget (Eb/N0) − Correlator Detector
85
Measured Using Measured Energy
Friis Est. at 10dB BW (4.6GHz) Friis Est. at 10dB BW (4.6GHz)
Friis Est. at RX peak power freq.(2GHz) 80 Friis Est. at RX peak power freq.(2GHz)
−120
75
−125
70
Energy (dBJ)
Eb/N0 (dB)
−130
65
−135
60
−140
55
−145
50
−150
45
0 1 0 1
10 10 10 10
Figure 5.25: Vivaldi Antenna Received Energy and Link Budget

Vivaldi Antennas Measured & Estimated Received Peak Power Vivaldi Antennas Sample Link Budget (E /N ) − Peak Detector
b 0
Measured Measured
85
−15
80
−20
75
Peak Power (dBW)
−25
70
Eb/N0 (dB)
−30
65
−35
60
−40
55
−45
50
−50
45
0 1 0 1
10 10 10 10
Figure 5.26: Vivaldi Antenna Received Peak Power and Link Budget
TimeDomain Antennas Measured & Estimated Received Energy TimeDomain Antennas Sample Link Budget (Eb/N0) − Correlator Detector
−135 Measured Using Measured Energy

60
−140
55
−145
50
Energy (dBJ)
−150
Eb/N0 (dB)
45
−155
40
−160
35
−165
30
−170
0 1 0 1
10 10 10 10
Figure 5.27: TimeDomain Antenna Received Energy and Link Budget

TimeDomain Antennas Measured & Estimated Received Peak Power TimeDomain Antennas Sample Link Budget (Eb/N0) − Peak Detector
−30 Measured Measured

65
−35 Friis Est. at RX peak power freq.(3GHz) Friis Est. at RX peak power freq.(3GHz)
60
−40
55
−45
Peak Power (dBW)
50
Eb/N0 (dB)
−50
45
−55
40
−60
35
−65
30
−70
0 1 0 1
10 10 10 10
Figure 5.28: TimeDomain Antenna Received Peak Power and Link Budget
5.6 Conclusions
For narrowband systems an accurate link budget method exists based on the Friis trans-
mission formula. However, that link budget is designed for a single frequency of operation.
Using the center frequency of the transmitted I-UWB pulse along with this traditional model
is not sufficient for most cases. Moreover, the Friis transmission formula is not able to cap-
ture all the spectral characteristics of the different UWB pulses and the antennas used in the
communications link. For that reason all the antenna related parameters in the Friis formula
were replaced with a single term called the antenna coupling gain GAP which fully captures
the pulse’s spectrum characteristics and its interaction with the used antennas. Therefore,
an antenna’s GAP is pulse dependent.
It was shown that GAP can be easily calculated by taking a sample time domain measurement
of the desired pulse and the antenna pair of the link. Furthermore, if the S21 parameters of
the antenna pair are available the calculation of the GAP for different pulses also is an easy
task because the prospective pulse can be mathematically generated and by using a simple
Matlab script the received pulse and GAP can be calculated.
The claim that the GAP is more accurate than the traditional method was verified by esti-
mating GAP for four pairs of UWB antennas both by using the time domain method and
the S21 method which were found to be in close agreement to each other. Furthermore, it
was demonstrated that the Friis formula can not capture all the cases by generating some
cases of Gaussian and Gaussian monocycle pulses with different spectrum characteristics.
It was found that the proposed method of using GAP consistently and accurately predicts the
link budget for both energy and peak power scenarios. On the other hand, the Friis formula,
though it was found for many cases involving the estimation of an energy-based link budget
to be accurate, does not provide sufficient accuracy for all cases. Furthermore, using the
Friis formula for the received peak power estimation was inaccurate suggesting that it can
not be used for calculating a peak power (detector) link budget. The predicted link budgets
were verified using the sample LOS and NLOS measurements taken in the validation section.
In this report, GAP it was only measured using one pulse. However, it was calculated for
several pulses. Future work can complement this work by estimating GAP for different pulses
using both proposed methods. This suggestion requires the availability of a pulser(s) that
can generate a variety of pulses.
The main focus of this chapter was to verify the received energy/power estimation component
of the proposed link budget. Future work can further investigate and verify other important
parameters of the link budget, such as receiver structure dependence and losses due to
pulse mismatch because of the channel distortion. In addition, it may propose a much
more thorough and elaborate link budget. Finally, if the right equipment are available a
complete communications link can be predicted with the link budget and be implemented
for verification.
Vita
From 1999 to 2002 Haris I. Volos attended the Higher Technical Institute in Nicosia, Cyprus.
Volos graduated with a Higher National Diploma (HND) in Electrical Engineering and was
awarded Best Performance in both Electronic and Power Engineering subjects. In 2002 Volos
was awarded an International Student Scholarship and attended Old Dominion University,
where he was a graduate of the honors college with a BS (summa cum laude) in Electrical
Engineering and a minor in Computer Engineering. In 2004 Volos was awarded the Cyprus
America Scholarship Program CASP Scholarship for MS studies at Virginia Polytechnic
Institute and State University.
95
Bibliography
[1] Opeermann, I.; Hämäläinen, H.; linatti, J., UWB Theory and Applications. West Sussex,
England: John Wiley & Sons Ltd 2004
[2] Joon-Yong Lee; Scholtz, R.A., ”Ranging in a dense multipath environment using an
UWB radio link,” Selected Areas in Communications, IEEE Journal on, vol.20, no.9pp.
1677- 1683, Dec 2002
[3] Chung, W.C.; Ha, D., ”An accurate ultra wideband (UWB) ranging for precision asset
location,” Ultra Wideband Systems and Technologies, 2003 IEEE Conference on , vol.,
no.pp. 389- 393, 16-19 Nov. 2003
[4] Hahafsza, B.R., Introduction to Radar Analysis, Boca Raton, Florida: CRC Press LLC,
1998
[5] Reed,J. H.; Ed., An Introduction to Ultra Wideband Communications Systems, Craw-
fordsville, Indiana: Prentice Hall, 2005.
[6] Proakis, J. G., Digital Communications. Boston, Massachusetts: McGraw-Hill, 4th ed.,
2001
[7] Rappaport, T. S., Wireless Communications: Principles and Practice. Upper Saddle
River, New Jersey: Prentice Hall, 2002
[8] Urkowitz, H., Signal Theory and Random Processes, Artech House, 1983.
[9] Scholtz, R.A.; Cramer, R.J.-M.; Win, M.Z., ”Evaluation of the propagation character-
istics of ultra-wideband communication channels,” Antennas and Propagation Society
International Symposium, 1998. IEEE , vol.2, no.pp.626-630 vol.2, 21-26 Jun 1998
[10] Donlan, B.M.; Venkatesh, S.; Bharadwaj, V.; Buehrer, R.M.;Jiann-An Tsai, ”The ultra-
wideband indoor channel,” Vehicular Technology Conference, 2004. VTC 2004-Spring.
2004 IEEE 59th , vol.1, no.pp. 208- 212 Vol.1, 17-19 May 2004
[11] Ghassemzadeh, S.S.; Greenstein, L.J.; Kavcic, A.; Sveinsson, T.; Tarokh, V., ”UWB
indoor path loss model for residential and commercial buildings,” Vehicular Technology
96
BIBLIOGRAPHY 97
Conference, 2003. VTC 2003-Fall. 2003 IEEE 58th , vol.5, no.pp. 3115- 3119 Vol.5, 6-9
Oct. 2003
[12] Molisch, A.F.; Foerster, J.R.; Pendergrass, M., ”Channel models for ultrawideband per-
sonal area networks,” Wireless Communications, IEEE [see also IEEE Personal Com-
munications] , vol.10, no.6pp. 14- 21, Dec. 2003

Thesis PDF

Uploaded by

Thesis PDF

Uploaded by

Ultra Wideband Ranging and Link Budget Design for Naval Crane

Thesis submitted to the Faculty of the

R. Michael Buehrer, Chair

Keywords: UWB, Ranging, Link Budget

I Ultra Wideband Ranging 3

2 UWB Ranging Theoretical Development 4

3 UWB-based Ranging:Hardware Demonstration 26

4 UWB Ranging Conclusions 46

II Ultra Wideband Link Budget Design 48

5 Ultra Wideband Link Budget Design 49

2.1 Problem Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3.1 Demonstration General Layout (not to scale) . . . . . . . . . . . . . . . . . . 26

5.1 VNA Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

3.1 Single-Side Sample Measurements (mm) . . . . . . . . . . . . . . . . . . . . 44

5.1 Conventional Free Space Link Budget . . . . . . . . . . . . . . . . . . . . . . 53

Ultra Wideband Ranging

UWB Ranging Theoretical

Figure 2.1: Problem Scenario

2.2 Using UWB Pulses for ranging

Figure 2.2: Ranging Sensors’ Location

2.3 Estimating Crate Orientation

2.3.1 Estimating Height above a Level Deck

Figure 2.3: Reﬂected Path Distance Estimation

n Vs x, d=10, h=10 a and b Vs x, d=10, h=10

(a) n = a + b Vs x (b) a and b Vs x

Figure 2.4: n = a + b Minimum Point

Referring to Figure 2.3, from trigonometry:

Taking the ﬁrst derivative with respect to x:

Setting the ﬁrst derivative to zero:

2.3.2 Estimating Height Above Deck with slope ϕ

sin(ϕ + θ) sin(180 − 2θ)

sin(θ − ϕ) sin(180 − 2θ)

Figure 2.5: Reﬂection With a Slope ϕ

sin (90 − (−ϕ + θ)) sin(90)

From Pythagorean Theorem (On the same triangle):

sin (90 − ϕ) sin(θ)

2.3.3 Determining Height and Orientation Relative to Deck

hc4 h1 hc1 hc1 h2 hc2 hc2 h3 hc3 hc3 h4 hc4

(b) Crate’s Sides

Figure 2.6: Crate’s Dimensions, Heights, and Slopes

To ﬁnd the heights of the corners the following procedure is performed:

2. Find the center heights of sides 1 and 3 using Equation (2.18).

5. Average the two heights for each corner.

6. Find two new vlaues of ϕl and ϕw using the four corners.

8. Repeat from 2, until the calculated angle ϕ values don’t change.

(a) ϕl = 8.38o, ϕw = 10.64o (b) ϕl = 14.09o, ϕw = 0o

Figure 2.7: Sample Plots

2.4 Simulation Tests

Figure 2.8: Crate Dimensions Used for the Simulation

• 11 diﬀerent heights from 5m to 10m.

• 6 diﬀerent slopes for each side from 0 to 20degrees.

Estimated h2 variance (m2)

(a) Height of Side 1 (b) Height of Side 2

Estimated h4 variance (m2)

(c) Height of Side 3 (d) Height of Side 4

Estimated h2 variance (m2)

(a) Height of Side 1 (b) Height of Side 2

Estimated h4 variance (m2)

(c) Height of Side 3 (d) Height of Side 4

Estimated h2 variance (m2)

(a) Height of Side 1 (b) Height of Side 2

Estimated h4 variance (m2)

(c) Height of Side 3 (d) Height of Side 4

Figure 2.11: Estimation Variance Vs Assumed Height (Crate Dimensions dw = 5m, dl =

Estimated h2 variance (m2)

(a) Height of Side 1 (b) Height of Side 2

Estimated h4 variance (m2)

(c) Height of Side 3 (d) Height of Side 4

2.5 Arrival Time Estimation Accuracy

Received Select time of the

Figure 2.13: Acquisition Process

2.6 Sample Link Budget using Radar Equation

Estimation σ CLRB and Simulation Results using Gaussian Pulses

Estimation Standard Deviation (m) −2

The radar equation is

Max Allowed Pulses (Na )= 3 × 106