You are on page 1of 103

The Florida State University

DigiNole Commons
Electronic Theses, Treatises and Dissertations

The Graduate School

11-30-2010

Study Of Correlations Between Microwave


Transmissions And Atmospheric Effects
Andrew James Stringer
Florida State University

Follow this and additional works at: http://diginole.lib.fsu.edu/etd


Recommended Citation
Stringer, Andrew James, "Study Of Correlations Between Microwave Transmissions And Atmospheric Effects" (2010). Electronic
Theses, Treatises and Dissertations. Paper 396.

This Thesis - Open Access is brought to you for free and open access by the The Graduate School at DigiNole Commons. It has been accepted for
inclusion in Electronic Theses, Treatises and Dissertations by an authorized administrator of DigiNole Commons. For more information, please contact
lib-ir@fsu.edu.

THE FLORIDA STATE UNIVERSITY


COLLEGE OF ENGINEERING

STUDY OF CORRELATIONS BETWEEN MICROWAVE TRANSMISSIONS AND


ATMOSPHERIC EFFECTS

By
ANDREW J. STRINGER

A Thesis submitted to the


Department of Electrical and Computer Engineering
in partial fulfillment of the
requirements for the degree of
Master of Science

Degree Awarded:
Fall Semester, 2010

The members of the Committee approve the thesis of Andrew J. Stringer defended on November
30th, 2010.

Dr. Simon Y. Foo


Professor Directing Thesis

Dr. Ming Yu
Committee Member

Dr. Bruce A. Harvey


Committee Member

Approved:

Dr. Simon Y. Foo, Chair, Department of Electrical and Computer Engineering

Dr. Ching-Jen Chen, Dean, College of Engineering.

The Graduate School has verified and approved the above-named committee members.
ii

ACKNOWLEDGEMENTS

I would like to thank and express my deepest appreciation to Dr. Simon Y. Foo and thank him
for his constant encouragement, criticism, perspectives, and ongoing inspiration. As a thesis
director, teacher, and friend to me, you have been an invaluable resource and have helped me
tremendously to complete this thesis.
I would like to thank Dr. Bruce A. Harvey as a valued committee member and for your guidance
and knowledge in rain attenuation models and wireless communications.
I also want to thank committee member Dr. Ming Yu for helping me challenge myself and enrich
my knowledge in computer programming.
I would like to extend a special thank you to William R. Allen, P.E. for his extended support,
criticism, and knowledge in wireless communications through the course of this project.
I would like to thank members of Florida Department of Transportation Traffic Engineering
Research Lab, specifically Ron Meyer, Vernell Johnson, and Derrick Vollmer, for their ongoing
efforts in helping make this project a success.
I would also like to thank Florida Department of Transportation District Three employee, Mark
Nallick for his programming knowledge and support.
I would also like to thank the Florida State University - College of Engineering Department of
Electrical and Computer Engineering, RCC Consultants, Inc., the Florida Department of
Transportation, and the RWIS and Clarus Initiative projects for their ongoing grants, assistance,
and support that made this research possible.
Finally, I would like to express my love for my parents, Michael and Barbara, my brothers, Nick
and Chris, and my partner, Christina Katopodis for their unfaltering support and encouragement,
and always believing in me. I could not have finished this manuscript without you. I love you
all.
iii

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................................ vi


LIST OF FIGURES ..................................................................................................................... vii
LIST OF ABBREVIATIONS ....................................................................................................... ix
ABSTRACT .................................................................................................................................. xi
1. INTRODUCTION .....................................................................................................................1
1.1. Overview ..........................................................................................................................1
1.2. Motivation ........................................................................................................................4
1.3. Problem Statement............................................................................................................4
1.4. Scope of Work ..................................................................................................................5

2. CRANE ATTENUATION MODELS ......................................................................................6


2.1. Global (Crane) Model.......................................................................................................6
2.2. Initial Two-Component Model .........................................................................................9
2.2.1. Volume Cell Contribution.........................................................................................9
2.2.2. Debris Contribution ................................................................................................12
2.2.3. Probability of Terrestrial Rain Rate ........................................................................13
2.2.4. Attenuation along a LOS Path ................................................................................14
2.3. Revised Two-Component Model ...................................................................................15
2.3.1. Model for Volume Cell Contribution .....................................................................15
2.3.2. Model for Debris Contribution ...............................................................................15

3. ITU ATTENUATION MODEL AND OTHER ATTENUATION MODELS.....................16


3.1. International Telecommunications Union Model ...........................................................16
3.2. Other Attenuation Models ..............................................................................................21

4. COMPUTER SIMULATION RESULTS AND KEY FINDINGS .......................................22


4.1. Data Acquisition .............................................................................................................22
4.2. Crane Models, ITU Model, and Path Loss 4.0 Analysis ................................................24
4.2.1. Analysis of Data Using Crane Models....................................................................24
iv

4.2.2. International Telecommunications Union Model Analysis ....................................25


4.2.3. Path Loss 4.0 Analysis ............................................................................................26
4.2.3.1. Greenville Analysis ......................................................................................27
4.2.3.2. Lake City DOT Analysis ..............................................................................29
4.2.3.3. SR-222 Analysis ...........................................................................................31
4.3. Correlation Analysis without Data Preprocessing ..........................................................33
4.4. Fast Fourier Transform and Power Spectrum Analysis .................................................38
4.4.1. Fast Fourier Transform Analysis ............................................................................38
4.4.2. FFT Spectrum Analysis ..........................................................................................39
4.4.3. Correlation Analysis ...............................................................................................41
4.5. Short Time Fourier Transform and Power Spectrum Analysis ......................................41
4.5.1. Short Time Fourier Transform Analysis .................................................................41
4.5.2. STFT Power Spectrum Analysis .............................................................................42
4.5.3. Correlation Analysis ...............................................................................................44
4.6. Discrete Wavelet Transform and Wavelet Decomposition Analysis .............................44
4.6.1. Wavelet Decomposition Analysis ...........................................................................44
4.6.2. Correlation Analysis ...............................................................................................51
4.7. Key Findings ..................................................................................................................52

5. CONCLUSION AND FUTURE WORK ...............................................................................56


5.1. Conclusion ......................................................................................................................56
5.2. Future Work and Recommendations ..............................................................................57

APPENDIX A: PROGRAM CODE ............................................................................................59


APPENDIX B: DEVICE SPECIFICATIONS AND DATASHEETS .......................................75
BIBLIOGRAPHY .........................................................................................................................88
BIOGRAPHICAL SKETCH ........................................................................................................90

LIST OF TABLES

Table 3.1: ITU Rain Rate Data for 0.001% Rain Fades in the Americas .....................................17
Table 3.2: Interpolated Regression Coefficients for 1-30 GHz ....................................................20
Table 3.3: ITU Rainfall Rates for Different Probabilities and Rain Regions .............................21
Table 4.1: Path Loss 4.0 Print Summary for Greenville ...............................................................28
Table 4.2: Path Loss 4.0 Print Summary for Lake City DOT.......................................................30
Table 4.3: Path Loss 4.0 Print Summary for SR-222 ...................................................................32
Table 4.4: Correlation Coefficients for Greenville .......................................................................33
Table 4.5: Correlation Coefficients for Lake City DOT ...............................................................33
Table 4.6: Correlation Coefficients for SR-222 ............................................................................34
Table 4.7: RSL Correlation Coefficients of the Chosen Sites ......................................................34
Table 4.8: RSL and Weather Parameter Cross-Correlation Coefficients for
Greenville .....................................................................................................................35
Table 4.9: RSL and Weather Parameter Cross-Correlation Coefficients for
Lake City DOT ............................................................................................................36
Table 4.10: RSL and Weather Parameter Cross-Correlation Coefficients for
SR-222 ......................................................................................................................37
Table 4.11: FFT Correlation Coefficients for Greenville .............................................................41
Table 4.12: Correlation Coefficients of Three Level Wavelet Decomposition for
Greenville ..................................................................................................................51
Table 4.13: Correlation Coefficients of Three Level Wavelet Decomposition for
Lake City DOT .........................................................................................................51
Table 4.14: Correlation Coefficients of Three Level Wavelet Decomposition for
SR-222 ......................................................................................................................52

vi

LIST OF FIGURES

Figure 1.1: A Typical Communication System ..............................................................................1


Figure 1.2: FDOT Statewide Telecommunications Network Deployment Map ............................3
Figure 2.1: Multiplier

in the Power-Law Relationship between Specific

Attenuation and Rain Rate ...........................................................................................7


Figure 2.2: Exponent

in the Power-Law Relationship between Specific

Attenuation and Rain Rate ...........................................................................................8


Figure 2.3: Edfs for the Joint Occurrence of Reflectivity and Square Root Area .......................10
Figure 2.4: Average Area of Volume Cells as Measured and Modeled Using an
Exponential Square Root Area Model........................................................................11
Figure 3.1: ITU Atmospheric Attenuation Prediction ..................................................................17
Figure 3.2: ITU Rain Regions for the Americas ...........................................................................18
Figure 3.3: ITU Rain Regions for Europe and Africa ..................................................................19
Figure 3.4: ITU Rain Regions for Asia .........................................................................................19
Figure 4.1: Sample Comma-Delimited Text File from the Control Module at
Greenville ESS site .....................................................................................................23
Figure 4.2: Weather Master 2000TM Example ..............................................................................23
Figure 4.3: Netboss Example ........................................................................................................24
Figure 4.4: ITU Model Rain Attenuation Prediction for Greenville Site .....................................25
Figure 4.5: ITU Model Rain Attenuation Prediction for Lake City DOT Site .............................26
Figure 4.6: Path Loss 4.0 Path Profile for Greenville ...................................................................27
Figure 4.7: Path Loss 4.0 Path Profile for Lake City DOT...........................................................29
Figure 4.8: Path Loss 4.0 Path Profile for SR-222 .......................................................................31
Figure 4.9: FFT of Greenville RSL and ESS data ........................................................................38
Figure 4.10: Enlarged Window of the FFT of Greenville RSL and ESS data ..............................39
Figure 4.11: Power Spectrum of Greenville RSL and ESS Data for One Day.............................40
Figure 4.12: Power Spectrum of Greenville RSL and ESS Data for a One Hour ........................40

vii

Figure 4.13: RSL STFT at 45 Angle for Greenville ESS Rotated Approximately 180 ............42
Figure 4.14: RSL STFT Power Frequency vs. Amplitude for Greenville ESS ............................43
Figure 4.15: RSL STFT Power Time vs. Amplitude for Greenville ESS.....................................43
Figure 4.16: Discrete wavelet transform illustration ....................................................................45
Figure 4.17: Stages of a Three Level Wavelet Decomposition ....................................................46
Figure 4.18: Wavelet Decomposition of Precipitation and RSL for Greenville Data ..................47
Figure 4.19: Wavelet Decomposition for RSL, RH, and T at Greenville ESS Site .....................48
Figure 4.20: Enlarged Wavelet Decomposition for Greenville Data ............................................48
Figure 4.21: Wavelet Decomposition of Precipitation and RSL for Lake City DOT Data ..........49
Figure 4.22: Enlarged Wavelet Decomposition for Lake City DOT data ....................................49
Figure 4.23: Wavelet Decomposition of Precipitation and RSL for SR-222 data ........................50
Figure 4.24: Enlarged Wavelet Decomposition for SR-222 data .................................................50
Figure 4.25: Greenville Data during First Week of April, 2010 ...................................................53
Figure 4.26: Three Level Wavelet Decomposition for Greenville Data .......................................54
Figure 4.27: Enlarged Three Level Wavelet Decomposition for Greenville Data .......................54

viii

LIST OF ABBREVIATIONS

Degrees Fahrenheit

BP

Barometric Pressure

BS

Base Station

CCIR

International Radio Consultative Committee

CWS

Columbia Weather Systems

dB

Decibel

DP

Dew Point

DFT

Discrete Fourier Transform

DWT

Discrete Wavelet Transform

GUI

Graphical User Interface

EDF

Empirical Distribution Function

EM

Electromagnetic

ESS

Environmental Sensor Station

FDOT

Florida Department of Transportation

FFT

Fast Fourier Transform

GHz

Gigahertz

HI

Heat Index

IEEE

Institute of Electrical and Electronics Engineers

ITS

Intelligent Transportation System

ITU

International Telecommunications Union

ITU-R

International Telecommunications Union Radio Communications

LOS

Line of Sight

QC

Quality Control

Precipitation

RF

Radio Frequency

RH

Relative Humidity

RSL

Received Signal Level

RWIS

Road Weather Information System


ix

RX

Receiver

SR-222

Gainesville Research Site

STFT

Short Time Fourier Transform

STN

Statewide Telecommunications Network

Temperature

TX

Transmitter

USDOT

United States Department of Transportation

WC

Wind Chill

WD

Wind Direction

WS

Wind Speed

WSA

Wind Speed Average

ABSTRACT

Understanding the effects of atmospheric conditions with respect to microwave propagation and
performance is critical to the design and placement of microwave antennas for modern
communication systems. Weather data acquisition in the state of Florida is underdeveloped and
the published effects of weather on microwave communications are limited to general models
based on large regional climate models. The goal of this research is to correlate atmospheric
conditions and microwave transmission via the existing Florida Department of Transportation
(FDOT) Road Weather Information System (RWIS) network, new Environmental Sensor Station
(ESS) sites, and Harris Corporation network management software Netboss. The microwave
radios in the FDOT microwave infrastructure through powerful Netboss scripting tools and
options are utilized to record the received signal level (RSL) output of the microwave radios for
signal analysis. This RSL data is analyzed and correlated with the acquired ESS weather data to
determine basic atmospheric effects on microwave propagation.
Methods for analysis of correlated data include existing atmospheric attenuation
models, such as the Global (Crane) and International Telecommunications Union (ITU) models,
and empirical methods such as the Fast Fourier Transform (FFT), Short Time Fourier Transform
(STFT), Discrete Wavelet Transform (DWT) and wavelet decomposition, and correlation
analysis of each method used. The data is treated as a discrete non-stationary signal. Results do
not show a clear correlation between receiver signal level (RSL) and weather parameters for
several of the test methods. Testing the correlation and cross correlation of the raw data yielded
weak correlation.

The simulation of rain attenuation via the ITU model displayed weak

insignificant results for the sets of RSL data. The FFT and STFT both incorporate too much
noise and distortion to accurately compute a correlation.
Wavelet decomposition shows a strong correlation between several weather
parameters and a weak correlation for others. This result confirms the wavelet decomposition
analysis and agrees with trends found in the RSL and weather parameters. Further analysis
points to multipath fading and atmospheric ducting.

During early hours of the morning,

reflections from moist surfaces, such as tree foliage and other terrestrial objects, water vapor and
dew will cause transmitted signals to reach the receive antenna out of phase, which will cause
xi

attenuation or gain while atmospheric ducting will cause gain in the RSL and is visible in the
acquired data. It is concluded that weather conditions such as water vapor, mist, and rising fog
have an effect on microwave propagation.

xii

CHAPTER 1
INTRODUCTION

1.1. Overview
As wireless systems become more complex, understanding the properties of electromagnetic
(EM) wave propagation becomes more important. Most wireless systems transmit data through
non-ideal environments which are filled with many attenuation factors and this creates a need for
modeling and understanding environmental effects on EM wave propagation. Typically there
are too many variables for a deterministic analysis, thus the need to employ statistical methods to
model channels are required to determine the mean or median effect likely to occur. The basics
of free-space propagation are consistent for all frequencies, but there are many variances in a
real-world communications link which is both sensitive and frequency dependent.

This

manuscript will focus on the microwave range (1 GHz to about 30 GHz) of radio frequencies
(RF), specifically in the 6.6 to 6.9 GHz range.

Figure 1.1: A Typical Communication System


The Florida Department of Transportation (FDOT) microwave telecommunications
infrastructure consists of 71 Line of Sight (LOS) microwave paths that cover most of Florida's
Interstate Highway System. This system provides voice and data communications for FDOT,
Florida Highway Patrol (FHP), and other public safety agencies. The microwave paths were
designed in accordance with accepted industry standards and techniques; however, particular
1

paths of the microwave system are experiencing more outages than the design anticipated. The
goal of this proposed project is to add new Environmental Sensor Stations (ESS) to the existing
FDOT Road Weather Information System (RWIS) and correlate the acquired weather data to
collected Received Signal Level (RSL) data to build a better understanding of atmospheric
effects on microwave transmission in the state of Florida at approximately 6.8 GHz. This
manuscript will provide a significant outlook on current attenuation modeling in the northern
region of the state of Florida due to environmental and atmospheric effects.
This project incorporates existing RWIS ESS sites via the FDOT Engineering and
Operations Office, Intelligent Transportation Systems (ITS) section, located in Tallahassee.
Columbia Weather Systems (CWS) Capricorn 2000TM data loggers and Weather Master 2000TM
software are used to collect and log atmospheric data, respectively. Three RWIS ESS sites and
six microwave sites will be utilized to gather crucial weather and microwave RSL data for
analysis. The FDOT microwave tower sites chosen for analysis are Greenville, Lake City DOT,
and Gainesville (interchange of SR-222 and I-75). See Figure 1.2 for chosen ESS sites in the
FDOT statewide microwave infrastructure deployment map.
The microwave RSL data is obtained via Netboss; a proprietary network management
software program developed by Harris Corporation that interfaces with the SCAN channel of the
FDOTs Harris DVM-6 Excel microwave radios in the FDOT microwave infrastructure. In
addition to many imbedded monitoring and maintenance features, Netboss also has powerful
scripting abilities and tools via a UNIX based VI editor. New scripts will be written in Netboss
to utilize the state of Floridas existing RWIS ESS sites to gather microwave RSL data for
analysis. Methods and models for the analysis of acquired data include Global (Crane) model,
Initial and Revised Two-Component model, International Telecommunications Union (ITU) rain
region model, and Fast Fourier Transform (FFT), Short Time Fourier Transform (STFT),
Discrete Wavelet Transform (DWT), and wavelet decomposition. The project work is conducted
with RCC Consultants, Inc. and the FDOT for access to the FDOT microwave communication
infrastructure, shelter sites, and the Traffic Engineering Research Lab (TERL) weather data
server and data loggers.

Figure 1.2: FDOT Statewide Telecommunications Network Deployment Map


This project involves a number of different sensor types, mountings, locations across the
state of Florida, data interpretation and correlation, considered analysis methods, and includes
many communication protocols for data acquisition and performance comparison purposes. In
addition to better understanding microwave transmission attenuation and performance, an added
benefit of the proposed project is that it also provides invaluable weather data to the United
States Department of Transportation (USDOT) Clarus initiative; a national weather data
acquisition initiative.
3

1.2. Motivation
Many research efforts have been devoted to modeling path loss propagation attenuation due to
atmospheric effects, specifically rain, water vapor, and fog, on microwave links by using
different methods ranging from analytical models and semi-empirical models, to observation
measurements. Most radio signal propagation models are developed using empirical methods,
based on fitting mathematical models to measured data. In recent years, few measurement-based
point rain rate attenuation models have been proposed and investigated. Leading models for path
loss attenuation due to atmospheric effects have been proposed by Robert K. Crane and the ITU
[1]-[5] and are in use in several path loss analysis programs by renowned RF manufacturers,
consulting firms, and engineering practices. These research works were primarily focused on
particular regions and a general model was developed and deployed for areas that do not produce
significant data.
Given the numerous weather conditions, and the lack of real-world observation modeling
in the state of Florida, it is desirable to correlate observations of Floridas atmospheric conditions
to the RSL of FDOTs statewide telecommunications network to better understand the impact
weather has on microwave transmission.

1.3. Problem Statement


There are many techniques and methods used to develop attenuation models which are later used
in path loss models. The Global (Crane) model and the ITU model are the most commonly used
models to calculate attenuation due to major atmospheric effects; mainly rain with some
discussions regarding water vapor and fog modeling on a terrestrial path link. Traditional
techniques for estimating losses due to atmospheric effects focus on the dominant source of
fading - rain attenuation. The focus of this project is the study of several atmospheric attributes
and their effect on microwave received signal levels, not on rain attenuation alone.

The

hypothesis of this research is that various atmospheric conditions such as relative humidity,
temperature, wind, and rain will have an impact on microwave transmission.

1.4. Scope of Work


The organization of this manuscript is as follows: Chapter 2 presents current attenuation path
loss models, focusing specifically on Robert K. Cranes volume cell and debris attenuation
models; Chapter 3 introduces the ITU rain attenuation model and provides some other commonly
used models based on observations, frequencies, regions, and estimations. The analyzed data
using selected models and observed data along with theoretical path loss models will be provided
in Chapter 4 which also includes comparisons with the empirical models and key findings from
correlated results. Finally, Chapter 5 provides a conclusion and recommendations for future
work.

CHAPTER 2
CRANE ATTENUATION MODELS

Different attenuation models are studied and used as a comparison method for the acquired data.
In this chapter the Global (Crane) Model, Initial Two-Component Model, and Revised TwoComponent Model are discussed in detail. Their relationship to the goal of this manuscript will
be discussed in Chapter 4.

2.1. Global (Crane) Model


The Global (Crane) Model was developed by Robert K. Crane (1980), a pioneer in rain
attenuation modeling, for use in Earth-space or terrestrial links. The Global model is based
entirely on geophysical observations of rain rate, rain structure, and the vertical variation of
atmospheric temperature.

None of the model constants are obtained from attenuation

measurements [2]. A statistical model is required to provide an accurate estimate of attenuation


due to rain being characteristically inhomogeneous on the horizontal plane. In Cranes model
the horizontal structure of rain is not dependent on the climate region. This is due to the fact that
the fluid dynamics parameters that are used to characterize flow are weakly dependent on
climate. This model uses the multiplier coefficient (k) and exponent () of the power-law
equation (2.1) for the approximation of spherical drops at an assumed temperature of 32 F and
the dielectric constant model for specific frequencies ranging from 1 to 1000 GHz. See Figures
2.1 and 2.2 for multiplier and exponent plots.
(2.1)
The polarization state of an antenna has little effect in determining the prediction of
attenuation along a terrestrial link, either experimentally observed or calculated using the k
multiplier and exponent plots.

Figure 2.1: Multiplier

in the Power-Law Relationship between Specific Attenuation and Rain

Rate. (Figure 4.3 from Ref. 2, courtesy of Wiley.)


The simplest path profile for attenuation due to rain rate is shown in equations 2.2 and
2.3. When this equation integrated it produces the observed median power law relationship,
which is the derivative of the power law relationship with respect to path length.
0

(2.2)
22.5

(2.3)

where
horizontal path attenuation (dB)
rain rate (mm/h)
path length (km)
specific attenuation, =

(dB/km)

and the remaining coefficients are the empirical constants of the piecewise exponential model:
ln

0.83

0.026

0.03

0.17ln
km
7

3.8

0.6

km
km
km
km

Cranes model provides a prediction for attenuation along a terrestrial Line of Sight (LOS) link
for the path-integrated rain rate given equiprobable value of rain rate.

Figure 2.2: Exponent

in the Power-Law Relationship between Specific Attenuation and Rain

Rate. (Figure 4.4 from Ref. 2, courtesy of Wiley.)


The Global Model employs data sets for various probabilities and availabilities that differ
from the ITU model, discussed later in section 3.1, and are only valid for distances up to 22.5
km. The Global Model does not employ an availability adjustment factor like the ITU model. If
the desired availability is not represented in the Crane data, it is possible to logarithmically
interpolate the given data to estimate the rain rate [1]. This method has been tested to provide
reasonable information, but is not sanctioned by Crane.

2.2. Initial Two-Component Model


The Two-Component Model for attenuation due to rainfall was initially based on the observation
of volume cells and debris, and an ad hoc procedure. These observations account for the spatial
correlations for each component and was eventually revised to account for vertical rainfall as
well as rainfall along a horizontal path. This model requires parameters for the two-component
rain rate distribution model and is therefore more complex if the global rain rate climate model is
not invoked, and thus the first step in the consideration of the more complex modeling problems
and the only step allowing for comparison with a significant body of measurements [2].
The Two-Component Model accounts for the contributions of heavy rain showers and
lighter intensity rain showers occurring in larger regions. RF propagation does not always
intersect a single cell or debris, or both along a LOS link; thus the model accounts for volume
cells and debris independently. The Two-Component Model assumes either a single volume
cell, only debris, or both, along a LOS link. This design is in place so a desired attenuation
threshold is not exceeded. The probability for each component, a volume cell of rain or debris, is
calculated and the results are summed for the total desired probability estimate.

2.2.1. Volume Cell Contribution


In this model the path-integrated, or terrestrial, rain-rate is given by
(2.4)
where
observed path-integrated value (km mm/h)
rain-rate profile along path (mm/h)
length of path (km)
The volume cell contribution for the path-integrated rain rate is approximated by
(2.5)
where
peak rain rate in volume cell
average dimension of volume cell with area
(see figs. 2.3 and 2.4) with

1.70 and

and rain rate,


0.002

adjustment factor required by definition of volume cell


9

The adjustment factor is included to represent the contribution of the debris close to a cell
but outside the region enclosing the -3 dB reflectivity value relative to the peak value (factor of
0.61 in rain rate) [2]. This factor is utilized when the path is much longer than the average
volume cell width. If the path is shorter than WC, the actual path length should be used for the
calculation of I and the adjustment factor should be unity.

Figure 2.3: Edfs for the Joint Occurrence of Reflectivity and Square Root Area. Data from
Kansas HIPLEX [Crane and Hardy, 1981]. (Figure 2.30 from Ref. 2, courtesy of Wiley.)
The path-integrated rain rate, RV, within a volume cell is modeled by
.

(2.6)

and
.

10

(2.7)

Figure 2.4: Average Area of Volume Cells as Measured and Modeled Using an Exponential
Square Root Area Model. Data from Kansas HIPLEX [Crane and Hardy, 1981]. (Figure 2.32
from Ref. 2, courtesy of Wiley.)
Equation 2.7 is the starting point in the particular application of the model where
is given and

and

are to be determined. The average dimension of volume cell,

, is

modeled by
(2.8)
Taking

min

yields
.

or

11

and

(2.9)

(2.10)

The initial two-component model is simplified by the assumption that all volume cells
have the same cross-sectional area. The area of influence of the volume cell about a point is
and the area of influence of a circular volume cell about a line of length

is

(2.11)

where is the average length of a line through a circular volume cell and given by
1
2
Crane approximates by

0.9

since both the area and shape of the cell are uncertain, where

Assuming only one volume cell can occur at random anywhere along the path, affect the
LOS link at any instance of time, and the random volume cell spatial distribution is uniformly
distributed, the probability of occurrence of the rain rate value for the center of a single volume
cell is given by

[2]. The probability of exceeding the specified occurrence of rain

rate for the center of a single volume cell is given by

(2.12)

2.2.2. Debris Contribution


To effectively calculate the debris contribution on a terrestrial path link, the spatial scale

has

to be associated with the rain within the debris. Crane and Hardy (1981) provided data on the
relationship between average rain rate and area for isolated echo areas. This data is used to
12

create a relationship between spatial scale

and the average rain rate within the debris. The

result is a regression line fit for the relationship area versus rain rate.
.

882
where

(km2)

(2.13)

(km)

(2.14)

is the debris area


.

29.7
1

The physical path length D or the debris scale length

, whichever results is the

smallest, is used in the calculation for a specified path integrated rain rate. For a long path,
.

Thus,
.
.

and
.

29.7

170

(km2)

(2.15)

For a path of length D,

min

(2.16)

29.7

(2.17)

(2.18)

2.2.3. Probability of Terrestrial Rain Rate


The Two-Component Model scaling parameters

and

are assumed to apply in all climate

regions due to the similarity in scale of the dynamic processes responsible for precipitation. The
probability for path integrated rain rate I is
(2.19)
The model cannot be used directly if the interest of probability is known and the value of I is
estimated. The values of probability must be calculated for a number of trial I values [2] then
interpolate or iteratively adjust the trial value of I until the interest of probability is estimated.

13

2.2.4. Attenuation Along a LOS Path


The attenuation along a LOS path is given by (2.20) and attenuation within a volume cell is
approximated by (2.21).
(2.20)
(2.21)
The adjustment factor to estimate additional attenuation outside a volume cell is
0.7

(2.22)

For a Gaussian volume cell profile, the errors in calculating attenuation caused by
assuming the

verses

3.5% for

relationship in equation 2.22 are

1.3 and

4.5% for

0.75 [2]. Thus for frequencies between 1 GHz and 100 GHz the error is less than 5% for
entire range of

(assuming Gaussian cells).

The two-component model estimates the rain rate in a volume cell and debris region and
calculates the probability of exceeding a certain threshold or attenuation value.
For a volume cell,

min

(2.24)

(2.23)

(2.25)

(2.26)

(2.27)

Neglecting the effect of the nonlinearity on the relationship between specific attenuation and the
average rain rate within a debris region yields

.
.

29.7

(mm/h)

(2.28)
(2.29)

Then,

(2.31)

29.7
14

(2.30)

min

(2.32)

(2.33)

The probability that the attenuation value a is exceeded is given by


(2.34)

2.3. Revised Two Component Model


The Revised Two-Component Model (R. K. Crane and H. C. Shieh; 1989) is an extension and
refinement of the initial model by Robert K. Crane. The refinements include a more realistic
treatment of the statistical variations and spatial correlations of rain within the cell and debris
components of the initial model [2]. The revised model has similar derivations as the initial twocomponent model, hence all intermediate steps and equations for the volume cell and debris
components will be omitted with the exception of the final attenuation and probability equations.

2.3.1. Model for Volume Cell Component


Rain cells often cause severe attenuation to transmitted signals over short time intervals. The
Revised Two-Component Model assumes constant specific attenuation with height and only
considers reduced attenuation on horizontal path links. The model also assumes that a spatial
rain rate profile along a horizontal line through a rain cell has a Gaussian distribution and the
occurrence for probability density for a rain cell is uniform. Thus, the attenuation is
2

(2.35)

and the probability of exceeding a specific attenuation is define as


A

, ,

(2.36)

2.3.2. Model for Debris Component


The probability density function for the debris component of the mixed rain rate process is
assumed to be jointly lognormal with the spatial correlation function for the variations in the
logarithm of the rain rate derived from radar observations [2].

ln

The final probability of exceeding a specified attenuation is the sum of

15

(2.37)

and

CHAPTER 3
ITU ATTENUATION MODEL

Different attenuation models were studied and used in a comparison method for the acquired
data. This chapter discusses, the International Telecommunications Union Model in detail along
with other attenuation models. The relationship of the ITU Model to the goal of this manuscript
will be discussed in chapter 4.

3.1. International Telecommunications Union Model


Nearly 100 years after the ITU was formed in 1865, several ITU members began focusing on
research and development of rain attenuation models and the effects the environment and
atmosphere have on RF propagation links. Similar to Cranes work, the ITU developed a global
rain model that incorporates rain region factors based on acquired meteorological data. The ITU
Model for a given availability on a horizontal or nearly horizontal communications link is to
determine the 99.999% fade depth [1]. Different fade depths are available and shown in Table
3.3. The five-nines data has lower confidence than four-nines data due to a smaller database,
however, the five-nines data will be viewed for this project, as five-nines is the industry standard
for public safety in Florida for LOS link reliability.

Atten0.001

(dB)

(3.1)

where
is the 99.999% rain rate for the rain region, in mm/h

is the specific attenuation in dB/km

is the link distance in km


and the distance factor r
1/ 1

(3.2)

with the effective path length


0

35

16

(km)

(3.3)

The specific attenuation is calculated by using the defined 99.99% rain rate region of the
corresponding region of interest. The ITU rain rate data for 0.001%, or five-nines, rain fades in
the Americas is shown in Table 3.1. The regression coefficients,

and , for frequencies 1-30

GHz and horizontal polarization are listed in Table 3.2. Rain rates based on geographical
regions are the most widely used and easily applied method for determining the rain rate [1].
Table 3.1: ITU rain rate data for 0.001% rain fades in the Americas
A
B
C
D
E
F
G
H
22
32
42
42
70
78
65
83
Source: Table 1 from Ref. 5, courtesy of the ITU.

J
55

K
100

L
150

M
120

N
180

P
250

Figure 3.1 shows specific attenuation of frequencies ranging from 1 GHz to 100 GHz due to
water vapor, dry air, and the sum of water vapor and dry air. Major specific attenuation is
apparent at 22.5 GHz and 60 GHz frequencies.

Figure 3.1: ITU Atmospheric Attenuation Prediction


17

The ITU model factors to model rain attenuation are not linear with distance, thus simply
multiplying the specific attenuation with distance will not calculate the correct estimate of the
attenuation over the LOS link. The ITU model is validated for frequencies up to at least 40 GHz
and distances up to 60 km [6]. The desired probability 100Availability expressed as a
percentage for latitudes greater than 30 degrees, North or South,
Atten/Atten0.001

0.12

0.546

0.043

(3.4)

0.07

0.855

0.139

(3.5)

and less than 30 degrees, North or South,


Atten/Atten0.001

ITU rain regions for the Americas, Europe and Africa, and Asia are shown in Figure 3.2, 3.3, and
3.4, respectively.

Figure 3.2: ITU Rain Regions for the Americas. (Figure 1 from Ref. 5, courtesy of the ITU.)

18

Figure 3.3: ITU Rain Regions for Europe and Africa. (Figure 2 from Ref. 5, courtesy of ITU.)

Figure 3.4: ITU Rain Regions for Asia. (Figure 3 from Ref. 5, courtesy of the ITU.)
19

Table 3.2: Interpolated Regression Coefficients for 1-30 GHz


f(GHz)
H
H
V
1
3.87 10
0.912
3.52 10
2
1.54 10
0.963
1.38 10
3
3.576 10
1.055
3.232 10
4
6.5 10
1.121
5.91 10
5
1.121 10
1.224
1.005 10
6
1.75 10
1.308
1.55 10
7
3.01 10
1.332
2.65 10
8
4.54 10
1.327
3.95 10
9
6.924 10
1.3
6.054 10
10
0.01
1.276
8.87E-3
11
0.014
1.245
0.012
12
0.019
1.217
0.017
13
0.024
1.194
0.022
14
0.03
1.173
0.027
15
0.037
1.154
0.034
16
0.043
1.142
0.039
17
0.05
1.13
0.046
18
0.058
1.119
0.053
19
0.066
1.109
0.061
20
0.075
1.099
0.069
21
0.084
1.091
0.077
22
0.093
1.083
0.085
23
0.103
1.075
0.094
24
0.113
1.068
0.103
25
0.124
1.061
0.113
26
0.135
1.052
0.123
27
0.147
1.044
0.133
28
0.16
1.036
0.144
29
0.173
1.028
0.155
30
0.187
1.021
0.167
Source: Table 10A.2 from Ref. 1, courtesy of John S. Seybold.

20

0.88
0.923
1.012
1.075
1.18
1.265
1.312
1.31
1.286
1.264
1.231
1.2
1.174
1.15
1.128
1.114
1.101
1.088
1.076
1.065
1.057
1.05
1.043
1.036
1.03
1.024
1.017
1.011
1.0068
1

Table 3.3: ITU Rainfall Rates for Different Probabilities and Rain Regions

Percentage
of Time (%)
1.0
0.3
0.1
0.03
0.01
0.003
0.001
Percentage
of Time (%)
1.0
0.3
0.1
0.03
0.01
0.003
0.001

A
<0.1
0.8
2
5
8
14
22

B
0.5
2
3
6
12
21
32

C
0.7
2.8
5
9
15
26
42

D
2.1
45
8
13
19
29
42

E
0.6
2.4
6
12
22
41
70

F
1.7
4.5
8
15
28
54
78

G
3
7
12
20
30
45
65

J
8
13
20
28
35
45
55

K
1.5
4.2
12
23
42
70
100

L
2
7
15
33
60
105
150

M
4
11
22
40
63
95
120

N
5
15
35
65
95
140
180

P
12
34
65
105
145
200
250

Q
24
49
72
96
115
142
170

H
2
4
10
18
32
55
83

Rainfall intensity exceeded (mm/h).

Source: Table 1 from Ref. 4, courtesy of the ITU.

3.2. Other Attenuation Models


Many other estimation and empirical attenuation models, due to rain and other atmospheric
effects, are available for RF attenuation; some similar to the ITU model, but provide different
methods for determining rain statistics. Some attenuation models are optimized for specific
applications or can only be adapted to specific geographical regions and could not be used in this
manuscript. Lin (1977) and succession of International Radio Consultative Committee (CCIR;
now International Telecommunications Union - Radio Communications or ITU-R) models have
merged to develop path-averaging models using regression analysis between simultaneous
attenuation and rain rate observations. Morita and Higuti (1978) assumed gamma distribution
and observed spatial correlation properties between point rain gauge observations to model the
rain rate distribution, but Crane argues the gamma distribution does not adequately represent the
observed dependence of effective path averaged factor on point rain rate at rain rates below 20
mm/h [2].

21

CHAPTER 4
COMPUTER SIMULATION RESULTS AND KEY FINDINGS

A variety of software is used to compile and process all acquired data. This chapter describes
software utilized in this project, specifically Weather Master 2000TM, MATLAB R2007b,
Netboss, and Microsoft Excel, and incorporates discussions of various methods of analysis. This
chapter also displays tables and figures with explanations, arguments, and supporting evidence
for each method used.

4.1. Data Acquisition


An array of software is utilized to acquire data from each site and store it in a format that can be
further processed. The Capricorn 2000TM weather station control module is a programmable
microprocessor with abundant on-board memory. The Capricorn 2000 Weather Display can
display weather information, perform complex computations, and store relatively large amounts
of weather data [10]. It incorporates a built-in circular data logger which can hold up to 511
records of sensor readings (samples) and High/Low information. The data logger can output
stored data in a comma-delimited text file as shown in Figure 4.1.
The Capricorn 2000TM control module at each site communicates with a proprietary
software, Weather Master 2000TM, on the FDOT ESS server located at the TERL in Tallahassee,
FL. The Weather Master 2000TM software has a graphical user interface (GUI) and incorporates
many weather statistics as shown in Figure 4.2, but the software was not reliable due to data
recording failures. This inconsistency created holes in the acquired data records and posed a
major problem for this project. The software bug was fixed after a series of updates and patches
provided by the manufacturer, and the missing data was filled by interpolation. This did not
solve the issue completely as some holes in the data were so large that interpolation could not
accurately convey the missing data. In this case data from external sources is used. Archived
weather data from www.weather.com and www.wunderground.com are used to assist in filling
some of the larger sections of missing data. Many MATLAB scripts were written to scan the
22

acquired data and interpolate any missing data. The MATLAB scripts are available in Appendix
A.
Date,Time,"Wind Speed","Wind Speed-Avg","Wind Direction","Rain Today","Relative Humidity","Adjusted Barometric
Pressure",Temperature1,"Wind Chill","Heat Index","Dew Point"
02/11/10,17:39,0,0,202.5,0,53,29.8500003814697,44.4799995422363,44.4799995422363,44.4799995422363,28.4268856048584
02/11/10,17:40,0,0,202.5,0,53,29.8500003814697,44.4799995422363,44.4799995422363,44.4799995422363,28.4268856048584
02/11/10,17:41,0,0,202.5,0,53,29.8500003814697,44.3699989318848,44.3699989318848,44.3699989318848,28.3247604370117

Figure 4.1: Sample comma-delimited text file from the control module at Greenville ESS site
The RSL data is obtained via Netboss. Netboss is a SCAN channel of the UNIX based
software, developed by Harris Communications, to interface with the Harris DVM6-45
microwave radios in FDOT shelters. Netboss has many scripting options using the VI Editor and
imbedded commands. Several scripts have been written to collect the RSL data from each site
and are available in Appendix A. See Figure 4.3 for a Netboss example.

Figure 4.2: Weather Master 2000TM Example


23

Figure 4.3: Netboss Example

4.2. Crane Models, ITU Model, and Path Loss 4.0 Analysis
Some models used for attenuation calculations and predictions were researched prior to data
acquisition, and are examined with the data to determine their reliability in the state of Florida.

4.2.1. Analysis of Data Using Crane Models


The Greenville and Monticello, Lake City DOT and US-41, and SR-222 and US-41 signal paths
chosen for research are 24.38 km, 22.27 km, and 37.59 km in length, respectively, and the
Global (Crane) Model, Initial Two-Component Model, and Revised Two-Component Model are
valid for distances up to 22.5 km. The most reliable site, in terms of working weather sensors, is
Greenville, and most analysis methods in this manuscript are computed using data from the
Greenville site. Due to the restriction of distance and the lack of accurate weather data, no
further analysis of data using Cranes models is computed.

24

4.2.2. International Telecommunications Union Model Analysis


The Greenville and Lake City DOT site data is analyzed using the ITU model. Given the
frequency of 6.835 GHz and a horizontal antenna polarization, the calculated linear regression
coefficients,

and , are 0.0028 and 1.3280, respectively. The linear regression coefficient

values are linearly interpolated using MATLAB. The program code is located in Appendix A.
The Greenville rain data is converted from inches per hour (in/h) to millimeters per hour (mm/h)
and the predicted rain attenuation is calculated for Greenville using the recorded mm/h rain rate.
The predicted rain attenuation is displayed in Figure 4.4 and Figure 4.5. The minimum and
maximum attenuation due to rain are 0 dB and 0.1549 dB, respectively. This very small amount
of attenuation has little effect on the received signal, and the RSL displays periodic attenuation
patterns that vary in amplitude much greater than the calculated rain attenuation. Research
points to other weather parameters causing the major attenuation cycles discussed in later
sections in this chapter.

Figure 4.4: ITU Model Rain Attenuation Prediction for Greenville Site

25

Figure 4.5: ITU Model Rain Attenuation Prediction for Lake City DOT Site
The code for the ITU model and regression coefficient interpolation can be found in Appendix A
of this manuscript.

4.2.3. Path Loss 4.0 Analysis


Path Loss 4.0 is used by the FDOT to determine the reliability of a communications link in the
Statewide Telecommunications Network (STN). The FDOT requires five-nines of reliability for
the STN. Tables 4.1 through 4.3 contain summary data from Path Loss 4.0. The reliability
method for analysis is the Vigants-Barnett method and the selected rain attenuation model is the
ITU-R P530-7. The ITU-R P530-7 is the full model name for the ITU model discussed in
Chapter 3. Figures 4.6 through 4.8 display a print profile of the sites that were analyzed. This
profile contains information regarding the antenna height, distance between sites, terrain layout,
and much more data that give engineers and designers a clear view of the current or future
site/system under analysis.

26

4.2.3.1. Greenville Analysis


The Greenville site is located one mile west of Greenville, FL on the Interstate 10 westbound
route. The majority of the terrestrial path between the Greenville and Monticello DOT sites is
populated with 60 ft trees, shown in green in Figure 4.7. There are some buildings located along
the path link, but their heights are only a fraction of that of the trees and thus can be ignored.
This, however, does not interfere with the LOS link due to the antenna heights; the first Fresnel
Zone is not breached. The LOS link is displayed in red and the bottom half of the first Fresnel
Zone is displayed in blue. The Path Loss 4.0 print summary, shown in Table 4.1, contains
information about the microwave radio used in this project among other site data. The FDOT
requires five-nines of reliability for the STN and based on the given criteria Path Loss 4.0
calculated Greenvilles annual multipath plus rain (%-sec) of 99.99432 and 1791.07 in
percentage and seconds, respectively. This is below FDOT standards and has been reported to
FDOT ITS engineers.

Figure 4.6: Path Loss 4.0 Path Profile for Greenville

27

Table 4.1: Path Loss 4.0 Print Summary for Greenville


Elevation (ft)
Latitude
Longitude
True azimuth ()
Antenna model
Antenna height (ft)
Antenna gain (dBi)
Radome loss (dB)
TX line type
TX line length (ft)
TX line unit loss (dB /100 ft)
TX line loss (dB)
Connector loss (dB)
Circ. branching loss (dB)
Other TX loss (dB)
Other RX loss (dB)
Frequency (MHz)
Polarization
Path length (mi)
Free space loss (dB)
Atmospheric absorption loss (dB)
Field margin (dB)
Net path loss (dB)
Radio model
TX power (watts)
TX power (dBm)
EIRP (dBm)
RX threshold criteria
RX threshold level (dBm)
RX signal (dBm)
Thermal fade margin (dB)
Climatic factor
C factor
Fade occurrence factor (Po)
Average annual temperature (F)
0.01% rain rate (mm/hr)
Flat fade margin - rain (dB)
Rain attenuation (dB)
Annual multipath + rain (%-sec)

Greenville
88.58
30 26 08.30 N
083 38 20.50 W
295.3
PA8-65D
187
42.3
0.6
E65 RFS
220
1.37
3.01
0.2
1.4

0.5
1.5
6835
Horizontal
15.15
136.65
0.22
1
64.67
DVM6 Excell
0.79
29
65.59
46.681 Mbps
-74.9
-35.67
39.23

63.67
DVM6 Excell
0.79
29
66.62
46.681 Mbps
-74.95
-34.67
40.28
2
6
3.46E-01
72
98
39.23
39.23
99.99432 - 1791.07

28

Monticello DOT
190.95
30 31 45.93 N
083 52 07.45 W
115.18
PA8-65D
123
42.3
0.6
E65 RFS
174
1.37
2.38
0.2
1.5

4.2.3.2. Lake City DOT Analysis


The Lake City DOT site is located at the Lake City DOT office complex in Lake City, FL. The
majority of the terrestrial path between the Lake City Dot and US-41 sites is populated with 60 ft
trees, shown in green in Figure 4.7. There are some buildings located along the path link, but
their heights are only a fraction of that of the trees and thus can be ignored. The tree line and
building heights do not interfere with the LOS link due to the antenna heights; the first Fresnel
Zone is not breached. The LOS link is displayed in red and the first Fresnel Zone is displayed in
blue. The Path Loss 4.0 print summary, shown in Table 4.2, contains information about the
microwave radio used in this project as well as other site data. The FDOT requires five-nines of
reliability for the STN and based on the given criteria Path Loss 4.0 calculated Lake City DOTs
annual multipath plus rain (%-sec) of 99.99789 and 664.49 in percentage and seconds,
respectively. This does not meet the five-nines FDOT standard, but FDOT ITS engineers state
that eleven minutes of annual downtime is not significant and can be ignored as other routing and
redundancy mechanisms are in place to keep the link active with such a small projected
downtime.

Figure 4.7: Path Loss 4.0 Path Profile for Lake City DOT

29

Table 4.2: Path Loss 4.0 Print Summary for Lake City DOT
Elevation (ft)
Latitude
Longitude
True azimuth ()
Antenna model
Antenna height (ft)
Antenna gain (dBi)
Radome loss (dB)
TX line type
TX line length (ft)
TX line unit loss (dB /100 ft)
TX line loss (dB)
Connector loss (dB)
Circ. branching loss (dB)
Other TX loss (dB)
RX filter loss (dB)
Frequency (MHz)
Polarization
Path length (mi)
Free space loss (dB)
Atmospheric absorption loss (dB)
Field margin (dB)
Net path loss (dB)
Radio model
TX power (watts)
TX power (dBm)
EIRP (dBm)
RX threshold criteria
RX threshold level (dBm)
RX signal (dBm)
Thermal fade margin (dB)
Climatic factor
C factor
Fade occurrence factor (Po)
Average annual temperature (F)
0.01% rain rate (mm/hr)
Flat fade margin - rain (dB)
Rain attenuation (dB)
Annual multipath + rain (%-sec)

Lake City
159.89
30 11 42.00 N
082 39 11.00 W
166.29
PA8-65D
186
42.3
0.6
E65 RFS
186
1.37
2.55
0.2
1.4

0.5
1.5
6855
Horizontal
13.84
135.94
0.2
1
64.24
DVM6 Excell
0.79
29
66.05
46.681 Mbps
-74
-35.24
38.76

63.24
DVM6 Excell
0.79
29
65.85
46.681 Mbps
-74.9
-34.24
40.66
2
6
2.67E-01
72
98
38.76
38.76
99.99789 - 664.49

30

US 41
86.09
29 59 59.00 N
082 35 54.00 W
346.32
PA8-65D
230
42.3
0.6
E65 RFS
230
1.37
3.15
0.2
1.5

4.2.3.3. SR-222 Analysis


The SR-222 site is located along Interstate 75 at the Exit 390 interchange, outside the
southbound on ramp in Gainesville, FL. The majority of the terrestrial path between the SR-222
and US-41 sites is populated with 60 ft trees, shown in green in Figure 4.8. There are some
buildings located along the path link, but their heights are only a fraction of that of the trees and
thus can be ignored.

The rest of the path link is filled with farmland and is treated as open land

in Path Loss 4.0. The tree line and farmland do not interfere with the LOS link due to the
antenna heights; the first Fresnel Zone is not breached. The LOS link is displayed in red and the
bottom half of the first Fresnel Zone in blue. The Path Loss 4.0 print summary, as shown in
Table 4.3, contains information about the microwave radio used in this project among as well as
site data. The FDOT requires five-nines of reliability for the STN. Based on the given criteria
Path Loss 4.0 calculated SR-222s annual multipath plus rain (%-sec) of 99.98588 and 4453.21
in percentage and seconds, respectively. This is below FDOT standards and has been reported to
FDOT ITS engineers.

Figure 4.8: Path Loss 4.0 Path Profile for SR-222

31

Table 4.3: Path Loss 4.0 Print Summary for SR-222


Elevation (ft)
Latitude
Longitude
True azimuth ()
Antenna model
Antenna height (ft)
Antenna gain (dBi)
Radome loss (dB)
TX line type
TX line length (ft)
TX line unit loss (dB /100 ft)
TX line loss (dB)
Connector loss (dB)
Circ. branching loss (dB)
Other TX loss (dB)
RX filter loss (dB)
Frequency (MHz)
Polarization
Path length (mi)
Free space loss (dB)
Atmospheric absorption loss (dB)
Field margin (dB)
Net path loss (dB)
Radio model
TX power (watts)
TX power (dBm)
EIRP (dBm)
RX threshold criteria
RX threshold level (dBm)
RX signal (dBm)
Thermal fade margin (dB)
Climatic factor
C factor
Fade occurrence factor (Po)
Average annual temperature (F)
0.01% rain rate (mm/hr)
Flat fade margin - rain (dB)
Rain attenuation (dB)
Annual multipath + rain (%-sec)

SR-222
121.5
29 41 15.52 N
082 26 45.85 W
337
PA8-65D
221
42.3
0.6
E65 FRS
221
1.37
3.03
0.2
1.4

0.5
1.5
6815
Horizontal
23.36
140.43
0.34
1
65.77
DVM6 Excell
0.79
29
67.47
46.681 Mbps
-74.9
-36.77
38.13

65.77
DVM6 Excell
0.79
29
66.53
46.681 Mbps
-74.9
-36.77
38.13
2
6
1.27E+00
72
98
38.13
38.13
99.98588 - 4453.21

32

US 41
86.09
29 59 59.00 N
082 35 54.00 W
156.92
PA8-65D
290
42.3
0.6
E65 FRS
290
1.37
3.97
0.2
1.5

4. 3. Correlation Analysis without Data Preprocessing


The correlation coefficients of the data for each site were calculated and are shown in Tables 4.4
through 4.7. The correlation coefficients of the RSL and other weather parameters such as wind
speed, relative humidity, temperature, precipitation, etc. for each site are very weak which
indicates that there is not a direct correlation between the RSL and weather parameters, and that
they are independent of each other. This does not hold true in observations and other studies. A
timing delay errors or non-synchronized timing errors may be the cause of the low correlation
values; a result from variations of antenna heights and sensor locations or preprocessing of the
data may be needed.
Table 4.4: Correlation Coefficients for Greenville
RSL
WS
WSA
WD
P
RH
BP
T
WC
HI
DP

RSL

WS

WSA

WD

RH

BP

WC

HI

DP

1
0.012
0.005
0.031
-0.054
-0.065
0.032
0.001
-0.004
-0.017
-0.068

0.012
1
0.202
-0.005
-0.023
-0.394
0.035
0.242
0.249
0.242
-0.110

0.005
0.202
1
-0.013
0.042
-0.122
0.088
-0.071
-0.053
-0.052
-0.199

0.031
-0.005
-0.013
1
-0.026
0.002
0.161
-0.061
-0.062
-0.083
-0.074

-0.054
-0.023
0.042
-0.026
1
0.155
-0.170
-0.063
-0.064
-0.088
0.078

-0.065
-0.394
-0.122
0.002
0.155
1
-0.081
-0.666
-0.659
-0.634
0.238

0.032
0.035
0.088
0.161
-0.170
-0.081
1
-0.077
-0.078
-0.071
-0.172

0.001
0.242
-0.071
-0.061
-0.063
-0.666
-0.077
1
0.988
0.961
0.548

-0.004
0.249
-0.053
-0.062
-0.064
-0.659
-0.078
0.988
1
0.972
0.553

-0.017
0.242
-0.052
-0.083
-0.088
-0.634
-0.071
0.961
0.972
1
0.560

-0.068
-0.110
-0.199
-0.074
0.078
0.238
-0.172
0.548
0.553
0.560
1

Table 4.5: Correlation Coefficients for Lake City DOT


RSL
WS
WSA
WD
P
RH
BP
T
WC
HI
DP

RSL

WS

WSA

WD

RH

BP

WC

HI

DP

1
0.005
0.008
0.058
-0.059
-0.086
0.052
0.066
0.079
0.085
-0.015

0.005
1
0.995
0.076
-0.045
-0.286
0.532
0.018
-0.036
0.013
-0.239

0.008
0.995
1
0.077
-0.046
-0.293
0.543
0.018
-0.036
0.014
-0.245

0.058
0.076
0.077
1
0.009
-0.227
0.069
0.191
0.236
0.227
-0.049

-0.059
-0.045
-0.046
0.009
1
0.186
-0.133
-0.060
-0.084
-0.102
0.102

-0.086
-0.286
-0.293
-0.227
0.186
1
-0.562
-0.295
-0.482
-0.407
0.648

0.052
0.532
0.543
0.069
-0.133
-0.562
1
-0.200
-0.010
-0.006
-0.654

0.066
0.018
0.018
0.191
-0.060
-0.295
-0.200
1
0.930
0.857
0.525

0.079
-0.036
-0.036
0.236
-0.084
-0.482
-0.010
0.930
1
0.923
0.307

0.085
0.013
0.014
0.227
-0.102
-0.407
-0.006
0.857
0.923
1
0.343

-0.015
-0.239
-0.245
-0.049
0.102
0.648
-0.654
0.525
0.307
0.343
1

33

Table 4.6: Correlation Coefficients for SR-222


RSL
WS
WSA
WD
P
RH
BP
T
WC
HI
DP

RSL

WS

WSA

WD

RH

BP

WC

HI

DP

1
0.099
0.077
-0.052
-0.090
0.045
0.057
0.044
0.065
0.046
0.059

0.099
1
0.580
-0.028
-0.076
0.130
0.146
-0.018
0.024
0.051
0.187

0.077
0.580
1
-0.057
-0.047
-0.150
0.155
-0.138
-0.160
-0.142
-0.153

-0.052
-0.028
-0.057
1
0.048
0.043
-0.240
0.028
0.039
0.054
0.047

-0.090
-0.076
-0.047
0.048
1
-0.091
-0.140
0.001
-0.037
-0.021
-0.097

0.045
0.130
-0.150
0.043
-0.091
1
0.263
-0.241
0.107
-0.073
0.958

0.057
0.146
0.155
-0.240
-0.140
0.263
1
-0.085
0.056
-0.005
0.268

0.044
-0.018
-0.138
0.028
0.001
-0.241
-0.085
1
0.913
0.931
-0.047

0.065
0.024
-0.160
0.039
-0.037
0.107
0.056
0.913
1
0.905
0.272

0.046
0.051
-0.142
0.054
-0.021
-0.073
-0.005
0.931
0.905
1
0.143

0.059
0.187
-0.153
0.047
-0.097
0.958
0.268
-0.047
0.272
0.143
1

The correlation coefficient matrix, as shown in Table 4.7, presents little correlation between the
selected research locations. This may be due to time-lag or non-synchronized issues and varying
antenna and sensor heights.
Table 4.7: RSL Correlation Coefficients of the Chosen Sites
Greenville
Lake City DOT
SR-222

Greenville
1
0.214
0.089

Lake City DOT


0.214
1
0.177

SR-222
0.089
0.177
1

A cross-correlation, the measure of similarity between two waveforms when a time-lag is


applied, is applied to the three sites since the correlation of RSL and weather data appears to be
very small. The output matrices (Tables 4.8, 4.9, and 4.10) of the sample cross-correlation
coefficients are similar to the output matrices for the correlation coefficients (Tables 4.4, 4.5, and
4.6 above). The values of the output matrices must be close to either +1 or -1 to confer a
relationship of dependence. The values of the output matrices for both the correlation coefficient
matrices and sample cross-correlation coefficient matrices are close to zero, thus affirming the
RSL and weather parameters are independent of one another. Preprocessing is needed to find a
correlation between the acquired data.

34

Table 4.8: RSL and Weather Parameter Cross-Correlation Coefficients for Greenville
WS
0.0095
0.0095
0.0096
0.0100
0.0094
0.0100
0.0097
0.0102
0.0100
0.0108
0.0108
0.0106
0.0116
0.0115
0.0117
0.0116
0.0116
0.0117
0.0121
0.0119
0.0120
0.0125
0.0119
0.0123
0.0129
0.0127
0.0137
0.0139
0.0134
0.0140
0.0152
0.0146
0.0153
0.0156
0.0157
0.0158
0.0159
0.0164
0.0163
0.0168
0.0170

WSA
0.0043
0.0043
0.0044
0.0043
0.0044
0.0045
0.0046
0.0047
0.0049
0.0051
0.0053
0.0054
0.0055
0.0056
0.0056
0.0055
0.0056
0.0056
0.0056
0.0056
0.0055
0.0054
0.0052
0.0052
0.0053
0.0055
0.0056
0.0058
0.0059
0.0059
0.0060
0.0061
0.0062
0.0063
0.0064
0.0064
0.0063
0.0062
0.0061
0.0060
0.0059

WD
0.0279
0.0278
0.0277
0.0273
0.0283
0.0278
0.0280
0.0289
0.0283
0.0277
0.0293
0.0291
0.0296
0.0290
0.0283
0.0287
0.0299
0.0311
0.0315
0.0310
0.0308
0.0313
0.0323
0.0323
0.0343
0.0332
0.0337
0.0351
0.0359
0.0354
0.0361
0.0357
0.0361
0.0356
0.0363
0.0356
0.0358
0.0354
0.0350
0.0346
0.0348

P
-0.0486
-0.0486
-0.0484
-0.0483
-0.0485
-0.0488
-0.0491
-0.0494
-0.0498
-0.0502
-0.0504
-0.0506
-0.0508
-0.0510
-0.0513
-0.0516
-0.0520
-0.0524
-0.0528
-0.0533
-0.0538
-0.0544
-0.0549
-0.0555
-0.0560
-0.0563
-0.0567
-0.0572
-0.0576
-0.0581
-0.0586
-0.0590
-0.0594
-0.0598
-0.0603
-0.0606
-0.0611
-0.0614
-0.0618
-0.0622
-0.0625

RH
-0.0537
-0.0542
-0.0548
-0.0553
-0.0559
-0.0565
-0.0571
-0.0576
-0.0582
-0.0587
-0.0593
-0.0599
-0.0604
-0.0610
-0.0616
-0.0621
-0.0628
-0.0633
-0.0639
-0.0645
-0.0650
-0.0656
-0.0662
-0.0667
-0.0672
-0.0677
-0.0682
-0.0687
-0.0691
-0.0696
-0.0700
-0.0704
-0.0708
-0.0713
-0.0717
-0.0721
-0.0725
-0.0729
-0.0734
-0.0737
-0.0741

BP
0.0326
0.0326
0.0325
0.0324
0.0324
0.0324
0.0323
0.0323
0.0323
0.0323
0.0323
0.0323
0.0323
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0322
0.0323
0.0323
0.0323
0.0324
0.0324
0.0324
0.0324
0.0324
0.0325
0.0324
0.0324
0.0324
0.0324

35

T
-0.0097
-0.0092
-0.0087
-0.0081
-0.0076
-0.0071
-0.0065
-0.0060
-0.0055
-0.0049
-0.0044
-0.0039
-0.0033
-0.0028
-0.0023
-0.0018
-0.0013
-0.0008
-0.0003
0.0001
0.0006
0.0010
0.0015
0.0019
0.0024
0.0029
0.0033
0.0037
0.0042
0.0046
0.0050
0.0053
0.0057
0.0061
0.0065
0.0068
0.0072
0.0076
0.0079
0.0083
0.0086

WC
-0.0140
-0.0135
-0.0129
-0.0124
-0.0118
-0.0113
-0.0108
-0.0103
-0.0097
-0.0091
-0.0086
-0.0081
-0.0075
-0.0070
-0.0065
-0.0060
-0.0055
-0.0050
-0.0045
-0.0041
-0.0036
-0.0032
-0.0027
-0.0023
-0.0019
-0.0014
-0.0010
-0.0006
-0.0001
0.0002
0.0006
0.0009
0.0013
0.0016
0.0019
0.0023
0.0026
0.0030
0.0033
0.0037
0.0040

HI
-0.0282
-0.0277
-0.0271
-0.0265
-0.0259
-0.0254
-0.0248
-0.0242
-0.0236
-0.0229
-0.0223
-0.0218
-0.0212
-0.0207
-0.0201
-0.0196
-0.0191
-0.0186
-0.0181
-0.0176
-0.0171
-0.0166
-0.0161
-0.0156
-0.0152
-0.0147
-0.0143
-0.0138
-0.0134
-0.0130
-0.0126
-0.0123
-0.0119
-0.0117
-0.0113
-0.0109
-0.0106
-0.0103
-0.0100
-0.0096
-0.0093

DP
-0.0696
-0.0695
-0.0694
-0.0693
-0.0692
-0.0692
-0.0691
-0.0690
-0.0688
-0.0686
-0.0686
-0.0685
-0.0684
-0.0683
-0.0683
-0.0682
-0.0683
-0.0682
-0.0682
-0.0681
-0.0681
-0.0680
-0.0680
-0.0680
-0.0680
-0.0678
-0.0679
-0.0678
-0.0677
-0.0677
-0.0676
-0.0676
-0.0676
-0.0677
-0.0676
-0.0675
-0.0675
-0.0675
-0.0675
-0.0674
-0.0674

Table 4.9: RSL and Weather Parameter Cross-Correlation Coefficients for Lake City DOT
WS
0.0045
0.0043
0.0040
0.0038
0.0036
0.0034
0.0033
0.0033
0.0033
0.0035
0.0037
0.0038
0.0040
0.0042
0.0043
0.0045
0.0047
0.0048
0.0048
0.0047
0.0046
0.0046
0.0047
0.0048
0.0050
0.0052
0.0054
0.0054
0.0054
0.0054
0.0054
0.0054
0.0055
0.0055
0.0055
0.0055
0.0054
0.0055
0.0056
0.0057
0.0058

WSA
0.0064
0.0063
0.0063
0.0063
0.0063
0.0063
0.0064
0.0065
0.0067
0.0069
0.0071
0.0073
0.0075
0.0077
0.0079
0.0081
0.0082
0.0083
0.0083
0.0082
0.0081
0.0081
0.0082
0.0084
0.0086
0.0088
0.0090
0.0091
0.0091
0.0090
0.0089
0.0089
0.0089
0.0089
0.0090
0.0091
0.0092
0.0093
0.0095
0.0096
0.0098

WD
0.0622
0.0622
0.0624
0.0623
0.0617
0.0619
0.0612
0.0614
0.0610
0.0606
0.0600
0.0602
0.0593
0.0597
0.0579
0.0582
0.0583
0.0584
0.0579
0.0582
0.0580
0.0581
0.0579
0.0580
0.0585
0.0586
0.0595
0.0595
0.0593
0.0586
0.0587
0.0588
0.0591
0.0589
0.0582
0.0584
0.0593
0.0588
0.0587
0.0590
0.0591

P
-0.0496
-0.0500
-0.0503
-0.0507
-0.0510
-0.0514
-0.0517
-0.0521
-0.0525
-0.0530
-0.0535
-0.0539
-0.0544
-0.0549
-0.0553
-0.0558
-0.0562
-0.0567
-0.0573
-0.0579
-0.0585
-0.0592
-0.0598
-0.0604
-0.0609
-0.0614
-0.0620
-0.0625
-0.0629
-0.0633
-0.0636
-0.0641
-0.0645
-0.0648
-0.0652
-0.0657
-0.0660
-0.0664
-0.0666
-0.0670
-0.0673

RH
-0.0795
-0.0797
-0.0800
-0.0803
-0.0806
-0.0809
-0.0813
-0.0815
-0.0819
-0.0822
-0.0825
-0.0828
-0.0831
-0.0834
-0.0837
-0.0841
-0.0844
-0.0847
-0.0849
-0.0852
-0.0856
-0.0858
-0.0860
-0.0862
-0.0864
-0.0866
-0.0868
-0.0870
-0.0872
-0.0873
-0.0875
-0.0876
-0.0877
-0.0879
-0.0880
-0.0881
-0.0882
-0.0883
-0.0884
-0.0884
-0.0885

BP
0.0519
0.0519
0.0519
0.0519
0.0519
0.0519
0.0519
0.0519
0.0520
0.0520
0.0521
0.0521
0.0521
0.0521
0.0521
0.0521
0.0522
0.0523
0.0523
0.0523
0.0523
0.0522
0.0522
0.0522
0.0523
0.0524
0.0524
0.0525
0.0525
0.0525
0.0525
0.0526
0.0525
0.0525
0.0526
0.0526
0.0527
0.0527
0.0527
0.0526
0.0526

36

T
0.0618
0.0620
0.0623
0.0625
0.0628
0.0630
0.0632
0.0634
0.0636
0.0639
0.0640
0.0642
0.0644
0.0645
0.0647
0.0649
0.0651
0.0653
0.0654
0.0656
0.0657
0.0658
0.0659
0.0659
0.0660
0.0661
0.0661
0.0661
0.0661
0.0661
0.0662
0.0661
0.0661
0.0661
0.0661
0.0660
0.0660
0.0660
0.0660
0.0659
0.0659

WC
0.0753
0.0756
0.0759
0.0762
0.0765
0.0768
0.0770
0.0771
0.0774
0.0776
0.0778
0.0779
0.0781
0.0782
0.0783
0.0785
0.0787
0.0789
0.0791
0.0792
0.0794
0.0795
0.0796
0.0797
0.0797
0.0798
0.0799
0.0799
0.0800
0.0800
0.0801
0.0801
0.0801
0.0801
0.0801
0.0801
0.0801
0.0802
0.0802
0.0801
0.0801

HI
0.0784
0.0788
0.0791
0.0795
0.0799
0.0802
0.0806
0.0809
0.0812
0.0816
0.0819
0.0822
0.0825
0.0827
0.0830
0.0834
0.0838
0.0840
0.0843
0.0846
0.0848
0.0850
0.0852
0.0853
0.0855
0.0856
0.0857
0.0857
0.0858
0.0860
0.0860
0.0861
0.0862
0.0863
0.0864
0.0865
0.0865
0.0866
0.0867
0.0867
0.0867

DP
-0.0136
-0.0136
-0.0136
-0.0137
-0.0137
-0.0137
-0.0138
-0.0138
-0.0139
-0.0139
-0.0140
-0.0140
-0.0141
-0.0143
-0.0144
-0.0145
-0.0145
-0.0146
-0.0147
-0.0147
-0.0149
-0.0149
-0.0150
-0.0151
-0.0152
-0.0153
-0.0154
-0.0155
-0.0156
-0.0157
-0.0158
-0.0159
-0.0160
-0.0161
-0.0162
-0.0163
-0.0163
-0.0164
-0.0165
-0.0165
-0.0166

Table 4.10: RSL and Weather Parameter Cross-Correlation Coefficients for SR-222
WS
0.0998
0.0999
0.0997
0.0999
0.0999
0.0999
0.1000
0.0995
0.0994
0.0996
0.0995
0.0993
0.0994
0.0995
0.0995
0.0994
0.0993
0.0992
0.0992
0.0991
0.0993
0.0991
0.0992
0.0992
0.0992
0.0992
0.0993
0.0996
0.0999
0.1000
0.0998
0.0997
0.0993
0.0990
0.0990
0.0989
0.0987
0.0985
0.0982
0.0984
0.0987

WSA
0.0756
0.0757
0.0758
0.0759
0.0759
0.0760
0.0761
0.0761
0.0761
0.0762
0.0763
0.0764
0.0765
0.0765
0.0766
0.0767
0.0768
0.0768
0.0768
0.0768
0.0768
0.0769
0.0769
0.0770
0.0771
0.0771
0.0771
0.0772
0.0772
0.0772
0.0771
0.0770
0.0770
0.0770
0.0769
0.0769
0.0769
0.0770
0.0770
0.0770
0.0770

WD
-0.0593
-0.0591
-0.0588
-0.0586
-0.0589
-0.0587
-0.0584
-0.0580
-0.0582
-0.0578
-0.0560
-0.0551
-0.0544
-0.0551
-0.0551
-0.0553
-0.0556
-0.0549
-0.0533
-0.0528
-0.0519
-0.0512
-0.0506
-0.0502
-0.0505
-0.0509
-0.0503
-0.0503
-0.0503
-0.0501
-0.0510
-0.0505
-0.0500
-0.0492
-0.0483
-0.0484
-0.0480
-0.0479
-0.0481
-0.0483
-0.0484

P
-0.0833
-0.0835
-0.0839
-0.0842
-0.0847
-0.0851
-0.0856
-0.0860
-0.0864
-0.0867
-0.0870
-0.0873
-0.0877
-0.0879
-0.0882
-0.0884
-0.0887
-0.0889
-0.0892
-0.0894
-0.0895
-0.0897
-0.0901
-0.0904
-0.0906
-0.0909
-0.0911
-0.0913
-0.0915
-0.0917
-0.0919
-0.0922
-0.0926
-0.0928
-0.0930
-0.0932
-0.0932
-0.0933
-0.0934
-0.0935
-0.0937

RH
0.0454
0.0454
0.0453
0.0453
0.0452
0.0452
0.0452
0.0452
0.0452
0.0452
0.0452
0.0452
0.0452
0.0452
0.0451
0.0451
0.0451
0.0451
0.0451
0.0450
0.0450
0.0450
0.0450
0.0449
0.0449
0.0449
0.0449
0.0449
0.0450
0.0450
0.0450
0.0452
0.0452
0.0452
0.0453
0.0453
0.0454
0.0454
0.0455
0.0455
0.0455

BP
0.0597
0.0596
0.0594
0.0593
0.0591
0.0590
0.0588
0.0586
0.0585
0.0584
0.0582
0.0580
0.0579
0.0577
0.0576
0.0574
0.0573
0.0571
0.0570
0.0568
0.0567
0.0566
0.0565
0.0564
0.0562
0.0560
0.0559
0.0557
0.0556
0.0554
0.0553
0.0552
0.0550
0.0549
0.0548
0.0546
0.0545
0.0544
0.0543
0.0541
0.0540

37

T
0.0364
0.0368
0.0372
0.0376
0.0380
0.0384
0.0388
0.0392
0.0396
0.0401
0.0404
0.0408
0.0412
0.0417
0.0421
0.0425
0.0429
0.0433
0.0437
0.0441
0.0445
0.0449
0.0452
0.0456
0.0460
0.0463
0.0466
0.0470
0.0473
0.0477
0.0480
0.0483
0.0486
0.0489
0.0492
0.0495
0.0497
0.0500
0.0502
0.0505
0.0507

WC
0.0558
0.0563
0.0567
0.0571
0.0575
0.0580
0.0584
0.0588
0.0593
0.0597
0.0601
0.0606
0.0610
0.0615
0.0619
0.0624
0.0629
0.0633
0.0638
0.0642
0.0646
0.0650
0.0654
0.0658
0.0662
0.0665
0.0669
0.0673
0.0677
0.0681
0.0685
0.0689
0.0693
0.0696
0.0700
0.0703
0.0706
0.0709
0.0712
0.0715
0.0718

HI
0.0382
0.0386
0.0390
0.0394
0.0398
0.0402
0.0406
0.0410
0.0414
0.0419
0.0423
0.0427
0.0431
0.0435
0.0439
0.0443
0.0448
0.0452
0.0456
0.0460
0.0463
0.0467
0.0470
0.0474
0.0477
0.0481
0.0485
0.0488
0.0491
0.0494
0.0497
0.0500
0.0503
0.0506
0.0508
0.0511
0.0513
0.0515
0.0518
0.0520
0.0522

DP
0.0582
0.0583
0.0583
0.0584
0.0585
0.0586
0.0586
0.0587
0.0588
0.0589
0.0589
0.0590
0.0591
0.0592
0.0592
0.0592
0.0593
0.0594
0.0594
0.0594
0.0594
0.0595
0.0595
0.0595
0.0596
0.0596
0.0598
0.0598
0.0598
0.0600
0.0601
0.0603
0.0604
0.0605
0.0605
0.0608
0.0609
0.0609
0.0611
0.0612
0.0613

4.4. Fast Fourier Transform and Power Spectrum Analysis


The Discrete Fourier Transform (DFT) decomposes a sequence of values in a function from their
time domain representation to their frequency domain representation.

The Fast Fourier

Transform (FFT) is a faster variation of the DFT algorithm and is able to compute the DFT and
its inverse. The FFT requires only
when

log

individual steps and transforming is worthwhile

log , where L is the vector length [12]. The FFT is defined as

0, ,

(4.1)

and the multidimensional FFT is defined as

(4.2)

4.4.1. Fast Fourier Transform Analysis


The multidimensional FFT was used to compute data in MATLAB R2007b and a sample of this
computation is presented in Figure 4.9.

Figure 4.9: FFT of Greenville RSL and ESS Data

38

The FFT is not recommended to analyze non-stationary signals since it cannot distinguish the
two or multiple signals very well. The FFT sees both signals as the same and constituted of the
same frequency components, as shown in Figures 4.9 and 4.10. Thus the FFT is not a suitable
tool for analyzing non-stationary signals or time-varying spectra. This information was found
after analysis was well under way and the rest of section 4.4 displays evidence for this argument.

Figure 4.10: Enlarged Window of the FFT of Greenville RSL and ESS Data

4.4.2. FFT Power Spectrum Analysis


The power spectrum of the FFT is very noisy and it is difficult to infer any correlation. Figures
4.11 and 4.12 present the power spectrum for Greenville over one day and one hour period,
respectively. From these figures it is clear that the power spectrum is not only distorted but also
a low method for determining any true correlation.

39

Figure 4.11: Power Spectrum of Greenville RSL and ESS data for One Day

Figure 4.12: Power Spectrum of Greenville RSL and ESS data for a One Hour
40

4.4.3. Correlation Analysis


The correlation analysis shows very high correlation between RSL and all weather parameters,
but this is only a strong correlation between the frequency components, not the spatial
correlation. See Table 4.11 below.
Table 4.11: FFT Correlation Coefficients for Greenville
RSL
RSL
WS
WSA
WD
P
RH
BP
T
WC
HI
DP

1
0.923
0.986
0.952
0.875
0.993
0.993
0.875
0.952
0.986
0.923

WS
0.923
1
0.968
0.980
0.980
0.958
0.958
0.984
0.981
0.966
0.998

WSA
0.986
0.968
1
0.984
0.934
0.997
0.997
0.939
0.986
0.998
0.966

WD
0.952
0.980
0.984
1
0.977
0.979
0.978
0.972
0.997
0.986
0.981

P
0.875
0.980
0.934
0.977
1
0.922
0.922
0.987
0.972
0.939
0.984

RH
0.993
0.958
0.997
0.979
0.922
1
1.000
0.922
0.978
0.997
0.958

BP
0.993
0.958
0.997
0.978
0.922
1.000
1
0.922
0.979
0.997
0.958

T
0.875
0.984
0.939
0.972
0.987
0.922
0.922
1
0.977
0.934
0.980

WC
0.952
0.981
0.986
0.997
0.972
0.978
0.979
0.977
1
0.984
0.980

HI
0.986
0.966
0.998
0.986
0.939
0.997
0.997
0.934
0.984
1
0.968

DP
0.923
0.998
0.966
0.981
0.984
0.958
0.958
0.980
0.980
0.968
1

4.5. Short Time Fourier Transform and Power Spectrum Analysis


The Short Time Fourier Transform (STFT) is a Fourier related transform that is used to
determine the sinusoidal frequency and phase content of local sections of a signal as it changes
over time.

This method is accurate only for a specific time and frequency resolution.

Heisenbergs uncertainty principle states the momentum and position of a moving particle cannot
be known simultaneously. This can be applied to signals and other discrete data. In the case of
frequency and time, the spectral component cannot be known at a given instant. This may cause
noise in the result of the STFT, either in the frequency or time resolutions. The power spectrum
is a function of frequency and is a deterministic function of time. It has dimensions of power per
Hz or energy per Hz and helps to identify periodicities, and is utilized to correlate RSL and
various weather conditions.

4.5.1. Short Time Fourier Transform Analysis


The STFT breaks the data to be transformed into block sections or windows along the signal
under analysis and performs the FT within the windows. The complex result is added to a

41

matrix, which records magnitude and phase for each point in time and frequency. The STFT can
be expressed as

(4.3)

The exponent determines the resolution of the frequency component in the STFT. When the
window or frame is small the time resolution is high, but the frequency resolution is low due to
the Heisenbergs uncertainty principle.

4.5.2. STFT Power Spectrum Analysis


The power spectrum of the STFT was computed for a small portion of data from the Greenville
ESS site. A 3-D plot of the STFT, time vs. frequency vs. power, is shown in Figure 4.13.

The

frequency component resolution is very well defined and has distinguishable amplitude or
power, as shown in Figure 4.14, but the time resolution is low. The time-axis is very noisy or
distorted. Figure 4.15 shows amplitude vs. time. The time values are very long and blend
together, thus a lot of noise or distortion is clearly present in the signal.

Figure 4.13: RSL STFT at 45 Angle for Greenville ESS Rotated Approximately 180

42

Figure 4.14: RSL STFT Frequency vs. Power for Greenville ESS

Figure 4.15: RSL STFT Time vs. Power for Greenville ESS
43

4.5.3. Correlation Analysis


The correlation analysis is not computed for the STFT due to the fact that the STFT did not yield
ranges from - to +. The STFT has

clear results. In the FFT the kernel window

windows of finite length, covering only a small portion of the signal, which in turn reduces the
frequency resolution [18]. The location of the exact frequency components that exist in the
signal is no longer known, only the band of frequencies that exist are known. An example of this
is the FFT example in section 4.4.1. The dilemma occurs in the choice of window size. When
the window is increased, the frequency resolution increase (and time resolution decreases) and
when the window is decreased the frequency resolution decreases (and time resolution
increases).

4.6. Discrete Wavelet Transform and Wavelet Decomposition Analysis


The Discrete Wavelet Transform (DWT) in MATLAB performs a single-level 1-D wavelet
decomposition with respect to a particular wavelet. The wavelet name chosen for this project is
Daubechies. In general the Daubechies wavelets are chosen to have the highest number A of
vanishing moments, (yet this does not imply the best smoothness) for given support width N=2A,
and among the 2A1 possible solutions the one is chosen whose scaling filter has extremal phase
[12]. The DWT provides sufficient information both for analysis and synthesis of the original
signal and with a reduction in computation time.

One level of decomposition and can

mathematically be expressed as follows:

(4.4)

(4.5)

where yhigh[k] and ylow[k] are the outputs of the highpass and lowpass filters, respectively, after
subsampling by 2 [18].

4.6.1. Wavelet Decomposition Analysis


The DWT (single-level wavelet decomposition) analyzes signals at different frequency bands at
different resolutions: coarse approximation and detailed information. The DWT incorporates
two sets of functions scaling functions (associated with lowpass filters) and wavelet functions
(associated with highpass filters).

The decomposition of a sampled signal into different

44

frequency bands is obtained by successive highpass and lowpass filtering of the sampled time
domain signal.

Figure 4.16: Discrete Wavelet Transform Illustration. (Figure 4.1 from Ref. 18, courtesy of R
Polikar.)
After filtering the signal, half of the samples are discarded, or subsampled by two, due to
the Nyquist - Shannon sampling theorem. As shown in Equation 4.6, the DWT continues to
filter the lower frequencies of half band filtering until the level of resolution is met. See Figure
4.16. The DWT is obtained by concatenating all coefficients starting from the last level of the
decomposition [18] and will have the same number of coefficients as the original signal.
Equation 4.7 is the reconstruction formula for each layer. Frequencies most prominent in the
45

original signal will appear as high amplitudes in the region of the DWT signal that include those
particular frequencies. Unlike the FFT, the DWT will not lose time localization of frequencies.

(4.6)

(4.7)

In Figure 4.17 the stages of a three level wavelet decomposition are presented

Figure 4.17: Stages of a Three Level Wavelet Decomposition


More than three levels could have been applied and as more levels are applied to the
wavelet decomposition, more of the input signal is filtered. This can theoretically dampen the
signal too much and the results would then appear as zero or near zero amplitude. Three levels
of decomposition are necessary to view the similarities between the RSL and weather
parameters. The three research sites for this project displayed a correlation between RSL at each
site location and their respected weather conditions after a three level wavelet decomposition
was calculated. Figure 4.18 - Figure 4.24 present various wavelet decomposition trials at various
scales. The precipitation parameter shows little correlation or temporal symmetry to the RSL.
This observation holds true for each site. The results for Greenville are shown in Figure 4.18.
The RSL and precipitation for Lake City DOT, as shown in Figure 4.21, has a smaller window
and the results more apparent.
46

Figure 4.18: Wavelet Decomposition of Precipitation and RSL for Greenville Data
Observations of the original data, with no preprocessing, presented a correlation between
the temperature, relative humidity, and received signal level. Attenuation is present during
increasing humidity and decreasing temperature with no presence of wind. It is normal to have a
decrease in signal strength during early morning hours and many observations made by FDOT
employees have confirmed this. The RSL, RH, and T wavelet decompositions for Greenville,
Lake City DOT, and SR-222 are shown in Figures 4.20, 4.22, 4.24, respectively. The three level
wavelet decomposition removed noise and distortion from the signal and presented the scaled
frequency components in the time domain allowing the attenuation and gain characteristics
viewable for analysis. Reviewing the data and wavelet analysis has shown that the major factors
in attenuation are wind speed, relative humidity, and temperature.

When the temperature

decreases and the relative humidity increases, the presence of high water vapor or fog occurs.
Studying of the data displays more attenuation when wind is not present. This leads to still or
slowly rising water vapor or fog and at the 6.8 GHz frequency these atmospheric conditions
cause visible attenuation in the signal. More on this will be discussed in Chapter 5.

47

Figure 4.19: Wavelet Decomposition for RSL, RH, and T at Greenville ESS Site

Figure 4.20: Enlarged Wavelet Decomposition for Greenville Data


48

Figure 4.21: Wavelet Decomposition of Precipitation and RSL for Lake City DOT Data

Figure 4.22: Enlarged Wavelet Decomposition for Lake City DOT Data
49

Figure 4.23: Wavelet Decomposition of Precipitation and RSL for SR-222 Data

Figure 4.24: Enlarged Wavelet Decomposition for SR-222 Data


50

4.6.2. Correlation Analysis


A three level wavelet decomposition of the data removed noise and distortion from the data.
Most correlation coefficients for all sites show a strong spatial correlation between RSL and
various weather parameters.

Tables 4.12 through 4.14 display the correlation coefficient

matrices of the data. The barometric pressure shows almost perfect correlation to the RSL of
each site. This is due to the minor fluctuations in BP due to its scaling factors. The amount of
rainfall and rain intensity based on the rainfall period has little effect on the RSL.
Table 4.12: Correlation Coefficients of Three Level Wavelet Decomposition for Greenville
RSL
RSL
WS
WSA
P
RH
BP
T
WC
HI
DP

1
-0.423
-0.809
-0.189
-0.971
-0.999
-0.991
-0.991
-0.986
-0.993

WS
-0.423
1
0.450
0.058
0.327
0.424
0.448
0.449
0.453
0.410

WSA
-0.809
0.450
1
0.175
0.770
0.810
0.798
0.800
0.795
0.792

P
-0.189
0.058
0.175
1
0.217
0.186
0.177
0.177
0.170
0.194

RH
-0.971
0.327
0.770
0.217
1
0.971
0.943
0.943
0.935
0.971

BP
-0.999
0.424
0.810
0.186
0.971
1
0.992
0.992
0.987
0.993

T
-0.991
0.448
0.798
0.177
0.943
0.992
1
1.000
0.999
0.993

WC
-0.991
0.449
0.800
0.177
0.943
0.992
1.000
1
0.999
0.993

HI
-0.986
0.453
0.795
0.170
0.935
0.987
0.999
0.999
1
0.991

DP
-0.993
0.410
0.792
0.194
0.971
0.993
0.993
0.993
0.991
1

Table 4.13: Correlation Coefficients of Three Level Wavelet Decomposition for Lake City DOT
RSL
RSL
WS
WSA
P
RH
BP
T
WC
HI
DP

1
-0.215
-0.219
-0.198
-0.950
-0.999
-0.991
-0.992
-0.988
-0.981

WS
-0.215
1
0.995
-0.001
0.118
0.220
0.216
0.210
0.215
0.167

WSA
-0.219
0.995
1
-0.001
0.119
0.224
0.220
0.214
0.219
0.169

P
-0.198
-0.001
-0.001
1
0.243
0.195
0.187
0.185
0.179
0.212

RH
-0.950
0.118
0.119
0.243
1
0.948
0.932
0.927
0.922
0.971

51

BP
-0.999
0.220
0.224
0.195
0.948
1
0.992
0.994
0.989
0.980

T
-0.991
0.216
0.220
0.187
0.932
0.992
1
0.999
0.997
0.986

WC
-0.992
0.210
0.214
0.185
0.927
0.994
0.999
1
0.998
0.982

HI
-0.988
0.215
0.219
0.179
0.922
0.989
0.997
0.998
1
0.981

DP
-0.981
0.167
0.169
0.212
0.971
0.980
0.986
0.982
0.981
1

Table 4.14: Correlation Coefficients of Three Level Wavelet Decomposition for SR-222
RSL
RSL
WS
WSA
P
RH
BP
T
WC
HI
DP

1
-0.430
-0.538
-0.231
-0.373
-0.998
-0.991
-0.992
-0.992
0.397

WS

WSA

-0.430
1
0.675
0.033
0.273
0.438
0.433
0.437
0.440
-0.018

-0.538
0.675
1
0.084
0.088
0.544
0.527
0.527
0.528
-0.332

P
-0.231
0.033
0.084
1
0.003
0.225
0.224
0.221
0.222
-0.176

RH

BP

-0.373
0.273
0.088
0.003
1
0.378
0.350
0.385
0.368
0.667

-0.998
0.438
0.544
0.225
0.378
1
0.994
0.995
0.994
-0.393

T
-0.991
0.433
0.527
0.224
0.350
0.994
1
0.999
0.999
-0.397

WC
-0.992
0.437
0.527
0.221
0.385
0.995
0.999
1
0.999
-0.367

HI

DP

-0.992
0.440
0.528
0.222
0.368
0.994
0.999
0.999
1
-0.378

0.397
-0.018
-0.332
-0.176
0.667
-0.393
-0.397
-0.367
-0.378
1

The relative humidity sensor for SR-222 malfunctioned during early stages of research.
The Capricorn 2000TM calculates the DP and HI using the RH value, thus this malfunction
effected the calculation for both the heat index and dew point. The correlation coefficients
display this discrepancy in Table 4.14. Further analysis will be discussed in Chapter 5.

4.7. Key Findings


The three level wavelet decomposition method of analysis proved most useful and presented an
accurate correlation of raw data observations to computed correlations. After review of the
recorded data, signal strength is most often attenuated during early to mid hours of the morning,
ranging from approximately 3:00 a.m. to 10:00 a.m. During these periods the main changes in
weather patterns are RH, T, and WS. In most cases when the RH increases, T decreases, and no
WS is present, the RSL fluctuates in amplitude. Fog and high amounts of water vapor are
present under conditions of high RH and low T. Without wind the water vapor or fog becomes
stagnant and acts as a wall for many microwave frequencies. Figure 4.25 shows a sample of raw
data for the Greenville site.
The ITU model is used to test attenuation due to rain rate. A plot of the three level
wavelet decomposition of the acquired data is shown in Figure 4.26. A correlation analysis of
the results did not present any strong correlations. For each site the precipitation shows a very
weak correlation of attenuation to RSL; specifically the mean value of the correlation coefficient
value is -0.206. Wind speed also presents a weak correlation coefficient; weaker than WSA, but
52

when analyzing the raw data a trend appears that when no wind is present the signal level
fluctuates more than when present, as can be seen during Day Seven in Figure 4.25.

Figure 4.25: Greenville Data during First Week of April, 2010


The ITU atmospheric attenuation prediction plot in Figure 3.1 displays higher attenuation
for 6.8 GHz and higher frequencies due to water vapor and dry air. Figures 4.26 and 4.27 show a
three level wavelet decomposition of the Greenville site data.

Figure 4.27 displays high

correlation of most weather attributes. Further analysis leads to the relationship of relative
humidity and temperature and the presence of wind. These are prime conditions for water vapor,
fog, and low-lying cloud formation. Based on the data collected those atmospheric conditions
are major factors in 6.8 GHz signal variation which may be a result of multipath fading and
atmospheric ducting.
Multipath is the propagation phenomenon where transmitted (TX) radio signals reach the
receiving (RX) antenna by two or more paths. Mountains, buildings, and other terrestrial objects
can cause multipath. When the temperature lowers and relative humidity rises, reflection from
moist surfaces, such as trees and other terrestrial objects, by water vapor and dew can cause
transmissions to reach the RX antenna with a phase shift causing attenuation, or sometimes gain.

53

Figure 4.26: Three Level Wavelet Decomposition for Greenville Data

Figure 5.27: Enlarged Three Level Wavelet Decomposition for Greenville Data
54

Attenuation in the RSL will occur when the TX signal and a reflected TX signal reach the RX
antenna at different phases, and a gain in the RSL will occur when the two signals arrive in
phase. These effects are referred to as destructive and constructive interference and phase
shifting of the signal which will cause Rayleigh fading.
An atmospheric duct in a horizontal layer in the lower atmosphere will also cause gain in
a RX signal. Atmospheric ducting along a horizontal path has vertical refractive index gradients
such that radio signals are guided or ducted. These signals tend to follow the curvature of the
Earth and experience less attenuation in the ducts than they would if the ducts were not present.
The duct acts as an atmospheric dielectric waveguide and limits the spread of the wave front to
only the horizontal dimension [23]. Thus, an atmospheric duct will cause gain in the RSL and is
visible in the acquired data.

55

CHAPTER 5
CONCLUSION AND FUTURE WORK

5.1. Conclusion
Studies of signal attenuation on microwave communications in the state of Florida are limited
and underdeveloped. The main focus of this manuscript is the correlation of microwave received
signal level (RSL) data and various weather parameters data, collected over a six month period.
Both RSL data and weather data used in this report were collected from three FDOT
telecommunication sites. The discrete wavelet transform and wavelet decomposition proved to
be useful tools in removing noise and distortion from a signal while displaying the frequency
components in the time domain with high resolution. After a three level wavelet decomposition
was performed a correlation analysis between RSL and atmospheric parameters was computed
and yielded clear results. Through these two methods of analysis the correlation of data was
easily calculated and a hypothesis for attenuation was formulated.
The RSL is effected by many of the weather parameters studied. Relative humidity,
temperature, and barometric pressure all held strong negative correlations when compared to the
RSL with correlation coefficient values -0.971, -0.991, and -0.999, respectively. The wind speed
and wind speed average also affected the RSL with weaker negative correlation values of -0.423
and -0.809, respectively.

Finally, precipitation held a weak negative correlation with a

correlation coefficient value of -0.189. A negative correlation value means if one variable
increases, then the other variable tends to decrease. This correlation points to the reduction in
signal strength as weather parameters vary, specifically relative humidity and inverse
temperature. Heat index, wind chill, and dew point use weather parameters with high correlation
coefficients in their calculation, specifically RH, T, and WSA. Hence, those parameters also
display a high correlation to RSL attenuation.
The study of raw data trends and the correlation of the wavelet decomposition prove the
hypothesis of this manuscript: microwave received signal strength is directly affected by
atmospheric conditions, specifically rise in relative humidity and fall in temperature. As T
decreases and RH increases, the air becomes saturated by water vapor and at higher densities
56

water vapor can become mist, fog, or a low-lying cloud which may result in multipath fading and
atmospheric ducting. During early hours of the morning reflection from moist surfaces, such as
trees and other terrestrial objects, by water vapor and dew will cause transmissions to reach the
RX antenna out of phase which will cause attenuation or gain and atmospheric ducting will cause
gain in the RSL and is visible in the acquired data.
It is concluded from the observed results of raw data and correlated wavelet
decompositions that weather conditions such as water vapor, mist, and rising fog have a direct
effect on microwave propagation and RSL attenuation.

5.2. Future Work and Recommendations


The current work and research direct the focus of this manuscript to prime conditions for water
vapor, fog and low-lying cloud formation. Further research into these areas is recommended.
The spatial diversity of weather conditions between microwave sites in the state of
Florida is extreme. Recommendations to improve research data and reliability would be the
installation of additional weather sensors between the microwave site locations, verses one ESS
collocated at the microwave site, as the average distance between the FDOT microwave sites
studied is approximately 26.4 km. Incorporating more ESS locations along the microwave path
link would provide more data observation points along the terrestrial path link, and would help in
determining the effect atmospheric conditions have on microwave propagation in order to
increase accuracy of future modeling and simulation. An annual data acquisition cycle (or
greater) would be ideal for better understanding the weather trends and create a higher chance of
recording data during a hurricane or other tropical occurrences.
Another recommendation would be to install weather sensors at the same heights as all
associated transmit (TX) and receive (RX) antennas at both ends of the signal path, and use
higher quality sensors.
During conditions for water vapor and low-lying cloud formation, moisture can form on
radome antennas, such as antennas used in this project, and cause reflection and multipath.
Further research into moisture levels of radomes during attenuation and gain events is
recommended.
Another recommendation is to study properties of metals used in microwave towers and
their reaction to temperature variances. The expansion and compression of tower members, due
57

to change in temperature, could have an effect on antenna alignment and cause variances in the
received signal level (RSL).
Final recommendations would be to analyze the data collected at all antenna locations
(main and diversity) and develop an improved model for antenna separation for a range of
microwave frequencies in the state of Florida. Current practices with respect to spatial diversity
of antenna are often reduced to Rule of Thumb measurements.

58

APPENDIX A
PROGRAM CODE

59

%-------------------------------------------------------------------------% This program will copy RSL data time values into separate cells (hours
% and minutes) for further data processing.
%-------------------------------------------------------------------------dataRow=1;
% row counter
for i=1:43139
% set to size of imported matrix
temp1=textdata(dataRow,1); % stores day cell into temp1
m=cell2mat(temp1);
% converts cell to matrix
s=split(':', m);
% splits the time cell
n1(i,1)=str2double(s(1));
% storing hour value
n1(i,2)=str2double(s(2));
% storing minute value
dataRow=dataRow+1;
% increments dataRow
end
MT=data(:,1);
% month temp
DT=data(:,2);
% day temp
RT=data(:,3);
% RSL temp
tempData=[MT DT n1 RT];
% new data matrix with hour & min included
%-------------------------------------------------------------------------%-------------------------------------------------------------------------% This program copies the temp matrix into newData without repeated
% time values.
%-------------------------------------------------------------------------%
Initial Constants
%-------------------------------------------------------------------------count1=1;
% initialize compare counter 1
count2=2;
% initialize compare counter 2
newDataCounter=1;
% initialize new data storage counter
repeatCounter=0;
% initialize repeat counter
repeatedMinutes=0;
% initialize repeated minute counter
%-------------------------------------------------------------------------%
Main Program
%-------------------------------------------------------------------------% WHILE IN FILE tempData DO...
while count2 < 43139
%---------------------------------------------------------------------% IF MINUTES ARE NOT SAME...
if tempData(count2,4)~=tempData(count1,4)
newData(newDataCounter,:)=tempData(count1,:);
% store data
newDataCounter=newDataCounter+1;
% increment counter
%---------------------------------------------------------------------% ELSE IF MINUTES ARE SAME, REMOVE THEM...
elseif tempData(count2,4)==tempData(count1,4)
repeatCounter=repeatCounter+1;
% increment counter
repeatedMinutes(repeatCounter,1)=newDataCounter;
% store data
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
newData(newDataCounter,:)=tempData(count1,:);
% store data
newDataCounter=newDataCounter+1;
% increment counter
end
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
end
%-------------------------------------------------------------------------% END OF FILE CLEAN UP...
newData(newDataCounter,:)=tempData(count1,:);
% store data in new matrix
newDataCounter=newDataCounter+1;
% increment counter
newData(newDataCounter,:)=tempData(count2,:);
% store data in new matrix
%--------------------------------------------------------------------------

60

%-------------------------------------------------------------------------% This program will linearly interpolate missing RSL data and create a new
% matrix containing the old data and interpolated data.
%-------------------------------------------------------------------------%
Initial Constants
%-------------------------------------------------------------------------count1=1;
% 1 of 2 compare counters
count2=2;
% 2 of 2 compare counters
monthCount=4;
% initializing month counter (to April)
dayCount=1;
% initializing day counter
hourCount=0;
% initializing hour counter
minuteCount=0;
% initializing minute counter
finalDataCounter=1; % counter for final data storage matrix
newDataCounter=1;
% counter for time split matrix: newData
%-------------------------------------------------------------------------%
Main Program
%-------------------------------------------------------------------------finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=newDataCounter+1; % incrementing counter
newDataCounter=newDataCounter+1;
% incrementing counter
minuteCount=minuteCount+1;
% incrementing counter
%-------------------------------------------------------------------------% WHILE IN FLIE newData DO...
while count2 < 42590
%---------------------------------------------------------------------% IF SAME MONTH THEN...
if newData(count2,1)==newData(count1,1)
%-----------------------------------------------------------------% IF SAME DAY THEN...
if newData(count2,2)==newData(count1,2)
%-------------------------------------------------------------% IF SAME HOUR THEN...
if newData(count2,3)==newData(count1,3)
%---------------------------------------------------------% IF NO 'HOLE' THEN...
if newData(count2,4)-newData(count1,4)==1
finalData(finalDataCounter,:)=...
newData(newDataCounter,:); % storing data
finalDataCounter=finalDataCounter+1;
newDataCounter=newDataCounter+1;
minuteCount=minuteCount+1;
count1=count1+1;
% incrementing counter
count2=count2+1;
% incrementing counter
else
%-----------------------------------------------------% ELSE INTERPOLATE 'HOLE'...
n=1;
% initialize temp counter variables
N=1;
% initialize temp counter variables
temp(n,1)=monthCount;
% store time split data in temp
temp(n,2)=dayCount;
% store time split data in temp
temp(n,3)=hourCount;
% store time split data in temp
temp(n,4)=minuteCount; % store time split data in temp
%-----------------------------------------------------% INTERPOLATION...
temp(n,5)=(newData(count2,5)-newData(count1,5))*...
((minuteCount-newData(count1,4))/(...
newData(count2,4)-newData(count1,4)))+...
newData(count1,5);
%-----------------------------------------------------% FILLING MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;

61

minuteCount=minuteCount+1;
%-----------------------------------------------------% WHILE 'HOLE(S)' EXIST, INTERPOLATE DATA...
while newData(count2,4)-temp(n,4)~=1
n=n+1;
temp(n,1)=monthCount;
% store time split in temp
temp(n,2)=dayCount;
% store time split in temp
temp(n,3)=hourCount;
% store time split in temp
temp(n,4)=minuteCount; % store time split in temp
%-------------------------------------------------% INTERPOLATION...
temp(n,5)=(newData(count2,5)-temp(N,5))*...
((minuteCount-temp(N,4))/...
(newData(count2,4)-temp(N,4)))+temp(N,5);
%-------------------------------------------------% FILL MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1;
N=N+1;
end
finalData(finalDataCounter,:)=...
newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;% increment counter
newDataCounter=newDataCounter+1;
% increment counter
minuteCount=minuteCount+1; % increment counter
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
end
%-------------------------------------------------------------% IF NEW HOUR THEN...
else
%---------------------------------------------------------% IF LAST MINUTE OF HOUR ISN'T THERE, THEN ADD IT...
if newData(count1,4)~=59
n=1;
% increment counter
N=1;
% increment counter
minuteCount=59;
temp(n,1)=monthCount;
% store time split data in temp
temp(n,2)=dayCount;
% store time split data in temp
temp(n,3)=hourCount;
% store time split data in temp
temp(n,4)=minuteCount; % store time split data in temp
%-----------------------------------------------------% INTERPOLATION...
temp(n,5)=(newData(count2,5)-newData(count1,5))*...
((minuteCount-newData(count1,4))/(60-...
newData(count1,4)))+newData(count1,5);
%-----------------------------------------------------% CHECKS AND ADDS PREVIOUS MINS (IF MISSING)...
while temp(N,4)-newData(count1,4)~=1
n=n+1; % increment counter
minuteCount=minuteCount-1; % decrement counter
temp(n,1)=monthCount;
% store temp data
temp(n,2)=dayCount;
% store temp data
temp(n,3)=hourCount;
% store temp data
temp(n,4)=minuteCount; % store temp data
%-------------------------------------------------% INTERPOLATION...
temp(n,5)=(temp(N,5)-newData(count1,5))*...
((minuteCount-newData(count1,4))/...
(temp(N,4)-newData(count1,4)))+...
newData(count1,5);

62

N=N+1; % increment counter


end
%-----------------------------------------------------% FILLING MATRIX W/ INTERP DATA BACKWARDS...
while N>0
finalData(finalDataCounter,:)=temp(N,:);
finalDataCounter=finalDataCounter+1;
N=N-1;
end
end
hourCount=hourCount+1; % increment counter
minuteCount=0; %initialize minute counter
%---------------------------------------------------------% IF FIRST MIN OF NEW HOUR NOT 0, THEN ADD IT...
if newData(count2,4)~=0
n=1;
% initialize counter
N=1;
% initialize counter
minuteCount=0; % initialize minute counter
temp(n,1)=monthCount;
% store count in temp variable
temp(n,2)=dayCount;
% store count in temp variable
temp(n,3)=hourCount;
% store count in temp variable
temp(n,4)=minuteCount; % store count in temp variable
%-----------------------------------------------------% INTERPOLATION...
temp(n,5)=abs((finalData(finalDataCounter-1,5)-...
newData(count2,5))*((minuteCount-...
newData(count2,4))/(finalData(...
finalDataCounter-1,4)-newData(count2,4)))+...
newData(count2,5));
%-----------------------------------------------------% FILLING MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1;
%-----------------------------------------------------% WHILE 'HOLE(S)' EXIST, INTERPOLATE DATA...
while newData(count2,4)-temp(n,4)~=1
n=n+1; % increment counter
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%-------------------------------------------------% INTERPOLATION...
temp(n,5)=(newData(count2,5)-temp(N,5))*...
((minuteCount-temp(N,4))/(newData(...
count2,4)-temp(N,4)))+temp(N,5);
%-------------------------------------------------% FILL MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1;
N=N+1; % increment counter
end
end
finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;% increment counter
newDataCounter=newDataCounter+1;
% increment counter
minuteCount=minuteCount+1; % increment counter
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
end

63

%-----------------------------------------------------------------% IF NEW DAY THEN...


else
%-------------------------------------------------------------% IF LAST MINUTE OF HOUR ISN'T THERE, THEN ADD IT...
if newData(count1,4)~=59
n=1;
% initializing counter
N=1;
% initializing counter
minuteCount=59;
% setting minute to 59
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%---------------------------------------------------------% INTERPOLATION...
temp(n,5)=(newData(count2,5)-newData(count1,5))*...
((minuteCount-newData(count1,4))/(60-...
newData(count1,4)))+newData(count1,5);
%---------------------------------------------------------% CHECKS AND ADDS PREVIOUS MINS (IF MISSING)...
while temp(N,4)-newData(count1,4)~=1
n=n+1; % increment counter
minuteCount=minuteCount-1;
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%-----------------------------------------------------% INTERPOLATION...
temp(n,5)=(temp(N,5)-newData(count1,5))*...
((minuteCount-newData(count1,4))/(temp(N,4)-...
newData(count1,4)))+newData(count1,5);
N=N+1; % increment counter
end
%---------------------------------------------------------% FILLING MATRIX W/ INTERP DATA BACKWARDS...
while N>0
finalData(finalDataCounter,:)=temp(N,:);
finalDataCounter=finalDataCounter+1;
N=N-1; % increment counter
end
end
hourCount=0;
% reinitializing counter
minuteCount=1;
% reinitializing counter
dayCount=dayCount+1;
% increment counter
%-------------------------------------------------------------% IF FIRST MIN OF NEW HOUR NOT 0, THEN ADD IT...
if newData(count2,4)~=0
n=1;
% reinitializing counter
N=1;
% reinitializing counter
minuteCount=0;
% reinitializing counter
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%---------------------------------------------------------% INTERPOLATION LOOP...
temp(n,5)=abs((finalData(finalDataCounter-1,5)-...
newData(count2,5))*((minuteCount-...
newData(count2,4))/(finalData(...
finalDataCounter-1,4)-newData(count2,4)))+...
newData(count2,5));

64

%---------------------------------------------------------% FILLING MATRIX W/ INTERP DATA/INCREMENT CONST...


finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
% increment counter
minuteCount=minuteCount+1; % increment counter
%---------------------------------------------------------% WHILE 'HOLE(S)' EXIST, INTERPOLATE DATA...
while newData(count2,4)-temp(n,4)~=1
n=n+1;
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%-----------------------------------------------------% INTERPOLATION...
temp(n,5)=(newData(count2,5)-temp(N,5))*...
((minuteCount-temp(N,4))/(newData(count2,4)-...
temp(N,4)))+temp(N,5);
%-----------------------------------------------------% FILL MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1; % increment minute counter
N=N+1; % increment counter
end
minuteCount=minuteCount+1; % increment counter
end
finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;
% increment counter
newDataCounter=newDataCounter+1;
% increment counter
hourCount=0;
% reinitialize counter
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
end
%---------------------------------------------------------------------% IF NEW MONTH THEN...
else
finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;
newDataCounter=newDataCounter+1;
minuteCount=0;
% reinitialize counter
hourCount=0;
% reinitialize counter
dayCount=1;
% reinitialize counter
monthCount=monthCount+1;
% increment counter
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
end
end
%-------------------------------------------------------------------------% END OF FILE CLEAN UP...
finalData(finalDataCounter,:)=newData(newDataCounter,:);
%-------------------------------------------------------------------------%-------------------------------------------------------------------------% This program will copy ESS data time values into separate cells (hours
% and minutes) for further data processing.
%-------------------------------------------------------------------------dataRow=1;
% row counter
for i=1:37027
% set to size of imported matrix
temp1=textdata(dataRow,1); % stores day cell into temp1
temp2=textdata(dataRow,2); % stores time cell into temp2
M=cell2mat(temp1);
% converts cell to matrix
S=split('/', M);
% splits the month/day cell

65

m=cell2mat(temp2);
s=split(':', m);
n1(i,1)=str2double(S(1));
n1(i,2)=str2double(S(2));
n1(i,3)=str2double(s(1));
n1(i,4)=str2double(s(2));
dataRow=dataRow+1;

%
%
%
%
%
%
%

converts cell to matrix


splits the time cell
storing month value
storing day value
storing hour value
storing minute value
increments dataRow

end
newData=[n1 data];
% new data matrix with hour & min included
%-------------------------------------------------------------------------%-------------------------------------------------------------------------% This program will linearly interpolate missing ESS data and create a new
% matrix containing the old data and interpolated data.
%-------------------------------------------------------------------------%
Initial Constants
%-------------------------------------------------------------------------count1=1;
% 1 of 2 compare counters
count2=2;
% 2 of 2 compare counters
monthCount=4;
% initializing month counter (to April)
dayCount=1;
% initializing day counter
hourCount=0;
% initializing hour counter
minuteCount=0;
% initializing minute counter
finalDataCounter=1; % counter for final data storage matrix
newDataCounter=1;
% counter for time split matrix: newData
%-------------------------------------------------------------------------%
Main Program
%-------------------------------------------------------------------------finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=newDataCounter+1; % incrementing counter
newDataCounter=newDataCounter+1;
% incrementing counter
minuteCount=minuteCount+1;
% incrementing counter
%-------------------------------------------------------------------------% WHILE IN FLIE newData DO...
while count2 < 37027
%---------------------------------------------------------------------% IF SAME MONTH THEN...
if newData(count2,1)==newData(count1,1)
%-----------------------------------------------------------------% IF SAME DAY THEN...
if newData(count2,2)==newData(count1,2)
%-------------------------------------------------------------% IF SAME HOUR THEN...
if newData(count2,3)==newData(count1,3)
%---------------------------------------------------------% IF NO 'HOLE' THEN...
if newData(count2,4)-newData(count1,4)==1
finalData(finalDataCounter,:)=...
newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;
newDataCounter=newDataCounter+1;
minuteCount=minuteCount+1;
count1=count1+1;
% incrementing counter
count2=count2+1;
% incrementing counter
else
%-----------------------------------------------------% ELSE INTERPOLATE 'HOLE'...
n=1;
% initialize temp counter variables
N=1;
% initialize temp counter variables
temp(n,1)=monthCount;
% store time split data in temp
temp(n,2)=dayCount;
% store time split data in temp
temp(n,3)=hourCount;
% store time split data in temp
temp(n,4)=minuteCount; % store time split data in temp

66

%-----------------------------------------------------% INTERPOLATION LOOP...


for i=5:14
temp(n,i)=(newData(count2,i)-newData(count1,i))*...
((minuteCount-newData(count1,4))/...
(newData(count2,4)-newData(count1,4)))+...
newData(count1,i);
end
%-----------------------------------------------------% FILLING MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1;
%-----------------------------------------------------% WHILE 'HOLE(S)' EXIST, INTERPOLATE DATA...
while newData(count2,4)-temp(n,4)~=1
n=n+1;
temp(n,1)=monthCount;
% store time split in temp
temp(n,2)=dayCount;
% store time split in temp
temp(n,3)=hourCount;
% store time split in temp
temp(n,4)=minuteCount; % store time split in temp
%-------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=(newData(count2,i)-temp(N,i))*...
((minuteCount-temp(N,4))/...
(newData(count2,4)-temp(N,4)))+temp(N,i);
end
%-------------------------------------------------% FILL MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1; % increment counter
N=N+1; % increment counter
end
finalData(finalDataCounter,:)=...
newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;
newDataCounter=newDataCounter+1;
% increment counter
minuteCount=minuteCount+1; % increment counter
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
end
%-------------------------------------------------------------% IF NEW HOUR THEN...
else
%---------------------------------------------------------% IF LAST MINUTE OF HOUR ISN'T THERE, THEN ADD IT...
if newData(count1,4)~=59
n=1;
% increment counter
N=1;
% increment counter
minuteCount=59;
temp(n,1)=monthCount;
% store time split data in temp
temp(n,2)=dayCount;
% store time split data in temp
temp(n,3)=hourCount;
% store time split data in temp
temp(n,4)=minuteCount; % store time split data in temp
%-----------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=(newData(count2,i)-newData(count1,i))*...
((minuteCount-newData(count1,4))/...
(60-newData(count1,4)))+newData(count1,i);

67

end
%-----------------------------------------------------% CHECKS AND ADDS PREVIOUS MINS (IF MISSING)...
while temp(N,4)-newData(count1,4)~=1
n=n+1; % increment counter
minuteCount=minuteCount-1; % decrement counter
temp(n,1)=monthCount;
% store temp data
temp(n,2)=dayCount;
% store temp data
temp(n,3)=hourCount;
% store temp data
temp(n,4)=minuteCount; % store temp data
%-------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=(temp(N,i)-newData(count1,i))*...
((minuteCount-newData(count1,4))/...
(temp(N,4)-newData(count1,4)))+...
newData(count1,i);
end
N=N+1; % incrementing counter
end
%-----------------------------------------------------% FILLING MATRIX W/ INTERP DATA BACKWARDS...
while N>0
finalData(finalDataCounter,:)=temp(N,:);
finalDataCounter=finalDataCounter+1;
N=N-1; % incrementing counter
end
end
hourCount=hourCount+1; % incrementing counter
minuteCount=0; %reinitializing minute counter
%---------------------------------------------------------% IF FIRST MIN OF NEW HOUR NOT 0, THEN ADD IT...
if newData(count2,4)~=0
n=1;
% initialize counter
N=1;
% initialize counter
minuteCount=0; % initialize minute counter
temp(n,1)=monthCount;
% store count in temp variable
temp(n,2)=dayCount;
% store count in temp variable
temp(n,3)=hourCount;
% store count in temp variable
temp(n,4)=minuteCount; % store count in temp variable
%-----------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=abs((finalData(finalDataCounter-1,i)-...
newData(count2,i))*((minuteCount-...
newData(count2,4))/(finalData...
(finalDataCounter-1,4)-newData(count2,4)))+...
newData(count2,i));
end
%-----------------------------------------------------% FILLING MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1;
%-----------------------------------------------------% WHILE 'HOLE(S)' EXIST, INTERPOLATE DATA...
while newData(count2,4)-temp(n,4)~=1
n=n+1; % increment counter
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var

68

%-------------------------------------------------% INTERPOLATION LOOP...


for i=5:14
temp(n,i)=(newData(count2,i)-temp(N,i))*...
((minuteCount-temp(N,4))/(newData(...
count2,4)-temp(N,4)))+temp(N,i);
end
%-------------------------------------------------% FILL MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1; % incrementing counter
N=N+1; % incrementing counter
end
end
finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;
newDataCounter=newDataCounter+1;
% incrementing counter
minuteCount=minuteCount+1; % incrementing minute counter
count1=count1+1;
% incrementing counter
count2=count2+1;
% incrementing counter
end
%-----------------------------------------------------------------% IF NEW DAY THEN...
else
%-------------------------------------------------------------% IF LAST MINUTE OF HOUR ISN'T THERE, THEN ADD IT...
if newData(count1,4)~=59
n=1;
% initializing counter
N=1;
% initializing counter
minuteCount=59;
% setting minute to 59
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%---------------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=(newData(count2,i)-newData(count1,i))*...
((minuteCount-newData(count1,4))/...
(60-newData(count1,4)))+newData(count1,i);
end
%---------------------------------------------------------% CHECKS AND ADDS PREVIOUS MINS (IF MISSING)...
while temp(N,4)-newData(count1,4)~=1
n=n+1; % increment counter
minuteCount=minuteCount-1;
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%-----------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=(temp(N,i)-newData(count1,i))*...
((minuteCount-newData(count1,4))/(temp(N,4)-...
newData(count1,4)))+newData(count1,i);
end
N=N+1; % increment counter
end
%---------------------------------------------------------% FILLING MATRIX W/ INTERP DATA BACKWARDS...

69

while N>0
finalData(finalDataCounter,:)=temp(N,:);
finalDataCounter=finalDataCounter+1;
N=N-1;
end
end
hourCount=0;
% reinitializing counter
minuteCount=1;
% reinitializing counter
dayCount=dayCount+1;
% incrementing day counter
%-------------------------------------------------------------% IF FIRST MIN OF NEW HOUR NOT 0, THEN ADD IT...
if newData(count2,4)~=0
n=1;
% reinitializing counter
N=1;
% reinitializing counter
minuteCount=0;
% reinitializing counter
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%---------------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=abs((finalData(finalDataCounter-1,i)-...
newData(count2,i))*((minuteCount-...
newData(count2,4))/(finalData(...
finalDataCounter-1,4)-newData(count2,4)))+...
newData(count2,i));
end
%---------------------------------------------------------% FILLING MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
% increment count
minuteCount=minuteCount+1; % incrementing counter
%---------------------------------------------------------% WHILE 'HOLE(S)' EXIST, INTERPOLATE DATA...
while newData(count2,4)-temp(n,4)~=1
n=n+1; % incrementing counter
temp(n,1)=monthCount;
% store count in temp var
temp(n,2)=dayCount;
% store count in temp var
temp(n,3)=hourCount;
% store count in temp var
temp(n,4)=minuteCount; % store count in temp var
%-----------------------------------------------------% INTERPOLATION LOOP...
for i=5:14
temp(n,i)=(newData(count2,i)-temp(N,i))*...
((minuteCount-temp(N,4))/(newData(...
count2,4)-temp(N,4)))+temp(N,i);
end
%-----------------------------------------------------% FILL MATRIX W/ INTERP DATA/INCREMENT CONST...
finalData(finalDataCounter,:)=temp(n,:);
finalDataCounter=finalDataCounter+1;
minuteCount=minuteCount+1; % increment minute counter
N=N+1; % increment counter
end
minuteCount=minuteCount+1; % increment counter
end
finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;
% increment counter
newDataCounter=newDataCounter+1;
% increment counter
hourCount=0;
% reinitialize counter
count1=count1+1;
% increment counter

70

count2=count2+1;
% increment counter
end
%---------------------------------------------------------------------% IF NEW MONTH THEN...
else
finalData(finalDataCounter,:)=newData(newDataCounter,:);
finalDataCounter=finalDataCounter+1;
% increment counter
newDataCounter=newDataCounter+1;
% increment counter
minuteCount=0;
% reinitialize counter
hourCount=0;
% reinitialize counter
dayCount=1;
% reinitialize counter
monthCount=monthCount+1;
% increment counter
count1=count1+1;
% increment counter
count2=count2+1;
% increment counter
end
end
%-------------------------------------------------------------------------% END OF FILE CLEAN UP...
finalData(finalDataCounter,:)=newData(newDataCounter,:);
%-------------------------------------------------------------------------
%-------------------------------------------------------------------------% Split function used to help evaluate broken down class types
% in MATLAB
%-------------------------------------------------------------------------function l = split(d,s)
%L=SPLIT(S,D) splits a string S delimited by characters in D. Meant to
%
work roughly like the PERL split function (but without any
%
regular expression support). Internally uses STRTOK to do
%
the splitting. Returns a cell array of strings.
%
%Example:
%
>> split('_/', 'this_is___a_/_string/_//')
%
ans =
%
'this'
'is'
'a'
'string'
[]
%
%Written by Gerald Dalley (dalleyg@mit.edu), 2004
l = {};
while (length(s) > 0)
[t,s] = strtok(s,d);
l = {l{:}, t};
end

%-------------------------------------------------------------------------% Interpolation to find the regression coefficients kH, kV, alphaH, and
% alphaV. The "top" and "bottom" values are taken from the ITU data base.
% Greenville - TX frequency.
%-------------------------------------------------------------------------ft=6;
% frequency top (GHz)
f=6.835;
% desired frequency (GHz)
fb=7;
% frequency bottom (GHz)
kHt=0.00175;
% kH top
kHb=0.00301;
% kH bottom
kVt=0.00155;
% kVtop
kVb=0.00265;
% kV bottom
alphaHt=1.308;
% alphaH top
alphaHb=1.332;
% alphaH bottom
alphaVt=1.265;
% alphaV top
alphaVb=1.312;
% alphaV bottom
kH=(kHb-kHt)*((f-ft)/(fb-ft))+kHt;
% kH
kV=(kVb-kVt)*((f-ft)/(fb-ft))+kVt;
% kV
% alphaH
alphaH=(alphaHb-alphaHt)*((f-ft)/(fb-ft))+alphaHt;
% alphaV

71

alphaV=(alphaVb-alphaVt)*((f-ft)/(fb-ft))+alphaVt;
kH;
kV;
alphaH;
alphaV;
%-------------------------------------------------------------------------% ITU rain model
%-------------------------------------------------------------------------theta=0.0272;
% elevation path angle
t=0;
% polarization tilt angle. 0 for horizontal (deg)
% the k-value
k=(kH+kV+(kH-kV)*((cos(theta))^2)*(cos(2*t)))/2;
k;
% the alpha value
alpha=(kH*alphaH+kV*alphaV+(kH*alphaH-kV*alphaV)*((cos(theta))^2)*(...
cos(2*t)))/(2*k);
alpha;
D=24.3815616;
% Greenville to Monticello distance (km)
% calc the atten matrix due to acquired data at GV
for i=1:263520
RR(1,1)=tempRain(i,1);
% actual rain rate for GV (mm/h)
d=35*(exp(-0.015*RR));
% effective path length (km)
r=(1/(1+D/d));
% distance factor
Atten(i,1)=k*(RR^(alpha))*D*r; % attenuation (dB)
end
%-------------------------------------------------------------------------% changing from "rain today" to mm/h
y=RSLandESS_GV(:,9)*25.4;
% Changing rain from in to (mm/h)
tempRain(1,1)=y(1,1);
count2=2;
count1=1;
while count2 < 263520
tempRain(count2,1)=y(count2,1)-y(count1,1);
count2=count2+1;
count1=count1+1;
end
% clearing negative values due to day changes
for i=1:263520
if tempRain(i,1) < 0
tempRain(i,1)=0;
end
end
%-------------------------------------------------------------------------% plot the attenuation
plot(Atten)
title('Predicted Rain Attenuation via ITU Model')
xlabel('Elasped Time (minutes)')
ylabel('Attenuation (dB)')

72

#SITE ONE SCRIPT: GREENVILLE


#!/bin/ksh
#header file
cat /dev/null > ESSRSL1.txt
#clear storage text file
cat /dev/null > ESSRxStatus1.txt
#clear storage text file
echo Day Month Day Time Year RSL >> ESSRSL1.txt
#put info into text file
limit=1
#setting limit
while [ $limit -le 10000 ]
#while limit is less than
do
scancont -region_num 23 -addr 33 -cmd 170 >> ESSRxStatus1.txt
#get RSL data and
awk '$7 == ">>"' ESSRxStatus1.txt | awk '{printf $8}' >> ESSRSL1.txt
#store selected
printf ' ' >> ESSRSL1.txt
#pieces into a
awk '$7 == ">>"' ESSRxStatus1.txt | awk '{printf $9}' >> ESSRSL1.txt
#temporary text
printf ' ' >> ESSRSL1.txt
#file
awk '$7 == ">>"' ESSRxStatus1.txt | awk '{printf $10}' >> ESSRSL1.txt
printf ' ' >> ESSRSL1.txt
awk '$7 == ">>"' ESSRxStatus1.txt | awk '{printf $11}' >> ESSRSL1.txt
printf ' ' >> ESSRSL1.txt
awk '$7 == ">>"' ESSRxStatus1.txt | awk '{printf $12}' >> ESSRSL1.txt
printf ' ' >> ESSRSL1.txt
awk '$5 == "m"' ESSRxStatus1.txt | awk '{printf $4}' >> ESSRSL1.txt
printf '\n' >> ESSRSL1.txt
cat /dev/null > ESSRxStatus1.txt
#clear temp text file
sleep 53
#sleep for 53 seconds
limit=`expr $limit + 1`
#increment limit counter
done
cp ESSRSL1.txt /tools/servermaint/Andrew/Data/RSLgreenvilleMain_`date
+%Y%m%d_%H%M%S`.txt
#time stamp data and save
cat /dev/null > ESSRSL1.txt
#clear temp text file
cat /dev/null > ESSRxStatus1.txt
#clear temp text file
#SITE TWO SCRIPT: LAKE CITY DOT
#!/bin/ksh
cat /dev/null > ESSRSL3.txt
cat /dev/null > ESSRxStatus2.txt
echo Day Month Day Time Year RSL >> ESSRSL3.txt
limit=1
while [ $limit -le 10000 ]
do
scancont -region_num 31 -addr 63 -cmd 170 >> ESSRxStatus2.txt
awk '$7 == ">>"' ESSRxStatus2.txt | awk '{printf $8}' >> ESSRSL3.txt
printf ' ' >> ESSRSL3.txt
awk '$7 == ">>"' ESSRxStatus2.txt | awk '{printf $9}' >> ESSRSL3.txt
printf ' ' >> ESSRSL3.txt
awk '$7 == ">>"' ESSRxStatus2.txt | awk '{printf $10}' >> ESSRSL3.txt
printf ' ' >> ESSRSL3.txt
awk '$7 == ">>"' ESSRxStatus2.txt | awk '{printf $11}' >> ESSRSL3.txt
printf ' ' >> ESSRSL3.txt
awk '$7 == ">>"' ESSRxStatus2.txt | awk '{printf $12}' >> ESSRSL3.txt
printf ' ' >> ESSRSL3.txt

73

awk '$5 == "m"' ESSRxStatus2.txt | awk '{printf $4}' >> ESSRSL3.txt


printf '\n' >> ESSRSL3.txt
cat /dev/null > ESSRxStatus2.txt
sleep 51
limit=`expr $limit + 1`
done
cp ESSRSL3.txt /tools/servermaint/Andrew/Data/RSLlakecityMain_`date
+%Y%m%d_%H%M%S`.txt
cat /dev/null > ESSRSL3.txt
cat /dev/null > ESSRxStatus2.txt
#SITE THREE SCRIPT: SR-222 (GAINESVILLE)
#!/bin/ksh
cat /dev/null > ESSRSL5.txt
cat /dev/null > ESSRSL6.txt
cat /dev/null > ESSRxStatus3.txt
echo Day Month Day Time Year RSL >> ESSRSL5.txt
echo Day Month Day Time Year RSL >> ESSRSL6.txt
limit=1
while [ $limit -le 10000 ]
do
scancont -region_num 35 -addr 67 -cmd 170 >> ESSRxStatus3.txt
awk '$7 == ">>"' ESSRxStatus3.txt | awk '{printf $8}' >> ESSRSL5.txt
printf ' ' >> ESSRSL5.txt
awk '$7 == ">>"' ESSRxStatus3.txt | awk '{printf $9}' >> ESSRSL5.txt
printf ' ' >> ESSRSL5.txt
awk '$7 == ">>"' ESSRxStatus3.txt | awk '{printf $10}' >> ESSRSL5.txt
printf ' ' >> ESSRSL5.txt
awk '$7 == ">>"' ESSRxStatus3.txt | awk '{printf $11}' >> ESSRSL5.txt
printf ' ' >> ESSRSL5.txt
awk '$7 == ">>"' ESSRxStatus3.txt | awk '{printf $12}' >> ESSRSL5.txt
printf ' ' >> ESSRSL5.txt
awk '$5 == "m"' ESSRxStatus3.txt | awk '{printf $4}' >> ESSRSL5.txt
printf '\n' >> ESSRSL5.txt
cat /dev/null > ESSRxStatus3.txt
sleep 47
limit=`expr $limit + 1`
done
cp ESSRSL5.txt /tools/servermaint/Andrew/Data/RSLsr222Main_`date
+%Y%m%d_%H%M%S`.txt
cat /dev/null > ESSRSL5.txt
cat /dev/null > ESSRSL6.txt
cat /dev/null > ESSRxStatus3.txt

74

APPENDIX B
DEVICE SPECIFICATIONS AND DATASHEETS

75

76

77

78

79

80

81

82

83

84

85

86

87

BIBLIOGRAPHY

[1]

J. S. Seybold, Introduction to RF Propagation, 2005, John Wiley and Sons, Inc.

[2]

R. K. Crane, Electromagnetic Wave Propagation Through Rain, 1996, John Wiley and
Sons, Inc.

[3]

A. F. Molisch, Wireless Communications, 2005, John Wiley and Sons, LTD.

[4]

ITU-R Recommendations, Characteristics of precipitation for propagation modeling,


ITU-R P.837-1, Geneva, 1994.

[5]

ITU-R Recommendations, Characteristics of precipitation for propagation modeling,


ITU-R P.837-3, Geneva, 2001.

[6]

ITU-R Recommendations, Propagation data and prediction methods required for the
design of terrestrial line of sight systems, ITU-R P.530-9, Geneva, 2001.

[7]

L. W. Couch, Digital and Analog Communication Systems (Seventh Edition), 2007,


Pearson and Prentice Hall companies.

[8]

B Sklar, Digital Communications: Fundamentals and Applications (2nd Edition), 2001,


Pearson and Prentice Hall companies.

[9]

A. C. Bostwick, et al., DVM6 Excell Digital VersaT1lity Microwave Radio (Volume II:
Reference Manual), 1997, Harris Corporation Farinon Division.

[10]

Columbia Weather Systems, Inc., CapricornTM 2000 Weather Station User Manual,
2005, http://www.columbiaweather.com/cap2000.html.

[11]

D. C. Montgomery, G C Runger, Applied Statistics and Probability for Engineers


(Fourth Edition), 2007, John Wiley and Sons, Inc.

[12]

G. Strand, T. Nguyen, Wavelets and Filter Banks, 1996, Wellesley Cambridge Press.

[13]

A. B. Carlson, P B Crilly, J C Rutledge, Communication Systems (Fourth Edition), 2002,


McGraw-Hill Companies, Inc.

[14]

J. J. Leader, Numerical Analysis and Scientific Computation, 2004, Pearson-Addison


Wesley.

[15]

R. Polikar, The Wavelet Tutorial Part I: Fundamental Concepts & an Overview of the
Wavelet Theory, 1996, http://users.rowan.edu/~polikar/wavelets/wtpart1.html.
88

[16]

R. Polikar, The Wavelet Tutorial Part II: The Fourier Transform & The Short Term
Fourier Transform, 1996, http://www.site.uottawa.ca/~qingchen/wavelet/p2.htm.

[17]

R. Polikar, The Wavelet Tutorial Part III: Multiresolution Analysis & the Continuous
Wavelet Transform, 1996,https://ccrma.stanford.edu/~unjung/mylec/WTpart3.html.

[18]

R. Polikar, The Wavelet Tutorial Part IV: Multiresolution Analysis: The Discrete Wavelet
Transform, 1996, http://users.rowan.edu/~polikar/WAVELETS/WTpart4.html.

[19]

G. Feldhake, T. Russell, Statistical Combination of Attenuations by Multiple Atmospheric


Effects, Antenna and Propagation Society International Symposium, P 1864-1867, 1996.

[20]

C. E. Mayer, B. E. Jaeger, Rain Attenuation Model Comparison and Validation, Online


Journal of Space Communications, Issue No. 2, 2002

[21]

L. Feral, J. Lemorton, L. Castanet, H. Sauvageot, Hycell - A new hybrid model of the rain
horizontal distribution for propagation studies, Radio Science, vol. 38, 2003.

[22]

ITU-R, Characteristics of precipitation for propagation modelling, ITU-R


Recommendations P 837-3, Supplement 1 to Vol. 2000, P series, part 1 and 2, 2001.

[23]

M. E. Thomas, Optical propagation in linear media: atmospheric gases and particles,


solid state components, and water, 2006, Oxford University Press.

89

BIOGRAPHICAL SKETCH

Andrew J. Stringer was born in Tallahassee, FL in 1983. He attended Florida State University
and received his Bachelor of Science in Electrical Engineering in 2007.

During his final

Semester as an undergraduate, Andrew worked for RCC Consultants, Inc. as an in-house


telecommunications consultant at the Florida Department of Transportation. Andrew returned to
Florida State University and received his Master of Science in Electrical Engineering in 2010.
Mr. Stringers research interests include atmospheric attenuation models and path loss models
for communication systems, next generation communication systems, and narrow banding.

90

You might also like