Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to...
Transcript of Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to...
![Page 1: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/1.jpg)
Modelling and Simulation
of
GPS Multipath Propagation
Bruce M Hannah
![Page 2: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/2.jpg)
Modelling and Simulation
of
GPS Multipath Propagation
Bruce M. Hannah
B.Eng. (Hons)
The Cooperative Research Centre for Satellite Systems
Queensland University of Technology
THIS DISSERTATION IS SUBMITTED IN PARTIAL FULFILMENT
OF THE REQUIREMENTS FOR THE AWARD OF THE DEGREE
DOCTOR OF PHILOSOPHY
March 2001
![Page 3: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/3.jpg)
![Page 4: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/4.jpg)
![Page 5: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/5.jpg)
![Page 6: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/6.jpg)
v
Statement of Authorship
The work contained in this thesis has not been previously submitted for a degree or
diploma at any other higher education institution. To the best of my knowledge and
belief, the thesis contains no material previously published or written by another
person except where due reference is made.
Signed:...................................................
Date: ......................................................
![Page 7: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/7.jpg)
vi
Key Words
Global Positioning System, GPS, Multipath, Radio Frequency, RF, Propagation,
Parabolic Equation, PE, Modelling, Simulation, Correlation, DLL, Discrimination,
Range Error, Mitigation, Reflection, Diffraction, Fresnel Zone.
![Page 8: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/8.jpg)
vii
Acknowledgements
This work was carried out in the Cooperative Research Centre for Satellite Systems
with financial support from the Commonwealth of Australia’s Cooperative Research
Centres Program. Other work presented in this dissertation was funded in part by the
British Council under a postgraduate bursary scheme. Additional funding was
provided by the QUT grant-in-aid scheme. The author thanks all for their financial
assistance.
Special thanks to Professor Kurt Kubik and Professor Miles Moody for providing
insightful respective supervision of this research work within the Space Centre for
Satellite Navigation and the Cooperative Research Centre for Satellite Systems.
The Radio Communications Research Unit at Rutherford Appleton Laboratory
Oxfordshire, UK, specifically Dr Mireille Levy for inspiring and maintaining the use
of the Parabolic Equation as an effective GPS propagation modelling tool.
Dr Rodney Walker of the Queensland University of Technology for his interest,
insight, inspiration and most all his friendship.
To my parents Mervyn and Maureen, my wife Tanya and her parents, Barry and Rae,
all of whom I owe a great personal debt for the investment they have made in me.
This dissertation is dedicated to them.
![Page 9: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/9.jpg)
viii
![Page 10: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/10.jpg)
ix
Abstract
Multipath remains a dominant error source in Global Positioning System (GPS)
applications that require high accuracy. With the use of differential techniques it is
possible to remove many of the common-mode error sources, but the error effects of
multipath have proven much more difficult to mitigate. The research aim of this work
is to enhance the understanding of multipath propagation and its effects in GPS
terrestrial applications, through the modelling of signal propagation behaviour and the
resultant error effects.
Multipath propagation occurs when environmental features cause combinations of
reflected and/or diffracted replica signals to arrive at the receiving antenna. These
signals, in combination with the original line-of-sight (LOS) signal, can cause
distortion of the receiver correlation function and ultimately the discrimination
function and hence errors in range estimation.
To date, a completely satisfactory mitigation strategy has yet to be developed. In the
search for such a mitigation strategy, it is imperative that a comprehensive
understanding of the multipath propagation environment and the resultant error effects
exists. The work presented here, provides a comprehensive understanding through the
use of new modelling and simulation techniques specific to GPS multipath.
This dissertation unites the existing theory of radio frequency propagation for the GPS
L1 signal into a coherent treatment of GPS propagation in the terrestrial environment.
To further enhance the understanding of the multipath propagation environment and
the resultant error effects, this dissertation also describes the design and development
of a new parabolic equation (PE) based propagation model for analysis of GPS
multipath propagation behaviour.
The propagation model improves on previous PE-based models by incorporating
terrain features, including boundary impedance properties, backscatter and time-
domain decomposition of the field into a multipath impulse response. The results
![Page 11: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/11.jpg)
x
provide visualisation as well as the defining parameters necessary to fully describe the
multipath propagation behaviour.
These resultant parameters provide the input for a correlation and discrimination
model for visualisation and the generation of resultant receiver error measurements.
Results for a variety of propagation environments are presented and the technique is
shown to provide a deterministic methodology against real GPS data.
The unique and novel combined modelling of multipath propagation and reception,
presented in this dissertation, provides an effective set of tools that have enhanced the
understanding of the behaviour and effect of multipath in GPS applications, and
ultimately should aid in providing a solution to the GPS multipath mitigation
problem.
![Page 12: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/12.jpg)
xi
Table of Contents
Chapter 1 Introduction and Overview.................................................................... 1
1.1 Introduction...................................................................................................... 1
1.2 Overview of Research Presented ..................................................................... 2
1.3 Research Contribution ..................................................................................... 5
1.4 References........................................................................................................ 6
Chapter 2 Nature of GPS Multipath Propagation ................................................. 9
2.1 Multipath Environment ..................................................................................... 9
2.2 Specular Reflection ......................................................................................... 10
2.2.1 Linear Reflection Coefficient Representation......................................... 10
2.2.2 Circular Reflection Coefficient Representation ...................................... 16
2.2.3 Ray-based Reflection Geometry: Relative Time Delay and Phase ......... 27
2.2.3.1 Summary of Amplitudes, Relative Delays and Phases ....................... 31
2.2.4 Multipath Modes and Coupled Reflection Coefficients ......................... 32
2.3 Rough Surface Scatter..................................................................................... 45
2.4 Fresnel Zones .................................................................................................. 49
2.5 Diffraction....................................................................................................... 54
2.5.1 Knife-Edge Diffraction ........................................................................... 55
2.6 GPS Fading Signal Characteristics ................................................................. 59
2.6.1 Signal Fade Envelopes ............................................................................ 59
2.6.2 Signal Fading Characteristics for Specular Reflection ........................... 72
2.7 GPS Receiver Context .................................................................................... 80
2.7.1 Aspects of Physical Antenna Location.................................................... 80
2.8 Path Loss in the Terrestrial Domain................................................................ 86
2.9 Summary ......................................................................................................... 87
2.10 References ................................................................................................... 90
Chapter 3 Overview of Propagation Modelling ................................................... 93
3.1 Review of Maxwell’s Equations.................................................................... 93
3.2 Overview of Computational Electromagnetics Techniques .......................... 96
3.2.1 Finite-Elements Technique ..................................................................... 96
![Page 13: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/13.jpg)
xii
3.2.2 Finite-Difference Time-Domain Technique............................................ 98
3.2.3 Finite-Difference Frequency-Domain Technique ................................. 102
3.2.4 Method of Moments .............................................................................. 102
3.2.5 Geometrical and Uniform Theory of Diffraction .................................. 104
3.2.6 Generalised Multipole Technique ......................................................... 105
3.2.7 Parabolic Equation Method................................................................... 106
3.3 Comparison of Modelling Techniques......................................................... 109
3.3.1 Requirements for Modelling GPS Signal Propagation ......................... 110
3.3.2 Comparison of EM Techniques............................................................. 110
3.4 References.................................................................................................... 112
Chapter 4 GPS Parabolic Equation Model ......................................................... 115
4.1 Development of the PE for Electromagnetic Propagation ............................ 115
4.2 The Free-Space Parabolic Equation .............................................................. 116
4.3 Limitations of Refractive Index Terms in PE Forms .................................... 119
4.3.1 Fourier Split-Step Solution of the SPE ................................................. 123
4.3.2 Phase Errors in Rational-Linear Approximation Forms of the PE........ 126
4.4 Numerical Implementation for GPS Satellite Propagation ........................... 128
4.4.1 Domain Sampling.................................................................................. 129
4.4.2 Incident Boundary Condition ................................................................ 130
4.4.3 Upper Boundary Condition ................................................................... 131
4.4.4 Lower Boundary Condition................................................................... 132
4.4.5 Implementation Algorithm.................................................................... 134
4.5 Implementing Arbitrary Terrain in the PE Model ......................................... 135
4.5.1 Boundary-Shift Technique for Arbitrary Terrain .................................. 136
4.6 Implementing Backscatter for a Two-Way PE Model .................................. 138
4.7 Summary ....................................................................................................... 141
4.8 References ..................................................................................................... 143
Chapter 5 Time Series Analysis with the FSPE.................................................. 149
5.1 Introduction.................................................................................................. 149
5.2 Implementation ............................................................................................ 150
5.3 Domain Considerations................................................................................ 155
![Page 14: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/14.jpg)
xiii
5.4 Deriving the MCIR from the PETA Model ................................................. 157
5.5 Multipath Phase Information from the PETA Model .................................. 159
5.6 Antenna Gain Pattern from Angle of Arrival Information .......................... 161
5.7 PETA Implementation Issues ...................................................................... 164
5.8 PETA Domain Representation and Performance ........................................ 166
5.9 Summary...................................................................................................... 168
5.10 References.................................................................................................... 169
Chapter 6 Model Validation................................................................................. 171
6.1 Validation of FSPE with an Exact Solution.................................................. 171
6.1.1 Phase Error ............................................................................................ 171
6.2 Forward Multipath Propagation .................................................................... 174
6.2.1 Static Test.............................................................................................. 174
6.2.2 Dynamic Tests....................................................................................... 177
6.3 Forward Diffraction ...................................................................................... 181
6.3.1 Static Test.............................................................................................. 181
6.3.2 Dynamic Test ........................................................................................ 185
6.4 BA/BB-Mode: Backscatter ........................................................................... 187
6.4.1 Static Test.............................................................................................. 188
6.4.2 Dynamic Test ........................................................................................ 192
6.5 Summary ....................................................................................................... 199
6.6 References ..................................................................................................... 200
Chapter 7 Simulation of GPS Propagation......................................................... 203
7.1 Introduction ................................................................................................... 203
7.2 Candidate Simulation Cases.......................................................................... 203
7.2.1 Pine Dam............................................................................................... 204
7.2.2 Caboolture Soccer Field........................................................................ 214
7.3 Summary ....................................................................................................... 220
7.4 References ..................................................................................................... 221
Chapter 8 SNR for Deriving a Height Observable............................................. 223
8.1 Introduction ................................................................................................... 223
![Page 15: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/15.jpg)
xiv
8.2 Estimating Antenna Height from the GPS S/N Ratio ................................... 225
8.3 Results ........................................................................................................... 228
8.4 North Pine Dam............................................................................................. 229
8.5 Bribie Island .................................................................................................. 230
8.6 System Aspects ............................................................................................. 230
8.7 Summary ....................................................................................................... 236
8.8 References ..................................................................................................... 237
Chapter 9 Receiver Correlation and Discrimination ......................................... 239
9.1 Introduction ................................................................................................... 239
9.2 Fundamentals of a GPS Receiver Model ...................................................... 240
9.2.1 Generic GPS Receiver Functions .......................................................... 241
9.2.1.1 Antenna and RF Section.................................................................... 241
9.2.1.2 Reference Oscillator and Frequency Synthesis ................................. 242
9.2.1.3 Down-conversion and IF ................................................................... 242
9.2.1.4 Signal Processing .............................................................................. 242
9.3 The Key GPS Receiver Elements.................................................................. 243
9.3.1 PN Code Generation.............................................................................. 243
9.3.2 Delay Lock Loop................................................................................... 245
9.4 A Receiver Model for use with the FSPE/PETA Model............................... 252
9.5 The Modelled Code Correlation Function and Multipath............................. 254
9.5.1 Variation of Relative Multipath Delay Time ........................................ 254
9.5.1.1 In-Phase Case .................................................................................... 254
9.5.1.2 Anti-phase Case................................................................................. 258
9.5.2 Variation of Relative Phase................................................................... 261
9.6 Error Envelopes for a Single multipath Signal.............................................. 265
9.6.1 Variation of Relative Time Delay ......................................................... 265
9.6.2 Variation of Relative Amplitude ........................................................... 268
9.6.3 Variation of Phase ................................................................................. 269
9.7 Error Envelopes for Two Multipath Signals ................................................. 270
9.8 Summary ....................................................................................................... 275
9.9 References ..................................................................................................... 276
![Page 16: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/16.jpg)
xv
Chapter 10 Conclusions.......................................................................................... 279
Chapter 11 Recommendations ............................................................................... 281
11.1 References ................................................................................................. 283
List of Appendices
Appendix A – Research Publications.................................................................... A-1
Appendix B – Raw GPS Multipath Data.............................................................. B-1
B.1 Data Results Fresh Water - North Pine Dam 2 December 1999....................B-1
(a) SV 5............................................................................................................B-1
(b) SV 6............................................................................................................B-2
(c) SV 8............................................................................................................B-2
(d) SV 9............................................................................................................B-3
(e) SV 10..........................................................................................................B-3
(f) SV 17..........................................................................................................B-4
(g) SV 24..........................................................................................................B-4
(h) SV 25..........................................................................................................B-5
(i) SV 26..........................................................................................................B-5
(j) SV 30..........................................................................................................B-6
B.2 Data Results Soil - Caboolture 30 November 1999.......................................B-7
(a) SV 5............................................................................................................B-7
(b) SV 6............................................................................................................B-8
(c) SV 8............................................................................................................B-8
(d) SV 17..........................................................................................................B-9
(e) SV 21..........................................................................................................B-9
(f) SV 23........................................................................................................B-10
(g) SV 26........................................................................................................B-10
(h) SV 30........................................................................................................B-11
B.3 Data Results Sea Water - Bribie Island 11 November 1999 ........................B-12
(a) SV 3..........................................................................................................B-12
(b) SV 6..........................................................................................................B-13
![Page 17: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/17.jpg)
xvi
(c) SV 10........................................................................................................B-13
(d) SV 17........................................................................................................B-14
(e) SV 21........................................................................................................B-14
(f) SV 22........................................................................................................B-15
(g) SV 23........................................................................................................B-15
(h) SV 30........................................................................................................B-16
Appendix C – MATLAB Code............................................................................... C-1
C.1 Propagation Modelling Code .........................................................................C-1
(a) GOPE.M.....................................................................................................C-1
(b) MPE.M.......................................................................................................C-7
(c) FIELDCALC.M........................................................................................C-22
(d) PETASETUP.M.......................................................................................C-26
(e) LOADPROFILE.M ..................................................................................C-27
C.2 GPS Receiver Modelling..............................................................................C-29
(a) RUNCORR.M..........................................................................................C-29
(b) RXCORR.M.............................................................................................C-32
(c) PNGEN.M................................................................................................C-37
C.3 GPS Multipath Data Acquisition .................................................................C-39
(a) ASHTEQC.M...........................................................................................C-39
(b) RIN2QC.M...............................................................................................C-41
(c) QC2MAT.M.............................................................................................C-43
(d) PLOTQC.M..............................................................................................C-46
![Page 18: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/18.jpg)
xvii
List of Figures
Figure 2.1 — Multipath environment ........................................................................... 9
Figure 2.2 — Linear reflection coefficients for Concrete ........................................... 12
Figure 2.3 — Linear reflection coefficients for Dry Ground ...................................... 13
Figure 2.4 — Linear reflection coefficients for Medium Dry Ground........................ 13
Figure 2.5 — Linear reflection coefficients for Wet Ground...................................... 14
Figure 2.6 — Linear reflection coefficients for Fresh Water...................................... 14
Figure 2.7 — Linear reflection coefficients for Sea Water......................................... 15
Figure 2.8 — Circular reflection coefficients for Concrete ........................................ 17
Figure 2.9 — Circular reflection coefficients for Dry Ground ................................... 17
Figure 2.10 — Circular reflection coefficients for Medium Dry Ground................... 18
Figure 2.11 — Circular reflection coefficients for Wet Ground................................. 18
Figure 2.12 — Circular reflection coefficients for Fresh Water ................................. 19
Figure 2.13 — Circular reflection coefficients for Sea Water .................................... 19
Figure 2.14 — Incident RHCP-RC for Concrete ........................................................ 21
Figure 2.15 — Incident RHCP-RC for Dry Ground ................................................... 22
Figure 2.16 — Incident RHCP-RC for Medium Dry Ground..................................... 22
Figure 2.17 — Incident RHCP-RC for Wet Ground................................................... 23
Figure 2.18 — Incident RHCP-RC for Fresh Water................................................... 23
Figure 2.19 — Incident RHCP-RC for Sea Water ...................................................... 24
Figure 2.20 — Incident LHCP-RC for Concrete ........................................................ 25
Figure 2.21 — Incident LHCP-RC for Dry Ground.................................................... 25
Figure 2.22 — Incident LHCP-RC for Medium Dry Ground..................................... 26
Figure 2.23 — Incident LHCP-RC for Wet Ground................................................... 26
Figure 2.24 — Incident LHCP-RC for Fresh Water ................................................... 27
Figure 2.25 — Incident LHCP-RC for Sea Water ...................................................... 27
Figure 2.26 — Forward scatter geometry ................................................................... 28
Figure 2.27 — Backscatter geometry 1....................................................................... 29
Figure 2.28 — Backscatter geometry 2....................................................................... 31
Figure 2.29 — Decoupled Polarisation Zone 1 BB-Mode.......................................... 33
Figure 2.30 — Coupled Polarisation Zone 1 BB-Mode.............................................. 34
Figure 2.31 — Modified Coupled RC Zone 1 Concrete............................................. 37
![Page 19: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/19.jpg)
xviii
Figure 2.32 — Modified Coupled RC Zone 1 Dry Ground ........................................ 37
Figure 2.33 — Modified Coupled RC Zone 1 Medium Dry Ground.......................... 38
Figure 2.34 — Modified Coupled RC Zone 1 Wet Ground........................................ 38
Figure 2.35 — Modified Coupled RC Zone 1 Fresh Water........................................ 39
Figure 2.36 — Modified Coupled RC Zone 1 Sea Water ........................................... 39
Figure 2.37 — Modified Coupled RC Zone 2 Concrete ............................................. 40
Figure 2.38 — Modified Coupled RC Zone 2 Dry Ground ........................................ 40
Figure 2.39 — Modified Coupled RC Zone 2 Medium Dry Ground.......................... 41
Figure 2.40 — Modified Coupled RC Zone 2 Wet Ground........................................ 41
Figure 2.41 — Modified Coupled RC Zone 2 Sea Water ........................................... 42
Figure 2.42 — Modified Coupled RC Zone 2 Fresh Water........................................ 42
Figure 2.43 — Resultant RC BB-Mode Concrete....................................................... 43
Figure 2.44 — Resultant RC BB-Mode Dry Ground.................................................. 43
Figure 2.45 — Resultant RC BB-Mode Medium Dry Ground ................................... 44
Figure 2.46 — Resultant RC BB-Mode Wet Ground ................................................. 44
Figure 2.47 — Resultant RC BB-Mode Fresh Water ................................................. 45
Figure 2.48 — Resultant RC BB-Mode Sea Water..................................................... 45
Figure 2.49 — Surface roughness geometry ............................................................... 46
Figure 2.50 — Rayleigh Roughness Criterion ............................................................ 47
Figure 2.51 — Rough Surface Reduction Factor ........................................................ 48
Figure 2.52 — Fresnel Zones for Reflection............................................................... 50
Figure 2.53 — First Fresnel Zone Dimensions (1-5 degs).......................................... 51
Figure 2.54 — First Fresnel Zone Dimensions (5-10 degs)........................................ 51
Figure 2.55 — First Fresnel Zone Dimensions (10-90 degs)..................................... 52
Figure 2.56 — Fresnel Zones for LOS........................................................................ 53
Figure 2.57 — Diffraction at Obstruction................................................................... 54
Figure 2.58 — Diffraction at Obstacle........................................................................ 55
Figure 2.59 — Knife-edge Diffraction Gain vs v........................................................ 57
Figure 2.60 — Diffraction Parameter and Gain .......................................................... 57
Figure 2.61 — Time of Arrival Error.......................................................................... 58
Figure 2.62 — Concrete F-mode................................................................................. 60
Figure 2.63 — Dry Ground F-mode............................................................................ 60
Figure 2.64 — Medium Dry Ground F-mode ............................................................. 61
![Page 20: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/20.jpg)
xix
Figure 2.65 — Wet Ground F-mode........................................................................... 61
Figure 2.66 — Fresh Water F-mode ........................................................................... 62
Figure 2.67 — Sea Water F-mode .............................................................................. 62
Figure 2.68 — Concrete BA-mode ............................................................................. 63
Figure 2.69 — Concrete2 BB-mode............................................................................ 64
Figure 2.70 — Dry Ground2 BB-mode ....................................................................... 64
Figure 2.71 — Medium Dry Ground2 BB-mode......................................................... 65
Figure 2.72 — Wet Ground2 BB-mode....................................................................... 65
Figure 2.73 — Concrete to Sea Water ........................................................................ 66
Figure 2.74 — Dry Ground to Sea Water ................................................................... 66
Figure 2.75 — Medium Dry Ground to Sea Water..................................................... 67
Figure 2.76 — Concrete to Fresh Water ..................................................................... 67
Figure 2.77 — Dry Ground to Fresh Water ................................................................ 68
Figure 2.78 — Medium Dry Ground to Sea Water..................................................... 68
Figure 2.79 — Wet Ground to Concrete..................................................................... 69
Figure 2.80 — Wet Ground to Dry Ground................................................................ 69
Figure 2.81 — Wet Ground to Medium Dry Ground ................................................. 70
Figure 2.82 — Medium DryGround to Concrete........................................................ 70
Figure 2.83 — Medium Dry Ground to Dry Ground .................................................. 71
Figure 2.84 — Dry Ground to Concrete ..................................................................... 71
Figure 2.85 — Linear Variation of Propagation Angle (Fwd).................................... 73
Figure 2.86 — Linear Variation (Reduced gradient) .................................................. 73
Figure 2.87 — Variation of Antenna Height............................................................... 74
Figure 2.88 — Backscatter from above ...................................................................... 75
Figure 2.89 — Equal Antenna Distance and Height................................................... 76
Figure 2.90 — Antenna Height > Distance................................................................. 77
Figure 2.91 — Antenna Height < Distance................................................................. 77
Figure 2.92 — Addition of Multipath Modes ............................................................. 78
Figure 2.93 — Variation of Relative Multipath Amplitude........................................ 79
Figure 2.94 — BA-mode Decorrelation Distance/Height Bound ............................... 82
Figure 2.95 — BA-mode Existence-Boundary........................................................... 83
Figure 2.96 — BA-mode Existence Region Metrics .................................................. 83
Figure 2.97 — BB-mode Existence Region................................................................ 84
![Page 21: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/21.jpg)
xx
Figure 2.98 — BB-Mode Decorrelation Distance....................................................... 85
Figure 2.99 — BB-mode Antenna Height Bounds for Decorrelation......................... 86
Figure 2.100 — Additional Path Loss......................................................................... 87
Figure 3.1 — 3-D FD-TD grid .................................................................................. 102
Figure 3.2 — PE solution domain............................................................................. 109
Figure 4.1 — Q-functions ......................................................................................... 127
Figure 4.2 — Phase error with common approximates............................................. 128
Figure 4.3 — FSPE Solution Domain....................................................................... 129
Figure 4.4 — Initial Field.......................................................................................... 131
Figure 4.5 — Upper Absorption Region................................................................... 132
Figure 4.6 — Reflection Coefficient in P-Space....................................................... 133
Figure 4.7 — Implementation Domain ..................................................................... 134
Figure 4.8 — Implementation Algorithm.................................................................. 135
Figure 4.9 — Solution Domain Representation ........................................................ 137
Figure 4.10 — Boundary Shift Technique ................................................................ 137
Figure 4.11 — Boundary-Shift Algorithm ................................................................ 138
Figure 4.12 — Backscatter implementation.............................................................. 140
Figure 4.13 — Backscatter Implementation Algorithm............................................ 141
Figure 5.1 — SINC pulse.......................................................................................... 150
Figure 5.2 — Input spectrum .................................................................................... 152
Figure 5.3 — Propagation domain ............................................................................ 155
Figure 5.4 — Corrections for spatial time reference................................................. 157
Figure 5.5 — Time delay estimation error ................................................................ 159
Figure 5.6 — LOS Phase Estimation Error............................................................... 160
Figure 5.7 — Forward Multipath Phase Estimation Error ........................................ 161
Figure 5.8 — LOS AOA Estimation Error................................................................ 163
Figure 5.9 — Forward Multipath AOA Estimation Error......................................... 163
Figure 5.10 — Non-aliased time response ................................................................ 165
Figure 5.11 — Aliased time response ....................................................................... 165
Figure 5.12 — Domain representation...................................................................... 166
![Page 22: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/22.jpg)
xxi
Figure 6.1 — Phase error in Narrow-Angle SPE...................................................... 172
Figure 6.2 — Phase error in FSPE............................................................................ 173
Figure 6.3 — FSPE Field vs Exact Solution............................................................. 174
Figure 6.4 — Field over flat perfect conductor......................................................... 175
Figure 6.5 — User Received Minimum L1 C/A Signal Level.................................. 176
Figure 6.6 — Time series for forward propagation .................................................. 176
Figure 6.7 — Dynamic situation............................................................................... 177
Figure 6.8 — FSPE field at 8 degrees....................................................................... 178
Figure 6.9 — PETA result at 8 degrees .................................................................... 178
Figure 6.10 — FSPE field and reconstructed PETA field comparison..................... 179
Figure 6.11 — Comparison of RC Magnitudes ........................................................ 180
Figure 6.12 — Fade Pattern Comparison.................................................................. 180
Figure 6.13 — Diffraction over terrain element ....................................................... 181
Figure 6.14 — Diffraction geometry......................................................................... 182
Figure 6.15 — Time series no terrain ....................................................................... 183
Figure 6.16 — Time series with terrain .................................................................... 184
Figure 6.17 — Dynamic diffraction situation ........................................................... 185
Figure 6.18 — FSPE field at 10 degrees.................................................................. 185
Figure 6.19 — FSPE field 5 degrees to 45 degrees .................................................. 186
Figure 6.20 — LOS and diffracted propagation time comparison............................ 187
Figure 6.21 — Forward propagation over terrain ..................................................... 188
Figure 6.22 — Back propagation from reflected interfaces...................................... 189
Figure 6.23 — Total propagated field....................................................................... 189
Figure 6.24 — Two-way field with 20 m high vertical reflector at 20m.................. 190
Figure 6.25 — Time series back-propagation........................................................... 190
Figure 6.26 — Model domain................................................................................... 191
Figure 6.27 — Stepped backscatter geometry........................................................... 192
Figure 6.28 — Stepped backscatter PETA results for 5 degrees .............................. 193
Figure 6.29 — Stepped backscatter PETA results for 12.5 degrees ......................... 194
Figure 6.30 — Stepped backscatter PETA results for 15 degrees ............................ 194
Figure 6.31 — Stepped backscatter PETA results for 5 to 15 degrees ..................... 195
Figure 6.32 — Stepped backscatter knife-edge geometry......................................... 195
![Page 23: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/23.jpg)
xxii
Figure 6.33 — Diffraction loss over B1 interface ..................................................... 196
Figure 6.34 — Normalised diffraction loss PETA vs CCIR for B1 interface........... 197
Figure 6.35 — PETA path delays ............................................................................. 198
Figure 6.36 — Reconstructed PETA field and FSPE field comparison ................... 198
Figure 6.37 — Reconstructed GO field and FSPE field comparison........................ 199
Figure 7.1 — Pine Dam Data Collection Site ........................................................... 204
Figure 7.2 — Data Collection Basis.......................................................................... 205
Figure 7.3 — Location Orientation ........................................................................... 205
Figure 7.4 — SV17 Results....................................................................................... 207
Figure 7.5 — SV17 AZ-EL Data .............................................................................. 207
Figure 7.6 — SV17 Fresnel Data .............................................................................. 208
Figure 7.7 — SV6 Results......................................................................................... 209
Figure 7.8 — SV6 AZ-EL Data ................................................................................ 210
Figure 7.9 — SV6 Fresnel Data ................................................................................ 210
Figure 7.10 — SV8 Results....................................................................................... 211
Figure 7.11 — SV8 AZ-EL Data .............................................................................. 211
Figure 7.12 — SV8 Fresnel Data .............................................................................. 212
Figure 7.13 — SV9 Results....................................................................................... 213
Figure 7.14 — SV9 AZ-EL Data .............................................................................. 213
Figure 7.15 — SV9 Fresnel Data .............................................................................. 214
Figure 7.16 — Caboolture Data Collection Site ....................................................... 215
Figure 7.17 — SV21 Results..................................................................................... 215
Figure 7.18 — SV21 AZ-EL Data ............................................................................ 216
Figure 7.19 — SV21 Fresnel Data ............................................................................ 216
Figure 7.20 — SV23 Results..................................................................................... 217
Figure 7.21 — SV23 AZ-EL Data ............................................................................ 217
Figure 7.22 — SV23 Fresnel Data ............................................................................ 218
Figure 7.23 — SV5 Results....................................................................................... 218
Figure 7.24 — SV5 AZ-EL Data .............................................................................. 219
Figure 7.25 — SV5 Fresnel Data .............................................................................. 219
Figure 8.1 — SV21 Bribie Island Tidal Variation .................................................... 223
![Page 24: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/24.jpg)
xxiii
Figure 8.2 — Relative Height Measurement ............................................................ 225
Figure 8.3 — Sampled received S/N......................................................................... 226
Figure 8.4 — Sampled elevation angle data ............................................................. 226
Figure 8.5 — Interference pattern ............................................................................. 228
Figure 8.6 — Height estimation above fresh water .................................................. 229
Figure 8.7 — Height estimation above sea surface................................................... 230
Figure 8.8 — System representation ......................................................................... 231
Figure 8.9 — Raw S/N sea water.............................................................................. 232
Figure 8.10 — Raw S/N fresh water......................................................................... 232
Figure 8.11 — Height Error Bound 0.1 Deg Error ................................................... 234
Figure 8.12 — Height Error Bound 0.01 Deg Error ................................................. 234
Figure 9.1 — Multipath Modelling Environment ..................................................... 239
Figure 9.2 — General form of GPS user equipment................................................. 240
Figure 9.3 — Generic GPS receiver.......................................................................... 241
Figure 9.4 — C/A Code Generation ......................................................................... 244
Figure 9.5 — C/A autocorrelation function .............................................................. 246
Figure 9.6 — Receiver tracking loops ...................................................................... 247
Figure 9.7 — Correlation process ............................................................................. 248
Figure 9.8 — DLL discriminator curves................................................................... 250
Figure 9.9 — Receiver Correlation Model ............................................................... 253
Figure 9.10 — DLL discriminator curves................................................................. 253
Figure 9.11 — Multipath-free Correlation and Discrimination ................................ 255
Figure 9.12 — 0.05 Chip delay In-phase Multipath.................................................. 256
Figure 9.13 — 0.5 chip delay In-phase Multipath .................................................... 257
Figure 9.14 — Correlation Distortion (0°)................................................................ 257
Figure 9.15 — Short-delay Multipath (180°)............................................................ 259
Figure 9.16 — Long-delay Multipath (180°) ............................................................ 260
Figure 9.17 — Correlation Distortion (180°)............................................................ 260
Figure 9.18 — 200ns relative delay , 0.5 Multipath Power Ratio (0°) ..................... 262
Figure 9.19 — 200ns relative delay, 0.5 Multipath Power Ratio (60°) .................... 262
Figure 9.20 — 200ns relative delay, 0.5 Multipath Power Ratio (90°) .................... 263
![Page 25: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/25.jpg)
xxiv
Figure 9.21 — 200ns relative delay, 0.5 Multipath Power Ratio (125°) .................. 264
Figure 9.22 — 200ns relative delay, 0.5 Multipath Power Ratio (180°) .................. 264
Figure 9.23 — ½-Chip 0.5 Multipath Power Ratio................................................... 265
Figure 9.24 — 1/20 Chip 0.5 Multipath Power Ratio............................................... 266
Figure 9.25 — ½-Chip 0.5 Multipath Power Ratio (2MHz)..................................... 267
Figure 9.26 — 1/20-Chip 0.5 Multipath Power Ratio (8MHz)................................. 267
Figure 9.27 — Variation of Multipath Power Ratio ½-Chip (0°) ............................. 268
Figure 9.28 — Variation of Multipath Power Ratio ½-Chip (180°) ......................... 269
Figure 9.29 — Range Error 0.5 Multipath Power Ratio (200 ns 0°-180°) ............... 270
Figure 9.30 — Dual Multipath Error (0° and 0°)...................................................... 271
Figure 9.31 — Dual Multipath Error (0° and 180°).................................................. 272
Figure 9.32 — Dual Multipath Error (180° and 0°).................................................. 272
Figure 9.33 — Dual Multipath Error (180° and 180°).............................................. 273
Figure 9.34 — S/N Fade Pattern ............................................................................... 274
List of Tables
Table 2.1 — Electrical Properties ............................................................................... 12
Table 2.2 — Geometric Delays and Phases ................................................................ 32
Table 2.3 — Coupled Reflection Coefficient Magnitudes.......................................... 36
Table 3.1 — Model Comparison............................................................................... 110
Table 4.1 — PE Coefficients..................................................................................... 122
Table 6.1 — Modelling Errors Forward.................................................................... 177
Table 6.2 — Modelling Errors Forward and Back.................................................... 191
Table 9.1 — C/A-Code Selection ............................................................................. 245
Table 9.2 — DLL discriminator algorithms.............................................................. 249
Table 9.3 — PLL discriminator algorithms .............................................................. 251
Table 9.4 — FLL discriminator algorithms .............................................................. 252
![Page 26: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/26.jpg)
1
Chapter 1 Introduction and Overview
1.1 Introduction
The Global Positioning System (GPS) is currently revolutionising the disciplines of
navigation and positioning [1, 2]. Since the first navigation satellite was
commissioned in June 1977, the research community has devoted a significant
amount of time and energy into reducing the errors associated with satellite based
positioning systems. Initially the best accuracy obtainable with the system was 30m
(in three dimensions and 95% of the time). With the use of differential techniques it is
possible to remove many of the common-mode error sources, but there is one error
source that has proven much more difficult to mitigate — multipath. Multipath is the
dominant error source in applications that require high accuracy [3, 4].
Examples of applications where high accuracy is required are numerous. There is a
concerted effort in the open-cut mining community to reduce human exposure to
hazardous tasks. The automation of mining machines such as haul trucks is a prime
example. The use of GPS to navigate these huge machines is mandated by the
accuracy and availability requirements. However, the environment in which these
machines operate is harsh for the operation of GPS. Indeed the principle error source
in this hostile propagation environment is multipath propagation.
Surveying also requires the highest accuracy and in an attempt to achieve the highest
precision possible all attempts are made at reducing potential error sources. The
development and expansion of the use of differential GPS (DGPS) is an example of
the user community desiring higher accuracy.
Further to these obvious requirements any reduction in error sources opens the way to
new applications of GPS. Although this work is concerned with terrestrial
applications, it is also noted that space-based applications of GPS are likewise limited
by localised multipath effects [5].
![Page 27: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/27.jpg)
2
There has been substantial research into the topic of multipath and it remains an active
research area. Work by Braasch, van Nee [3, 4] and others [6-11] has provided an
understanding of multipath effects for defined and controlled multipath parameters.
The derivation of multipath propagation parameters, amplitude, phase, phase rate-of-
change, and delay, from actual propagation models has, however, received little
attention.
This research describes the use of modelling techniques, to gain an insight into
multipath propagation behaviour, and to extract the relevant multipath parameters for
simulated environments. In conjunction with receiver modelling, the use of a
propagation model — that more realistically represents actual propagation conditions
— provides the potential to further the understanding of multipath and its related
effects on GPS receivers. A comparative analysis of this model system with results
from actual receiver measurements (observables), obtained in field trials, provides
verification of an accurate understanding of the underlying principles used in the
simulation of the multipath environment.
This research aims to provide an improved understanding of the behaviour and effects
of Global Positioning System (GPS) signal propagation in precision application
environments, through the use of novel modelling and simulation techniques.
1.2 Overview of Research Presented
The research work and specific contributions presented in this dissertation are as
follows:
The fundamental issues of multipath propagation, and the concepts necessary for
development of modelling techniques are presented. Previous work has not offered
complete definition or explanation of the specific multipath environment for GPS.
The concept of reflection and diffraction are not new, but in this work a more
![Page 28: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/28.jpg)
3
thorough analysis of reflection and diffraction for the GPS L1 (1.575 GHz) signal is
given.
Specifically the scattering of the right-hand-circularly-polarised (RHCP) GPS signal is
represented in linear and circular reflection coefficients. A novel representation, more
suitable for single boundary propagation modelling, is presented. This coupled
polarisation reflection coefficient makes use of the fact that both co- and cross-
polarised representations have approximately similar phase shifts. The axial-ratio, or
effective left-hand-circular-polarisation (LHCP) rejection ratio of the GPS antenna is
then incorporated, with the cross-polarised component, into a single reflection
coefficient that may be used in propagation models designed for modelling of linear
polarisation propagation.
Aspects of rough surface effects, Fresnel zones, and diffraction are examined in detail.
As are the physical dimensions, in which GPS multipath propagation has an effect on
the receiver, for defined propagation states.
Providing an insight into multipath propagation behaviour; an enhanced Parabolic
Equation (PE)-based GPS propagation model is developed specifically for multipath
analysis of the GPS L1 signal. This work is based on a newly developed Free-Space
PE propagation technique. The model presented is an extension of an initial PE model
developed by Walker. This improved model includes the effects of backscatter and the
more realistic coupled GPS reflection coefficient for modelling of terrain boundary
effects.
In addition to providing relevant propagation data, the new model provides a
visualisation of the complete propagation of the GPS signal and its interaction with
the localised terrain. The visualisation aspects of the model are a valuable aid in the
understanding of GPS multipath propagation.
To extract the necessary multipath parameters, amplitude, phase, time delay, a Fourier
time synthesis technique has been developed. The complete PE-based Time Analysis
![Page 29: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/29.jpg)
4
(PETA) model provides a comprehensive description of the multipath RF
environment.
In addition, as for the FSPE on which it is based, the PETA provides a visualisation of
the multipath channel impulse response. This visualisation allows an interpretation of
the propagation mechanisms and their interaction in the modelled domain. The
resultant data output also provides the necessary input for a generic GPS correlation
model that provides range error estimation for the given propagation environment.
The FSPE and PETA models have been validated against exact solutions and accepted
standards, and multipath data was collected from a variety of sites and a comparison
made with the simulations provide by the FSPE model. In all cases the FSPE and
PETA models showed excellent agreement with accepted theoretical aspects of
propagation and the real-world data collected in field sessions.
The undesirable effect of signal fading on the GPS L1 signal was investigated and
utilised to provide a novel observable for relative antenna height above a reflecting
surface.
Receiver concepts are introduced and an open loop GPS code-correlation receiver
model is implemented. The results of the FSPE/PETA models form the input to the
receiver model and provide an insight into the effects of multipath propagation on the
code measurement process.
In this work a novel modelling environment has been researched, developed and
implemented. The Free-Space Parabolic Equation model, the PE-based Time Analysis
model and the GPS correlation/discrimination model provide a comprehensive suite
of modelling and simulation tools for the investigation of GPS multipath propagation
and its resultant error effects. It is recommended that further investigation be made of
any benefit in pursuing 3D modelling. In addition a thorough comparative analysis
should be made — using the modelling tools developed in this work — of multipath
data collected from a large variety of propagation environments.
![Page 30: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/30.jpg)
5
1.3 Research Contribution
The program of work carried out and presented in this dissertation has resulted in the
contribution of the following:
A thorough analysis of the terrestrial multipath propagation environment for the GPS
L1 signal. The nature of the propagation mechanisms and the resultant behaviour
including reflection, diffraction, rough surface effects and the spatial context of the
environment are examined for the first time, in such a way as to bring coherence to
our understanding of GPS multipath propagation.
The development and implementation of a new electromagnetic propagation model
specifically for GPS multipath investigation. The model builds on the accepted
parabolic equation propagation model with the inclusion of boundary conditions that
deal with the right-hand circularly polarised nature of the GPS signal. In addition
backscatter has been incorporated to provide a more realistic representation of real-
world propagation environments.
A novel time-analysis model that uses the GPS parabolic-equation propagation
method to provide a complete description and visualisation of the multipath
propagation behaviour.
The development of an open-loop correlation/discrimination model that allows the
variation of correlation/discrimination design parameters. Combined with the
propagation models, this novel comprehensive modelling system permits the testing
of performance for particular implementations in defined and repeatable multipath
environments.
![Page 31: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/31.jpg)
6
1.4 References
[1] National Academy of Public Administration and National Research Council,
"The Global Positioning System: Charting the Future," May 1995 1995.
[2] National Research Council, "The Global Positioning System - A Shared
National Asset." Washington DC: National Academy Press, 1995.
[3] M. S. Braasch, "Multipath Effects," in Global Positioning System: Theory and
Applications, vol. 1, B. W. Parkinson and J. R. Spilker Jr., Eds. Washington:
American Institute of Aeronautics and Astronautics, 1996, pp. 547-568.
[4] R. D. J. van Nee, "Multipath and Multi-Transmitter Interference in Spread-
Spectrum Communication and Navigation Systems," in Faculty of Electrical
Engineering, Telecommuncation and Traffic Control Systems Group. Delft:
Delft University of Technology, 1995, pp. 205.
[5] P. Axelrad and L. M. Ward, "Spacecraft Attitude Estimation Using the Global
Positioning System: Methodology and results for RADCAL," Journal of
Guidance, Control, and Dynamics, vol. 19, pp. 1201-1209, 1996.
[6] E. Breeuwer, "Modelling and Measuring GPS Multipath Effects," in Faculty
of Electrical Engineering. Delft: Delft University of Technology, 1992.
[7] B. Eissfeller and J. O. Winkel, "GPS Dynamic Multipath Analysis in Urban
Areas," presented at The 9th International Technical Meeting of The Satellite
Division of The Institute of Navigation., Kansas City, Missouri, 1996.
[8] A. El-Rabbany, "Temporal Characteristics of Multipath Errors," presented at
8th International Technical Meeting of The Satellite Division of The Institute
of Navigation., Palm Springs, California, 1995.
[9] T. Lo and J. Litva, "Use of a Highly Deterministic Multipath Signal Model in
Low-Angle Tracking," presented at IEE Proceedings-F [Radar and Signal
Processing], 1991.
[10] H. Leung and T. Lo, "A Spatial Temporal Dynamical Model for Multipath
Scattering From the Sea," IEEE Transactions on Geoscience & Remote
Sensing, vol. 33, pp. 441-448, 1995.
![Page 32: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/32.jpg)
7
[11] S. U. Hwu, B. P. Lu, R. J. Panneton, and B. A. Bourgeois, "Space Station GPS
Antennas Multipath Analysis," presented at IEEE Antennas and Propagation
Society International Symposium, Newport Beach, California, 1995.
![Page 33: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/33.jpg)
8
THIS PAGE IS INTENTIONALLY BLANK
![Page 34: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/34.jpg)
9
Chapter 2 Nature of GPS Multipath Propagation
To further the understanding of GPS multipath propagation and to develop and
implement an accurate GPS propagation model for simulation studies, it is necessary
that the environmental aspects of radio frequency propagation at the GPS L1
frequency be well understood. In this chapter the fundamental issues of environmental
factors, reflection, diffraction and rough surface effects on the GPS L1 signal are
examined in detail. The relevant concepts are examined within the context of being
ultimately incorporated into a GPS multipath propagation model.
2.1 Multipath Environment
Multipath is the unwanted distortion, of the direct line-of-sight satellite signal, by
localised reflected and/or diffracted signals. An example of these multipath signals is
shown in Figure 2.1.
Reflected path
Edge-diffracted path
Reflected pathDirect path
GPS antenna
Figure 2.1 — Multipath environment
The nature of the localised terrain determines the composition of the radio frequency
environment. To understand this multipath environment, it is necessary to consider
not only the physical relationship of the GPS receiver to the surrounding terrain
![Page 35: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/35.jpg)
10
elements (geometry), but also the propagation characteristics (electromagnetic
properties ) of the terrain.
The term multipath obviously describes the separate propagation paths taken by the
reflected/diffracted signals. Since the multipath signals travel additional distances
they are delayed relative to the line-of-sight (LOS) signal. This relative time delay, is
one of the defining parameters for describing the characteristics of multipath [1, 2].
In addition to the relative time delay, multipath is characterised by its amplitude,
phase, and phase rate-of-change, all relative to the LOS signal. The relative phase is a
function of the additional path length and the electrical properties of the
reflecting/diffracting medium, whilst the phase rate-of-change accounts for the
changing multipath propagation environment—due to the relative satellite-user
dynamics. Finally, the relative amplitude of the multipath signals is determined by the
nature of the reflecting surface structure. Critical aspects of the scattering of the GPS
L1 signal are presented in the following sections.
2.2 Specular Reflection
The theory of specular reflection is well understood [3-5], and aspects of the theory
are of importance for the understanding of multipath propagation of the Right-Hand
Circularly Polarised (RHCP) GPS L1 signal. In the following sections the concept of
the reflection coefficient is examined.
2.2.1 Linear Reflection Coefficient Representation
As was outlined previously, the behaviour of multipath is determined by the geometry
and electrical properties of the propagation environment.
The reflection coefficients, derived from the Fresnel equations for a smooth earth
surface provide information on the nature of reflected signals. The GPS signal is
![Page 36: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/36.jpg)
11
RHCP, and since circular polarisation is the vector sum of the linearly polarised
waves (horizontal and vertical), it is appropriate to initially consider the reflection
coefficients for the linear cases. The reflection coefficients for horizontal and vertical
polarisation are given respectively as,
ΓH =− −
+ −
sin cos
sin cos
θ ε θ
θ ε θ
2
2(2.1)
and
ΓV =− −
+ −
ε θ ε θ
ε θ ε θ
sin cos
sin cos
2
2(2.2)
where
ε εσ
ωε= −r j0
(2.3)
is the complex dielectric constant with assumed time dependence in tje ω− .
Substituting for ω and ε0, in equation (2.3) gives
ε ε λσ= −r j60 (2.4)
The calculation of each linear reflection coefficient is now straightforward, for a given
frequency, grazing angle (θ), dielectric constant and conduction value for the
reflecting surface medium. The resultant complex reflection coefficient then defines
the relative amplitude and phase of the specular reflection, and is of the general form,
Γ = ≤ ≤−ρ ρφe j , 0 1 (2.5)
![Page 37: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/37.jpg)
12
By now using equations (2.2) and (2.3), the linear reflection coefficients for
representative materials at 1 GHz (man-made and natural) given in Table 2.1 [6] can
be plotted.
Material Conductivity Relative Permittivity
Concrete 2 x 10-5 3
Dry Ground 1 x 10-5 4
Medium Dry Ground 4 x 10-2 7
Wet Ground 2 x 10-1 30
Fresh Water (fresh) 2 x 10-1 80
Sea Water (sea) 4 20
Table 2.1 — Electrical Properties
Figure 2.2 through to Figure 2.7, show the magnitude and phase of the linear
reflection coefficients as a function of propagation angle, for the given materials, at
the GPS L1 frequency, 1.575 GHz.
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient - Concrete
Mag
nitu
de
HV
Figure 2.2 — Linear reflection coefficients for Concrete
![Page 38: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/38.jpg)
13
0 10 20 30 40 50 60 70 80 900
50
100
150
200P
hase
(de
gs)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient - Dry Ground
Mag
nitu
de
HV
Figure 2.3 — Linear reflection coefficients for Dry Ground
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient - Medium Dry Ground
Mag
nitu
de
HV
Figure 2.4 — Linear reflection coefficients for Medium Dry Ground
![Page 39: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/39.jpg)
14
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient - Wet Ground
Mag
nitu
de
HV
Figure 2.5 — Linear reflection coefficients for Wet Ground
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient - Fresh Water
Mag
nitu
de
HV
Figure 2.6 — Linear reflection coefficients for Fresh Water
![Page 40: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/40.jpg)
15
0 10 20 30 40 50 60 70 80 900
50
100
150
200P
hase
(de
gs)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900.2
0.4
0.6
0.8
1Reflection Coefficient - Sea Water
Mag
nitu
de
HV
Figure 2.7 — Linear reflection coefficients for Sea Water
As can be seen, the variation of the reflection coefficient magnitude is different for
each linear polarisation. The horizontal component for all media decreases smoothly
for increasing propagation angles, with a near constant 180 degree phase shift for all
propagation angles. The vertical component magnitude, however, decreases quickly
and has a near constant 180 degree phase shift for angles less than the Brewster angle.
For propagation angles greater than the Brewster angle the vertical reflection
coefficient magnitude increases with a near constant zero degree phase shift.
On considering the vector addition of the two linear reflection coefficients, it is
obvious that the resultant polarisation of the GPS RHCP will be elliptical when the
coefficients are different, circular when they are equal, and linearly polarised when the
vertical component goes to zero. The nature of the final polarisation is determined by
the relative phase relationship of each linear component upon reflection.
For angles less than the Brewster angle the reflected signal is right-hand-elliptically-
polarised (RHEP), becoming increasingly more elliptic as the Brewster angle is
approached. Generally at the Brewster angle the polarisation approaches that of pure
horizontal polarisation. For angles greater than the Brewster angle the reflected signal
is left-hand-elliptically-polarised (LHEP) with the ellipticity decreasing as
propagation angles increase.
![Page 41: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/41.jpg)
16
The implication for modelling the GPS RHCP signal using the linear reflection
coefficients, is that two separate (horizontal and vertical) boundary models would
need to be implemented. An alternative representation of specular reflection
coefficients is given in the following section.
2.2.2 Circular Reflection Coefficient Representation
As an alternative representation of the reflection of the GPS RHCP L1 signal, the
resultant reflection can be considered as the sum of two circularly polarised (CP)
signals; one that maintains the co-polarisation (original) and a cross-polarisation
(opposite) component. The copolar ( oΓ ) and crosspolar ( xΓ ) reflection coefficients, as
a function of the horizontal and vertical reflection coefficients, are respectively given
as [7]:
2vh
o
Γ+Γ=Γ (2.6)
and
2vh
x
Γ−Γ=Γ (2.7)
From the previous analysis of the linear reflection coefficients for RHCP, it is an
intuitive observation that for propagation angles less than the Brewster angle the
copolar component predominates and for angles greater than the Brewster angle the
crosspolar component becomes dominant. This understanding is verified in the
following plots of copolar and crosspolar reflection coefficients for the same media as
those used for the linear reflection coefficient case.
![Page 42: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/42.jpg)
17
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient: circ (original & cross) - Concrete
Mag
nitu
de
orig cross
0 10 20 30 40 50 60 70 80 900
50
100
150
200P
hase
(de
gs)
Propagation angle (degs)
Figure 2.8 — Circular reflection coefficients for Concrete
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient: circ (original & cross) - Dry Ground
Mag
nitu
de
orig cross
Figure 2.9 — Circular reflection coefficients for Dry Ground
![Page 43: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/43.jpg)
18
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient: circ (original & cross) - Medium Dry Ground
Mag
nitu
de
orig cross
Figure 2.10 — Circular reflection coefficients for Medium Dry Ground
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient: circ (original & cross) - Wet Ground
Mag
nitu
de
orig cross
Figure 2.11 — Circular reflection coefficients for Wet Ground
![Page 44: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/44.jpg)
19
0 10 20 30 40 50 60 70 80 900
50
100
150
200P
hase
(de
gs)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient: circ (original & cross) - Fresh Water
Mag
nitu
de
orig cross
Figure 2.12 — Circular reflection coefficients for Fresh Water
0 10 20 30 40 50 60 70 80 900
50
100
150
200
Pha
se (
degs
)
Propagation angle (degs)
0 10 20 30 40 50 60 70 80 900
0.2
0.4
0.6
0.8
1Reflection Coefficient: circ (original & cross) - Sea Water
Mag
nitu
de
orig cross
Figure 2.13 — Circular reflection coefficients for Sea Water
The behaviour of the reflected signal is more easily recognised in the CP
representation, than in the linear case. For circular polarisation, copolar and crosspolar
reflection coefficients are of nearly equal magnitude at the Brewster angle. When the
magnitudes of the CP components are different the resultant polarisation will be
elliptic. This confirms that for propagation angles less than the Brewster angle the
resultant reflection is RHEP, becoming increasingly elliptic, until near linear
![Page 45: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/45.jpg)
20
polarisation occurs at the Brewster angle. For angles greater than the Brewster angle
the signal is highly LHEP becoming less elliptic for increasing propagation angles,
and ultimately, fully LHCP at 90 degrees.
On further examination of the circular reflection coefficients it is apparent that the
phase of both the copolar and crosspolar components tend towards 180 degrees. From
this observation an assumption can be made that both circular components, that
represent the original signal reflection, experience the same phase shift. However, to
model the complete reflection of the RHCP GPS signal it would still be necessary to
model two separate boundary conditions to account for the two separate polarisation
cases and then couple the resultant fields at the antenna location.
In considering the actual signal reception at the GPS antenna of a single reflected
signal, a further assumption can be made to simplify the boundary condition
requirements to a single effective CP reflection coefficient. This can be achieved by
making use of the fact that the GPS antenna is designed in an optimal sense, to
receive RHCP signals. The antenna thus rejects the LHCP or cross-polarised
component, at some level. By incorporating the LHCP rejection ratio into the cross-
polarised component, a simplified coupled single boundary representation is given,
one which can be readily implemented in a GPS propagation model. This effective
reflection coefficient for an incident RHCP signal can be written as:
πρρ jx
K
cR e−−
+=Γ 2010 (2.8)
where
cρ is the co-polarisation circular reflection coefficient magnitude (RHCP)
xρ is the cross-polarised circular reflection coefficient magnitude (LHCP)
and
![Page 46: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/46.jpg)
21
K is the GPS antenna LHCP rejection ratio in dB—which may be a constant or
modelled as a function of elevation angle (and azimuth angle). In general the gain
characteristics of high quality GPS antennas are essentially omnidirectional for both
polarisations. With this assumption the rejection ratio is essentially constant and the
use of K as a constant is justified. Typical values for commercial GPS antennas are
axial ratios of 3dB which equate to LHCP rejection ratios of about 10-11dB.
The effective coupled reflection coefficient magnitude for an incident L1 GPS RHCP
signal is given in the following figures, for constant rejection ratios ranging from 0 to
30 dB, in 3 dB increments.
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled RC Mag: Concrete
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.14 — Incident RHCP-RC for Concrete
30 dB
0 dB
![Page 47: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/47.jpg)
22
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled RC Mag: Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.15 — Incident RHCP-RC for Dry Ground
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled RC Mag: Medium Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.16 — Incident RHCP-RC for Medium Dry Ground
30 dB
0 dB
30 dB
0 dB
![Page 48: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/48.jpg)
23
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled RC Mag: Wet Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.17 — Incident RHCP-RC for Wet Ground
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled RC Mag: Fresh Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.18 — Incident RHCP-RC for Fresh Water
30 dB
0 dB
30 dB
0 dB
![Page 49: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/49.jpg)
24
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled RC Mag: Sea Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.19 — Incident RHCP-RC for Sea Water
In addition to the coupled reflection coefficients for incident RHCP, it is necessary to
consider the resultant coupled reflection coefficient for an incident LHCP signal. A
component of LHCP will be present after a single reflection of RHCP, if this LHCP
signal component is incident upon another reflection boundary, then the resultant
coupled LHCP reflection coefficient is given by:
πρρ jxc
K
L e−−
+=Γ 2010 (2.9)
here
cρ is the co-polarisation circular reflection coefficient magnitude (LHCP)
xρ is the cross-polarised circular reflection coefficient magnitude (RHCP)
and again
K is the GPS antenna LHCP rejection ratio in dB.
30 dB
0 dB
![Page 50: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/50.jpg)
25
The effective coupled reflection coefficient magnitude for an incident L1 LHCP signal
is given in the following figures, for rejection ratios ranging from 0 to 30 dB, in 3 dB
increments.
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled LHCP Incident RC Mag: Concrete
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.20 — Incident LHCP-RC for Concrete
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled LHCP Incident RC Mag: Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.21 — Incident LHCP-RC for Dry Ground
![Page 51: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/51.jpg)
26
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled LHCP Incident RC Mag: Medium Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.22 — Incident LHCP-RC for Medium Dry Ground
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled LHCP Incident RC Mag: Wet Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.23 — Incident LHCP-RC for Wet Ground
![Page 52: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/52.jpg)
27
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled LHCP Incident RC Mag: Fresh Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.24 — Incident LHCP-RC for Fresh Water
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Coupled LHCP Incident RC Mag: Sea Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.25 — Incident LHCP-RC for Sea Water
2.2.3 Ray-based Reflection Geometry: Relative Time Delay and Phase
Before considering reflection coefficients for specific GPS multipath propagation
cases we first need to develop an understanding of propagation situations that describe
the majority of GPS multipath propagation scenarios. In this section we consider
geometric optics to investigate simple ray-based geometry and formulate equations for
![Page 53: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/53.jpg)
28
the defining multipath parameters of relative amplitude, delay and phase, for three
primary modes of multipath propagation. These equations form a canonical set of
reflection situations that can be used in the verification of multipath parameters
derived from a GPS L1 propagation model. They also form the basis for an
understanding of the nature of multipath propagation in terrestrial environments.
Consider the forward-scatter problem confined to an arbitrary two-dimensional
domain with a flat reflecting lower boundary, Figure 2.26. The GPS antenna is located
at point P, at a distance d from the left-hand boundary, height h above the reflecting
surface, and with a Line-of-sight (LOS) signal propagating into the domain at angle θ.
h
-h
θ θ
d
θ
P
Pi
θ
∆R
LOS
Figure 2.26 — Forward scatter geometry
Using image theory, the reflected signal must travel an additional distance ∆R to the
image point Pi. This additional path length is given by:
θsin2hR =∆ (2.10)
Also the distance of propagation of the LOS signal into the domain—from the
arbitrary left-hand-side reference boundary—is given by
Dd
p =cosθ
(2.11)
![Page 54: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/54.jpg)
29
Now consider the backscatter problem with the addition of a vertical reflecting
surface forming the right-side boundary, Figure 2.27.
h
-h
θ
P
Pi
2θ
x
l1
l2
l3
l5
l4
LOS
Figure 2.27 — Backscatter geometry 1
For this given geometry the following inequality must be satisfied:
xh
>tanθ
The region in which this occurs will be known as Zone 1. The individual path lengths
are then given by:
lx
1
2=
cos
cos
θθ
, l2 =x
cosθ, l3 = 2hsinθ +
x cos2θcosθ
, l4 =x
cosθ−
h
sinθ, l5 =
h
sinθ,
and the total path-length differences, relative to the line-of sight, for the two multipath
propagation paths are given by,
∆R xa = 2 cosθ (2.12)
![Page 55: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/55.jpg)
30
and
∆R h xb = +2 2sin cosθ θ (2.13)
Here the subscripts a and b represent the multipath, either arriving from above or
below the horizontal plane containing the antenna.
For the special case of a geometry with xh
=tanθ
(corner reflection), the path lengths
are:
lx
1
2=
cos
cos
θθ
, l2 = l3 = l5 =h
sinθ=
x
cosθ, l4 = 0
and the total path difference is,
∆Rx h
= =2 2
cos sinθ θ(2.14)
Finally, if xh
<tanθ
, the case is such that the reflected signal arriving from below the
horizontal is produced by a different geometric arrangement, as shown in Figure 2.28.
![Page 56: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/56.jpg)
31
h
θP
2θ
x
l1
l2
l3
l5
l4
LOS
Figure 2.28 — Backscatter geometry 2
For this case, the region of occurrence will be known as Zone 2, and the relevant paths
are:
lx
1
2=
cos
cos
θθ
, l2 = l5 =x
cosθ, l3 = 2x cosθ −
h cos2θsinθ
, l4 =h
sinθ−
x
cosθ
However, the total path differences are the same as those given by equations (2.10)
and (2.11).
2.2.3.1 Summary of Amplitudes, Relative Delays and Phases
In summary, the amplitudes, relative time delays, and phases, for each of these simple
multipath propagation environments, can be given in terms of the effective reflection
coefficient and the geometric relationship of the antenna and the terrain.
![Page 57: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/57.jpg)
32
Path
(Mode)
Amplitude Relative TimeDelay
Total Phase Retardation
Line-of-sight
(LOS)
E0 1
c
d
cos
*
θ
2πλ θ
d
cos
*
Forward
scatter (F)
ρg E0 ( )12
ch sinθ φ
πλ
θg
h+
4sin
Backscattera
(BA)
ρr E0 ( )12
cx cosθ φ
πλ
θr
x+
4cos
Backscatterb
(BB)
0Egr ρρ ( )12 2
ch xsin cosθ θ+ ( )φ φ
πλ
θ θr g h x+ + +4
sin cos
*These values are relative to the arbitrary incident boundary of Figure 2.26
Table 2.2 — Geometric Delays and Phases
Here d is the horizontal distance from the left-side arbitrary boundary to the antenna
location, h is the vertical height of the antenna and x is the horizontal distance from
the vertical reflector to the antenna location. The speed of propagation is given as the
reference speed of light C. The reflection coefficients, ρg , φg , ρr , and φr , refer to the
ground and back-reflector respectively. For our derived effective coupled CP
reflection coefficients, the phase terms are all assumed as being π.
2.2.4 Multipath Modes and Coupled Reflection Coefficients
The three types of multipath propagation characterised in the previous section can be
considered as primary modes of GPS multipath propagation. F-Mode represents the
forward reflection of the signal from a ground-bounce source, BA-Mode is the back-
reflection that arrives from above the antenna. F-mode multipath propagation is a case
for direct application of the coupled RHCP reflection coefficient at the lower incident
boundary. For BA-Mode, the reflection coefficient magnitude terms are derived from
the same coupled RHCP reflection coefficient but using the transpose of the
![Page 58: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/58.jpg)
33
propagation angle (π/2-θ). That is for a propagation angle of 10 degrees the mode BA-
mode reflection coefficient is given by the coupled incident RHCP reflection
coefficient for an incident angle of 80 degrees at the backscatter vertical interface.
BB-Mode specifies the back-reflection arriving from below. In addition the BB-Mode
has two zones (Zone 1 and Zone 2) that specify the nature of the production of this
mode of multipath. In Zone 1 the first reflection is a BA-mode from the backscatter
interface incident upon the lower boundary and reflected to the antenna spatial
location. Consider Figure 2.29 where the circular reflection coefficients are decoupled
in polarisation and treated separately.
incident RHCP (Eo)
x and ρρ ′′c
(LHCP)
(RHCP)
xρρ
′′c
x and ρρc
(LHCP) and
(RHCP) and
cxx
xx
ρρρρρρρρ
′′′′
c
cc
Figure 2.29 — Decoupled Polarisation Zone 1 BB-Mode
Coupling the final resultant magnitudes at the antenna location with a LHCP rejection
factor of k gives:
( ) 0Ekk cxxcxxcc ρρρρρρρρ ′+′+′+′ (2.15)
![Page 59: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/59.jpg)
34
Now for BA-mode the back-reflector has the coupled RHCP incident reflection
coefficient magnitude applied. Therefore all reflected signals from the backscatter
interface in BA-mode are coupled correctly and the field travelling down towards the
lower boundary is fundamentally correct. However at the lower boundary the
reflection coefficient magnitude must be such that the resultant magnitude is the same
as that given in Equation (2.15) for the decoupled polarisation.
incident RHCP (Eo)
xρρ ′+′ kc
( ) 0Ek xc ′+′
σ
( ) 0Ekk cxxcxxcc ′+′+′+′
Figure 2.30 — Coupled Polarisation Zone 1 BB-Mode
If an arbitrary reflection coefficient is applied at the lower boundary then
( ) cxxcxxccxc kkk ρρρρρρρρσρρ ′+′+′+′=′+′ (2.16)
and
xc ηρρσ += (2.17)
where
![Page 60: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/60.jpg)
35
The required lower boundary reflection coefficient magnitude for BB-mode in Zone 1
is seen to be a coupled incident RHCP reflection coefficient with the cross-
polarisation component modified by the ratio of the incident LHCP and incident
RHCP coupled reflection coefficient for the back-scatter interface. This ratio increases
the cross-polarised influence on the final field magnitude compensating for the initial
coupling of polarisation states at the back-scatter interface.
Following the same procedure for Zone 2 gives:
xc ρµρβ ′+′= (2.18)
where
Likewise the BB-mode Zone 2 back-scatter reflection coefficient is a coupled
coefficient modified by the ratio of the incident LHCP to incident RHCP coupled F-
mode reflection coefficients. The associated modes and reflection coefficient
magnitudes are given in Table 2.3.
′+′′+′
=xc
xc
k
k
ρρρρη
++=
xc
xc
k
k
ρρρρµ
![Page 61: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/61.jpg)
36
Multipath Mode Coupled RC Magnitude Boundary
F-Mode xc kρρ + Lower in forward direction
BA-Mode xc kρρ ′+′ Back Reflector at positive
angles of incidence
BB-Mode Zone 1 xxc
xcc k
k ρρρρρρ
′+′′+′
+ Lower in back direction
BB-Mode Zone 2 xxc
xcc k
k ρρρρρρ ′
+++′ Back Reflector at negative
angles of incidence
Table 2.3 — Coupled Reflection Coefficient Magnitudes
The nature of the modified coupled reflection coefficients for BB-Mode in Zone 1 are
illustrated in Figure 2.31 to Figure 2.36 given below. It should be noted that these
reflection coefficients are contrived in the sense that they are necessitated by the
initial coupling of polarisation through the use of a coupled reflection coefficient at
the initial reflection boundary.
We recall that the coupling of the polarisation is done such that a single boundary
condition is satisfied and in terms of modelling there is no longer any need to treat the
polarisation boundary conditions separately, thus simplifying the boundary modelling.
![Page 62: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/62.jpg)
37
0 10 20 30 40 50 60 70 80 900
0.5
1
1.5
2
2.5
3
3.5
4Coupled BB-Mode Zone 1 RC Mag: Concrete
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.31 — Modified Coupled RC Zone 1 Concrete
0 10 20 30 40 50 60 70 80 900
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5Coupled BB-Mode Zone 1 RC Mag: Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.32 — Modified Coupled RC Zone 1 Dry Ground
![Page 63: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/63.jpg)
38
0 10 20 30 40 50 60 70 80 900
1
2
3
4
5
6
7Coupled BB-Mode Zone 1 RC Mag: Medium Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.33 — Modified Coupled RC Zone 1 Medium Dry Ground
0 10 20 30 40 50 60 70 80 900
2
4
6
8
10
12
14Coupled BB-Mode Zone 1 RC Mag: Wet Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.34 — Modified Coupled RC Zone 1 Wet Ground
![Page 64: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/64.jpg)
39
0 10 20 30 40 50 60 70 80 900
2
4
6
8
10
12
14
16
18Coupled BB-Mode Zone 1 RC Mag: Fresh Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.35 — Modified Coupled RC Zone 1 Fresh Water
0 10 20 30 40 50 60 70 80 900
2
4
6
8
10
12
14
16
18Coupled BB-Mode Zone 1 RC Mag: Sea Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.36 — Modified Coupled RC Zone 1 Sea Water
The coupled RC for BB-Mode in Zone 2 is again the propagation angle transpose of
the coupled reflection coefficient magnitudes and are included for completeness.
![Page 65: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/65.jpg)
40
0 10 20 30 40 50 60 70 80 900
0.5
1
1.5
2
2.5
3
3.5
4Coupled BB-Mode Zone 2 RC Mag: Concrete
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.37 — Modified Coupled RC Zone 2 Concrete
0 10 20 30 40 50 60 70 80 900
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5Coupled BB-Mode Zone 2 RC Mag: Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.38 — Modified Coupled RC Zone 2 Dry Ground
![Page 66: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/66.jpg)
41
0 10 20 30 40 50 60 70 80 900
1
2
3
4
5
6
7Coupled BB-Mode Zone 2 RC Mag: Medium Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.39 — Modified Coupled RC Zone 2 Medium Dry Ground
0 10 20 30 40 50 60 70 80 900
2
4
6
8
10
12
14Coupled BB-Mode Zone 2 RC Mag: Wet Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.40 — Modified Coupled RC Zone 2 Wet Ground
![Page 67: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/67.jpg)
42
0 10 20 30 40 50 60 70 80 900
2
4
6
8
10
12
14
16
18Coupled BB-Mode Zone 2 RC Mag: Sea Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.41 — Modified Coupled RC Zone 2 Sea Water
0 10 20 30 40 50 60 70 80 900
2
4
6
8
10
12
14
16
18Coupled BB-Mode Zone 2 RC Mag: Fresh Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.42 — Modified Coupled RC Zone 2 Fresh Water
The resultant reflection coefficient for the double reflection mode specified by BB-
Mode multipath is identical for Zone 1 and Zone 2, and is simply the product of the
two incident boundary reflection coefficient magnitudes. Indeed the modified
coupling at the second incident boundary ensures that the resultant product represents
the decoupled polarisation case. The resultant reflection coefficients for common
![Page 68: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/68.jpg)
43
material (that is both boundaries have the same characteristics) BB-Mode are shown
in the figures below for LHCP rejections ranging from 0dB to 30dB.
0 10 20 30 40 50 60 70 80 900
0.05
0.1
0.15
0.2
0.25
0.3
0.35Resultant Coupled BB-Mode RC Mag: Concrete
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.43 — Resultant RC BB-Mode Concrete
0 10 20 30 40 50 60 70 80 900
0.05
0.1
0.15
0.2
0.25
0.3
0.35Resultant Coupled BB-Mode RC Mag: Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.44 — Resultant RC BB-Mode Dry Ground
![Page 69: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/69.jpg)
44
0 10 20 30 40 50 60 70 80 900
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5Resultant Coupled BB-Mode RC Mag: Medium Dry Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.45 — Resultant RC BB-Mode Medium Dry Ground
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7Resultant Coupled BB-Mode RC Mag: Wet Ground
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.46 — Resultant RC BB-Mode Wet Ground
![Page 70: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/70.jpg)
45
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8Resultant Coupled BB-Mode RC Mag: Fresh Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.47 — Resultant RC BB-Mode Fresh Water
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9Resultant Coupled BB-Mode RC Mag: Sea Water
Mag
nitu
de
Propagation Angle (degs)
0 3 6 912151821242730
Figure 2.48 — Resultant RC BB-Mode Sea Water
2.3 Rough Surface Scatter
The coupled reflection coefficients account only for specular reflection. For many
practical cases of GPS L1 propagation the surface cannot be considered as smooth. In
these cases the signal is scattered in a non-coherent, non-directional sense. If the
![Page 71: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/71.jpg)
46
surface contains features that lead to diffuse scattering of the incident wave the
coupled reflection coefficients for the GPS RHCP L1 signal require modification.
In explaining diffuse scatter from rough surfaces we consider the reflection of rays
incident on an irregular surface, Figure 2.49.
∆H
ψ
ψ ψ
Figure 2.49 — Surface roughness geometry
The phase difference, between the two rays reflecting from the different levels, is
given by [8],
ψλ
πφ sin4 H
H
∆=∆ (2.19)
Setting the phase difference to π, full cancellation of the reflected components in the
forward direction occurs. If there is no energy transferred in the forward direction then
diffuse scattering of the signal in other directions must have occurred. The Rayleigh
criterion (where the phase difference is considered π/2) for the roughness of a surface
can be written as [9],
ψλ
sin8≥∆H (2.20)
![Page 72: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/72.jpg)
47
In the case where ∆H is less than this criterion, the surface roughness effects are
considered negligible—specular reflection occurs—and the coupled reflection
coefficient applies without modification. In general ∆H is considered as the standard
deviation of the surface height about the local mean value within the first Fresnel
zone (which is examined in the next section).
For GPS L1 signal propagation a plot of ∆H for all propagation angles provides only a
qualitative indication of the likelihood of the resultant scatter being diffuse or
specular, Figure 2.50.
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Surface Roughness Criterion >delta-H
delta
-H:
Sur
face
Irre
gula
rity
(m)
Propagation Angle (degs)
Figure 2.50 — Rayleigh Roughness Criterion
For cases where the surface roughness needs to be accounted for, the effective
specular reflection coefficient, is modified by applying a magnitude reduction factor
to the coefficient. The modified reflection coefficient is then given as [10]:
Γ=Γ ss ρ (2.21)
where;
region of increaseddiffuse scattering
![Page 73: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/73.jpg)
48
−= 2
02
2
1
2
1exp HHs I φφρ (2.22)
is the rough surface specular reduction factor, and 0I the modified Bessel function of
zero order.
A plot of sρ at the GPS L1 frequency for RMS surface roughness heights of 0-10cm
at 2.5cm steps, is given below.
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Rough Surface Scatter
Diff
use
Sca
tter
Red
uctio
n F
acto
r
Propagation Angle (degs)
Figure 2.51 — Rough Surface Reduction Factor
This treatment of rough surface scattering accounts only for the reduction of the
coherent scatter in the specular direction. Generally for a rough surface, the reflected
signal has an additional component that accounts for the nature of the diffuse scatter.
Experimental results indicate that the diffuse component is statistically random with a
Rayleigh distribution. In this work, only the rough surface specular reduction factor
will be considered for implementation of rough surface effects in the developed
propagation model through the use of equation (2.22).
2.5cm
5.0cm7.5cm10cm
0cm
![Page 74: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/74.jpg)
49
2.4 Fresnel Zones
In geometrical optics representations of propagation, specular reflection is considered
to occur from some single geometric point. However, the source illuminates a large
portion of the Earth, and there is contribution from an area of the surface, to the total
reflected signal. In understanding the nature of GPS L1 propagation it is worthwhile
asking the question, ”which regions of the scattering surface most contribute to the
total field at the GPS receiving antenna”.
In Figure 2.52, we consider a smooth plane illuminated by the GPS source at A and
received at the GPS antenna located at B. The interleaving space can be sub-divided
by a family of Fresnel ellipsoids. These ellipsoids have focal points located at A and
B, such that at any point M on any one ellipsoid, the following relation holds [11]:
2
λnABMBAM +=+ (2.23)
Where n is an integer with n=1, defining the first Fresnel ellipsoid and so forth. The
Fresnel zone concept for the case of reflection is such that the path of reflection is
between the source located at A, and an image of the antenna located at B’. This is
illustrated in Figure 2.52.
![Page 75: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/75.jpg)
50
A
B’
B
M
h
y
x
n=1 2 3
Figure 2.52 — Fresnel Zones for Reflection
The ellipses generated in the xy-plane, by the intersection of the Fresnel ellipsoids,
determine the Fresnel zones. The nth zone is defined as the area between the ellipses
obtained from ellipsoids n and n-1 respectively. It is generally accepted that the first
Fresnel zone contributes most to the reflected incident energy, when the reflecting
surface is much larger than the first Fresnel zone.
The defining measurements, radius and semi-major axis, of the first Fresnel zones for
the GPS L1 frequency, as a function of antenna height and propagation angle are
respectively [11]:
θλ
sin
hRF = (2.24)
![Page 76: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/76.jpg)
51
and θsin
FSMA
RL = (2.25)
The following plots give an indication of the dimension of the first Fresnel zone for
antenna heights ranging from 1 to 5 metres.
1 1.5 2 2.5 3 3.5 4 4.5 50
2
4
6
8
1st
Zone
Rad
ius
(m)
GPS First Fresnel Zone
1 1.5 2 2.5 3 3.5 4 4.5 50
100
200
300
400
500
Sem
i-maj
or a
xis
(m)
Propagation Angle
1m2m3m4m5m
Figure 2.53 — First Fresnel Zone Dimensions (1-5 degs)
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 101
1.5
2
2.5
3
3.5
1st
Zone
Rad
ius
(m)
GPS First Fresnel Zone
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 100
10
20
30
40
Sem
i-maj
or a
xis
(m)
Propagation Angle
Figure 2.54 — First Fresnel Zone Dimensions (5-10 degs)
![Page 77: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/77.jpg)
52
10 20 30 40 50 60 70 80 900
0.5
1
1.5
2
2.5
1st
Zone
Rad
ius
(m)
GPS First Fresnel Zone
10 20 30 40 50 60 70 80 900
5
10
15
Sem
i-maj
or a
xis
(m)
Propagation Angle
Figure 2.55 — First Fresnel Zone Dimensions (10-90 degs)
The size and location of the first Fresnel zone (for reflection) with relation to the GPS
antenna allows interpretation of the level of multipath that could be expected to result.
When there is a surface that is dimensionally larger than the first zone, we can expect
specular reflection to occur. The nature of the surface determines the level of the
reflection.
![Page 78: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/78.jpg)
53
GPS SV
Fresnel Zones
Obstruction in firstFresnel zone
GPS antenna
Figure 2.56 — Fresnel Zones for LOS
Considering now Fresnel zones for line-of-sight transmission, Figure 2.56, the nature
of the propagation is determined by obstructions within the Fresnel zones. Signal
blocking is apparent if there is an obstructing object larger than the first Fresnel zone.
If there is an obstruction lying within the first Fresnel zone that is not larger than the
zone (but larger in relation to the wavelength) then the expected multipath
propagation behaviour is that of diffraction. Diffraction of the GPS L1 signal is
considered in the next section.
![Page 79: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/79.jpg)
54
2.5 Diffraction
As was seen in the previous sections, specular reflection is the primary multipath
producer. There is however, one other propagation characteristic that can lead to
secondary multipath effects—that of signal diffraction. Whenever a signal encounters
an obstructing object, some of the energy is diffracted at the edges of the object—
effectively bending the signal around the edge. This results in the area behind the
obstructing object not being completely shadowed from the direct signal (as would be
expected from geometric optics), and some residual energy will exist within the
expected shadow region.
Constant Phase Fronts
Diffracted Signal
LOS
Obstruction
Shadow Boundary
Figure 2.57 — Diffraction at Obstruction
The most obvious effect of diffraction on the GPS L1 signal is the shadowing effect of
the obstacle. The expectation is that the signal level will be reduced and the satellite
being tracked will become unusable because of lack of signal strength. From a
multipath perspective however the main effect of diffraction of the GPS L1 signal is
the very presence of a signal in a shadow region. The nature of the signal is such that
![Page 80: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/80.jpg)
55
it cannot have travelled by the direct LOS path, since the GPS antenna LOS is blocked
by the obstacle. The signal must therefore have travelled some additional distance
around the obstruction.
2.5.1 Knife-Edge Diffraction
For diffraction of the GPS L1 frequency we consider the case of knife-edge
diffraction. The geometric parameters that define the knife-edge diffraction problem
are incorporated in a single dimensionless Fresnel diffraction parameter denoted by v.
There are many representations of v depending on the given geometric relationships.
For GPS it is most appropriate to utilise the following geometric description.
LOS
θβ
h
ha
d2
d1
l
da
Constant phase fronts
Figure 2.58 — Diffraction at Obstacle
The angle between the line-of-sight vector (propagation angle θ ) and the diffracted
ray is given by,
θβ −
= −
a
a
d
h1tan (2.26)
![Page 81: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/81.jpg)
56
with
222 aa dhd += (2.27)
and
d d1 2= cosβ (2.28)
In the GPS far field the Fresnel diffraction parameter can then be written as:
v hd
=2
2λ(2.29)
where
h d= 2 sin β (2.30)
For the knife-edge case the signal distance between the LOS and the diffracted path is
seen to be d2-d1. If the dynamic range of the GPS receiver allows tracking of the
diffracted signal then it will have an error in time-of-travel estimation of:
( )it ddc
−= 2
1ε (2.31)
To determine if the receiver can track the signal we need to quantify the expected loss
in the diffraction region. For values of v greater than –1 the loss can be approximated
by [11];
![Page 82: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/82.jpg)
57
( ) ( )vvvJ +++= 1log204.6 2 (2.32)
Which is plotted below in Figure 2.59 for 31 ≥≥ v- .
-1 -0.5 0 0.5 1 1.5 2 2.5 3-25
-20
-15
-10
-5
0
5
J(v)
(db
)
v
Figure 2.59 — Knife-edge Diffraction Gain vs v
As a descriptive example we place a GPS receiver 5 metres from and below a knife-
edge diffractor. The resultant plots of parameter v and the diffraction gain are given
below, Figure 2.60.
10 15 20 25 30 35 40 45 50-2
0
2
4
6
Diff
ract
ion
Par
amet
er (
v)
10 15 20 25 30 35 40 45 50-30
-20
-10
0
Diff
ract
ion
Gai
n (d
b)
Propagation Angle (degs)
Figure 2.60 — Diffraction Parameter and Gain
![Page 83: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/83.jpg)
58
The critical angle at which the GPS signal is expected to be in the clear is 45 degrees.
From the plot the loss at this critical angle is approximately 6dB. In addition the
signal is not in the clear (that is at 0dB or line-of-sight) until approximately 50
degrees.
If the receiver’s dynamic range is such that it can track signals down 30 dB then the
diffracted signal could be tracked (for a satellite that is setting). The LOS time-of-
travel error for tracking this example diffracted signal is given below in Figure 2.61.
10 15 20 25 30 35 40 45 500
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Err
or (
ns)
Propagation Angle (Degs)
Figure 2.61 — Time of Arrival Error
This time error is low but for this example the estimate of range is in error by a
maximum of 1.2 metres. In a stand-alone application of GPS this range error would
not have a influential effect on the total position solution. It would however be a
substantial range error for DGPS applications.
![Page 84: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/84.jpg)
59
2.6 GPS Fading Signal Characteristics
2.6.1 Signal Fade Envelopes
Having established a set of coupled reflection coefficients for the GPS L1 signal for
the three primary modes (F, BA, BB) of specular multipath propagation, it is now
possible to consider the effect of these selected modes on the received GPS signal
strength.
The total electric field above the earth, due to the interference of the direct LOS signal
with a single reflected signal component, is termed the interference region [12]. The
interference region field can be written as,
( ) ( ) ( )E = + − +E f E f ei ij k R
0 0θ ρ θ φ∆ (2.33)
where f(θi) and f(θr) represent the field distribution, as a function of the angle of
incidence, and angle of reflection respectively. In terrestrial propagation these factors
would account for the transmission antenna’s radiation pattern. For the GPS satellite
propagation problem, these pattern shape terms are not required—as at the Earth’s
surface the far-field is essentially uniform plane-wave only.
In equation (2.33), the total phase retardation of the reflected component is the sum of
the phase shift upon reflection (φ), and the phase shift due to the additional path
length (∆R) taken by the multipath signal—which is simply k∆R.
The concept of the interference region provides a visualisation of the two-ray
problem. It is obvious that the fading pattern of the signal will be bounded by the in-
phase and anti-phase limits. The normalised received signal strength envelopes of the
GPS L1 signal with one dominant reflection can then be written as:
![Page 85: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/85.jpg)
60
( ) ( )Γ+→Γ− 1log201log20 (2.34)
Considering first F-mode—a single forward reflection. Plots (Figure 2.62 to Figure
2.67)of the signal strength fade envelopes reveal the bounds of the fading received
signal strength for this mode of GPS L1 multipath affected signal.
0 10 20 30 40 50 60 70 80 90-50
-40
-30
-20
-10
0
10Fade Envelopes: Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.62 — Concrete F-mode
0 10 20 30 40 50 60 70 80 90-50
-40
-30
-20
-10
0
10Fade Envelopes: Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.63 — Dry Ground F-mode
![Page 86: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/86.jpg)
61
0 10 20 30 40 50 60 70 80 90-50
-40
-30
-20
-10
0
10Fade Envelopes: Medium Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.64 — Medium Dry Ground F-mode
0 10 20 30 40 50 60 70 80 90-50
-40
-30
-20
-10
0
10Fade Envelopes: Wet Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.65 — Wet Ground F-mode
![Page 87: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/87.jpg)
62
0 10 20 30 40 50 60 70 80 90-50
-40
-30
-20
-10
0
10Fade Envelopes: Fresh Water
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.66 — Fresh Water F-mode
0 10 20 30 40 50 60 70 80 90-50
-40
-30
-20
-10
0
10Fade Envelopes: Sea Water
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.67 — Sea Water F-mode
From the above plots it is apparent that for the F-mode, single forward-scatter
multipath signal, the effect on the received signal strength is more conspicuous at
lower angles of propagation.
For the BA-mode case, the fade envelopes are transposed in propagation angle and it
is obvious that the effect on the signal strength envelope is opposite. That is, deeper
![Page 88: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/88.jpg)
63
fades of the signal strength occur at higher propagation angles. This is evident in
Figure 2.68 given as an example of the mode BA fade envelope for a concrete back
reflector.
0 10 20 30 40 50 60 70 80 90-50
-40
-30
-20
-10
0
10Fade Envelopes: Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.68 — Concrete BA-mode
The fade envelopes for all other media in BA-mode can be read directly as the
propagation angle transpose of the F-mode plots, and are not repeated here.
For mode BB the total reflection coefficient of the multipath signal is the product of
the ground and back reflector individual modified coupled reflection coefficients. The
four plots shown below indicate the expected signal envelopes for the case of the
ground and back-reflector being of the same medium (hence the symmetry).
![Page 89: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/89.jpg)
64
0 10 20 30 40 50 60 70 80 90-1.5
-1
-0.5
0
0.5
1
1.5BB-Mode Fade Envelope: Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6 LHCP Rejection12 LHCP rejection18 LHCP rejection
Figure 2.69 — Concrete2 BB-mode
0 10 20 30 40 50 60 70 80 90-2
-1.5
-1
-0.5
0
0.5
1
1.5BB-Mode Fade Envelope: Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6 LHCP Rejection12 LHCP rejection18 LHCP rejection
Figure 2.70 — Dry Ground2 BB-mode
![Page 90: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/90.jpg)
65
0 10 20 30 40 50 60 70 80 90-3
-2
-1
0
1
2
3BB-Mode Fade Envelope: Medium Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6 LHCP Rejection12 LHCP rejection18 LHCP rejection
Figure 2.71 — Medium Dry Ground 2 BB-mode
0 10 20 30 40 50 60 70 80 90-8
-6
-4
-2
0
2
4BB-Mode Fade Envelope: Wet Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6 LHCP Rejection12 LHCP rejection18 LHCP rejection
Figure 2.72 — Wet Ground2 BB-mode
From these plots the fading of the signal strength due to the double reflection of the
BB-mode multipath is greatest at 45 degrees. For the representative media given here,
the fades for concrete are unlikely to exceed 1 dB, for dry ground 1.5 dB, for medium
dry ground 2 dB, and finally for wet ground 6 dB. In general the fading expected due
to BB-mode multipath is not as large as the fades for F and BA mode multipath.
However the fading remains fairly constant across a large range of propagation angles,
![Page 91: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/91.jpg)
66
whereas the fading for the other two modes is dominant at the extremes of the
propagation incident angle range. Additional BB-mode fade envelopes are included
here for completeness. They indicate the variation of the two propagation media.
0 10 20 30 40 50 60 70 80 90-5
-4
-3
-2
-1
0
1
2
3Fade Envelopes: Sea Water-Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.73 — Concrete to Sea Water
0 10 20 30 40 50 60 70 80 90-5
-4
-3
-2
-1
0
1
2
3Fade Envelopes: Sea Water-Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.74 — Dry Ground to Sea Water
![Page 92: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/92.jpg)
67
0 10 20 30 40 50 60 70 80 90-5
-4
-3
-2
-1
0
1
2
3Fade Envelopes: Sea Water-Medium Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.75 — Medium Dry Ground to Sea Water
0 10 20 30 40 50 60 70 80 90-5
-4
-3
-2
-1
0
1
2
3Fade Envelopes: Fresh Water-Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.76 — Concrete to Fresh Water
![Page 93: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/93.jpg)
68
0 10 20 30 40 50 60 70 80 90-5
-4
-3
-2
-1
0
1
2
3Fade Envelopes: Fresh Water-Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.77 — Dry Ground to Fresh Water
0 10 20 30 40 50 60 70 80 90-5
-4
-3
-2
-1
0
1
2
3Fade Envelopes: Fresh Water-Medium Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.78 — Medium Dry Ground to Sea Water
![Page 94: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/94.jpg)
69
0 10 20 30 40 50 60 70 80 90-4
-3
-2
-1
0
1
2
3Fade Envelopes: Wet Ground-Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.79 — Wet Ground to Concrete
0 10 20 30 40 50 60 70 80 90-4
-3
-2
-1
0
1
2
3Fade Envelopes: Wet Ground-Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.80 — Wet Ground to Dry Ground
![Page 95: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/95.jpg)
70
0 10 20 30 40 50 60 70 80 90-4
-3
-2
-1
0
1
2
3Fade Envelopes: Wet Ground-Medium Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.81 — Wet Ground to Medium Dry Ground
0 10 20 30 40 50 60 70 80 90-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2Fade Envelopes: Medium Dry Ground-Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.82 — Medium DryGround to Concrete
![Page 96: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/96.jpg)
71
0 10 20 30 40 50 60 70 80 90-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2Fade Envelopes: Medium Dry Ground-Dry Ground
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.83 — Medium Dry Ground to Dry Ground
0 10 20 30 40 50 60 70 80 90-2
-1.5
-1
-0.5
0
0.5
1
1.5Fade Envelopes: Dry Ground-Concrete
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
6dB LHCP Rejection 12dB LHCP Rejection18dB LHCP Rejection
Figure 2.84 — Dry Ground to Concrete
In this section the bounds of the GPS signal fade pattern have been examined for the
three primary modes of multipath propagation. In the next section the nature of the
fading pattern within the envelope bounds is examined.
![Page 97: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/97.jpg)
72
2.6.2 Signal Fading Characteristics for Specular Reflection
For the case of m specular multipath signals (ignoring receiver antenna gain pattern),
the received signal strength at the GPS receiver can be written as ;
( )
+= ∑ +
m
i
ji
j iLOSLOS eeAS ψφφ ρ0 (2.35)
where for the ith multipath signal, ρi is the magnitude of the reflection coefficient, ψi
is the total phase retardation due to the reflection and subsequent additional path
distance travelled, and φLOS is the phase of the LOS signal.
We now consider the normalised case of a single forward multipath signal. From
Table 2.2 the nature of the total received signal strength is a function of the reflection
coefficient amplitude, the height of the GPS antenna, and the propagation angle from
the satellite.
We consider first a linear variation of the propagation angle from 1-90 degrees in 900
measurement cycles (epochs). The height of the receiving antenna is set at 0.5 meters
with Multipath to LOS Ratio (MPR) of 1 (ie the multipath signal is of equal amplitude
to the LOS signal). The reflection coefficient amplitude has not been considered for
this descriptive example.
![Page 98: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/98.jpg)
73
0 100 200 300 400 500 600 700 800 900-40
-30
-20
-10
0
10H=0.5: MPR=1.0
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
0 100 200 300 400 500 600 700 800 9000
20
40
60
80
100A
ngle
(de
gs)
Measurement Epochs
Figure 2.85 — Linear Variation of Propagation Angle (Fwd)
The gradient of the linear variation in the propagation angle is now reduced by half
such that in the same 900 epochs the propagation angle varies from 0-45 degrees.
0 100 200 300 400 500 600 700 800 900-40
-30
-20
-10
0
10H=0.5: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
0 100 200 300 400 500 600 700 800 9000
20
40
60
80
100
Ang
le (
degs
)
Measurement Epochs
Figure 2.86 — Linear Variation (Reduced gradient)
Note that the peaks of the fading pattern peak at 6 dB—since the direct plus multipath
add up to twice the strength of the direct signal only. From these plots we also note
that the periodicity of the fading is a function of the rate-of-change of propagation
![Page 99: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/99.jpg)
74
angle. This is expected, since the rate-of-change in phase of the single multipath
signal, as a function of the rate-of-change in propagation angle, is given by:
dt
dh
dt
d θθλπψ
= cos
4(2.36)
This equation explains one aspect of the fading pattern—in that the phase-rate varies
as the cosine of the propagation angle. So for a given positive rate-of-change of
propagation angle the fade separation increases as the rate-of-change of the multipath
phase decreases.
It is also apparent from equation (2.36) that the phase rate is also directly related to
the height of the antenna above the reflecting surface. This can be seen in Figure 2.87
where for a given rate-of-change in propagation angle the fading periodicity increases
directly as a function of antenna height.
0 10 20 30 40 50 60 70 80 90-40
-35
-30
-25
-20
-15
-10
-5
0
5
10H=0.2: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)0 10 20 30 40 50 60 70 80 90
-40
-35
-30
-25
-20
-15
-10
-5
0
5
10H=0.5: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
0 10 20 30 40 50 60 70 80 90-40
-35
-30
-25
-20
-15
-10
-5
0
5
10H=1: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)0 10 20 30 40 50 60 70 80 90
-40
-35
-30
-25
-20
-15
-10
-5
0
5
10H=2: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Propagation Angle (degs)
Figure 2.87 — Variation of Antenna Height
![Page 100: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/100.jpg)
75
For the case of BA-mode, backscatter from above, the phase-rate is given by
dt
dx
dt
d θθλπψ
−= sin
4(2.37)
Here the phase rate-of-change varies as the negative sine of the propagation angle. In
a similar fashion as for the forward scatter case we can infer that for a given positive
rate-of-change of propagation angle the fade separation now decreases as the rate-of-
change of the multipath phase increases. This is illustrated in Figure 2.88.
0 100 200 300 400 500 600 700 800 900-40
-30
-20
-10
0
10X=0.5: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Measurement Epochs
0 100 200 300 400 500 600 700 800 9000
20
40
60
80
100
Ang
le (
degs
)
Measurement Epochs
Figure 2.88 — Backscatter from above
Comparing Figure 2.85 with Figure 2.88 for the same variation of propagation angle it
is possible (for these two cases) to infer where the multipath signal has originated
from.
Similarly for BB-mode, the backscatter from below, the phase rate-of-change is given
as:
![Page 101: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/101.jpg)
76
( )dt
dxh
dt
d θθθλπψ
sincos4 −= (2.38)
The resultant fading pattern for an antenna located 3m from the ground and 3m from
the backscatter interface is shown in Figure 2.89.
0 100 200 300 400 500 600 700 800 900-40
-30
-20
-10
0
10X=3 H=3: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Measurement Epochs
0 100 200 300 400 500 600 700 800 9000
20
40
60
80
100
Ang
le (
degs
)
Measurement Epochs
Figure 2.89 — Equal Antenna Distance and Height
For equal height and distance, and with linear variation of propagation angle, we note
that the first 450 epochs are dominated by the variation due to the antenna height. The
cosine variation being apparent. The final 450 epochs are dominated by the distance
from the backscatter interface with its sine variation. This dominant variation depends
solely on the relevant distances. Decreasing the distance to the backscatter interface
will make the antenna height the predominant factor in the fading pattern, Figure 2.90.
![Page 102: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/102.jpg)
77
0 100 200 300 400 500 600 700 800 900-40
-30
-20
-10
0
10X=1 H=3: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)Measurement Epochs
0 100 200 300 400 500 600 700 800 9000
20
40
60
80
100A
ngle
(de
gs)
Measurement Epochs
Figure 2.90 — Antenna Height > Distance
Likewise, if the antenna height is reduced relative to the distance to the backscatter
interface this distance factor has increased influence on the resultant fade pattern,
Figure 2.91.
0 100 200 300 400 500 600 700 800 900-40
-30
-20
-10
0
10X=3 H=1: MPR=1
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Measurement Epochs
0 100 200 300 400 500 600 700 800 9000
20
40
60
80
100
Ang
le (
degs
)
Measurement Epochs
Figure 2.91 — Antenna Height < Distance
Having examined these three cases, it is apparent that individually it would be
possible to infer the nature of the reflection from the resultant fade pattern. However
![Page 103: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/103.jpg)
78
the true nature of the multipath environment may be such that all three of these basic
multipath propagation modes exist at the same time and are received by the GPS
antenna. The increasing complexity of the resultant signal strength fading pattern is
obvious as more multipath modes are introduced, Figure 2.92.
0 100 200 300 400 500 600 700 800 900-40
-20
0
X=3 H=3: MPR=1
dB
0 100 200 300 400 500 600 700 800 900-40
-20
0
dB
0 100 200 300 400 500 600 700 800 900-40
-20
0
dB
Measurement Epochs
Figure 2.92 — Addition of Multipath Modes
In the first plot the resultant signal fading pattern is due only to mode F—a single
forward multipath signal. The next plot is the addition of modes F and BA—the
backscatter from above multipath with the forward multipath. Lastly the backscatter
from below (mode BB) is also included. As can be seen the variations that were
apparent for the individual cases are no longer recognisable.
Having examined the periodicity of the fading, we now consider the effect on the
received signal strength of variation in the MPR. For the forward scatter case, the
resultant variation of signal strength fade depth is shown in Figure 2.93, for an
antenna height of 0.5 metres.
![Page 104: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/104.jpg)
79
0 100 200 300 400 500 600 700 800 900-40
-35
-30
-25
-20
-15
-10
-5
0
5
10Variation of Relative MP Amplitude (H=0.5)
Nor
mal
ised
Sig
nal S
tren
gth
(dB
)
Measurement Epochs
00.20.40.60.8 1
Figure 2.93 — Variation of Relative Multipath Amplitude
From a generalised viewpoint of signal reception the critical point of note is the effect
of the signal fades on the receiver’s ability to search for or track the received GPS
signal. The signal-to-noise ratio at deep fades may be so low that in the search mode
the receiver cannot lock onto the desired signal, and if it is tracking a signal it may
lose lock for a period of time. All of these effects are undesirable.
The exact nature of the received signal strength for the three primary multipath modes
is the product of the appropriate fade envelope as given in Section 2.6.1 and the
fading patterns examined in this section.
![Page 105: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/105.jpg)
80
2.7 GPS Receiver Context
Having examined the nature of multipath propagation of the GPS L1 signal it is
necessary to consider these findings in the context of the location and limits of the
GPS receiver.
2.7.1 Aspects of Physical Antenna Location
In the previous sections the nature of GPS multipath has been examined. It is apparent
from the three primary modes of specular reflection that the relative delay of the
multipath signal depends solely on the physical relationship of the reflector to the
antenna. For the GPS SPS with a standard ½-chip correlator spacing, signals delayed
by more than 1.5 chips (1466 nanoseconds) are effectively decorrelated and as such
have no impact on the accuracy of the range measurement [1, 13, 14]. For the three
primary modes of specular reflection this decorrelation limit provides a basis for
bounding the physical relationship of the antenna to the terrain within the multipath
environment.
Considering F-mode multipath, the height bound for decorrelation of the multipath
signal can be derived from Table 2.2 and is written approximately as:
θsin
220=Fh (2.39)
This height represents the height of the antenna above a region of terrain that contains
a relatively flat area larger than the first Fresnel zone. The intervening terrain
introduces only blockage and diffractive effects and the multipath is not otherwise
constrained by the physical distance to the area of reflection. Equation (2.39) shows
that the minimum height for decorrelation of the multipath signal occurs at 90 degrees
(zenith) where the required height for decorrelation is 220 metres or higher. For most
land-based applications of GPS, the antenna height would be well below this level. At
![Page 106: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/106.jpg)
81
lower elevation angles the decorrelation height increases markedly and the likelihood
of decorrelation decreases dramatically. It is apparent then that F-mode multipath is
likely to introduce range errors at practical antenna heights, ignoring receiver antenna
gain pattern effects.
For a narrow-correlator with 1/10-chip spacing, signals delayed by more than 1.1
chips (1075 nanoseconds) are effectively decorrelated [15, 16]. The height bound for
decorrelation of the multipath signal in this case is given approximately as:
θsin
160=Fh (2.40)
In this case the minimum decorrelating antenna height is 160 metres.
Similarly for the BA-mode the relative delay is dependant upon the distance of the
antenna from the back-reflector. For a standard correlator receiver, the bound of
decorrelation distance is given approximately by:
θcos
220=BAx (2.41)
For a back-reflector of height Hb, the distance bound must satisfy the inequality
θtan0 BAb xHh −≤≤ (2.42)
In most land-based applications the height of the back-reflector is unlikely to exceed
30 metres. This height represents the height of a typical tall urban structure but is
arbitrary in choice and is selected from a purely practical viewpoint. This results in
values of decorrelation distance from the back-reflector and corresponding height
given in Figure 2.94.
![Page 107: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/107.jpg)
82
220 220.2 220.4 220.6 220.8 221 221.2 221.4 221.6 221.8 2220
5
10
15
20
25
30BA-Mode Decorrelation Bound
Ant
enna
Hei
ght
(m)
Distance (m)
0 degs
1 degs
2 degs
3 degs
4 degs
5 degs
6 degs
7 degs
0 degs
1 degs
2 degs
3 degs
4 degs
5 degs
6 degs
7 degs
Figure 2.94 — BA-mode Decorrelation Distance/Height Bound
The implication of these results—for an arbitrary 30 metre high back reflector—is that
if the antenna is located more than the required bounded distance from the back-
reflector, then the multipath is either decorrelated or non-existent (as far as the GPS
receiver is concerned).
For the signal to exist the antenna must be located below the bound shown, for the
corresponding decorrelation distance. If the antenna is above this height then it is not
located within the back-reflected signal region and is not received. At propagation
angles greater than 8 degrees decorrelation will not occur, since the maximum
distance for the multipath—relative to the existence (or shadow) boundary, Figure
2.95— can never exceed the decorrelation distance. For the narrow-correlator receiver
the equivalent decorrelation distance limits are from 160 metres at 0 degrees to 163
metres at 10.5 degrees.
The actual spatial region of existence for the BA-mode multipath signal is shown
below in Figure 2.95.
![Page 108: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/108.jpg)
83
Hb
ExistenceBoundary
θ
Figure 2.95 — BA-mode Existence-Boundary
At all propagation angles the BA-mode multipath signal exists only within the lower
region (ignoring diffractive effects) defined by the existence-boundary. The relevant
distances related to the limit of the back-reflector height (30 metres) and the
propagation angle is illustrated in Figure 2.96 below.
0 20 40 60 80 100 120 140 160 180 200 2200
5
10
15
20
25
30BA-Mode Existence Boundary
Bac
k R
efle
ctor
(m
)
Distance (m)
5 degs20 degs35 degs60 degs75 degs80 degs
Figure 2.96 — BA-mode Existence Region Metrics
It is obvious, that to avoid BA-mode multipath, the antenna needs to as far as
practicable from any vertical interface, and that the multipath mode is more evident at
low propagation angles.
![Page 109: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/109.jpg)
84
For the BB-mode multipath the existence region is shown in Figure 2.97.
Hb
UpperExistenceBoundary
θ
LowerExistenceBoundary
Zone 2
Zone 1
Figure 2.97 — BB-mode Existence Region
This region is defined by an upper and a lower boundary, which bounds the antenna
height for a given distance from the back-reflector. In addition the region is defined by
two distinct zones related to the process of the generation of the BB-mode multipath
signal (see Section 2.2.3). For the lower height bound the decorrelation distance is
given by
θcos220=BBLx (2.43)
Likewise, the upper height bound limits the decorrelation distance as:
( ) θθ cossin220 bBBU Hx −= (2.44)
![Page 110: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/110.jpg)
85
The height of the antenna as a function of distance from the back-reflector must
satisfy
bBBUBBL Hxhx +≤≤ θθ tantan (2.45)
for the BB-mode multipath signal to be received at the antenna. Applying the bounds
given in equation (2.45) the resultant decorrelation distances are shown in Figure
2.98.
0 10 20 30 40 50 60 70 80 900
50
100
150
200
250BB-Mode Decorrelation Distance
Dis
tanc
e (m
)
Propagation Angle (degs)
At Lower Height BoundAt Upper Height Bound
Figure 2.98 — BB-Mode Decorrelation Distance
The corresponding height bounds are shown below in Figure 2.99.
![Page 111: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/111.jpg)
86
0 10 20 30 40 50 60 70 80 900
50
100
150
200
250BB-Mode Decorrelation Height
Hei
ght
(m)
Propagation Angle (degs)
At Lower BoundAt Upper Bound
Figure 2.99 — BB-mode Antenna Height Bounds for Decorrelation
2.8 Path Loss in the Terrestrial Domain
In general the free space path loss in a radio link is specified by the Friss transmission
formula given by [12],
=
2log10
rP
P
t
r α(2.46)
where α is a function of the frequency and the effective apertures of the transmit and
receive antennas, and r is the separation distance.
In modelling the propagation of GPS in a terrestrial solution domain we need to know
if free space transmission loss is required in the implementation. The GPS satellite is
located approximately 20000 km from the receiving antenna. In addition the
minimum received signal strength specified for the GPS L1 signal is –160 dBw. By
equating these values an estimate of additional path loss can be made, Figure 2.100.
![Page 112: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/112.jpg)
87
0 100 200 300 400 500 600-160.3
-160.25
-160.2
-160.15
-160.1
-160.05
-160
-159.95
Min
imum
Sig
nal L
evel
(dB
w)
Additional Path Distance (km)
Path Loss - Over 20,000km
Figure 2.100 — Additional Path Loss
We see that for additional path distances of 100 km the additional path loss is less
than 0.05 dB. In terms of the physical dimension of a model solution domain the
exclusion of path loss in the model would result in very small errors in the estimate of
the received signal strength. In general the solution domain would not exceed a few
kilometres and the path loss variation over these distances would be overwhelmed by
actual variation of the incident received field strength being modelled.
From a multipath perspective the additional path loss due to the additional distance
travelled by the multipath signal, relative to the LOS path is inconsequential.
2.9 Summary
In this chapter the nature of multipath propagation of the GPS L1 signal has been
examined. The theory of specular reflection, rough surface effects, Fresnel zones, and
diffraction have been applied directly to aid in describing the expected propagation
behaviour of the GPS L1 signal.
Considering specular reflection, an effective coupled CP reflection coefficient was
developed that combines the nature of the reflected RHCP GPS signal with the
![Page 113: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/113.jpg)
88
rejection capability of the GPS antenna. By integrated these parameters, a single
effective circular polarisation reflection coefficient, can be used to describe the nature
of the specular reflection. From the viewpoint of modelling GPS signal propagation,
the necessary constraint of modelling two separate polarisation modes of propagation
has been removed. The adaptation of a single-mode propagation model is thus more
readily achievable.
Three primary propagation modes for specular multipath were introduced. The F-
mode, which describes the specular reflection of the GPS signal from the forward
direction over the ground. The BA-mode, which describes specular reflection from a
vertical interface located behind the GPS antenna, resulting in a back-reflected signal
arriving from above the horizontal plane containing the antenna. Lastly the BB-mode
that describes a double reflected signal that includes a forward and a backward
component of specular reflection, arriving from below the horizontal plane containing
the antenna.
The individual nature of these three primary multipath modes were examined in
relation to the limit of signal fading, as shown in the fade envelopes for the received
signal strength. The fading nature of the modes were also examined from the
perspective of phase-rate of change of the received GPS signal.
The effect of rough surfaces on the reflection of the GPS L1 signal was examined. In
general, rough surfaces result in a reduction of the specular reflection. The rough
surface reduction factor was derived for the GPS L1 signal. In considering GPS
multipath propagation, rough surface effects are considered desirable as they limit the
effect of specular reflection.
Fresnel zone descriptions were derived for the GPS L1 signal. These descriptions
provide a geometric interpretation of the spatial metrics of the physical terrain in
which the GPS L1 signal is propagating. The requirements for specular reflection,
signal blockage and diffraction were given physical meaning. Diffraction of the GPS
L1 signal was investigated and the nature of this multipath examined.
![Page 114: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/114.jpg)
89
Finally the nature of the three modes of multipath propagation were examined within
the context of the physical relationship of the GPS antenna with the terrain, and the
effect of decorrelation of the multipath signals. It was found that in general F-mode
multipath is unlikely to be decorrelated. That BA-mode multipath has a very confined
region of influence, spatially and in terms of propagation angle range. For this mode it
was found that as long as the antenna was located more than approximately 220
metres from the reflecting interface then the multipath mode either did not exist or
was effectively decorrelated. For BB-mode the region of influence is much greater
due to the large region of existence of this multipath mode. However the double-
reflected nature of the BB-mode multipath limits it’s ultimate effect on the GPS
receiver.
These propagation conditions form the theoretical platform for the development of a
model tailored to the propagation of the GPS L1 signal.
![Page 115: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/115.jpg)
90
2.10 References
[1] M. S. Braasch, "On the Characteristics of Multipath Errors in Satellite-Based
Precision Approach and Landing Systems," in Department of Electrical and
Computer Engineering. Athens: Ohio University, 1992, pp. 203.
[2] R. D. J. van Nee, "Multipath Effects on GPS Code Phase Measurements,"
Navigation: Journal of The Institute of Navigation, vol. 39, pp. 177-190, 1992.
[3] H. R. Reed and C. M. Russell, Ultra High Frequency Propagation, Second ed.
New York: Chapman and Hall, 1966.
[4] S. Ramo, J. R. Whinnery, and T. van Duzer, Fields and Waves in
Communication Electronics. New York: John Wiley & Sons, 1984.
[5] H. Bremmer, Terrestrial Radio Waves: Theory of Propagation. London:
Elsevier Publishing Company, 1949.
[6] ITU-R, "Electrical Characteristics of the Surface of the Earth,"
Recommendation Rec. 527-3, 1992.
[7] W. L. Flock, "Propagation Effects on Satellite Systems at Frequencies Below
10 GHz: A Handbook for Satellite Systems Design," NASA December 1987.
[8] P. Beckman and A. Spizzichino, The Scattering of Electromagnetic Waves
From Rough Surfaces. Norwood: Artech House, 1987.
[9] D. E. Kerr, "Propagation of Short Radio Waves." Boston: Boston Technical
Publishers, 1964.
[10] CCIR International Radio Consultive Commitee, "Report 1008 Reflection
from the Surface of the Earth," International Telecommunications Union,
Dubrovnik 1986.
[11] CCIR International Radio Consultive Commitee, "Report 715-2 Propagation
by Diffraction," International Telecommunications Union, Dubrovnik 1986.
[12] J. Doble, Introduction to Radio Propagation for Fixed and Mobile
Communications. Boston: Artech House, 1996.
[13] M. S. Braasch, "Isolation of GPS Multipath and Receiver Tracking Errors,"
Navigation: Journal of The Institute of Navigation, vol. 41, pp. 415-434, 1994.
[14] M. S. Braasch, "GPS and DGPS Multipath Effects and Modeling," in ION
GPS-95 Tutorial: Navtech Seminars, 1995.
![Page 116: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/116.jpg)
91
[15] A. J. van Dierendonck, P. Fenton, and T. Ford, "Theory and Performance of
Narrow Correlator Spacing in a GPS Receiver," presented at The Institute of
Navigation National Technical Meeting, San Diego, CA, 1992.
[16] A. Montalvo and A. Brown, "A Comparison of Three Multipath Mitigation
Approaches for GPS Receivers," presented at 8th International Technical
Meeting of The Satellite Division of The Institute of Navigation., Palm
Springs, California, 1995.
![Page 117: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/117.jpg)
92
THIS PAGE IS INTENTIONALLY BLANK
![Page 118: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/118.jpg)
93
Chapter 3 Overview of Propagation Modelling
This Chapter introduces some of the more popular techniques that can be applied to
solving electromagnetic problems. The requirements for modelling GPS signal
propagation introduces some restrictions on which models are appropriate. In
particular the size of the domain in which the modelling takes place is of primary
importance for modelling GPS signal propagation. The size of the domain will be of
the order of several tens of metres up to hundreds of metres, in both height and range.
In characterising multipath propagation the relative delay of signals needs to be
accurately modelled. With time-domain based models the time information is
inherent, but for frequency-domain models a Fourier-synthesis technique, for deriving
time-domain information, needs to be considered. In this Chapter the applicability of
various methods is compared, for the modelling of GPS signal propagation.
3.1 Review of Maxwell’s Equations
This section provides a brief review of Maxwell’s equations of electromagnetics and
the development of the Helmholtz wave equation [1, 2]. As electromagnetic problems
require the solution or approximate (numerical) solution of these equations it is
appropriate that they be reviewed here. The development is started by defining the
following parameters for a linear, isotropic medium.
B = µ H
D = ε E
J = σ E
Here µ is the permeability, ε is the permittivity, and σ is the conductivity of the
medium. The vector components are:
D — Electric Flux Density ( Displacement current)
![Page 119: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/119.jpg)
94
B — Magnetic Flux Density
E — Electric Field Intensity
H — Magnetic Field Intensity
J — Current Density
Maxwell’s equations in differential form are given by:
∇⋅ D = ρ (3.1)
∇ ⋅ B = ∇ ⋅H = 0 (3.2)
∇ × = − = −EB H∂
∂ µ∂∂t t
(3.3)
∇ × = + = +H JD
EE∂
∂ σ ε∂∂t t
(3.4)
and in integral form by:
D dS⋅ =∫ ∫S VdVρ (3.5)
B dS⋅ =∫S0 (3.6)
E dl B dS⋅ = − ⋅∫ ∫∂
∂ t S(3.7)
H dl J dS D dS⋅ = ⋅ + ⋅∫∫ ∫S St
∂∂
(3.8)
The development of the Helmholtz wave equation is started by using the vector
identity, ( )∇ × ∇ × = ∇ ∇⋅ − ∇A A A2 . The curl equation in H can then be rewritten as:
![Page 120: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/120.jpg)
95
( )∇ × ∇ × = ∇ ∇⋅ − ∇H H H2
now substituting (3.2) and (3.4) gives
( )− ∇ = ∇ × +
= ∇ × +
2H E
EEσ ε
∂∂
σ ε∂∂t t
and substituting (3.3) results in
∇ = +22
2HH H
µσ∂∂
µε∂∂t t
(3.9)
Similarly for the case of E in a source-free region (∇ ⋅ E = 0 )
∇ = +22
2EE E
µσ∂∂
µε∂∂t t
(3.10)
The differential forms for the time-periodic case ( e j tω time dependence) are reduced
by replacing ∂∂ t
with jω and ∂∂
2
2t with −ω 2 giving the vector wave equations,
∇ =2 2H Hγ and ∇ =2 2E Eγ (3.11)
whereγ µσ ω µεω α β= − = +j j2
This is known as the propagation constant with real part α, the attenuation factor, and
imaginary part β, the phase shift constant. Now withγ = jk the three-dimensional
Helmholtz’s wave equations for phasor fields are given by:
∇2E + k2E = 0 and ∇2H + k 2H = 0 (3.12)
![Page 121: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/121.jpg)
96
3.2 Overview of Computational Electromagnetics Techniques
Computational techniques have revolutionised the way in which electromagnetic
problems are analysed. Electrical engineers rely on efficient and accurate computer
models to analyse and evaluate electromagnetic behaviour of antennas, propagation,
scattering and component designs.
Although most electromagnetic problems ultimately involve solving only one or two
partial differential equations subject to boundary constraints, very few practical
problems can be solved without computer-based methods [3]. Computational
electromagnetics involves the development of numerical algorithms for the solution of
Maxwell’s equations, and their subsequent use in analysing electromagnetic problems.
Whereby, analytical techniques make simplifying assumptions about the geometry of
the problem in order to apply a closed-form solution, numerical techniques attempt to
solve the fundamental field equations directly, subject to the boundary constraints
posed by the geometry.
Numerical techniques generally require more computation than analytical techniques
but they are very powerful EM analysis tools. Without making a priori assumptions
about which field interactions are most significant, numerical techniques analyse the
entire geometry, and calculate the solution based on a full-wave analysis [4]. A
number of different numerical techniques for solving electromagnetic problems are
available. The following sections provide an introduction to several of the most
widely used techniques in computational electromagnetics.
3.2.1 Finite-Elements Technique
An increasing availability of computer resources coupled with a desire to model more
complex electromagnetic problems has resulted in a wave of renewed interest in
Finite Element (FE) methods for solving EM radiation problems. FE techniques [3-8]
require the entire domain to be divided into several sub-domains or elements. Certain
![Page 122: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/122.jpg)
97
boundary conditions are satisfied at the boundaries and each element may have
completely different material properties from those of neighbouring elements.
The first step in FE analysis is to divide the configuration into a number of small
elements, the selection of the element shape is based upon the geometry, material
constants, excitations and boundary constraints of the problem. In each finite element,
a simple variation of the field quantity is assumed. The corners of the elements are
known as nodes. The basis of the FE analysis is to determine the field quantities at the
nodes.
Most FE methods use variational techniques, minimising or maximising an
expression that is known to be stationary about the true solution. Generally, finite-
element analysis techniques solve for the unknown field quantities by minimising an
energy functional. The energy functional is an expression describing all the energy
associated with the configuration being analysed. For 2-D, time-harmonic problems
this functional is a surface integral while for 3-D problems, it can be represented as a
volume integral,
FH E J E
= + −⋅
∫µ ε
ω
2 2
2 2 2v jdv (3.13)
The first two terms in the integrand represent the energy stored in the magnetic and
electric fields and the third term is the energy dissipated (or supplied) by conduction
currents. Expressing H in terms of E and setting the derivative of this functional with
respect to E equal to zero, an equation of the form f(J, E) = 0 is obtained. An
approximation(k-th order) of this function is then applied at each of the N nodes and
boundary conditions are satisfied, resulting in a system of equations,
J
J
J
y y y
y y y
y y y
E
E
En
n
n
n n nn n
1
2
11 12 1
21 22 2
1 2
1
2
M
K
K
M M OM
K
M
=
(3.14)
![Page 123: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/123.jpg)
98
The values of J in Equation (3.14) are referred to as the source terms. They represent
the known excitations. The elements of the Y-matrix are functions of the problem
geometry and boundary constraints. Since each element only interacts with elements
in its own neighbourhood, the Y-matrix is generally sparse. The terms of the vector E
represent the unknown electric field at each node. These values are obtained by
solving the system of equations.
In general, finite element techniques are appropriate for modelling complex
inhomogeneous configurations. However, they do not model unbounded radiation
problems as effectively as other techniques. The major advantage that finite element
methods have over other EM modelling techniques stems from the fact that the
electrical and geometric properties of each element can be defined independently.
This permits the problem to be set up with a large number of small elements in
regions of complex geometry and fewer, larger elements in relatively open regions.
Thus it is possible to model configurations that have complicated geometry’s and
many arbitrarily shaped dielectric regions in a relatively efficient manner
3.2.2 Finite-Difference Time-Domain Technique
Finite difference time domain (FD-TD) techniques [3-6, 8] require the entire domain
volume to be meshed. Normally, this mesh must be uniform, so that the mesh density
is determined by the smallest detail of the configuration. The FD-TD is, as the name
suggests, a time-domain technique well-suited to transient analysis problems. The
technique is best explained by the numerical FD-TD solution of the one-dimensional
scalar wave equation which is given by:
∂∂
∂∂
2
22
2
2
u
tc
u
x= (3.15)
where ( )u u x t= , .
Now defining a function of the form u x t F x ct G x ct( , ) ( ) ( )= + + − , where F and G
are arbitrary, and differentiating twice with respect to t and x provides
![Page 124: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/124.jpg)
99
( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( )
∂∂
∂∂
∂∂
∂∂
u
t
dF x ct
d x ct
x ct
t
dG x ct
d x ct
x ct
tcF x ct cG x ct
u
tc F x ct c c G x ct c F x ct c G x ct
= ++
⋅+
+ −−
⋅−
= ′ + − ′ −
= ′′ + − − ′′ − = ′′ + + ′′ −
( )
( )
( )
( )2
22 2 2
and
( ) ( ) ( ) ( )
( ) ( )
∂∂
∂∂
∂∂
∂∂
u
x
dF x ct
d x ct
x ct
x
dG x ct
d x ct
x ct
xF x ct G x ct
u
xF x ct G x ct
= ++
⋅+
+ −−
⋅−
= ′ + + ′ −
= ′′ + + ′′ −
( )
( )
( )
( )2
2
Substituting these equations into the equation (3.15) gives
c2 ′ ′ F x + ct( ) + c2 ′ ′ G x − ct( ) = c2 ′ ′ F x + ct( )+ ′ ′ G x − ct( )[ ]
F and G are known as propagating wave solutions and the identity holds regardless of
their choice. After some time, ∆t, the wave solution F must have moved to the left (-x
direction) since the argument of F has increased by c∆t, the spatial part of the
argument has to correspondingly decrease by c∆t. The converse is true for G and
therefore F(x+ct) and G(x-ct) are leftward and rightward travelling waves
respectively. The factor c represents the wave velocity in the ±x direction. A Taylor’s
series expansion of u x tn( , ) for a fixed tn , about the space point xk to ( )x xk + ∆ gives:
u x + ∆x( )tn
= u x k ,t n+ ∆x ⋅
∂u
∂x x k ,tn
+∆x 2
2⋅∂ 2u
∂x2x k ,tn
+∆x 3
6⋅∂3u
∂x3xk ,t n
+∆x4
24⋅
∂4u
∂x4ξ1,t n
The last term is an error term where ξ1 is a space point located somewhere in the
interval ( )x x xk k, + ∆ . Now for the Taylor’s series expansion of u x tn( , ) for a fixed tn ,
about the space point xk to ( )x xk − ∆ .
![Page 125: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/125.jpg)
100
( )u x x u xu
x
x u
x
x u
x
x u
xt x tx t x t x t t
nk n
k n k n k n n
+ = − ⋅ + ⋅ − ⋅ + ⋅∆ ∆ ∆ ∆ ∆,
, , , ,
∂∂
∂∂
∂∂
∂∂ ξ
2 2
2
3 3
3
4 4
42 6 242
Here ξ 2 in the error term represents a space point in the interval ( )x x xk k, − ∆ .
Adding these two expressions results in
( ) ( )u x x u x x u xu
x
x u
xt x tx t t
nk n
k n n
+ + − = + ⋅ + ⋅∆ ∆ ∆ ∆2
122
2
2
4 4
4
3
,, ,
∂∂
∂∂ ξ
With ξ 3 a space point in the new interval ( )x x x xk k− +∆ ∆, . Rearranging to give an
expression for the second derivative
( ) ( ) ( )( )
( )[ ]∂∂
β2
2 2
22u
x
u x x u x u x x
xx
x t
k
tk nn
,
=+ − + −
+∆ ∆
∆∆ (3.16)
Equation (3.16) is a second-order, central-difference approximation to the second
partial space derivative of u, with β representing the error term. Using subscript k for
spatial position and superscript n for observation epoch gives
( )( )[ ]∂
∂β
2
21 1
2
22u
x
u u u
xx
x t
kn
kn
kn
k n,
=− +
++ −
∆∆ (3.17)
Similarly the second-order, central difference approximation to the second partial time
derivative of u is given by
( )( )[ ]∂
∂β
2
2
1 1
2
22u
t
u u u
tt
x t
kn
kn
kn
k n,
=− +
++ −
∆∆ (3.18)
For these ukn is a wave or field quantity calculated at the spatial point x k xk = ∆ and
epoch t n tn = ∆ .
![Page 126: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/126.jpg)
101
Substituting into the scalar wave equation (3.15),
( )( )[ ] ( )
( )[ ]u u u
tt c
u u u
xxk
nkn
kn
kn
kn
kn+ −
+ −− ++ =
− ++
1 1
2
2 2 1 12
22 2
∆∆
∆∆β β (3.19)
and solving for the latest value of u at the spatial point k gives
( )( )
( )[ ] ( )[ ]u c tu u u
xu u t xk
n kn
kn
kn
kn
kn+ + − −=
− +
+ − + +1 2 1 1
21 2 22
2∆∆
∆ ∆β β (3.20)
This is an explicit second-order expression for ukn+1 in which all wave quantities given
in the RHS of the equation are known from previous values at epochs tn and tn−1 . The
FD-TD solution is thus obtained by solving ukn+1 for all space points and iterating in
time.
Similarly the FD-TD method can be applied to solve Maxwell’s time-dependent curl
equations (2.3, 2.4), repeated here
t∂∂µ H
E −=×∇ and t∂
∂εσ EEH +=×∇ (3.21)
The domain is represented as two interleaved grids. Each grid contains respective
points of electric and magnetic field values. With the spatial domain as represented in
Figure 3.1, a first-order central-difference approximation of curl equation in E (2.21)
is expressed as:
[ ] [ ]1
4 1 2 3 40
01
01
∆ ∆ ∆x yE E E E
tH Hz
nx
nz
ny
ny
ny
n+ − − = − −+ −µ(3.22)
This equation (3.22) is solved for Hy0n+1 . The same technique is applied to the
alternate curl equation of the magnetic field, and in this way, at each time step, the
![Page 127: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/127.jpg)
102
magnetic and electric field components are alternately solved on their respective grids.
Thus the field is time-step propagated throughout the grid until either a steady-state
solution or the desired response is obtained.
Ex1
Ey1
Ez1
Ex2
Ey2
Ey3Ex4
Hx0
Ez4
Ez3
Hz0
Hy0
z
y
x
scatterer
Figure 3.1 — 3-D FD-TD grid
3.2.3 Finite-Difference Frequency-Domain Technique
Although conceptually the Finite Difference Frequency Domain (FD-FD) method [3-
6, 8] is similar to the Finite Difference Time Domain (FD-TD) method, from a
practical standpoint it is more closely related to the finite element method. Like FD-
TD, this technique results from a finite difference approximation of Maxwell’s curl
equations. However, the time-harmonic versions of these equations are employed, and
since there is no time stepping it is not necessary to keep the mesh spacing uniform.
Therefore optimal FD-FD meshes generally resemble optimal finite element meshes.
The FD-FD technique generates a system of linear equations, where the corresponding
matrix is sparse like that of the finite element method. Although it is conceptually
much simpler than the finite element method, very little attention has been devoted to
this technique in the literature.
3.2.4 Method of Moments
Like finite-element analysis, the moment method [3-6, 8, 9]] is a technique for solving
complex integral equations (2.7, 2.8) by reducing them to a system of simpler linear
![Page 128: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/128.jpg)
103
equations. In contrast to the approach of the finite element method however, moment
methods employ a technique known as the method of weighted residuals. Harrington
[10]] first demonstrated the power and flexibility of this numerical technique for
solving problems in electromagnetics.
All weighted residual techniques begin by establishing a set of trial solution functions
with one or more variable parameters. The residuals are a measure of the difference
between a trial solution and the true solution. The variable parameters are determined
in a manner that guarantees a best fit of the trial functions based on a minimisation of
the residuals. The equation solved by moment method techniques is generally a form
of the Electric Field Integral Equation (EFIE) or the Magnetic Field Integral Equation
(MFIE). Both of these equations can be derived from Maxwell’s equations by
considering the problem of a field scattered by a perfect conductor (or a lossless
dielectric). As an example, the EFIE is written in the form,
( )E J= f e (3.23)
where E is the incident field and J is the induced current. The form of the integral
equation used determines which types of problems a moment-method technique is
best suited to solve. For example one form of the EFIE may be particularly well suited
for modelling thin-wire structures, while another form is better suited for analysing
metal plates. Usually these equations are expressed in the frequency domain, however
the method of moments can also be applied in the time domain. The first step in the
moment-method solution process is to expand J as a finite sum of basis (or
expansion) functions,
J ==∑ J bi ii
M
1
(3.24)
where bi is the i-th basis function and Ji is an unknown coefficient. Next, a set of M
linearly independent weighting functions, wj, are defined. An inner product of each
weighting function is formed with both sides of the equation being solved. This
![Page 129: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/129.jpg)
104
results in a set of M independent equations, and by expanding J using equation (3.24)
the result is a set of M equations in M unknowns.
[ ] [ ][ ]E J= Z (3.25)
Where the vector E contains the known incident field quantities and the terms of the
Z-matrix are functions of the geometry. The unknown coefficients of the induced
current are the terms of the J vector. These values are obtained by solving the system
of equations. Other parameters such as the scattered electric and magnetic fields can
be calculated directly from the induced currents.
Depending on the form of the field integral equation used, moment methods can be
applied to configurations of conductors only, homogeneous dielectrics only, or very
specific conductor-dielectric geometry’s. Moment method techniques applied to
integral equations are not very effective when applied to arbitrary configurations with
complex geometry’s or inhomogeneous dielectrics. Moment method techniques do an
excellent job of analysing a wide variety of important three-dimensional
electromagnetic radiation problems, particularly the modelling of wire antennas or
wires attached to large conductive surfaces. They are also widely used for antenna and
electromagnetic scattering analysis.
3.2.5 Geometrical and Uniform Theory of Diffraction
The Geometrical Theory of Diffraction was proposed by Keller [11], as an extension
and improvement to the classic ray-based high-frequency approximation, Geometric
Optics (GO), by introducing additional rays (diffraction coefficients) to account for
diffraction [12]. Diffraction is a local phenomena at high frequencies and the
behaviour of the diffracted wave at edges, corners, and surfaces can be determined
from an asymptotic form of the exact solution for simpler canonical problems. For
example, the diffraction around a sharp edge is found by considering the asymptotic
form of the solution for an infinite wedge. Both GO and GTD are only accurate when
the dimensions of objects being analysed are large relative to the wavelength of the
![Page 130: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/130.jpg)
105
field. In general, as the wavelengths of an electromagnetic field approach zero, the
fields can be determined using geometric optics.
In GTD the high-frequency approximate solutions for general scatterers are derived
from a set of exact solutions which in turn are derived for a selection simple
geometric shapes. The common canonical solutions are for infinite wedges,
conducting half sheets, circular disks, cylinders, spheres, and others. Therefore the
problem geometry needs to constructed around these various shapes before a solution
can be found.
The Uniform Theory of Diffraction (UTD) is an extension of the GTD made by
implementing improvements to the diffraction coefficients [13]. The use of uniform
coefficients solves one limitation of the GTD, the infinities produced by the
asymptotic evaluation of the Fresnel integral used in deriving the diffraction
coefficients.
GTD and UTD are approximation methods, but are still analytical; their physical
meaning being quite clear, in contrast to other computational electromagnetics
techniques. No special techniques are required for the implementation of GTD/UTD
codes, but implementations that require detailed terrain are very complex. and have
high computational loads.
3.2.6 Generalised Multipole Technique
The Generalised Multipole Technique (GMT) [4, 5] is a relatively new method for
analysing EM problems. It is a frequency domain technique that (like the method of
moments) is based on the method of weighted residuals. However, this method is
unique in that the expansion functions are analytic solutions of the fields generated by
sources located some distance away from the surface where the boundary condition is
being enforced.
![Page 131: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/131.jpg)
106
Moment methods generally employ expansion functions representing quantities such
as charge or current that exist on a boundary surface. The expansion functions of the
GMT are spherical wave field solutions corresponding to multipole sources. By
locating these sources away from the boundary, the field solutions form a smooth set
of expansion functions on the boundary and singularities on the boundary are avoided.
Like the method of moments, a system of linear equations is developed and then
solved to determine the coefficients of the expansion functions that yield the best
solution. Since the expansion functions are already field solutions, it is not necessary
to do any further computation to determine the fields. Conventional moment methods
determine the currents and/or charges on the surface first and then must integrate
these quantities over the entire surface to determine the fields. This integration is not
necessary at any stage of the GMT solution.
There is little difference in the way dielectric and conducting boundaries are treated
by the GMT. The same multipole expansion functions are used. For this reason, a
general purpose implementation of the GMT models configurations with multiple
dielectrics and conductors much more readily than a general purpose moment-method
technique. On the other hand, moment method techniques, which employ expansion
functions that are optimised for a particular type of configuration (e.g. thin wires), are
generally much more efficient at modelling that specific type of problem.
3.2.7 Parabolic Equation Method
The parabolic equation was first proposed for use in radiowave propagation by
Leontovich and Fock in 1946 [14, 15]. However at that time there were no suitable
numerical techniques for solving the equation and hence the method was not pursued.
Past research has seen the parabolic equation being applied to seismology and
underwater acoustics problems [16]. From this work and with advances in numerical
techniques there has been increased interest in applying the method to electromagnetic
propagation problems. An excellent reference may found in [17].
![Page 132: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/132.jpg)
107
The parabolic equation is a reduction of the wave equation making use of the fact that
horizontal variations in refractive index occur at a slower rate than vertical variations.
The starting point for the development of an electromagnetic parabolic equation is
with the Helmholtz wave equation (3.12) for a field component, ( )ψ x z, with
assumed time dependence e− jωt .
∇ + =20
2 2 0ψ ψk n (3.26)
where k0
2=
πλ
is the free space wavenumber, and nk
k=
0
is the refractive index.
Since the propagation is above a spherical earth, a spherical co-ordinate system would
be appropriate. However, the problem can be greatly simplified if some assumptions
are made. The first assumption is that propagation takes place over a flat earth; thus
allowing the use of a cylindrical co-ordinate system. The field is then assumed
invariant in azimuth (azimuthal symmetry). Equation (3.26) can now be expressed in
cylindrical co-ordinates (r,φ, z)—with the φ - co-ordinate terms removed. For the
problem of radiowave propagation the co-ordinate r represents range distance from
the source, and z represents the height above the earth.
∂ ψ∂
∂ψ∂
∂ ψ∂
ψ2
2
2
2 02 21
0r r r z
k n+ + + = (3.27)
It is now assumed that the solution of Equation (3.27) is in terms of a Hankel function
that satisfies the Bessel differential equation
( ) ( ) ( ) ( ) ( ) ( )∂∂
∂∂
2 10
2
10
02 1
0
10
H k r
r r
H k r
rk H k ro o
o+ + = (3.28)
![Page 133: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/133.jpg)
108
and takes the form
( ) ( ) ( ) ( )ψ r z u r z H k r, ,= 01
0 (3.29)
which is an outgoing cylindrical wave solution. The envelope function u(r, z) is
assumed to be slowly varying in range. Substituting the trial solution (3.29) into
equation (3.27) and using the Hankel function property given by equation (3.28)
gives:
( ) ( )( ) ( ) ( )∂
∂∂
∂∂∂
∂∂
2
2 10
10
2
2 02 22 1
1 0u
r H k r
H k r
r r
u
r
u
zk n u
o
o+ +
+ + − = (3.30)
Using the far-field assumption,k0r ⟩⟩ 1, the Hankel function is then given by the phase
factor
( ) ( )H k r eojk r1
00≈ (3.31)
Substitution of this approximation into equation (3.30) yields the simplified elliptic
equation,
( )∂∂
∂∂
∂∂
2
2 0
2
2 02 22 1 0
u
rjk
u
r
u
zk n u+ + + − = (3.32)
There are several methods for deriving various parabolic forms from equation (3.32).
One method is to assume that,
2 0
2
2jku
r
u
r
∂∂
∂∂
⟩⟩ (3.33)
which is valid since the main radial dependence of the field is contained in the e jkr
term of the Hankel function, while the envelope function u is slowly varying with
range [18]. With this approximation the following wave equation is obtained.
![Page 134: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/134.jpg)
109
( )2 1 00
2
2 02 2jk
u
r
u
zk n u
∂∂
∂∂
+ + − = (3.34)
This is the Standard Parabolic Equation (SPE) and is limited in propagation angle
(≈15°) by the paraxial approximation given by Equation (3.33). Equation (3.34) can
be solved by marching the solution out in range, see Figure 3.2, since it is
fundamentally an open boundary problem.
∆x
∆z
0,0
Figure 3.2 — PE solution domain
3.3 Comparison of Modelling Techniques
The use of any particular computational electromagnetic technique is application
dependent. Although it is theoretically possible to base all computational
electromagnetics codes on one numerical method, it is obvious that this is not possible
from a practical viewpoint. The selection of an appropriate numerical technique for
modelling GPS signal propagation is the subject of this section.
![Page 135: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/135.jpg)
110
3.3.1 Requirements for Modelling GPS Signal Propagation
The study, to be undertaken, is concerned with modelling the propagation of GPS
signals in the local vicinity. The effects of secondary path signals also need to be
modelled. Therefore the main criterion is the size of the domain over which the
modelling is to take place. The wavelength of the GPS L1 frequency is of the order of
19 cm. For representative domains of say, 100 metres, the total size of the domain,
expressed in wavelengths, will be of a size of about 500.
3.3.2 Comparison of EM Techniques
Table 3.1 [3] gives an indication of the applicability of the EM modelling techniques
discussed in the previous sections, as a function of performance for domains ranging
in size from less than a wavelength to domains greater than 100 wavelengths.
Table 3.1 — Model Comparison
As can be seen, the methods that are applicable to the requirements of modelling GPS
signal propagation, solely based on domain size, are the high-frequency techniques
and the parabolic equation method. The high-frequency techniques include the ray-
based methods of GTD and UTD.
Modelling Approach <λ/10 Åλ Å10λ Å100λ >100λParabolic Equation Method á á á á áHigh-Frequency Techniques à à á á áMethod of moments á á à à àFD-TD and FE methods à á á á àHybrid Methods á á á à à
![Page 136: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/136.jpg)
111
The GTD/UTD implementation, requires that complex arbitrary terrain shapes be
represented as combinations of canonical problems, and therefore, to include the
appropriate diffraction coefficients requires a complex representation of the terrain
elements. To more fully characterise GPS multipath propagation accurate
representations of terrain are required. It is the intention in this work to use accurate
Digital Terrain Models (DTM) as input to the EM modelling code. The use of
GTD/UTD models preclude the use of DTMs as input for the terrain.
In addition to terrain implementation problems, there are many different propagation
modes, representing families of rays, that need to be considered and calculated in ray-
based code. Thus the resultant GTD/UTD code is complex and has long development
times. Finally these high-frequency methods are only suitable when the structures are
larger than a wavelength.
From these considerations it is determined that the most applicable technique is the
Parabolic Equation method. Not only does it allow for the large electrical domain
sizes required for the analysis but it has no apparent limitations for the analysis of
smaller-scale structures. In addition, it will allow the use of DTMs for terrain input, is
simple to implement, and is not nearly as computationally intensive as GTD/UTD
implementations
![Page 137: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/137.jpg)
112
3.4 References
[1] S. Y. Liao, Microwave Devices and Circuit Theory, 2nd ed. Englewood Cliffs
N.J.: Prentice Hall, 1985.
[2] D. K. Cheng, Field and Wave Electromagnetics. Reading, MA: Addison-
Wesley, 1989.
[3] K. Umashankar and A. Taflove, Computational Electromagnetics. Boston,
MA: Artech House, 1993.
[4] T. H. Hubling, “Survey of Numerical Electromagnetic Modeling Techniques,”
University of Missouri-Rolla TR91-1-001.3, 1991.
[5] C. Hafner, The Generalized Multipole Technique for Computational
Electromagnetics. Boston: Artech House, 1990.
[6] N. Ida, Numerical Modeling for Electromagnetic Non-Destructive Evaluation.
London: Chapman & Hall, 1995.
[7] P. P. Silvester and R. L. Ferrari, Finite Elements for Electrical Engineers,
Second ed. Cambridge: Cambridge University Press, 1990.
[8] E. Yamashita, “Analysis Methods for Electromagnetic Problems,” . Boston:
Artech House, 1990.
[9] J. T. Johnson, R. T. Shin, J. C. Eidson, L. Tsang, and J. A. Kong, “Method of
Moments Model for VHF Propagation,” IEEE Transactions on Antennas &
Propagation, vol. 45, pp. 115-125, 1997.
[10] R. F. Harrington, Field Computation by Moment Methods. New York:
Macmillan, 1968.
[11] J. B. Keller, “Geometrical Theory of Diffraction,” Journal of the Optical
Society of America, vol. 52, pp. 116-130, 1962.
[12] M. Ando, “The Geometrical Theory of Diffraction,” in Analysis Methods for
Electromagnetic Problems , E. Yamashita, Ed. Boston: Artech House, 1990,
pp. 213-242.
[13] D. A. McNamara, C. W. I. Pistorius, and J. A. G. Malherbe, Introduction to
the Uniform Geometrical Theory of Diffraction. Boston: Artech House, 1990.
![Page 138: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/138.jpg)
113
[14] M. A. Leontovich and V. A. Fock, “Solution of the Problem of Propagation of
Electromagnetic Waves along the Earth's Surface by the Method of Parabolic
Equations,” Journal of Physics of the USSR, vol. 10, pp. 13-24, 1946.
[15] V. A. Fock, Electromagnetic Diffraction and Propagation Problems. Oxford:
Pergamon, 1965.
[16] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, Computational
Ocean Acoustics. New York: AIP Press, 1994.
[17] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, “Parabolic
Equations,” in Computational Ocean Acoustics , R. T. Beyer, Ed. New York:
AIP Press, 1994.
[18] B. W. Parkinson and J. R. Spilker Jr., “Global Positioning System: Theory and
Applications Volume I,” in Progress in Astronautics and Aeronautics Series,
vol. 163, P. Zarchen, Ed. Washington: American Institute of Aeronautics and
Astronautics, 1996.
[19] C. D. McGillem and G. R. Cooper, Continuous and Discrete Signal and
System Analysis, Third ed. Philadelphia: HRW Saunders, 1991.
[20] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, “Broadband
Modelling,” in Computational Ocean Acoustics , R. T. Beyer, Ed. New York:
AIP Press, 1994.
[1] D. K. Cheng, Field and Wave Electromagnetics. Reading, MA: Addison-
Wesley, 1989.
[2] S. Y. Liao, Microwave Devices and Circuit Theory, 2nd ed. Englewood Cliffs
N.J.: Prentice Hall, 1985.
[3] K. Umashankar and A. Taflove, Computational Electromagnetics. Boston,
MA: Artech House, 1993.
[4] T. H. Hubling, "Survey of Numerical Electromagnetic Modeling Techniques,"
University of Missouri-Rolla TR91-1-001.3, September 1 1991.
[5] C. Hafner, The Generalized Multipole Technique for Computational
Electromagnetics. Boston: Artech House, 1990.
[6] N. Ida, Numerical Modeling for Electromagnetic Non-Destructive Evaluation.
London: Chapman & Hall, 1995.
[7] P. P. Silvester and R. L. Ferrari, Finite Elements for Electrical Engineers,
Second ed. Cambridge: Cambridge University Press, 1990.
![Page 139: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/139.jpg)
114
[8] E. Yamashita, "Analysis Methods for Electromagnetic Problems." Boston:
Artech House, 1990.
[9] J. T. Johnson, R. T. Shin, J. C. Eidson, L. Tsang, and J. A. Kong, "Method of
Moments Model for VHF Propagation," IEEE Transactions on Antennas and
Propagation, vol. 45, pp. 115-125, 1997.
[10] R. F. Harrington, Field Computation by Moment Methods. New York:
Macmillan, 1968.
[11] J. B. Keller, "Geometrical Theory of Diffraction," Journal of the Optical
Society of America, vol. 52, pp. 116-130, 1962.
[12] M. Ando, "The Geometrical Theory of Diffraction," in Analysis Methods for
Electromagnetic Problems, E. Yamashita, Ed. Boston: Artech House, 1990,
pp. 213-242.
[13] D. A. McNamara, C. W. I. Pistorius, and J. A. G. Malherbe, Introduction to
the Uniform Geometrical Theory of Diffraction. Boston: Artech House, 1990.
[14] M. A. Leontovich and V. A. Fock, "Solution of the Problem of Propagation of
Electromagnetic Waves along the Earth’s Surface by the Method of Parabolic
Equations," Journal of Physics of the USSR, vol. 10, pp. 13-24, 1946.
[15] V. A. Fock, Electromagnetic Diffraction and Propagation Problems. Oxford:
Pergamon, 1965.
[16] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, Computational
Ocean Acoustics. New York: AIP Press, 1994.
[17] M. F. Levy, Parabolic Equation Methods for Electromagnetic Wave
Propagation, 1 ed. London: IEE, 2000.
[18] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, "Parabolic
Equations," in Computational Ocean Acoustics, R. T. Beyer, Ed. New York:
AIP Press, 1994.
[19] B. W. Parkinson and J. R. Spilker Jr., "Global Positioning System: Theory and
Applications Volume I," in Progress in Astronautics and Aeronautics, vol.
163, P. Zarchen, Ed. Washington: American Institute of Aeronautics and
Astronautics, 1996.
![Page 140: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/140.jpg)
115
Chapter 4 GPS Parabolic Equation Model
4.1 Development of the PE for Electromagnetic Propagation
The parabolic equation method was first proposed, as a solution method for
propagation of electromagnetic waves, by Leontovich and Fock [1, 2]. Since that time
the PE method has been used for physics, seismic, atmospheric and underwater
acoustics, and electromagnetics applications [3].
The potential of the PE method was not fully realised until an efficient numerical
technique known as the Fourier split-step was introduced, to the ocean acoustics
community, by Hardin and Tappert [4]. The PE method was then proposed as a useful
modelling technique for seismic wave propagation [5] as well as receiving increased
attention in ocean acoustics [6].
The acoustic PE method was modified for electromagnetics by Ko et al [7]. Both
Craig and Dockery [8, 9] continued research on the PE—using the method to model
electromagnetic wave propagation in the troposphere. Development of the PE method,
for tropospheric propagation modelling, continued in the late 1980’s, early 1990’s,
with Craig [10-14], Levy [15-18], Dockery[19], and others [20-25], refining the
standard PE for communications and radar predictions.
The PE research for ocean acoustics was, in the meantime, developing methods for
dealing with wide-angle propagation [26-36], a requirement for particular ocean
environment problems.
Further significant contributions to the development of the PE method have been
made. In particular the implementation of backscatter [37-41] and the inclusion of
terrain interactions [42-50] have extended the usefulness of the PE method, for
various applications. An excellent reference on the PE method can be found in [51].
![Page 141: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/141.jpg)
116
The use of the PE method for modelling GPS signal propagation is one such
application to be enabled by this ongoing research and significantly, was first
introduced by Walker [52, 53] for modelling the operation of GPS in harsh
environments. In this chapter the PE modelling method for GPS is introduced.
Limitations of the standard PE forms are investigated. Improvements of the original
PE model are made including finite impedance conditions at the boundary and two-
way propagation through the inclusion of backscatter.
4.2 The Free-Space Parabolic Equation
We start with the simplified elliptic equation introduced in Chapter 3 as equation
(3.32) and repeated here for clarity.
( ) 012 2202
2
02
2
=−+++ unkz
u
r
ujk
r
u
∂∂
∂∂
∂∂
(4.1)
By defining the operators
rP
∂∂= and
2
2
20
2 1
zknQ
∂∂+= (4.2)
the simplified elliptic equation given by equation (4.1) can now be written in the
form,
( )[ ] 012 2200
2 =−++ uQkPjkP (4.3)
The factorisation of this equation permits it to be formulated into incoming and
outgoing wave components.
![Page 142: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/142.jpg)
117
( )( ) [ ] 0,00000 =−++−+ uQPjkuQjkjkPQjkjkP
(4.4)
where
[ ] QPPQQP −=, (4.5)
is the commutator of the P and Q operators. If the refractive index n is independent of
range then the operators commute and the final term in equation (4.4) is zero. This is
true for free-space propagation but is also assumed valid for weak range dependence
and thus the commutator term can be neglected. The outgoing component of equation
(4.4) is now selected, giving
( ) 010 =−+ uQjkPu (4.6)
By replacing the operator notation, the result is a one-way wave equation, exact for
range-independent environments, within the limits imposed by the far-field
approximation, and is given by
01
12
2
20
20 =
+−+ u
zknjk
r
u
∂∂
∂∂
(4.7)
This equation is evolutionary in range but is not easily solved, due to the inclusion of
the refractive index term, n. In this work we are concerned only with propagation
effects from localised terrain interaction, and not atmospheric effects. If the refractive
index is set to 1 (free-space propagation) then the result is the Free-Space Parabolic
Equation (FSPE) which is given by
![Page 143: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/143.jpg)
118
01
112
2
20
0 =
+−+ u
zkjk
r
u
∂∂
∂∂
(4.8)
Replacing r with x, to emphasise that this is a two-dimensional problem, and by
taking a Taylor series expansion of the square-root operator we can make use of the
Fourier transform property
( ) ( ) ( )pXjpzxdz
d nF
n
n
⇔ (4.9)
This allows an efficient Fourier transform based stepping technique to be used, where
the solution at a range-step ( x∆ ) is given by
( ) ( )[ ]
=∆+
−−∆
− zxuFeFzxxuk
pxjk
,,11
12
2
(4.10)
Here p is the vertical wave-number and is related to k by θsinkp = , with θ , the
propagation angle relative to the horizontal. The p-domain defines the angular
spectrum (Chapter 3) of the field, and together with the z-domain (spatial domain),
they form a Fourier transform pair. The exponential term in equation (4.10) is often
referred to as the PE propagator—which may take different forms for different PE
implementations. Specifically for this free-space implementation it will be known as
the FSPE propagator.
As can be seen from the form of equation (4.10) an initial field condition is defined at
x=0, and the solution marched out in range in discrete steps using Fourier transforms.
The solution of the FSPE is an exact solution and is not limited in either propagation
angle or range. This is not true for PE forms that include refractive index terms. In the
![Page 144: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/144.jpg)
119
following section alternate, traditional PE models are examined for their limitations in
modelling GPS signal propagation.
4.3 Limitations of Refractive Index Terms in PE Forms
For propagation in media, various forms of parabolic equation can be derived from
equation (4.7) by using approximations for the pseudo-differential operator Q. The
resultant non-free-space parabolic equations may then be solved by a variety of
numerical techniques, but none are as efficient as the Fourier-step technique used to
solve the FSPE.
The development of these alternate PE forms is initiated by writing the square-root
operator Q given in equation (4.2) as
qQ += 1 (4.11)
where
µε +=q (4.12)
with
12 −= nε and 2
2
20
1
zk ∂∂µ = (4.13)
These abbreviations allow the operator to be cast in a form that implies something
about the nature of the medium and the angle of propagation. This is shown in the
following development.
Consider a trial plane-wave solution, for the reduced function, of the form
![Page 145: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/145.jpg)
120
( )θθ sincos zrjkeu ±= (4.14)
The medium wave-number can be related to the horizontal (kr) and vertical (kz or p )
wave-numbers by the dispersion relation
222zr kkk += (4.15)
giving
θsinkkz = (4.16)
where θ is the angle of propagation with respect to the horizontal.
Using the trial solution given by equation (4.14) in equation (4.7) allows evaluation of
the differential operator µ.
20
2
k
kz−=µ (4.17)
by making use of equation (4.16) and substituting for the refractive index, n, gives
θµ 22 sinn−= (4.18)
Thus the operator µ is seen to be a function of the propagation angle and refractive
index, while ε is a function only of the refractive index. A further implication
becomes apparent when q is reformed.
![Page 146: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/146.jpg)
121
( ) θµε 222 sin1 nnq −−=+= (4.19)
Using Snell’s law, the refractive index can be related to the propagation angles
through
θθ
cos
cos 0=n (4.20)
giving
( ) 022
20
2
sin1sin1cos
cos θθθθ −=−−=q (4.21)
The implication of this result is that the operator q is only a function of propagation
angle from the source.
Approximations of the Q-operator can be written in a general rational-linear form as
qba
qbaqQ
11
001++≈+= (4.22)
This equation and equation (4.7) provide the basis for the development of a general
parabolic wave equation given by
02
2
2
3
=+++z
uDCu
rz
uB
r
uA
∂∂
∂∂∂
∂∂ (4.23)
where
![Page 147: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/147.jpg)
122
( )
( ) ( )( )( )( )01
0
201010
20
1
211
1
1
bbk
jD
nbbaajkC
k
bB
nbaA
−=
−−+−=
=
−+=
This equation can be solved by a variety of methods including finite-differences,
finite-elements, and in certain cases, the Fourier split-step technique [4]. Various
coefficients have been used in parabolic equation implementations of equation (4.23),
in acoustics, seismology, and electromagnetic propagation studies. The following
table summarises the coefficients for three of the most common.
Coefficients Tappert Claerbout (Padé 1) Greene
a0 1.0 1.00 0.99987
b0 0.5 0.75 0.79624
a1 1.0 1.00 1.00000
b1 0.0 0.25 0.30102
Table 4.1 — PE Coefficients
The first set of coefficients are attributed to Tappert and are the first two terms of a
Taylor series expansion of Q given by
K+−+=+82
112qq
q
The requirement for convergence of this series is that 1<q . Substitution of these
coefficients into the general form given by equation (4.23) gives
( ) 012 2202
2
0 =−++ unkz
u
r
ujk
∂∂
∂∂
(4.24)
![Page 148: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/148.jpg)
123
Which is known as the Standard (or narrow-angle—due to the convergence
requirement) Parabolic Equation (SPE).
Greene derived the coefficients used for his implementation by optimisation of the
rational-linear approximation for defined angle intervals [54]. The coefficients were
chosen to minimise the maximum error of the square-root approximation.
The coefficients given by Claerbout [5] are those of the first term of a Padé series
expansion of Q in the form of
( )∑=
+++
+=m
i
m
mi
mi qOb
qaQ
1
12
,
,
11 (4.25)
All of these forms provide approximate solutions to the full-wave equation and as
such are limited in their usefulness for modelling GPS signal propagation. These
limitations are examined in the following sections.
4.3.1 Fourier Split-Step Solution of the SPE
By making use of operator formalism a solution scheme for the standard parabolic
equation is developed [55].
Equation (4.24) can rewritten, with the range variable r replaced by x (highlighting the
fact that it is a two-dimensional problem), in the following compact form
[ ]uBAjx
u +=∂∂
(4.26)
where
![Page 149: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/149.jpg)
124
( )12
2 −= nk
A and 2
2
2
1
zkB
∂∂=
The solution of equation (4.26) at a range step, x + ∆x, is
( ) ( ) ( )zxuezxxu xBAj ,, ∆+=∆+ (4.27)
A split-form of the exponential operator, ( ) xBAje ∆+
, is now introduced.
( ) xjBxjAxBAj eee ∆∆∆+ ≅ (4.28)
giving
( ) ( )zxueezxxu xjBxjA ,, ∆∆=∆+ (4.29)
The split form of this equation is exact only when the operators A and B commute.
This is the case if the refractive index, n, is assumed constant. The form of this
splitting now allows a relatively straight forward numerical solution. Since A is a
multiplication operator, a Fourier transform solution of the B-term is sought.
Let
( ) ( )zxxgezxxu xjA ,, ∆+=∆+ ∆ (4.30)
where g is given by
( ) ( )zxuezxxg zkxj
,,2
2
2
1
∂∂
∆
=∆+ (4.31)
![Page 150: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/150.jpg)
125
Making use of the Fourier transform property
( ) ( ) ( )pXjpzxdz
d nF
n
n
⇔ (4.32)
the Fourier transform of equation (4.31) can be written directly as
( ) ( )pxUepxxGp
k
xj
,,2
2
∆−
=∆+ (4.33)
This allows the solution to g to written in terms of two Fourier transforms
( ) ( )[ ]g x x z F e F u x zj x
kp
+ =
−−
∆∆
, ,1 22
(4.34)
The complete solution of the reduced function, u, also includes the refractive effects
of the medium—through the refractive index term, n—and is given by
( ) [ ] ( )[ ]
=∆+
∆−−∆−
zxuFeFezxxup
k
xjxn
jk
,,22
211
2 (4.35)
It is now evident why Tappert coined the terms Fourier Split-Step for this algorithm,
since the solution involves Fourier techniques and accounts for diffraction and
refraction in two distinct (split) steps.
The solution of g given by equation (4.34) is the solution of the SPE for propagation
in free-space (n = 1) and only accounts for diffractive effects. Note that the PE
Propagator, for this solution, varies from that given for the FSPE solution, equation
(4.10). This variation of PE-propagator is crucial and it will become evident in the
next section how this effects the field solution.
![Page 151: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/151.jpg)
126
4.3.2 Phase Errors in Rational-Linear Approximation Forms of the PE
Now equations (3.18 and 3.28) show that the exact form (Helmholtz) of the square-
root operator Q, and hence the phase of the solution [3], is given by
θ2sin1−=Q (4.36)
Which is valid for a plane wave propagating at angle θ in free space.
Likewise, substituting for q, equation (4.22), into the general rational-linear form
given by equation (4.23) allows the phases for each case to be written directly [32] as
QTappert = −1 05 2. sin θ (4.37)
QClaerbout Pade,
. sin
. sin1
2
2
1 0 75
1 0 25=
−−
θθ
(4.38)
QGreene =−
−0 99987 0 79624
1 0 30102
2
2
. . sin
. sin
θθ
(4.39)
A plot of these various Q-approximates over the angle interval {-90 degrees, +90
degrees} gives an insight into the asymptotic behaviour of each approximation. This
is shown in Figure 4.1.
![Page 152: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/152.jpg)
127
-100 -80 -60 -40 -20 0 20 40 60 80 1000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Q Approximations
Angle (deg)
Q F
unct
ion
Exact Tappert Greene Claerbout
Figure 4.1 — Q-functions
As can be seen the approximations for Q are only valid for a narrow range of angles,
centred about zero degrees (horizontal propagation). The divergence of the
approximations from the exact Q function results in phase errors within the stepped
solution of each PE approximate form.
These associated phase errors can be defined as,
aa QQ −=ξ (4.40)
where Qa represents each of the approximations given above, and Q the exact function
given by equation (4.36). These phase errors, plotted as a function of positive
propagation angle, see Figure 4.2, give a numerical interpretation of the angular
limitations associated with each of these approximations.
![Page 153: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/153.jpg)
128
Tappert
..............
Greene
Claerbout
Figure 4.2 — Phase error with common approximates
For an arbitrary phase error limit of 0.002 the Tappert, Claerbout (Padé1), and Greene
approximations have acceptable propagation angle limits of approximately 20°, 36°,
and 47° respectively.
From this it is seen that the requirements for modelling of GPS signal propagation (up
to 90 degrees) is not easily met by these non-free-space PE forms. This provides
justification for using the FSPE, which has no phase error limitations, related to
propagation angles (achieved by not including refractive atmospheric effects). In
addition the Fourier-step solution, given by equation (4.10), is highly efficient, indeed
optimal, when there are no refractive index variations and terrain can be modelled
with non-complex boundary conditions. There are alternate implementations, such as
a split-step Padé method [29], that allows wider-angle limits, and is highly efficient
for more complex propagation problems.
4.4 Numerical Implementation for GPS Satellite Propagation
Having established that the FSPE solution, given by equation (4.10), will form the
basis for a GPS propagation model, we can now consider the implementation
requirements of the model. The Fourier step solution is an open boundary problem
![Page 154: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/154.jpg)
129
that is solved by modifying the previous solution and marching it to the next array
position, using equation (4.10).
4.4.1 Domain Sampling
The representative solution domain for the FSPE model is shown in Figure 4.3.
∆x
∆z
0,0
zmax
range
Figure 4.3 — FSPE Solution Domain
The first step in the implementation is to select the sampling rates for the spatial
domain parameters x and z. For the vertical spatial sampling the sampling rate must
satisfy the Nyquist sampling criterion [56] where an analogy between the
frequency⇔time domains of Fourier analysis in signal processing, and the spatial-
frequency⇔spatial-distance domains for this modelling, is made. Therefore the
sampling must satisfy,
max2
2
pz
π≤∆ (4.41)
![Page 155: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/155.jpg)
130
where for a selected maximum propagation angle
maxmax sinθkp = (4.42)
In the corresponding, angular spectrum domain (p-domain) the sample spacing must
satisfy
zNp
∆=∆ π2
(4.43)
where N is the total sample number in the Fast Fourier Transform (FFT), and should
be a power of two for efficiency. If the maximum propagation angle is taken to be 90
degrees (zenith) then ∆z is simply λ/2.
The selection of the spatial sampling in range (∆x) is made such that the solution of
the FSPE is within acceptable error bounds. The error can be made as small as desired
by making ∆x sufficiently small. Jensen et al [57] suggest that the only safe way to
ensure numerically accurate PE results is through a convergence test , where ∆x and
∆z are systematically reduced until a stable answer is obtained.
4.4.2 Incident Boundary Condition
The left-hand boundary represents the incident, or initial, field condition and as such
must represent the field values entering the domain from the selected GPS satellite.
The most effective way to implement the incident boundary condition is to use an
initial plane-wave field representation at the first array position (x = 0)—following
the implementation given by Walker [53]—this is satisfied by
( ) ( )θθ sin0
sin0
zjkzjkinc eEeEE −= − (4.44)
![Page 156: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/156.jpg)
131
This plane-wave incident field represents a direct signal from the satellite in addition
to a ground reflected signal modelled as occurring from terrain in front of the initial
boundary. It is an arbitrary choice to include the reflected field components in the
initial boundary condition, and if they are considered of no consequence in the
subsequent analysis they may be omitted. The nature of this initial field is represented
in Figure 4.4
GPS Satellitepropagation
path
θ
Direct SignalConstant Phase Fronts
Reflected SignalConstant Phase Fronts
Z
X
X=0
- Initial Field Value
Figure 4.4 — Initial Field
There are other incident field conditions that may be utilised, Levy [58] suggested that
the field can be represented as incoming energy, thus allowing the vertical domain
size to be reduced and improving computation times. In this work the plane-wave
field condition is adequate (in a computational sense) for the modelling of GPS signal
propagation.
4.4.3 Upper Boundary Condition
The upper boundary is required to absorb all of the energy incident upon it. If the field
were simply set to zero the upper boundary would act as an ideal conductor and all the
![Page 157: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/157.jpg)
132
incident energy would be reflected back into the domain and corrupt the results. The
solution proposed by Craig [14], and implemented by Walker [53] for modelling GPS
propagation, was to set a domain height, 2Zmax, as being twice the maximum height of
interest and to apply a Hamming or Hanning window in this extended region,
effectively attenuating the signal with filtering techniques.
Figure 4.5 shows the implementation of the upper boundary condition for the FSPE.
Here the incident plane-wave initial field is attenuated by a Hanning window in the
upper region (Zmax to 2Zmax) of the implementation-domain. We define the solution-
domain as the positive first half of the implementation domain.
0 500 1000 1500 2000 2500-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1Initial Field - 10 deg
Fie
ld L
evel
Z-Samples
Hanning WindowInitial Field
Figure 4.5 — Upper Absorption Region
4.4.4 Lower Boundary Condition
For the lower boundary the effective reflection coefficient is implemented through the
use of FFT’s in the angular spectrum domain. The incident field is used to form the
source image where the 180 degree phase shift is then implemented. The magnitude
of the reflection coefficient is calculated for the complete propagation angular
spectrum (0 degrees - 90 degrees). This is then applied directly, to the p-space
(angular spectrum space) image representation of the field, to form the product. The
![Page 158: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/158.jpg)
133
resultant modified image field is then propagated with the source to form the FSPE
solution at the next range step. The p-space application of the effective reflection
coefficient is illustrated in Figure 4.6 for a plane wave at 10 degrees propagation
angle incident upon a fresh water boundary.
0 10 20 30 40 50 60 70 80 900
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1N
orm
alis
ed V
alue
Propagation Angle (degrees)
Reflection Coefficient in P-Space
Angular Spectrum (0-90)Coefficient Magnitude
Figure 4.6 — Reflection Coefficient in P-Space
For this case the resultant angular spectrum completely includes the coupled reflection
coefficient derived in Chapter 2 for GPS signal propagation. As was shown in Chapter
2, the effective reflection coefficient can also be modified to include any rough
surface effects.
It should be noted that the image method proposed here is an approximation. A more
correct method would involve a mixed Fourier transform as proposed by Kuttler and
Dockery [59, 60].
![Page 159: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/159.jpg)
134
4.4.5 Implementation Algorithm
The complete implementation domain is shown below in Figure 4.7.
FieldSource Solution Domain
UpperAbsorption Region
2Zmax
LowerAbsorption Region
ImageSourceincludingEffective
Reflection Coefficient
-2Zmax
Zmax
-Zmax
0 X
Z
HanningWindow
HanningWindow
Image Domain
Figure 4.7 — Implementation Domain
The solution-domain specifies the region of interest in the modelling being
undertaken. The implementation methodology is outlined in the algorithm flow-chart
shown below in Figure 4.8.
![Page 160: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/160.jpg)
135
FFTReflectionCoefficient
InverseFFT
SurfaceRoughness
ModifiedImage Field
Total Field
Image Field
Source Field
Combine Fields
FFTFSPE
PropagatorInverse
FFT
PropagationAngle
Initial Field
ApplyHanning Window
Figure 4.8 — Implementation Algorithm
At this point the FSPE modelling is in its basis form and does not include the effect of
terrain interactions. The following section introduces the implementation of terrain
interaction into the basic FSPE model.
4.5 Implementing Arbitrary Terrain in the PE Model
The Fourier split-step method has been extensively used for modelling tropospheric
propagation over smooth earth and sea profiles. There is now however increased
interest in modelling propagation over arbitrary terrain profiles. The modelling of
![Page 161: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/161.jpg)
136
propagation over arbitrary terrain is a mandatory requirement for the GPS propagation
model being developed as part of this research.
The following section introduces the Boundary-Shift (BS) technique for inclusion of
terrain interaction.
4.5.1 Boundary-Shift Technique for Arbitrary Terrain
The boundary-shift technique, for handling arbitrary terrain within the PE code,
involves the shifting of the field array (aperture) either up or down to account for the
shift in the boundary position, and thus satisfy the terrain boundary conditions. The
field aperture immediately to the left of any obstructing terrain is stored then shifted
down according to the height of the terrain element. The lower elements — those that
would propagate into the terrain — are discarded and zeros inserted at the top of the
array to maintain the correct number of elements. This modified field array is then
propagated to the next array, with the FSPE Fourier-step technique. At negative
terrain transitions, the reverse procedure is applied. The array is shifted up by the
corresponding height, with the top elements discarded, and zeros inserted at the
element positions where the field is obscured by the terrain. The result of the
boundary shifting technique is simply a restructuring of the domain representation to
that of a field propagating over a plane earth while accounting for diffractive effects
over terrain.
The development of the boundary-shift technique is based on intuitive concepts and
approximations, more so than any sound mathematical or physical formulations [61].
The method has however shown excellent agreement with other more mathematically
correct methods.
The implementation is straight forward and allows use of the Fourier step routine, for
propagation over a smooth earth, to be used (without modification) to propagate the
field, with aperture shifts made at the appropriate range steps. The technique is
presented graphically in Figure 4.9 and Figure 4.10. Figure 4.9 is the representation of
![Page 162: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/162.jpg)
137
the problem within the solution-domain space. The implementation of the boundary
shift method is shown in Figure 4.10.
The boundary shift method, for this simple example, is explained as follows:
The field aperture immediately to the left of the first terrain block (array 1) is stored
then shifted down four elements. The four lower elements—those that would
propagate into the terrain—are discarded and zeros inserted at the top of the array to
maintain the correct number of elements. This modified field array is then propagated
to the second array, with the Fourier split-step. At the negative terrain transition, the
reverse technique is applied. The array is shifted up, with the four top elements
discarded, and zeros inserted at the element positions where the field is obscured by
the terrain. Similarly the technique is applied for all terrain transitions. The result of
the boundary shifting technique is seen, in Figure 4.10, as a restructuring of the
domain representation to that of a field propagating over a plane earth.
∆x
∆z
0,0Range
Height
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
N∆z
Figure 4.9 Figure 4.10
![Page 163: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/163.jpg)
138
The inclusion of the boundary-shift technique into the basic FSPE model involves
modification of the source field element as given in Figure 4.8. This modification is
implemented as follows.
Field Array
TerrainTransition
ShiftField Array
Up
ShiftField Array
DownField Array
Source Field
-ve +ve
No
Figure 4.11 — Boundary-Shift Algorithm
4.6 Implementing Backscatter for a Two-Way PE Model
In the development of the FSPE, it was necessary to assume that the field was
outgoing only. For the one-way FSPE model the incident boundary condition is set to
represent the incoming plane-wave propagation from a representative GPS satellite.
The FSPE method propagates the field components in the +x direction with the
Fourier-step technique. The terrain profile is accounted for by the boundary shift
method (aperture shift) as outlined in the previous section. The one-way restriction
![Page 164: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/164.jpg)
139
can be lifted by using a store and forward method of back-propagating field
components.
In this implementation a method is proposed whereby the field components that will
propagate into terrain are identified, stored and utilised as initial boundary conditions.
The PE is then propagated, with these initial boundary conditions, in the reverse
direction with the terrain mirrored and accounted for with the same boundary-shift
techniques, as used in the one-way implementation. This method has been utilised by
Levy [41] for electromagnetic propagation over terrain, and by Collins [38, 39] for the
analogous problem in underwater acoustics.
The steps for implementation of a two-way PE model derived from a one-way model
are as follows;
1. The field is propagated with the one-way PE model in the forward (+x) direction
2. The field components that will propagate into terrain (potential back-scatterers),
are identified.
3. These field values and indexes to their positions within the domain are stored for
later use.
4. The terrain profile and domain are mirrored vertically such that the one-way
implementation can again be used without modifying the existing PE model code.
5. The one-way PE is then used to propagate the stored field values—with
appropriate application of reflection coefficients and surface roughness—which
are added into the model as initial field conditions of the back-propagation.
6. The field components of the forward and back implementations are then added to
provide the resultant full field.
The backscatter implementation in the solution-domain is shown graphically in Figure
4.12.
![Page 165: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/165.jpg)
140
a b
Figure 4.12 — Backscatter implementation
The forward field condition is shown as (a) with the mirrored reverse case indicated
as (b).
The use of this technique is justified by image theory, where the components at a
vertical interface would travel to an image of the domain mirrored vertically about the
vertical reflector. In addition, the method is complementary to the boundary-shift
technique, where the down-shifted components normally discarded, are stored for use
as the initial field values for a two-way PE implementation. It should be noted that
reflection coefficients can be directly applied to the backscattered field components.
The implementation algorithm for the two-way FSPE model is given in Figure 4.13.
+x -x
![Page 166: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/166.jpg)
141
MoreArrays
Insert Array
Reverse FSPE
StoredReverse
Field Arrays
Total ReverseField
StoredForward
Field
Sum Fields
Total Field
YES
No
Figure 4.13 — Backscatter Implementation Algorithm
4.7 Summary
In this chapter the Free-Space Parabolic Equation propagation model for GPS was
introduced. The two-dimensional FSPE model accounts for reflection coefficients,
terrain interaction, diffractive effects and backscatter. Its efficiency is in the nature of
the solution using Fourier stepping techniques. The solution-domain is unbounded in
range and as such the total field can be calculated at any desired range position.
In addition the FSPE model, being a full field solution, provides information on the
field at any discrete spatial position within the solution domain. Although this is a
positive aspect in a model for analysis of GPS multipath effects there are drawbacks.
The most apparent limitation for multipath analysis is that the FSPE, as introduced in
this chapter, provides only field information, and as such does not describe the GPS
![Page 167: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/167.jpg)
142
multipath environment. It does however, provide the ability to visualise the field at
any spatial position within the solution domain.
The key multipath parameters of relative amplitude, time delay, phase and phase rate
are lost with the assumed time dependence removed. This time dependence can be
reinstated and the issue of deriving the Multipath Channel Impulse Response (MCIR
from the FSPE is covered in the next chapter.
![Page 168: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/168.jpg)
143
4.8 References
[1] V. A. Fock, Electromagnetic Diffraction and Propagation Problems. Oxford:
Pergamon, 1965.
[2] M. A. Leontovich and V. A. Fock, "Solution of the Problem of Propagation of
Electromagnetic Waves along the Earth’s Surface by the Method of Parabolic
Equations," Journal of Physics of the USSR, vol. 10, pp. 13-24, 1946.
[3] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, Computational
Ocean Acoustics. New York: AIP Press, 1994.
[4] R. H. Hardin and F. D. Tappert, "Applications of the Split-Step Fourier
Method to the Numerical Solution of Nonlinear and Variable Coefficient
Wave Equations," SIAM Review, vol. 15, pp. 423, 1973.
[5] J. F. Claerbout, Fundamentals of Geophysical Data Processing. New York:
McGraw Hill, 1976.
[6] F. R. DiNapoli and R. L. Deavenport, "Numerical Methods of Underwater
Acoustic Propagation," in Ocean Acoustics, J. A. DeSanto, Ed. Berlin:
Springer-Verlag, 1979.
[7] H. W. Ko, J. W. Sari, M. E. Thomas, P. J. Herchenroeder, and P. J. Martone,
"Anomalous Propagation and Radar Coverage Through Inhomogeneous
Atmospheres," presented at AGARD CP-346, 1984.
[8] K. H. Craig, "Propagation Modelling in the Troposphere: Parabolic Equation
Method," Electronics Letters, vol. 24, pp. 1136-1139, 1988.
[9] G. D. Dockery, "Modeling Electromagnetic Wave Propagation in the
Troposphere using the Parabolic Equation," IEEE Transactions on Antennas
and Propagation, vol. 36, 1988.
[10] K. H. Craig and M. F. Levy, "Field Strength Forecasting with the Parabolic
Equation: Wideband Applications," presented at Sixth International
Conference on Antennas and Propagation ICAP 89, Coventry, UK, 1989.
[11] K. H. Craig and M. F. Levy, "Recent Developments in Propagation
Forecasting: Channel Characterisation for Radar Systems," presented at IEE
Colloquium on ’Radar Clutter and Multipath Propagation’ (Digest No.62),
London, UK, 1989.
![Page 169: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/169.jpg)
144
[12] K. H. Craig and M. F. Levy, "A Forecasting System using the Parabolic
Equation- Application to Surface-to-Air Propagation in the Presence of
Elevated Layers," presented at Operational Decision Aids for Exploiting or
Mitigating Electromagnetic Propagation Effects (AGARD-CP-453), San
Diego, CA, USA, 1989.
[13] K. H. Craig and M. F. Levy, "A PC-based Microwave Propagation Forecasting
Model," presented at Seventh International Conference on Antennas and
Propagation ICAP 91, York, UK, 1991.
[14] K. H. Craig and M. F. Levy, "Parabolic Equation Modelling of the Effects of
Multipath and Ducting on Radar Systems," presented at IEE Proceedings-F
[Radar and Signal Processing], 1991.
[15] M. F. Levy and K. H. Craig, "Case Studies of Transhorizon Propagation:
Reliability of Predictions using Radiosonde Data," presented at Sixth
International Conference on Antennas and Propagation ICAP 89, Coventry,
UK, 1989.
[16] M. F. Levy and K. H. Craig, "Assessment of Anomalous Propagation
Predictions using Minisonde Refractivity Data and the Parabolic Equation
Method," presented at Operational Decision Aids for Exploiting or Mitigating
Electromagnetic Propagation Effects (AGARD-CP-453), San Diego, CA,
USA, 1989.
[17] M. F. Levy and K. H. Craig, "Millimeter-Wave Propagation in the
Evaporation Duct," presented at Atmospheric Propagation in the UV, Visible,
IR and MM-Wave Region and Related Systems Aspects (AGARD-CP-454),
Copenhagen, Denmark, 1990.
[18] M. F. Levy, "PE Modelling of Radiowave Propagation over the Sea,"
presented at IEE Colloquium on ’The Interaction of Radiowaves with the Sea
Surface’ (Digest No.037). London, UK, 1990.
[19] G. D. Dockery, "Method for Modelling Sea Surface Clutter in Complicated
Propagation Environments," IEE Proceedings-F [Radar and Signal
Processing], vol. 137, pp. 73-79, 1990.
[20] J. P. Reilly and G. D. Dockery, "Influence of Evaporative Ducts on Radar Sea
Return," IEE Proceedings-F, vol. 137, pp. 80-88, 1990.
![Page 170: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/170.jpg)
145
[21] P. L. Slingsby, "Modelling Tropospheric Ducting Effects on VHF/UHF
Propagation," IEEE Transactions on Broadcasting, vol. 37, pp. 25-34, 1991.
[22] A. E. Barrios, "Parabolic Equation Modeling in Horizontally Inhomogeneous
Environments," IEEE Transactions on Antennas and Propagation, vol. 40, pp.
791-797, 1992.
[23] D. Rouseff, "Simulated Microwave Propagation Through Tropospheric
Turbulance," IEEE Transactions on Antennas and Propagation, vol. 40, pp.
1076-1083, 1992.
[24] H. V. Hitney, "Hybrid Ray Optic and Parabolic Equation Methods for Radar
Propagation Modeling," presented at International Conference Radar 92,
Brighton, UK, 1992.
[25] W. P. M. N. Keizer and R. B. Boekema, "Within the Horizon Propagation
Measurements over sea at 10.5 GHz," presented at IEEE Antennas and
Propagation Society International Symposium, Chicago, IL, USA, 1992.
[26] M. D. Collins, "Applications and Time-Domain Solution of Higher-Order
Parabolic Equations in Underwater Acoustics," Journal of the Acoustic Society
of America, vol. 86, pp. 1097-1102, 1989.
[27] M. D. Collins, "Benchmark Calculations for Higher-Order Parabolic
Equations," Journal of the Acoustic Society of America, vol. 87, pp. 1535-
1538, 1990.
[28] M. D. Collins and E. K. Westwood, "A Higher-Order Energy-Conserving
Parabolic Equation for Range-Dependent Ocean Depth, Sound Speed, and
Density," Journal of the Acoustic Society of America, vol. 89, pp. 1068-1075,
1991.
[29] M. D. Collins, "A Split-Step Padé Solution for the Parabolic Equation
Method," Journal of the Acoustic Society of America, vol. 93, pp. 1736-1742,
1993.
[30] R. A. Dalrymple, L. C. Munasinghe, D. H. Wood, and J. T. Kirby, "A Very-
Wide-Angle Acoustic Model for Underwater Sound Propagation," Journal of
the Acoustic Society of America, vol. 88, pp. 1863-1876, 1990.
[31] N. Dodd, "Efficient Higher-Order Finite-Difference Schemes for Parabolic
Models," Coastal Engineering, vol. 28, pp. 57-92, 1996.
![Page 171: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/171.jpg)
146
[32] R. J. Hill, "Wider-Angle Parabolic Wave Equation," Journal of the Acoustic
Society of America, vol. 79, pp. 1406-1409, 1986.
[33] G. H. Knightly, D. Lee, and D. F. St. Mary, "A Higher-Order Parabolic Wave
Equation," Journal of the Acoustic Society of America, vol. 82, pp. 580-587,
1987.
[34] D. J. Thomson and N. R. Chapman, "A Wide-Angle Split-Step Algorithm for
the Parabolic Equation," Journal of the Acoustic Society of America, vol. 74,
pp. 1848-1854, 1983.
[35] D. J. Thomson, "Wide-Angle Parabolic Equation Solutions to Two Range-
Dependent Benchmark Problems," Journal of the Acoustic Society of America,
vol. 87, pp. 1514-1520, 1990.
[36] D. Yevick and D. J. Thomson, "Split-Step/Finite-Difference and Split-
Step/Lanczos Algorithms for Solving Alternative Higher-Order Parabolic
Equations," Journal of the Acoustic Society of America, vol. 96, pp. 396-405,
1997.
[37] P. P. Borsboom and M. F. Levy, "Scattering with Parabolic Equation Methods:
Application to RCS Computation," presented at IEE Colloquium on Common
Modelling Techniques for Electromagnetic Wave and Acoustic Wave
Propagation, London, UK, 1996.
[38] M. D. Collins and R. B. Evans, "A Two-way Parabolic Equation for Acoustic
Backscattering in the Ocean," Journal of the Acoustic Society of America, vol.
91, pp. 1357-1368, 1992.
[39] M. D. Collins, "A Two-Way Parabolic Equation for Elastic Media," Journal of
the Acoustic Society of America, vol. 93, pp. 1815-1825, 1993.
[40] M. F. Levy, "Parabolic Equation Modelling of Backscatter from the Rough
Sea Surface," presented at Target and Clutter Scattering and their Effects on
Military Radar Performance (AGARD-CP-501), Ottawa, Ont., Canada, 1991.
[41] M. F. Levy and P. P. Borsboom, "Radar Cross-section Computations using the
Parabolic Equation Method," Electronics Letters, vol. 32, pp. 1234-1236,
1996.
[42] A. E. Barrios, "Terrain Modelling using the Split-step Parabolic equation
Method," presented at International Conference Radar 92, Brighton, UK,
1992.
![Page 172: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/172.jpg)
147
[43] A. E. Barrios, "A Terrain Parabolic Equation Model for Propagation in the
Troposphere," IEEE Transactions on Antennas and Propagation, vol. 42, pp.
90-98, 1994.
[44] M. F. Levy, "Parabolic Equation Modelling of Propagation over Irregular
Terrain," presented at Seventh International Conference on Antennas and
Propagation ICAP 91, York, UK, 1991.
[45] M. F. Levy, "Horizontal Parabolic Equation Solution of Radiowave
Propagation Problems on Large Domains," IEEE Transactions on Antennas
and Propagation, vol. 43, pp. 137-144, 1995.
[46] S. W. Marcus, "A Parabolic Approximation Method for Propagation
Prediction in an Inhomogeneous Atmosphere over Irregular Terrain,"
presented at 17th Convention of Electrical and Electronics Engineers in Israel,
Tel Aviv, Israel, 1991.
[47] C. Mattiello, "Use of Parabolic Equation for Computing Diffraction by Terrain
Undulations," Cselt Tech Rep, vol. 21, pp. 947-963, 1993.
[48] R. J. McArthur, "Propagation Modelling over Irregular Terrain using the Split-
step Parabolic Equation Method," presented at International Conference Radar
92, Brighton, UK, 1992.
[49] D. J. Donohue and J. R. Kuttler, "Modeling Radar Performance over Terrain,"
John Hopkins APL Technical Digest, vol. 18, pp. 279-287, 1997.
[50] C. C. Lin and J. P. Reilly, "A Site-Specific Model of Radar Terrain
Backscatter and Shadowing," John Hopkins APL Technical Digest, vol. 18,
pp. 432-447, 1997.
[51] M. F. Levy, Parabolic Equation Methods for Electromagnetic Wave
Propagation, 1 ed. London: IEE, 2000.
[52] R. A. Walker, "Numerical Modelling of GPS Signal Propagation:
Development of a Numerical Electromagnetic Wave Propagation Model for
the Modelling of GPS Positioning," Space Centre for Satellite Navigation,
SIDC Technical Report 002-95, 1996.
[53] R. A. Walker, "Operation and Modelling of GPS Sensors in Harsh
Environments," in School of Electrical and Electronic Systems Engineering:
Queensland University of Technology, 1999.
![Page 173: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/173.jpg)
148
[54] R. R. Greene, "The Rational Approximation to the Acoustic Wave with
Bottom Interaction," Journal of the Acoustic Society of America, vol. 76, pp.
1764-1773, 1984.
[55] M. Cayer, B. Philibert, M. Lecours, and D. Dion, "Analysis of the Fourier
Split-Step Method for Resolution of Radio Propagation Over the Sea,"
presented at 1994 Canadian Conference on Electrical and Computer
Engineering, Halifax, Canada, 1994.
[56] A. V. Oppenheim and R. W. Schafer, Discrete-Time Signal Processing.
Englewood Cliffs: Prentice Hall, 1989.
[57] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, "Parabolic
Equations," in Computational Ocean Acoustics, R. T. Beyer, Ed. New York:
AIP Press, 1994.
[58] M. F. Levy, "Transparent Boundary Conditions for Parabolic Equation
Solutions of Radiowave Propagation Problems," IEEE Transactions on
Antennas and Propagation, vol. 45, pp. 66-72, 1997.
[59] J. R. Kuttler and G. D. Dockery, "Theoretical Description of the Parabolic
Approximation/Fourier Split-Step Method of Representing Electromagnetic
Propagation in the Troposphere," Radio Science, vol. 26, pp. 381-393, 1991.
[60] G. D. Dockery and J. R. Kuttler, "An Improved Impedance-Boundary
Algorithm for Fourier Split-Step Solutions of the Parabolic Wave Equation,"
IEEE Transactions on Antennas and Propagation, vol. 44, pp. 1592-1599,
1996.
[61] A. E. Barrios, "Terrain and Refractivity Effects on Non-Optical Paths,"
presented at AGARD Electromagnetic Wave Propagation Panel Symposium,
Rotterdam, The Netherlands, 1993.
![Page 174: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/174.jpg)
149
Chapter 5 Time Series Analysis with the FSPE
In the previous chapter the FSPE method for modelling GPS signal propagation was
introduced. The major limitation of the full-field FSPE solution is the lack of defining
multipath information. The FSPE provides only the solved field value within the
solution domain. The defining parameters of relative delay, amplitude, and phase are
fully incorporated within the solution. This defining multipath information can,
however, be extracted from the FSPE field results. This chapter introduces the PE-
based Time Analysis (PETA) model, for extraction of the multipath parameters from
the FSPE full-field solution.
5.1 Introduction
The determination of relative propagation path delays is important in the
understanding of GPS multipath errors. The superposition of delayed replicas of the
direct ranging signal leads to distortion of the signal at the GPS receiver antenna [1].
The receiver requires an undistorted signal to provide an accurate estimate of the
pseudorange to the satellite. In trying to understand the impact these multipath signals
have on the receiver it is necessary to characterise the multipath signal. Two of the
most important multipath parameters are; the relative time delay between the direct
and multipath signals, and the relative amplitude of the delayed signals[2, 3].
Together these parameters form the Multipath Channel Impulse Response (MCIR)
[4].
Determining the MCIR is an example of a pulse propagation problem which can be
solved, via the frequency domain, by the use of Fourier synthesis of the modelled
FSPE results [5]. This method is attractive since the FSPE propagation model that has
been developed for analysis of GPS signals lends itself readily to processing by this
technique. The technique involves Fourier synthesis based on a number of FSPE
calculations over a band of frequencies and the solution of the time-dependent field
![Page 175: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/175.jpg)
150
equation can be obtained by the Fourier transformation of the PE field solution,
namely:
( ) ( ) ( ) dfefzxufStzxu ftj π2,,,, ∫∞
∞−
= (5.1)
where S(f) is the spectrum of a source pulse and u(x,z,f) is the spatial transfer function
derived from the FSPE modelling process. This integral is evaluated using Fast
Fourier Transform (FFT) techniques at the spatial point of interest in the model
solution domain, that is, the antenna location. For this work we have chosen as our
source, a sinc pulse—of a duration to be determined by modelling requirements—
modulated at the GPS L1 frequency. The MCIR is the output of the PE-based Time
Analysis (PETA), and is given as a time series of time delayed, and attenuated source
pulses. The factors for determining the computational load are made clear as the
technique is further explained.
5.2 Implementation
Consider a time-domain sinc pulse as depicted in Figure 5.1.
Figure 5.1 — SINC pulse
![Page 176: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/176.jpg)
151
This sinc pulse is defined as:
( ) ( )Wt
WtWth
ππ
2
2sin2= (5.2)
here the function ( )usinc is defined as ( )u
u
ππsin
.
Now n/2W (-∞<n<∞) represents the zero crossings of the pulse. The Fourier transform
[6] of this pulse is given by
( ) ( ){ }
=
W
fWtWFfH
2rect=2sinc2 (5.3)
here rect is defined as rectangular function of frequency such that:
( )elsewhere0
2
11rect
=
<= xx
and W is defined as the half bandwidth of the signal.
The selection of a sinc pulse as the input signal guarantees that the spectrum of input
frequencies require no modification to their respective amplitudes, thus simplifying
the implementation. Since GPS signal propagation is to be modelled, the function is
shifted such that it is centred at the GPS L1 frequency—1.575.42 GHz (f1). This
frequency shift is achieved by convolving the rectangular function with a delta
function at f1. The inverse Fourier transform of the shifted rectangular function is the
sinc pulse modulated by 12 fje π . The frequency spectrum is now given by
( ) ( ){ }
−=
W
ffeWtWFfS fj
2rect=2sinc2 12 1π (5.4)
![Page 177: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/177.jpg)
152
Figure 5.2 shows the complete source frequency spectrum.
f1 f1+Wf1-W
|S(f)|
0f
Figure 5.2 — Input spectrum
Assuming now, that the time response is required at some spatial point p(x, z), in a
two-dimensional, frequency-domain, electromagnetic propagation model, and that the
time response will be modelled within a time window of length T; the time and
frequency axes can be discretised as
( )11,0,min −=∆+= Nktkttk K
and
( )( ) ( )12/1,0,1112/, −−−−−=∆= NNlflf l KK
The first zero crossing points of the sinc pulse are used to define its pulse width. The
relationship between the time domain pulse width (τ) and the bandwidth (2W) of the
rectangular frequency function is given by
W =1
τ(5.5)
Since the half-bandwidth cannot exceed the central frequency, the time-domain pulse
width satisfies:
![Page 178: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/178.jpg)
153
nse
63.0957542.1
1 =≥τ
Therefore for the modelling of GPS L1 pulse propagation a pulse cannot be smaller
than 0.63 nanoseconds in width at the first zero crossing points. In this work one
nanosecond is selected as an appropriate pulse width in which to resolve separately
propagated pulses.
Now consider the spacing of samples in the frequency domain:
∆ fT
=1
(5.6)
With the frequency spacing now known, the total number of frequencies ( M )
required, the sample number (L) of the mid-spectrum frequency (f1), and the required
value of each discrete frequency can be found. The total number of frequencies
required is given by,
∆
=f
WM
2round (5.7)
and the sample number of the mid-spectrum frequency,
∆
=f
fL 1round (5.8)
The minimum frequency can now be defined by,
fM
Lf ∆
−=
2min (5.9)
![Page 179: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/179.jpg)
154
giving the required discrete frequency spectrum,
Mkfkff k K1,0,min =∆+= (5.10)
To satisfy the Nyquist sampling criterion the sampling frequency in the time domain
must be at least twice the maximum frequency of the source spectrum.
max2
1
ft =∆ (5.11)
where
fM
Lf ∆
+=
2max (5.12)
The time window length is given by
tNttT ∆=−= minmax (5.13)
In order to use efficient FFT algorithms it is necessary that N is a power of 2. For a
specified window length T the minimum number of time samples, N, is found.
t
TN
∆=min (5.14)
The next highest power of two is then selected as the required sample number.
min2 2 NN x ≥= (5.15)
![Page 180: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/180.jpg)
155
Finally it is now necessary to calculate the time-domain sample spacing.
2N
Tt =∆ (5.16)
5.3 Domain Considerations
Now consider a solution domain (in range and height) that contains a portion of the
spatial plane x, z, referenced to the origin and bounded by zmax and xmax, Figure 5.3.
This domain is an arbitrary selection but is chosen to represent, as closely as possible,
a two-dimensional electromagnetic domain in which the propagation occurs. The
domain can include elements of terrain [7].
θi
ζζ’
θr
z
x
E+ E-
H+ H-
domainzmax
xmax(0,0)
Figure 5.3 — Propagation domain
In a simplified plane-wave model representation, the lower boundary, represented by
the x-axis, is considered a perfect conductor, and as such all EM energy that impinges
is reflected according to the governing laws. Two directions are specified, such that
the incident and reflected waves have directions of travel, ζ and negative ζ’,
respectively. Therefore, for the case of a Transverse Electric (TE) plane wave, the
incident and reflected fields at any point x, z (z > 0) are given by [8]:
![Page 181: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/181.jpg)
156
( )ii xzjkinc eEE θθ cossin +−
+= (5.17)
( )rr xzjkrefl eEE θθ cossin −
−= (5.18)
The condition imposed on the lower boundary (z = 0) is such that the field must be
zero for all x. Therefore the two amplitudes and angles must be equal giving the total
field anywhere in the domain as:
( ) ( )θθθθ cossin0
cossin0
xzjkxzjk eEeEE −+− −= (5.19)
Figure 5.4 illustrates the concept of the time reference spatial corrections necessary in
the implementation. Since the reference plane-wave source point in the domain is at
the origin (point A), an adjustment is required so that all of the calculated field values
are referenced, at t = 0, to the arbitrary left boundary of the domain (x = 0), for all
values of z. This is achieved by shifting the plane-wave reference from point A to
point C. The first step is to correct for the height of the antenna (A to B), then to
correct for the distance, from the arbitrary boundary, to the antenna (B to C).
![Page 182: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/182.jpg)
157
planes of constant phase
arbitrary boundary
θ
∆z
∆x
∆z sinθ
∆x(sinθtanθ)
transmission distance referenced to (0,0) at t = 0
A
B
C
antenna location
Domain
Figure 5.4 — Corrections for spatial time reference
The required correction to the total field calculations is then of the form;
( )θθθ tansinsin xjkzjkc eeuu ∆∆= (5.20)
Here the calculated field is corrected for a zero time reference relative to the spatial
point where the line-of-sight signal enters the domain.
5.4 Deriving the MCIR from the PETA Model
These equations now form the basis for time series analysis where the FSPE
propagation model is executed for each of the frequencies and the spectrum recreated
from the resultant field values at the point of interest. The inverse FFT of the
constructed spectrum will result in the complex MCIR which is given by
( )∑=
−=M
iii tMCIR
1
sinc τ (5.21)
![Page 183: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/183.jpg)
158
In estimating the relative amplitude and time delay from the MCIR a gradient ascent
technique is used to locate the peaks. The sample numbers of these peaks are then
used to find the value of amplitude and time delay, of the signal pulses, within the
selected time window. These respective parameters are given by
[ ]iki MCIRabs=α (5.22)
tkii ∆=τ (5.23)
Here ki is the sample number of the ith peak and ∆t is the time sample spacing.
The complex field in terms of the MCIR parameters, at a spatial point (x,z), is then
given by the addition of the decomposed plane waves.
( ) ( )∑=
++=M
i
ftji
ftjPETA
iieezx1
22 0, φππ αψ (5.24)
Here the first term represents the line-of-sight signal with a propagation time of t0,
from an arbitrary domain incident boundary at, x=0. The summation term represents,
the M multipath signals, where αi and ti represent respectively, the ith multipath
amplitude and time of arrival. The phase term, φi, is the resultant phase shift due to the
boundary reflection(s) for the ith multipath signal. This equation can be normalised by
assuming zero reference phase for the LOS signal. This normalisation is simply a
change from absolute time delay, as presented by the PETA, to relative time delay,
and is given by,
( ) ( )∑=
++=M
i
fjiPETA
iiezx1
21, φτπαψ (5.25)
![Page 184: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/184.jpg)
159
where τi is the time delay relative to the LOS signal [9].
An example of time delay estimation error from the PETA based on an exact plane-
wave solution, and for a canonical geometric optics forward propagation problem, is
shown in Figure 5.5.
5 10 15 20 25 30-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08Relative Delay Error
Del
ay e
rror
(ns
)
Propagation angle (degs)
Figure 5.5 — Time delay estimation error
The mean relative time delay estimation error is 0.00 ns with a standard deviation of
0.02 ns.
5.5 Multipath Phase Information from the PETA Model
The MCIR directly provides estimates of αI and τi. The phase term, φi, is not so readily
extracted as it remains embedded in the complex MCIR information. The total MCIR
phase is a combination of the transformed frequency-domain phase of each
component making up the total PE field. This composite phase is dependant on the
distance travelled (time delay) and the phase upon reflection (if reflection has
occurred) of each plane wave present in the model. The determination of the phase
term, φi, can be achieved by taking the difference between the MCIR phase results
![Page 185: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/185.jpg)
160
from the PETA and a reference phase produced by a plane-wave MCIR at the
estimated amplitude and time delay of the signal path under consideration. The
residual phase, for the given time delay, is then the phase variation resulting from
terrain interaction. This procedure can be written as
( )[ ] ( )[ ]ii kiiREFkiiiPETAi ,MCIRphase,,MCIRphase −= φφ (5.26)
Here the MCIR derived from the PETA is a combined function of amplitude, time
delay, and the phase term. The reference result differs in that it does not include the
unknown phase term and the differencing of the time-domain phases, at each pulse
peak, gives the desired phase term.
Estimation of the phase of the LOS signal, using the technique outlined, for a simple
forward propagation problem, is shown in Figure 5.6.
5 10 15 20 25 30-25
-20
-15
-10
-5
0
5
10
15LOS Phase
Pha
se (
degr
ees)
Propagation angle (degs)
Figure 5.6 — LOS Phase Estimation Error
The mean is 0.39 degrees with a standard deviation of 5.41 degrees, where the true
phase for the LOS is zero degrees. For the single specular reflection the error in phase
estimation is shown in Figure 5.7.
![Page 186: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/186.jpg)
161
5 10 15 20 25 30165
170
175
180
185
190
195
200
205Reflection Phase
Pha
se (
degr
ees)
Propagation angle (degs)
Figure 5.7 — Forward Multipath Phase Estimation Error
In this case the mean phase is 180.3 degrees with a standard deviation of 5.56 degrees.
The expected phase of the reflected signal is 180 degrees for the effective reflection
coefficient.
5.6 Antenna Gain Pattern from Angle of Arrival Information
The multipath parameters derived from the PETA model provide relative amplitude,
time delay, and phase of each multipath component. To more fully model a realistic
situation it is necessary to include the effects of the GPS antenna gain pattern [10]. No
effective way of including the antenna gain pattern within the FSPE has yet been
found, and the gain pattern is presently implemented post-modelling.
The gain pattern specifies the relative gain of the antenna in a particular orientation,
namely horizontal or vertical. The horizontal gain pattern specifies a field distribution
factor in azimuth, while the vertical gain pattern specifies a field distribution factor in
elevation. For the two-dimensional PETA model it is only the latter that needs to be
considered. The incorporation of the antenna pattern gives the field as [11]:
![Page 187: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/187.jpg)
162
( ) ( ) ( ) ( )∑=
++=M
i
ftji
ftjPETA
iiefefzx1
22 0, φππ αθθψ (5.27)
where the term f(θ) represents the receiving antenna’s vertical gain factor. To
incorporate this factor it is necessary to have an estimation of the Angle of Arrival
(AOA) for each signal component. In practice this is achieved with an array of
antennas, such as the Multiple Emitter Location and Signal Parameter Estimation
technique introduced by Schmidt [12]. However, with the PETA model the results are
time-based and it is possible to determine the AOA from the relative time
information. This is achieved by selected a field cell spatially diverse from the
antenna location, and comparing relative time shifts, of each plane-wave component,
as provided by the PETA model. For a point vertically displaced from the antenna
location the AOA is given by
( )
−
∆= ∆+− zzx
izx
iAOA z
C ,,1sin ττϕ (5.28)
where ∆z is the vertical displacement, with its sign determining relative displacement
(positive for above, negative for below). The AOA thus derived is then used to
incorporate the antenna gain characteristics relative to each signal component in post
processing.
Examples of errors in AOA estimates, using this technique, are given respectively in
Figure 5.8 and Figure 5.9. For the LOS estimate the mean error is 0 degrees with a
standard deviation of 2.91 degrees.
![Page 188: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/188.jpg)
163
5 10 15 20 25 30-8
-6
-4
-2
0
2
4
6
8LOS Angle of Arrival Error
AO
A e
rror
(de
gs)
Propagation angle (degs)
Figure 5.8 — LOS AOA Estimation Error
5 10 15 20 25 30-8
-6
-4
-2
0
2
4
6Reflection Angle of Arrival Error
AO
A e
rror
(de
gs)
Propagation angle (degs)
Figure 5.9 — Forward Multipath AOA Estimation Error
For a single forward reflection the error in AOA estimate has a mean of 0.01 degrees
with a standard deviation of 2.68 degrees.
![Page 189: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/189.jpg)
164
5.7 PETA Implementation Issues
Some observations on resolution and computational load of this Fourier-based time
analysis technique are now made. Firstly the required resolution of the time domain
solution should be considered as necessary for the analysis being made. If the
resolution is not considered adequate then other parameters, of the technique, require
adjustment to achieve the required resolution. The most obvious way to increase the
time resolution is to increase the sample number. This can be achieved by
oversampling in the time-domain.
Secondly the effect of aliasing (periodicity of the time window T), introduced by the
discretisation in frequency, needs to be considered. The actual time response (in terms
of the inverse FFT), in the selected time window {tmin, tmin + T} is given by [5],
( ) ( )[ ]( )( )∑∑
≠
−
=
−− +−
∆=0
12/
0
2
2 ,,,,Re2,,2
2min
nk
N
l
N
tj
ftjllk nTtzxueefzxuftzxu
k
l
ππαπ (5.29)
The last term, of equation (5.29), represents the aliasing from the periodic time
windows. Although the time window needs to be made as small as possible, to reduce
the computational load of the technique, aliasing can only be avoided by making
certain the time window selected will contain all of the time response information.
These conflicting requirements can be met by selected tmin from the expected
propagation time requirements of the problem being analysed. The total length of the
time window is then selected, likewise, by considering the range and geometry of the
propagation environment. An example of a non-aliased time-window is shown in
Figure 5.10.
![Page 190: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/190.jpg)
165
0 5 10 15 20 25 30 35 40 45 500
0.2
0.4
0.6
0.8
1
Absolute time delay (ns)
peak 1 delay: 037.0955 nspeak 2 delay: 045.5206 nstime delay: 8.4251 ns
Figure 5.10 — Non-aliased time response
We note the absolute propagation times are 37.0955ns for the LOS and 45.5206ns for
a single forward reflection. This is for the simulation of a spatial point 10 metres into
the modelling domain and 5 metres above the reflecting surface.
If the time-window is now halved in length aliasing will occur. This aliasing effect is
shown in Figure 5.11
0 5 10 15 20 250
0.2
0.4
0.6
0.8
1
Absolute time delay (ns)
peak 1 delay: 012.3326 nspeak 2 delay: 020.7659 nstime delay: 8.4333 ns
Figure 5.11 — Aliased time response
![Page 191: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/191.jpg)
166
Here the LOS time of arrival is 12.3326ns aliased in the 25ns window such that the
true LOS time is the addition of the modelled delay and the aliasing window length,
giving 37.3326ns. Likewise the true reflection time is 45.7659ns, being the addition of
the modelled time and aliasing window length.
5.8 PETA Domain Representation and Performance
The propagation domain is represented by a two-dimensional plane that is specified
by the azimuthal direction to the satellite, the maximum height, and the maximum
range to be modelled. The antenna can be located at any point on the plane, above the
terrain. Terrain information can be derived from Digital Terrain Models (DTMs) [13].
The model domain is depicted in Figure 5.12.
Forward propagation
Back- propagation
Antenna
Satelliteazimuth
(0,zmax)
(0,0)(xmax,zmax)
(xmax,0)
Terrain
Incidentboundary
Satellite signal
Figure 5.12 — Domain representation
The definitions of forward and back-propagation are relative to the directions
specified in Figure 5.12.
![Page 192: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/192.jpg)
167
Since the time information is essentially contained within the phase variation at each
frequency, it is necessary that the rapid phase variation be reinstated. This is achieved
by modifying the reduced PE field result as follows
( ) ( ) jkxezxuzx −= ,,ψ (5.30)
The phase must be reinstated not only in the forward direction but also for each
individual backscattering interface. This requirement can only be met if each back-
scatterer is treated as an individual FSPE propagation problem. The total field is then
simply the addition of the forward and backward fields.
The model simulation time for single frequency field values, with forward
propagation only, is given by the proportionality
AkT wayPE θ∝−1 (5.31)
where, k is the wave-number, θ is the propagation angle, and A is the area of the
domain plane. With inclusion of back-scatter this increases to
( ) wayPE
wayPE TLT −− += 12 1 (5.32)
for L back-scatterers. For the PETA the simulation time is
wayPE
pulse
winPETA TT −= 22
ττ
(5.33)
where τwin is the width of the time analysis window, and τpulse is the source pulse width.
![Page 193: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/193.jpg)
168
5.9 Summary
In this chapter the PETA was introduced as an effective method for deriving the
MCIR and all relevant multipath parameters from the FSPE. It was shown that the
method is capable of accurately estimating the relative amplitude, time delay and
phase of the multipath signals.
The method as an adjunct to the FSPE provides a comprehensive modelling technique
for the analysis of GPS multipath propagation.
![Page 194: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/194.jpg)
169
5.10 References
[1] L. R. Weill, "GPS Multipath Mitigation by Means of Correlator Reference
Waveform Design," presented at The National Technical Meeting of The
Institute of Navigation., Santa Monica, CA, 1997.
[2] R. D. J. van Nee, "Multipath and Multi-Transmitter Interference in Spread-
Spectrum Communication and Navigation Systems," in Faculty of Electrical
Engineering, Telecommuncation and Traffic Control Systems Group. Delft:
Delft University of Technology, 1995, pp. 205.
[3] M. S. Braasch, "GPS and DGPS Multipath Effects and Modeling," in ION
GPS-95 Tutorial: Navtech Seminars, 1995.
[4] H. Hashemi, "Impulse Response Modeling of Indoor Radio Propagation
Channels," IEEE Journal on Selected Areas in Communications, vol. 11, pp.
967-978, 1993.
[5] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, "Broadband
Modelling," in Computational Ocean Acoustics, R. T. Beyer, Ed. New York:
AIP Press, 1994.
[6] C. D. McGillem and G. R. Cooper, Continuous and Discrete Signal and
System Analysis, Third ed. Philadelphia: HRW Saunders, 1991.
[7] J. T. Hviid, J. B. Andersen, J. Toftgard, and J. Bojer, "Terrain-based
Propagation Model for Rural Area - An Integral Equation Approach," IEEE
Transactions on Antennas and Propagation, vol. 43, pp. 41-46, 1995.
[8] D. K. Cheng, Field and Wave Electromagnetics. Reading, MA: Addison-
Wesley, 1989.
[9] R. J. C. Bultitude, P. Melancon, H. Zaghloul, G. Morrison, and M. Prokki,
"The Dependence of Indoor Radio Channel Multipath Characteristics on
Transmit Receive Ranges," IEEE Journal on Selected Areas in
Communications, vol. 11, pp. 979-990, 1993.
[10] S. U. Hwu, B. P. Lu, R. J. Panneton, and B. A. Bourgeois, "Space Station GPS
Antennas Multipath Analysis," presented at IEEE Antennas and Propagation
Society International Symposium, Newport Beach, California, 1995.
![Page 195: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/195.jpg)
170
[11] H. Bremmer, Terrestrial Radio Waves: Theory of Propagation. London:
Elsevier Publishing Company, 1949.
[12] R. O. Schmidt, "Multiple Emitter Location and Signal Parameter Estimation,"
IEEE Transactions on Antennas and Propagation, vol. 34, pp. 276-280, 1986.
[13] R. A. Walker, "Operation and Modelling of GPS Sensors in Harsh
Environments," in School of Electrical and Electronic Systems Engineering:
Queensland University of Technology, 1999.
![Page 196: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/196.jpg)
171
Chapter 6 Model Validation
In the previous chapters, the methodology for modelling GPS multipath propagation
was introduced, namely the Free Space Parabolic Equation (FSPE) model and the PE-
based Time Analysis (PETA) model. The combination of the FSPE and PETA
provides a comprehensive tool for the analysis and visualisation of the behaviour of
multipath propagation for terrestrial applications of GPS. In this chapter an analyse of
the results of the FSPE/PETA implementation is made. The FSPE-PETA modelling
system is implemented in MATLAB.
6.1 Validation of FSPE with an Exact Solution
With the MATLAB implementation of the FSPE completed, a validation of the results
is made by comparison with results from an exact field solution. By validating against
an exact plane-wave field solution any implementation errors are easily determined [1,
2]. The theory of the PE model is well tested [3, 4] and no attempt is made within this
work to retest the underlying validity of the PE as a EM modelling tool [5]. The
testing performed here is to show that the particular implementation of the FSPE and
the PETA, for modelling and simulation of GPS multipath propagation, is valid, and
that the models are suitable for their intended purpose within reasonable bounds of
accuracy [6].
6.1.1 Phase Error
Before completing the validation of the free-space parabolic equation propagation
model, it is worthwhile considering the problem of phase error in the standard PE
forms [7]. In chapter 3 the limitation of including the refractive index term in the
standard PE was examined. To highlight this limitation on angle of propagation, the
SPE (the standard parabolic equation) was implemented and compared to an exact
solution.
![Page 197: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/197.jpg)
172
Figure 6.1 represents plots of the phase variation over range, in a narrow-angle
Tappert PE implementation [2] (dashed) and an exact plane-wave solution [8] (solid)
for propagation at 40 degrees over an ideal conductor.
0 10 20 30 40 50 60 70 80 90 100-4
-3
-2
-1
0
1
2
3
4SPE vs Exact Phase Variation over Range
range (m)
phas
e (r
ads)
exactSPE
Figure 6.1 — Phase error in Narrow-Angle SPE
The results of an error analysis shows a mean error -0.04 rads (-2.3 degs), and an error
standard deviation of 2.59 radians (148 degs). The obvious nature of the error is that it
is cumulative and increases as a direct function of range, as the SPE solution is
marched out in range. As mentioned previously the narrow-angle Tappert PE, or SPE,
is therefore unacceptable if accurate phase information is required at high propagation
angles.
The FSPE is now compared for the same propagation problem, see Figure 6.2.
![Page 198: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/198.jpg)
173
0 10 20 30 40 50 60 70 80 90 100-4
-3
-2
-1
0
1
2
3
4
range (m)
phas
e (r
ads)
FSPE vs Exact Phase Variation over Range
exactFSPE
Figure 6.2 — Phase error in FSPE
There is no error in the FSPE solution with a mean error of 0.00, error standard
deviation 0.00 using the same error analysis as for the SPE. The FSPE as given by
equation (4.8 Chapter 4) is shown to be an exact solution for the GPS propagation
problem and is not limited by angle of propagation but achieves this by neglecting
refractive index effects.
For the GPS propagation problem — local terrestrial multipath, where we are within
less than 300 metres of the receiver — refractive index effects are negligible and can
be ignored. Results of comparison of the FSPE field and an exact solution is shown in
Figure 6.3.
![Page 199: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/199.jpg)
174
5 10 15 20 25 30-70
-60
-50
-40
-30
-20
-10
0
10
Propagtion angle (degs)
Fie
ld L
evel
(dB
)
Signal Strength (x=25,z=3)
Exact FSPE (-30dB)
Figure 6.3 — FSPE Field vs Exact Solution
In this test example the signal strength has been simulated using both the FSPE and
exact plane-wave solution. The simulation is of a GPS satellite rising in elevation
(linearly) from 5 degrees to 30 degrees. The simulated antenna position is located 25
metres from the incident boundary and at a height 3 metres above the local terrain.
The FSPE solution has been offset by 30 dB for clarity of comparison.
Again, the FSPE solution agrees extremely well with the exact solution of the wave
equation [9].
6.2 Forward Multipath Propagation
6.2.1 Static Test
Having established that the FSPE is valid against an exact plane-wave solution we set
about testing the PETA against an equivalent exact solution in the time-domain — the
results derived from geometric interpretation of propagation problems [10].
![Page 200: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/200.jpg)
175
The FSPE model was tested for the forward propagation case over a perfect
conductor. Here the domain size is 40 metres in range and height.
Figure 6.4 — Field over flat perfect conductor
The standing wave pattern represents the constructive and destructive interference of
the incident field and the ground reflections and is defined, in the interference region
[11], as
( ) ( ) ( )φθρθ +∆−+= Rkjri efEfE 00E (6.1)
It is interesting to note that for this geometry the distance between minima is 1.147
metres. Here f(θi) and f(θr) represent the field distribution, as a function of the angle of
incidence, and angle of reflection respectively. In terrestrial propagation these factors
would account for the transmission antenna’s radiation pattern. For the GPS satellite
propagation problem, these elevation angle-dependent pattern shape terms account for
the variation of received power due to the transmitting antenna array design as shown
in Figure 6.5.
![Page 201: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/201.jpg)
176
Figure 6.5 — User Received Minimum L1 C/A Signal Level
The Fourier synthesis technique (PETA) was applied to the FSPE model and the
results for a spatial position, x = 20 m, z = 10 m, are shown in Figure 6.6.
Figure 6.6 — Time series for forward propagation
A comparison of results for the modelled and calculated delays is given in Table 6.1.
![Page 202: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/202.jpg)
177
Delay Calculated (ns) Modelled (ns) Error (ns)
Forward 66.9213 66.9189 0.0024 (0.004%)
Forward Scatter 5.8104 5.8350 0.0246 (0.4%)
Table 6.1 — Modelling Errors Forward
It should be noted that the results are for propagation over a perfect conductor and
hence the multipath signals are of equal amplitude to the line-of-sight (LOS) incident
GPS signal. The results show excellent agreement between the PETA results and
those derived from geometric interpretation of the problem.
6.2.2 Dynamic Tests
We now consider the case of simulated GPS satellite motion and forward specular
reflection, as depicted in Figure 6.7,. Again in this multipath situation we have the
direct LOS signal and a single multipath signal arriving at the antenna.
5 m
Figure 6.7 — Dynamic situation
The modelling is for a GPS satellite rising in elevation from 5 degrees to 10 degrees.
Figure 6.8 and Figure 6.9 show, respectively, the calculated FSPE field, and the PETA
result for a propagation angle of 8 degrees.
![Page 203: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/203.jpg)
178
Figure 6.8 — FSPE field at 8 degrees
This plot of the field strength again shows the classical interference region pattern,
with constructive and destructive interference clearly evident in height. The distance
between minima is calculated at 0.684 metres.
0 5 10 15 20 25 30 35 400
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Relative time delay (ns)
peak 1 delay: 016.8750 nspeak 2 delay: 021.5039 nstime delay: 4.6289 ns
Figure 6.9 — PETA result at 8 degrees
The time-domain analysis clearly shows the LOS and the multipath signals. Each of
the multipath parameters is extracted from the PETA results at 0.1 degree increments,
![Page 204: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/204.jpg)
179
and by using equation (5.24 Chapter 5), we can reconstruct the total field. Figure 6.10
shows a comparison of the PETA estimated field compared to the full field solution as
given by the FSPE propagation model.
5 6 7 8 9 10-40
-35
-30
-25
-20
-15
-10
-5
0
5
10
Propagation angle (degs)
Fie
ld a
mpl
itude
(dB
)
Figure 6.10 — FSPE field and reconstructed PETA field comparison
This figure again shows the classical fading pattern for a single multipath reflection.
The results from the PETA reconstruction are in good agreement with the full field
result given by the FSPE. having compared the PETA and FSPE field results we now
make a comparison of the PETA multipath relative amplitude with the implemented
reflection coefficient of concrete for an antenna height of 2 metres, F-mode
propagation from 1 to 25 degrees. The results are shown below in Figure 6.11.
![Page 205: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/205.jpg)
180
0 5 10 15 20 250
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Reflection Coefficient Magnitude: circ(combined) - Concrete
Mag
nitu
de
Propagation angle (degs)
PETA Calculated
Figure 6.11 — Comparison of RC Magnitudes
The reflection coefficient, as implemented in the FSPE and ultimately in the PETA, is
in strong agreement with the calculated reflection coefficient for this problem—LHCP
rejection of 6dB. The fade pattern for this problem is also tested against that of an
exact interpretation of the refection coefficient, Figure 6.12.
0 5 10 15 20 25-30
-25
-20
-15
-10
-5
0
5
10FSPE Fade Pattern
Fie
ld L
evel
(dB
)
Propagation Angle (degs)
FSPE Calculated Envelope
Figure 6.12 — Fade Pattern Comparison
![Page 206: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/206.jpg)
181
Again the FSPE-PETA implementation of the boundary refection coefficients is in
excellent agreement with the expected results.
6.3 Forward Diffraction
6.3.1 Static Test
The FSPE propagation model was tested in the diffracted signal shadow area behind a
terrain block with a propagation angle of 2 degrees. The terrain block may represent a
building for the purpose of this study. The height of the terrain block is 20 metres and
is situated in the domain 20 metres from the left side boundary. The total domain
range is 60 metres and the domain height is 40 metres. The spatial point is selected as
being 19 metres above the baseline terrain and 50 metres in range giving a diffracted
ray angle of approximately 5 degrees. This places the spatial point in the geometric
optics shadow zone, a zone that is treated correctly for diffraction by this PE model.
The field plot for this situation is shown in Figure 6.13.
Figure 6.13 — Diffraction over terrain element
![Page 207: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/207.jpg)
182
The diffractive effects that are expected can be quantified by considering the
geometrical arrangement represented here.
LOS
αβ
h
y1
y2
d2
d1y3
l
x1 x2
x3
Figure 6.14 — Diffraction geometry
The implementation of the model used for this problem calculates time delay relative
to the line-of-sight. The total distance modelled is the line-of-sight ray-path to the
intersection of the perpendicular marked h, (this distance is represented on the
diagram as l ), plus the distance from this point to the spatial point of interest in the
domain (distance d1) . The total diffracted ray path distance is l+d2, giving the path
difference as d2-d1. The relevant calculations are as follows.
The angle between the line-of-sight vector (propagation angle α) and the diffracted
ray is given by,
β α=
−−tan 1 2
2
y
x(6.2)
with
d x y2 22
22= + (6.3)
![Page 208: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/208.jpg)
183
and
d d1 2= cosβ (6.4)
The relative time delay between the line-of-sight and the diffracted signal is then
given by
( )path delayd
c= −2 1 cosβ (6.5)
The path delay for this particular problem is calculated as 0.070226 ns. Time spacing
resolution for this modelling was 0.00244 ns.
Figure 6.15 shows the time series of the line-of-sight without terrain obstruction. Note
that the ground reflection is present in this plot.
Figure 6.15 — Time series no terrain
The amplitude for this plot has been normalised, with a maximum recorded value of
25.76e-3. The line-of-sight delay to the spatial point of interest is modelled at
![Page 209: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/209.jpg)
184
166.7725 ns with forward-scatter delay of 4.4189 ns. Actual time delay calculations
for this case are, 166.7683 ns with forward-scatter delay of 4.4206 ns.
Modelling the time series with the inclusion of the terrain block resulted in the plot
shown in Figure 6.16.
Figure 6.16 — Time series with terrain
Again this plot has been normalised against the maximum amplitude, for this case
3.23e-3. The forward delay is seen to be 166.8457 ns. Note there is no dominant
ground reflection, as would be expected due to the terrain obstruction. However it is
interesting to note the reflected diffracted signal to the right of the diffracted solely
signal.
The modelled diffraction delay is simply the difference between the line-of-sight
propagation time and the propagation time of the diffracted signal. Therefore, the time
delay for this case is 0.0732 ns, this compares reasonably, with the calculated delay of
0.0702 ns. It must be remembered however that the time resolution for this particular
case is ±0.00244 ns.
![Page 210: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/210.jpg)
185
6.3.2 Dynamic Test
We now include satellite dynamics and present the case of a GPS satellite rising over
a terrain obstruction, from an initial elevation angle of 5 degrees to a final angle of 15
degrees. In this case we can expect diffraction effects to dominate. The situation is
depicted in Figure 6.17.
5 m
3 m
2 m
Figure 6.17 — Dynamic diffraction situation
A plot of the instantaneous PE field, for a propagation angle of10 degrees, is given in
Figure 6.18.
Figure 6.18 — FSPE field at 10 degrees
![Page 211: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/211.jpg)
186
Diffractive effects are evident behind the terrain obstacle. These diffractive effects are
clearly presented in Figure 6.19.
5 10 15 20 25 30 35 40 45-20
-15
-10
-5
0
5Diffracted Field Value
Fie
ld a
mpl
itude
(dB
)
Propagation angle (degs)
Figure 6.19 — FSPE field 5 degrees to 45 degrees
The field result for an elevation change of 5 degrees to 15 degrees, exhibits the usual
features of diffraction by an obstacle, with field strength rising, overshooting and
oscillating about the 0dB LOS level. At about 35 degrees we see the diffractive effects
reducing and ground reflection interference starting to dominate.
![Page 212: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/212.jpg)
187
5 10 1523.4
23.5
23.6
23.7
23.8
23.9
24
24.1
24.2
24.3Diffraction Time Delay
Tim
e de
lay
(ns)
Propagtion angle (degs)
Figure 6.20 — LOS and diffracted propagation time comparison
Figure 6.20 presents a comparison of the delay of the diffracted signal (upper plot) to
that of the unobstructed line-of-sight (lower plot) as provided from the PETA results.
This clearly indicates the additional path length caused by the diffraction of the signal
around the terrain edge. If a receiver has a dynamic range of better than 20 dB then it
is able to acquire and maintain track of the diffracted signal. At 5 degrees the
diffracted path delay is 0.2 ns representing approximately a 6 cm range error. The
convergence of the curves corresponds to the fact that the path becomes line-of-sight
as the satellite rises.
6.4 BA/BB-Mode: Backscatter
This section considers a two-way FSPE model developed from the one-way model
with the inclusion of backscatter. Results are presented for the static case and for
satellite motion.
Propagation angle (degs)
![Page 213: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/213.jpg)
188
6.4.1 Static Test
Figure 6.21 shows the field for a 5 degree forward (one-way) GPS signal, propagating
over simple terrain elements. The interference region (standing wave) pattern,
generated by the incident and reflected fields, is clearly evident in the unobstructed
areas of the domain. The shadowing/diffraction effects of the propagation over the
terrain blocks is also shown.
Figure 6.21 — Forward propagation over terrain
The back propagation for this example is shown in Figure 6.22. Here the field values
are for the back reflected field only. The interference pattern in front of the block at
the 40 metre range is as expected. There is no field after the 100 metre range since
there are no reflected interfaces after this point. The amplitude of the reflected field
components has been arbitrarily set at 80% of the incident components at the
reflecting boundary and with a 180 degree phase shift.
![Page 214: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/214.jpg)
189
Figure 6.22 — Back propagation from reflected interfaces
The total field (two-way) propagation for this example is shown in Figure 6.23. This
is resultant of the addition of the forward and back fields.
Figure 6.23 — Total propagated field
The time delay analysis using the Fourier synthesis technique is now tested with a
domain consisting of a single 20 metre high vertical reflector located at a 20 metre
range point. The total resultant two-way field is shown in Figure 6.24.
![Page 215: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/215.jpg)
190
Figure 6.24 — Two-way field with 20 m high vertical reflector at 20m
The time series for the spatial point, x = 10 m, z = 10 m, is shown below—Figure
6.25.
Figure 6.25 — Time series back-propagation
The stored and reflected field components were arbitrarily set to 80% of the incident
field level. This was done to provide clarity in the normalised plot of Figure 6.25, and
also to demonstrate that the backscatter algorithm can implement reflection
coefficient conditions on the reflecting boundary. This will allow the reflections, as
![Page 216: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/216.jpg)
191
implemented in the backscatter algorithm, to have the correct amplitude and phase
variation, as determined by the electrical characteristics of the modelled terrain.
A comparison between calculated and modelled results is given in Table 6.2. Here
Backscattera refers to backscatter arriving from above the horizontal, and Backscatterb
refers to backscatter arriving from below.
Delay Calculated (ns) Modelled (ns) Error (ns)
Forward 33.4607 33.4961 0.0354 (0.1%)
Forward Scatter 5.8104 5.7617 0.0487 (0.8%)
Backscattera 66.4130 66.4062 0.0068 (0.01%)
Backscatterb 72.2234 72.2656 0.0422 (0.06%)
Table 6.2 — Modelling Errors Forward and Back
The modelled domain is shown below in Figure 6.26.
10 m 10 m
20 m
10 m
reflector
forward Bb
Ba
LOS
Figure 6.26 — Model domain
![Page 217: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/217.jpg)
192
6.4.2 Dynamic Test
To demonstrate satellite dynamics we present a more complicated arrangement that
may occur in urban environments. The situation is depicted in Figure 6.27.
5 m 5 m
6 m
2 m
2 m
5 m
B1
B2
Figure 6.27 — Stepped backscatter geometry
We call this a stepped backscatter, where in addition to the forward scatter, we have
signals scattered in the reverse propagation direction, from two distinct interfaces.
Figure 6.28 shows the time-domain results for a 5 degree propagation angle. This
modelling case presents all elements of the previous propagation examples, namely;
forward reflection, backscatter and diffractive effects.
![Page 218: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/218.jpg)
193
0 10 20 30 40 50 60 70 800
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Relative time delay (ns)
Nor
mal
ised
am
plitu
de
Propagation angle - 5 °
peak 1 delay: 016.7383 nspeak 2 delay: 019.6484 nspeak 3 delay: 049.9609 nspeak 4 delay: 052.8613 ns
Figure 6.28 — Stepped backscatter PETA results for 5 degrees
Here we see the LOS, forward scatter, and two additional multipath signals reflected
from interface B1. The first of these multipath signals is identified as backscatter from
above, that is, the LOS is reflected from the B1 interface and arrives at the antenna
location from a positive elevation angle. The next signal is backscatter from below,
and is reflection of the LOS from a combination of ground and interface. At 5 degrees
refection the B2 interface is obstructed by the B1 step. Close examination of Figure
6.28 shows some low level signal from the B2 interface, but diffractive effects have
reduced its influence.
The propagation mechanisms in this situation become evident at higher propagation
angles. In Figure 6.29 we show the PETA results for 12.5 degrees.
![Page 219: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/219.jpg)
194
0 10 20 30 40 50 60 70 800
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Relative time delay (ns)
Nor
mal
ised
am
plitu
de
Propagation angle - 12.5 °
peak 1 delay: 017.1680 nspeak 2 delay: 024.3457 nspeak 3 delay: 049.7168 nspeak 4 delay: 056.8945 nspeak 5 delay: 062.7148 ns
Figure 6.29 — Stepped backscatter PETA results for 12.5 degrees
Here we see the expected variation in influence of the reflecting interfaces. The
multipath signal from B1 (above) is now affected by diffraction effects, and the
multipath signal from B2 (above) is beginning to dominate. At 15 degrees (Figure
6.30) the effect is more pronounced and the reflection from B2 is essentially line-of-
sight. We note the reflection from B1 (below) is unaffected, and that there is reflection
from B2 (below).
0 10 20 30 40 50 60 70 800
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Relative time delay (ns)
Nor
mal
ised
am
plitu
de
Propagation angle - 15 °
peak 1 delay: 017.3828 nspeak 2 delay: 025.9863 nspeak 3 delay: 049.5996 nspeak 4 delay: 058.1836 nspeak 5 delay: 062.5000 ns
Figure 6.30 — Stepped backscatter PETA results for 15 degrees
![Page 220: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/220.jpg)
195
The total multipath propagation situation is presented in Figure 6.31. Here we see the
full influence of the diffractive effects for this situation.
3040
5060 70
80
2010
05
15
10
Time Delay (ns)
PropagationAngle (degs)
Figure 6.31 — Stepped backscatter PETA results for 5 to 15 degrees
We now compare these diffraction results to the theory of knife-edge diffraction. The
edge represented is the top of the B1 interface with geometry as shown below in
Figure 6.32.
5 m 2 m
5 m
T
R
d1
d2
h
Figure 6.32 — Stepped backscatter knife-edge geometry
In this representation the energy reflected from the B2 interface is assumed to act as
an electromagnetic radiation source, hence it appears as a transmitter (T). The receiver
![Page 221: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/221.jpg)
196
(R) is located at the indicated spatial position, with d1 and d2 respectively
representing the distance of transmitter and receiver from the diffracting knife-edge. It
is now possible to apply the CCIR recommended formula [13] for knife-edge
obstacles, which is given by
+=
21
112
ddhv
λ(6.6)
which, for large values of d1 reduces to
v hd
=2
2λ(6.7)
For v greater than –1 the approximate value of loss in decibels is given by
( ) ( )
−++−+= 1.011.0log209.6 2 vvvJ (6.8)
A plot of the diffraction loss for the given geometry is shown in Figure 6.33.
11 11.5 12 12.5 13 13.5 14 14.5 150.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3Knife Edge Diffraction Loss
Rel
ativ
e am
plitu
de
Propagtion angle (degs)
Figure 6.33 — Diffraction loss over B1 interface
Propagation angle (degs)
![Page 222: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/222.jpg)
197
We note from Figure 6.33 that the propagation is not line-of-sight until about 14.4
degrees. A plot of the normalised CCIR diffraction loss is made in comparison with
the loss as given by the PETA results, Figure 6.34.
5 10 150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Back-Scatter Amplitudes
Rel
ativ
e am
plitu
de
Propagtion angle (degs)
Figure 6.34 — Normalised diffraction loss PETA vs CCIR for B1 interface
Here we see that for the B1 interface diffraction, the CCIR (circles) and PETA (line)
results are in good agreement. The additional plot given in Figure 6.34 is the PETA
diffraction results for the B2 interface.
The total plot of PETA derived time delays, for the stepped backscatter situation, is
shown in Figure 6.35.
Propagation angle (degs)
![Page 223: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/223.jpg)
198
0 10 20 30 40 50 60 70 805
6
7
8
9
10
11
12
13
14
15Multipath Time Profile
Pro
paga
tion
angl
e (d
egs)
Time dealy (ns)
Figure 6.35 — PETA path delays
Starting from the left we have the LOS, forward scatter, B1 backscatter (above), B1
backscatter (below) and B2 backscatter (above) delay profiles. Results are now
presented for the reconstruction of the total field from the PETA results. In Figure
6.36 the reconstructed field (diamonds) shows excellent agreement with the FSPE
field result.
5 10 15-20
-15
-10
-5
0
5
10
15
Propagation angle (degs)
Fie
ld a
mpl
itude
(dB
)
Figure 6.36 — Reconstructed PETA field and FSPE field comparison
Time delay (ns)
![Page 224: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/224.jpg)
199
The RMS error is 1.953 dB. A comparison of the FSPE results to those expected by
geometric optics (GO) is now shown in Figure 6.37. The geometric optics solution is
simply the field calculated from the trigonometric ray paths and their relative delays,
which do not account for any diffraction. Here the RMS error is 4.8391 dB and it is
obvious that by neglecting diffraction effects large errors are produced in the
estimation of the field for this situation.
5 10 15-25
-20
-15
-10
-5
0
5
10
15
Propagation angle (degs)
Fie
ld a
mpl
itude
(dB
)
Figure 6.37 — Reconstructed GO field and FSPE field comparison
6.5 Summary
Results were presented that show the validity of the FSPE and PETA in a variety of
static and dynamic GPS modelling scenarios. The visualisation of the full-field and
the time-domain results given by the FSPE-PETA model were shown to give an
insight into the behaviour of GPS multipath in a variety of static and dynamic GPS
simulations.
![Page 225: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/225.jpg)
200
6.6 References
[1] G. D. Dockery, “Development and Use of Electromagnetic Parabolic Equation
Propagation Models For Us Navy Applications,” Johns Hopkins APL
Technical Digest, vol. 19, pp. 283-292, 1998.
[2] F. B. Jensen, W. A. Kuperman, M. B. Porter, and H. Schmidt, “Parabolic
Equations,” in Computational Ocean Acoustics, R. T. Beyer, Ed. New York:
AIP Press, 1994.
[3] G. D. Akrivis, V. A. Dougalis, and N. A. Kampanis, “Error Estimates for
Finite Element Methods for a Wide-Angle Parabolic Equation,” Applied
Numerical Mathematics, vol. 16, pp. 81-100, 1994.
[4] G. D. Akrivis, V. A. Dougalis, and G. E. Zouraris, “Error Estimates for Finite-
Difference Methods for a Wide-Angle Parabolic Equation,” SIAM Journal of
Numerical Analysis, vol. 33, pp. 2488-2509, 1996.
[5] M. A. Leontovich and V. A. Fock, “Solution of the Problem of Propagation of
Electromagnetic Waves along the Earth's Surface by the Method of Parabolic
Equations,” Journal of Physics of the USSR, vol. 10, pp. 13-24, 1946.
[6] M. D. Collins, “Comparison of algorithms for Solving Parabolic Wave
Equations,” Journal of the Acoustic Society of America, vol. 100, pp. 178-182,
1996.
[7] M. D. Collins, “A Split-Step Padé Solution for the Parabolic Equation
Method,” Journal of the Acoustic Society of America, vol. 93, pp. 1736-1742,
1993.
[8] S. Ramo, J. R. Whinnery, and T. van Duzer, Fields and Waves in
Communication Electronics. New York: John Wiley & Sons, 1984.
[9] R. A. Walker, “Operation and Modelling of GPS Sensors in Harsh
Environments,” in School of Electrical and Electronic Systems Engineering:
Queensland University of Technology, 1999.
[10] M. Ando, “The Geometrical Theory of Diffraction,” in Analysis Methods for
Electromagnetic Problems, E. Yamashita, Ed. Boston: Artech House, 1990,
pp. 213-242.
![Page 226: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/226.jpg)
201
[11] D. E. Kerr, “Propagation of Short Radio Waves,” . Boston: Boston Technical
Publishers, 1964.
[12] J. Doble, Introduction to Radio Propagation for Fixed and Mobile
Communications. Boston: Artech House, 1996.
[13] CCIR International Radio Consultive Commitee, “Report 715-2 Propagation
by Diffraction,” International Telecommunications Union, Dubrovnik 1986.
![Page 227: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/227.jpg)
202
THIS PAGE IS INTENTIONALLY BLANK
![Page 228: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/228.jpg)
203
Chapter 7 Simulation of GPS Propagation
In the previous chapters the FSPE and PETA were introduced as novel methods for
the modelling and simulation of GPS multipath propagation. In this chapter we
examine collected GPS data and FSPE simulations. The complete data results can be
found in Appendix B.
7.1 Introduction
The previous chapter validated the FSPE against an exact solution; validated the
PETA to geometric optics; and validated the PETA to the FSPE field results. This
combinational validation allows us now to compare the FSPE to measured GPS signal
data, with confidence in the underlying modelling techniques used for the following
simulation cases.
7.2 Candidate Simulation Cases
A data collection exercise was initiated for testing of the FSPE model. Several
candidate sites were examined and it was determined the most appropriate multipath
data could be collected in a classic F-mode situation. Water features were considered
the most appropriate as the surface would be uniform and only affected by wind
turbulence. This turbulence would generate rough surface effects [1] and for pure
specular reflection a calm situation was required [2].
With this in mind — and with a desire to test different reflection coefficients — two
sites were selected; one a fresh-water feature and the other a flat grass playing field .
The comparison of the results of the GPS signal data and the FSPE simulations are
given in the following sections. The full data results can be found in Appendix B, as
well as data for a third site at Bribie Island for the investigation of sea water multipath
and tidal variation effects.
![Page 229: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/229.jpg)
204
7.2.1 Pine Dam
The first data collection exercise was made at Pine Dam, north of Brisbane in
Queensland, Australia. The large fresh-water surface provided an excellent
environment for collection of specular multipath data. The data was collected on a day
with very little wind, thus the surface was considered specular, and surface roughness
effects could be neglected. In Figure 7.1 the set-up for the data collection at Pine Dam
is shown.
Figure 7.1 — Pine Dam Data Collection Site
The antenna was mounted on a tripod with the antenna oriented vertically, such that
the full antenna gain pattern, see Figure 7.2, incorporated the expected specular
multipath reflection [3]. The height of the antenna above the water level was
measured at approximately 1.26 metres.
![Page 230: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/230.jpg)
205
Water Feature
AntennaGain Pattern
LOS
MP
1.26m
Figure 7.2 — Data Collection Basis
The orientation of the antenna was selected such that as much of the water feature as
possible was within the radiation pattern of the GPS receiver antenna. The bore-sight
orientation was also selected for the greatest horizontal distance across the water
feature before terrain was encountered in the opposite shoreline. In this way the
maximum number of satellites would be propagating within the known modelling
domain. The orientation of the GPS receiver antenna is illustrated in Figure 7.3 below.
N
Shoreline
Antenna Boresight
Water Surface
NW
Figure 7.3 — Location Orientation
![Page 231: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/231.jpg)
206
The limit in azimuth that is permitted for reflection off the water surface is
approximately SSW to approximately NNE. This represents from approximately –165
degrees to approximately +25 degrees azimuth angle range.
Data was collected using an Ashtech Z-Surveyor GPS receiver and the Ashtech
700700 L1/L2 antenna. The data collection interval was set at 1 second epochs. The
data was analysed using a combination of TEQC from UNAVCO [4] and MATLAB
scripts. After approximately 2 hours of data collection four satellites were seen to
have maintained full specular reflection with the desired surface. The three satellites
were SV’s 17, 6, 8, and 9.
Before going further it is worthwhile noting that it is usual and correct to term the
signal strength reported by a GPS receiver as carrier-to-noise C/No, which is typically
in units of dB-Hz, thus inferring some knowledge of the tracking bandwidth. This
refers to the fact that the GPS signal is direct sequence spread spectrum and the
receiver estimates the C/No since it is a stochastic process. .In this work we are
seeking to validate the FSPE and hence the exact nature of the signal is not important,
and we revert back to the more generic term of signal-to-noise.
In addition, the trend of the power received at the GPS receiver is a function of many
variables and as such is difficult to estimate. In this validation no attempt has been
made to trend the simulated incident power other than the power variation specified in
the Standard Positioning Service specification. The simulated values (which are
inherently normalised to a SNR of 0 dB) are shifted in relative terms by an estimated
received incident signal strength.
The simulation results for SV 17 are shown below in Figure 7.4.
![Page 232: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/232.jpg)
207
0 500 1000 1500 2000 2500 3000 3500 4000 450042
44
46
48
50
52
54
Epoch Samples
S/N
(dB
)
S/N Comparison SV17-Fresh Water
MeasuredModelled
Figure 7.4 — SV17 Results
The total data simulated in this case was for a total period of 4500 epochs or 75
minutes. For the first 3000 epochs or 50 minutes the simulation is in excellent
agreement with the data. The LHCP rejection ratio was selected to be 11 dB (typical
range for a 3dB axial ratio specification). SV 17 is seen to be rising from
approximately 10 degrees to approximately 32 degrees in elevation for the simulation
period, Figure 7.5. In azimuth the SV is moving from North-West to the West,
starting approximately in-line with the bore-sight at North-West.
0 500 1000 1500 2000 2500 3000 3500 4000 45000
10
20
30
40Az-El SV17
Ele
vatio
n (d
eg)
0 500 1000 1500 2000 2500 3000 3500 4000 4500-80
-70
-60
-50
Azi
mut
h (d
eg)
Figure 7.5 — SV17 AZ-EL Data
![Page 233: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/233.jpg)
208
The geometric point of reflection; which represents the centre of the Fresnel zone is
very close to the antenna, ranging in distance from approximately 7 metres to 2 metres
horizontally distant from the antenna for the observation period, Figure 7.6.
0 500 1000 1500 2000 2500 3000 3500 4000 45000
5
10Specular Reflection Point SV17 :antenna height = 1.26m
Ran
ge D
ista
nce
(m)
0 500 1000 1500 2000 2500 3000 3500 4000 45000.5
1
1.5
1st
Zone
Rad
ius
(m)
0 500 1000 1500 2000 2500 3000 3500 4000 45000
5
10
sem
i-maj
or a
xis
(m)
Measurement Epochs
Figure 7.6 — SV17 Fresnel Data
We note the Fresnel zone dimensions are quite small. The differing value of the
simulation and data from epoch 3000 appears to be a variation in the received signal
strength of SV17, with an apparent downward trend. The result could also be a factor
of the proximity of the shore line impinging within the Fresnel zone of the higher
elevation angles. Any variation in the surface condition; height, reflection coefficient,
or surface roughness will affect the simulation results.
The second set of simulation data is that for SV6, Figure 7.7.
![Page 234: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/234.jpg)
209
0 500 1000 1500 2000 2500 3000 3500 400020
25
30
35
40
45
50
Epoch Samples
S/N
(dB
)
S/N Comparison SV6-Fresh Water
MeasuredModelled
Figure 7.7 — SV6 Results
We note that the fades in the signal strength data recorded, are of such depth that the
tracking loop has in fact lost lock on two occasions. The first is approximately
between epochs 510 and 800, with the second occurring between epochs 2800 and
3200. We note that the loss of tracking exists for approximately 5-6 minutes on each
occasion.
The simulation again shows excellent agreement with the data. We note that this SV
is also rising in elevation (approximately 25 degrees to 42 degrees), and moving
approximately from South-West to approximately South-South-West, Figure 7.8.
![Page 235: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/235.jpg)
210
0 500 1000 1500 2000 2500 3000 3500 400020
25
30
35
40
45Az-El SV6
Ele
vatio
n (d
eg)
0 500 1000 1500 2000 2500 3000 3500 4000-150
-140
-130
-120
Azi
mut
h (d
eg)
Figure 7.8 — SV6 AZ-EL Data
As for the previous SV the Fresnel zone becomes progressively smaller and closer to
the antenna location as the SV rises in elevation angle, Figure 7.9. The slight variation
in the simulation results near the last peak in the fading pattern would appear to be the
local variable terrain being included in the reflective properties of the water surface
and affecting the multipath signal slightly.
0 500 1000 1500 2000 2500 3000 3500 40001
2
3Specular Reflection Point SV6 :antenna height = 1.26m
Ran
ge D
ista
nce
(m)
0 500 1000 1500 2000 2500 3000 3500 40000.4
0.6
0.8
1st
Zone
Rad
ius
(m)
0 500 1000 1500 2000 2500 3000 3500 40000.5
1
1.5
2
sem
i-maj
or a
xis
(m)
Measurement Epochs
Figure 7.9 — SV6 Fresnel Data
![Page 236: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/236.jpg)
211
Our third simulation is for SV8, Figure 7.10.
0 500 1000 1500 2000 2500 3000 3500 4000 450044
46
48
50
52
54
56
Epoch Samples
S/N
(dB
)
S/N Comparison SV8-Fresh Water
MeasuredModelled
Figure 7.10 — SV8 Results
The simulation provides excellent agreement with the results from epoch 1700. We
again note that at the higher propagation angles that there is some difference between
the simulation and the actual received signal strength results.
0 500 1000 1500 2000 2500 3000 3500 4000 450010
15
20
25
30
35Az-El SV8
Ele
vatio
n (d
eg)
0 500 1000 1500 2000 2500 3000 3500 4000 4500-70
-60
-50
-40
-30
Azi
mut
h (d
eg)
Figure 7.11 — SV8 AZ-EL Data
![Page 237: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/237.jpg)
212
This would appear to indicate that the closeness of the reflection zone — at the higher
elevation angles — to the antenna is again the underlying cause. This is apparent since
the relative peak-to-fade of the simulation and the data does not vary significantly. If
there is a variation between the simulated and the actual reflection coefficient then
this is indicated within the comparison as variation in the depth of the fades. The only
other probable cause for the variation in the simulation, is a variation in the received
signal power at this elevation angle range. This is possible, but would be extremely
difficult to include in a propagation simulation unless the complete power profile is
available.
0 500 1000 1500 2000 2500 3000 3500 4000 45002
4
6
8Specular Reflection Point SV8 :antenna height = 1.26m
Ran
ge D
ista
nce
(m)
0 500 1000 1500 2000 2500 3000 3500 4000 45000.5
1
1.5
1st
Zone
Rad
ius
(m)
0 500 1000 1500 2000 2500 3000 3500 4000 45000
2
4
6
sem
i-maj
or a
xis
(m)
Measurement Epochs
Figure 7.12 — SV8 Fresnel Data
The final simulation for Pine Dam is shown below in Figure 7.13, for SV9.
![Page 238: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/238.jpg)
213
0 500 1000 1500 2000 2500 3000 3500 400038
40
42
44
46
48
50
52
54
56
Epoch Samples
S/N
(dB
)
S/N Comparison SV9-Fresh Water
MeasuredModelled
Figure 7.13 — SV9 Results
Again the simulation using the FSPE is in excellent agreement with the recorded data.
The SV is setting, starting at approximately 20 degrees before disappearing below the
horizon. The SV is moving across from approximately the bore-sight of the antenna at
NW around to the North, Figure 7.14.
0 500 1000 1500 2000 2500 3000 3500 40000
5
10
15
20
25Az-El SV9
Ele
vatio
n (d
eg)
0 500 1000 1500 2000 2500 3000 3500 4000-25
-20
-15
-10
-5
0
Azi
mut
h (d
eg)
Figure 7.14 — SV9 AZ-EL Data
![Page 239: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/239.jpg)
214
We also note that the Fresnel zone dimensions are increasing and the distance to the
zone is increasing, Figure 7.15.
0 500 1000 1500 2000 2500 3000 3500 40000
50
100Specular Reflection Point SV9 :antenna height = 1.26m
Ran
ge D
ista
nce
(m)
0 500 1000 1500 2000 2500 3000 3500 40000
2
4
1st
Zone
Rad
ius
(m)
0 500 1000 1500 2000 2500 3000 3500 40000
100
200
sem
i-maj
or a
xis
(m)
Measurement Epochs
Figure 7.15 — SV9 Fresnel Data
7.2.2 Caboolture Soccer Field
The second data collection and simulation exercise was undertaken at a grass field
located in the township of Cabooture. The experimental set-up and the nature of the
environment is shown in Figure 7.16.
![Page 240: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/240.jpg)
215
Figure 7.16 — Caboolture Data Collection Site
The data collected permitted the simulation of SV’s 21, 23, and 5. The simulated
fading patterns from the FSPE and the elevation, azimuth and Fresnel zone data is
given in the following figures.
4000 4500 5000 5500 6000 6500 7000 750025
30
35
40
45
50
55
60
Epoch Samples
S/N
(dB
)
S/N Comparison SV21-Wet Ground
MeasuredModelled
Figure 7.17 — SV21 Results
![Page 241: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/241.jpg)
216
4000 4500 5000 5500 6000 6500 7000 75000
5
10
15
20
25Az-El SV21
Ele
vatio
n (d
eg)
4000 4500 5000 5500 6000 6500 7000 7500-55
-50
-45
-40
-35
Azi
mut
h (d
eg)
Figure 7.18 — SV21 AZ-EL Data
4000 4500 5000 5500 6000 6500 7000 75000
50
100Specular Reflection Point SV21 :antenna height = 1.3m
Ran
ge D
ista
nce
(m)
4000 4500 5000 5500 6000 6500 7000 75000
2
4
6
1st
Zone
Rad
ius
(m)
4000 4500 5000 5500 6000 6500 7000 75000
100
200
300
sem
i-maj
or a
xis
(m)
Measurement Epochs
Figure 7.19 — SV21 Fresnel Data
![Page 242: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/242.jpg)
217
0 500 1000 1500 2000 2500 3000 350036
38
40
42
44
46
48
50
52
54
56
Epoch Samples
S/N
(dB
)
S/N Comparison SV23-Wet Ground
MeasuredModelled
Figure 7.20 — SV23 Results
0 500 1000 1500 2000 2500 3000 35005
10
15
20
25
30Az-El SV23
Ele
vatio
n (d
eg)
0 500 1000 1500 2000 2500 3000 3500-32
-30
-28
-26
-24
Azi
mut
h (d
eg)
Figure 7.21 — SV23 AZ-EL Data
![Page 243: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/243.jpg)
218
0 500 1000 1500 2000 2500 3000 35000
5
10
15Specular Reflection Point SV23 :antenna height = 1.3m
Ran
ge D
ista
nce
(m)
0 500 1000 1500 2000 2500 3000 35000.5
1
1.5
21s
t Zo
ne R
adiu
s (m
)
0 500 1000 1500 2000 2500 3000 35000
10
20
sem
i-maj
or a
xis
(m)
Measurement Epochs
Figure 7.22 — SV23 Fresnel Data
1500 2000 2500 3000 3500 4000 4500 5000 5500
25
30
35
40
45
50
Epoch Samples
S/N
(dB
)
S/N Comparison SV5-Wet Ground
MeasuredModelled
Figure 7.23 — SV5 Results
![Page 244: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/244.jpg)
219
1500 2000 2500 3000 3500 4000 4500 5000 55000
10
20
30
40Az-El SV5
Ele
vatio
n (d
eg)
1500 2000 2500 3000 3500 4000 4500 5000 55006
8
10
12
14A
zim
uth
(deg
)
Figure 7.24 — SV5 AZ-EL Data
1500 2000 2500 3000 3500 4000 4500 5000 55000
10
20Specular Reflection Point SV5 :antenna height = 1.3m
Ran
ge D
ista
nce
(m)
1500 2000 2500 3000 3500 4000 4500 5000 55000.5
1
1.5
2
1st
Zone
Rad
ius
(m)
1500 2000 2500 3000 3500 4000 4500 5000 55000
10
20
30
sem
i-maj
or a
xis
(m)
Measurement Epochs
Figure 7.25 — SV5 Fresnel Data
In all three cases for the grass field at Caboolture the FSPE simulation provided
excellent agreement to the received signal strength.
![Page 245: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/245.jpg)
220
7.3 Summary
In this chapter results of comparisons of simulated propagation with real collected
data has been shown. In all cases the FSPE model has provided simulations that are in
excellent agreement with actual received GPS signals.
Two representative environments were chosen, one representing fresh water and the
other wet ground. Results from the FSPE and the PETA models were in excellent
agreement with GPS multipath propagation measurements in these simple
environments. This provides the basis for simulation of more complex environments.
![Page 246: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/246.jpg)
221
7.4 References
[1] P. Beckman and A. Spizzichino, The Scattering of Electromagnetic Waves
From Rough Surfaces. Norwood: Artech House, 1987.
[2] CCIR International Radio Consultive Commitee, “Report 1008 Reflection
from the Surface of the Earth,” International Telecommunications Union,
Dubrovnik 1986.
[3] A. Kavak, G. Xu, and W. Vogel, “GPS Multipath Fade Measurements to
Determine L-Band Ground Reflectivity Properties,” presented at 20th NASA
Propagation Experimenters Meeting, Pasadena, 1996.
[4] UNAVCO/UCAR, “TEQC,” , 2000 Feb 29 ed. P.O. Box 3000 Boulder,
Colorado 80307-3000: UNAVCO, 2000.
![Page 247: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/247.jpg)
222
THIS PAGE IS INTENTIONALLY BLANK
![Page 248: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/248.jpg)
223
Chapter 8 SNR for Deriving a Height Observable
In this chapter an analysis is made of the fading effect on the received GPS signal
caused by multipath over defined surfaces. This analysis suggest that it is possible to
use this effect, of multipath on the signal-to-noise ratio (SNR) or more correctly the
carrier-to-noise C/No, to derive information about the reflecting surface.
8.1 Introduction
In the process of collecting and analysing GPS multipath data it became apparent that
variation in the signal strength fading pattern contained extractable information that
described aspects of the reflecting surface [1, 2]. Consider the comparison of GPS
SNR data and the modelled simulation for propagation over tidal sea water, Figure
8.1.
0 1000 2000 3000 4000 5000 6000 700025
30
35
40
45
50
55
60
Figure 8.1 — SV21 Bribie Island Tidal Variation
To accurately model the fading pattern for this situation, accurate tidal information
was required such that the relative height between the reflecting surface and the GPS
![Page 249: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/249.jpg)
224
antenna could be used in the model. Without this accurate information the modelling
results were very inaccurate. This highlighted the fact that the fading pattern is a direct
function of relative antenna height, as was shown by equation (2.36) given in Chapter
2.
With this in mind, a technique was investigated that makes use of the often
undesirable fading effect of GPS multipath on the SNR, to determine the relative
change in height between a reflecting surface and the phase centre of a GPS antenna.
Others [3, 4] have used multipath fading characteristics to infer antenna height but
these techniques have relied on expensive or multiple receivers . This relative height
change technique does not rely on special receiver techniques and can be applied to a
number of practical applications from observations of the local tide heights, or
observation of the change in water level in a dam.
Land subsidence observations [5] could also be made by deliberately placing ground
reflectors to produce multipath. As the antenna moves with respect the located ground
reflector a change in height would be recorded. The advantage that this technique has
over standard GPS positioning, is that an expensive carrier phase receiver is not
required [6]. So for slope monitoring applications, networks of cheap OEM GPS
receivers could feasibly be used to determine subsidence.
This work shows that processing of the SNR from a standard GPS receiver yields cm
level accuracy for relative height change. We assume that a single reflection is caused
by the reflecting surface. The GPS antenna is orientated such that the gain pattern
includes both the line-of-sight and the reflection from the surface. The signal received
is a combination of these two signals and the fade pattern in the received S/N depends
on the antenna height. Analysis of these fade patterns over a period of time allows an
estimation of the height variation. Results are presented verifying the technique.
![Page 250: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/250.jpg)
225
Reflected path
Direct path
GPS antenna
Relative Height
Figure 8.2 — Relative Height Measurement
Figure 8.2 shows the geometry and the relative height measurement that is estimated
with this technique.
8.2 Estimating Antenna Height from the GPS S/N Ratio
The model chosen for this work is a simple two-ray plane-wave representation that
assumes that there is [7-9]:
• A single specular reflection from the reflecting surface.
• No diffuse scattering.
• No variation in the reflecting surface height at the various geometric reflection
points.
Under these assumptions the complex received signal is given by:
( )RjkSVSV eAS ∆Γ+= 1 (8.1)
The amplitude of the LOS svA , and the reflection coefficient magnitude ρ , only
affect the size of the variation of the peaks and fades in the received signal. Therefore
only the phase upon reflection, φ , affects the signal shape at any given epoch
(elevation angle sample).
![Page 251: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/251.jpg)
226
The received signal-to-noise data is now sampled at the peaks and fades and the
corresponding angle of propagation determined from the elevation angle to the
satellite in use. The fade pattern for SV3 over sea water is shown in Figure 8.3. Here
the signal-to-noise data has been averaged with a sliding 101-point averaging window.
The corresponding samples for the peaks and fades are then determined.
0.85 0.9 0.95 1 1.05 1.1 1.15
x 104
25
30
35
40
45
50
55
60Averaged S/N SV3
S/N
(dB
)
epochs
Raw AveragedSample
Figure 8.3 — Sampled received S/N
The elevation angle and corresponding sampling ( iθ ) is shown in Figure 8.4.
0.85 0.9 0.95 1 1.05 1.1 1.15
x 104
0
2
4
6
8
10
12
14
16
18
20Elevation Angle SV3
Ele
vatio
n A
ngle
(de
gs)
epochs
SV ElevationSample
Figure 8.4 — Sampled elevation angle data
![Page 252: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/252.jpg)
227
We can now solve, at each peak/fade sample, for the local antenna height ih , above
the reflecting surface using
samples ...3,2,1sin2
mik
nh
i
iii =−=
θφπ
(8.2)
where
( )111 −− −+= iiii signnn θθ (8.3)
and 0n , the starting count, is some initial integer value that is even for a peak or odd
for a fade. The sign of the change in propagation angle determines whether n is
incremented or decremented by 1 from the previous value.
In this way as each peak and fade passes through the antenna’s vertical phase-centre
position an integer count is maintained. The initial value for 0n is determined from a
priori knowledge of the antenna height at the first corresponding peak or fade. Once
this initialisation is set the procedure is independently capable of estimating antenna
height from the signal-to-noise ratio.
To clarify the technique, a visual representation of a FSPE derived forward scatter
interference pattern is shown in Figure 8.5.
![Page 253: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/253.jpg)
228
Figure 8.5 — Interference pattern
This is for a low elevation satellite, and represents a domain 25 metres in height and
range. The lower boundary is the plane reflecting surface. An antenna is represented at
3.9 metres in height and 15 metres into the domain. The peaks in the interference
pattern are seen as the dark portions, with the fades lightest. For this particular
representation the antenna is seen to be located at a peak for this propagation angle.
Note that the distance between minima for this case is 2.63 metres. If the height
estimation technique was initialised at this point, the starting value of 0n would be
two, as it is the second peak in the pattern above the reflecting boundary. Peaks in the
interference pattern occur at n = 0, 2, 4…, and correspondingly the first fade is at n =
1, with further fades at n = 3, 5, 7 …etc.
8.3 Results
The technique was validated in a number of experiments. Results are presented for
both still water (where there is no variation of the surface height) and marine
environments.
![Page 254: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/254.jpg)
229
8.4 North Pine Dam
The first experimental arrangement was the determination of the antenna height above
a water feature with no tidal variation. Data was collected at a fresh water dam with
the antenna oriented vertically and facing the main water surface.
The local height from the antenna to the water surface was measured at approximately
1.3 metres. The estimation of antenna height from three SV’s (6, 8 and 9) is shown in
Figure 8.6.
0 500 1000 1500 2000 2500 3000 3500 4000
1.255
1.26
1.265
1.27
Antenna Height Estimation
Ant
enna
Hei
ght
(met
res)
epochs
SV6SV8SV9
Figure 8.6 — Height estimation above fresh water
As the reflection surface height is essentially static it is appropriate to take some
statistics of the resultant height estimation data. The mean of the resultant data is an
estimated antenna height of 1.26 metres, and the standard deviation is 0.0044 metres.
![Page 255: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/255.jpg)
230
8.5 Bribie Island
Data was collected at a tidal coastal site, with a relatively smooth sea surface. Three
SV’s, 10, 21 and 22 were observed with direct reflection from the sea surface. The
estimation of antenna height is shown in Figure 8.7.
3500 4000 4500 5000 5500 6000 6500 7000 7500 80003.3
3.4
3.5
3.6
3.7
3.8
3.9
4
Ant
enna
Hei
ght
(m)
epochs
Antenna Height Estimation
SV10SV21SV22
Figure 8.7 — Height estimation above sea surface
The total tidal variation for the estimation period was a rise of 53 cm [10]. The
estimates of antenna height agree with this variation showing a total decrease in
antenna height of approximately 57 cm for SV22. The other two SV’s observed did
not cover the full observation period of SV22 but do provide confirmation of a
decrease in relative antenna height, corresponding to an increase in the sea-surface
height.
8.6 System Aspects
The presented technique could be implemented as a surface monitor for a variety of
applications. A system-level representation is shown below in Figure 8.8.
![Page 256: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/256.jpg)
231
monitored surface
Suitable SV’s
GPS Rx
surface heightestimation
reflector
S/N
ephemeris data
initialheight
estimate
Figure 8.8 — System representation
Here a GPS receiver is located at a monitoring site and an estimate of antenna height
(surface height) is provided for a selected surface area. However, there are a number
of system level aspects that are required in order to make this technique work. These
requirements are:
• An initial height estimate must be provided.
• The location of the antenna must be precisely known.
• Current ephemeris is required.
• The monitored surface must have a known, preferably invariable reflection
coefficient.
The initial antenna height is required for the initialisation of the estimation technique
as mentioned previously. The antenna location and ephemeris is required so accurate
SV elevation angle information can be generated. Finally, the amount of multipath
must be controlled. It is not desirable to have a very strong multipath signal that
causes the GPS receiver to lose lock, as this would not allow any measurements to be
made. Loss of lock is evident in Figure 8.9 (reflection from salt water), where large
decreases in received signal strength have exceeded the dynamic range of the
receiver’s tracking loops. The resulting data gaps make it difficult to accurately
predict the true position of the fades.
![Page 257: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/257.jpg)
232
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 21
S/N
(dB
)
Figure 8.9 — Raw S/N sea water
Figure 8.10 shows a corresponding fade pattern for SV8 over fresh water. The
reduced reflection coefficient allows the receiver to maintain lock for the full
observation period.
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 8
S/N
(dB
)
Figure 8.10 — Raw S/N fresh water
![Page 258: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/258.jpg)
233
The technique also lends itself to the intentional placing of a ground reflector onto to
surface to be monitored. In this case, careful consideration of the shape and properties
of that ground reflector must be made.
From Figure 8.6 it is seen that multiple observations, from separate satellites in
differing elevation ranges, are combined to produce the best estimate of the GPS
antenna height. To provide a large number of height estimations it is desirable that the
variation of satellite elevation is high, for any given SV.
The accuracy of the height estimation is also a function of the satellite elevation angle.
Equation (8.2) can be recast in the form
FBh = (8.4)
where
θsin2
1
kF = and φπ −= nB (8.5)
If there is an error in the estimate of θ, then ε±=′ FF and the new estimate of height
is given by Bhh ε±=′ . For a given dθ, and h, the error in height is then
approximately bounded by
θε
d
dF
F
hh ±= (8.6)
From equation (8.6) we see that the error in height is a function of the actual height h,
the SV elevation angle θ, and the size of the angle error, dθ. A plot of the error
bounds for a height of 1m, an error in θ of 0.1 degree, for an elevation range of 1
degree to 20 degrees is given in Figure 8.11.
![Page 259: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/259.jpg)
234
0 2 4 6 8 10 12 14 16 18 20-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
SV Elevation (degs)
Hei
ght
Err
or (
m)
1m Height Error Bounds: 0.1 deg error
Figure 8.11 — Height Error Bound 0.1 Deg Error
Reducing the error in θ to 0.01 degree, results in the height error bounds given below
in Figure 8.12.
0 2 4 6 8 10 12 14 16 18 20-0.01
-0.008
-0.006
-0.004
-0.002
0
0.002
0.004
0.006
0.008
0.01
SV Elevation (degs)
Hei
ght
Err
or (
m)
1m Height Error Bounds: 0.01 deg error
Figure 8.12 — Height Error Bound 0.01 Deg Error
As expected there is a ten-fold reduction in the error bounds. From this error analysis
the following observations can be made:
![Page 260: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/260.jpg)
235
• Larger relative heights have proportionally larger errors for a given error in the
estimation of the SV elevation angle.
• The error in height estimation decreases inversely to the SV elevation angle.
• The error in height estimation decreases proportionally to the error in the
estimation of the SV elevation angle.
From Figure 8.3 and Figure 8.4, it is obvious that the error in estimation of the SV
elevation angle, is dependent upon accurate location of the peaks and fades in the
received S/N. In addition, the rate of change, and accuracy of the elevation data
ultimately sets the resolution of the elevation angle error.
![Page 261: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/261.jpg)
236
8.7 Summary
In this chapter use was made of the fading effect on the received GPS signal caused
by multipath over defined surfaces. We saw that it is possible — through
undemanding techniques — to make use of GPS multipath effects on the received
signal-to-noise ratio (SNR), to derive height information about the reflecting surface.
![Page 262: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/262.jpg)
237
8.8 References
[1] A. Kavak, G. Xu, and W. Vogel, "GPS Multipath Fade Measurements to
Determine L-Band Ground Reflectivity Properties," presented at 20th NASA
Propagation Experimenters Meeting, Pasadena, 1996.
[2] K. D. Anderson, "Determination of Water Level and Tides Using
Interferometric Observations of GPS Signals," Journal of Atmospheric and
Oceanic Technology, vol. 17, pp. 1118–1127, 2000.
[3] C. E. Cohen, "Attitude Determination Using GPS," in The Department of
Aeronautics and Astronautics. Stanford: Stanford University, 1993, pp. 184.
[4] S. Wu, T. Meehan, and L. Young, "The Potential Use of GPS Signals as
Ocean Altimetry Observables," presented at ION National Technical Meeting,
Santa Monica, 1997.
[5] R. A. Walker, "Multipath Issues in GPS Monitoring," presented at
International Workshop on Advances in GPS Deformation Monitoring, Perth,
Australia, 1998.
[6] E. D. Kaplan, "Understanding GPS: Principles and Applications," in Mobile
Communications Series, J. Walker, Ed. Boston: Artech House, 1996.
[7] P. Beckman and A. Spizzichino, The Scattering of Electromagnetic Waves
From Rough Surfaces. Norwood: Artech House, 1987.
[8] CCIR International Radio Consultive Commitee, "Report 715-2 Propagation
by Diffraction," International Telecommunications Union, Dubrovnik 1986.
[9] CCIR International Radio Consultive Commitee, "Report 1008 Reflection
from the Surface of the Earth," International Telecommunications Union,
Dubrovnik 1986.
[10] G. J. Broadbent, "Tidal Readings - Bongaree Jetty," Maritime Services
Branch, Queensland Transport Maritime Division 1999.
![Page 263: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/263.jpg)
238
THIS PAGE IS INTENTIONALLY BLANK
![Page 264: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/264.jpg)
239
Chapter 9 Receiver Correlation and Discrimination
In the previous chapters the FSPE and PETA models have been shown to provide a
novel and worthwhile simulation environment, that provides not only a unique
environment for visualisation but also all of the necessary defining parameters for
describing multipath propagation within the simulation domain. In this chapter the
value of incorporating a GPS receiver model with the propagation modelling
environment is investigated.
9.1 Introduction
The advantage of the FSPE/PETA modelling method, as shown in previous chapters,
is that the exact multipath nature of a complicated environment can be understood and
decomposed. By combining a receiver model with the propagation models a complete
software-based satellite to user modelling system is developed. A block diagram of
this system is shown in Figure 9.1
Satellite
Signal
FSPE/PETAGPS Propagation
Model
RXModel
RXParameters
MultipathParameters
MultipathAnalysis
Environment
MultipathParameters
ErrorParametersMitigation
Figure 9.1 — Multipath Modelling Environment
![Page 265: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/265.jpg)
240
No assumptions will be made regarding the nature of the input signal to the receiver,
the only inputs to the system will be the receiver parameters (discrimination type, chip
spacing, bandwidth, sampling), satellite parameters (elevation, azimuth, code), signal
parameters (frequency, power) and the environmental parameters (terrain
characteristics).
9.2 Fundamentals of a GPS Receiver Model
A typical GPS user equipment configuration is shown in Figure 9.2. The separation of
the functional blocks, the GPS Receiver, and the Application Processing is based on
the premise that not all GPS receivers perform navigation processing. The application
may be that of time transfer, data collection or differential surveying. For this reason it
is appropriate to consider that the user equipment performs two separate functions.
Firstly, tracking of received signals, utilising some form of Phase or Frequency Lock
Loop (PLL/FLL) for carrier recovery and a Delay Lock Loop (DLL) for code
recovery, and secondly the processing of the observables provided by the receiver, to
provide application-specific requirements.
GPS Receiver
Application Processing
pseudoranges carrier phase nav data
position velocity
User Clock
GPSSatellite Signals
Observables
Figure 9.2 — General form of GPS user equipment
The following sections will look more closely at the operation of the GPS receiver,
the technological aspects and their implementations.
![Page 266: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/266.jpg)
241
9.2.1 Generic GPS Receiver Functions
The system level functions of a GPS receiver vary, in only a few aspects, from the
fundamental functions performed by many modern RF-based communications
systems. A generic GPS receiver would consist of the following functional elements
as shown in Figure 9.3: antenna, Radio Frequency (RF) amplification, reference
oscillator, frequency synthesis, down-conversion, Intermediate Frequency (IF) section,
and signal processing.
RFSection
DownConverter
FrequencySynthesiser
ReferenceOscillator
IFSection
SignalProcessing
ClocksLO
pseudorangesdelta pseudorangeintegrated Doppler
Figure 9.3 — Generic GPS receiver
These functions are described in various excellent GPS-specific sources including [1,
2], and since the first six functions are also common to other RF systems the reader
can find information regarding these topics in this literature. All functions are briefly
described here for completeness, but it is the signal processing function that forms the
core of a GPS receiver and it is this function that is the focus of later sections.
9.2.1.1 Antenna and RF Section
The GPS signals, received from all GPS satellites, are right-hand circularly polarised
(RHCP), and of very low power (-160 dBw) [3]. The antenna and RF sections
therefore need to maximise the signal reception. The antenna needs to have a near
hemispherical gain pattern, so as to maximise the number of satellites for tracking.
![Page 267: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/267.jpg)
242
The antenna may be designed as an active device with some amplification occurring
before the signal is sent to the RF section.
The RF section usually consists of filtering and amplification functions. The filtering
is used to reduce the effects of out-of-band noise and interference. The amplification
is achieved with a low-noise-amplifier (LNA), with the gain selected so as to establish
the designed receiver noise figure.
9.2.1.2 Reference Oscillator and Frequency Synthesis
The reference oscillator provides the time and frequency reference for the GPS
receiver. The parameters of the reference oscillator (size, stability, and phase noise)
are trade-offs between cost and performance. The higher the stability, the more costly
the oscillator becomes. In GPS receivers the reference oscillator is used by the
frequency synthesiser to generate all Local Oscillators (LO) and clocks.
9.2.1.3 Down-conversion and IF
The local oscillators are used in the downconverter to convert the RF signal down to a
lower Intermediate Frequency (IF). The IF section provides additional amplification,
filtering and provides a conditioned signal that can be used by the signal processing
section.
9.2.1.4 Signal Processing
The signal processing section is really the core function in a GPS receiver. It performs
multi-channel acquisition and carrier/code tracking of satellites, navigation data
demodulation, code-phase (pseudorange), carrier-phase (delta pseudorange) and
carrier frequency (integrated Doppler) measurements, in addition to extracting the
Signal-to-Noise Ratio (SNR) of the received signals.
![Page 268: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/268.jpg)
243
9.3 The Key GPS Receiver Elements
The key elements of a GPS receiver are located in the signal processing section. These
elements provide the ability for the receiver to acquire and track the GPS ranging-
code. There are four fundamental sub-sections necessary.
1. Pseudo-Random Noise (PRN) code generator — all of the possible satellite PRN
codes need to be replicated within the receiver, for in the acquisition and tracking
of the GPS ranging signal. The generation of these codes is achieved with code
generators similar in design to those used on the GPS satellites.
2. Signal Acquisition Process — acquires the satellite signal so that tracking may be
performed.
3. Delay Lock Loop (DLL) — tracks the code-phase of the received signal.
4. Phase or Frequency lock Loop (PLL/FLL) — tracks the carrier-phase of the
received signal.
These elements will be examined in the following sections.
9.3.1 PN Code Generation
The GPS C/A-code is a 1023 bit (chip) Gold code, with chipping rate of 1.023 MHz.
This gives a code with a period of 1 millisecond and a chip width of 977.5
nanoseconds. The receiver PN C/A coder implementation is specified in ICD-GPS-
200 and the GPS SPS Signal Specification. The implementation of a two-tap selection
C/A-code generator is given in Figure 9.4.
![Page 269: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/269.jpg)
244
1 2 3 4 5 6 7 8 9 10
1 2 3 4 5 6 7 8 9 10
PHASE SELECTOR
G1 GENERATOR
G2 GENERATOR
S1
S2
TAP SELECTIONG2i
G1
G2
XGi
÷ 1024
÷ 20
÷ 10
RESETALL
ONES
RESET
RESET
CLOCK
CLOCK
10.23 Mbps
X1 EPOCH
C/A CODE
G EPOCH
DATA CLOCK
50 bps
1 kbps
SYNCSYNC
+
+
+
+
Figure 9.4 — C/A Code Generation
Following this implementation it is seen that the C/A-code is the Modulo-2 sum of
two 1023 chip linear sequences, G1 and G2i. Here G2i is a delayed G2 sequence, with the
effective delay selected by the Modulo-2 sum of two phases from the G2 shift register.
The G1 and G2 sequences are generated by 10-stage shift registers initialised with the
vector 1111111111, and specified by the following polynomials
13101 ++= XXG and
123689102 ++++++= XXXXXXG
The specified code phase assignments for the selection of the G2i sequence, provide a
total of thirty-six codes for GPS use. The code phase assignments, as a function of the
GPS PRN code number, are summarised in Table 9.1.
![Page 270: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/270.jpg)
245
PRN Signal Number (N) Code Phase Selection
1, 2, 3, 4 ( ) ( )51 +⊕+ NN
5, 6 ( ) ( )44 +⊕− NN
7, 8, 9 ( ) ( )16 +⊕− NN
10, 11, 12, 13, 14, 15, 16 ( ) ( )78 −⊕− NN
17, 18, 19, 20, 21, 22 ( ) ( )1316 −⊕− NN
23 ( ) ( )2022 −⊕− NN
24, 25, 26, 27, 28 ( ) ( )1820 −⊕− NN
29, 30, 31, 32, 33 ( ) ( )2328 −⊕− NN
34 ( ) ( )2430 −⊕− NN
35, 36 ( ) ( )2834 −⊕− NN
Table 9.1 — C/A-Code Selection
9.3.2 Delay Lock Loop
GPS signal acquisition and tracking is a two-dimensional process. The receiver must
correlate the received code with a shifted internally generated code replica as well as
maintain lock with the carrier of the signal.
In the code-phase dimension, when the code-phases are matched there is maximum
correlation. Minimum correlation occurs when the replica code is offset by more than
one chip. The C/A-code autocorrelation function is given in Equation (9.1),
![Page 271: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/271.jpg)
246
( ) ( ) ( )RG
CAG
itt
t Gi
t dtτ τ= ==∫ +
1
1023 01023
Τ(9.1)
and the resultant plot of the autocorrelation function is shown in Figure 9.5.
0 1977.5 x 10 6
-1/1023
A2 RG(t)
t (ms)
Figure 9.5 — C/A autocorrelation function
The code-phase tracking loop is of a similar form to that used in other Direct
Sequence Spread Spectrum (DSSS) systems—a Delay Lock Loop (DLL). The DLL
used can be either coherent or noncoherent. The coherent DLL requires parallel
tracking of the carrier-phase whilst the noncoherent DLL requires no carrier tracking.
The form of the DLL is determined by the type of discriminator used.
In addition to detecting the signal in the code-phase dimension, the receiver must also
detect the signal in the carrier-phase dimension. The tracking of the carrier-phase can
be achieved by a variety of tracking loops that are of either PLL, Costas PLL, or FLL
form. In each of these forms, there is a variety of discriminator functions that can be
utilised. A generic form of the carrier and code tracking loops is shown in Figure 9.6.
![Page 272: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/272.jpg)
247
PN CodeGenerator
Accumulator
Accumulator
Accumulator
Accumulator
Accumulator
Accumulator
Early Prompt LateCarrierNCO
SINMap
COSMap
Carrier LoopDiscriminator
Carrier LoopFilter
CodeNCO
Code LoopFilter
Code LoopDiscriminator
carrier aiding
I
Q
E
P
L
QE
E
P
L
QL
QP
IE
IP
IL
IPQP
Figure 9.6 — Receiver tracking loops
The input is digital IF, and the carrier is stripped or wiped-off by mixing with a
replica carrier (plus carrier Doppler, to account for relative receiver dynamics). The
outputs are in-phase (I) and quadrature-phase (Q) samples. This signal is then
collapsed back to baseband by code stripping using early/late correlators. The
baseband signal is then processed by some form of discriminator. The code and carrier
loops are controlled by Numerically Controlled Oscillators (NCO).
The early/late correlators provide the input to the discriminator, which is implemented
by one of several different early minus late forms. The code correlation process is
illustrated in Figure 9.7.
![Page 273: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/273.jpg)
248
-1/2-1 0 +1E
P
L
Chip offset
Normalised correlator output
-1/2-1 0 +1
E
P
L
Chip offset
Received code
Rx generated codes
Early
Prompt
Late
Figure 9.7 — Correlation process
This shows how the early, prompt, and late correlation envelopes vary, for one chip of
the C/A-code, as the phase of the replica code is varied. The early minus late
discrimination of these envelopes produces a discrimination characteristic (tracking
error) that controls the closed loop operation of the DLL. Some common DLL
discrimination algorithms [4, 5] are given in Table 9.2.
![Page 274: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/274.jpg)
249
Discriminator Algorithm Description
( ) ( )I I I Q Q QE L P E L P− + −∑ ∑ Dot Product Power — noncoherent
( ) ( )∑∑ +−+ 2222LLEE QIQI Early - Late Power — noncoherent
( ) ( )∑∑ +−+ 2222LLEE QIQI Early - Late Envelope — noncoherent
( ) ( )( ) ( )∑∑
∑∑+++
+−+2222
2222
LLEE
LLEE
QIQI
QIQI Normalised Early - Late Envelope —
noncoherent
sign IP( ) IE − IL( )∑ Early - Late — coherent
Table 9.2 — DLL discriminator algorithms
Each consecutive noncoherent algorithm increases in computational load, with the
dot-product having the lowest, and the normalised envelope the highest. The coherent
algorithm is included here but its performance is marginal at low SNR and is
obviously unusable when carrier tracking is not possible. The noncoherent DLL
implementations are the most robust. The discriminator characteristic (also known as
the s-curve) for the envelope and normalised envelope discriminators are shown in
Figure 9.8.
![Page 275: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/275.jpg)
250
-0.5-1.5
+1.5+0.5ε/T - chips
D(ε) - chips
-1.0
+1.0
+1.0
-1.0
envelope
normalised envelope
Figure 9.8 — DLL discriminator curves
The discriminator output is the relative tracking error of the DLL and is filtered and
applied to the code NCO, which in turn provides the necessary adjustment to the code
generator phase for correct code-phase alignment of the replica and received codes.
For the carrier tracking loop there are, again, several implementations that may be
used. If the receiver does not need to demodulate the navigation message (pure phase
tracking) then a pure PLL implementation can be used. Since however, most receivers
will need to demodulate this data, a different PLL (Costas PLL) is often used. Other
implementations may make use of Frequency Lock Loops (FLL). The FLL achieves
carrier wipe-off by replicating the approximate frequency of the carrier.
The common algorithms [4, 5], for the true PLL and Costas PLL implementations, are
presented in Table 9.3.
![Page 276: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/276.jpg)
251
PLL Discriminator
Algorithm
Phase Error Description
( )sign I QP P⋅ sinφ Decision-Directed PLL/Costas
I QP P⋅ sin2φ Generic PLL/Costas
Q
IP
P
tanφ Tangent PLL/Costas
TANQ
IP
P
−
1
φ Arctangent — two-quadrant (Costas)
four-quadrant (PLL)
Table 9.3 — PLL discriminator algorithms
The performance of these discriminator algorithms is different under variable SNR
conditions. The decision-directed, generic, and tangent discriminator algorithms are
identical for the pure PLL and Costas PLL implementations. The decision-directed
algorithm is near optimal for high SNR and has the least computational burden. The
Generic algorithm is near optimal at low SNR with moderate computational burden,
while the tangent algorithm is suboptimal, but performs reasonable well at high and
low SNR, with high computational burden. In addition, the tangent algorithm must
check for divide by zero errors near ± 90 degrees. The final algorithm—arctangent—
is an optimal Maximum Likelihood (ML) estimator at high and low SNR. In the pure
PLL implementation it is a four-quadrant arctangent, while for the Costas PLL, it is
two-quadrant.
The common FLL discriminator algorithms [5] are given in Table 9.4.
![Page 277: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/277.jpg)
252
FLL Discriminator Algorithm Frequency
Error
Description
( ) ( )sign I I Q Q I Q I Q
t tP P P P P P P P1 2 1 2 1 2 2 1
2 1
⋅ + ⋅ ⋅ ⋅ − ⋅−
( )[ ]sin 2 2 1
2 1
φ φ−−t t
Sign(dot)
cross
I Q I Q
t tP P P P1 2 2 1
2 1
⋅ − ⋅−
( )sin φ φ2 1
2 1
−−t t
Cross
( ) ( )[ ]( )
tan I I Q Q I Q I Q
t t
P P P P P P P P− ⋅ + ⋅ ⋅ − ⋅
−
11 2 1 2 1 2 2 1
2 1 360
,( )
φ φ2 1
2 1 360
−−t t
Arctangent
cross, dot
Table 9.4 — FLL discriminator algorithms
9.4 A Receiver Model for use with the FSPE/PETA Model
The aspects of the GPS receiver, covered in the preceding sections, are sufficient for
the development and implementation of a GPS receiver model. Models that have been
developed for the characterisation of the effects of multipath propagation, include
those of Braasch [6], van Nee [7], and others [8-21]. These models were developed
using the well defined fundamentals of the GPS receiver; signal down-conversion via
mixing, correlation, and discrimination. In this work, the primary focus is on the
investigation and characterisation of multipath effects on the code-phase tracking of
the ranging signal. In this context it is sufficient that the receiver model is developed
around a variable, generic DLL implementation, shown in Figure 9.9. The MATLAB
implementation can be found in Appendix C.
![Page 278: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/278.jpg)
253
CHANNELC/A-CODE
GEN
N Multipath signal parameters[relative amplitudes, delays, phases]
+
RECEIVERC/A-CODE
GEN
EARLYCORRELATOR
PROMPTCORRELATOR
LATECORRELATOR
PE L
[Correlator spacing]
LOW-PASS
FILTER
[Cut-off frequency]
DISCRIMINATOR
[Discriminator select]
Range error
N
1
LOS
Figure 9.9 — Receiver Correlation Model
The resultant correlation and discrimination allows the visualisation of these functions
for any multipath scenarios produced by FSPE/PETA model-based simulations. The
various discrimination functions as outlined above were implemented in the receiver
model and are shown below, Figure 9.10.
-1.5 -1 -0.5 0 0.5 1 1.5-1.5
-1
-0.5
0
0.5
1
1.5DLL Discrimination Functions
Nor
mal
ised
Dis
crim
inat
ion
Val
ue
Chip Offset
E-L Norm E-L Env E-L Power Dot Product
Figure 9.10 — DLL discriminator curves
![Page 279: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/279.jpg)
254
These discrimination functions are as expected for the various implementation types,
and it can be assumed that the correlation and discrimination functions are correctly
implemented within the model.
9.5 The Modelled Code Correlation Function and Multipath
Having established a DLL model that provides visualisation and calculation of
correlation and discrimination functions (and hence the range error), we can consider
the actual effect that multipath propagation has on the correlation function shape and
the consequential tracking error generated by the receiver discrimination function. In
general the correlation function is distorted, but with use of these modelling
techniques a unique visualisation of the corruption of the signal can be made. The
visualisation is a valuable aid in increasing our understanding of multipath effects and
how they may be mitigated.
9.5.1 Variation of Relative Multipath Delay Time
The first modelling situation presented is the case of variation in relative time delay of
a fixed amplitude single multipath signal. Firstly we consider the two limiting cases of
multipath error effect on the correlation function shape. That is; when the multipath
signal is completely in phase (zero degrees phase difference) with the LOS signal and
when it is 180 degrees out of phase with the LOS signal. These two cases; the in-
phase and anti-phase, are presented below.
9.5.1.1 In-Phase Case
In this first series of correlation and discrimination plots for the in-phase case, the
relative multipath-to-LOS amplitude is one. This implies that the multipath signal is
of equal amplitude to the LOS signal. The correlator modelled is a standard ½-chip
correlator and infinite bandwidth is selected (no pre-correlation filtering), this is done
![Page 280: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/280.jpg)
255
to provide clarity in the plots. The first plot, Figure 9.11, is the multipath-free
correlation of the LOS signal and the resultant discrimination function for this
condition.
Figure 9.11 — Multipath-free Correlation and Discrimination
As we see the discrimination or tracking error is zero for this case as is expected. In
the next plot, Figure 9.12, a single multipath signal is present as well as the LOS
signal.
![Page 281: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/281.jpg)
256
Figure 9.12 — 0.05 Chip delay In-phase Multipath
The total correlation function of the combined multipath and LOS signal is shown to
have increased from the previous normalised value of one. Although the
discrimination function shows very little distortion there is in fact a fractional offset
of the zero-crossing point from the zero chip offset. In other words there is a tracking
error whenever the discrimination function has a non-zero-offset for a prompt arrival
of the LOS signal.
We now increase the in-phase multipath relative delay and observe that the distortion
of the discrimination function has increased dramatically, Figure 9.13.
![Page 282: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/282.jpg)
257
Figure 9.13 — 0.5 chip delay In-phase Multipath
The range offset that the DLL tracks is the zero-crossing offset which in this case is a
positive offset. This infers that the range estimate for this case is larger than the true
range. We now consider the correlation function and describe the distortion points
caused by in-phase multipath.
-2000 -1500 -1000 -500 0 500 1000 1500 2000-0.5
0
0.5
1
1.5
2
2.5x 10
4 Correlation Function
Time Offset (ns)
Cod
e C
orre
latio
n
Delayed Prompt Combined
Figure 9.14 — Correlation Distortion (0°)
C
A
B
D
E
F
![Page 283: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/283.jpg)
258
Consider the figure of a multipath corrupted correlation function shown above, Figure
9.14. The following points are observed and correspond to those indicated on the
figure:
A. The undistorted leading portion of the combined correlation function has the same
gradient as the LOS correlation function.
B. The gradient of the combined correlation function above the multipath breakpoint
(the breakpoint occurs at the point where the multipath signal arrives after the
LOS) is the linear combination of the LOS and multipath correlation functions.
C. The actual peak of the correlation is not displaced or distorted (for the infinite
bandwidth case).
D. The second breakpoint occurs here. Note that the gradient immediately following
this point is substantial different to that just prior to the peak of the correlation
function.
E. The absolute value of the gradient on the rear side of the combined correlation
function is the same value as the absolute of the leading edge gradient (B). If these
gradients are extended to intercept then the displacement of the tracked peak
becomes obvious.
F. The last position of the combined correlation function is undistorted after the third
and final breakpoint.
9.5.1.2 Anti-phase Case
In the plot of correlation and discrimination given below in Figure 9.15, the combined
signal is made up of a prompt LOS and a very short delayed multipath signal that
arrives out-of-phase with a relative amplitude of 0.5. That is the multipath has power
half that of the LOS signal.
![Page 284: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/284.jpg)
259
Figure 9.15 — Short-delay Multipath (180°)
As for the in-phase case we note that the short-delay multipath has very little apparent
effect on the discrimination and hence the tracking or zero-crossing point.
The true distortion becomes apparent as the relative multipath delay is increased,
Figure 9.16.
![Page 285: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/285.jpg)
260
Figure 9.16 — Long-delay Multipath (180°)
As for the in-phase case there is an offset of the zero-crossing point. For the anti-
phase multipath the offset is negative and the range estimate from the tracking loop
will be shorter than the true range. Again we consider this correlation distortion.
-2000 -1500 -1000 -500 0 500 1000 1500 2000-1
-0.5
0
0.5
1
1.5
2x 10
4 Correlation Function
Time Offset (ns)
Cod
e C
orre
latio
n
Delayed Prompt Combined
Figure 9.17 — Correlation Distortion (180°)
A
C
B
D
E
F
![Page 286: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/286.jpg)
261
Consider Figure 9.17, which represents the anti-phase case for a single multipath
delayed relative to the line-of-sight by approximately 100 nanoseconds, and at a
relative amplitude of 0.5. The observations are restricted to the shape of the
correlation function and do not include any mathematical interpretation of the
function, other than geometric. Firstly the relative delay of the multipath signal is
represented at the breakpoints on the leading and trailing slopes, this is also true of the
in-phase case. Indeed if it is possible to track the breakpoints (A and B, or C and D, or
E and F) then the relative delay could be inferred.
These breakpoints occur as singular transitions in the slope of the function. The
function includes six breakpoints (A-F) within the 4-delta spacing used in the figure.
Obviously point C represents the peak correlation point, and we note again that the
multipath does not alter the location of this peak.
9.5.2 Variation of Relative Phase
We now consider the distortion of the correlation function and the resultant
discrimination function for variation in the relative phase of the multipath signal. In
in-phase and anti-phase limiting cases were examined above. Consider the case of a
multipath signal arriving 200 nanoseconds after the line-of-sight signal from the GPS
satellite and at ½ of the LOS signal power.
![Page 287: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/287.jpg)
262
-3 -2 -1 0 1 2 3-2
-1
0
1
2Correlation Function
Nor
mal
ised
Cor
rela
tion
LOSMP P
-3 -2 -1 0 1 2 3-2
-1
0
1
2E-L Envelope Discriminator
Offset (chips)
Nor
mal
ised
Dis
crim
inat
ion
Figure 9.18 — 200ns relative delay , 0.5 Relative Multipath Ratio (0°)
In Figure 9.18, shown above, the multipath arrives in-phase relative to the LOS signal.
There is a small positive offset in the discrimination function as is expected. We now
retard the relative phase to 60 degrees relative to the LOS, Figure 9.19.
-3 -2 -1 0 1 2 3-2
-1
0
1
2Correlation Function
Nor
mal
ised
Cor
rela
tion
LOSMP P
-3 -2 -1 0 1 2 3-2
-1
0
1
2E-L Envelope Discriminator
Offset (chips)
Nor
mal
ised
Dis
crim
inat
ion
Figure 9.19 — 200ns relative delay, 0.5 Relative Multipath Ratio (60°)
![Page 288: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/288.jpg)
263
Note the effect on the multipath correlation function is a reduction in the correlation
value. This is the same effect expected if the relative amplitude of the multipath is
also reduced without altering the relative phase relationship between the multipath
and the LOS signal. The reduction in the multipath correlation reduces the distortion
of the combined correlation and hence the tracking error of the discrimination
function is also reduced. The chosen delay of 200 ns represents a delay that would
occur in forward scatter for an antenna height above a reflector of 30 m, with the
satellite at zenith. It is also representative of an antenna distance of 30 m from a back
reflector at a satellite angle of zero degrees.
Further change in the relative phase relationships highlights the reduction of
correlation relative to the phase relationship of the multipath and LOS signal, Figure
9.20.
-3 -2 -1 0 1 2 30
0.5
1
1.5
2Correlation Function
Nor
mal
ised
Cor
rela
tion
LOSMP P
-3 -2 -1 0 1 2 3-2
-1
0
1
2E-L Envelope Discriminator
Offset (chips)
Nor
mal
ised
Dis
crim
inat
ion
Figure 9.20 — 200ns relative delay, 0.5 Relative Multipath Ratio (90°)
Here the multipath signal is in quadrature phase with the LOS signal and there is zero
correlation of the multipath signal. Further retardation of the relative phase leads to a
relative change of the correlation function, which is equivalent to a anti-phase
multipath signal arriving at reduced amplitude, Figure 9.21.
![Page 289: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/289.jpg)
264
-3 -2 -1 0 1 2 3-2
-1
0
1
2Correlation Function
Nor
mal
ised
Cor
rela
tion
LOSMP P
-3 -2 -1 0 1 2 3-2
-1
0
1
2E-L Envelope Discriminator
Offset (chips)
Nor
mal
ised
Dis
crim
inat
ion
Figure 9.21 — 200ns relative delay, 0.5 Relative Multipath Ratio (125°)
Finally we arrive at the 180 degree relative phase retardation or the anti-phase case,
where our multipath signal is in full negative correlation, Figure 9.22.
-3 -2 -1 0 1 2 3-2
-1
0
1
2Correlation Function
Nor
mal
ised
Cor
rela
tion
LOSMP P
-3 -2 -1 0 1 2 3-2
-1
0
1
2E-L Envelope Discriminator
Offset (chips)
Nor
mal
ised
Dis
crim
inat
ion
Figure 9.22 — 200ns relative delay, 0.5 Relative Multipath Ratio (180°)
![Page 290: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/290.jpg)
265
9.6 Error Envelopes for a Single multipath Signal
We now consider the examples given above for a single multipath signal, over a range
of variables for the limiting cases (i.e. in-phase and anti-phase). This includes the
variation over the complete range of correlated delays, relative multipath signal
strength, and phase for a fixed delay.
9.6.1 Variation of Relative Time Delay
In this section we validate the receiver model by generating known results from
previous work, notably the error curves for GPS receivers developed by Braasch [22,
23].
0 200 400 600 800 1000 1200 1400 1600-80
-60
-40
-20
0
20
40
60
80C/A-code; 0.5 MP/LOS; 1/2-chip Dot Product
Relative MP Delay (ns)
Pse
udor
ange
Err
or (
m)
Figure 9.23 — ½-Chip 0.5 Relative Multipath Ratio
We have not generated every possible combination for the relative multipath ratio
value, but have selected 0.5 relative multipath ratio as a representative case. In Figure
9.23 we have the range error envelope (0° and 180° limiting cases) for an infinite pre-
![Page 291: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/291.jpg)
266
correlation bandwidth (remembering that the DLL model developed allows any pre-
correlation bandwidth to be selected) for a standard ½ chip correlator spaced DLL.
0 200 400 600 800 1000 1200 1400 1600-8
-6
-4
-2
0
2
4
6
8C/A-code; 0.5 MP/LOS; 1/20-chip Dot Product
Relative MP Delay (ns)
Pse
udor
ange
Err
or (
m)
Figure 9.24 — 1/20 Chip 0.5 Relative Multipath Ratio
The equivalent range error envelope for a narrow correlation DLL is given above in
Figure 9.24. Again we note that the model developed allows any correlation sampling
spacing (relative to the sampling resolution) to be selected. We now incorporate pre-
correlation filtering into the model.
![Page 292: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/292.jpg)
267
0 200 400 600 800 1000 1200 1400 1600-80
-60
-40
-20
0
20
40
60
80C/A-code; 0.5 MP/LOS; 1/2-chip Dot Product (2Mhz)
Relative MP Delay (ns)
Pse
udor
ange
Err
or (
m)
Figure 9.25 — ½-Chip 0.5 Relative Multipath Ratio (2MHz)
The pre-correlation bandwidth for the standard ½-chip correlation spacing is usually 2
Megahertz for the most commercially available receivers [1]. In Figure 9.25 above we
have selected this bandwidth and we see the resultant variation of the range error
envelope. Figure 9.26 shows the equivalent case for the accepted narrow correlation
design, where the pre-correlation bandwidth is 8 Megahertz [24].
0 200 400 600 800 1000 1200 1400 1600-8
-6
-4
-2
0
2
4
6
8C/A-code; 0.5 MP/LOS; 1/20-chip Dot Product (8Mhz)
Relative MP Delay (ns)
Pse
udor
ange
Err
or (
m)
Figure 9.26 — 1/20-Chip 0.5 Relative Multipath Ratio (8MHz)
![Page 293: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/293.jpg)
268
In all of the above examples the results of the implemented correlation/discrimination
model are similar to the results presented in a variety of published sources see [22, 25-
28] for examples.
9.6.2 Variation of Relative Amplitude
We now consider the variation of relative amplitude of a single multipath signal over
a range of relative delays.
Figure 9.27 — Variation of Relative Multipath Ratio ½-Chip (0°)
![Page 294: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/294.jpg)
269
Figure 9.28 — Variation of Relative Multipath Ratio ½-Chip (180°)
Figure 9.27 and Figure 9.28 show the level of complex information that can be
derived from the developed DLL model. Here the X-axis is relative time delay of the
multipath signal with respect to the line-of-sight signal, and the Y-axis is the relative
amplitude. We see that for fixed delays the range error varies in a non-linear manner,
although the range errors at fixed amplitudes are piece-wise linear in form.
9.6.3 Variation of Phase
In the cases given above we have considered the range error for the limiting cases of
in-phase and anti-phase multipath signals. The variation of relative phase was shown
to distort the discrimination function in a non-linear manner.
![Page 295: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/295.jpg)
270
0 20 40 60 80 100 120 140 160 180-60
-50
-40
-30
-20
-10
0
10
20C/A-Code 1:180 degs,0.5MPR, 200ns
Relative Code Phase (degs)
Ran
ge E
rror
(m
)
Figure 9.29 — Range Error 0.5 Relative Multipath Ratio (200 ns 0°-180°)
As an example of the nature of this variation we present the case of phase variation,
from zero to 180 degrees phase retardation, for a fixed 200 nanosecond relative delay
half-power multipath signal, Figure 9.29. This example shows the wideband nature of
the range error (note the non-sinusoidal shape) found by Braasch in his research [22].
9.7 Error Envelopes for Two Multipath Signals
In the above section we considered the effect of the relative parameters that define
multipath on the receiver and the corresponding range error for a single multipath
signal. In real-world situations the effects of multiple multipath signals on the receiver
need to be considered [29].
The model developed in this work is not limited in the number of multipath signals
that may be present with the LOS signal. As an example of this multiple multipath
capability we present the limiting cases (range error envelopes) for two multipath
signals, where both multipath signals are half-power relative to the LOS signal.
![Page 296: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/296.jpg)
271
For all of the following figures the X-axis is relative delay in nanoseconds of first
multipath (MP1) with respect to the LOS signal and the Y-axis is the relative delay in
nanoseconds of the second multipath (MP2) with respect to the first multipath signal
(MP1). Both multipath signals are 0.5 relative amplitude with respect to the LOS
signal.
Figure 9.30 — Dual Multipath Error (0 ° and 0°)
![Page 297: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/297.jpg)
272
Figure 9.31 — Dual Multipath Error (0 ° and 180°)
Figure 9.32 — Dual Multipath Error (180° and 0°)
![Page 298: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/298.jpg)
273
Figure 9.33 — Dual Multipath Error (180 ° and 180°)
These are limiting cases for a standard ½-chip correlator of infinite bandwidth for
C/A-code PRN1, and infer the envelope of the range error for stated phase
relationships. Occurrences of these limiting phase relationships are as follows
(assuming zero reference phase for LOS and 180 degree phase shift upon reflection of
each multipath signal):
1. MP1 and MP2 in-phase with LOS: when relative delay of MP1 with-respect-to
LOS is 0.3174ns (1/2f) or some multiple of 0.6348ns (1/f) from this initial delay,
and relative delay of MP2 writ MP1 is some multiple of 0.6348ns (1/f).
2. MP1 in-phase and MP2 anti-phase with LOS: when relative delay of MP1 wrt
LOS is 0.3174ns (1/2f) or some multiple of 0.6348ns (1/f) from this initial delay,
and relative delay of MP2 with-respect-to MP1 is some odd multiple of 0.3174ns
(1/2f).
3. MP1 anti-phase and MP2 in-phase with LOS: when relative delay of MP1 with-
respect-to LOS is 0.6348ns (1/f) or some multiple of 0.6348ns (1/f) from this
initial delay, and relative delay of MP2 with-respect-to MP1 is some odd multiple
of 0.3174ns (1/2f).
![Page 299: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/299.jpg)
274
4. MP1 anti-phase and MP2 anti-phase with LOS: when relative delay of MP1 with-
respect-to LOS is 0.6348ns (1/f) or some multiple of 0.6348ns (1/f) from this
initial delay, and relative delay of MP2 with-respect-to MP1 is some multiple of
0.6348ns (1/f).
These relationships are evident when a plot of signal strength for the single multipath
case is made, Figure 9.34.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2-8
-6
-4
-2
0
2
4in-phase
anti-phase
S/N Single 0.5 MPR Multipath
S/N
(db
)
Relative delay (ns)
Figure 9.34 — S/N Fade Pattern
In this figure the signal is the addition of the normalised LOS with a single 0.5
(relative multipath ratio) multipath, with 180 degree phase shift upon reflection. The
peaks represent the case when both signals are in-phase and the fades indicate anti-
phase relationships.
For the three-dimensional plots of range error, the true behaviour of the error within
the envelopes is such that the plots would be of such complexity as to render them
impossible to interpret. The complete plot of range error would involve the transition
of the error from one state to any other state of the three remaining limiting cases. In
addition there are two other intermediate states; where the relative phase of either or
both multipath results in zero contribution to the range error. This occurs when the
multipath signal is in quadrature-phase with the LOS. For this case the multipath
![Page 300: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/300.jpg)
275
correlation is zero and the range error bounds to a single multipath occurrence. This
bound is also met when the second multipath is effectively decorrelated at delays
exceeding 1500ns from the LOS (relative delay of MP1+ relative delay of MP2 =
1500ns). For the limiting cases these zones of decorrelation of the second multipath
signal are clearly evident in the figures given above (Figure 9.30-Figure 9.33).
9.8 Summary
In this chapter the developed correlation/discrimination model has been shown to
provide complete control of multipath variables. The resultant multipath errors are
given directly and any combination of multipath signals is possible.
The implementation of the model is not limited by the need to solve iterative
equations and is very closely modelled on the physical implementation that would be
seen in a modern GPS receiver.
The model provides the ability to provide complete visualisation of the correlation,
discrimination functions and the resultant range error results.
The model presented here can utilise directly the resultant output of the PETA
propagation model and as such provides the ability to generate realistic interpretations
of resultant effects of multipath propagation on a GPS receiver with selected
particular parameters.
Indeed it is the logical next step, in future research work, that the modelling is
combined fully, and that a comparison of measured and modelled data be made for a
large variety of propagation environments.
![Page 301: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/301.jpg)
276
9.9 References
[1] B. W. Parkinson and J. R. Spilker Jr., "Global Positioning System: Theory and
Applications Volume I," in Progress in Astronautics and Aeronautics, vol.
163, P. Zarchen, Ed. Washington: American Institute of Aeronautics and
Astronautics, 1996.
[2] E. D. Kaplan, "Understanding GPS: Principles and Applications," in Mobile
Communications Series, J. Walker, Ed. Boston: Artech House, 1996.
[3] NAVSTAR Joint Program Office, "Technical Characteristics of the
NAVSTAR GPS," 1991.
[4] A. J. van Dierendonck, "GPS Receivers," in Global Positioning System:
Theory and Applications Volume I, vol. 163, Progress in Astronautics and
Aeronautics, B. W. Parkinson and J. R. Spilker Jr., Eds. Washington:
American Institute of Aeronautics and Astronautics, 1996, pp. 329-407.
[5] P. W. Ward, "Satellite Signal Aquisition and Tracking," in Understanding
GPS: Principles and Applications, Mobile Communications Series, E. D.
Kaplan, Ed. Boston: Artech House, 1996.
[6] M. S. Braasch, "GPS Multipath Model Validation," presented at IEEE 1996
Position and Location Symposium, Atlanta, Georgia, 1996.
[7] R. D. J. van Nee, "Multipath and Multi-Transmitter Interference in Spread-
Spectrum Communication and Navigation Systems," in Faculty of Electrical
Engineering, Telecommuncation and Traffic Control Systems Group. Delft:
Delft University of Technology, 1995, pp. 205.
[8] G. D. Akrivis, V. A. Dougalis, and N. A. Kampanis, "Error Estimates for
Finite Element Methods for a Wide-Angle Parabolic Equation," Applied
Numerical Mathematics, vol. 16, pp. 81-100, 1994.
[9] J. R. Auton and J. Cruz, "Simulating GPS Receiver Measurement Errors,"
presented at The 9th International Technical Meeting of The Satellite Division
of The Institute of Navigation., Kansas City, Missouri, 1996.
[10] P. Axelrad, C. J. Comp, and P. F. MacDoran, "SNR-Based Multipath Error
Correction for GPS Differential Phase," IEEE Transactions on Aerospace and
Electronic Systems, vol. 32, pp. 650-660, 1996.
![Page 302: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/302.jpg)
277
[11] M. E. Cannon, G. Lachapelle, and G. Lu, "Kinematic Ambiguity Resolution
with High-Precision C/A Code Receiver," Journal of Surveying Engineering
ASCE, vol. 119, pp. 147-155, 1993.
[12] A. J. R. M. Coenen and A. J. de Vos, "FFT-Based Interpolation for Multipath
Detection in GPS/GLONASS Receivers," Electronics Letters, vol. 28, pp.
1787-1788, 1992.
[13] D. Doris and A. Benhallam, "On Correlation Processes Reducing Multipath
Errors in the L1 GPS Receiver," presented at The 9th International Technical
Meeting of The Satellite Division of The Institute of Navigation., Kansas City,
Missouri, 1996.
[14] G. Lachapelle, M. E. Cannon, and G. Lu, "A Comparison of P-Code and High
Performance C/A Code GPS Receivers for on the Fly Ambiguity Resolution,"
Bulletin Geodesique, vol. 67, pp. 185-192, 1993.
[15] G. Lachapelle, M. E. Cannon, G. Lu, and B. Loncarevic, "Shipborne GPS
Attitude Determination During MMST-93," IEEE Journal of Oceanic
Engineering, vol. 21, pp. 100-105, 1996.
[16] W. Lippencott, T. Milligan, and D. Igli, "Method for Calculating Multipath
Environment and Impact on GPS Receiver Solution Accuracy," presented at
ION National Technical Meeting, Santa Monica, California, 1996.
[17] A. Montalvo and A. Brown, "A Comparison of Three Multipath Mitigation
Approaches for GPS Receivers," presented at 8th International Technical
Meeting of The Satellite Division of The Institute of Navigation., Palm
Springs, California, 1995.
[18] G. D. Morley and W. D. Grover, "Improved Location Estimation with Pulse-
Ranging in Presence of Shadowing and Multipath Excess-Delay Effects,"
Electronics Letters, vol. 31, pp. 1609-1610, 1995.
[19] J. Shi and M. E. Cannon, "Critical Error Effects and Analysis in Carrier Phase-
Based Airborne GPS Positioning over Large Areas," Bulletin Geodesique, vol.
69, pp. 261-273, 1995.
[20] B. J. H. van den Brekel and R. D. J. van Nee, "GPS Multipath Mitigation by
Antenna Movements," Electronics Letters, vol. 28, pp. 2286-2288, 1992.
![Page 303: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/303.jpg)
278
[21] W. Zhuang and J. M. Tranquilla, "Effects of Multipath and Antenna on GPS
Observables," IEE Proceedings - Radar, Sonar and Navigation, vol. 142, pp.
267-275, 1995.
[22] M. S. Braasch, "On the Characteristics of Multipath Errors in Satellite-Based
Precision Approach and Landing Systems," in Department of Electrical and
Computer Engineering. Athens: Ohio University, 1992, pp. 203.
[23] M. S. Braasch, "GPS and DGPS Multipath Effects and Modeling," in ION
GPS-95 Tutorial: Navtech Seminars, 1995.
[24] A. J. van Dierendonck, P. Fenton, and T. Ford, "Theory and Performance of
Narrow Correlator Spacing in a GPS Receiver," presented at The Institute of
Navigation National Technical Meeting, San Diego, CA, 1992.
[25] M. S. Braasch, "Multipath Effects," in Global Positioning System: Theory and
Applications, vol. 1, B. W. Parkinson and J. R. Spilker Jr., Eds. Washington:
American Institute of Aeronautics and Astronautics, 1996, pp. 547-568.
[26] L. R. Weill, "GPS Multipath Mitigation by Means of Correlator Reference
Waveform Design," presented at The National Technical Meeting of The
Institute of Navigation., Santa Monica, CA, 1997.
[27] L. R. Weill, "Conquering Multipath: The GPS Accuracy Battle," GPS World,
vol. 8, pp. 59-66, 1997.
[28] L. R. Weill, "Achieving Theoretical Accuracy Limits for Pseudoranging in the
Presence of Multipath," presented at The 8th International Technical Meeting
of The Satellite Division of The Institute of Navigation., Palm Springs, CA,
1995.
[29] C. Macabiau, B. Roturier, E. Chatre, and R. Yazid, "N-Multipath Performance
of GPS Receivers," presented at Position Location and Navigation
Symposium, San Diego, 2000.
![Page 304: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/304.jpg)
279
Chapter 10 Conclusions
Multipath remains a dominant error source in Global Positioning System (GPS)
applications that require high accuracy. Multipath propagation occurs when
environmental features cause combinations of reflected and/or diffracted replica
signals to arrive at the GPS receiving antenna. These signals, in combination with the
original line-of-sight (LOS) signal, can cause distortion of the receiver correlation
function and hence errors in range estimation.
With the use of differential techniques it is possible to remove many of the common-
mode error sources, but the error effects of multipath have proven much more difficult
to mitigate.
The research aim of this work was to enhance the understanding of multipath in GPS
terrestrial applications. This was achieved through the use of novel models of signal
propagation behaviour and its effects. To this end, the work presented in this
dissertation describes the research, development, implementation and validation of:
• a Free-Space Parabolic Equation (FSPE) based propagation model for analysis of
multipath propagation field behaviour, and;
• a PE-based Time Analysis (PETA) model that accurately provides defining
multipath propagation information namely; relative delay , amplitude and phase of
multipath signals in realistic terrestrial propagation environments, and;
• a correlation/discrimination model that processes the results of the PETA (or
derived input) into resultant error effects.
In addition the existing theory of radio frequency propagation, for the GPS L1 signal,
has been united into a coherent treatment of GPS propagation in the terrestrial
environment.
![Page 305: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/305.jpg)
280
The propagation models developed and implemented herein improve on previous PE-
based models by incorporating terrain features, including boundary impedance
properties with a novel coupled circular reflection coefficient, backscattering and
time-domain decomposition of the terrestrial field into a unique multipath impulse
response.
The results provide visualisation as well as the parameters necessary to fully describe
the multipath propagation behaviour. These resultant parameters may be used as input
to the newly developed correlation/discrimination model, for the visualisation and the
generation of resultant error parameters.
Results for a variety of propagation environments were presented and the technique
was shown to provide a deterministic methodology against real GPS data and
accepted solutions. In addition a novel method was introduced for the use of the
undesirable effects of multipath propagation on the received GPS signal for the
determination of relative antenna height.
The comprehensive modelling system developed allows the testing of various
mitigation scenarios such that the effectiveness of each strategy can be ascertained.
The various parameters of the correlation and discrimination model can be adjusted
and tested using real-world propagation input. This flexible approach can also be
extended into the propagation domain where multiple antenna location scenarios can
also be investigated.
The unique and novel combined modelling of multipath propagation and reception,
presented in this dissertation, provides an effective set of tools that can be used to
further the understanding of the behaviour and effect of multipath in GPS
applications, and ultimately should aid in providing a solution to this problem.
![Page 306: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/306.jpg)
281
Chapter 11 Recommendations
The author’s recommendations for further research in the topic area are presented
below.
In the work presented, a novel and unique modelling and simulation system for
terrestrial GPS multipath propagation has been introduced. The concept of a coupled
circular reflection coefficient allows the use of a single boundary condition in the
Free-Space PE and PE-based time analysis models.
The inclusion of an accurate antenna gain model in the FSPE would ultimately
provide improved simulation accuracy for the models and the resultant simulations[1].
Although the author has utilised time-domain information [2] to derive an angle of
arrival for the signals incident upon the antenna — as a means to implement the
antenna gain pattern — it is felt that a more inclusive methodology [3] could be
adopted for future implementations of the FSPE. The gain pattern could, for instance,
be implemented as part of the field image which is propagated in the FSPE. In
addition the LHCP rejection ratio of the antenna should be more correctly modelled as
a function of elevation angle.
An obvious extension to the models presented in this work is the development of
three-dimensional model implementations. The additional information and improved
accuracy brought by three-dimensional modelling may not be considered worthwhile,
in view of the much greater computational load, but nonetheless the question needs to
be asked. The benefit of a three-dimensional implementations [4, 5] over that
provided by the present two-dimensional FSPE/PETA implementation therefore
requires further investigation.
The FSPE and the PETA implementations used in this work were written in
MATLAB. This environment is excellent as a research and development tool but both
models would benefit greatly in terms of speed and performance from a software
implementation such as in C or C++.
![Page 307: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/307.jpg)
282
It is also evident that there is lead-on work in the area of using GPS multipath as an
additional observable. Although not completely novel [6], but considered at similar
times, we saw in Chapter 7 how it is possible to make use of GPS multipath by
manipulation of the received Signal-to-Noise Ratio [7]. A more rigorous and thorough
analysis (which is outside of the scope of this work) is required to extend this concept.
The correlation model developed in Chapter 8 provides the basis for comparison of
modelled receiver errors with measured results. Further research should be made to
provide complete verification and validation of the correlation models presented.
As previously introduced, the concept of temporal variation of correlation samples
and subsequent biasing of the discrimination function requires further investigation.
Finally the complete simulation system developed, which incorporates the
FSPE/PETA GPS multipath propagation models and the DLL receiver model should
be utilised in an extensive investigation of GPS multipath using recorded multipath
measurements from a large number of propagation domains.
![Page 308: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/308.jpg)
283
11.1 References
[1] F. Amoroso and W. W. Jones, "Geometric Model for DSPN Satellite
Reception in the Dense Scatterer Mobile Environment," IEEE Transactions on
Communications, vol. 41, pp. 450-453, 1993.
[2] B. M. Hannah, R. A. Walker, and K. Kubik, "Parabolic Equation-Based Time
Analysis of GPS Multipath Propagation," presented at International
Conference of Spatial Information Science and Technology, Wuhan, China,
1998.
[3] S. U. Hwu, B. P. Lu, R. J. Panneton, and B. A. Bourgeois, "Space Station GPS
Antennas Multipath Analysis," presented at IEEE Antennas and Propagation
Society International Symposium, Newport Beach, California, 1995.
[4] W. M. O’Brien, E. M. Kenny, and P. J. Cullen, "An Efficient Implementation
of a Three-Dimensional Microcell Propagation Tool for Indoor and Outdoor
Urban Environments," IEEE Transactions on Vehicular Technology, vol. 49,
pp. 622-630, 2000.
[5] C. A. Zelley and C. C. Constantinou, "A Three-Dimensional Parabolic
Equation Applied to VHF/UHF Propagation over Irregular Terrain," IEEE
Transactions on Antennas and Propagation, vol. 47, pp. 1586-1596, 1999.
[6] K. D. Anderson, "Determination of Water Level and Tides Using
Interferometric Observations of GPS Signals," Journal of Atmospheric and
Oceanic Technology, vol. 17, pp. 1118–1127, 2000.
[7] B. M. Hannah and R. A. Walker, "Determination of Tide Height Variation
using GPS Multipath," presented at 4th International Symposium on Satellite
Navigation Technology & Applications, Brisbane, Australia, 1999.
![Page 309: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/309.jpg)
284
THIS PAGE IS INTENTIONALLY BLANK
![Page 310: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/310.jpg)
A-1
Appendix A Research Publications
Kubik, K, Hannah, B., Sang, J., Hahn, M., “Attitude Sensing System Using Stereo
Imagery Tracking”, Griffith University Contract Report for Electro Optic Systems
Pty. Ltd. May 1997.
Hannah, B., Walker, R. and Kubik, K., “Towards a Complete Virtual Multipath
Analysis Tool”, Proceedings of The 11th International Technical Meeting of The
Satellite Division of The Institute of Navigation, Nashville TN, USA, September
1998.
Hannah, B., Walker, R. and Kubik, K., “Parabolic Equation-Based Time Analysis of
GPS Multipath Propagation”, International Conference of Spatial Information Science
and Technology, Wuhan, China, December 13-16 1998.
Hannah, B., “Parabolic Equation Research at RAL”, Cooperative Research Report,
British Council, December 1998.
Hannah, B., Walker, R. and Kubik, K., “Electromagnetic Propagation Modelling for
GPS”, SS04, Cooperative Research Centre for Satellite Systems Conference, Paradise
Wirrina Cove Resort, South Australia, CRCSS Technical Memoranda 99/1, 16-19
February, 1999.
Hannah, B. and Walker, R., “Determination of Tide Height Variation using GPS
Signal-to-Noise Ratio”, The 4th International Symposium on Satellite Navigation
Technology and Applications, Brisbane, Australia, 20-23 July 1999.
Hannah, B., “PE-Based Fourier Synthesis Time Analysis of GPS Multipath
Propagation”, Interim PhD report, July 1999.
![Page 311: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/311.jpg)
A-2
Hannah, B., Kubik, K., and Walker, R., “Propagation Modelling of GPS Signals”,
Quo Vadis Geodesia Symposium, Technical Report Nr. 1999.6, Department of
Geodesy and Geoinformatics, Universität Stuttgart, October 1999.
Hannah, B. and Walker, R. “Dual Multipath Error Envelopes for SPS GPS”, CRCSS
Centre Conference 2000, Adelaide, South Australia, CRCSS Technical Memoranda
00/1, 15-17 February, 2000.
Walker, R. and Hannah, B. and Kubik, K., “Deterministic GPS Multipath Mitigation
for Spacecraft Precise Orbit Determination Applications”, SS02, Cooperative
Research Centre for Satellite Systems Conference, Paradise Wirrina Cove Resort,
South Australia, CRCSS Technical Memoranda 99/1, 16-19 February, 1999.
Walker, R., Hannah, B. and Kubik, K. “Multipath Issues in GPS Monitoring”,
International Workshop Proceedings for Advances in Deformation Monitoring, Curtin
University of Technology, Perth Western Australia, 24-25 September 1998. Invited
paper.
Walker, R. and Hannah, B., “Deterministic GPS Multipath Mitigation for Spacecraft
Precise Orbit Determination Applications”, The 4th International Symposium on
Satellite Navigation Technology and Applications, Brisbane, Australia, 20-23 July
1999.
Walker, R. and Hannah, B. “Deterministic GPS Multipath Mitigation for Spacecraft
Applications”, CRCSS Centre Conference 2000, Adelaide, South Australia, CRCSS
Technical Memoranda 00/1, 15-17 February, 2000.
![Page 312: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/312.jpg)
B-1
Appendix B Raw GPS Multipath Data
B.1 Data Results Fresh Water - North Pine Dam 2 December 1999
Antenna Height 1.26 metres.
(a) SV 5
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 5
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 5
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 5
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 5
Ele
vatio
n (d
egs)
![Page 313: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/313.jpg)
B-2
(b) SV 6
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 6
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 6
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 6
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 6
Ele
vatio
n (d
egs)
(c) SV 8
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 8
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 8
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 8
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 8
Ele
vatio
n (d
egs)
![Page 314: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/314.jpg)
B-3
(d) SV 9
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 9
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 8
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 9
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 9
Ele
vatio
n (d
egs)
(e) SV 10
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 10
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 10
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 10
Ele
vatio
n (d
egs)
L1 MULTIPATH N/A
![Page 315: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/315.jpg)
B-4
(f) SV 17
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 17
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 17
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 17
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 17
Ele
vatio
n (d
egs)
(g) SV 24
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 24
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 24
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 24
Ele
vatio
n (d
egs)
L1 MULTIPATH N/A
![Page 316: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/316.jpg)
B-5
(h) SV 25
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 25
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 25
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 25
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 24
Ele
vatio
n (d
egs)
(i) SV 26
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 26
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 26
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 26
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 26
Ele
vatio
n (d
egs)
![Page 317: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/317.jpg)
B-6
(j) SV 30
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
epochs
L1 S/N SV 30
S/N
(dB
)
0 500 1000 1500 2000 2500 3000 3500 4000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 30
Ran
ge E
rror
(m
)
0 500 1000 1500 2000 2500 3000 3500 4000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 30
Azi
mut
h (d
egs)
0 500 1000 1500 2000 2500 3000 3500 40000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 30
Ele
vatio
n (d
egs)
![Page 318: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/318.jpg)
B-7
B.2 Data Results Soil - Caboolture 30 November 1999
Antenna height 1.3 metres.
(a) SV 5
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 5
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 5
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 5
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 5
Ele
vatio
n (d
egs)
![Page 319: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/319.jpg)
B-8
(b) SV 6
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 6
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 6
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 6
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 6
Ele
vatio
n (d
egs)
(c) SV 8
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 8
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 8
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 8
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 8
Ele
vatio
n (d
egs)
![Page 320: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/320.jpg)
B-9
(d) SV 17
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 17
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 17
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 17
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 17
Ele
vatio
n (d
egs)
(e) SV 21
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 21
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 21
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 21
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 21
Ele
vatio
n (d
egs)
![Page 321: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/321.jpg)
B-10
(f) SV 23
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 23
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 23
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 23
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 23
Ele
vatio
n (d
egs)
(g) SV 26
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 26
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 26
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 26
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 26
Ele
vatio
n (d
egs)
![Page 322: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/322.jpg)
B-11
(h) SV 30
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
epochs
L1 S/N SV 30
S/N
(dB
)
0 1000 2000 3000 4000 5000 6000 7000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 30
Ran
ge E
rror
(m
)
0 1000 2000 3000 4000 5000 6000 7000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 30
Azi
mut
h (d
egs)
0 1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 30
Ele
vatio
n (d
egs)
![Page 323: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/323.jpg)
B-12
B.3 Data Results Sea Water - Bribie Island 11 November 1999
Antenna Height 3.4 metres (approx – tidal variation)
(a) SV 3
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 3
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 3
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 3
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 3
Ele
vatio
n (d
egs)
![Page 324: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/324.jpg)
B-13
(b) SV 6
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 6
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 6
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 6
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 6
Ele
vatio
n (d
egs)
(c) SV 10
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 10
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 10
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 10
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 10
Ele
vatio
n (d
egs)
![Page 325: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/325.jpg)
B-14
(d) SV 17
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 17
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 17
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 17
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 17
Ele
vatio
n (d
egs)
(e) SV 21
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 21
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 21
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 21
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 21
Ele
vatio
n (d
egs)
![Page 326: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/326.jpg)
B-15
(f) SV 22
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 22
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 22
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 22
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 22
Ele
vatio
n (d
egs)
(g) SV 23
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 23
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 23
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 23
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 23
Ele
vatio
n (d
egs)
![Page 327: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/327.jpg)
B-16
(h) SV 30
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
epochs
L1 S/N SV 30
S/N
(dB
)
0 2000 4000 6000 8000 10000 12000-5
-4
-3
-2
-1
0
1
2
3
4
5
epochs
L1 Multipath SV 30
Ran
ge E
rror
(m
)
0 2000 4000 6000 8000 10000 12000
-300
-200
-100
0
100
200
300
epochs
Satellite Azimuth SV 30
Azi
mut
h (d
egs)
0 2000 4000 6000 8000 10000 120000
10
20
30
40
50
60
70
80
90
epochs
Satellite Elevation SV 30
Ele
vatio
n (d
egs)
![Page 328: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/328.jpg)
C-1
Appendix C MATLAB Code
C.1 Propagation Modelling Code
(a) GOPE.M
%***************************************************************************% GOPE - control file for FSPE and PETA modelling%%% Bruce Hannah% Version date 14/8/00%***************************************************************************
clear % clear the workspace to start
%***************************************************************************% The following code is an example of using the code with data files% created from data collection exercises modified to MATLAB format by% teqc routines.% When not required code is commented out%***************************************************************************
%sv_num=17; % set SV number%start_epoch=1; % start of data epoch%end_epoch=5000; % end of data epoch%epoch_step=50; % data resolution%epoch_range=start_epoch:end_epoch; % set epoch range%resampled_epoch_range=...% start_epoch:epoch_step:end_epoch; % resample data%cd g:\BMH_data\30_11_98\mat_files; % set data path%load cab_ele % load elevation data%ele_data=sv_data(sv_num,:); % save elevation data%load cab_sn1 % load SNR data%sn_data=sv_data(sv_num,:); % save SNR data%clear sv_data % clear data
%***************************************************************************% start timing of simulation and zero floating point operation counter%***************************************************************************ticflops(0)
%***************************************************************************% global variable declarations%***************************************************************************
global TERRAIN_PROFILE % stores dtm terrain profileglobal TRUEglobal FALSEglobal MHZglobal GHZglobal Cglobal RADIANSglobal DEGREES
![Page 329: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/329.jpg)
C-2
%***************************************************************************% constants%***************************************************************************
TRUE=1; % logical true keywordFALSE=0; % logical false keywordMHZ=1e6; % megahertz constantGHZ=1e9; % gigahertz constantC=3e8; % speed of light in a vacuum (m/s)RADIANS=pi/180; % convert degs to rads (ex: 10*RADIANS)DEGREES=180/pi; % convert rads to degs (ex: pi*DEGREES)
%***************************************************************************% control flags%*************************************************************************** % if setfield_plot=1; % PE field is plottedtime_plot=1; % time series is plottedterrain=0; % terrain is loadedbackscatter=0; % backscatter algorithm is implemented in mpe.mtime_analysis=0; % time analysis routine is implemented in mpe.mini_reflection=1; % initial reflection component includedsv_power=0; % sv power level profile usedfilter_time=0; % a hanning window is applied for time filteringsave_spectrum=0; % angular spectrum saved for later usesave_results=0; % results are saved in MAT filesave_images=0; % images are saved in format specifieddo_fwd_rays=0; % forward rays are plotted on the field plotdo_rev_rays=0; % reverse rays are plotteddo_interp=0; % interpolates dtm data if necessarydo_pie_progress=0; % user feedback of calculation timeuser_ant=0; % user selectable antenna positioning
%***************************************************************************% variable selections%***************************************************************************
domain_height=10; % default domain heightx_max=20; % default domain range distancedbmax=10; % maximum scale value for field displaydbmin=-40; % minimum scale value for field displayt_amp=1.2; % maximum amplitude scale for time displayf=1.575*GHZ; % centre frequency of analysislb_roughness=0; % surface roughness height for lower boundarybs_roughness=0; % surface roughness height for backscatterersrange_step=1; % selected range step size for PE calculationsrange_dist=6; % default location of antenna in x-dimensionant_height=1; % default location of antenna in z-dimensionangle_range=1; % user selects single angle or% angle_range=2:0.05:10; % or range of arbitrary linear values or% angle_range=... % or load angle info from data to simulate% ele_data(resampled_epoch_range);LHCP_reject=6; % antenna rejection of LHCP signals in dBrc_type=’concrete’;% reflection coefficient material type flag % if empty or 0 then not invoked % use only the following strings: % sea_water, fresh_water, ground_dry, % ground_med, ground_wet, concrete
if time_plot % catch flags for missed settings time_analysis=1; % user must want time analysisendif backscatter terrain=1; % user must have terrain selected to use bsend
![Page 330: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/330.jpg)
C-3
%***************************************************************************% setup for image saving%***************************************************************************
if save_images % save image option ON prntype=’-dbitmap’; % set saved image format dirpath= ’c:\bruce\PEfiles\movfiles\’; % set path dirname=[’F_mode2’];%,num2str(sv_num)]; % directory name for images if ~exist([dirpath,dirname],’dir’) % check if directory exists eval([’!mkdir ’,dirpath,dirname]); % if not make directory end eval([’cd ’,dirpath,dirname]); % change to new directoryend
%***************************************************************************% terrain and backscatter setup routines%***************************************************************************
if terrain % use terrain option ON [TERRAIN_PROFILE,domain_height,x_max]=... % load the terrain information loadprofile(range_step,do_interp); fwd_delta_h=get_heights(TERRAIN_PROFILE); % load delta heights ofterrain if backscatter % backscatter option ON rev_delta_h=-fliplr(fwd_delta_h); % reversed delta heights scatterer=find(fwd_delta_h>0); % index to back scatterers endend
%***************************************************************************% get antenna height and distance from user input if option selected%***************************************************************************
if user_ant&terrain % available only with terrain figure % open figure hold on terrain_x=0:range_step:(x_max-2*range_step); % set up terrain display dk_green=[ 0.3 .5 0.3]; % set terrain block colour col=dk_green; for index=1:length(terrain_x) % fill in terrain profile h=fill([terrain_x(index) terrain_x(index)... terrain_x(index)+range_step... terrain_x(index)+range_step],... [0 TERRAIN_PROFILE(index) TERRAIN_PROFILE(index) 0],col); set(h,’EdgeColor’,col); end; set(gca,’YLim’,[0 domain_height],’XLim’,[0 x_max-2*range_step]); title(’Use the mouse to position the Antenna’); xlabel(’Range (m)’); ylabel(’Domain height (m)’);
ans=ginput(1); % get user selection range_dist=round(ans(1)/range_step)*range_step; % set range toantenna ant_height=ans(2); % set height of antenna
%bring the antenna height back down to height above terrain
ant_height=ant_height-TERRAIN_PROFILE(round(range_dist/range_step)); close;end
![Page 331: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/331.jpg)
C-4
%***************************************************************************% time analysis setup routines if option selected% calls petasetup.m%***************************************************************************
if time_analysis % time analysis option ON: load time analysisparameters [f_max,f_array,lower_pad,upper_pad,ts,tmin]=... petasetup(f,x_max,range_dist,range_step,... ant_height,angle_range,backscatter,terrain);else % time analysis option OFF: limit to centre frequency f_array=f; % frequency array is centre frequency f_max=f; % maximum frequency is centre frequencyend
%***************************************************************************% sampling in the z (height) domain% at 90 degrees dz is half the wavelength of the maximum frequency% other selections are allowable within nyquist bound or better%***************************************************************************
dz=0.01; % C/(4*f_max) is the alternative
%***************************************************************************% adjust terrain profile for dz sample resolution%***************************************************************************
if terrain TERRAIN_PROFILE=round(TERRAIN_PROFILE/dz)*dz;end
%***************************************************************************% FSPE setup routines%***************************************************************************
ant_height=round(ant_height/dz)*dz; % antenna height at dz stepcell_X=round(range_dist/range_step); % antenna range cellcell_Z=round(ant_height/dz); % antenna height cellif time_analysis % arrays for time analysis delay_time=zeros(length(angle_range),8); % initialise time delay delay_amp=zeros(length(angle_range),8); % initialise amplitude delay_phase=zeros(length(angle_range),8); % initialise phase delay_aoa=zeros(length(angle_range),8); % initialise angle of arrivalend
angle_index=1; % initialise loop index (angle)
%***************************************************************************% Run Simulation for all angles through calls to mpe.m (Multiple PE)%***************************************************************************
for theta=angle_range*RADIANS %======START OF SIMULATION====== disp([’Running FSPE for ’,... num2str(angle_range(angle_index)),... ’ degs :’, rc_type]); mpe;
![Page 332: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/332.jpg)
C-5
%***************************************************************************% Store PETA results for each multipath signal% delay_time is delay% delay_amp is normalised amplitude {0 1}% delay_phase is phase% delay_aoa is angle of arrival% t_out is MCIR data%***************************************************************************
if time_analysis delay_time(angle_index,1:length(delay))=delay(1,:); delay_amp(angle_index,1:length(mp_mag))=mp_mag(1,:); delay_phase(angle_index,1:length(phase_diff))=phase_diff(1,:); delay_aoa(angle_index,1:length(aoa))=aoa(1,:); t_out(:,:,angle_index)=time_out; end
%***************************************************************************% Store FSPE results for each propagation angle% field_result is total FSPE field returned by mpe.m% field_pe is absolute field value at antenna location (f)% field_db is field value in dB%***************************************************************************
field_pe(angle_index)=result_field(cell_Z,cell_X); field_db(angle_index)=20*log10(abs(field_pe(angle_index)));
%***************************************************************************% set-up routines to save images of simulation%***************************************************************************
if (save_images)&(field_plot|time_plot) if angle_index<10 filename=[’ image0’,num2str(angle_index)]; else filename=[’ image’,num2str(angle_index)]; end eval([’print ’,prntype,filename]); close end angle_index=angle_index+1;end %=======END OF SIMULATION=======
%***************************************************************************% Simulation timing and operations report%***************************************************************************
runtime=toc; % time of simulationdisp([’simulation time: ’,num2str(runtime),’s. ’,’FLOPS: ’,num2str(flops)]);
%***************************************************************************% Save results to MAT file if required%***************************************************************************
if save_results fname=input(’save results filename?’,’s’); if time_analysis t_out=squeeze(t_out); % remove single dimension in t_out array eval([’save ’,fname,’_td’,... ’ delay_time delay_amp ... delay_phase delay_aoa ... angle_range time_axis ... t_out field_db field_pe’]); else eval([’save ’,fname,’_fd’,’ field_db field_pe angle_range’]); endend
![Page 333: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/333.jpg)
C-6
%***************************************************************************% Memory cleanup%***************************************************************************
clear globalclear do_* db*clear field_plot time_plot terrain backscatter time_analysis... filter_time save_spectrum save_results save_images user_ant angle_index... aoa delay domain_height f_max lower_pad upper_pad mp_mag phase_diff ... stored_field full_field t_amp time_out theta
![Page 334: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/334.jpg)
C-7
(b) MPE.M
%***************************************************************************% MPE - called by gope.m for range of propagation angles%% Free-space parabolic equation Fourier step method% Options:% Backscatter% Time series analysis option%% This file is called recursively by gope so that a range of propagation% angles can be used to provide a simulation of GPS satellite motion.% Depending on options selected it will calculate field results% with and without terrain, backscattering, variable sv power profiles,% or reflection coefficients. Time domain analysis is also carried out% if selected.%% Results are saved and/or displayed as required.%% Bruce M. Hannah% Cooperative Research Centre for Satellite Systems% Queensland University of Technology% Version date 22/8/00%***************************************************************************
%***************************************************************************% global variables%***************************************************************************
global TRUE; % Logical true keywordglobal FALSE; % Logical false keywordglobal MHZ; % Megahertzglobal GHZ; % Gigahertzglobal C; % Speed of light in a vacuum (m/s)global RADIANS; % Convert degs to rads (ex: 10*RADIANS)global DEGREES; % Convert rads to degglobal TERRAIN_PROFILE; % Terrain profile data
%***************************************************************************% Linearised user minimum received signal level profile% as per ICD-GPS-200 6.3.1 if option selected%***************************************************************************
if sv_power if theta*DEGREES<=4 sv_power_ratio=1; elseif theta*DEGREES<=20 sv_power_ratio=10^(((1.1/16)*(theta*DEGREES-4))/20); elseif theta*DEGREES<=32 sv_power_ratio=10^((1.1+(0.75/12)*(theta*DEGREES-20))/20); elseif theta*DEGREES<=40 sv_power_ratio=10^((1.85+(0.16/8)*(theta*DEGREES-32))/20); elseif theta*DEGREES<=48 sv_power_ratio=10^((2.01-(0.14/8)*(theta*DEGREES-40))/20); elseif theta*DEGREES<=60 sv_power_ratio=10^((1.87-(0.66/12)*(theta*DEGREES-48))/20); elseif theta*DEGREES<=90 sv_power_ratio=10^((1.21-(1.21/30)*(theta*DEGREES-60))/20); endelse sv_power_ratio=1;end
bck_call=FALSE; % flag for use with backscatter routine
![Page 335: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/335.jpg)
C-8
%***************************************************************************% z_max: maximum height of domain allowing for entry of line-of-sight signal% into the area of interest. This is achieved by taking into account the% angle of propagation and then doubling the domain height.%***************************************************************************
z_max=2*(domain_height+((x_max*1.1)*tan(theta)));
%***************************************************************************% setup sampling and initilise arrays, dz is calculated in gope.m%***************************************************************************
N_minus_1=nextpow2(z_max/dz); % number of z samplesN=N_minus_1+1;if save_spectrum U=zeros(2^N,x_max/range_step); % initialise spectrum arrayendif backscatter % with backscatter initialise % an initial fwd field array fwd_initial=zeros(2^N_minus_1,x_max/range_step);endu=zeros(2^N_minus_1,x_max/range_step); % initialise generic field arrayfwd_field=zeros(2^N_minus_1,x_max/range_step); % initialise forward fieldarrayu_len=length(u(:,1)); % length of z domain in samplesz=dz:dz:((2^N_minus_1)*dz); % generate array of z elementsz=reshape(z,u_len,1);
%***************************************************************************% generate Hanning window and modify the shape of the window so that it% tapers off the aperture field for values between Zmax and 2*Zmax.%***************************************************************************
hn=hanning(u_len);hn(1:u_len/2)=ones(size(hn(1:u_len/2)));
%***************************************************************************% setup the sampling in the angular-spectrum or p-domain%***************************************************************************
dp=2*pi/((2^N)*dz);p=0:dp:(((2^N_minus_1)-1)*dp);
%***************************************************************************% fft setup%***************************************************************************
fft_size=2^N;
clear N u_len % memory clean up
%***************************************************************************% Do the calculations for the frequencies specified in f_array. These %frequencies are specified for the Fourier-based time analysis. If time% analysis is not being done then the f_array is simply the GPS frequency% specified in gope.m%***************************************************************************
for array_index=1:length(f_array) ========Start field calculations========== % if selected option if do_pie_progress % provide visual feedback l=length(f_array)-array_index+1; % of calculation progress pie([array_index l]); title(’Percent Complete’); pause(0.01); set(gca,’Selected’,’on’); end
frequency=f_array(array_index); % get current frequency k=2*pi*frequency/C; % calculate wavenumber
![Page 336: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/336.jpg)
C-9
%**************************************************************************% setup reflection coefficients for modifying the angular-spectrum% or p-space%**************************************************************************
% Set the 0 to 90 degrees theta limit from p-space
rc_theta=asin((0:2*dp:k)/k);
% Electrical properties of mediums if rc_type if strcmp(rc_type,’sea_water’) cond=4; ep_r=20; elseif strcmp(rc_type,’fresh_water’) cond=1e-3; ep_r=80; elseif strcmp(rc_type,’ground_dry’) cond=1e-5; ep_r=4; elseif strcmp(rc_type,’ground_med’) cond=1e-3; ep_r=7; elseif strcmp(rc_type,’ground_wet’) cond=1e-2; ep_r=30; elseif strcmp(rc_type,’concrete’) cond=2e-5; ep_r=3; end
ep=ep_r-(j*60*(C/frequency)*cond);
sin_term=sin(rc_theta); cos_term=sqrt(ep-(cos(rc_theta).^2)); refl_h=(sin_term-cos_term)./(sin_term+cos_term); refl_v=(ep*sin_term-cos_term)./(ep*sin_term+cos_term); po=(refl_h+refl_v)/2; px=(refl_h-refl_v)/2;
% Co-polarised rc magnitude: forward propagation on lower boundary co_mag=abs(po);
% Cross-polarised rc magnitude: forward propagation on lower boundary cross_mag=abs(px);
% Coupled forward reflection coefficient for RHCP incidence fmode_rc_R=(co_mag+(10^(-LHCP_reject/20))*cross_mag);
% Coupled forward reflection coefficient for LHCP incidence fmode_rc_L=((10^(-LHCP_reject/20))*co_mag+cross_mag);
%***************************************************************************% Surface roughness reduction factor for lower boundary%***************************************************************************
g_term=(4*pi*lb_roughness*sin(rc_theta))/(C/frequency); bessel_factor=besseli(0,0.5*(g_term.^2));
% Rough surface reduction factor for lower boundary fwd_sr_factor=exp(-0.5*(g_term.^2)).*bessel_factor;
% Total combined rc mag for fwd prop over lower boundary fwd_rc_lb=fmode_rc_R.*fwd_sr_factor;
clear cond sin_term cos_term refl_h refl_v po px g_term... bessel_factor ep_r
![Page 337: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/337.jpg)
C-10
%***************************************************************************% backscatter reflection coefficients%***************************************************************************
if backscatter
% Theta at the vertical backscatter interface % for the calculation of reflection coefficients % is 90 degrees - propagation angle bs_theta=pi/2-rc_theta;
% Calculate bs reflection coefficient magnitudes sin_term=sin(bs_theta); cos_term=sqrt(ep-(cos(bs_theta).^2)); refl_h=(sin_term-cos_term)./(sin_term+cos_term); refl_v=(ep*sin_term-cos_term)./(ep*sin_term+cos_term); po=(refl_h+refl_v)/2; px=(refl_h-refl_v)/2;
% Co-polarised component at bs interface bs_co_mag=abs(po);
% Cross-polarised component at bs interface bs_cross_mag=abs(px);
% BA-mode rc for RHCP incidence on bs interface bamode_rc_R=(bs_co_mag+(10^(-LHCP_reject/20))*bs_cross_mag);
% BA-mode rc for LHCP incidence on bs interface bamode_rc_L=((10^(-LHCP_reject/20))*bs_co_mag+bs_cross_mag);
% Coupling factor for BB-mode zone 1 z1_factor=bamode_rc_L./bamode_rc_R;
% Coupling factor for BB-mode zone 2 z2_factor=fmode_rc_L./fmode_rc_R;
%***********************************************************************% surface roughness reduction factor on backscatter interface%***********************************************************************
g_term=(4*pi*bs_roughness*sin(bs_theta))/(C/frequency); bessel_factor=besseli(0,0.5*(g_term.^2)); bs_sr_factor=exp(-0.5*(g_term.^2)).*bessel_factor;
%***********************************************************************% Calculation of coupled and modified% coupled reflection coefficient magnitudes%***********************************************************************
% Lower boundary modified coupled rc in back direction bck_rc_lb=(co_mag+z1_factor.*cross_mag).*fwd_sr_factor;
% Backscatter interface coupled rc for BA-mode propagation bamode_rc_bs=bamode_rc_R.*bs_sr_factor;
% Backscatter interface modified couplied rc BB-mode propagation bbmode_rc_bs=(bs_co_mag+z2_factor.*bs_cross_mag).*bs_sr_factor;
clear sin_term cos_term refl_h refl_v po px bs_co_mag... bs_cross_mag bamode_rc_R bamode_rc_L z1_factor z2_factor... g_term bessel_factor bs_sr_factor fmode_rc_L fmode_rc_R... bs_theta
end
![Page 338: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/338.jpg)
C-11
%***************************************************************************% Calculation of coupled reflection coefficient for initial reflected% incident signal%***************************************************************************
if ini_reflection
% Calculate initial incident rc from initial propagation angle sin_term=sin(theta); cos_term=sqrt(ep-(cos(theta).^2)); refl_h=(sin_term-cos_term)./(sin_term+cos_term); refl_v=(ep*sin_term-cos_term)./(ep*sin_term+cos_term); po=(refl_h+refl_v)/2; px=(refl_h-refl_v)/2; orig_mag=abs(po); cross_mag=abs(px); inc_rc_mag=(orig_mag+(10^(-LHCP_reject/20))*cross_mag);
g_term=(4*pi*lb_roughness*sin(theta))/(C/frequency); bessel_factor=besseli(0,0.5*(g_term.^2));
% Rough surface reduction factor mod_factor=exp(-0.5*(g_term.^2)).*bessel_factor;
% initial value for incident reflection inc_rc_mag=inc_rc_mag*mod_factor;
clear sin_term cos_term refl_h refl_v po px... orig_mag cross_mag g_term bessel_factor mod_factor end
% If no medium defined do total reflection % This does not couple polarisations else fwd_rc_lb=ones(1,length(rc_theta)); bck_rc_lb=ones(1,length(rc_theta)); bamode_rc_bs=ones(1,length(rc_theta)); bbmode_rc_bs=ones(1,length(rc_theta)); inc_rc_mag=1; end
rc_mag=fwd_rc_lb;
clear ep co_mag cross_mag fmode_rc_L fmode_rc_R fwd_sr_factor... fwd_rc_lb rc_theta
%***************************************************************************% setup initial field and pe propagator%***************************************************************************
incident=exp(-j*k*z*sin(theta)); % incident plane-wave reflected=-exp(j*k*z*sin(theta)); % reflected plane-wave
if ini_reflection u(:,1)=(sv_power_ratio*incident)+... (sv_power_ratio*inc_rc_mag*reflected); % initial reduced field hasreflection clear inc_rc_mag else u(:,1)=(sv_power_ratio*incident); % just initial incident field end
% FSPE propagator propagator=exp(i*k*range_step*(sqrt(1-p.^2/k^2)-1));
% Image of FSPE propagator mirror_prop=fliplr(propagator);
% Combined propagator for FSPE solution combined_prop=[propagator,mirror_prop];
![Page 339: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/339.jpg)
C-12
% Initial propagation is in forward direction % Therefore terrain jumps are for forward terrain if terrain delta_h=fwd_delta_h; end
%***************************************************************************% The forward propagated reduced field is calculated by the fieldcalc% routine%***************************************************************************
fieldcalc;
%***************************************************************************% The FSPE calculates the reduced field which has the rapid phase variation% in x removed. For time-domain analysis this phase variation needs% to be reinstated.%***************************************************************************
index=1;
for x=range_step:range_step:x_max fwd_field(:,index)=u(:,index)*exp(j*k*x); index=index+1; end
if terrain stored_count=bin_count; end
%***************************************************************************% If the backscatter option is selected we need to find the initial% backscatter field components and propagate them using the FSPE% field calculation (fieldcalc) routine.%***************************************************************************
if backscatter
% Identify backscatter interface locations and element sizes for back_index=1:length(scatterer) ref_x_index=scatterer(back_index); num_cells=round((1/dz)*delta_h(ref_x_index)); fwd_initial(1:num_cells,ref_x_index)=... fwd_field(1:num_cells,ref_x_index); end
% Mirror initial backscatter fields so fieldcalc can be used bck_initial=fliplr(fwd_initial); clear back_index num_cells ref_x_index [row_pos,col_pos,field_value]=find(bck_initial); row_index=find(row_pos==1); col_pos=col_pos(row_index); delta_h=rev_delta_h;
% Insert initial bs field values and calculate field % for each backscattering interface for index=1:length(col_pos)
u(:,1)=bck_initial(:,1); bck_call=col_pos(index); row_start=row_index(index);
if index==length(col_pos) row_end=length(row_pos); else row_end=row_index(index+1)-1; end
num_rows=row_end-row_start+1;
% The lower boundary reflection coefficient is now % the modified coupled rc for backscatter
![Page 340: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/340.jpg)
C-13
rc_mag=bck_rc_lb;
% Use fieldcalc to solve field fieldcalc;
x=0; col_index=col_pos(index);
% Reinstate phase variation for backscatter for phase_index=col_pos(index)*range_step:range_step:x_max u(:,col_index)=u(:,col_index)*exp(j*k*x); x=x+range_step; col_index=col_index+1; end
clear phase_index
% For all caculated fields in the reverse direction % the total reverse field is the summation of all solved fields if index==1 rev_field=u; else rev_field=rev_field+u; end
end
% Back propagation completed bck_call=FALSE;
% The total back propagated field is the mirrored reverse FSPE result bck_field=fliplr(rev_field);
%**************************************************************************% Back propagation is being done therefore the total field is addition of% both forward and back propagated fields.%**************************************************************************
total_field=bck_field+fwd_field;
clear num_rows rev_field bck_initial bck_field field_value row_pos... row_start col_pos col_index bamode_rc_bs bbmode_rc_bs bck_rc_lb... row_end row_index
else
**************************************************************************% Back propagation is not being done therefore the total field is the% forward field only. %**************************************************************************
total_field=fwd_field;
end
clear delta_bins bin_count incident reflected fwd_field... combined_prop propagator mirror_prop
![Page 341: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/341.jpg)
C-14
%***************************************************************************% To correct the time domain results to the incident boundary% we must first correct for the relative height of the antenna location% to the incident field origin point.If there is terrain the relative height% is the height difference between the height of the terrain% at the incident boundary and the terrain height at the antenna,% plus the actual antenna height above the terrain.% If there is no terrain the relative height is simply the antenna height.% This is the first correction term applied to the field result.%% To correct to the line-of-sight entry point on the incident boundary% a correction is made for the antenna distance.% This is the second correction term to the field result.
% In addition a correction is applied to shift the time domain results% relative to the minimum time set for the time window. This correction is% the last correction term applied to the field result.%% The resultant field value is stored for each frequency in the array% field_cell.%% An angle of arrival estimation uses the variation of time arrival of the% signals by looking at the time results one cell above the antenna.% This result is stored in aoa_cell.%***************************************************************************
if time_analysis
if terrain del_h=ant_height+(round(TERRAIN_PROFILE(cell_X)/dz)*dz)-... (round(TERRAIN_PROFILE(1)/dz)*dz); else del_h=ant_height; end
% Phase correction field_cell(array_index)=total_field(cell_Z,cell_X)... *exp(j*k*del_h*sin(theta))... *exp(j*k*range_dist*(1/cos(theta)-cos(theta)))... *exp(-j*2*pi*frequency*tmin); aoa_cell(array_index)=total_field(cell_Z+1,cell_X)... *exp(j*k*del_h*sin(theta))... *exp(j*k*range_dist*(1/cos(theta)-cos(theta)))... *exp(-j*2*pi*frequency*tmin);
clear del_h
end
%***************************************************************************% When the frequency is the centre frequency store the field result% at antenna location%***************************************************************************
if frequency==f stored_field=total_field; result_field=stored_field(cell_Z,cell_X); end
clear total_field
end
![Page 342: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/342.jpg)
C-15
%===========================================================================%===========Field calculations at all frequencies completed=================%===========================================================================
clear sv_power_ratio k frequency dp fft_size array_index ... fwd_initial hn N_minus_1 p z bck_call u
% Visual progress indication stoppedif do_pie_progress closeend
%***************************************************************************% Time analysis is done here. The frequency-domain spectrum is created from% the stored field results of each frequency. The inverse fft then gives the% required time-domain results. Residual phase and angle of arrival (aoa)for% each peak are also estimated.%***************************************************************************
if time_analysis
% Time results filtered if option selected % Filtering results in widening of time pulses % Resolution is therefore reduced with filtering if filter_time win_fun=transpose(hanning(length(field_cell))); field_cell=win_fun.*field_cell; aoa_cell=win_fun.*aoa_cell; threshold=0.1; % threshold for filtered results clear win_fun else threshold=0.4; % threshold value for peak search end
pos_spec=[lower_pad,field_cell,upper_pad]; % create frequency spectrum td=fft(pos_spec); % do fft to get time-domain result time_out=abs((td)); % get the amplitude time spectrum t_step=2*ts; % the sampling is twice since using half the samples max_value=max(time_out); % find the maximum value peak time_out=time_out./max_value; % prenormalise time_out for peak search samples=find(time_out>threshold); % get all the samples above thethreshold max_index=1; % initialise the index of peaks sample_num=0; % initilise sample storage array store=FALSE; % recursive detection flag
%***************************************************************************% Use an gradient ascent technique to find the peak of each pulse% above the threshold.%***************************************************************************
for index=1:length(samples)-1 % get gradient grad=time_out(samples(index+1))-time_out(samples(index));
if grad>0 % if a positive gradient find_max=samples(index+1); % store the sample number store=TRUE; % and set store flag else % if the gradient is negative
if store % and there is a stored sample, this a peak sample_num(max_index)=find_max; % the sample number of the peakis stored store=FALSE; % reset flag to search for more peaks max_index=max_index+1; % increment the peak index for the nextpeak end
end end
num_peaks=length(sample_num); % number of peaks above the threshold
![Page 343: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/343.jpg)
C-16
if num_peaks>0 % if there is at least one peak norm_amp1=time_out(sample_num(1)); % get the prenormalised peak valueof peak 1 time_out=time_out./norm_amp1; % normalise the time output to thefirst peak
%**************************************************************************% The time delays, amplitudes and phases of the peaks are now determined.%**************************************************************************
delay=(tmin/1e-9)+(t_step/1e-9*(sample_num-1)); % array of time delays mp_mag=time_out(sample_num); % array of corresponding magnitudes td_phase=angle(td(sample_num)); % array of phase at each peak
for phase_index=1:length(td_phase) % for each delay
field_ref=exp(j*2*pi*f_array*(delay(phase_index)*1e-9-tmin));
pos_spec_ref=[lower_pad,field_ref,upper_pad]; td_ref=fft((pos_spec_ref)); % reference phase of nth delay td_ref_phase=angle(td_ref(sample_num(phase_index)));
%***********************************************************************% The resultant phase from reflection is the difference between the% actual phase and the reference phase created solely using the% absolute time delay.%***********************************************************************
phase_diff(phase_index)=td_ref_phase-td_phase(phase_index); end
for index=1:num_peaks delay_txt=num2str(delay(index),’%9.4f’); delay_len=length(delay_txt);
if delay_len==7 delay_txt=[’0’,delay_txt]; elseif delay_len==6 delay_txt=[’0’,’0’,delay_txt]; elseif delay_len==5 delay_txt=[’0’,’0’,’0’,delay_txt]; end
text_str(index,:)=[delay_txt,’ ns’]; disp([’peak ’,num2str(index),’: ’,text_str(index,:)]); end
clear index delay_txt delay_len text_str
end
![Page 344: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/344.jpg)
C-17
%***************************************************************************% This is the routine that looks at the cell above the antenna location% and calculates angle of arrival of the peaks in the time spectrum%***************************************************************************
aoa_spec=[lower_pad,aoa_cell,upper_pad]; % create frequency spectrum aoa_td=fft(aoa_spec); % do fft to get time-domain result aoa_time_out=abs((aoa_td)); % get the amplitude time spectrum aoa_max_value=max(aoa_time_out); % find the maximum value peak % prenormalise time_out for peak search aoa_time_out=aoa_time_out./max_value; % get all the samples above the threshold aoa_samples=find(aoa_time_out>threshold); max_index=1; % initialise the index of peaks aoa_sample_num=0; % initilise sample storage array store=FALSE; % recursive detection flag
%***************************************************************************% Use an gradient ascent technique to find the peak of each pulse% above the threshold.%***************************************************************************
for index=1:length(aoa_samples)-1 grad=aoa_time_out(aoa_samples(index+1))-... aoa_time_out(aoa_samples(index)); % store the gradient
if grad>0 % if a positive gradient find_max=aoa_samples(index+1); % store the sample number store=TRUE; % and set store flag else % if the gradient is negative
if store % and there is a stored sample, this a peak % sample number of the peak is stored
aoa_sample_num(max_index)=find_max; store=FALSE; % reset flag to search for more peaks % increment the peak index for the next peak max_index=max_index+1; end
end end
% number of peaks above the threshold num_aoa_peaks=length(aoa_sample_num);
if num_aoa_peaks>0 % if there is at least one peak norm_amp1=time_out(aoa_sample_num(1)); % normalise the time output aoa_time_out=aoa_time_out./norm_amp1; % to the first peak
%**************************************************************************% The time delay for aoa analysis is now determined.%**************************************************************************
% array of time delays aoa_delay=... (tmin/1e-9)+(t_step/1e-9*(aoa_sample_num-1)); end
%**************************************************************************% Angle of arrival (aoa) estimates are found for each pulse in the time% domain by finding the difference between the arrival times in the two% cells. The inverse sin of the relative distance travelled by the pulse% gives the aoa estimate.%***************************************************************************
aoa=asin(C*(delay-aoa_delay)*1e-9/dz)*DEGREES;
time_axis=(0:t_step/1e-9:length(time_out)*t_step/1e-9)+tmin/1e-9;
![Page 345: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/345.jpg)
C-18
%***************************************************************************% Plotting Routines - Time plot only%***************************************************************************
if time_plot&~field_plot plot(time_axis,[time_out,0]); axis([round(min(time_axis)) round(max(time_axis)) 0 t_amp]); xlabel(’Absolute time delay (ns)’); x_text_pos=time_axis(round(length(time_out)/1.64));
for index=1:num_peaks text(x_text_pos,1-index*0.05,... [’peak ’,num2str(index),’ delay: ’,text_str(index,:)]); end
if num_peaks>1
for index=2:num_peaks delay_diff=delay(index)-delay(1); text(x_text_pos,1-(index+num_peaks-1)*.05,... [’time delay: ’,num2str(delay_diff),’ ns’]); end
end clear index text_str x_text_pos delay_diff end clear max_index find_max grad aoa_* max_value field_cell field_ref... norm_amp1 num_peaks num_aoa_peaks pos_spec pos_spec_ref samples... sample_num store t_step td td_phase td_ref td_ref_phase threshold... tmin tsend
%**************************************************************************% Setup rountine for field alone or field and time plots together.%***************************************************************************
if field_plot|(field_plot&time_plot)
%***************************************************************************% Shift the field to account for the boundary shift technique. Zeros are% inserted at the terrain positions.%***************************************************************************
if terrain
x_index=1;
num_zeros=round((dz+TERRAIN_PROFILE(x_index))/dz);
for x=0:range_step:x_max-2*range_step
if x_index==1 stored_bins=0; else stored_bins=stored_count(x_index-1); end
num_zeros=num_zeros+stored_bins;
clear stored_bins
if ~(num_zeros==0) clear temp temp(abs(num_zeros))=0; field=stored_field((1:(length(stored_field(:,x_index))-... length(temp))),x_index); us=size(field); field=reshape(field,1,us(1)); temp_field=[temp,field]; stored_field(:,x_index)=temp_field(:); clear field temp temp_field us end
![Page 346: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/346.jpg)
C-19
x_index=x_index+1; end
clear stored_count x_index
end
num_z_elem=round(domain_height/dz); % number of Z elements X=0:range_step:x_max-2*range_step; % X array of range elements Z=0:dz:(num_z_elem)*dz; % Z array of height elements warning off % suppress log of zero warnings log_field=20*log10(abs(stored_field(1:length(Z),1:length(X)))); warning on clear stored_field num_z_elem num_zerosend
%***************************************************************************% Field plot only%***************************************************************************
if field_plot&~time_plot pcolor(X,Z,log_field); colormap(jet); caxis([dbmin dbmax]); shading interp; colorbar;
clear X Z log_field
if terrain % overlay the terrain on the field plot hold on terrain_x=0:range_step:(x_max-2*range_step); dk_green=[ 0.3 .5 0.3]; col=dk_green;
for step=1:length(terrain_x); h=fill([terrain_x(step) terrain_x(step) ... terrain_x(step)+range_step terrain_x(step)+range_step],... [0 TERRAIN_PROFILE(step) TERRAIN_PROFILE(step) 0],col); set(h,’EdgeColor’,col); end
hold off y_ant=ant_height+TERRAIN_PROFILE(cell_X);
clear step col dk_green h terrain_x
else y_ant=ant_height; end
if do_fwd_rays fr_height=plot_fwd_rays(range_dist,range_step,... ant_height,domain_height,theta,terrain); end
if do_rev_rays&backscatter [bra_dist,brb_dist,brb_height]=plot_rev_rays... (range_dist,ant_height,domain_height,theta,... x_max,scatterer,range_step,cell_X); end
hold on plot(range_dist,y_ant,’kv’); % plot antenna location markers plot(range_dist,y_ant,’k*’); hold off
xlabel(’Range (m)’); ylabel(’Height (m)’); title([’theta = ’,num2str(theta*DEGREES),' °']);end
![Page 347: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/347.jpg)
C-20
%***************************************************************************% Field and time plotted together.%***************************************************************************
if field_plot&time_plot subplot(2,1,1),pcolor(X,Z,log_field); colormap(jet); caxis([dbmin dbmax]); shading interp; colorbar;
clear log_field X Z
if terrain hold on terrain_x=0:range_step:(x_max-2*range_step); dk_green=[ 0.3 .5 0.3]; col=dk_green; for step=1:length(terrain_x);
h=fill([terrain_x(step) terrain_x(step) ... terrain_x(step)+range_step terrain_x(step)+range_step],... [0 TERRAIN_PROFILE(step) TERRAIN_PROFILE(step) 0],col); set(h,’EdgeColor’,col); end
hold off y_ant=ant_height+TERRAIN_PROFILE(cell_X);
clear step col dk_green h terrain_x
else y_ant=ant_height; end
if do_fwd_rays fr_height=plot_fwd_rays(range_dist,range_step,ant_height,... domain_height,theta,terrain); end
if do_rev_rays&backscatter [bra_dist,brb_dist,brb_height]=plot_rev_rays(range_dist,... ant_height,domain_height,theta,x_max,scatterer,range_step,cell_X); end
hold on plot(range_dist,y_ant,’kv’); plot(range_dist,y_ant,’k*’); hold off
title([’Theta = ’,num2str(theta*DEGREES),' °']); xlabel('Range (m)'); ylabel('Height (m)');
subplot(2,1,2),plot(time_axis,[time_out,0]);
axis([round(min(time_axis)) round(max(time_axis)) 0 t_amp]); % setamplitude limit
if do_fwd_rays t_los=(range_dist/(C*cos(theta)))*1e9; t_fr=t_los+(2*fr_height*sin(theta)/C)*1e9; hold on plot(t_los,0,'gd') plot(t_fr,0,'kd') hold off clear t_los t_fr end
if exist('bra_dist')&bra_dist
for index=1:length(bra_dist)
![Page 348: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/348.jpg)
C-21
t_bra=(range_dist*1e9/(C*cos(theta)))+... (2*bra_dist(index)*cos(theta)*1e9/C); hold on plot(t_bra,0,’bd’) hold off clear t_bra end
end
if exist(’brb_dist’)&brb_dist
for index=1:length(brb_dist) t_brb=(range_dist*1e9/(C*cos(theta)))+... (2*brb_dist(index)*cos(theta)*1e9/C)+... (2*brb_height(index)*sin(theta)*1e9/C); hold on plot(t_brb,0,’cd’) hold off clear t_brb end
end
xlabel(’Absolute time delay (ns)’); ylabel(’Relative amplitude’);end
clear rev_delta_h y_ant x delta_h fwd_delta_h fr_height... brb_dist brb_height bra_dist
![Page 349: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/349.jpg)
C-22
(c) FIELDCALC.M
%***************************************************************************% FIELDCALC - Free-space parabolic field solution% Called by mpe.m for each angle and frequency if doing time analysis%% Bruce Hannah% version date 22/8/00%***************************************************************************
clear bin_count delta_bins;
%***************************************************************************% March FSPE solution out in range%***************************************************************************
x_index=1;
for x=0:range_step:(x_max-(2*range_step))
%***************************************************************************% Filter upper limit of solution domain with Hanning Window%***************************************************************************
u(:,x_index)=u(:,x_index).*hn;
%***************************************************************************% Previous field from field array%***************************************************************************
prev_field=u(:,x_index);
ps=size(prev_field); % Reshape field prev_field=reshape(prev_field,1,ps(1));
%***************************************************************************% Set up parameters for boundary shifting%***************************************************************************
if terrain if delta_h(x_index)~=0 delta_bins=round(delta_h(x_index)/dz); if (delta_bins>0) prev_field=prev_field((abs(delta_bins)+1):length(prev_field)); prev_field(length(prev_field)+abs(delta_bins))=0; end if (delta_bins<0) clear temp; temp(abs(delta_bins))=0; prev_field=[temp,prev_field(1:... (length(prev_field)-length(temp)))]; end else delta_bins=0; end bin_count(x_index)=delta_bins; end
%***************************************************************************% Create image field and apply 180 deg phase shift%***************************************************************************
odd_part=fliplr(-prev_field);
![Page 350: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/350.jpg)
C-23
%***************************************************************************% Create the image angular spectrum%***************************************************************************
odd_ang_spec=fft(odd_part);
%***************************************************************************% This is the 0:90 degree region for the image spectrum%***************************************************************************
pos_ang_spec=odd_ang_spec(1:length(rc_mag));
%***************************************************************************% Apply reflection coefficient for the reflection% In forward direction this is F-mode reflection coefficient% In back direction (when there is backscatter) this is the% modified coupled reflection coefficient to compensate for% polarisation coupling at the backscatter interface%***************************************************************************
mod_pos_spec=(rc_mag.*abs(pos_ang_spec)).*exp(j*angle(pos_ang_spec));
%***************************************************************************% Insert modified spectrum into total spectrum%***************************************************************************
odd_ang_spec(1:length(rc_mag))=mod_pos_spec;
%***************************************************************************% Inverse fft to give total image field with correct rc applied%***************************************************************************
odd_part=ifft(odd_ang_spec);
clear odd_ang_spec pos_ang_spec mod_pos_spec
%***************************************************************************% Combine incident and image fields to create total field%***************************************************************************
combined=[prev_field,odd_part];
%***************************************************************************% Then take the FFT of it for total angular spectrum%***************************************************************************
U_x=fft(combined,fft_size);
%*************************************************************************** % Save incident spectrum if required%***************************************************************************
if save_spectrum if (x_index==1) U(:,1)=U_x(:); end end
%*************************************************************************** % U_x is the angular spectrum for the previous step. % We now multiply by the Propagator%***************************************************************************
U_x=U_x.*combined_prop;
if save_spectrum U(:,x_index+1)=U_x(:); end
![Page 351: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/351.jpg)
C-24
%***************************************************************************% Take the IFFT to create FSPE propagated field%***************************************************************************
u_x=ifft(U_x,fft_size);
clear U_x combined
%***************************************************************************% Only want the upper half of the domain (solution domain)%***************************************************************************
u_x=u_x(1:(2^(N_minus_1)));
%***************************************************************************% This is now the next field array%***************************************************************************
u(:,x_index+1)=u_x(:);
%***************************************************************************% Insert the backscattered field components with reflection coefficients% The postive angles of the angular spectrum at the bs interface are bb-mode% The negative angles are ba-mode% Each region of angular spectrum is seperated into -90-0 and 0-90 degree% The appropriate reflection coefficient is then applied in p-space%***************************************************************************
if bck_call==x_index+1
%*************************************************************************% Apply 180 deg phase shift to field components%*************************************************************************
u(1:num_rows,x_index+1)=-field_value(row_start:row_end);
%*************************************************************************% Full angular spectrum (p-space) of backscattered field%*************************************************************************
bs_p_space=fft(u(:,x_index+1));
%*************************************************************************% Angles of propagation at 0:90 degs are bb-mode%*************************************************************************
pos_ang_spec=bs_p_space(1:length(bbmode_rc_bs));
%*************************************************************************% Angles of propagation at -90:0 degs are ba-mode%*************************************************************************
neg_ang_spec=bs_p_space(length(bs_p_space)-... length(bamode_rc_bs)+1:length(bs_p_space));
%*************************************************************************% Apply bb-mode reflection coeff magnitude (includes surface roughness)%*************************************************************************
mod_pos_spec=(bbmode_rc_bs’.*... abs(pos_ang_spec)).*exp(j*angle(pos_ang_spec));
%*************************************************************************% Apply ba-mode reflection coeff magnitude%*************************************************************************
mod_neg_spec=(fliplr(bamode_rc_bs)’.*... abs(neg_ang_spec)).*exp(j*angle(neg_ang_spec));
![Page 352: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/352.jpg)
C-25
%*************************************************************************% Insert modified positive angular spectrum components%*************************************************************************
bs_p_space(1:length(bbmode_rc_bs))=mod_pos_spec;
%*************************************************************************% Insert modified negative angular spectrum components%*************************************************************************
bs_p_space(length(bs_p_space)-length(bbmode_rc_bs)+... 1:length(bs_p_space))=mod_neg_spec;
%*************************************************************************% IFFT gives modified elements of the backscattered field%*************************************************************************
u(:,x_index+1)=ifft(bs_p_space);
clear bs_p_space pos_ang_spec neg_ang_spec mod_pos_spec mod_neg_spec
end
x_index=x_index+1;
end
clear x_index rc_mag prev_field odd_part u_x ps
![Page 353: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/353.jpg)
C-26
(d) PETASETUP.M
%***************************************************************************% PETASETUP – PE-based time analysis set-up function% Called by gope.m if doing time analysis%% Bruce Hannah% version date 22/8/00%***************************************************************************
function [f_max,f_array,lower_pad,upper_pad,ts,tmin]=... petasetup(f,x_max,range_dist,range_step,... ant_height,angle_range,backscatter,terrain)
global TERRAIN_PROFILEglobal C;
theta_max=max(angle_range)*pi/180;theta_min=min(angle_range)*pi/180;pulse_width=1e-9; % required width to resolve multipath delays
% round in 10 units below LOS timing
tmin=floor(1e9*range_dist/(10*C*cos(theta_min)))*10e-9;if backscatter % allow time to end of domain and back to ant location tmax=ceil(1e9*(x_max+(x_max-range_dist))/(10*C))*10e-9;else % allow only for forward scatter tmax=ceil((1e9*range_dist/(10*C*cos(theta_max)))... +(2e9*ant_height*sin(theta_max)/(10*C)))*10e-9; if terrain tmax=ceil((1e9*range_dist/(10*C*cos(theta_max)))... +(2e9*(ant_height+TERRAIN_PROFILE(round(range_dist/range_step)))... *sin(theta_max)/(10*C)))*10e-9;% allow additional time for terrain endendtime_window=tmax-tmin;M=round(f*time_window);time_window=M/f;f_step=1/time_window;
if time_window<200e-9 os_factor=128;elseif time_window<500e-9 os_factor=64;elseif time_window<1000e-9 os_factor=32;end
b=f_step*round((1/pulse_width)/f_step); % 2*b is the bandwidth of RECTf_max=f+b; % maximum frequencyf_min=f-b; % minimum frequencymin_fs=(os_factor*f_max); % Nyquist samplingmax_ts=1/min_fs; % maiximum time stepN_min=time_window/max_ts; % minimum sample sizex=nextpow2(N_min); % set at power of twoN_ir=2^x;ts=time_window/N_ir; % time domain sample spacingf_array=f_min:f_step:f_max; % create full array of frequencieslower_pad=zeros(1,round((f_min/f_step)-1)); % pad outupper_pad=zeros(1,(N_ir/2)-length(lower_pad)-length(f_array));
![Page 354: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/354.jpg)
C-27
(e) LOADPROFILE.M
%***************************************************************************% LOADPROFILE – function for loading of DTM terrain data%% Bruce Hannah% Version date 14/8/00%***************************************************************************
function[TERRAIN_PROFILE,domain_height,x_max]=loadprofile(range_step,do_interp)
% function loadprofile.m loads the dtm terrain profile for FSPE-PETA
% dtm range resolution as provided in the dtmdtm_step=1; % dtm range resolution as provided in the dtm
% additional height for the domain above terrain with terrain option ONmedium_height=10;
%***************************************************************************% The following code is an example of created DTM profiles for testing%***************************************************************************
flat(1:10)=0;back_reflector=20;block1(1:3)=3;block2(1:5)=6;block3(1:15)=3;block4(1:15)=1;wedge=0:0.2:5;profile1=[0,0.05,0,0.05,0,0.05,0,0.05,0,0.05,0,0.05,0,0.05,0,0.05,0];profile2=[flat,block1,flat];profile3=[flat,flat,flat,block,flat,flat,block,flat];profile4=[flat,4,flat,flat];profile5=[[4 4 4 3 2 1],flat,[1 2 3 4 4 5 6 6 6 ]];
%***************************************************************************% The DTM profile is loaded at this point%***************************************************************************
dtm=profile3;
%***************************************************************************%sampling of dtm for use by PE%***************************************************************************
dtm=dtm-min(dtm)+1;domain_height=(max(dtm)-1)+medium_height;terrain_scale=range_step/dtm_step;
%***************************************************************************%undersample dtm, if dtm has higher resolution than is used by PE%***************************************************************************
if terrain_scale>1 terrain_index=1; for scale_index=1:terrain_scale:length(dtm) TERRAIN_PROFILE(terrain_index)=dtm(scale_index); terrain_index=terrain_index+1; end
%***************************************************************************% oversample dtm by interpolating between points if dtm resolution is not as% high as required by PE%***************************************************************************
elseif terrain_scale<1 terrain_index=1;
![Page 355: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/355.jpg)
C-28
for dtm_index=2:length(dtm) %counts all the elements in the dtm %measure the size and sign of terrain excursions h=dtm(dtm_index)-dtm(dtm_index-1); %for terrain transitions, interpolate if h~=0&do_interp for h_step=0:h*terrain_scale:h TERRAIN_PROFILE(terrain_index)=dtm(dtm_index-1)+h_step; terrain_index=terrain_index+1; end %when no transitions just oversample at set terrain height else; for count=1:1/terrain_scale TERRAIN_PROFILE(terrain_index)=dtm(dtm_index-1); terrain_index=terrain_index+1; end end end%when resolutions are the same just load the dtm into the terrain profileelse TERRAIN_PROFILE=dtm;endx_max=length(TERRAIN_PROFILE)*range_step;
![Page 356: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/356.jpg)
C-29
C.2 GPS Receiver Modelling
(a) RUNCORR.M
%***************************************************************************% RUNCORR – control file for rxcorr%% Bruce Hannah% Version date 14/8/00%***************************************************************************
%clear;tic
%***************************************************************************% option selections for correlation and display%***************************************************************************
save_images=0;filter_on=0;corr_select=0; % 0=standard 1 chip spacing % 1=narrow 0.1 chip spacingplot_select=11; % 0=no plotting % 1=plot prompt correlation % 2=plot early correlation % 3=plot late correlation % 4=plot LOS correlation % 5=plot MP correlation % 6=plot early & late % 7=plot prompt, early & late % 8=plot prompt, LOS & MP % 9=plot discriminator function % 10=plot prompt & discriminator % 11=plot prompt, LOS, MP & discriminatordiscrim_type=3; % 1=dot product (E-L)P [coherent & noncoherent] % 2=early minus late power (E^2-L^2) [noncoherent] % 3=early minus late envelope (E-L) [noncoherent] % 4=early minus late envelope normalised (E-L)/(E+L) [noncoherent]
%***************************************************************************% defaults%***************************************************************************
C=3e8; % speed of propagation of the C/A-codeprn_num=3; % prn number of the c/a codeca_code_period=1e-3; % c/a code is 1 ms long
filter_order=2; % filter order; for filtfilt order is twice this
cutoff_freq=1e6; % 2 MHz precorrealtion filter for standard correlatorcorr_spacing=0.5; % standard 1 chip (1/2 chip early and 1/2 chip late) % correlator spacingchip_samples=40; % this is the resample rate for the C/A-code with % the standard correlatorsample_mult=1;lag_chips=2; % the lag in chips for the correlation function display
if corr_select % if narrow correlator sample_mult=2; cutoff_freq=8e6; % 8 MHz pre-correlation filter for narrow correlator corr_spacing=0.05; % 1/10th chip narrow correlator spacing % (1/20 chip early and 1/20 chip late) % resample rate of C/A-code for narrow is 10x that for standard chip_samples=chip_samples*sample_mult;
![Page 357: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/357.jpg)
C-30
lag_chips=0.1; % the lag in chips for the correlation functiondisplayend
%***************************************************************************% the rounded lag in integer samples for the correlation function%***************************************************************************
lag=round(lag_chips*chip_samples);
sample_time=1e-3/(1023*chip_samples);sample_freq=1/sample_time;
if ~exist(’prompt_code’)
%***************************************************************************% generate the (1023 chip x chip_samples) sampled c/a code from pngen.m%***************************************************************************
prompt_code=pngen(prn_num,chip_samples);end
%***************************************************************************% set up filtering%***************************************************************************
if filter_on filt_cutoff=cutoff_freq/(sample_freq/2);else filt_cutoff=1; % no filtering of codeend
[B,A]=butter(filter_order,filt_cutoff);
%***************************************************************************% lowpass filter the prompt C/A-code to replicate the receiver% precorrelation filtering.%***************************************************************************
filtered_prompt_code=filtfilt(B,A,prompt_code);
clear B A filt_cutoff sample_freq % free up memory
%***************************************************************************% set up for saving images%***************************************************************************
if save_images % save image option ON prntype=’-dbitmap’; % set saved image format dirpath= ’c:\bruce\PEfiles\movfiles\’; % set path dirname=[’corr\run3’]; % directory name for storing images if ~exist([dirpath,dirname],’dir’) % check if directory already exists eval([’!mkdir ’,dirpath,dirname]); % if not make directory end eval([’cd ’,dirpath,dirname]); % change to new directory to save imagesend
%***************************************************************************% default delay parameters% note: rxcorr accepts unlimited array of multipath signals%***************************************************************************
% the relative amplitude of the multipath code to the LOS codeMP_delay_amp=[0.5];
% the delay time of the multipath code to the LOS codeMP_delay_time=[5*sample_time];
% phase of multipath code in degreesMP_delay_phase=[0];
![Page 358: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/358.jpg)
C-31
index=1;
%***************************************************************************% run rxcorr for set parameters%***************************************************************************
for MP_delay_time=0:10*sample_time*sample_mult:1600e-9 [discrim,tau_error,e_p_l]=... rxcorr(MP_delay_time,MP_delay_amp,MP_delay_phase,LOS_delay_time,... LOS_delay_amp,prompt_code,filtered_prompt_code,chip_samples,... sample_time,corr_spacing,discrim_type,lag,plot_select);
range_error(diff_index,time_index)=C*tau_error; e_p_l_sample(time_index,:)=e_p_l;
%***************************************************************************% range error and early,late and prompt sample data%***************************************************************************
range_error(index)=C*tau_error; e_p_l_sample(index,:)=e_p_l;
if save_images filename=[’ mp_out_time2_’,num2str(index)]; eval([’print ’,prntype,filename]); close end index=index+1;end
toc
![Page 359: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/359.jpg)
C-32
(b) RXCORR.M
%***************************************************************************% RXCORR – receiver correlation/discrimnation function% Called by runcorr.m% delay_time,MP_delay_amp,MP_delay_phase are the relative multipath% parameters. Prompt_code is the generated C/A-code for a given PRN number.%% Bruce Hannah% version date 22/8/00%***************************************************************************
function [discrim,tau_error,e_p_l]=... rxcorr2(MP_delay_time,MP_delay_amp,MP_delay_phase,... LOS_delay_time,LOS_delay_amp,prompt_code,... filtered_prompt_code,chip_samples,sample_time,... corr_spacing,discrim_type,lag,plot_select)
corr_spacing_samples=corr_spacing*chip_samples;
zero_lag_sample=lag+1; % the sample number of the zero lag correlation
%***************************************************************************% early and late sample numbers%***************************************************************************
early_sample=zero_lag_sample-(corr_spacing*chip_samples);late_sample=zero_lag_sample+(corr_spacing*chip_samples);
%***************************************************************************% generate the early and late local reference codes%***************************************************************************
early_code=[prompt_code(corr_spacing_samples+1:length(prompt_code)),... prompt_code(1:corr_spacing_samples)];
late_code=[prompt_code(length(prompt_code)-... corr_spacing_samples+1:length(prompt_code)), prompt_code(1:length(prompt_code)-corr_spacing_samples)];
%***************************************************************************% the LOS, multipath delayed, and attenuated replicas are generated% the delayed code is addition of delayed replicas of the prompt (LOS) code%***************************************************************************
LOS_code=filtered_prompt_code; %initialise LOS as the unshifted prompt
if LOS_delay_time<0
LOS_delay_chips=-LOS_delay_time/(1e-3/1023); LOS_delay_samples=round(LOS_delay_chips*chip_samples); LOS_code=LOS_delay_amp*[filtered_prompt_code(LOS_delay_samples+... 1:length(filtered_prompt_code)),filtered_prompt_code(... 1:LOS_delay_samples)];% the LOS code arrives earlier
elseif LOS_delay_time>0
LOS_delay_chips=LOS_delay_time/(1e-3/1023); LOS_delay_samples=round(LOS_delay_chips*chip_samples); LOS_code=LOS_delay_amp*[filtered_prompt_code(length(filtered_prompt_... code)-LOS_delay_samples+1:length(filtered_prompt_code)),... filtered_prompt_code(1:length(filtered_prompt_code)-... LOS_delay_samples)]; % the LOS code arrives laterend
![Page 360: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/360.jpg)
C-33
%***************************************************************************% initialise the Multipath code as all zero%***************************************************************************
MP_code=zeros(size(LOS_code));
%***************************************************************************% generate the Multipath code as the linear combination of all multipath% signals%***************************************************************************
for index=1:length(MP_delay_time)
MP_delay_chips=MP_delay_time(index)/(1e-3/1023);
% the delay in samples for the multipath code MP_delay_samples=round(MP_delay_chips*chip_samples); MP_code=MP_code+(cos(MP_delay_phase(index)*pi/180)*... MP_delay_amp(index)*[LOS_code(length(LOS_code)-... MP_delay_samples+1:length(LOS_code)),... LOS_code(1:length(LOS_code)-MP_delay_samples)]);end
%***************************************************************************% generate the autocorrelation function% and normalise for display%***************************************************************************
%***************************************************************************% this is the cross-correlation of the prompt filtered c/a code with the% unfiltered reference%***************************************************************************
corr_auto=xcov(prompt_code,filtered_prompt_code,lag);
%***************************************************************************% the reference autocorrelation coeff at the zeroth lag%***************************************************************************
norm_corr_coeff=corr_auto(zero_lag_sample);
clear corr_auto
%***************************************************************************% the true input code to the receiver is now the addition of the LOS% and the delayed MP codes%***************************************************************************
input_code=LOS_code+MP_code;
%***************************************************************************% ontime correlation of the combined input code%***************************************************************************
corr_prompt=xcov(prompt_code,input_code,lag)/norm_corr_coeff;
%***************************************************************************% late correlation of the combined input code%***************************************************************************
corr_late=(xcov(late_code,input_code,lag)/norm_corr_coeff);
%***************************************************************************% early correlation of the combined input code%***************************************************************************
corr_early=(xcov(early_code,input_code,lag)/norm_corr_coeff);
![Page 361: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/361.jpg)
C-34
%***************************************************************************% LOS correlation%***************************************************************************
corr_LOS=xcov(prompt_code,LOS_code,lag)/norm_corr_coeff;
%***************************************************************************% correlation of the delayed multipath code%***************************************************************************
corr_MP=xcov(prompt_code,MP_code,lag)/norm_corr_coeff;
e_p_l=[corr_prompt(early_sample) corr_prompt(zero_lag_sample) ... corr_prompt(late_sample)];
%***************************************************************************% selected receiver discrimination%***************************************************************************
if discrim_type==1 discrim=(corr_early-corr_late).*corr_prompt; discrim_text=’Dot Product’;elseif discrim_type==2 discrim=(corr_early.^2-corr_late.^2); discrim_text=’E-L Power’;elseif discrim_type==3 discrim=(corr_early-corr_late); discrim_text=’E-L Envelope’;elseif discrim_type==4 discrim=(corr_early-corr_late)./(corr_early+corr_late); discrim_text=’E-L Normalised Envelope’;elseif discrim_type==5 discrim=(corr_early+corr_late)-corr_prompt; discrim_text=’NEW BIASED’;end
%***************************************************************************% plot selection and set up%***************************************************************************
if plot_select if plot_select==4 min_corr=round(min(corr_LOS)*11)/10; max_corr=round(max(corr_LOS)*11)/10; elseif (plot_select==5) min_corr=round(min(corr_MP)*11)/10; max_corr=round(max(corr_MP)*11)/10; elseif (plot_select==8)|(plot_select==11)|(plot_select==12) min_corr=round(min([min(corr_LOS) min(corr_MP)min(corr_prompt)])*11)/10; max_corr=round(max([max(corr_LOS) max(corr_MP)max(corr_prompt)])*11)/10; else min_corr=round(min(corr_prompt)*11)/10; max_corr=round(max(corr_prompt)*11)/10; endend
max_discrim=round(max(discrim)*11)/10;
min_corr=-2;max_corr=2;max_discrim=2;
![Page 362: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/362.jpg)
C-35
%***************************************************************************% track zero crossing of discrimination function%***************************************************************************
pos_error_sample=min(find(discrim(early_sample:late_sample)>0))+... early_sample-1; % the first sample that is positiveneg_error_sample=max(find(discrim(early_sample:late_sample)<0))+... early_sample-1; % the last sample that is negative
%***************************************************************************% timing error measurement%***************************************************************************
if (discrim(zero_lag_sample)<1e-6)&(discrim(zero_lag_sample)>-1e-6) tau_error=0;elseif neg_error_sample>=zero_lag_sample tau_error=sample_time*((neg_error_sample-zero_lag_sample)... +(discrim(neg_error_sample)/(discrim(neg_error_sample)-discrim(pos_error_sample))));elseif pos_error_sample<=zero_lag_sample tau_error=-sample_time*((zero_lag_sample-pos_error_sample)... +(discrim(pos_error_sample)/(discrim(pos_error_sample)-discrim(neg_error_sample))));end
%***************************************************************************% plot results if required%***************************************************************************
if plot_select
plot_prompt=’plot((-lag:lag)/chip_samples,corr_prompt,’’b’’);’; plot_early=’plot((-lag:lag)/chip_samples,corr_early,’’r’’);’; plot_late=’plot((-lag:lag)/chip_samples,corr_late,’’g’’);’; plot_discrim=’plot((-lag:lag)/chip_samples,discrim,’’b’’);’; plot_LOS=’plot((-lag:lag)/chip_samples,corr_LOS,’’g’’);’; plot_MP=’plot((-lag:lag)/chip_samples,corr_MP,’’r’’);’; corr_axes=’axis([-lag/chip_samples lag/chip_samples min_corr ... max_corr]);’; discrim_axes=’axis([-lag/chip_samples lag/chip_samples ... -max_discrim max_discrim]);’; corr_title=’title(’’Correlation Function’’);’; discrim_title=’title([discrim_text,’’ Discriminator’’]);’; x_text=’xlabel(’’Offset (chips)’’);’; y_corr=’ylabel(’’Normalised Correlation’’);’; y_discrim=’ylabel(’’Normalised Discrimination’’);’;
if plot_select==1 % prompt plot_control=[plot_prompt,corr_axes,corr_title,x_text,y_corr,... ’grid on;’];
elseif plot_select==2 % early plot_control=[plot_early,corr_axes,corr_title,x_text,y_corr,... ’grid on;’];
elseif plot_select==3 % late plot_control=[plot_late,corr_axes,corr_title,x_text,y_corr,... ’grid on;’];
elseif plot_select==4 % LOS plot_control=[plot_LOS,corr_axes,corr_title,x_text,y_corr,’grid on;’];
elseif plot_select==5 % MP plot_control=[plot_MP,corr_axes,corr_title,x_text,y_corr,’grid on;’];
elseif plot_select==6 % early & late plot_control=[plot_early,’hold on;’,plot_late,corr_axes,corr_title,... x_text,y_corr,’grid on;legend(’’E’’,’’L’’);’];
elseif plot_select==7 % early, late & prompt plot_control=[plot_early,’hold on;’,plot_late,plot_prompt,...
![Page 363: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/363.jpg)
C-36
corr_axes,corr_title,x_text,y_corr,’grid on;legend... (’’E’’,’’L’’,’’P’’);’];
elseif plot_select==8 % LOS, MP & prompt plot_control=[plot_LOS,’hold on;’,plot_MP,plot_prompt,corr_axes,... r_title,x_text,y_corr,’grid on;legend(’’LOS’’,’’MP’’,’’P’’);’];
elseif plot_select==9 % discriminator plot_control=plot_discrim;
elseif plot_select==10 % prompt & discriminator plot_control=[’subplot(2,1,1);’,plot_prompt,corr_axes,corr_title,... y_corr,’grid on;’,’subplot(2,1,2);’,plot_discrim,discrim_axes,... discrim_title,x_text,y_discrim,’grid on;’];
elseif plot_select==11 % LOS, MP, prompt & discriminator plot_control=[’subplot(2,1,1);’,plot_LOS,’hold on;’,plot_MP,... plot_prompt,corr_axes,corr_title,y_corr,’grid on;legend... (’’LOS’’,’’MP’’,’’P’’);’,’subplot(2,1,2);’,’hold on;’... plot_discrim,discrim_axes,discrim_title,x_text,y_discrim,... ’grid on;’];
elseif plot_select==12 % early, late & discriminator plot_control=[’subplot(2,1,1);’,plot_early,’hold on;’,plot_late,... corr_axes,corr_title,y_corr,’grid on;legend... (’’E’’,’’L’’);’,’subplot(2,1,2);’,’hold on;’... plot_discrim,discrim_axes,discrim_title,x_text,y_discrim,... ’grid on;’];
end
eval(plot_control);
end
![Page 364: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/364.jpg)
C-37
(c) PNGEN.M
%***************************************************************************% PNGEN – C/A-code generationo function% Called by runcorr.m
% 1023 bit C/A code generator function% input:% prn(SV) number via prn_num% chip_sampling via chip_samples% output:% sampled C/A code via ca_code% generates the 1023 bit C/A code available in prn_code% saves prn_code to MAT file if save_code flag set%% Bruce Hannah% version date 22/8/00%***************************************************************************
function ca_code=pngen(prn_num,chip_samples)
save_code=0;
N=1023; % number of bitsM=10; % number of shift registersca_code=[];
if prn_num<5 % prn phase tap lookup table S=[prn_num+1,prn_num+5];elseif prn_num<7 S=[prn_num-4,prn_num+4];elseif prn_num<10 S=[prn_num-6,prn_num+1];elseif prn_num<17 S=[prn_num-8,prn_num-7];elseif prn_num<23 S=[prn_num-16,prn_num-13];elseif prn_num<24 S=[prn_num-22,prn_num-20];elseif prn_num<29 S=[prn_num-20,prn_num-18];elseif prn_num<34 S=[prn_num-28,prn_num-23];end
% initialise registers
for k=1:M reg_g1(k)=1; % initialise G1 generator registers reg_g2(k)=1; % initialise G2 generator registersend
%***************************************************************************% see p14-17 GPS SPS signal specification for details%***************************************************************************
for clock=1:N input_g1=xor(reg_g1(3),reg_g1(10)); input_g2=xor(xor(xor(reg_g2(2),reg_g2(3)), xor(reg_g2(6),reg_g2(8))),xor(reg_g2(9),reg_g2(10))); G1(clock)=reg_g1(10); % output of G1 stored G2(clock)=reg_g2(10); % output of G2 stored G2i(clock)=xor(reg_g2(S(1)),reg_g2(S(2))); % phase selector output stored prev_reg_g1=reg_g1; % store current G1 register states prev_reg_g2=reg_g2; % store current G2 register states for k=2:M reg_g1(k)=prev_reg_g1(k-1); % shift registers 2-10 of G1 reg_g2(k)=prev_reg_g2(k-1); % shift registers 2-10 of G2 end
![Page 365: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/365.jpg)
C-38
reg_g1(1)=input_g1; % register 1 bit set from given input reg_g2(1)=input_g2;end
prn_code=xor(G1,G2i); % prn code generated from G1 and G2i
for k=1:N if prn_code(k)==0 prn_code(k)=1; % map code from [0,1] to [1,-1] else prn_code(k)=-1; end for n=1:chip_samples chip_code(k,n)=prn_code(k); end ca_code=[ca_code,chip_code(k,:)];end
%***************************************************************************% code saved to MAT file%***************************************************************************
if save_code~=0 var_name=’prn_code’; filename=[’prn’,num2str(prn_num),’code’]; save(filename,var_name);end
![Page 366: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/366.jpg)
C-39
C.3 GPS Multipath Data Acquisition
(a) ASHTEQC.M
%***************************************************************************% ASHTEQC Converts ashtech download files (B,E) to rinex using teqc% The location of teqc.exe must be specified in ’teqc_dir’% The location of the data directory must be specified in ’data_dir’% Ashtech B, E files must exist in a directory named ’ash_files’% If not already existing a directory is created called ’rinex_files’% The resultant obs and/or nav rinex files are placed here%% User specifies observables and file creation options (nav or obs or both)%% Bruce Hannah% version date 3/2/00%***************************************************************************
clear
teqc_dir=’g:\teqc\’; % location of teqc.exedata_dir=’g:\BMH_data\1_04_00\’; % set the path where data is locatedash_dir=[data_dir,’ash_files’];rinex_dir=[data_dir,’rinex_files\’];
eval([’cd ’,ash_dir]); % change to existing ashtech file directory
disp(’convert Ashtech files to RINEX format’)disp(’’)
while 1
if ~exist(rinex_dir,’dir’) % check if directory already exists eval([’!mkdir ’,rinex_dir]); % if not make rinex directory end
l1=’’; % initilise strings l2=’’; c1=’’; p1=’’; p2=’’; d1=’’; d2=’’; s1=’’; s2=’’; nav=’’; n_file=’’; obs=’’; obs_str=’’;
dir
disp(’option 1: obs + nav’) disp(’option 2: obs only’) disp(’option 3: nav only’) opt=input(’enter option number (q to quit): ’,’s’);
if strcmp(opt,’q’) % quit break else b_file=input(’enter B-file name: ’,’s’); out_file=input(’enter output file name (no extension): ’,’s’); opt=str2num(opt); if opt==1|opt==2 % if obs required disp(’enter observables (1 include 0 exclude)’) % user includes or excludes the observations from % the command line string
![Page 367: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/367.jpg)
C-40
if input(’L1: ’) l1=’L1+’; end if input(’L2: ’); l2=’L2+’; end if input(’C1: ’); c1=’C1+’; end if input(’P1: ’); p1=’P1+’; end if input(’P2: ’); p2=’P2+’; end if input(’D1: ’); d1=’D1+’; end if input(’D2: ’); d2=’D2+’; end if input(’S1: ’); s1=’S1+’; end if input(’S2: ’); s2=’S2+’; end % command line string for observations obs_str=[l1,l2,c1,p1,p2,d1,d2,s1,s2]; obs_str(length(obs_str))=’ ’; % replace last ’+’ with a space obs=’ -O.obs ’; % obs option file_ext=’.98o’; % obs file extension if opt==1 % do both nav and obs nav=’ +nav ’; % include nav option n_file=[out_file,’.98n’]; % nav file name end
else % nav file conversion only nav=’n ’; file_ext=’.98n ’; end eval([’!’,teqc_dir,’teqc -ash d’,nav,rinex_dir,n_file,obs,obs_str,... b_file,’ > ’,rinex_dir,out_file,file_ext])
endend
![Page 368: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/368.jpg)
C-41
(b) RIN2QC.M
%***************************************************************************% RIN2QC Teqc Quality Check that creates UNAVCO compact files from rinex% data files. The location of teqc.exe must be specified in ’teqc_dir’% The location of the data directory must be specified in ’data_dir’% Rinex files must exist in a directory named ’rinex_files’% If not already existing a directory is created called ’qc_files’% where the resultant qc files are saved%% Bruce Hannah% version date 3/2/00%***************************************************************************
clear
data_dir=’g:\BMH_data\1_04_00\’; % specify data location pathteqc_dir=’g:\teqc\’; % specify location of teqc.exe
rinex_dir=[data_dir,’rinex_files\’]; % rinex directory pathqc_dir=[data_dir,’qc_files\’]; % qc directory path
if ~exist(qc_dir,’dir’) % check if qc directory already exists eval([’!mkdir ’,qc_dir]); % if not make qc directoryend
eval([’cd ’,rinex_dir]); % change to rinex directory
while 1 dir % display rinex files f_name=input(’enter RINEX obs filename (q to quit): ’,’s’);
if strcmp(f_name,’q’) break else eval([’!g:\teqc\teqc +qc ’,f_name]);
% move the qc files to the qc directory
qc_name=f_name(1:length(f_name)-4);
if exist([qc_name,’.azi’],’file’) eval([’!copy ’,qc_name,’.azi ’,qc_dir]); eval([’!del ’,qc_name,’.azi’]); end
if exist([qc_name,’.ele’],’file’) eval([’!copy ’,qc_name,’.ele ’,qc_dir]); eval([’!del ’,qc_name,’.ele’]); end
if exist([qc_name,’.mp1’],’file’) eval([’!copy ’,qc_name,’.mp1 ’,qc_dir]); eval([’!del ’,qc_name,’.mp1’]); end
if exist([qc_name,’.mp2’],’file’) eval([’!copy ’,qc_name,’.mp2 ’,qc_dir]); eval([’!del ’,qc_name,’.mp2’]); end
if exist([qc_name,’.sn1’],’file’) eval([’!copy ’,qc_name,’.sn1 ’,qc_dir]); eval([’!del ’,qc_name,’.sn1’]); end
if exist([qc_name,’.sn2’],’file’) eval([’!copy ’,qc_name,’.sn2 ’,qc_dir]); eval([’!del ’,qc_name,’.sn2’]);
![Page 369: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/369.jpg)
C-42
end
if exist([qc_name,’.ion’],’file’) eval([’!copy ’,qc_name,’.ion ’,qc_dir]); eval([’!del ’,qc_name,’.ion’]); end
if exist([qc_name,’.iod’],’file’) eval([’!copy ’,qc_name,’.iod ’,qc_dir]); eval([’!del ’,qc_name,’.iod’]); end
endend
![Page 370: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/370.jpg)
C-43
(c) QC2MAT.M
%***************************************************************************% QC2MAT Convert UNAVCO compact plot files to matlab data format% User will be prompted for a teqc UNAVCO format file% Data is then loaded from UNAVCO compact format% and saved in matlab .mat format for later plotting using plotqc.mat%% The location of the data directory must be specified in ’data_dir’% QC files must exist in a directory named ’qc_files’% If not already existing a directory is created called ’mat_files’% where the resultant files are saved%% The variables ’xxx_data’, ’total_sats’, ’epoch’, ’t_samp’, ’start-time’% are stored in the .mat file. xxx indicates the data type stored% for example, azi file data will be stored in variable ’azi_data’%% Bruce Hannah% version date 7/2/00%***************************************************************************
clear
dirpath=’g:\BMH_data\1_04_00\’; % set path
qc_dir=[dirpath,’qc_files\’]; % qc directory pathmat_dir=[dirpath,’mat_files\’]; % mat directory path
if ~exist(mat_dir,’dir’) % check if directory already exists eval([’!mkdir ’,mat_dir]); % if not make mat directoryend
while 1 fid=-1; eval([’cd ’,qc_dir]); % change to qc directory dir % display qc files f_name=input(’enter qc file name (q to quit): ’,’s’);
if (strcmp(f_name,’q’)) break else f_front=f_name(1:length(f_name)-4); f_ext=f_name(length(f_name)-2:length(f_name)); [fid,message]=fopen(f_name,’rt’); disp(message); end
if fid~=-1 line=fgetl(fid); % read in the first header line s1=sscanf(line,’%s’);
if strcmp(s1,’COMPACT’) % check for UNAVCO file compatibility compact_format=1; else disp(’this file is not UNAVCO compact format’) fclose(fid); compact_format=0; end
if compact_format % if a compatible file
line=fgetl(fid); % read in the second header line (not used) % read in the third header line (time sample rate) line=fgetl(fid); t_samp=sscanf(line,’%*s %g’); % get the sample time line=fgetl(fid); % read the fourth header line start_time=sscanf(line,’%*s %g’); % get the start time % all other lines are sat or observation data
![Page 371: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/371.jpg)
C-44
epoch=1; % initialise epoch count counter=1; % initialise decade counter sat_line=1; % initialise sat line flag true obs_line=0; % initialise obs line flag false visible_sats(1:32)=0; % initialise array of visible_sats line=fgetl(fid); % go to first observation line
while 1 % while not the end of file if sat_line % look for sats % read in the number of sats and sv numbers sat_info=sscanf(line,’%d’);
if sat_info(1)==0 % if there are no sats % set data for all sats as no data (non number) sv_data(1:32,epoch)=nan; line=fgetl(fid); % move to the next line if ~isstr(line) % if end of the file, stop reading file break end sat_line=1; % set flag to check for sat data again epoch=epoch+1; % increment epoch counter else % otherwise there are sats sat_line=0; % sat line flag is reset obs_line=1; % observation line flag is set num_sats=sat_info(1); % number of sats is read for y=2:num_sats+1 % satellite order is read sat_order(y-1)=sat_info(y); % and stored in sat_order % for each sat line read flag each visible sat visible_sats(sat_order(y-1))=1; end sat_info=0; % reset sat_info for next sat line read end
end
if obs_line % if there is a line of observation data % initialise data for all sats as no data (non number) sv_data(1:32,epoch)=nan; line=fgetl(fid); % go to the observation line obs=sscanf(line,’%f’); % read the observations for y=1:num_sats % for each sat % store data in variable sv sv_data(sat_order(y),epoch)=obs(y); end obs_line=0; % reset observation line flag line=fgetl(fid); % go to the next line
if ~isstr(line) % if end of the file, stop reading file break else % else the line is a repeat obs flag ornew sat info flag=sscanf(line,’%d’); % read the line if flag==-1 % if it’s a data flag obs_line=1; % there are more obs for these sats % otherwise there’s a new satellite configuration else % set the sat line flag to read the new sat info sat_line=1; end end
epoch=epoch+1; % increment the epoch counter end
if epoch/100==counter disp([num2str(100*counter),’ epochs of ’,... f_name,’ processed’]); counter=counter+1; end
end disp([num2str(epoch),’ total epochs processed’]);
![Page 372: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/372.jpg)
C-45
fclose(fid); total_sats=find(visible_sats); if strcmp(f_ext,’azi’) data_type=’azi_data’; azi_data=sv_data; elseif strcmp(f_ext,’ele’) data_type=’ele_data’; ele_data=sv_data; elseif strcmp(f_ext,’ion’) data_type=’ion_data’; ion_data=sv_data; elseif strcmp(f_ext,’iod’) data_type=’iod_data’; iod_data=sv_data; elseif strcmp(f_ext,’mp1’) data_type=’mp1_data’; mp1_data=sv_data; elseif strcmp(f_ext,’mp2’) data_type=’mp2_data’; mp2_data=sv_data; elseif strcmp(f_ext,’sn1’) data_type=’sn1_data’; sn1_data=sv_data; elseif strcmp(f_ext,’sn2’) data_type=’sn2_data’; sn2_data=sv_data; end % save info in MAT file eval([’save ’,mat_dir,f_front,’_’,f_ext,’ ’,data_type,... ’ total_sats epoch t_samp start_time’]) end endend
![Page 373: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/373.jpg)
C-46
(d) PLOTQC.M
%***************************************************************************
% PLOTQC Plot azi,ele,sn1,sn2,mp1,mp2,ion,iod qc’d data% PLOTQC Loads and plots data from MATLAB .mat file format% The location of the data directory must be specified in ’data_dir’% The mat files must exist in a directory named ’mat_files’%% Bruce Hannah% version date 18/1/00%***************************************************************************
data_dir=’g:\BMH_data\30_11_00\’; % set data directory patheval([’cd ’,data_dir,’mat_files’]); % change to mat file directory
while 1 clear % clear all variables dir
f_name=input(’enter mat filename ("q" to quit): ’,’s’); if strcmp(f_name,’q’) break else f_ext=f_name(length(f_name)-2:length(f_name)); eval([’load ’,f_name],[’disp(lasterr)’]) end
data_type=[f_ext,’_data’]; % thsi is for early data
if exist(’sv_data’)==1 data_type=’sv_data’; end
if exist(data_type)==1 disp([’data for SV: ’,num2str(total_sats)]) disp(’type "r" to return to top, "a" to plot all, "sv" for sat data’)
while 1 % plot routine until user quits using ’q’ key_in=input(’: ’,’s’); % keyboard prompt line for display
if ~(strcmp(key_in,’r’)|strcmp(key_in,’a’)|strcmp(key_in,’sv’)) key_in=str2num(key_in); if (key_in>=1)&(key_in<=32) eval([’plot(’,data_type,’(key_in,:))’]); sv_text=num2str(key_in); plot_flag=1; else disp(’not a correct SV number, or command’) plot_flag=0; end elseif strcmp(key_in,’a’) line_hdl=plot(1:epoch,eval(data_type)); legend(line_hdl(total_sats),num2str(transpose(total_sats))); sv_text=num2str(total_sats); plot_flag=1; elseif strcmp(key_in,’sv’) disp([’SV: ’,num2str(total_sats)]) elseif strcmp(key_in,’r’) break end
if plot_flag xlabel(’epochs’) if strcmp(f_ext,’azi’) title([’Satellite Azimuth SV ’,sv_text]) ylabel(’Azimuth (degs)’)
![Page 374: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/374.jpg)
C-47
axis([0 epoch -360 360]); elseif strcmp(f_ext,’ele’) title([’Satellite Elevation SV ’,sv_text]) ylabel(’Elevation (degs)’) axis([0 epoch 0 90]); elseif strcmp(f_ext,’ion’) title([’L2 Ionospheric Observable SV ’,sv_text]) ylabel(’Range Error (m)’) elseif strcmp(f_ext,’iod’) title([’Derivative of L2 Ionospheric ... Observable SV ’,sv_text]) ylabel(’Range Error (m)’) elseif strcmp(f_ext,’mp1’) title([’L1 Multipath SV ’,sv_text]) ylabel(’Range Error (m)’) axis([0 epoch -5 5]); elseif strcmp(f_ext,’mp2’) title([’L2 Multipath SV ’,sv_text]) ylabel(’Range Error (m)’) axis([0 epoch -5 5]); elseif strcmp(f_ext,’sn1’) title([’L1 S/N SV ’,sv_text]) ylabel(’S/N (dB)’) axis([0 epoch 0 60]); elseif strcmp(f_ext,’sn2’) title([’L2 S/N SV ’,sv_text]) ylabel(’S/N (dB)’) axis([0 epoch 0 60]); end zoom on end end endend
![Page 375: Modelling and Simulation of GPS Multipath Propagationmultipath have proven much more difficult to mitigate. The research aim of this work is to enhance the understanding of multipath](https://reader031.fdocuments.us/reader031/viewer/2022011913/5faaa201e144e853d00d33d5/html5/thumbnails/375.jpg)