University of Southern California Dissertations and Theses

Double-directional channel sounding for next generation wireless communications

(USC Thesis Other)

Double-directional channel sounding for next generation wireless communications

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content DOUBLE-DIRECTIONAL CHANNEL SOUNDING FOR NEXT GENERATION WIRELESS
COMMUNICATIONS
by
Hussein Hammoud
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ELECTRICAL ENGINEERING)
August 2024
Copyright 2024 Hussein Hammoud

Acknowledgements
First and foremost, I would like to express my deepest gratitude to my Ph.D. research advisor, Dr. Andreas
F. Molisch, for his unwavering support and guidance since my first day at USC. I am also deeply thankful
to my qualification exam and dissertation committee members: Dr. Alan Willner, Dr. Keith Chugg, Dr.
Leana Golubchik, Dr. Constantine Sideris, and Dr. Yue Wang, for their invaluable insights and support.
I extend my heartfelt thanks to my fellow research group members at USC who have contributed to
the work presented in this thesis: Guillermo Castro, Jorge Gomez Ponce, Bassel Abou Ali Modad, Naveed
Ahmed Abbasi, Thomas Choi, Yuning Zhang, Zihang Cheng, Celalettin Umit Bas, Rui Wang, and Seun
Sangodoyin. It has been a pleasure working with all of you. Additionally, I am grateful to our administrative staff: Susan Wiedem, Gerrielyn Ramos, Corine Wong, and Diane Demetras, for their assistance
throughout my Ph.D. journey.
I would like to acknowledge our collaborators from the Austrian Institute of Technology, the Technical
University of Wien, and the Technical University of Brno, namely Markus Hoder, Faruk Pasic, Radek
Zavorka, Herbert Groll, Dr. Ales Prokes, Dr. Thomas Zemen, and Dr. Christoph Mecklenbrauker. Working
with you has been a great pleasure, and I look forward to future collaborations.
My heartfelt thanks go to my friends in Los Angeles: Hasan Hamad, Mohamad Awada, Michella Rustom, Alexios Rustom, Feng Ling, Akash Panda, and Javier Bourquez. Despite the demanding schedules
and the challenging work-life balance of our Ph.D. programs, your support and friendship have meant the
world to me.
ii

I am deeply grateful to my family: my parents, Sanaa and Ali, my sisters Nour, Rim, and Ghinwa, my
brother Mohammad, and my parents-in-law, Jessica and Cruz, as well as my brothers-in-law, Mostafa, Ali,
Cruz, CJ, and Caleb, for their unconditional love and support. Lastly, I would like to thank my beloved
Alyssa. Your love, support, and patience have encouraged me to pursue my dreams. I am forever grateful
to have you in my life.
iii

Table of Contents
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Chapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivation and Trends in Wireless Communications . . . . . . . . . . . . . . . . . . . . . 1
1.2 Overview of Wireless Channels and Impact on System Design . . . . . . . . . . . . . . . . 5
1.2.1 Wireless propagation phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.2 Condensed channel parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Overview of Channel Sounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Channel sounding: What and why? . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.2 Time-domain vs frequency-domain sounders . . . . . . . . . . . . . . . . . . . . . 12
1.3.3 SISO vs MIMO sounders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Evaluation Methods: Fourier, MUSIC, SAGE . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Thesis Outline and Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5.1 Real-Time Millimeter-Wave Double-Directional Channel Sounder . . . . . . . . . . 16
1.5.2 ReRoMA-based Vehicle-to-Vehicle Measurement . . . . . . . . . . . . . . . . . . . 17
1.5.3 Sub-GHz Measurements and Modelling . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Other Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Chapter 2: Real-Time Millimeter-Wave Double-Directional Channel Sounder . . . . . . . . . . . . 20
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 Main Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Sounder Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.1 ReRoMA Mechanical Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.2 Prototype mm-wave sounder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Calibration Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4.1 Antenna and system calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4.2 Time stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.3 Comparison to stepper motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.5 Evaluation Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.6 Reference Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.6.1 Dynamic sphere reflector measurements . . . . . . . . . . . . . . . . . . . . . . . . 43
iv

2.6.2 Rx moving away from Tx measurement . . . . . . . . . . . . . . . . . . . . . . . . 47
2.6.3 Cart-to-cart T-intersection scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Chapter 3: ReRoMA-based Vehicle-to-Vehicle Measurement . . . . . . . . . . . . . . . . . . . . . . 54
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Main Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3 Measurement Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.1 Driving in convoy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.2 Driving on opposite lanes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3.3 Overtaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.4 Evaluation Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.5.1 Power delay profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.5.2 Angular power profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.3 Pathloss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.5.4 RMS delay spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.5.5 Angular spreads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.5.6 Power distribution among MPCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5.7 Stationarity time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Chapter 4: Sub-GHz Measurements and Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.2 Main Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3 Sounder Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.3.1 Antenna Array . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.3.2 RF Chain Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.3.3 Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.4 Measurement Campaign . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5 Processing and Evaluations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.5.1 Path loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.5.2 RMS Delay spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.5.3 Angular spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.5.4 Power distribution over MPC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.6.1 Power delay profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.6.2 Angular power spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.6.3 Path loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.6.4 RMS delay spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.6.5 Angular spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.6.6 Power distribution of MPCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.7 Sidelink Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.7.1 Sidelink background and Diversity methods . . . . . . . . . . . . . . . . . . . . . . 113
4.7.2 Simulation Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.7.3 Sidelink sample simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.7.4 Diversity gains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.7.5 Comparison to standard channel models . . . . . . . . . . . . . . . . . . . . . . . . 119
v

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
vi

List of Tables
2.1 Sounder parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1 Comparison of our results to existing mm-wave literature . . . . . . . . . . . . . . . . . . . 78
4.1 Pathloss mean and standard deviation for all floors across the different scenarios . . . . . . 102
4.2 P Lo2i
linear modelling parameters and the corresponding confidence intervals . . . . . . 105
4.3 Shadowing parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.4 RMS delay spread mean and standard deviation for all floors across the different scenarios 107
4.5 Azimuth of Arrival spread mean and standard deviation for all floors across the different
scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.6 Azimuth of Departure spread mean and standard deviation for all floors across the
different scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.7 Elevation of Arrival spread mean and standard deviation for all floors across the different
scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.8 Elevation of Departure spread mean and standard deviation for all floors across the
different scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.9 5D Kappa for all floors across the different scenarios . . . . . . . . . . . . . . . . . . . . . . 112
4.10 Mean SNR gain for different diversity techniques . . . . . . . . . . . . . . . . . . . . . . . 119
vii

List of Figures
1.1 5G standard key performance indicators, [1] . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Radio spectrum allocation in the United States, [2] . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Multipath propagation in wireless communications, [3] . . . . . . . . . . . . . . . . . . . . 5
1.4 SISO sounders: (a) Non VNA-based, (b) VNA-based, [15] . . . . . . . . . . . . . . . . . . . 11
1.5 MIMO array sounders types: (a) Full array, (b) Virtual array, (c) Switched array, [15] . . . . 13
2.1 ReRoMA sample configuration diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 High-level Sounder Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 ReRoMA Implementation in Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 Reflector tape pattern on (a) Rx and (b) Tx . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5 Detailed Tx chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.6 Detailed Rx chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.7 Timing sequence and triggering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.8 Google Earth view of measurement environment. Tx and Rx locations are marked in red
and blue respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.9 MATLAB view of measurement environment. Surrounding trees (green) and building
walls (blue). Red and blue dots corresponds to Tx and Rx locations respectively. . . . . . . 37
2.10 Omni-PDP for the 1 hour measurement, overlapped . . . . . . . . . . . . . . . . . . . . . . 37
2.11 APS comparison for the 12x12 MIMO snapshot generated by (a) Manual rotation, (b)
Motor rotation with sub-sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
viii

2.12 Reference Measurements Scenarios. Scenario 1 is metallic sphere moving along the linear
path from marked start to marked end point. Scenario 2 and 3 are rotation of metallic
sphere reflector around Tx and Rx respectively. . . . . . . . . . . . . . . . . . . . . . . . . 43
2.13 Dynamic omni-PDP vs time for Linear Sphere movement scenario . . . . . . . . . . . . . . 44
2.14 Dynamic evaluation in the linear sphere movement scenario, (a) ADPS at Tx, (b) joint
APS. Shown in grey are the projections on each of the dimensions (marginal data). . . . . 45
2.15 Dynamic evaluation in the circular sphere around Tx movement scenario, (a) ADPS at Tx,
(b) joint APS.Shown in grey are the projections on each of the dimensions (marginal data). 46
2.16 Dynamic omni-PDP vs time for Circular Sphere around Tx movement scenario . . . . . . . 47
2.17 Moving Rx measurement scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.18 Dynamic evaluation in the Moving Rx away from Tx scenario, (a) ADPS at Rx, (b) ADPS
at Tx. Shown in grey are the projections on each of the dimensions (marginal data).
Reflection from building B can be identified in (b) as the component starting at around
(delay, AoD) = (9, 45), while building C can be identified in (b) as the component starting
at around (delay, AoD) = (20, 160) and can be tracked afterwards. . . . . . . . . . . . . . . 49
2.19 Dynamic omni-PDP vs time for Rx Moving Away from Tx scenario . . . . . . . . . . . . . 49
2.20 Measurement environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.21 Omni-PDP for the T intersection scenario at time (a) t=0 s (NLOS) , (b) t=30 s (LOS), (c)
t=55 s (NLOS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.22 Dynamic PDP evolution vs time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.23 Dynamic evaluation of the ADPS at (a) Tx, (b) Rx . . . . . . . . . . . . . . . . . . . . . . . 52
3.1 Convoy driving scenario, map view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 Opposite side and overtaking scenarios, map view . . . . . . . . . . . . . . . . . . . . . . . 60
3.3 Convoy driving scenario, reflections identification . . . . . . . . . . . . . . . . . . . . . . . 65
3.4 Dynamic omni-directional PDP vs time for opposite sides scenario . . . . . . . . . . . . . . 66
3.5 Overtaking driving scenario, reflections identification . . . . . . . . . . . . . . . . . . . . . 67
3.6 Dynamic joint APS vs time for convoy driving scenario . . . . . . . . . . . . . . . . . . . . 68
3.7 Dynamic joint APS vs time for opposite sides driving scenario . . . . . . . . . . . . . . . . 68
3.8 Dynamic joint APS vs time for overtaking driving scenario . . . . . . . . . . . . . . . . . . 69
ix

3.9 CDF plot of the omni-directional pathloss in dB . . . . . . . . . . . . . . . . . . . . . . . . 69
3.10 Linear fit for the omni-directional pathloss vs log(d) . . . . . . . . . . . . . . . . . . . . . . 70
3.11 CDF plot of the RMS delay spread in ns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.12 RMS delay spread vs log(distance) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.13 CDF plot of the AoA angular spread in Fleury’s definition . . . . . . . . . . . . . . . . . . . 72
3.14 CDF plot of the AoD angular spread in Fleury’s definition . . . . . . . . . . . . . . . . . . 72
3.15 AoA Angular spread vs log(distance) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.16 AoD Angular spread vs log(distance) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.17 CDF plot of the Kappa parameter in dB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.18 Kappa vs log(distance) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.19 CDF plot of the stationarity time Tstat parameter in seconds for the omni-directional
non-normalized PDPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.20 CDF plot of the stationarity time Tstat parameter in seconds for the max-dir nonnormalized PDPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.21 CDF plot of the stationarity time Tstat parameter in seconds for the omni-directional
TOA-normalized PDPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.22 CDF plot of the stationarity time Tstat parameter in seconds for the max-dir TOAnormalized PDPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.1 Full system block diagram of the channel sounder. . . . . . . . . . . . . . . . . . . . . . . 85
4.2 (a) Single Antenna Element top/side view, (b) HFSS Simulated Antenna Pattern . . . . . . 86
4.3 Array Design - Top/side view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.4 Antenna Radiation Pattern of a single antenna element on the array . . . . . . . . . . . . . 87
4.5 Tx Antenna Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.6 Map of the measurement scenario showing the TX positions for LOS (red contour) and
NLOS (blue contour) scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.7 Floorplan of the building in which the RXs were placed. Positions marked in blue were
used for the LOS and the NLOS measurements, whereas positions marked in red were
used for the "deep LOS" measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
x

4.8 Sample PDP for LOS scenario, 2nd floor position 4, dT x−Rx = 33m . . . . . . . . . . . . . . 98
4.9 Sample PDP for DeepLOS scenario, 2nd floor position 4, dT x−Rx = 41m . . . . . . . . . . . 98
4.10 Sample PDP for NLOS scenario, 2nd floor position 4, dT x−Rx = 75m . . . . . . . . . . . . . 99
4.11 Sample APS for LOS scenario, 2nd floor position 4, dT x−Rx = 33m . . . . . . . . . . . . . . 100
4.12 Sample APS for DeepLOS scenario, 2nd floor position 4, dT x−Rx = 41m . . . . . . . . . . . 100
4.13 Sample APS for NLOS scenario, 2nd floor position 4, dT x−Rx = 75m . . . . . . . . . . . . . 101
4.14 Side-by-side sample view of the main propagation processes involved in NLOS positions . 101
4.15 Parameters definition for the Pathloss modelling . . . . . . . . . . . . . . . . . . . . . . . . 103
4.16 (a) 3D scatter plot for LOS pathloss modelling, (b) Projection of model onto angular axis,
(c) Project of model onto distance axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.17 (a) 3D scatter plot for DeepLOS pathloss modelling, (b) Projection of model onto angular
axis, (c) Project of model onto distance axis . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.18 Shadowing models for (a) LOS, (b) DeepLOS . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.19 Excess NLOS pathloss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.20 RMS Delay Spread models for (a) LOS/DeepLOS, (b) NLOS . . . . . . . . . . . . . . . . . . 108
4.21 RMS Delay Spread vs distance for (a) LOS/DeepLOS, (b) NLOS . . . . . . . . . . . . . . . . 108
4.22 CDF for the angular spreads computed from Restricted Beamformer across the 3
measurement scenarios, (a) AoA, (b) AoD, (c) EoA, (d) EoD . . . . . . . . . . . . . . . . . . 111
4.23 CDF for the angular spreads computed from Full Beamformer across the 3 measurement
scenarios, (a) AoA, (b) AoD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.24 CDF for the κ across the 3 measurement scenarios for both Omni and Max Dir (a) LOS, (b)
DeepLOS, (c) NLOS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.25 κ5D vs distance for (a) Max Dir, (b) Omni . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.26 Sample BLER vs SNR Sidelink result for one measurement position . . . . . . . . . . . . . 117
4.27 CDF of SNR gains evaluated from Sidelink Simulations vs directly from the channel
transfer function, for LOS and NLOS scenarios. Evaluations done for second floor of EEB
only. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
xi

4.28 CDF for the SNR gains for LOS, DeepLOS and NLOS scenarios. Ensembles consist of all
measurement points (all floors). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.29 SNR gain distribution for the different diversity techniques for 3GPP and Measured
channel, LOS and NLOS scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
xii

Abstract
For any wireless system to function optimally, it is imperative to have an in-depth understanding of the
wireless propagation channel. An accurate model of the channel allows the communication system to
either mitigate adverse effects or exploit beneficial characteristics. Propagation channel measurements,
also known as channel sounding, is the most accurate way to acquire the true characteristics of such
environment. The first part of this dissertation presents a novel channel sounder design for high-frequency
communications (mm-wave/terahertz) and the use of such a sounding mechanism to measure and model
double-directional dynamic vehicular channels.
Due to the unique propagation characteristics at those frequencies, it is anticipated that most of the
future millimeter-wave systems will utilize beam-forming antenna arrays to overcome the higher pathloss
that occurs at such frequencies. Consequently, the angular spectrum and its temporal evolution are vital for
the efficient design of such systems. While such information about the channels can be obtained by the use
of (full-/switched-/phased-) antenna arrays, the cost and the availability of such arrays or the required electrical components to make such system work are prohibitive. Hence, we designed and built a mechanical
structure called ReRoMA, short for redirecting rotating mirror arrangement, which is capable of obtaining directionally-resolved measurements in dynamic environment, capturing a high-angular-resolution
snapshot of the environment in about one second, orders of magnitude faster than the common practice of
using rotating-horn antennas. ReRoMA is low-cost and flexible, as it requires only a single radio frequency
chain at each link end and performs mechanical beam-steering, while still achieving fast measurements by
xiii

physically separating the signal generation/transmission and the beam-steering components. We outline
in this work the fundamental concept of ReRoMA, describe in detail its implementation and demonstrate
its application and accuracy in static and dynamic reference measurement scenarios.
Furthermore, we present results from a vehicle-to-vehicle measurement campaign using ReRoMAbased channel sounder operating at 60 GHz, investigation various aspects of the channel at such frequencies such as moving scatterers, blocking objects, foliage, in addition to statistical evaluations of propagation
parameters such as pathloss, delay spread, angular spreads, power distribution across multi-path components (MPCs), and stationarity region, all of which are essential for wireless system design and testing at
such frequencies. This measurement campaign covered scenarios such as driving in convoy, driving on
opposide sides of a six-lane road, and overtaking. We observe pathloss with pathloss coefficient around
1.9, root-mean square (RMS) delay spread between 5 and 120 ns, angular spread values under 20 degrees,
power distribution over MPCs between 6 and 15 dB, and provided comparison of our measurement results
against the literature.
In the second part of this dissertation, we present a channel sounder operating in the Sub-GHz band
based on a 16x16 MIMO system, and discuss results from an extensive measurement campaign of the
channels between users in the 5-floors of a California office building and a multi-antenna base station
located either in the street in front of the office building (i.e., Line-of-Sight (LOS) to the building, or a
different street (Non-Line-Of- Sight (NLOS) to the building). Similarly to before, we evaluate the statistics
of the channel and observe that in LOS scenarios, there is a 10 dB variation of the pathloss between the
ground and the 5th floor, while such variation was not observed in NLOS cases. The delay spreads followed
a similar trend where there is a variation of 10 ns with the increase in the Rx height for LOS scenario
and remained approximately the same for NLOS scenarios. Angular spreads are high for the indoor unit,
independent of the height difference. As for the power distribution across MPCs, we have observed κ
xiv

values of about 8 dB for LOS scenarios and -1 dB for NLOS scenarios, assuming beamforming to maximize
the received power.
Furthermore, we present simulation results for LTE Sidelink (the standard adopted by PSOs for D2D
communications) performance over the measured I2O channels, and provide comparison between these
measurement results and the 3GPP channel models. We also provide an “applique” on top of the standard
(i.e., requiring no changes in the standard) that uses multiple-antenna combining to provide SNR gains
compared to the standard SISO D2D Sidelink channel. We observe on average around 10 dB SNR gain for
applying Maximum Ratio Combining (MRC) compared to a best-case SISO channel.
xv

Chapter 1
Introduction
1.1 Motivation and Trends in Wireless Communications
Wireless communication has become an integral part of modern society, transforming the way we connect and interact. From its humble beginnings with radio broadcasts in the early 20th century, wireless
technology has evolved into a sophisticated network of communication systems, enabling everything from
mobile phone calls to high-speed internet access. This evolution has been driven by the need for mobility,
convenience, and the growing demand for data services. The fundamental principle behind wireless communication is the transmission of electromagnetic waves through the air, allowing devices to communicate
without physical connections. This has enabled a wide range of applications, including mobile telephony,
satellite communication, Wi-Fi, and Bluetooth.
Cellular communication systems specifically have gone through several generations, each bringing significant improvements in speed, capacity, and functionality. The first generation (1G) introduced analog
voice communication, while the second generation (2G) brought digital voice and basic data services. The
third generation (3G) enabled mobile internet access, and the fourth generation (4G) provided high-speed
internet and multimedia services. A comparative study between all four generations is available in [105].
The fifth generation (5G) currently under mass deployment has challenged pre-existing technologies by
promising even faster speeds, lower latency, and support for a massive number of connected devices. An
1

overview about the standards, trials, challenges, deployment and practice can be found in [104]. Talking specifically about the promises behind the adaptation of 5G technologies, key performance indicators
(KPIs) can be seen in Fig. 1.1 and include peak data rates up to 20 Gbps, ultra-reliable low-latency communication (URLLC) with latencies as low as 1 ms, and massive machine-type communications (mMTC)
supporting up to 1 million devices per square kilometer. These KPIs are essential for supporting a wide
range of applications, from enhanced mobile broadband (eMBB) to the Internet of Things (IoT). The higher
data rates enable new services such as augmented reality (AR) and virtual reality (VR), which require substantial bandwidth. The low latency is critical for applications that need real-time responses, such as
autonomous driving and remote surgery. The ability to support a massive number of devices is vital for
the IoT, where billions of devices need to communicate efficiently. To achieve these KPIs, 5G employs
several key technologies, including new air interfaces, advanced antenna technologies, and enhanced core
network architectures. The use of millimeter-wave frequencies, massive MIMO, and beamforming are
some of the significant advancements in 5G that help meet these requirements.
Spectrum scarcity is a critical issue in wireless communications (see Fig. 1.2). The lower frequency
bands (below 6 GHz) are already congested with various services, making it challenging to find new spectrum for 5G. The move towards higher frequency bands, particularly the millimeter-wave [95, 23] and
Terahertz [63, 106] spectrum, provides a solution by offering large swathes of available bandwidth. However, these higher frequencies also come with their own set of challenges, including higher propagation
losses and reduced coverage areas. Massive MIMO [70, 75] is a crucial technology that has been a topic of
research interest in the last decades because of its ability to possibly mitigate some of these challenges. By
using a large number of antennas at the base station, massive MIMO can create narrow, targeted beams
that improve spectral efficiency and increase the capacity of the network. Beamforming thus allows the
system to focus the signal towards a specific user, reducing interference by pointing spectral nulls towards non-desired users and therefore increasing the overall network performance. The combination of
2

Figure 1.1: 5G standard key performance indicators, [1]
higher frequency bands and massive MIMO is one of the key factors that allows 5G to deliver the high data
rates that were promised. However, the introduction of such new technologies also necessitates significant changes in network design and deployment, including the need for more base stations and advanced
signal processing techniques to manage the complex propagation environment. These propagation environments have to be explored and properly modelled to provide key parameters that are useful for system
design, testing, and evaluation under realistic conditions.
In addition to requirements imposed by cellular systems, ad-hoc networks, including device-to-device
(D2D) [57] and vehicle-to-vehicle (V2V) [32] communications, are crucial for the future of wireless communications. D2D communication is particularly useful in scenarios where traditional network infrastructure
is unavailable or overloaded. For example, in disaster recovery situations, where the network infrastructure
may be damaged, D2D communication can provide a means for first responders and affected individuals
to communicate directly. In urban areas with high user density, D2D can help offload traffic from the
3

U.S. DEPARTMENT OF COMMERCE
NAT OI NAL TELECOMMUNICATIONS & INFORMATION ADMINISTRAT OI N
MOBILE (AERONAUTICAL TELEMETERING)
S)
5.68
5.73
5.90
5.95
6.2
6.525
6.685
6.765
7.0
7.1
7.3
7.35
8.1
8.195
8.815
8.965
9.040
9.4
9.5
9.9
9.995
10.003
10.005
10.1
10.15
11.175
11.275
11.4
11.6
11.65
12.05
12.10
12.23
13.2
13.26
13.36
13.41
13.57
13.6
13.8
13.87
14.0
14.25
14.35
14.990
15.005
15.010
15.10
15.6
15.8
16.36
17.41
17.48
17.55
17.9
17.97
18.03
18.068
18.168
18.78
18.9
19.02
19.68
19.80
19.990
19.995
20.005
20.010
21.0
21.45
21.85
21.924
22.0
22.855
23.0
23.2
23.35
24.89
24.99
25.005
25.01
25.07
25.21
25.33
25.55
25.67
26.1
26.175
26.48
26.95
26.96
27.23
27.41
27.54
28.0
29.7
29.8
29.89
29.91
30.0
UNITED
STATES
THE RADIO SPECTRUM
NON-GOVERNMENT EXCLUSIVE
GOVERNMENT EXCLUSIVE GOVERNMENT/ NON-GOVERNMENT SHARED
RADIO SERVICES COLOR LEGEND
ACTIVITY CODE
NOT ALLOCATED RADIONAVIGATION FIXED
MARITIME MOBILE
FIXED
MARITIME MOBILE
FIXED
MARITIME MOBILE
Radiolocation RADIONAVIGATION
FIXED
MARITIME
MOBILE
Radiolocation
FIXED
MARITIME
MOBILE FIXED
MARITIME
MOBILE
AERONAUTICAL
RADIONAVIGATION
AERONAUTICAL
RADIONAVIGATION
Aeronautical
Mobile
Maritime
Radionavigation
(Radio Beacons)
MARITIME
RADIONAVIGATION
(RADIO BEACONS)
Aeronautical
Radionavigation (Radio Beacons)
3
9
14
19.95
20.05
30
30
59
61
70
90
110
130
160
190
200
275
285
300
3 kHz 300 kHz
300 kHz 3 MHz
3 MHz 30 MHz
30 MHz 300 MHz
3 GHz
300 GHz
300 MHz
3 GHz
30 GHz
Aeronautical
Radionavigation
(Radio Beacons)
MARITIME
RADIONAVIGATION
(RADIO BEACONS)
Aeronautical
Mobile
Maritime
Radionavigation
(Radio Beacons)
AERONAUTICAL
RADIONAVIGATION
(RADIO BEACONS)
AERONAUTICAL
RADIONAVIGATION
(RADIO BEACONS)
Aeronautical
Mobile
Aeronautical Mobile RADIONAVIGATION
AERONAUTICAL
RADIONAVIGATION
MARITIME
MOBILE
Aeronautical
Radionavigation
MOBILE (DISTRESS AND CALLING)
MARITIME MOBILE
MARITIME
MOBILE
(SHIPS ONLY)
MOBILE
AERONAUTICAL
RADIONAVIGATION
(RADIO BEACONS)
AERONAUTICAL
RADIONAVIGATION
(RADIO BEACONS)
BROADCASTING
(AM RADIO)
MARITIME MOBILE (TELEPHONY)
MARITIME MOBILE (TELEPHONY)
MOBILE (DISTRESS AND CALLING)
MARITIME
MOBILE
LAND MOBILE
MOBILE
FIXED
STANDARD FREQ. AND TIME SIGNAL (2500kHz)
STANDARD FREQ. AND TIME SIGNAL
Space Research
MARITIME
MOBILE
LAND MOBILE
MOBILE
FIXED
AERONAUTICAL
MOBILE (R)
STANDARD FREQ.
AERONAUTICAL MOBILE (R)
AERONAUTICAL
MOBILE (OR)
AERONAUTICAL
MOBILE (R)
FIXED
MOBILE**
Radiolocation
FIXED MOBILE*
AMATEUR
FIXED
FIXED
FIXED
FIXED
FIXED
MARITIME
MOBILE
MOBILE*
MOBILE*
MOBILE
STANDARD FREQ. AND TIME SIGNAL (5000 KHZ)
AERONAUTICAL MOBILE (R)
AERONAUTICAL MOBILE (OR)
STANDARD FREQ. Space Research
MOBILE**
AERONAUTICAL MOBILE (R)
AERONAUTICAL MOBILE (OR)
FIXED MOBILE*
BROADCASTING
MARITIME MOBILE
AERONAUTICAL MOBILE (R)
AERONAUTICAL MOBILE (OR)
FIXED Mobile
AMATEUR SATELLITE AMATEUR
AMATEUR
FIXED
Mobile
MARITIME MOBILE
MARITIME
MOBILE
AERONAUTICAL MOBILE (R)
AERONAUTICAL MOBILE (OR)
FIXED
BROADCASTING
FIXED
STANDARD FREQ. AND TIME SIGNAL (10,000 kHz)
STANDARD FREQ. Space Research
AERONAUTICAL MOBILE (R)
AMATEUR
FIXED
Mobile*
AERONAUTICAL MOBILE (R)
AERONAUTICAL MOBILE (OR)
FIXED
FIXED
BROADCASTING
MARITIME
MOBILE
AERONAUTICAL MOBILE (OR)
AERONAUTICAL MOBILE (R)
RADIO ASTRONOMY
Mobile*
AMATEUR
BROADCASTING
AMATEUR AMATEUR SATELLITE
Mobile* FIXED
BROADCASTING
STANDARD FREQ. AND TIME SIGNAL (15,000 kHz)
STANDARD FREQ. Space Research
FIXED
AERONAUTICAL MOBILE (OR)
MARITIME
MOBILE
AERONAUTICAL MOBILE (R)
AERONAUTICAL MOBILE (OR)
FIXED
FIXED
BROADCASTING
STANDARD FREQ. Space Research
FIXED
MARITIME MOBILE
Mobile FIXED
AMATEUR AMATEUR SATELLITE
BROADCASTING
FIXED
AERONAUTICAL MOBILE (R)
MARITIME MOBILE
FIXED
FIXED
FIXED
Mobile*
MOBILE**
FIXED
STANDARD FREQ. AND TIME SIGNAL (25,000 kHz)
STANDARD FREQ. Space Research
LAND MOBILE
MARITIME MOBILE
LAND MOBILE
MOBILE**
RADIO ASTRONOMY
BROADCASTING
MARITIME MOBILE
LAND MOBILE
FIXED MOBILE**
FIXED
MOBILE**
MOBILE
FIXED
FIXED
FIXED
FIXED
FIXED
LAND MOBILE
MOBILE**
AMATEUR AMATEUR SATELLITE
MOBILE
LAND MOBILE
MOBILE
MOBILE
FIXED
FIXED
MOBILE
MOBILE
FIXED
FIXED
LAND
MOBILE
LAND
MOBILE
LAND
MOBILE
LAND MOBILE
Radio Astronomy
RADIO ASTRONOMY
LAND MOBILE
FIXED
FIXED
MOBILE
MOBILE
MOBILE
LAND MOBILE
FIXED
LAND
MOBILE
FIXED
FIXED
MOBILE
MOBILE
LAND
MOBILE AMATEUR
BROADCASTING
(TV CHANNELS 2-4)
FIXED MOBILE
FIXED MOBILE
FIXED MOBILE
FIXED MOBILE
AERONAUTICAL RADIONAVIGATION
BROADCASTING
(TV CHANNELS 5-6)
BROADCASTING
(FM RADIO)
AERONAUTICAL
RADIONAVIGATION
AERONAUTICAL
MOBILE (R)
AERONAUTICAL MOBILE
AERONAUTICAL MOBILE
AERONAUTICAL
MOBILE (R)
AERONAUTICAL
MOBILE (R)
AERONAUTICAL MOBILE (R)
MOBILE FIXED
AMATEUR
BROADCASTING
(TV CHANNELS 7-13)
MOBILE
FIXED
MOBILE
FIXED
MOBILE SATELLITE
FIXED
MOBILE
SATELLITE
MOBILE
FIXED
MOBILE
SATELLITE
MOBILE
FIXED MOBILE
AERONAUTICAL RADIONAVIGATION
STD. FREQ. & TIME SIGNAL SAT. (400.1 MHz)
MET. SAT.
(S-E)
SPACE RES.
(S-E)
Earth Expl.
Satellite (E-S)
MOBILE SATELLITE (E-S)
FIXED MOBILE RADIO ASTRONOMY
RADIOLOCATION Amateur
LAND MOBILE
Meteorological
Satellite (S-E)
LAND MOBILE
BROADCASTING
(TV CHANNELS 14 - 20)
BROADCASTING
(TV CHANNELS 21-36)
RADIO ASTRONOMY
TV BROADCASTING
RADIOLOCATION
FIXED
Amateur
AERONAUTICAL
RADIONAVIGATION
MOBILE** FIXED
AERONAUTICAL
RADIONAVIGATION
Radiolocation
Radiolocation MARITIME RADIONAVIGATION
MARITIME
RADIONAVIGATION Radiolocation
Radiolocation
Radiolocation
RADIOLOCATION RADIOLOCATION
Amateur
AERONAUTICAL
RADIONAVIGATION
(Ground)
RADIOLOCATION Radio- location
AERO. RADIONAV.(Ground) FIXED SAT. (S-E) RADIO- LOCATION Radio- location
FIXED
FIXED
SATELLITE
(S-E)
FIXED
AERONAUTICAL
RADIONAVIGATION
MOBILE
FIXED MOBILE
RADIO ASTRONOMY Space Research (Passive)
AERONAUTICAL
RADIONAVIGATION
RADIOLOCATION Radio- location
RADIONAVIGATION
Radiolocation
RADIOLOCATION Radiolocation
Radiolocation
Radiolocation
RADIOLOCATION
RADIOLOCATION
MARITIME
RADIONAVIGATION
MARITIME
RADIONAVIGATION
METEOROLOGICAL
AIDS
Amateur
Amateur
FIXED
FIXED
SATELLITE (E-S) MOBILE
FIXED
SATELLITE (E-S)
FIXED
SATELLITE (E-S) MOBILE
FIXED
FIXED
FIXED
FIXED
MOBILE
FIXED SPACE RESEARCH (E-S)
FIXED
Fixed MOBILE SATELLITE (S-E) FIXED SATELLITE (S-E)
FIXED SATELLITE (S-E)
FIXED
SATELLITE (S-E)
FIXED
SATELLITE (S-E)
FIXED
SATELLITE (E-S)
FIXED
SATELLITE (E-S)
FIXED
SATELLITE
(E-S)
FIXED
SATELLITE
(E-S)
FIXED
FIXED
FIXED
FIXED
FIXED
FIXED
FIXED
MET.
SATELLITE (S-E)
Mobile
Satellite (S-E)
Mobile
Satellite (S-E)
Mobile
Satellite (E-S) (no airborne)
Mobile Satellite
(E-S)(no airborne)
Mobile Satellite (S-E)
Mobile
Satellite (E-S)
MOBILE
SATELLITE (E-S)
EARTH EXPL.
SATELLITE(S-E)
EARTH EXPL.
SAT. (S-E)
EARTH EXPL.
SATELLITE (S-E)
MET.
SATELLITE
(E-S)
FIXED
FIXED
SPACE RESEARCH (S-E) (deep space only)
SPACE RESEARCH (S-E)
AERONAUTICAL
RADIOLOCATION
RADIONAVIGATION
Radiolocation
Radiolocation
Radiolocation
Radiolocation
MARITIME
RADIONAVIGATION
Meteorological
Aids RADIONAVIGATION
RADIOLOCATION Radiolocation
RADIOLOCATION
Radiolocation
Radiolocation Amateur
Amateur Amateur Satellite
RADIOLOCATION
FIXED
FIXED
FIXED
FIXED
FIXED
SATELLITE
(S-E)
FIXED
SATELLITE
(S-E)
Mobile **
SPACE RESEARCH
(Passive)
EARTH EXPL.
SAT. (Passive)
RADIO
ASTRONOMY
SPACE
RESEARCH (Passive)
EARTH EXPL.
SATELLITE (Passive)
RADIO
ASTRONOMY
BROADCASTING
SATELLITE
AERONAUTICAL RADIONAV. Space Research (E-S)
Space
Research
Land Mobile
Satellite (E-S)
Radiolocation RADIO- LOCATION
RADIO
NAVIGATION
FIXED
SATELLITE (E-S)
Land Mobile
Satellite (E-S)
Land Mobile
Satellite (E-S) Fixed Mobile FIXED SAT. (E-S)
Mobile
Fixed
FIXED
Mobile FIXED
MOBILE
Space Research
Space Research
Space Research
SPACE RESEARCH
(Passive) RADIO ASTRONOMY EARTH EXPL. SAT. (Passive)
Radiolocation
RADIOLOCATION Radiolocation
FX SAT (E-S)
FIXED SATELLITE (E-S) FIXED
FIXED
FIXED MOBILE
EARTH EXPL.
SAT. (Passive)
MOBILE
Earth Expl. Satellite (Active)
Standard
Frequency and
Time Signal Satellite (E-S)
Earth
Exploration
Satellite
(S-S)
MOBILE FIXED
MOBILE
FIXED
Earth
Exploration
Satellite (S-S)
FIXED MOBILE FIXED SAT (E-S)
FIXED SATELLITE (E-S) MOBILE SATELLITE (E-S)
FIXED
SATELLITE
(E-S)
MOBILE
SATELLITE
(E-S)
Standard
Frequency and
Time Signal Satellite (S-E)
Stand. Frequency and Time Signal Satellite (S-E) FIXED MOBILE
RADIO
ASTRONOMY
SPACE
RESEARCH
(Passive)
EARTH
EXPLORATION
SAT. (Passive)
RADIONAVIGATION
RADIONAVIGATION INTER-SATELLITE
RADIONAVIGATION
RADIOLOCATION Radiolocation
SPACE RE.
.(Passive)
EARTH EXPL.
SAT. (Passive) FIXED MOBILE
FIXED MOBILE
FIXED MOBILE
Mobile Fixed
FIXED
SATELLITE (S-E)
BROADCASTING BCST SAT.
FIXED MOBILE
FX
SAT(E-S) MOBILE FIXED
EARTH
EXPLORATION
SATELLITE
FI XED
SATELLITE (E-S)
MOBILE
SATELLITE (E-S)
MOBILE FIXED
SPACE
RESEARCH
(Passive)
EARTH
EXPLORATION
SATELLITE
(Passive)
EARTH
EXPLORATION
SAT. (Passive)
SPACE
RESEARCH
(Passive)
INTERSATELLITE RADIO- LOCATION
SPACE
RESEARCH FIXED
MOBILE
FIXED
MOBILE
SATELLITE
(E-S)
MOBILE
SATELLITE
RADIO
NAVIGATION
RADIONAVIGATION
SATELLITE
EARTH
EXPLORATION
SATELLITE
FIXED
SATELLITE
(E-S)
MOBILE FIXED FIXED SATELLITE (E-S)
AMATEUR AMATEUR SATELLITE
AMATEUR AMATEUR SATELLITE
Amateur
Satellite Amateur
RADIOLOCATION
MOBILE FIXED MOBILE SATELLITE
(S-E)
FIXED
SATELLITE
(S-E)
MOBILE FIXED
BROADCASTING
SATELLITE
BROADCASTING
SPACE
RESEARCH
(Passive)
RADIO
ASTRONOMY
EARTH
EXPLORATION
SATELLITE
(Passive)
MOBILE
FIXED
MOBILE FIXED RADIO- LOCATION FIXED SATELLITE (E-S)
MOBILE
SATELLITE
RADIONAVIGATION
SATELLITE
RADIONAVIGATION
Radiolocation
EARTH EXPL.
SATELLITE (Passive)
SPACE RESEARCH
(Passive)
FIXED
FIXED
SATELLITE
(S-E)
SPACE
RESEARCH
(Passive)
RADIO
ASTRONOMY
EARTH
EXPLORATION
SATELLITE
(Passive)
FIXED
MOBILE
MOBILE
INTERSATELLITE
RADIOLOCATION
INTERSATELLITE
Radiolocation
MOBILE
MOBILE
SATELLITE
RADIONAVIGATION
RADIONAVIGATION
SATELLITE
AMATEUR AMATEUR SATELLITE
Amateur Amateur Satellite RADIO- LOCATION
MOBILE FIXED FIXED SATELLITE (S-E)
MOBILE FIXED FIXED SATELLITE (S-E)
EARTH
EXPLORATION
SATELLITE (Passive)
SPACE RES.
(Passive)
SPACE RES.
(Passive)
RADIO
ASTRONOMY
FIXED
SATELLITE
(S-E)
FIXED
MOBILE FIXED
MOBILE FIXED
MOBILE FIXED
MOBILE FIXED
MOBILE FIXED
SPACE RESEARCH
(Passive)
RADIO
ASTRONOMY
EARTH
EXPLORATION
SATELLITE (Passive)
EARTH
EXPLORATION
SAT. (Passive)
SPACE
RESEARCH
(Passive)
INTERSATELLITE
INTERSATELLITE
INTERSATELLITE
INTERSATELLITE
MOBILE
MOBILE
MOBILE
MOBILE
SATELLITE
RADIONAVIGATION
RADIONAVIGATION
SATELLITE
FIXED
SATELLITE
(E-S)
FIXED
FIXED
EARTH
EXPLORATION SAT.
(Passive)
SPACE RES.
(Passive)
SPACE
RESEARCH
(Passive)
RADIO
ASTRONOMY
EARTH
EXPLORATION
SATELLITE
(Passive)
MOBILE FIXED
MOBILE FIXED
MOBILE FIXED
FIXED
SATELLITE (S-E)
FIXED
SATELLITE(S-E)
FIXED
SATELLITE (S-E)
EARTH EXPL.
SAT. (Passive)
SPACE RES.
(Passive)
Radiolocation
Radiolocation
RADIOLOCATION
AMATEUR AMATEUR SATELLITE
Amateur Amateur Satellite
EARTH EXPLORATION
SATELLITE (Passive) SPACE RES. (Passive)
MOBILE
MOBILE
SATELLITE
RADIONAVIGATION
RADIONAVIGATION
SATELLITE
MOBILE
MOBILE
FIXED
RADIOASTRONOMY
FIXED
SATELLITE
(E-S)
FIXED
3.0
3.025
3.155
3.230
3.4
3.5
4.0
4.063
4.438
4.65
4.7
4.75
4.85
4.995
5.003
5.005
5.060
5.45
MARITIME
MOBILE
AMATEUR AMATEUR SATELLITE
FIXED Mobile
MARITIME MOBILE
STANDARD FREQUENCY & TIME SIGNAL (20,000 KHZ)
Space Research
AERONAUTICAL MOBILE (OR)
AMATEUR SATELLITE AMATEUR
MET. SAT. (S-E) MOB. SAT. (S-E) SPACE RES. (S-E) SPACE OPN. (S-E)
MET. SAT. (S-E) Mob. Sat. (S-E) SPACE RES. (S-E) SPACE OPN. (S-E)
MET. SAT. (S-E) MOB. SAT. (S-E) SPACE RES. (S-E) SPACE OPN. (S-E)
MET. SAT. (S-E) Mob. Sat. (S-E) SPACE RES. (S-E) SPACE OPN. (S-E)
MOBILE
FIXED
FIXED Land Mobile
FIXED MOBILE
LAND MOBILE
LAND MOBILE
MARITIME MOBILE
MARITIME MOBILE
MARITIME MOBILE
MARITIME MOBILE
LAND MOBILE
FIXED MOBILE MOBILE SATELLITE (E-S)
Radiolocation
Radiolocation
LAND MOBILE
AMATEUR
MOBILE SATELLITE (E-S) RADIONAVIGATION SATELLITE
MET. AIDS
(Radiosonde)
METEOROLOGICAL AIDS (RADIOSONDE)
SPACE RESEARCH
(S-S) FIXED MOBILE
LAND MOBILE
FIXED
LAND MOBILE
FIXED
FIXED
RADIO ASTRONOMY
RADIO ASTRONOMY METEOROLOGICAL AIDS (RADIOSONDE)
METEOROLOGICAL
AIDS (Radiosonde)
METEOROLOGICAL
SATELLITE (s-E)
Fixed
FIXED
MET. SAT.
(s-E)
FIXED
FIXED
AERONAUTICAL MOBILE SATELLITE (R) (space to Earth)
AERONAUTICAL RADIONAVIGATION RADIONAV. SATELLITE (Space to Earth)
AERONAUTICAL MOBILE SATELLITE (R) (space to Earth) Mobile Satellite (S- E)
RADIO DET. SAT. (E-S) MOBILESAT(E-S) AERO. RADIONAVIGATION
AERO. RADIONAV.
AERO. RADIONAV.
RADIO DET. SAT. (E-S)
RADIO DET. SAT. (E-S)
MOBILE SAT. (E-S)
MOBILE SAT. (E-S) Mobile Sat. (S-E)
RADIO ASTRONOMY
RADIO ASTRONOMY MOBILE SAT. (E-S)
FIXED MOBILE
FIXED
FIXED
(LOS) MOBILE (LOS) SPACE RESEARCH (s-E)(s-s) SPACE OPERATION (s-E)(s-s) EARTH EXPLORATION SAT. (s-E)(s-s)
Amateur
MOBILE Fixed RADIOLOCATION
AMATEUR
RADIO ASTRON. SPACE RESEARCH EARTH EXPL SAT
FIXED SAT.
(S-E)
FIXED
MOBILE
FIXED
SATELLITE (S-E)
FIXED MOBILE FIXED SATELLITE (E-S)
FIXED
SATELLITE
(E-S)
MOBILE FIXED
SPACE
RESEARCH (S-E) (Deep Space)
AERONAUTICAL RADIONAVIGATION
EARTH
EXPL. SAT.
(Passive)
300
325
335
405
415
435
495
505
510
525
535
1605
1615
1705
1800
1900
2000
2065
2107
2170
2173.5
2190.5
2194
2495
2501
2502
2505
2850
3000
RADIOLOCATION
BROADCASTING
FIXED
MOBILE
AMATEUR
RADIOLOCATION
MOBILE FIXED MARITIME MOBILE
MARITIME MOBILE (TELEPHONY)
MARITIME
MOBILE
LAND
MOBILE MOBILE FIXED
30.0
30.56
32.0
33.0
34.0
35.0
36.0
37.0
37.5
38.0
38.25
39.0
40.0
42.0
43.69
46.6
47.0
49.6
50.0
54.0
72.0
73.0
74.6
74.8
75.2
75.4
76.0
88.0
108.0
117.975
121.9375
123.0875
123.5875
128.8125
132.0125
136.0
137.0
137.025
137.175
137.825
138.0
144.0
146.0
148.0
149.9
150.05
150.8
152.855
154.0
156.2475
157.0375
157.1875
157.45
161.575
161.625
161.775
162.0125
173.2
173.4
174.0
216.0
220.0
222.0
225.0
235.0
300
ISM – 6.78 ± .015 MHz ISM – 13.560 ± .007 MHz ISM – 27.12 ± .163 MHz
ISM – 40.68 ± .02 MHz
ISM – 24.125 ± 0.125 GHz 30 GHz
ISM – 61.25 ± .250 GHz ISM – 122.5 ± .500 GHz ISM – 245.0 ± 1GHz
300.0
322.0
328.6
335.4
399.9
400.05
400.15
401.0
402.0
403.0
406.0
406.1
410.0
420.0
450.0
454.0
455.0
456.0
460.0
462.5375
462.7375
467.5375
467.7375
470.0
512.0
608.0
614.0
698
746
764
776
794
806
821
824
849
851
866
869
894
896
901901
902
928
929
930
931
932
935
940
941
944
960
1215
1240
1300
1350
1390
1392
1395
1400
1427
1429.5
1430
1432
1435
1525
1530
1535
1544
1545
1549.5
1558.5
1559
1610
1610.6
1613.8
1626.5
1660
1660.5
1668.4
1670
1675
1700
1710
1755
1850
2000
2020
2025
2110
2155
2160
2180
2200
2290
2300
2305
2310
2320
2345
2360
2385
2390
2400
2417
2450
2483.5
2500
2655
2690
2700
2900
3000
MARITIME MOBILE SATELLITE
(space to Earth) MOBILE SATELLITE (S-E)
RADIOLOCATION RADIONAVIGATION SATELLITE (S-E)
RADIOLOCATION Amateur
Radiolocation AERONAUTICAL RADIONAVIGATION
SPA CE RESEARCH ( Passive) EARTH EXPL SAT (Passive) RADIO ASTRONOMY
MOBILE
MOBILE ** FIXED-SAT (E-S)
FIXED
FIXED
FIXED**
LAND MOBILE (TLM)
MOBILE SAT.
(Space to Earth)
MARITIME MOBILE SAT.
(Space to Earth)
Mobile
(Aero. TLM)
MOBILE SATELLITE (S-E)
MOBILE SATELLITE
(Space to Earth) AERONAUTICAL MOBILE SATELLITE (R) (space to Earth)
3.0
3.1
3.3
3.5
3.6
3.65
3.7
4.2
4.4
4.5
4.8
4.94
4.99
5.0
5.15
5.25
5.35
5.46
5.47
5.6
5.65
5.83
5.85
5.925
6.425
6.525
6.70
6.875
7.025
7.075
7.125
7.19
7.235
7.25
7.30
7.45
7.55
7.75
7.90
8.025
8.175
8.215
8.4
8.45
8.5
9.0
9.2
9.3
9.5
10.0
10.45
10.5
10.55
10.6
10.68
10.7
11.7
12.2
12.7
12.75
13.25
13.4
13.75
14.0
14.2
14.4
14.47
14.5
14.7145
15.1365
15.35
15.4
15.43
15.63
15.7
16.6
17.1
17.2
17.3
17.7
17.8
18.3
18.6
18.8
19.3
19.7
20.1
20.2
21.2
21.4
22.0
22.21
22.5
22.55
23.55
23.6
24.0
24.05
24.25
24.45
24.65
24.75
25.05
25.25
25.5
27.0
27.5
29.5
29.9
30.0
ISM – 2450.0 ± 50 MHz
30.0
31.0
31.3
31.8
32.0
32.3
33.0
33.4
36.0
37.0
37.6
38.0
38.6
39.5
40.0
40.5
41.0
42.5
43.5
45.5
46.9
47.0
47.2
48.2
50.2
50.4
51.4
52.6
54.25
55.78
56.9
57.0
58.2
59.0
59.3
64.0
65.0
66.0
71.0
74.0
75.5
76.0
77.0
77.5
78.0
81.0
84.0
86.0
92.0
95.0
100.0
102.0
105.0
116.0
119.98
120.02
126.0
134.0
142.0
144.0
149.0
150.0
151.0
164.0
168.0
170.0
174.5
176.5
182.0
185.0
190.0
200.0
202.0
217.0
231.0
235.0
238.0
241.0
248.0
250.0
252.0
265.0
275.0
300.0
ISM – 5.8 ± .075 GHz
ISM – 915.0 ± 13 MHz
INTER-SATELLITE RADIOLOCATION SATELLITE (E-S)
AERONAUTICAL
RADIONAV.
PLEASE NOTE: THE SPACING ALLOTTED THE SERVICES IN THE SPECTRUM SEGMENTS SHOWN IS NOT PROPORTIONAL TO THE ACTUAL AMOUNT
OF SPECTRUM OCCUPIED.
AERONAUTICAL
MOBILE
AERONAUTICAL
MOBILE SATELLITE
AERONAUTICAL
RADIONAVIGATION
AMATEUR
AMATEUR SATELLITE
BROADCASTING
BROADCASTING
SATELLITE
EARTH EXPLORATION
SATELLITE
FIXED
FIXED SATELLITE
INTER-SATELLITE
LAND MOBILE
LAND MOBILE
SATELLITE
MARITIME MOBILE
MARITIME MOBILE
SATELLITE
MARITIME
RADIONAVIGATION
METEOROLOGICAL
AIDS
METEOROLOGICAL
SATELLITE
MOBILE
MOBILE SATELLITE
RADIO ASTRONOMY
RADIODETERMINATION
SATELLITE
RADIOLOCATION
RADIOLOCATION SATELLITE
RADIONAVIGATION
RADIONAVIGATION
SATELLITE
SPACE OPERATION
SPACE RESEARCH
STANDARD FREQUENCY
AND TIME SIGNAL
STANDARD FREQUENCY
AND TIME SIGNAL SATELLITE
RADIO ASTRONOMY
FIXED
MARITIME MOBILE
FIXED
MARITIME
MOBILE Aeronautical
Mobile
STANDARD FREQ. AND TIME SIGNAL (60 kHz) FIXED Mobile*
STAND. FREQ. & TIME SIG.
MET. AIDS
(Radiosonde) Space Opn. (S-E) MOBILE. SAT. (S-E)
Fixed
Standard
Freq. and
Time Signal
Satellite (E-S)
FIXED
STANDARD FREQ. AND TIME SIGNAL (20 kHz)
Amateur
MOBILE
FIXED
SAT. (E-S) Space Research
ALLOCATION USAGE DESIGNATION
SERVICE EXAMPLE DESCRIPTION
Primary FIXED Capital Letters
Secondary Mobile 1st Capital with lower case letters
U.S. DEPARTMENT OF COMMERCE
National Telecommunications and Information Administration
Office of Spectrum Management
October 2003
MOBILE BROADCASTING
TRAVELERS INFORMATION STATIONS (G) AT 1610 kHz
59-64 GHz IS DESIGNATED FOR
UNLICENSED DEVICES
Fixed
AERONAUTICAL
RADIONAVIGATION
SPACE RESEARCH (Passive)
* EXCEPT AERO MOBILE (R)
** EXCEPT AERO MOBILE WAVELENGTH
BAND
DESIGNATIONS
ACTIVITIES
FREQUENCY
3 x 107m 3 x 106m 3 x 105m 30,000 m 3,000 m 300 m 30 m 3 m 30 cm 3 cm 0.3 cm 0.03 cm 3 x 105Å 3 x 104Å 3 x 103Å 3 x 102Å 3 x 10Å 3Å 3 x 10-1Å 3 x 10-2Å 3 x 10-3Å 3 x 10-4Å 3 x 10-5Å 3 x 10-6Å 3 x 10-7Å
0 10 Hz 100 Hz 1 kHz 10 kHz 100 kHz 1 MHz 10 MHz 100 MHz 1 GHz 10 GHz 100 GHz 1 THz 1013Hz 1014Hz 1015Hz 1016Hz 1017Hz 1018Hz 1019Hz 1020Hz 1021Hz 1022Hz 1023Hz 1024Hz 1025Hz
THE RADIO SPECTRUM
3 kHz MAGNIFIED ABOVE 300 GHz
VERY LOW FREQUENCY (VLF)
Audible Range AM Broadcast FM Broadcast Radar Sub-Millimeter Visible Ultraviolet Gamma-ray Cosmic-ray
Infra-sonics Sonics Ultra-sonics Microwaves Infrared
P LS X C Radar
Bands
LF MF HF VHF UHF SHF EHF INFRARED VISIBLE ULTRAVIOLET X-RAY GAMMA-RAY COSMIC-RAY
X-ray
ALLOCATIONS
FREQUENCY
BROADCASTING FIXED MOBILE*
BROADCASTING FIXED BROADCASTING FIXED Mobile
FIXED BROADCASTING
BROADCASTING FIXED
FIXED
BROADCASTING
FIXED
BROADCASTING FIXED
BROADCASTING
FIXED
BROADCASTING
FIXED
BROADCASTING FIXED
BROADCASTING FIXED
FIXED
FIXED
FIXED
FIXED
LAND
MOBILE
FIXED
AERONAUTICAL MOBILE (R)
AMATEUR SATELLITE
AMATEUR
MOBILE SATELLITE (E-S)
FIXED
Fixed Mobile Radiolocation FIXED MOBILE
LAND MOBILE MARITIME MOBILE
FIXED
LAND MOBILE
FIXED
LAND MOBILE
RADIONAV-SATELLITE
FIXED MOBILE
FIXED LAND MOBILE
MET. AIDS
(Radio- sonde) SPACE OPN. (S-E) Earth Expl Sat (E-S) Met-Satellite (E-S) MET-SAT. (E-S) EARTH EXPL SAT. (E-S)
Earth Expl Sat (E-S) Met-Satellite (E-S) EARTH EXPL SAT. (E-S) MET-SAT. (E-S)
LAND MOBILE
LAND MOBILE FIXED
LAND MOBILE
FIXED
FIXED
FIXED LAND MOBILE
LAND MOBILE
FIXED LAND MOBILE
LAND MOBILE
LAND MOBILE
LAND MOBILE
MOBILE FIXED
MOBILE FIXED
BROADCAST MOBILE FIXED
MOBILE FIXED
FIXED LAND MOBILE
LAND MOBILE
FIXED LAND MOBILE
AERONAUTICAL MOBILE
AERONAUTICAL MOBILE
FIXED LAND MOBILE
LAND MOBILE
LAND MOBILE FIXED
LAND MOBILE FIXED
MOBILE FIXED
FIXED
FIXED
MOBILE
FIXED
FIXED
FIXED
BROADCAST
LAND MOBILE
LAND MOBILE
FIXED
LAND MOBILE
METEOROLOGICAL
AIDS
FX Space res. Radio Ast E-Expl Sat
FIXED MOBILE**
MOBILE SATELLITE (S-E) RADIODETERMINATION SAT. (S-E)
Radiolocation MOBILE FIXED
Amateur Radiolocation
AMATEUR
FIXED MOBILE
B-SAT FX MOB Fixed Mobile Radiolocation
RADIOLOCATION
MOBILE **
Fixed (TLM) LAND MOBILE
FIXED (TLM) LAND MOBILE (TLM)
FIXED-SAT (S-E) FIXED (TLM)
MOBILE
MOBILE SAT.
(Space to Earth) Mobile **
MOBILE** FIXED
MOBILE
MOBILE SATELLITE (E-S)
SPACE OP.
(E-S)(s-s)
EARTH EXPL.
SAT. (E-S)(s-s)
SPACE RES.
(E-S)(s-s) FX. MOB.
MOBILE FIXED
Mobile
R- LOC.
BCST-SATELLITE Fixed Radiolocation
B-SAT R- LOC. FX MOB Fixed Mobile Radiolocation
FIXED MOBILE** Amateur RADIOLOCATION
SPACE RES..(S-E)
MOBILE FIXED
MOBILE SATELLITE (S-E)
MARITIME MOBILE
Mobile
FIXED
FIXED
BROADCAST MOBILE FIXED
MOBILE SATELLITE (E-S)
FIXED
FIXED MARITIME MOBILE FIXED
FIXED MOBILE**
FIXED MOBILE**
FIXED SAT (S-E) AERO. RADIONAV.
FIXED
SATELLITE (E-S)
Amateur- sat (s-e)
Amateur MOBILE FIXED SAT(E-S)
FIXED FIXED SATELLITE (S-E)(E-S)
FIXED FIXED SAT (E-S) MOBILE
Radiolocation RADIO- LOCATION FIXED SAT.(E-S)
Mobile**
Fixed Mobile FX SAT.(E-S) L M Sat(E-S)
AERO RADIONAV FIXED SAT (E-S)
AERONAUTICAL RADIONAVIGATION
RADIOLOCATION
Space Res.(act.)
RADIOLOCATION Radiolocation
Radioloc. RADIOLOC. Earth Expl Sat Space Res.
Radiolocation BCST SAT.
FIXED FIXED SATELLITE (S-E)
FIXED SATELLITE (S-E)
EARTH EXPL. SAT. FX SAT (S-E) SPACE RES.
FIXED SATELLITE (S-E)
FIXED SATELLITE (S-E)
FIXED SATELLITE (S-E) MOBILE SAT. (S-E)
FX SAT (S-E) MOBILE SATELLITE (S-E)
FX SAT (S-E) STD FREQ. & TIME MOBILE SAT (S-E)
EARTH EXPL. SAT. MOBILE FIXED SPACE RES.
FIXED MOBILE
MOBILE** FIXED
EARTH EXPL. SAT. FIXED MOBILE** RAD.AST SPACE RES.
FIXED MOBILE
INTER-SATELLITE
FIXED
RADIO ASTRONOMY SPACE RES. (Passive)
AMATEUR AMATEUR SATELLITE
Radiolocation Amateur RADIO- LOCATION Earth Expl. Satellite (Active)
FIXED
INTER-SATELLITE RADIONAVIGATION
RADIOLOCATION SATELLITE (E-S) INTER-SATELLITE
FIXED
SATELLITE
(E-S)
RADIONAVIGATION
FIXED
SATELLITE
(E-S)
FIXED
MOBILE SATELLITE (E-S) FIXED SATELLITE (E-S)
MOBILE FIXED Earth Exploration Satellite (S-S)
std freq
& time e-e-sat (s-s) MOBILE FIXED
e-e-sat MOBILE
SPACE
RESEARCH (deep space)
RADIONAVIGATION INTER- SAT SPACE RES.
FIXED MOBILE SPACE RESEARCH
(space-to-Earth)
SPACE
RES.
FIXED
SAT. (S-E)
MOBILE FIXED
FIXED-SATELLITE
MOBILE FIXED FIXED SATELLITE MOBILE SAT.
FIXED
SAT
MOBILE
SAT.
EARTH
EXPL
SAT (E-S)
Earth
Expl. Sat (s - e) SPACE RES. (E-S)
FX-SAT
(S-E)
FIXED MOBILE BROAD- CASTING BCST SAT.
RADIO
ASTRONOMY FIXED MOBILE** FIXED SATELLITE (E-S)
MOBILE
SATELLITE (E-S) FIXED SATELLITE (E-S)
MOBILE RADIONAV. SATELLITE
FIXED MOBILE MOB. SAT(E-S) RADIONAV.SAT.
MOBILE
SAT (E-S).
FIXED MOBILE FX SAT(E-S)
MOBILE FIXED
INTER- SAT EARTH EXPL-SAT (Passive) SPACE RES.
INTER- SAT SPACE RES. EARTH-ES
INTER- SAT EARTH-ES SPACE RES. MOBILE FIXED
EARTH
EXPLORATION
SAT. (Passive)
SPACE
RES. MOBILE FIXED INTER - SAT
FIXED MOBILE
INTERSAT RADIO- LOC. MOBILE FIXED EARTH EXPLORATION SAT. (Passive)
MOBILE FIXED
INTERSATELLITE FIXED MOBILE**
MOBILE** INTER- SATELLITE
MOBILE INTER- SATELLITE
RADIOLOC. Amateur
Amateur Sat. Amateur RADIOLOC.
AMATEUR SAT AMATEUR RADIOLOC.
SPACE
RESEARCH
(Passive)
EARTH
EXPL SAT.
(Passive)
FIXED MOBILE INTER- SATELLITE SPACE RESEARCH
(Passive)
EARTH
EXPL SAT.
(Passive)
Amatuer FIXED MO- BILE INTER- SAT. SPACE RES. E A R T H EXPL . SAT
INTERINTER-SAT. INTER-SAT.
SATELLITE
MOBILE FIXED
FX-SAT (S - E) BCST - SAT.
B- SAT. MOB** FX-SAT
SPACE RESEARCH
SPACE
RES..
This chart is a graphic single-point-in-time portrayal of the Table of Frequency Allocations used by the
FCC and NTIA. As such, it does not completely reflect all aspects, i.e., footnotes and recent changes
made to the Table of Frequency Allocations. Therefore, for complete information, users should consult the
Table to determine the current status of U.S. allocations.
Figure 1.2: Radio spectrum allocation in the United States, [2]
cellular network, improving overall network performance. On the other hand, V2V communication is a
key component of intelligent transport systems (ITS) [12]. By allowing vehicles to communicate directly
with each other, V2V can enhance safety through applications such as collision avoidance, lane change
assistance, and cooperative adaptive cruise control. V2V communication also supports the development of
autonomous driving technologies by providing vehicles with real-time information about their surroundings. Both D2D and V2V communications rely on the proximity of devices and the ability to establish
direct links without going through a central infrastructure. This requires advanced protocols and mechanisms for discovery, link establishment, and data transfer. Additionally, security and privacy are critical
concerns in ad-hoc networks, as the absence of a central authority makes it challenging to enforce policies
4

Figure 1.3: Multipath propagation in wireless communications, [3]
and protect against malicious activities. Most importantly, and in-line with the work explored in this thesis, ad-hoc wireless propagation channels are inherently different than cellular channels and thus require
additional exploration and modelling.
1.2 Overview of Wireless Channels and Impact on System Design
Hand-in-hand with wireless propagation channels comes the term of multipath propagation. Multipath
propagation, as can be seen in Fig. 1.3, refers to when the signals take multiple paths to reach the receiver. These paths can arrive with different delays, amplitudes and phases, causing fading (patterns of
constructive interference) and (in wideband systems) inter-symbol interference (ISI), possibly degrading
the quality of the received signal and increasing the error rate, and thus significantly affecting the performance of wireless systems. Understanding the multipath propagation environment and its effect on the
communication link is thus essential for designing intelligent systems based on signal processing that can
mitigate these effects.
5

The double-directional channel impulse response is a powerful tool for characterizing the multipath
environment and thus provides a complete description of the channel, describing all the parameters of
the different multipath components (MPCs). By considering both the directions of departure and arrival,
it provides a detailed picture of how signals propagate through the environment. This in general can be
represented as detailed in [84] as:
h(t, τ, Ω, Ψ) = X
L
l=1
ρlδ(Ψ − Ψl)δ(Ω − Ωl)δ(τ − τl) (1.1)
where l is the path index, ρl
, τl are the complex gain and delay of the l
th path, Ωl
is the angle of departure
(AoD) Ω = [ϕ, θ] of the l-th path, where ϕ, θ are the azimuth and elevation, respectively, and Ψl
is the
angle of arrival (AoA); similarly Ψ = [ϕ, θ]. From that, the multiple-input multiple-output (MIMO) channel
transfer function H(f) can be captured by using multiple antennas (see details in sec 1.3.3) at the Tx
side (ϕTx,i
, θTx,j
) and the receiver end (ϕRx,k, θRx,l
) as well. This data can thus be arranged in a fivedimensional matrix with dimensions [Nf × NTx
az × NTx
el × NRx
az × NRx
el ] where Nf
is the number of
subcarriers, NTx
az , NTx
el , NRx
az , and NRx
el are the number of positions in azimuth of departure, elevation of
departure, azimuth of arrival and elevation of departure respectively, as detailed in [84, Chapter 6]
1.2.1 Wireless propagation phenomena
The propagation of each of the signal copies can undergo a multitude of propagation phenomena [84], out
of which we list the following:
1. Free space propagation:
Free space propagation refers to the ideal scenario where a signal travels in a straight line without
encountering any obstacles. The power of the signal decreases with the square of the distance from
6

the transmitter. More specifically, the received power Pr is related to the transmitted power Pt by
the Friis transmission equation (under the assumption of isotropic antennas at both ends):
Pr = Pt(
λ
4πd)
2
(1.2)
where λ is the wavelength of the signal, and d is the distance between the transmitter and the
receiver. The validity of Friis’ law is restricted to the far field of the antennas, i.e. the Tx and Rx
antennas have to be placed at least at Rayleigh distance, defined as:
dR =
2L
2
a
λ
(1.3)
where La is the largest dimension of the antenna.
2. Reflection:
Reflection (or more specifically in the context of this section, specular reflection) occurs when a
signal bounces off a smooth large (compared to the wavelength) surface such as a building or the
ground. The strength of the reflected signal depends on the properties of the surface, such as its
material, as well as the angle of incidence. The polarization of the incident and reflected field might
be different: (i) in the case of linearly polarized fields, the reflected field’s polarization vector will
remain linear but will endure a phase shift depending on the angle of incidence, and (ii) in the case
of circularly polarized fields, the reflected field’s polarization will remain circular, but will flip the
handedness (from left-hand circular polarization to right-hand circular polarization and vice-versa).
7

3. Diffraction:
Diffraction involves the bending of waves around the edges of an obstacle. It is particularly important
in environments with obstacles, such as buildings or hills, that block the direct path. The ability of
a signal to diffract around an obstacle depends on the wavelength of the signal and the size of the
obstacle. Lower frequency signals, with longer wavelengths, tend to diffract more effectively than
higher frequency signals.
4. Scattering:
Scattering happens when a signal encounters small objects or rough surfaces, causing the signal to
scatter in multiple directions. It is caused by objects or roughnesses that are comparable in size to
the wavelength of the signal.
Due to the variety of processes, together with the presence of multiple objects, different paths are
possible. Furthermore, the specific processes impact the strength of the MPC that goes through them,
impacting the pattern of constructive and destructive interference (or fading) at the receiver.
1.2.2 Condensed channel parameters
All these processes impact the multipath propagation, which is described by the double-directional channel
response. While this channel response is the most general formulation, more compact parameters can be
derived from it. The most popular are:
1. Pathloss:
Pathloss represents the reduction in signal strength due to the change of strength in the MPCs (it
by definition averages out the small-scale variations, i.e. variations due to the changing interference patterns between the MPCs). Accurate models of pathloss are essential for designing wireless
networks, as they help predict the coverage area and ensure that the signal strength is sufficient for
8

reliable communication. It is commonly defined at the sum of the powers of the different MPCs, averaging of small-scale fading (over the stationarity region) and assuming isotropic antennas. Pathloss
is typically modeled as a function of frequency, distance, and other scenario-specific parameters
such as height, penetration angle (in Outdoor-to-indoor propagation), etc.
2. Delay dispersion:
Delay dispersion refers to the spread of arrival times of multipath components. It can cause intersymbol interference (ISI) in digital communication systems, where the tail of one symbol overlaps
with the head of the next symbol. This is an issue when we discuss multi-path propagation in
wideband systems, where delay dispersion (channel impulse response not being a delta anymore)
can be equivalently described by what is termed "frequency selectivity", referring to variations in the
channel transfer function over the bandwidth of interest. Delay dispersion can be mitigated by using
equalizers that compensate for the different delays of the multipath components. The most widely
used condensed parameter corresponding to delay dispersion is the "root-mean square (RMS) delay
spread" of the channel. Delay dispersion has a big impact on wireless system design: for example,
in OFDM systems, the duration of the cyclic prefix may be chosen to be longer than the maximum
excess delay to prevent ISI.
3. Angular dispersion:
Angular dispersion describes the spread of angles of departure and arrival of multipath components.
It affects the spatial characteristics of the received signal. Understanding the angular dispersion of
the channel helps in designing antenna arrays and beamforming algorithms that can effectively
capture and combine the multipath components. This could in turn lead to a much more robust
communication link through the use of angular diversity (intelligent combination of signals propagating via different directions), a better interference management through the use of beamforming
(pointing main beam to desired users and spectral nulls to non-desired ones), or to a better overall
9

channel capacity through the use of spatial multiplexing (sending independent streams of data across
the different MPCs). The condensed parameter corresponding to angular dispersion is termed the
angular spread, which can be computed in azimuth or in elevation from the angular power spectrum,
as explained in [39].
4. K-factor and power distribution among MPCs:
The K-factor [111] is the ratio of the power in the direct path to the power in the scattered paths. A
high K-factor indicates a strong component (line-of-sight or strong reflection path) , while a low Kfactor indicates that the signal does not have dominant paths. The power distribution among MPCs
results in a distribution of the narrowband signal amplitude within one stationarity region that is
typically modeled as Rician.
5. Channel stationarity refers to the time duration over which the channel characteristics remain constant, i.e. the duration over which the wide sense stationary uncorrelated scattering (WSS-US) assumption is valid. Stationarity can be defined for power-delay profiles, angular power spectra, Kfactor, etc. In general, this is usually done based on a "sliding-window" correlation approach, where
the correlation can be done along the time, frequency, or spatial axis. If the correlation is above a
certain threshold, stationarity is assumed, otherwise it is not. In dynamic environments, such as
vehicular communication, the channel characteristics can change rapidly. Understanding the stationarity (in addition to the coherence time) of the channel helps in designing adaptive communication systems that can adjust their parameters to maintain reliable communication. This includes
a combination of slow and fast link adaptation, where slow adaptation can be done based on the
stationarity time and includes long-term channel state information (CSI) feedback (for reduction in
overhead), resource allocation and scheduling, and second-order-CSI beamforming based methods,
while fast link adaptation is based on coherence time and could include adaptive modulation and
coding schemes, scheduling frequent channel estimation and tracking, in addition to equalization.
10

(a) (b)
Figure 1.4: SISO sounders: (a) Non VNA-based, (b) VNA-based, [15]
1.3 Overview of Channel Sounding
1.3.1 Channel sounding: What and why?
Whether we are looking into new and higher frequency bands for operations, or new unexplored scenarios
and technologies to be applied to already-established communications in the lower frequency bands, the
theoretical performance limits of wireless systems are determined by the wireless propagation channel, as
well as the actual performance of any practical system operating in such channels [84]. For this reason,
suitable models for the propagation channels that are based on extensive and accurate measurements, are
required for the design and performance evaluation of any wireless system. Such measurements are what
we refer to as "Channel Sounding".
Channel sounding thus involves transmitting a known signal, such as a pulse or a multitone signal,
and measuring the received signal. By analyzing the received signal, the impulse response of the channel
can be determined, and from that, parameters such as power, delay and angles can be extracted.
Generally speaking, channel sounders can be categorized as either time-domain or frequency-domain,
and can further be classified as either SISO or MIMO sounders.
11

1.3.2 Time-domain vs frequency-domain sounders
Time-domain channel sounders (Fig. 1.4a) operate by transmitting short pulses or sequences of pulses
and measuring the time it takes for the echoes of these pulses to return from various reflecting objects.
This approach allows for the direct measurement (or, in the case of sequences of pulses, measurement
via correlation at the receiver) of the impulse response of the channel, which can then be used to infer
the multipath characteristics of the environment. Frequency-domain channel sounders (Fig. 1.4b), on the
other hand, measure the channel transfer function directly. This can be done by either by sweeping a range
of frequencies and measuring the frequency response of the channel at each tone by the use of a vector
network analyzer (VNA), or by the use of chirp sequences or multitone waveforms. Chirping is done by
the use of a transmit waveform whose instantaneous frequency changes linearly with time, covering the
bandwidth of interest. Such sounders thus sweeps through the different frequencies, measuring them at
different times. Multi-tone sounding works alternatively by measuring the channel on different frequencies
at the same time. This is usually done by the use of a waveform that has ideally a low crest factor and a
constant power spectral density in the band of interest, such as the Zadoff-Chu sequences.
VNA-based sounders are easy to calibrate and don’t require efforts for synchronization (as the Tx
and Rx usually reside in the same equipment), however their applications can be fairly limited to short
separation distance between Tx and Rx (which can be extended by the use of fiber links), and mostly static
scenarios as the time span of a single frequency sweep is usually several hundred milliseconds. However,
since non VNA-based sounders consist of separate units for the transmitter and the receiver, requiring
strict synchronization by the use of a shared trigger or highly stable dedicated frequency references, their
design can be significantly more difficult, but their flexibility make them capable of covering almost any
practical scenarios, especially highly dynamic communication environments.
12

(a) (b) (c)
Figure 1.5: MIMO array sounders types: (a) Full array, (b) Virtual array, (c) Switched array, [15]
1.3.3 SISO vs MIMO sounders
The further classification of SISO vs MIMO is self explanatory: SISO channel sounders utilize single antenna elements, and they can be either omnidirectional or directional: (i) omnidirectional SISO sounders
use antennas that radiate or receive signals uniformly in all directions, and (ii) directional SISO sounders
employ highly directional antennas that focus the transmitted or received energy in specific directions.
While SISO measurements can lead of modelling of path loss and delay dispersion, angular dispersion can
only be obtained with a channel sounder that employs antenna arrays, where the antennas can either be
offset from each other in space (translational arrays), or can consist of antennas pointing into different
directions (rotational arrays) [65]. In either case, the arrays can be real, switched, or virtual. Real arrays
(Fig. 1.5a) consists of building multiple single-antenna single-RF chains, all working in parallel. This approach is the fastest making it the most suitable for dynamic channels, however the cost is the highest
and calibration is non-trivial. Switched arrays (Fig. 1.5c) consists of building multiple antenna elements
accompanied by a single RF chain and a fast electronic switch. As long as the switching across all Tx-Rx
antenna pairs happen within the coherence time of the channel, such a method is equivalent to the real
arrays. This method is again suitable for dynamic channels however the lack of fast high-power switches
in new frequency bands is a significant limitation. Virtual arrays (Fig. 1.5b) consist of moving a single
antenna element by the use of a positioner, usually with a stepper motor. While this method is by far
the cheapest and easiest to implement, the mechanical movement introduces a significant delay, making
13

it only suitable for static channels. A review of channel sounding techniques suitable especially for highfrequencies can be found in [71]. A more detailed overview about channel sounders and their application
in the literature can be found in Chapter 2.
1.4 Evaluation Methods: Fourier, MUSIC, SAGE
Measurement data evaluation can significantly vary depending on multiple factors such as complexity,
resolution, and robustness. We enumerate the following popular methods:
1. Fourier methods:
Fourier transform allows the estimation of channel characteristics in terms of delay and angles of
departure/arrival in an approach that is simple but that may lack resolution. Delay resolution is
limited to the inverse of the sounding waveform’s bandwidth, and angular resolution is limited to
the aperture size of the antenna (or antenna array). While delay profile can be estimated by the
amplitudes of the inverse fourier transform of the channel transfer function, the angle estimation is
a little more evolved. This is usually estimated by the use of beamformers [101], such as the Bartlett
beamformer. The Bartlett beamformer works by steering the main direction of the beams across
a pre-selected angular grid, then calculating the output power for each direction by projecting the
steering vectors into the covariance matrix of the received signal. The output is thus an estimation
of the angular power spectrum and can be used for angular estimation or angular dispersion evaluation. While such a method is computationally simple and provides a basic level of beamforming
performance, it has limitations in terms of resolution and interference suppression. For the angular
resolution, only signals that are separated by at least the aperture of the antennas can be resolved,
otherwise they will be indistinguishable.
14

2. MUSIC:
MUSIC [102], short for multiple signal classification, is a subspace-based method that provides highresolution estimates of the directions of arrival of multipath components. By decomposing the covariance matrix of the received signal into signal and noise subspaces, MUSIC can accurately estimate the angles of the MPCs for both departure and arrival even in the presence of noise. The
effectiveness of MUSIC lies in its ability to exploit the orthogonality between the signal and noise
subspaces. By projecting the received signal onto the noise subspace, MUSIC identifies the directions
of MPCs corresponding to the peaks in the resulting pseudospectrum.
3. CLEAN:
CLEAN [54] is a greedy matching pursuit algorithm that works by sequentially estimating and removing the dominant MPC, also termed successive interference/order cancellation. It does that by
estimating a single specular MPC via correlation with the received waveform (in the first iteration)
or its residue (in the later iterations), and its delay, synthesizing the contribution of this MPC into
the channel impulse response, updating the residue by subtracting this MPC contribution from the
previous residue, and then checking for convergence based on total number of MPCs extracted or
total residual energy. CLEAN is thus a simple, easy to implement, and conceptually straightforward
algorithm, and works really well in scenarios where the signal consists of a few strong well-separated
components. It does however require very accurate system and antenna calibrations.
4. SAGE:
SAGE [38], short for space-alternating generalized expectation-maximization, is a powerful iterative algorithm that, by alternating between the estimation of signal impact in different parameter
domains, estimates the parameters of MPCs including delays, angles, and amplitudes. SAGE operates
by iteratively updating the estimates of the parameters in a manner that maximizes the likelihood
15

function. This involves separating the contributions of individual MPCs and refining their parameter estimates in each iteration. The method is highly effective in resolving MPCs that are closely
spaced in time, frequency, or angle. CLEAN is typically used as the initialization step of SAGE, thus
it builds on it in terms of complexity and performance. Similar to CLEAN, far field communication
is assumed, and accurate system and antenna calibrations are required.
1.5 Thesis Outline and Contributions
1.5.1 Real-Time Millimeter-Wave Double-Directional Channel Sounder
In this chapter, we present a novel patent pending real-time sounding method based on redirecting rotating mirror arrangement (ReRoMA). Up until now, the common practice to investigate the directional
characteristics of millimeter-wave channels has been the use of rotating horn antennas, or the expensive
phased-arrays. The sounding method presented here can be an attachment to be placed on top of any
single-antenna sounder, giving it directional capabilities. This design strikes a new tradeoff between cost,
complexity, and measurement speed, and is flexible in frequency and envisioned to be especially useful
in the exploration of the terahertz spectrum. Along with the system design, we present verification and
calibration details and some reference measurement scenarios to serve as proof-of-concept.
Related Publications
• Hussein Hammoud, Yuning Zhang, Zihang Cheng, Seun Sangodoyin, Hofer Markus, Faruk Pasic,
Thomas M. Pohl, Radek Zavorka, Ales Prokes, Thomas Zemen, Christoph Mecklenbrauker, Andreas
F. Molisch, “A novel low-cost channel sounder for double-directionally resolved measurements in
the mmwave band,” IEEE Transactions on Wireless Communications, under review, 2024.
• Hussein Hammoud, Yuning Zhang, Zihang Cheng, Seun Sangodoyin, Hofer Markus, Faruk Pasic,
Thomas M. Pohl, Radek Zavorka, Ales Prokes, Thomas Zemen, Christoph Mecklenbrauker, Andreas
16

F. Molisch, “A novel low-cost channel sounder for double-directionally resolved measurements in
the mmwave band,” IEEE International Conference on Communications., 2024.
• A. F. Molisch, Christoph Mecklenbraeuker, Thomas Pohl, Hussein Hammoud, Yuning Zhang,
“Spinning directional antenna in centimeter and millimeter wave bands,” Patent WO 2022155493
A1, sep 2022, available at: https://patents.google.com/patent/WO2022155493A1/en.
1.5.2 ReRoMA-based Vehicle-to-Vehicle Measurement
In this chapter, we use the previously described ReRoMA sounding method and build a 60 GHz channel
sounder able to capture a 36x72 MIMO snapshot in one second. We use the sounder to perform an extensive
vehicular measurement campaign for different scenarios encountered while driving such as driving in
convoy, opposite sides of a 6-lanes street, and overtaking. We present results for time-varying pathloss,
delay spread, angular spreads on both the Tx and the Rx, power distribution among MPCs, and stationarity
time.
Related Publications
• Hussein Hammoud, Yuning Zhang, Zihang Cheng, Seun Sangodoyin, Hofer Markus, Faruk Pasic,
Thomas M. Pohl, Radek Zavorka, Ales Prokes, Thomas Zemen, Christoph Mecklenbrauker, Andreas
F. Molisch, “Double-Directional V2V Channel Measurement using ReRoMA at 60 GHz,” IEEE Transactions on Vehicular Technologies, under review, 2024.
• Molisch, Andreas F., Hussein Hammoud, Yuning Zhang, METRANS Transportation Center, and
Pacific Southwest Region. Measurement and Modeling of Broadband Millimeter-Wave Signal Propagation Between Intelligent Vehicles. No. PSR-18-08. METRANS Transportation Center (Calif.),
2021.
17

1.5.3 Sub-GHz Measurements and Modelling
In this chapter, we describe a sub-GHz 16x16 MIMO dynamic channel sounder and evaluate the results from
an extensive measurement campaign covering scenarios of interest for the public safety organizations such
as an emergency responder communicating with a street-level command post without cellular base-station
coverage, (mimicking a firefighter communicating with a firetruck). We additionally explore the additional
robustness introduced by employing multi-antenna diversity techniques.
Related Publications
• Hussein Hammoud, Zihang Cheng, Jorge Gomez-Ponce, Seun Sangodoyin, Jason Kahn, Andreas F.
Molisch, “Propagation Channel Measurements for Indoor-to-Outdoor Communications for Deviceto-Device Public Safety Applications,” IEEE Open Journal of the Communications Society, under
review, 2024.
• Hussein Hammoud, Pawan K. Venkatesh, Jorge Gomez-Ponce, Seun Sangodoyin, Jason Kahn, and
Andreas F. Molisch. "LTE Sidelink Indoor-to-Outdoor Device-to-Device Channel Measurements and
Simulations for Public Safety Applications." In 2023 IEEE 97th Vehicular Technology Conference
(VTC2023-Spring), pp. 1-7. IEEE, 2023.
1.6 Other Publications
Related Publications
• Zhang, Yuning, Thomas Choi, Zihang Cheng, Issei Kanno, Masaaki Ito, Jorge Gomez-Ponce, Hussein Hammoud et al. "Large-scale Outdoor Cell-free mMIMO Channel Measurement in an Urban
Scenario at 3.5 GHz." arXiv preprint arXiv:2405.20617 (2024).
• Hofer, Markus, David Löschenbrand, Jiri Blumenstein, Herbert Groll, Stefan Zelenbaba, Benjamin
Rainer, Laura Bernadó, Hussein Hammoud et al.. "Wireless vehicular multiband measurements in
18

centimeterwave and millimeterwave bands." In 2021 IEEE 32nd Annual International Symposium on
Personal, Indoor and Mobile Radio Communications (PIMRC), pp. 836-841. IEEE, 2021.
• Groll, Herbert, Erich Zöchmann, Markus Hofer, Hussein Hammoud, Seun Sangodoyin, Thomas
Zemen, Jiri Blumenstein, Ales Prokes, Andreas F. Molisch, and C. F. Mecklenbräuker. "60 GHz V2I
channel variability for different elevation angle switching strategies." In 2020 14th European Conference on Antennas and Propagation (EuCAP), pp. 1-5. IEEE, 2020.
• Prokes, Ales, Jiri Blumenstein, Josef Vychodil, Tomas Mikulasek, Roman Marsalek, Erich Zöchmann, Herbert Groll,Hussein Hammoud et al. "Multipath propagation analysis for vehicle-toinfrastructure communication at 60 GHz." In 2019 IEEE Vehicular Networking Conference (VNC),
pp. 1-8. IEEE, 2019.
19

Chapter 2
Real-Time Millimeter-Wave Double-Directional Channel Sounder
2.1 Introduction
Motivated by the need for ever-increasing data rates and number of users, there is an ongoing trend in the
wireless industry to exploit higher frequency bands, due to the large amount of fallow spectrum available
in these bands. This trend results, e.g., in the use of mmWave bands for 5G deployment and the anticipated
use of (sub)THz bands in 6G. Hence, both industry and academic researchers have shown a heightened
interest in systems operating within these frequency ranges. It is axiomatic that wireless communication
system development requires a deep understanding of, and models for, the wireless propagation channel
characteristics in the band of operation. Consequently, extensive and accurate channel measurements at
high frequencies are critical for the effective development and testing of any robust and efficient system
for communication, localization, sensing, or combinations thereof, in 5G/6G.
At higher frequencies, there is an inherent increase in the isotropic pathloss, making essential the
use of adaptive arrays to improve the signal-to-noise ratio (SNR) [84, Chapter 4]. Furthermore, antenna
arrays can also be used for the implementation of single-user multiple-input multiple-output (SU-MIMO)
or multi-user MIMO (MU-MIMO) schemes to directly increase the throughput. For all these aspects, we
need the double-directional channel characteristics, i.e., models and measurements need to be performed
directionally resolved at both link ends.
20

Directional channel measurements require the use of multiple antennas, which can either be offset
from each other in space (translational arrays), or can consist of directional antennas pointing into different directions (rotational arrays). In either case, the arrays can be real, switched, or virtual [84, Chapter 9].
In particular, virtual arrays using mechanically rotating horn antennas have been widely used in measurement campaigns over the past decade. Their main drawback is that the mechanical movement severely
limits the measurement speed, which in turn limits both the number of points that can be measured in a
campaign, and the scenarios that can be measured. Despite this, they remain a valuable and popular tool
for detailed directional channel analysis.
The goal of the current paper is to overcome the limitations of mechanically-rotated horn arrays while
retaining their advantages; in other words, to develop a flexible, cost-effective double-directional channel
sounder that can perform dynamic channel measurements and measure order-of-magnitude more points
per hour than a traditional rotated-horn setup. We achieve this goal by an approach we call ReRoMA (redirecting rotating mirror arrangement) [82]. It is based on the principle of separating the signal generation
and transmission via a single fixed antenna, from a mechanically moving part that redirects the resulting
beam. As depicted in Fig. 2.1, components like the horn antenna and cables remain stationary (no rotation), pointing upwards. The transmitted electromagnetic (EM) waves are redirected by a mirror inclined
at an angle α, with α = 45◦
enabling the beam to sweep across the horizontal plane. This mirror is placed
inside a swiftly rotating tube, which changes the beam direction over time. As none of the rotating tube’s
components are connected to any electronic components, rotation can be much faster than in traditional
sounders where the horn itself, and the cable attached to it, are rotated, see next section for more details.
As a matter of fact, a full ”MIMO snapshot", i.e., a combination of all transmit and receive directions, can
be obtained in about 1 s, compared to tens of minutes with traditional rotating horns.
So far, mmWave propagation has been investigated in different scenarios using several different types
of channel sounders. The simplest channel sounders typically use single omnidirectional antennas on
21

Figure 2.1: ReRoMA sample configuration diagram
both the transmitter and receiver end such as the work in [108, 66]. These channel sounders are helpful in
gathering pathloss, shadowing and delay-spread information. However, they lack directional information,
and the large isotropic pathloss at mmWave frequencies can result in poor SNR. The latter problem may be
overcome by using directional antennas (with fixed orientation) at one [xu2000measurements , 69, 133]
or both [51] link ends, but this adds the problem of not detecting MPCs that fall outside the beamwidth of
the used antennas, and still does not provide directional information.
For directional sounding, antenna arrays are required. Real arrays use multiple radio frequency (RF)
chains to excite/capture multiple array elements simultaneously. While this allows the fastest measurements, such arrays are costly, difficult to implement and calibrate, and for mmWave systems are thus typically limited to small numbers of array elements [123]. Switched arrays also use multiple physical array
elements but only a single RF chain, which is sequentially connected to the different antenna elements via
an electronic switch; though at higher frequencies the availability, power limitations, and attenuation of
the switches can become challenging. Since electronic switches are fast, measurement of all combinations
of transmit and receive elements (called a ”MIMO snapshot") can be done on the order of a few, or a few
22

tens, of milliseconds, depending on the array sizes. Examples for switched sounders at mmWave frequencies include [25, 110]. Improvements of the SNR can be achieved by switching between horn antennas, as
realized in [109]. Ref. [91] presents a hybrid switched/real sounder.
Phased arrays switch beams sequentially into different directions and can generally achieve higher
transmit power than switched arrays. A number of channel sounders based on this principle have been
presented over the past years [14, 16, 107, 26, 27, 30, 29]. Typical number of antenna elements at each link
end are between 8 and 256. Measurement times of a MIMO snapshot are similar to the switched-array case.
Main challenges are availability and/or cost of phased arrays, in particular in newly-explored frequency
bands.
Virtual arrays, such as those created by mechanically moving antenna elements across different preset
location, measure different locations at different times. They are often combined with vector network
analyzers (VNAs) to scan over frequency [45, 55, 44], but can also be combined with time-domain sounding
[88]. Related to this are mechanically-steerable rotational arrays, created, e.g., by rotating horn antennas
with stepper motors, sampling the channel from different angles. They can be combined with VNAs to
scan over frequency for each horn orientation [49, 79, 89], transmission of pseudonoise sequences that
are received with a correlation receiver [95, 124, 72, 56, 62, 35], or multitone sounding [73, 36]. Such
sounders have been used for measurements at a variety of frequencies in the mmWave band, e.g., 28 GHz
[95, 56, 62, 79], 60 GHz [73, 49], 73 GHz [90] and at higher frequencies [7, 8, 124, 33], as well as multiband [129, 89, 35]. They have been used for measurements in a wide variety of environments, ranging
from indoor residential [31], indoor office [73, 62, 79, 129], shopping malls [56, 89], to outdoor device–todevice scenarios [94, 8], microcells [95, 124, 7] and even macrocells [130, 34]. Rotating horn sounders are
significantly less expensive and easier to calibrate than full or phased arrays, and are useful in particular
when investigating new frequency bands where phased arrays might not be available.
23

A key challenge is that these channel sounders are slow, due to the necessary mechanical movement,
typically involving a stepper motor, when transitioning between different antenna positions. Moreover,
the mechanical influences on cabling, such as cable twisting during horn rotation, can adversely affect
the performance of high-frequency hardware. Although rotary joints can address this issue, their implementation comes with a substantial cost, particularly beyond 40 GHz, and they are prone to failure during
prolonged operation [37]. While placing the entire sounder on a rotating platform resolves cabling concerns, it only allows for relatively slow rotation speeds (maximum 300 RPM) and has been implemented
solely for narrowband sounding at 28 and 60 GHz [34, 92].
A system somewhat closer in spirit to the current paper was presented by some of us in [43] where
a rotating hemispherical shell, with an aperture cut into it, was placed over an omnidirectional antenna
feed. This rotation allows mechanical beam steering but was only designed for beamforming at one link
end, and rotation is slow and not appropriate for dynamic measurements (full rotation on one link end
took several seconds). Furthermore, the fundamental operating principle was different, as in this approach
the beam was formed by a rotating entity, while in the current paper an already-formed beam is redirected
by the rotating entity.
2.2 Main Contributions
Our primary scientific contributions are as follows:
• We present a novel approach to channel sounding utilizing the ReRoMA concept. This method
strikes a unique balance between measurement speed, cost/complexity, and the requirement for
specialized components. Notably, it maintains the simplicity of traditional rotating-horn sounders
while enhancing measurement speed by at least two orders of magnitude.
24

AWG Tx IF
Chain
Tx RF
Chain
LO IF LO RF
Distribution
Amplifier GPSDO
Digitizer Rx RF
Chain
Rx IF
Chain
LO RF LO IF
Distribution
Amplifier GPSDO
Figure 2.2: High-level Sounder Diagram
• We detail the design, components, and calibration essential for the functioning of ReRoMA and
introduce a functioning prototype demonstrating its capabilities.
• Utilizing this prototype, which operates at 60 GHz, we conduct double-directional reference and calibration measurements using a metallic sphere reflector and demonstrate the accuracy of the sounder.
Additionally, we perform measurements in a dynamic measurement scenario (moving receiver (Rx)).
2.3 Sounder Design
2.3.1 ReRoMA Mechanical Structure
Our system’s mechanical structure at the transmitter (Tx)∗
contains two principal components: (i) a fixed
object (FO): this encompasses all the necessary electronics for generating the sounding signal, which is
then transmitted upward through a horn antenna, which is also part of the FO, as depicted in Fig. 2.1. (ii)
a moving object (MO) which redirects the waves emanating from the horn into time-varying directions,
thus enabling the sounder to autonomously perform a comprehensive scan of the entire 360◦
azimuth
plane. The MO consists of a cylindrical tube that is rotated by a belt drive powered by a direct current
(DC) motor, see Fig. 2.3. The choice of this mechanism was guided by its ability to provide rotation that is
∗
the Rx is completely analogous
25

Figure 2.3: ReRoMA Implementation in Hardware
both consistent (essential for accuracy and resolution) and fast (much faster than manual or stepper-motor
methods).
A mirror set at an angle in the tube reflects the electromagnetic (EM) waves propagating vertically
upwards from the horn at the base of the tube. The waves are redirected into the horizontal plane and
exit through a slit on the side of the tube. The azimuth angle of the reflection depends on the rotation
angle of the tube and thus changes quickly with time. Although our current design uses reflection into
the horizontal plane, the mirror’s angle is adjustable and can be tailored to suit different measurement
scenarios. For instance, aligning the mirror downwards might be more suitable for measuring channels
between an elevated device (e.g., base station (BS)) and devices on the ground (e.g., user equipments (UEs)).
Remark: Precise alignment of the horn antenna relative to the rotating tube is crucial. The antenna
beam should precisely illuminate the mirror’s center, facilitating beam reflection through the tube slit while
minimizing internal reflections and waveguiding effects, which might lead to beam broadening and sidelobes. To achieve this, we designed a metallic mount that securely positions the antenna element beneath
the tube’s base, featuring high-precision adjustment screws that allow fine-tuned alignment adjustments
of the horn antenna based on calibration with a laser pointer.
26

The mechanical rotation principle at the Rx and Tx requires a careful trade off between angular and
temporal resolution for the intended identification of radio propagation mechanisms in (slowly) mobile
scenarios in the mmWave band.
We can achieve a mechanically safe and stable rotation speed of NRx = 2080 rpm = 34.67 Hz with our
prototype where rpm indicated rotations-per-minute. Hence, a complete Rx rotation takes TRx = 1/NRx =
28.8 ms.
We aim at sampling close to critical with respect to the antennas angular beam width. Hence, we chose
a temporal sampling time of Ts = 200 µs, with measurements of the two polarizations interlaced. Thus,
the effective sampling time for each polarization is 2Ts
, and we obtain MRx = 1/(2NRxTs) = 72 angular
impulse response measurements per revolution at the Rx side, spaced ∆βRx = 360/MRx = 5◦
. Hence, for
our Rx beamwidth of βRx = 9◦ we achieve a normalized sampling rate of ∆βRx/βRx = 0.55 .
The Tx performs a slower stable †
rotation at a speed of NTx = 57.8 rpm = 0.963 Hz. Hence a complete
rotation takes TTx = 1/NTx = 1.038 s. The number of measured angular positions on the Tx side per Tx
revolution is thus MTx = NTx/NRx = 36 spaced at ∆βTx = 360/MTx = 10◦
. To adhere to the principle
of capturing more than one sample per 3 dB beamwidth, a different horn antenna is used for the Tx, with
a 3 dB beamwidth of βTx = 25◦
, achieving a normalized sampling rate of ∆βTx/βTx = 0.4 . Worth to
note however that sampling at 3 dB beamwidth is a minimum requirement; sampling at faster than 3 dB
beamwidth would lead to a better parameter estimation accuracy and a higher SNR, as long as the hardware
on the receiver side (digitization and data streaming) is able to support it. Rotating at such different speeds
between the Tx and the Rx allows sampling the channel in an angular pattern similar to the stepper-motorbased sounders, where the Rx steps into all angular positions for each Tx position ‡
. This implies that the
full MIMO snapshot capture time (i.e. the time during which all 36×72 combinations of Tx and Rx positions
†
Proper calibration of the rotation speeds might be necessary especially at the Tx side because of possible instability at low
speeds. Such calibration in addition to precise angular encoding (see next section) would alleviate this problem.
‡Note however that while they are similar, it is not exactly the same. In ReRoMA, neither the Tx nor the Rx "stops" in certain
angular positions, they are always in continuous rotation. However, they can be assumed stationary during the capture duration
(Tx and Rx rotate 0.005◦
and 0.4◦
during a digitizer capture respectively.)
27

are measured) is 1.038 s. While this duration is shorter than the typical stationarity times of channels (the
period during which channel statistics remain constant) envisioned for our measurement environment
and speeds [19], it exceeds the coherence time (approximately the inverse of the Doppler spread) of most
channels. This indicates that our sounder is not capable of measuring the Doppler spectrum.
To achieve the high rotational speed, ReRoMA uses a ”free-running" motor. Deviations from the nominal (intended) orientations are acceptable as long as they are measured and incorporated in the reconstruction of the double-directional impulse response. Hence, we measure the angular orientation αTx[m] and
αRx[m] at the moment of signal capture t = mTs using an optical photoelectric sensor system, m ∈ Z
+
indicated the discrete impulse response index. We use a polarized reflective tape affixed to the MO, with
the sensor itself mounted atop the FO. The use of polarized tape ensures enhanced resistance to ambient
light interference.
To achieve high precision we place multiple pieces of tape along the tube circumference, resulting in a
distinct (coded) sequence of dark, non-reflective and light, reflective sections. The pattern is designed with
the following two constraints: (i) Low auto-correlation of the pattern sequence except at zero-shift, thus
allowing precise correlation between sensor readings and physical tube direction, and (ii) dark and light
region width based on the sensor resolution, i.e. matching the width of the polarized light beam emitted
by the photoelectric sensor. The patterns used in our prototype at Tx and Rx§
are shown in Fig. 2.4.
The sensor information is gathered using National Instruments (NI) DAQ boards, which operate at
different sampling rates for the Tx and Rx due to their varying rotation speeds. The sampling times are
θTx = 8.3 µs and θRx = 0.4 µs. The optical sensor data is stored together with a GPS timestamp to align
the sensor orientation with the collected impulse response measurement during post processing.
In principle any polarization of the Tx horn can be used in ReRoMA. Linear polarization is widely used
for horn antennas and thus seem like a natural choice. However, in this case the polarization vector of the
§Note that the same tape pattern could have been used on both the Tx and the Rx. The difference in our design was only for
convenience and distinction between Tx and Rx both in hardware and in processing codes.
28

Figure 2.4: Reflector tape pattern on (a) Rx and (b) Tx
redirected wave alters its orientation angle depending on the position of the mirror: for example, if the
E-vector of the wave emanating from the horn is along the x-axis, the vector of the reflected wave will be
along the z-axis at 0 degree azimuth of the tube, but along the x-axis when the tube is at 90 degree azimuth.
It is thus beneficial to employ circular polarization, because the redirected wave has the same handedness
for all tube orientations.
At the same time, it is crucial to remember that circular polarization changes its handedness upon
reflection. Thus, assuming the same circular polarization direction employed on both the Tx and Rx, we
would only correctly capture multipath components (MPCs) that have undergone an even number of reflections. An MPC reflected an odd-number of times would be attenuated by the cross-polarization ratio
at the Rx. Therefore, to fully characterize the channel, we must receive both types of circular polarization
- left-handed circular polarization (LHCP) and right-handed circular polarization (RHCP) - at the Rx end.
For this purpose, we switch between the two polarizations, measuring them quasi-simultaneously. Specifically, the switching happens every-other capture using a trigger signal with period of 200 µs. While the
switching between polarizations might be alternatively implemented also at the Tx, this would require
29

polarization switches with higher power handling capabilities. Additionally, in order to be able to estimate the full polarimetric channel matrix, polarization switches would be needed on both Tx and Rx. Such
system is envisioned in future extensions of this work.
In our prototype we have used the Mi-wave Series 145 polarization switch [81]. The switching of
this device occurs in a ferrite section which is made up of a Faraday rotator. It consists of a small circular
ferrite rod which is supported by a Teflon cylinder in a thin-walled stainless steel waveguide. A coil wound
around this waveguide provides the magnetic excitation in the ferrite where different excitation current
values rotate the incident electric field by different angles. The excitation currents required for switching
between LHCP and RHCP were carefully calibrated using a VNA, and the fast switching between current
values was done by a current-mirror transistor circuit in combination with an op-amp to allow for the
required current swing.
2.3.2 Prototype mm-wave sounder
This subsection discusses the creation and detection of the sounding signal. The fundamental design of
our ReRoMA channel sounder is not dependent on the details of these processes - the sounding signal
exciting the Tx horn can be generated in a large variety of ways. Instead, the forthcoming explanation is
intended to provide a clearer understanding of the operational prototype’s performance and the outcomes
(including their limitations) obtained from its measurements. This helps in interpreting the measurement
results and evaluating the sounder’s effectiveness in various scenarios. A high-level schematic for our
sounder can be seen in Fig. 2.2.
For our sounding signal, we use multi-tone signals that resembles Zadoff-Chu sequences found in LTE
and NR (new radio). Our sequences are designed to provide flat spectrum and low peak-to-average power
ratio (PAPR) of signals even when they are filtered and oversampled [119]. The sounding signal spans a
bandwidth of BW = 200 MHz, composed of Nsc = 400 subcarriers. This results in a subcarrier spacing
30

of ∆fsc = BW/Nsc = 500 kHz which enables the clear identification of propagation distances up to a
maximum of Dmax = cτmax = c/∆fsc = 600m where c is the speed of light. We anticipate this range to
be more than sufficient for all our measurement scenarios.
In terms of distance resolution, the 200 MHz bandwidth allows for a precision of ∆d = c∆τ =
c/BW = 1.5m Worth to note that While we currently use this 200 MHz bandwidth primarily as a proof
of concept, the ReRoMA system itself would only impose a limit in the bandwidth of the horn antennas,
enabling possibly tens of GHz if the signal generation electronics can provide that.
The sounding waveform is digitally shifted to passband. For the measurement, it is initially triggered
by a 1 PPS (pulse per second) signal from a GPS-disciplined clock and then continues to repeat indefinitely.
Each cycle of this waveform lasts for Twf = 2 µs.
The diagram of the Tx RF chain is shown in Fig. 2.5. The previously mentioned pre-loaded sounding
waveform is output from the AWG and upconverted by an initial mixing stage to a first intermediate
frequency (IF) with center frequency of 3.7GHz where one of the sidebands is filtered before the signal is
upconverted to the 60 GHz band, where a band-pass filter is used to confine the transmitted frequencies
within the 59 − 61GHz band.
In the RF stage, a power amplifier boosts the signal power to 22 dBm, after which a polarizer forces
the signal’s polarization to LHCP. Finally, the signal is sent from the Tx antenna. For our experiments, we
utilized a conical horn antenna with a beamwidth of 25◦
.
The Rx RF chain’s layout is depicted in Fig. 2.6. The signal transmitted through the wireless channel
is received using a conical horn antenna with a beamwidth of 9
◦
. The antenna output then enters a dualfunction device that acts as both a polarization-switch and polarizer, capable of alternating between LHCP
and RHCP. Following this, the signal passes through a cascade of a bandpass filter, a low-noise amplifier
(LNA) and a variable attenuator that allows for the fine-tuning of the input power into the downconverter
based on the specific measurement scenario.
31

AWG
BPF
59-61 GHz
Attenuator
12dB
LNA
23 dB DC Blocker
LO
3.7 GHz
BPF
3.3-3.7 GHz
LPF
100 GHz
LO
14.25 GHz
x4
Frequency
Multiplier
PA
30 dB Polarizer
GPSDO
Figure 2.5: Detailed Tx chain
Digitizer
BPF
59-61 GHz
Variable
Attenuator
LNA
40 dB
LO
3.7 GHz
BPF
3.3-3.7 GHz
LO
14.25 GHz
x4
Frequency
Multiplier
Polarizer
GPSDO
Polarization
Switch
LNA
23 dB
Figure 2.6: Detailed Rx chain
The output of the attenuator is downconverted to 3.7GHz IF where further filtering is applied. Subsequently, a second down-mixing process brings the center frequency to 300 MHz frequency. This step is
followed by another amplification using a second LNA, which is tailored to match the voltage swing requirements of the digitizer. Finally, the processed signal is captured and stored by a National Instruments
PXIe-5162 digitizer for later analysis and processing.
A summary of all sounder parameters is given in Table 2.1.
32

Parameter Symbol Value
Frequency subcarriers Nsc 400
Waveform duration Twf 2 µs
RF frequency range fstart − fend 60.3-60.5 GHz
Measured bandwidth BW 200 MHz
Maximum bandwidth BWmax 1 GHz
Tx Antenna 3dB Beamwitdh βTx 25◦
Rx Antenna 3dB Beamwitdh βRx 9
◦
Tx/Rx rotation range [ϕstart : ϕend[ [−180◦
: 180◦
[
Tx rotation resolution ∆βTx 10◦
Tx antenna positions MTx 36
Rx rotation resolution ∆βRx 5
◦
Rx antenna positions MRx 72
Capture trigger period Ts 200 µs
SIMO snapshot duration TSIMO 28 ms
MIMO snapshot duration TMIMO 1.038 s
Dynamic range DR 45 dB
Sampling rate Rs 1.25 GSps
Table 2.1: Sounder parameters
The timing for capturing the signal is synchronized using a GPS-disciplined 10 MHz and 1 PPS signals.
Initially, the capture process is activated by the 1 PPS signal, enabling the synchronization of the AWG’s
transmission and the digitizer’s reception with the precision of the 1 PPS signal. Following this initial 1 PPS
trigger, a subsequent trigger, derived from the 10 MHz signal, governs the sampling process, operating
with a 200 µs period. This secondary trigger maintains accurate synchronization between the Tx and
Rx. Additional clock drift experienced during the measurement campaign can be corrected for in postprocessing by linear interpolation of the drift in between reference LOS calibration measurements.
The chosen trigger period is set to facilitate the capture of the received signal by the digitizer each
time the receive direction rotates by half the beamwidth of the antenna. Since the sampling periodicity
is considerably larger than the duration of the sounding signal, the deadtimes during reception can be
33

Figure 2.7: Timing sequence and triggering
used for streaming the sampled signal to an external recording devices. Such duty cycling would not
be necessary when the backplane of the digitizer supports a data rate that is sufficient for continuous
streaming.
To coordinate the switching of the polarization switch, another trigger signal is employed, featuring a
100 µs delay relative to the primary 200 µs trigger. This switching occurs during the digitizer’s deadtime.
This method ensures that the polarization switch transitions are integrated into the capture cycle, avoiding any interference with the ongoing data acquisition. Fig. 2.7 illustrates the various triggers and their
respective roles in the signal capture process is provided in.
2.4 Calibration Measurements
2.4.1 Antenna and system calibration
The back-to-back (B2B) calibration of our setup was done as follows: We first measure the over-thewaveguide (HOTW(fk)) system response by disconnecting the horn antennas from the setup in Fig. 2.2 and
connecting the corresponding ports with a cascade of waveguides, where fk is the frequency index. This
34

allows to measure the full frequency response of our system in addition to these waveguide extensions.
The frequency response of the waveguide extensions (HWaveguides(fk)) is then measured with a VNA. The
B2B calibration response, per polarization direction o as chosen by the polarization switch at the Rx side,
can therefore be calculated as
HB2B|o
(fk) =
HOTW|o
(fk)
HWaveguides(fk)
(2.1)
We next calibrate the antenna characteristic, by performing pattern measurements, for the Tx and
the Rx separately, in an anechoic chamber. The device-under-test (DUT) is rotated in azimuth in steps
of 5
◦
, while a reference horn antenna with known characteristic is used at the other link end. These
measurements were performed twice, once with the DUT being the horn antenna alone, and once with the
DUT being the whole (FO+MO) setup of one link end, to observe the pattern differences between those two
situations. We note that the beampattern shape was the same for the two measurements, with a slightly
more attenuation (∼ 1 dB) caused by the mirror reflection. We use the latter pattern measurement, denoted
as bT(ϕ, f), bR(ϕ, f) in our evaluation as outlined in Sec. 2.5.
In addition to the standard calibration procedures described above, and because of the difference of our
sounding method compared to other sounders, we performed two additional verification measurements
to verify (i) the stability of our sounder, and (ii) the comparability of our results to the traditional steppermotor rotating horn sounders.
2.4.2 Time stability
The first measurement evaluates the time and power stability of the sounder over a long duration. We
measured a stationary environment for one hour, with a full MIMO snapshot captured every one second,
using a single GPSDO (GPS disciplined oscillator) connecting to distribution amplifiers on the Tx and the
Rx sides. The measurement was performed in the Epstein Family Plaza at USC. We show a satellite view
35

Figure 2.8: Google Earth view of measurement environment. Tx and Rx locations are marked in red and
blue respectively.
of the environment in Fig. 2.8.
¶
. A rough schematic for the environment was generated in MATLAB, see
Fig. 2.9. This simplified geometry will be used for comparison with measurements in the remainder of the
paper, as it accounts for the dominant propagation mechanisms in the channel.
To verify the stability of the sounder, we plot and overlap the noise-thresholded omnidirectional powerdelay profiles (Omni-PDP), calculated by selecting the direction of the maximum power contribution per
delay bin, of all of the MIMO snapshots captured within the one hour duration, and we show the results
in Fig. 2.10. We clearly observe the highly stable performance of the sounder where all of the APDPs are
well overlapped. The main difference comes from APDP samples that are close to the noise-threshold, 30
dB or more below the peak amplitude.
¶This environment is used in subsequent measurements but with addition of objects acting as reflectors
36

0 10 20 30 40 50 60 70
x-axis (m)
0
10
20
30
40
50
60
70
80
90
100
y-axis (m)
Rx
Tx
Figure 2.9: MATLAB view of measurement environment. Surrounding trees (green) and building walls
(blue). Red and blue dots corresponds to Tx and Rx locations respectively.
0 10 20 30 40 50 60 70 80
distance [m]
-110
-100
-90
-80
-70
-60
Figure 2.10: Omni-PDP for the 1 hour measurement, overlapped
2.4.3 Comparison to stepper motor
The second measurement was to verify the angular stability as well as agreement with the well-established
stepper-motor rotating horn approach. We thus measured the same scenario once with our fast-rotating
37

motors sampling the full MIMO channel every one second, and a second time by manually stepping the
angle of rotation of the tube to mimic a stepper motor approach. Due to the time and effort needed to
perform the manual measurement, only 12 angular steps (one step every 30◦
) were measured at both Tx
and Rx, thus creating a 12 × 12 MIMO snapshot. The measurement was done in the same location as the
previous (long-term) measurement (see Fig 2.9).
To provide a fair comparison between the two different scenarios, we sub-sample the 72 × 36 MIMO
snapshot captured by the motor rotation down to a 12×12 MIMO snapshot, i.e. taking every sixth angular
step on the Rx, and every third on the Tx. Fig. 2.11 shows the APS (angular power spectrum), calculated
as the total power per Tx-Rx antenna pair, generated by the two scenarios; we only plot the angular
range corresponding to five positions on the Rx side, namely the antenna directions where we actually see
reflectors heavily contributing to the channel response. We calculate the RMSE between the two generated
APSs and note a difference of less than 1 dB, confirming the equivalence between our sounding approach
and the stepper-motor rotating horn approach.
2.5 Evaluation Procedure
As outlined in the preceding sections, the objective of this work is to analyze and characterize the doubledirectional propagation channel between two link ends, measured from a Tx to an Rx, though channel
reciprocity allows determination of the reverse direction as well. The dynamic, complex double-directional
channel transfer function can be written as the sum of the contributions from N MPCs [83]:
H(f, φR,q, φT,p) = X
N
l=1
αlbT(ϕT,l − φT,p, f)bR(ϕR,l − φR,q, f)e
−j2πfτl
(2.2)
where αl
, τl
, ϕT,l and ϕR,l are the complex amplitude, propagation delay, direction-of-departure (DoD)
and direction-of-arrival (DoA) of the l
th path, respectively, and bT,p(φT,p, f) and bR,q(φR,q, f) are the
38

Manual rotation 12x12 MIMO APS
8 9 10 11 12
Rx antenna index
2
4
6
8
10
12
Tx antenna index
[X,Y] [10 4]
Index -55.0738
(a)
Motor rotation sub-sampled 12x12 MIMO APS
8 9 10 11 12
Rx antenna index
2
4
6
8
10
12
Tx antenna index
[X,Y] [10 4]
Index -54.921
(b)
Figure 2.11: APS comparison for the 12x12 MIMO snapshot generated by (a) Manual rotation, (b) Motor
rotation with sub-sampling
transmitter and receiver complex antenna gain pattern centered at the p
th and q
th orientation, where the
horn points into φT,p, φR,q respectively.∥ These angles are time-dependent, but change on a scale that is
much longer than the duration of the sounding signal xwav(t); for convenience of notation we henceforth
do not explicitly write this time dependence. This model assumes that wave propagation occurs only in
the horizontal plane, i.e., elevation is not considered.
The received signal at time t is the convolution of the sounding sequence with the channel impulse
response and the system (back-to-back) impulse response, which can also be written as
rmeas(t, φR,q, φT,p) = F
−1
f
{Rmeas(f, φR,q, φT,p)} = F
−1
f
{H(f, φR,q, φT,p)HB2B(f)Xwav(f)} (2.3)
where F
−1
f
is the inverse fast Fourier transform (IFFT) with respect to f, HB2B(f) is the back-to-back
transfer function of the system and Xwav(f) is the sounding waveform.
∥Note that the calibration pattern is measured only once, and has its maximum in direction 0 for both Tx and Rx.
39

For our measurement, taking into account the datasets collected during the measurement campaign
(which include timestamped ADC captures and timestamped Tx and Rx sensor data), the received signal
ˆrmeas on a rectangular grid are approximated from the measured sample values rmeas as
ˆrmeas(nTs, φˆR,q, φˆT,p) = F
−1
f
{Rˆ meas(f, φˆR,q, φˆT,p)} ≈ rmeas[nTs, φR,q, φT,p] (2.4)
where Ts is the ADC trigger period set at 200µ s, and ( ˆφR,q, φˆT,p) are calculated by projecting the measured (Rx,Tx) angular orientation (ωR(nTs), ωT(nTs)) onto the nearest rectangular grid point( ˆφR,q, φˆT,p)
where this rectangular grid was created by taking the center points of the angular orientation between two
consecutive temporal samples on both the Tx and the Rx. A more elaborate approach of interpolating the
diagonal grid into a rectangular grid by means of non-uniform and sparse sampling is envisioned for future
work. The resulting separation between grid points is 10◦
and 5
◦ on the Tx and Rx side respectively. In
this context, we operate under the assumption that the tubes on both the Tx and Rx sides remain stationary during the brief capture duration of 16 µs, corresponding to 8 repetitions (that are averaged) of the
2 µs-duration waveform for a single Tx-Rx directional pair.
The measurement data for the two polarization directions can therefore be arranged in 3-D matrices
whose entries are Rˆ meas|co(fk, φˆR,q, φˆT,p) and Rˆ meas|cross(fk, φˆR,q, φˆT,p), with the dimensions Nf×NRx×
NTx where fk is the frequency index, φˆR,q, φˆT,p are the rectangular antenna positions we have projected
our angular data onto, Nf
is the number of frequency points, and NRx and NTx are the total number of
directions on Rx and Tx sides, respectively. We then eliminate the effects of the system transfer function
(the transfer function of our sounder) and the antenna patterns for both the Tx and the Rx as described in
the calibration section to eventually compute the calibrated directional channel transfer function H as
H(fk, φˆR,q, φˆT,p) =
Rˆ meas|co(fk, φˆR,q, φˆT,p)
HB2B|co(fk)bT(0, fk)bR(0, fk)
+
Rˆ meas|cross(fk, φˆR,q, φˆT,p)
HB2B|cross(fk)bT(0, fk)bR(0, fk)
. (2.5)
40

From this, the directional power delay profile (PDP) is computed as
Pdirec(τ, φˆR,q, φˆT,p) = |F −1
fk
{H(fk, φˆR,q, φˆT,p) · Whann(fk)}|2
(2.6)
where F
−1
fk
is the inverse fast Fourier transform (IFFT) with respect to fk, and Whann is a Hann window
applied in the frequency domain, which suppresses sidelobes in the delay domain at the expense of slightly
broadening the main lobe of the impulse response. Finally, we apply noise thresholding and delay gating,
similar to [8]:
P(τ, φˆR,q, φˆT,p) =



Pdirec if (τ ≤ τgate) ∧ (Pdirec ≥ Pλ)
0 otherwise
(2.7)
where τgate is the delay gating value selected to avoid incorporation of points with longer delay than what
can be created in the considered environment, as well as points with “wrap-around" effect of the IFFT, and
Pλ is the noise threshold. The thresholding eliminates noise-only delay bins, whose contributions would
distort estimation of delay- and angular-dispersion parameters [42]. For our current measurements, τgate
is set to 250 m excess runlength, while Pλ is selected to be 6 dB above the average noise power level of the
PDP.
For analyzing the channel behavior from an “omni-directional" perspective, we synthesize the omniPDPs by an approach similar to [56], i.e. by reconstructing the omni-directional pattern from the full
double-directional capture by selecting the maximum power component (the direction of the highest contribution) per delay bin
Pomni(τ ) = max
φˆR,q,φˆT,p
P(τ, φˆR,q, φˆT,p) (2.8)
41

In addition, to analyze the channel behavior from an angular perspective, we generate the angular and
delay power spectrum (ADPS) at both the Tx and the Rx as
ADPSTx(τ, φˆT,p) = X
φˆR,q
P(τ, φˆR,q, φˆT,p) (2.9)
ADPSRx(τ, φˆR,q) = X
φˆT,p
P(τ, φˆR,q, φˆT,p) (2.10)
In addition to the Fourier-resolution parameter extraction methods, and in order to accurately estimate
the location of some reflectors in the channel and compare it to the ground-truth based on the environment, we use a high-resolution parameter extraction (HRPE) method, namely MUSIC (MUltiple SIgnal
Classification) [102]. This algorithm was specifically chosen because of its effectiveness and robustness
for accurate DoA and DoD estimations for MPCs in the channel. In quasi-static scenarios where a limited
number of MPCs are available, algorithms such as CLEAN or SAGE could be aswell used to extract the
main MPCs in the channel. However, care needs to be taken when such conditions are not met due to the
sampling speed of our prototype; we are sampling at speeds lower than the (Doppler-)Nyquist rate, which
indicates that the phase relationships that we normally have as the basis for such HRPE algorithms are not
present.
2.6 Reference Measurements
In addition to the calibration measurements performed to verify the functionality of the sounder, we have
performed multiple proof-of-concept reference measurements designed to verify the sounder’s ability for
dynamic characterization of the channel and the ability to track different reflectors and clusters of MPCs
over time. These measurements were done in the same environment as those in the previous section, but
with the addition of artificial reflectors that allowed to determine the ground-truth angles and delays of
the associated MPCs from geometric considerations.
42

Figure 2.12: Reference Measurements Scenarios. Scenario 1 is metallic sphere moving along the linear path
from marked start to marked end point. Scenario 2 and 3 are rotation of metallic sphere reflector around
Tx and Rx respectively.
2.6.1 Dynamic sphere reflector measurements
The first set of these reference measurements used a spherical metallic reflector mounted on top of a tripod
and moving along specific tracks as shown in Fig. 2.12. In the first measurement of that set, the sphere
moved on a linear track on one side of the Tx/Rx positions at a speed of 1 m/s. The scenario was measured
for 80 seconds, corresponding to a total of 80 MIMO snapshots. The measurement was performed at night
to avoid time-variant reflectors (passing-by humans) except for the metallic sphere. The convention that
was used for the coordinates system when describing all angles is shown in the compass in Fig. 2.12, which
is adapted throughout the rest of the paper.
Following the process of Sec. 2.5, we evaluate the omni-PDP, where all delays are multiplied by the
speed of light, so that they correspond to distances. We then stack all of these PDPs to visualize the
dynamic evolution of PDP vs time, as can be seen in Fig. 2.13. We can clearly observe the line-of-sight
(LoS) component that never changes throughout the measurement, since nothing has obstructed that path
43

Dynamic omni-PDP vs time
10 20 30 40 50 60 70 80
Time(s)
0
10
20
30
40
50
60
70
80
90
100
delay(m)
-90
-85
-80
-75
-70
-65
-60
-55
Figure 2.13: Dynamic omni-PDP vs time for Linear Sphere movement scenario
during our reference scenario. We also observe the sphere-reflector MPC starting at a delay higher than the
LoS delay and coming closer to the LoS as the measurement progresses, reaching a minimum excess delay
compared to the LoS when it is at the mid-position between Tx and Rx, after which the delay increases
again. We also observe no significant changes in other MPCs seen in the channel, caused by either the
surrounding buildings or the trees as shown in Fig. 2.12.
We also show in Fig. 2.14 the dynamic evolution of the ADPS for the Tx and the joint Tx-Rx APS. First
of all, we notice how each of the MPCs clusters can be dynamically tracked in angle and in delay without
ambiguity. We clearly observe the stationary MPCs that remain constant throughout the measurement,
and the track followed by the spherical reflector that can be individually seen in the ADPS at the Tx Fig.
2.14a, and also in the joint APS, Fig. 2.14b.
In addition to the Fourier-resolution evaluation for this (and subsequent) reference measurements, we
also analyze our sounder’s capability in reflector localization and location-tracking across the different
44

(a) (b)
Figure 2.14: Dynamic evaluation in the linear sphere movement scenario, (a) ADPS at Tx, (b) joint APS.
Shown in grey are the projections on each of the dimensions (marginal data).
MIMO snapshots by localizing the reflector MPCs and accurately estimating their three-dimensional coordinates (τref, ϕT,ref, ϕR,ref), where τref is estimated from a 10−fold oversampling applied to the omni-PDP
(with a corresponding delay resolution of 15 cm), and the angular pair (ϕT,ref, ϕR,ref) is estimated from the
MUSIC spectrum with a 1
◦
angular resolution in both axes. We then compare these coordinates estimates
to the ground-truth coordinates of our reflectors as calculated from the measurement geometry.
For the first reference scenario described above, we observe RMSE (root mean square error) of ∼ 18 cm
in delay, ∼ 10◦
in Tx angle, and ∼ 4
◦
in Rx angle. We conjecture that the larger errors in Tx angle are
caused by the larger beamwidth of the Tx horn.
The second reference measurement with the sphere reflector consists of moving the reflector in a
circular track around the Tx, in order to verify the angular precision of the sounder at the Tx side. The
measurement track can also be observed in Fig. 2.12. The measurement was done with the sphere reflector
rotating around the Tx with a rotational speed of 5
◦
/s, measuring the full circular track in 72 seconds.
We again visualize the data in terms of dynamic evolution of ADPS at the Tx and dynamic evolution of
the joint Tx-Rx APS, in addition to the omni-PDP vs time, as can be seen in Figs. 2.16 and 2.15. Starting with
the PDP evolution, we see the main LOS MPC coming at the correct delay (30 m), the reflector starting
45

(a) (b)
Figure 2.15: Dynamic evaluation in the circular sphere around Tx movement scenario, (a) ADPS at Tx, (b)
joint APS.Shown in grey are the projections on each of the dimensions (marginal data).
at an evaluated delay of 34.5 m in comparison to the ground-truth delay of 34.2 m, which is consistent
with the accuracy of the previous reference measurement. We also note that it is evolving in a way that is
consistent with the trajectory the reflector has taken around the Tx. We also observe how the LOS power
drops significantly when the reflector is in a location where it blocks the LOS path. The rest of the MPCs
stay constant throughout the duration of the measurement.
As for the angular visualization, we can clearly observe the stationary components throughout the
measurement, in addition to no ambiguity in tracking the sphere reflector in its full-circle track around the
Tx. As for the high-resolution evaluations with MUSIC, we note RMSE of ∼ 14 cm in delay, ∼ 10.5
◦
in Tx
angle, and ∼ 5
◦
in Rx angle, which are again consistent with the accuracies of the previous measurement.
The third reference measurement with the sphere is similar to the preceding one, but now with the
sphere moving in a circle around the Rx, see Fig. 2.12. We omit discussing the detailed results of this
scenario because of similarities with the previous reference scenario. We do note however the RMSE
results of ∼ 14.5 cm in delay, ∼ 10◦
in Tx angle, and ∼ 4.8
◦
in Rx angle, which are again within our
sounder’s accuracy estimates.
46

Dynamic omni-PDP vs time
10 20 30 40 50 60 70
Time(s)
0
10
20
30
40
50
60
70
80
90
100
delay(m)
-90
-85
-80
-75
-70
-65
-60
-55
Figure 2.16: Dynamic omni-PDP vs time for Circular Sphere around Tx movement scenario
2.6.2 Rx moving away from Tx measurement
In addition to the previously measured reference scenarios, we have measured a dynamic scenario where
the Tx and Rx were mounted on carts (about 1.5 m above ground), and performed a measurement that
mimics a car-to-car environment where one car is driving away from the other one. A diagram representation of the measurement environment is shown in Fig. 2.17. The Tx stays stationary throughout the
measurement, while the Rx starts moving from a 4 m distance with respect to the Tx at a speed of 1 m/s
until it stops at 56 m away from the Tx. The LoS path is maintained throughout the measurement, while
other propagation paths may vary depending on surrounding objects.
We again visualize the results in terms of dynamic evolution of omni-PDP and the dynamic ADPSs at
both Tx and Rx, see Figs. 2.19 and 2.18. The un-obstructed LoS path is present throughout the measurement,
with increasing delay corresponding to the increasing displacement of the Rx cart. The wall reflection
MPCs start at a wide angle compared to the LoS path; (ϕT, ϕR) = (45◦
, −45◦
) for reflection on building
B, and (ϕT, ϕR) = (168◦
, −168◦
) for reflection on building C. These MPCs then get closer to the LoS
path in both angle and delay as the Rx gets further away from the Tx, which conforms again with the
geometry of the scenario. At the final position (when the Rx is 56 m away from the Tx), we observe
the reflection from building C still present at (ϕT, ϕR) = (110◦
, −110◦
), while the MPC reflected on
47

Tx
Rx
Building
A
Building
B
Building
C
Building
D
24m
32m
14m
10m 14m 15m
38m
20m
12m
52m
Rx
Figure 2.17: Moving Rx measurement scenario
building B has disappeared. We observe the birth of an additional double-reflection MPC reflected on
building C then A before reaching the receiver. We also observe many MPCs that are present throughout
the measurement which were traced to be originated as diffractions caused by cylindrical lampposts and
trees in the measurement environment. An additional scenario (T-intersection with intermittent LOS) is
presented and analyzed in our conference paper [47].
2.6.3 Cart-to-cart T-intersection scenario
We have also measured a dynamic scenario where the Tx and Rx were mounted on carts (about 1.5 m
above ground), and performed a measurement that mimics a car-to-car T-intersection environment. The
measurement was performed on the campus of the University of Southern California (USC) in downtown
Los Angeles, CA, USA; a map of the environment are shown in Fig. 2.20. The Tx stays stationary throughout the measurement, while the Rx starts from a non-line-of-sight (NLOS) position with respect to the Tx,
48

(a) (b)
Figure 2.18: Dynamic evaluation in the Moving Rx away from Tx scenario, (a) ADPS at Rx, (b) ADPS at Tx.
Shown in grey are the projections on each of the dimensions (marginal data). Reflection from building B
can be identified in (b) as the component starting at around (delay, AoD) = (9, 45), while building C can be
identified in (b) as the component starting at around (delay, AoD) = (20, 160) and can be tracked afterwards.
Dynamic omni-PDP vs time
10 20 30 40 50 60
Time(s)
0
10
20
30
40
50
60
70
80
90
100
delay(m)
-90
-85
-80
-75
-70
-65
-60
-55 Figure 2.19: Dynamic omni-PDP vs time for Rx Moving Away from Tx scenario
and progresses along a linear track where it gets into a line-of-sight (LOS) scenario for a short period of
time before going into NLOS again.
The first 8 seconds of this measurement scenario had the Rx stationary (for calibration purposes),
after-which the Rx movement speed was fixed at 1 m/s. The total measurement time is 60 seconds.
Following the process of Sec. III, we evaluate the omni-PDP, where all delays are multiplied by speed
of light, so that they correspond to distances. Sample results for 3 measurement positions indicated with
49

Tx
Rx
Building
A
Building
B
Building
C
Building
D
24m
32m
14m
10m 14m 15m
38m
20m
12m
52m
Rx
Figure 2.20: Measurement environment
yellow circles on Fig. 2.20 are shown in Fig. 2.21 where the main MPCs, their strengths and their corresponding delays are indicated. We move from a NLOS scenario in Fig. 2.21(a) (position 1) to a LOS scenario
Fig. 2.21(b) (position 30) to again a NLOS scenario Fig. 2.21(c) (position 55), which is consistent with the
fact that the first-arriving MPC is the strongest in Fig. 2.21(b) but not Figs. 2.21(a), (c). We also note that
the delays seen in the PDPs accurately correspond to the delays expected from the environment. More
specifically, for position 1, we observe a weak component coming at delay 25.5 m, which corresponds to
the diffraction along the edge of building B. A much stronger component arrives at delay = 39 m, which
corresponds to the component emitted from the Tx, reflected on building C before reaching the Rx. For
position 30, we observe the LOS peak coming at delay = 12 m, as well as several MPCs caused by reflections
of building C or reflections caused by the metallic lamp-posts signified by the green circles in Fig. 2.20.
We go again into the NLOS scenario with position 55, where we see the weak diffraction around the edge
50

0 50 100 150
distance [m]
-115
-110
-105
-100
-95
-90
-85
-80
-75
PDP Omni
X 25.5
Y -97.2352
X 39
Y -79.7244
X 45
Y -93.7081
X 51
Y -94.2091
X 70.5
Y -92.0279
(a)
0 50 100 150
distance [m]
-120
-110
-100
-90
-80
-70
-60
PDP Omni
X 12
Y -60.8055 X 16.5
Y -72.6206
X 24
Y -77.4293
X 33
Y -77.3744
X 36
Y -83.5225
X 54
Y -93.5443
(b)
0 50 100 150
distance [m]
-115
-110
-105
-100
-95
-90
PDP Omni
X 25.5
Y -97.5778
X 39
Y -90.2562 X 42
Y -91.8228
(c)
Figure 2.21: Omni-PDP for the T intersection scenario at time (a) t=0 s (NLOS) , (b) t=30 s (LOS), (c) t=55 s
(NLOS)
Dynamic omni-PDP vs time
10 20 30 40 50 60
Time(s)
0
10
20
30
40
50
60
70
80
90
100
delay(m)
-90
-85
-80
-75
-70
-65
-60
-55
Figure 2.22: Dynamic PDP evolution vs time
of building A, in addition to the strong reflection caused by building D. We also show the dynamic PDP
evolution vs. time in Fig. 2.22 where we can clearly see the position for which we have moved from NLOS
into LOS and then back to NLOS again. Additionally, while we are under LOS conditions, we can observe
the temporal side-lobes (stemming from the inverse Fourier transform of the finite-support spectrum) seen
before the delay of the LOS that are visible because of the LOS path being strong, and are therefore absent
(eliminated by the thresholding) in the case where the quasi-LOS component is weak. Moreover, we are
able to track the reflections off of the buildings, most notably buildings C and D.
0We also show in Fig. 2.23 the dynamic evolution of the ADPS for both the Tx and the Rx. First of all,
we clearly notice how each of the MPCs clusters can be dynamically tracked properly in angle and in delay
51

(a) (b)
Figure 2.23: Dynamic evaluation of the ADPS at (a) Tx, (b) Rx
without ambiguity. We also show in grey the projections of the ADPS on each of the planes, to allow for
easier distinction of the tracks followed by each of the MPCs. The main contributing MPCs are the LOS
path, the MPC reflected against building C, and the MPC reflected against building D. We can clearly see
the birth and death for all of these clusters thus verifying the directional capabilities of our sounder.
2.7 Conclusion
In this section, we have introduced a new concept for channel sounding, called ReRoMA, which retains
the simplicity, low cost and flexibility of rotating-horn channel sounders while drastically improving measurement speed. Our prototype enables scanning of the full 360◦
range in azimuth at both Tx and Rx,
allowing the capture of a MIMO snapshot with 36 × 72 orientation combinations within the stationarity
time of a dynamic channel. A series of reference measurements with a defined-location spherical reflector
verified the angular and delay accuracy of our sounder, showing delay accuracy of 17 cm, and angular accuracy at Tx and Rx of 10 and 4.5 degree, respectively. Moreover, we have performed a proof-of-concept
measurement of a dynamic cart-to-cart scenario at 60 GHz mimicking a car moving away from another,
and shown the ability of the sounder to directionally resolve and track the different propagation paths in
52

angle and delay. Some calibration and reference sample measurements using this sounder will be made
publicly available on the WiDeS website.
The ReRoMA concept is flexible and can be extended in a variety of ways. Firstly, adjustment to other
frequency bands is easily possible. Similarly, increased measurement bandwidth just requires change of
the baseband signal generation/capture and RF filter bandwidth. Different tradeoffs between measurement
speed and Tx beamwidth can lead to higher angular resolution. Finally, fully polarimetric measurements
can be achieved by using a (high-power) polarization switch at the Tx as well.
53

Chapter 3
ReRoMA-based Vehicle-to-Vehicle Measurement
3.1 Introduction
Intelligent vehicles, which facilitate autonomous and semi-autonomous driving, are anticipated to dominate transportation over the next two decades. While much research focuses on on-board sensors and
their intelligent processing, another crucial communication is vehicle-to-vehicle (V2V) communication,
enabling coordinated driving maneuvers.
Inter-vehicle coordination is essential for improving traffic flow and reducing accidents. For instance,
during a lane change, if a car can communicate its intention to merge to the vehicles in the target lane, it
can do so with less inter-vehicle distance, reducing the need for unpredictable acceleration or deceleration.
Furthermore, convoys of cars and trucks driving with small gaps can enhance fuel efficiency and traffic
flow, made possible by reliable V2V communication. From a safety perspective, V2V communication can
alert drivers to obstacles and dangers beyond the "line-of-sight" of their own sensors but detectable by
other vehicles’ sensors. For example, a vehicle can notify others of a tree blocking a lane around a bend or
a stalled vehicle that could cause a chain-reaction accident.
To enhance the integration of information from other vehicles with on-board sensors, it is beneficial
to transmit raw sensor data (e.g., current lidar scans) to surrounding cars. This transmission must be lowlatency, which precludes the use of highly efficient data compression algorithms that introduce delays.
54

Consequently, the data rate required for these transmissions increases to 100-1000 Mbit/s per car, exceeding
current bandwidth capabilities. Thus, new applications must utilize mm-wave frequencies, where several
Gigahertz of bandwidth are available. For instance, in 2016, the FCC in the US allocated 14 GHz of spectrum
between 20 and 100 GHz for new communication applications.
Developing new V2V systems requires a thorough understanding of the V2V propagation channel. Key
metrics such as path loss, shadowing, delay dispersion, and angular spread must be known to develop reliable and efficient signaling methods. However, much of this information remains unknown for mm-wave
frequencies. While some measurement campaigns have been conducted, they have not gathered all relevant data. Specifically, mm-wave systems need adaptive arrays to counteract significant free-space path
loss, but current measurements do not provide the necessary directional channel properties due to limitations in existing channel sounders. Furthermore, measurements in many critical scenarios are lacking.
The goal of this work is to fill these gaps by (i) conducting measurement campaigns in various environments using a novel channel sounder capable of capturing the relevant information and parameters of
interest, and (ii) developing channel models that align with physical reality and provide all necessary data
for system development.
Channel sounding involves exciting the propagation channel with a known signal, measuring the received signal, and extracting the "channel response" from this data. The capabilities of the channel sounder
determine the extent of information obtained; for instance, if the transmitter (Tx) and/or receiver (Rx) uses
omnidirectional antennas, no information about the directions of the multipath components (MPCs) can
be gathered. Narrow-band measurements, which use an excitation signal with a small bandwidth, cannot
provide details about MPC delays but offer a better signal-to-noise ratio, enabling a greater measurement
range between Tx and Rx.
There is extensive literature on V2V channel measurements and modeling below 6 GHz, covering
various propagation environments, including urban, suburban, highway, and rural areas, see [86, 114, 80,
55

77] and references therein. Such campaigns are mostly conducted by having two vehicles drive in the
same or opposite directions on the same street. These campaigns typically derive average path loss and
delay dispersion models from the measured results; the validity of the Wide-Sense-Stationary Uncorrelated
Scattering assumption has also been investigated [19, 18, 96]. Directional characteristics for such channels
have been measured, e.g. in [97, 6, 120].
V2V propagation channel measurements for the mm-wave band have been conducted since the early
2000s under varying conditions and using channel sounders of different capabilities. Descriptions of
such measurement campaigns and modelling are available in the surveys [50, 85] and references therein.
Pathloss, shadowing, and fading characteristics were measured primarily to model the propagation losses
in the mm-wave frequency band. Many of these measurements used omnidirectional antennas [41, 52] to
capture the MPCs coming from all directions. However, due to the higher isotropic free-space pathloss at
mm-wave frequencies, such measurements have a reduced range (maximum distance between Tx and Rx);
furthermore such measurements do not provide any directional (angular) information.
The measurement range can be improved by the use of directional antennas. In most measurements,
horn antennas with a fixed beam directions are used. They can be equipped either at the Tx as in [133,
131, 132] (where measurements were carried out in an urban street scenario at 60 GHz), or at the Rx [53]
(dual-band measurements at 3.2 GHz and 34.3 GHz in an urban street scenario), or at both ends [115, 117]
(dual-band at 2.4 GHz and 39 GHz in four different dynamic V2V scenarios). Other studies focused on the
effect of blockage [103, 126, 21, 126], or modelling the small-scale fading [116, 100] statistics in multiple
bands. Although these studies have been important in terms of modelling power and delay dispersion over
longer distances and with better signal-to-noise ratio, they lose important information about potential
MPCs reflected on surrounding structures, since MPCs arriving or departing at angles outside the horn
antenna’s beamwidth are not visible.
56

To capture directional information, the use of antenna arrays, whether real, virtual or switching/phased
is needed. A detailed survey of the different types of sounders is given in [48] and will not be repeated
here. Real arrays, which have a full RF chain for each antenna element, are cost-prohibitive at mm-wave
frequencies. Switched arrays [93] or phased arrays [14, 16, 107, 26, 27, 30, 29] provide an attractive alternative. Still, to our knowledge, the only switched sounder used for V2V studies is the one of AT&T [30, 29,
59]. It has been used in multiple studies for V2V and V2X environments because of its ability to measure
channels at 28 and 39 GHz, and to capture a full MIMO snapshot in 6 ms. However, the high cost and
development effort for this sounder, made more extreme by the requirement that the sounder is rugged
enough for operation while being driven on rough surfaces, prevents duplication of this principle.
The virtual array principle, where a single antenna element is mechanically translated or rotated, allows construction of sounders with much lower cost and is thus extremely popular for mm-wave frequencies [95, 45]. However, they are only suitable for static scenarios due to the long capture duration because
of the necessary mechanical movement. The main limitation for the measurement speed is the stepper motor moving the antenna. In a recent paper [47, 48] we presented a rotating-horn sounder that is suitable
for dynamic scenarios with a capture duration within the stationarity time of the channel. This is achieved
by employing a patent-pending redirecting rotating mirror arrangement (ReRoMA) [82]. It is important to
note however that such a sounder is not able to measure a MIMO snapshot within a coherence time of the
channel. Rather, ReRoMA-based sounders provide a new trade-off between cost, complexity, and measurement speed, making it an attractive method for channel sounding and modelling when switched/phased
sounders are not available due to reasons of cost, power limitations, or component availability.
3.2 Main Contributions
Our primary scientific contributions are as follows:
57

• Using our ReRoMA-based channel sounder operating at 60 GHz, we perform extensive doubledirectional V2V channel measurement campaigns for different driving scenarios, namely convoy,
overtaking, and driving on opposite sides.
• We verify evaluation results by comparing against the geometry of the environment, and discuss
sample results for physical insights.
• We perform statistical evaluation of the measurement data and estimate omni-directional and directional parameters such as pathloss, RMS delay spread, angular spreads, MPC power distribution,
and stationarity time.
3.3 Measurement Scenarios
We have conducted measurement campaigns in three relevant scenarios: (i) an urban scenario where the
cars, driving in convoy, are surrounded by buildings, vegetation, and traffic, (ii) an urban scenario on a
wide 6-lanes street where cars and the cars are driving in opposite direction, (iii) an urban scenario in
which the cars are overtaking each other. We describe these locations in more detail in the following.
3.3.1 Driving in convoy
This environment covered a driving track starting and ending in front of Vivian Hall of Engineering (VHE)
on the USC University Park Campus (UPC), Los Angeles, CA, USA. A detailed outline of the track is shown
in Fig. 3.1. The start and end points are shown in textboxes on the figure. The length of the arrows describes
the distances travelled until the cars come to a full stop (i.e. at a stop sign). The environment is diverse
in terms of surroundings, with some sections surrounded by vegetation, some by buildings, and some by
trucks and roadwork equipment parked on the sides of the street. The operators of the trucks tried to keep
58

Figure 3.1: Convoy driving scenario, map view
a constant separation distance of 10-15 meters while maintaining a steady driving speed of around 4m/s.
The total measurement duration was about 15mins, generating 874 MIMO snapshots.
3.3.2 Driving on opposite lanes
This environment covered a section of Vermont street outside of USC-UPC, stretching between the West
36th Place and the Jefferson streets. A map-view of the track is shown in Fig. 3.2. The Tx and Rx paths are
shown in orange and brown arrows respectively. The operators of the trucks synchronized their drivingoff time via walkie-talkies, driving along their respective tracks while maintaining a speed of around 4m/s.
The measurement lasted for around 2.5mins, generating 150 MIMO snapshots.
59

Figure 3.2: Opposite side and overtaking scenarios, map view
3.3.3 Overtaking
This environment covered the same section of Vermont street as the opposite-lane scenario. The Tx stays
stationary at the red X mark shown on the map, while the Rx’s path is identical to the previous scenario, and
is shown as brown-colored arrow. The operators of the Rx truck tried to maintain a speed of around 4m/s
throughout the duration of the measurement. The measurement lasted for around 2.5mins, generating 150
MIMO snapshots.
3.4 Evaluation Procedure
As outlined in the preceding sections, the objective of this work is to analyze and characterize the doubledirectional propagation channel between the Tx and the Rx. Using all of the datasets captured with our
equipment, we are able to organize the data of each MIMO snapshot into a three-dimensional matrix
Hmeas(f, ϕRx, ϕTx) where f represents the frequency points over the 200 MHz bandwidth, while ϕRx and
60

ϕTx represent the azimuth orientations of the Rx and Tx respectively. The dimensions of Hmeas are Nsc ×
MTx×MRx. The effects of the system (including antenna within the rotating contraption) transfer function
are then eliminated from the measurement data, and we obtain the calibrated directional channel transfer
function Hmeas(f, ϕRx, ϕTx) per MIMO snapshot. For further detail on the performed data processing,
please refer to [48].
From this, the directional power delay profile (PDP) is computed as
Pdirec(τ, ϕRx, ϕTx)) = |F −1
f
{H(f, ϕRx, ϕTx) · Whann(f)}|2
(3.1)
where F
−1
f
is the inverse fast Fourier transform (IFFT) with respect to f and Whann(f) is a Hann window
applied in the frequency domain for delay-domain sidelobes suppression. We then apply noise thresholding
and delay gating, similar to [42], as
P(τ, ϕRx, ϕTx) =



Pdirec if (τ ≤ τgate) ∧ (Pdirec ≥ Pλ)
0 otherwise
(3.2)
where τgate is the delay gating value selected to avoid incorporation of points with longer delay than what
can be created in the considered environment, as well as points with “wrap-around" effect of the IFFT, and
Pλ is the noise threshold. For our current measurements, τgate is set to 350 m excess runlength, while Pλ
is selected to be 9 dB above the average noise power level of the PDP.
To analyze the channel behavior from an "omnidirectional" perspective, we synthesize the omni-PDPs
using an approach similar to that in [56]. This involves reconstructing the omnidirectional pattern from
61

the full double-directional capture by selecting the component with the maximum power (the direction
with the highest contribution) for each delay bin:
Pomni(τ ) = max
ϕRx,ϕTx
P(τ, ϕRx, ϕTx) . (3.3)
We also consider the PDP in the directional maximum power,
Pmax−dir(τ ) = P(τ, ϕRx,max, ϕTx,max) . (3.4)
where ϕRx,max, ϕTx,max is the direction combination providing the largest path gain as defined in Eq. 3.5.
Using the omni-directional PDPs generated for each MIMO snapshot, we proceed to compute several
condensed parameters that are commonly used to characterize propagation channels, such as path gain (the
reciprocal of the path loss), delay spread, angular spread, power distribution among MPCs, and stationarity
time for the PDPs, see [84, chapter 5-7].
Path gain is computed as the sum of the power in each delay bin in the PDP
P Gq =
X
τ
Pq(τ ) (3.5)
where q can stand for either ”omni" or ”max-dir".
Delay spread is the square root of the second central moment of the PDP
Sτ,q =
sP
τ Pq(τ )τ
2
P
τ Pq(τ )
− T2 where T =
P
τ Pq(τ )τ
P
τ Pq(τ )
(3.6)
62

Angular spreads give a condensed quantification of the angular dispersion in the channel. We first
generate the angular and delay power spectrum (ADPS) at both the Tx and the Rx as
ADPSTx(τ, ϕTx) = X
ϕRx
P(τ, ϕRx, ϕTx) (3.7)
ADPSRx(τ, ϕRx) = X
ϕTx
P(τ, ϕRx, ϕTx) . (3.8)
It is important to note that performing noise thresholding and delay gating before computing the ADPSs is
imperative in order to suppress the accumulation of noise in directions where no significant MPCs occur.
From the ADPSs, we can obtain the single-directional angular power spectrum (APS) at both the Tx
and the Rx by integrating over delays τ . We then proceed to compute the angular spreads from the APSs
by applying Fleury’s definition [39] as
Sϕ =
sR
|exp(jϕ) − µϕ|
2AP S(ϕ)dϕ
R
AP S(ϕ)dϕ
with µϕ =
R
exp(jϕ)AP S(ϕ)dϕ
R
AP S(ϕ)dϕ
(3.9)
The angular spreads following this definition will be unit-less, ranging between 0 and 1, where 0 corresponds to no spread in the channel, and 1 corresponds to power coming uniformly from all directions.
The power distribution over MPCs is another very important parameter for the channel analysis as
it captures the ratio of the strongest MPC versus the other MPCs present in the channel. We define a
parameter κ as
κ =
Pomni(˜τ1)
Pτ˜N
τ˜=˜τ2
Pomni(˜τ )
(3.10)
where τ˜k is the location of the k-th local maximum of the omni-directional PDP Pomni(˜τ ), ordered by
magnitude, so that τ˜1 signifies the location of the largest local maximum. Note that while this parameter
63

has some similarity to the Rice factor, it is not identical, since (due to the limited resolution of the setup)
we cannot extract the strongest MPC, but rather determine magnitude of PDP peaks.
The stationarity of the PDP is measured in units of MIMO snapshots where we compute the crosscorrelation between subsequent PDPs and we mark them as part of the same stationarity region if the
correlation is above a certain threshold α. We mark the end of a region and the start of a new one whenever the correlation falls under α, and we calculate the length of each region based on how many MIMO
snapshots it contains. For our analysis, we have considered stationarity region results for α = 0.7 and 0.9.
Under that principle, we evaluate the stationarity regions for both omni-directional and max-dir PDPs. We
additionally distinguish between the PDPs with/without Time-of-arrival (TOA)-normalization to the LOS
component for both omni and max-dir cases. TOA-normalization in that sense would keep the LOS at a
constant distance, and only changes in the other components of the PDP could lead to non-stationarities.
3.5 Results
Based on the evaluation procedure described in the previous section, we evaluate all 1,174 measurement
positions that were recorded in our campaign, covering a total of 874 MIMO snapshots for the convoy
driving scenario, and 150 MIMO snapshots for each of overtaking and driving in opposite directions. The
results obtained are summarized below.
3.5.1 Power delay profiles
We first show the omni-directional PDP. For the presentation of the results, all delays are multiplied by
the speed of light so that they correspond to distance. The dynamic evolution of these PDPs vs time are
shown in Figs. 3.3, 3.4 and 3.5.
Looking at the convoy driving scenario first, we clearly observe the attempt at keeping a constant
separation distance between the two vehicles. We also observe that in comparison to Fig. 3.1, the positions
64

Figure 3.3: Convoy driving scenario, reflections identification
at which the vehicles stopped (marked by the arrow-ending points on the map) correspond to the time
instants for which the Tx-Rx distance significantly decreased. The minimum Tx-Rx separation distance
was seen at t = 370s, corresponding to the instant at which the cars undergo the U-turn as can be seen
in Fig. 3.1. In general, line-of-sight (LOS) was maintained throughout the measurement, and separation
distance ranged between 4m and 18m. In addition to the LOS component, we observe significant MPCs
with delays up to 65m that were traced back to be surrounding buildings and vehicles high enough to be
captured by our antenna beam (height of two meters), as confirmed by the video recordings. An example
mapping between surrounding objects and the MPCs seen in the PDPs is shown in Fig. 3.3. Worth to note
here that other components seen in the PDPs can be traced and explained from geometry aswell, however
they were omitted for space reasons.
Next we look at the overtaking and driving on opposite sides scenarios as described in Fig. 3.2. We
observe major similarities in the dynamic omni-directional PDPs between the two scenarios as can be seen
65

Dynamic omni-PDP vs time
20 40 60 80 100 120 140
Time(s)
0
50
100
150
200
250
300
delay(m)
-105
-100
-95
-90
-85
-80
-75
-70
-65
-60
-55
Figure 3.4: Dynamic omni-directional PDP vs time for opposite sides scenario
in Figs. 3.4 and 3.5, which is expected given the similarity in the environment for the two scenarios. In
general, LOS has been maintained throughout the measurement duration for both scenarios. An example
mapping between surrounding objects and the MPCs seen in the PDPs for the overtaking scenario is shown
in Fig. 3.5. We observe significant contributions from the metallic light-posts on the sidewalks placed at
regular intervals. We also see the LOS starting at high delay for either scenario, then decreasing monotonically at a rate matching the vehicles’ speeds until it hits a minimum at time t = 75s, corresponding to
the position for which the Tx-Rx separation is minimized. We omit to show an example for the driving on
opposite sides case due to the geographical similarity between the two scenarios.
3.5.2 Angular power profiles
In terms of APSs, we show the dynamic evolution of the joint Tx-Rx APSs with time for the duration of
the measurements, for all three scenarios in Figs. 3.6, 3.7, and 3.8. We also show the marginals, i.e., the
projections of the 3D plot on all three planes for better visualization of the different contributing MPCs.
Starting with the driving in convoy scenario in Fig. 3.6, we clearly observe every turn that was taken
by the cars as we proceed in the measurement. We observe limited contribution from MPCs coming from
66

Figure 3.5: Overtaking driving scenario, reflections identification
different directions, due to the environment being dominated by the short-distance LOS MPC. Such contributions were tracked back through the video footage and were identified as reflections coming from
surrounding trucks and/or buildings. It is worth to note here that an additional threshold was placed on
the data, suppressing weak MPCs. This was done to be able to visualize the LOS MPC throughout the
duration of the measurement (i.e. other components were removed by the threshold to not crowd the plot
or add any ambuiguity in following the track).
For the driving on opposite sides and the overtaking scenarios, we observe considerable similarity
between the dynamic APSs, similar to our observations for the PDPs. There is a main cluster of MPCs
in the LOS direction, but also major contributions from surrounding lamposts placed regularly on both
67

Figure 3.6: Dynamic joint APS vs time for convoy driving scenario
Figure 3.7: Dynamic joint APS vs time for opposite sides driving scenario
sides of the street. Additionally, depending on the measurement positions, there are noticeable reflections
caused by the United States Postal Service (USPS) building located at the intersection between Vermont
and W 36th Street.
3.5.3 Pathloss
We proceed with the analysis of the statistics (over the ensemble of MIMO snapshots) of the channel
parameters, starting with pathloss. The omni-directional pathloss values are reported in the CDF plot given
in Fig. 3.9 for all 3 scenarios. We also plot in Fig. 3.10 the pathloss values for all positions vs the logarithm
68

Figure 3.8: Dynamic joint APS vs time for overtaking driving scenario
-145 -140 -135 -130 -125 -120 -115 -110 -105
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot of omni-directional pathloss in dB
Overtaking
Opposite sides
convoy
Figure 3.9: CDF plot of the omni-directional pathloss in dB
of the distance, where the distance was measured as the Euclidean distance of the Tx and Rx cars whose
coordinates are captured by a positioning sensor. The observations for all measurements show a close
agreement with free space propagation model (FSPL) with small variation around the mean. The pathloss
coefficient evaluated as the slope from the linear fit in Fig. 3.10 is at 1.91 for the omni-directional case and
1.90 for the max-dir case. This is smaller than the FSPL coefficient of 2, which is physically reasonable, as
the total received signal consists of several reflected MPCs in addition to the LOS component that provides
power decaying with d
2
.
69

6 8 10 12 14 16 18 20 22 24 26
10 * log10(Tx-Rx distance (m))
-145
-140
-135
-130
-125
-120
-115
-110
-105
-100
Pathloss (dB)
Pathloss vs log-distance
Max-dir
Max-dir least-square fit
Omni
Omni least-square fit
Figure 3.10: Linear fit for the omni-directional pathloss vs log(d)
3.5.4 RMS delay spread
The RMS delay spread ranges between 5 - 35 ns for overtaking, 7 - 55 ns for opposite sides, and 30 - 110 ns for
the driving in convoy scenario. The CDF plots for the three scenarios can be seen in Fig. 3.11. We observe
significantly higher delay spread for the driving in convoy scenario where we see more contribution from
MPCs of varying path-lengths. We conjecture that this is not a matter of the ”convey" driving style, but
rather of the different type of street in which the overtaking/opposite driving was happening, with the
latter much wider, and fewer buildings and parked cars by the roadside. A plot showing the dependence
of the RMS delay spread on the log-distance can be seen in Fig. 3.12. We observe an increase of the delay
spread with distance, which is consistent with the well-established results for cellular environments. We
also observe that the slope of the increase is considerably larger in the convoy environment than in the
opposite/overtaking environment.
3.5.5 Angular spreads
For the statistics of the angular spread, we show the CDF plots for both angles of departures (AoD) and
angles of arrivals (AoA) for all three scenarios in Figs. 3.13 and 3.14. We note relatively low spreads for
both the departure and arrival angles which is expected given the measurement scenarios (LOS maintained
70

0 20 40 60 80 100 120
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot Delay spread
Overtaking omni
Opposite sides omni
convoy omni
Overtaking max-dir
Opposite sides max-dir
convoy max-dir
Figure 3.11: CDF plot of the RMS delay spread in ns
6 8 10 12 14 16 18 20 22 24 26
10 * log10(Tx-Rx distance (m))
0
20
40
60
80
100
120
RMS Delay Spread (ns)
RMS Delay spread vs distance
Convoy omni
Opposite sides omni
Overtaking omni
Convoy max-dir
Opposite sides max-dir
Overtaking max-dir
Figure 3.12: RMS delay spread vs log(distance)
throughout the measurement). We show the dependence of the angular spreads for both the Tx and Rx
vs the log-distance in Figs 3.15 and 3.16. It can be observed that there is relatively little dependence of
angular spread on the distance, with the exception of the opposite-driving scenario’s AoA distribution. It
is also worth to note that despite this being a peer-to-peer measurement scenario where it is expected for
the statistics for the AoA and AoD to be similar, we only observe that similarity in the convoy driving
scenario. We speculate that this is due to the fact that only during the convoy scenario the Tx and the Rx
have followed the same path with the same surrounding; for the other two scenarios, the Rx was driving
71

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot AoA (Fleury)
Overtaking
Opposite sides
convoy
Figure 3.13: CDF plot of the AoA angular spread in Fleury’s definition
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot AoD (Fleury)
Overtaking
Opposite sides
convoy
Figure 3.14: CDF plot of the AoD angular spread in Fleury’s definition
on the side of the road that is much more rich in surrounding structures and trees, causing significantly
higher angular spreads on the Rx.
3.5.6 Power distribution among MPCs
From physical considerations of the measured scenarios, and given what was observed in the previous
evaluations about low delay and angular spreads, we expect the values of κ to be large (on a dB scale)
since the power is mostly concentrated in the LOS component. This is indeed borne out by the evaluations
as can be seen in the kappa CDF plots in Fig. 3.17 for all three scenarios. In general, the values of κ
72

6 8 10 12 14 16 18 20 22 24 26
10 * log10(Tx-Rx distance (m))
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
AoA Angular Spread
AoA angular spread vs distance
Convoy
Opposite sides
Overtaking
Figure 3.15: AoA Angular spread vs log(distance)
6 8 10 12 14 16 18 20 22 24 26
10 * log10(Tx-Rx distance (m))
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
AoD Angular Spread
AoD angular spread vs distance
Convoy
Opposite sides
Overtaking
Figure 3.16: AoD Angular spread vs log(distance)
range between 12 and 15 dB in the convoy driving scenario where the distance between the Tx and Rx was
mostly maintained constant. Larger ranges are observed for the overtaking and driving on opposite sides
scenarios where we see κ values ranging between 6 and 14 dB. A plot showing the dependence of κ on the
log-distance between the Tx and the Rx is shown in Fig. 3.18 for all three scenarios.
3.5.7 Stationarity time
Finally, we evaluate the stationarity time of the omni-directional PDP for the three driving scenarios and
show the results in the CDF plot for Tstat in Fig. 3.19, for correlation thresholds α = 0.7 and 0.9. We observe
73

4 6 8 10 12 14 16
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot Kappa (dB)
Overtaking
Opposite sides
convoy
Figure 3.17: CDF plot of the Kappa parameter in dB
6 8 10 12 14 16 18 20 22 24 26
10 * log10(Tx-Rx distance (m))
4
6
8
10
12
14
16
Kappa (dB)
Kappa vs distance
Convoy
Opposite sides
Overtaking
Figure 3.18: Kappa vs log(distance)
that for the convoy driving scenario, in 50% of the cases, our evaluated stationarity time is around 5s for
α=0.9 threshold, and between 5s and 20s for α=0.7 threshold. This shows that indeed for our measurement
scenarios, our sampling duration falls comfortably within the stationarity region of the PDP of the channel.
For the opposite-side and overtaking scenarios, the stationarity times are much shorter, since the delay of
the strongest (i.e., LOS) peak changes as the vehicles are moving. As a matter of fact, the stationarity
times are almost completely governed by these evolutions, so that the specific values say less about the
environment and more about the distance between the vehicles as function of time. We also show the
stationarity time of the max-dir PDPs for the three driving scenarios in Fig. 3.20, again for correlation
74

thresholds α = 0.7 and 0.9. We observe significantly shorter stationarity time for the max-dir PDPs since for
this case, and given that our scenarios are all LOS, stationarity is a direct function of the distance between
the vehicles and any fluctuation in the distance would have major effect on the correlation between the
PDPs.
To be able to visualize the effects of the environment on the stationarity time of the PDPs without such
an influence of the LOS component, we additionally provide results for the TOA-normalized PDPs where
we normalize the PDPs in delay to the LOS component. We show the results for both omni-directional and
max-dir PDPs for the three driving scenarios and for correlation thresholds α = 0.7 and 0.9. Starting with
the omni-directional PDPs in Fig. 3.21, we see now much larger range for the stationarity time, especially
in the opposite sides driving scenario, where we see a stationarity region of length about 120s, roughly
80% of the measurement scenario duration. We observe shorter Tstat values for the convoy scenario, where
we see that after normalizing by the LOS delay, surrounding structures start having much larger influence
and thus any major changes in the structures surrounding the vehicles causes non-stationarities.
Finally, we also look at the stationarity time for the TOA-normalized max-dir PDPs and show the
results in 3.22, where now that we are looking at only the maximum direction (corresponding to the LOS
direction in such scenarios), and after calibrating by the changes in the LOS delay, we observe extremely
large values of Tstat.
3.6 Conclusion
In this section, we presented the results of an extensive double-directional measurement campaign for
mm-wave V2V channels, using the ReRoMA sounding principle. We have provided an overview of the
measurement methodology and environments, as well as the signal processing done to extract parameters
of interest in V2V wireless communication system design. We have measured the dynamic channel at
60 GHz in convoy, overtaking, and driving on opposite sides scenarios. We have observed pathloss with
75

0 10 20 30 40 50 60 70 80 90
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot for Tstat in seconds - raw omni-PDPs
Convoy alpha = 0.7
Opposite sides alpha = 0.7
Overtaking alpha = 0.7
Convoy alpha = 0.9
Opposite sides alpha = 0.9
Overtaking alpha = 0.9
Figure 3.19: CDF plot of the stationarity time Tstat parameter in seconds for the omni-directional nonnormalized PDPs
0 10 20 30 40 50 60 70 80
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot for Tstat in seconds - raw max-dir-PDPs
Convoy alpha = 0.7
Opposite sides alpha = 0.7
Overtaking alpha = 0.7
Convoy alpha = 0.9
Opposite sides alpha = 0.9
Overtaking alpha = 0.9
Figure 3.20: CDF plot of the stationarity time Tstat parameter in seconds for the max-dir non-normalized
PDPs
pathloss coefficient nomni=1.91 and nmax−dir=1.90, RMS delay spread ranges in 5-120ns and 3-25ns for omni
and max-dir respectively, angular spread values between 0.05 and 0.35 under Fleury’s unitless definition
(corresponding to roughly 3-20 deg spreads). These values are compared to existing mm-wave literature
in ??. Additionally, we have evaluated power distribution among MPCs as a parameter κ and found its
76

0 20 40 60 80 100 120
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot for Tstat in seconds - TOA-normalized omni-PDPs
Convoy alpha = 0.7
Opposite sides alpha = 0.7
Overtaking alpha = 0.7
Convoy
Overtaking
Opposite sides
Figure 3.21: CDF plot of the stationarity time Tstat parameter in seconds for the omni-directional TOAnormalized PDPs
0 20 40 60 80 100 120 140 160 180
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
F(x)
CDF plot for Tstat in seconds - TOA-normalized max-dir-PDPs
Convoy alpha = 0.7
Opposite sides alpha = 0.7
Overtaking alpha = 0.7
Convoy alpha = 0.9
Opposite sides alpha = 0.9
Overtaking alpha = 0.9
Figure 3.22: CDF plot of the stationarity time Tstat parameter in seconds for the max-dir TOA-normalized
PDPs
values ranging between 6 and 15 dB. And finally, we have evaluated the stationarity time of the PDPs with
and without TOA-normalization to the LOS component and shown that our measurement duration falls
comfortably within the stationarity time of the channel. Up to the authors’ knowledge, no such results are
available in the literature.
77

Table 3.1: Comparison of our results to existing mm-wave literature
Author Sounder type Carrier Freq. Path loss coeff Scenario Delay Spread AoA spread AoD spread
Our work ReRoMA 60 GHz 1.90 Urban 5-120ns omni
3-25ns max-dir 3-20deg 3-20deg
Hoellinger et al. Omnidirectional 26 GHz 1.92 Urban 51 ns N/A N/A
Suburban 80 ns N/A N/A
Zöchmann et al. Omni/ Fixed Horn 60 GHz N/A Urban 1-4 ns N/A N/A
Uyrus et al. Fixed Horn 26.5 GHz 1.49 Parking garage N/A N/A N/A
38.5 GHz 1.58 Parking garage N/A N/A
Park et al. Omni/Fixed Horn 28 GHz N/A Parking Lot 0.17-3.21 ns 7.03-9.09 deg
Chopra (AT&T) et al. Phased Arrays 28 GHz 1.77 Urban N/A N/A N/A
30 GHz N/A Urban 19 ns 27.3 deg 26.9 deg
Boban et al. Rotating Horn 60 GHz N/A Urban 16.3 ns 19.7 deg 19 deg
73 GHz N/A Urban 10.2 ns 25.6 deg 25.1 deg
73 GHz N/A Highway 6.1 ns 23.4 deg 23 deg
Bas et al. Phased array 27.85 GHz 2.051 Urban (Campus) 50-100 ns 18-26 deg 14-23 deg
Yamamoto et al. Fixed Horn 59.1 GHz 0.4 - 1.8 Highway N/A N/A N/A
Takahashi et al. Fixed Horn 60 GHz 1.7 - 2.2 Highway N/A N/A N/A
Wang et al. Fixed Horn 73 GHz 2.7 Urban N/A N/A N/A
Sanchez et al. Fixed Horn 38 GHz N/A Urban 6.52 ns N/A N/A
60 GHz N/A Urban 5.92 ns N/A N/A
These results can serve for the establishment of double-directional channel models for 5G/6G V2V communications systems at mm-wave frequencies, which in turn enable wireless system designers to exploit
delay and spatial diversity and/or multi-stream communications.
78

Chapter 4
Sub-GHz Measurements and Modelling
4.1 Introduction
One of the most active research fields today is that of ad hoc networks, where a collection of communications devices (wired or wireless, mobile or stationary) wish to communicate, but have no fixed infrastructure available, or no pre-determined organization of available links [122]. Ad hoc networks are thus
especially suited for use in situations where infrastructure is either not available, or not robust enough in
times of emergencies. These features make ad-hoc networks useful for communications by public safety
organizations (PSOs), such as emergency responders, since their systems cannot rely on traditional cellular infrastructure, because these systems might be needed after natural or man-made disasters that render
the cellular infrastructure nonfunctional. The basic deployment mode is thus known as infrastructure-free
device-to-device (D2D) communications.
D2D communications between emergency responders and mobile command posts has a long history.
In recent years, there has been a trend to move from proprietary systems to systems following commercial
cellular standards. The fourth generation (4G) Long Term Evolution (LTE) technology of the Third Generation Partnership Project (3GPP) [112] was the first to explicitly foresee a D2D mode, called Sidelink, that
was intended for emergency responders. In early 3GPP releases for D2D technology, it assumed singlestream systems (since LTE devices may have single transmit antennas), and only supported broadcast
79

communication. Later 4G releases introduced unicast communication in addition to enhanced features
such as carrier aggregation and higher modulation schemes. 3GPP fifth generation New Radio (5G NR)
Release 16, introduced support for Sidelink with 2X2 MIMO, enabling transmission of up to 2 layers [5].
Originally, 5G NR Sidelink was only used for V2X technology, but more recently 3GPP has studied how
it can be used for public safety communications. Further enhancements and possibly higher-dimensional
MIMO can be anticipated for future 3GPP 5G releases and also beyond into 6G.
A key deployment scenario for PSOs is the communication between emergency responders inside a
building to a command post set up at street level either directly in front of the building, or a few blocks
away. We call the associated propagation channel here indoor-to-outdoor (I2O), to distinguish it from
the more common outdoor-to-indoor (O2I) channel in which a base station that is in an elevated (rooftop)
position communicates with an indoor user. We furthermore stress that - in contrast to most of the 5G/B5G
developments - interest in emergency-responder communications focuses on the frequency range below
1 GHz, as the longer wavelengths provide greater robustness against outages, a key feature for emergency
communications.
The wireless propagation channel determines the theoretical performance limits of wireless systems,
as well as the actual performance of any practical system operating in such a channel [84]. For this reason,
suitable channel models that are based on extensive and accurate measurements, are required for the
design and performance evaluation of any wireless system. Notably, the channel characteristics depend
on the environment and deployment scenario in which the system operates. Thus, even though channel
models have been developed for more than 60 years, the emergence of new systems, frequency bands, and
deployment scenarios necessitates new channel measurement campaigns and models derived therefrom.
This also holds for the area of emergency communications, for which the relevant channels need to
be further explored. In particular, while the O2I channel has been investigated in considerable details in
the literature, treatment of the I2O channel is much rarer. [9] shows only results for one Tx and two Rx
80

locations. [11, 121] provide pathloss and sample impulse responses, but their main emphasis is on channel
characterization for localization systems. Extensive pathloss measurements for a variety of I2O scenarios
relevant for PSOs are described in [127] at 100 − 700 and at 4900 MHz; some results at lower frequencies
(200 and 600 MHz) are also presented in [61]. Delay dispersion for I2O scenarios with a variety of building
types at two frequency bands (725−800 MHz and 4.9−5 GHz) are described in [78]. Pathloss, small-scale
fading, as well as delay dispersion for propagation from a car into one office building at 4.9 GHz are in
[24]. Quasi-deterministic models are presented in [98, 46] at 800 MHz and 2.6 GHz respectively. More
recently, more work has been done on I2O channel measurements, with focus on Internet-of-Things (IoT)
and smart grid communication systems [67, 28, 68, 99, 58, 64]. These papers presented measurements
results in the 700-868 MHz frequency bands and modelled parameters such as pathloss and delay spread,
showing a pathloss exponent between 2.5 and 2.3, and delay spread ranging between 20ns and 300ns.
While these papers provide important results, they did not analyze the impact of the height of the indoor
station (i.e., which floor) on the channel, nor did they consider outdoor locations that might not have LoS
to the building of interest. Most importantly, they did not measure multi-antenna characteristics, and thus
are not suitable for evaluating the potential performance of modern systems as described above.
One such system is LTE sidelink as described previously, however, since the standard is already established, it is very difficult to modify it, since agreement from most manufacturers and stakeholders would
have to be obtained. Hence, the goal of the research in the current paper (which followed the targets of
a larger research program of/for PSO organizations) was finding a way to increase the reliability of the
communication system without needing to change the standard. In other words, we aim to create an "applique" that can sit on top of the standard, and increase its robustness without the need to change the
standard itself. The performance of this applique needs then to be evaluated in realistic propagation channels that represent the actual deployment scenarios and environments. In particular, we are interested
in the Indoor-to-Outdoor scenario (I2O), where a device carried by an emergency responder is located
81

indoors, typically on a higher floor, and wants to communicate with a command post located outdoor at
street level. Neither a test with generic channel models (such as uncorrelated Rayleigh channels), nor with
the standard 3GPP channels, which do not represent the envisioned deployment scenarios, are acceptable
for such testing; note that this scenario is different from the well-explored outdoor-to-indoor scenario
where an elevated outdoor base station communicates with an indoor device. We thus performed extensive measurement campaigns for the I2O case and evaluated the performance improvement achieved by
our applique in these measured channels.
While the New Radio (NR) sidelink for D2D and Vehicle-to-everything (V2X) have been investigated
in considerable detail in the literature, there has been a dearth of work done in LTE Sidelink for D2D
communications. [60] provides an overview of 3GPP-system based D2D communication. [118] shows the
Block Error Rate (BLER) performance of sidelink in a SISO channel with only additive white Gaussian noise
(AWGN). Other work on LTE Sidelink focuses on multi-hop communications [13, 20], neighbor discovery
[22, 76], synchronization [74], or channel estimation [87]. Although many of these papers investigated
increasing the reliability and robustness of Sidelink, it was still taken as a SISO standard, where the transmitter (Tx) and the receiver (Rx) are each equipped with a single antenna. In addition, there has been
no evaluations of these results over real measured channels, on the contrary, if any verification is done,
is usually done over 3GPP standardized channel models such as D2D in different scenarios (outdoor and
indoor), or outdoor-to-indoor (O2I) models.
Moreover, while the O2I channel has been investigated in considerable details in the literature, treatment of the I2O channel is much rarer. [9] shows only results for one Tx and two Rx locations. [11, 121]
provide pathloss and sample impulse responses, but their main emphasis is on channel characterization
for localization systems. Extensive pathloss measurements for a variety of I2O scenarios relevant for PSOs
are described in [127] at 100 − 700 and at 4900 MHz; some results at lower frequencies (200 and 600 MHz)
are also presented in [61]. Delay dispersion for I2O scenarios with a variety of building types are described
82

in [78]. Pathloss, small-scale fading, as well as delay dispersion for propagation from a car into one office
building are in [24]. Quasi-deterministic models are presented in [98, 46]. While these papers provide
important results, these papers did not analyze the impact of the height of the indoor station (i.e., which
floor) on the channel, nor did they consider outdoor locations that might not have LoS to the building of
interest. Most importantly, they did not measure multi-antenna characteristics, and thus were not suitable for
evaluating the performance of our applique.
4.2 Main Contributions
The main contributions of our work are
• Description of a multi-antenna channel sounder for measurement in the public safety band around
800 MHz. This sounder is based on a switched-array principle and thus capable of measuring channel
dynamics and MIMO characteristics.
• Sample results of power delay profiles (PDP), and angular power spectra (APS) from the measurement campaign for I2O channels and their physical interpretations.
• Statistical results for pathloss and rms delay spread for line-of-sight (LoS), and ”deep LoS” (outdoor
station has LOS to the building, but indoor station is on the side of the building facing away from
the ground station), as well as for non-LoS (NLOS), where the outdoor station does not have LoS to
the building in which the indoor station is located.
• Statistical results for the angular spreads in both azimuth and elevation, and on both the Tx (departure) and Rx (arrival) side, for the three scenarios (LOS, DeepLOS, NLOS).
• Statistical results for the power distribution across multipath components (MPCs), with analysis
done for both maximum-power-beam direction and the omni-directional characteristics.
83

• Proposal of a completely standards-compliant performance improvement approach for LTE-Sidelink
by exploiting well-established antenna diversity techniques.
• Presentation of sample measurements for I2O channels and giving physical interpretations of the
results.
• Presentation of simulation results of Sidelink standard over the measured I2O channels and compare
to results obtained with standard 3GPP channel models.
• Statistical results for the achievable SNR gains of our applique compared to the SISO model.
4.3 Sounder Design
We first describe the channel sounder we constructed for these investigations. It is designed to not only
allow for the measurement of pathloss and delay spread, but also of the multi-antenna (MIMO) characteristics of the channel, and thus direction of arrival (DoA) and direction of departure (DoD). We thus choose an
architecture that combines a multi-tone sounding of the transfer function with a switched-array principle;
this approach represents a good compromise between cost and measurement speed; it was introduced in
[113] and has been successfully used since by various groups. The block diagram is shown in Fig. 4.1; this
implementation is similar in spirit to a previous one from our group [119], though with different arrays,
operating frequency, and bandwidth.
4.3.1 Antenna Array
The arrays at Tx and Rx have identical structure, consisting of an 8-faced cylindrical array with two vertically displaced elements on each face, see Fig. 4.3. The antenna elements were designed as parasitic patches
(see Fig. 4.2a) that provide large elevation beamwidth and good backlobe suppression, as can be seen in
the pattern as simulated by the EM simulation software HFSS [4] in Fig. 4.2b. The horizontal and vertical
84

Figure 4.1: Full system block diagram of the channel sounder.
spacings as measured in-between the phase centers of the antenna elements are 0.5λ each. The radiation
pattern of one element is shown in Fig. 4.4. We can see that an element has a fairly omni-directional
characteristic in the range where it is pointing at (-45 deg to 45 deg in azimuth), with a large backlobe suppression. We also see the fairly large beamwidth in elevation (about 70 degrees), which is necessary since
MPCs at the ground station can have steep elevation angles, especially if the indoor device is on a higher
floor. To further enhance the backlobe suppression of our antenna elements, the patches are mounted on
a cylindrical metallic structure that effectively extends the backplane. The antenna arrays were calibrated
in the anechoic chamber at University of Southern California (USC), over 4π spherical angles in steps of
5 deg in azimuth and 5 deg elevation. This calibration is used for the compensation of the antenna patterns
according to Sec. III. The measured pattern is similar to the result of the simulation of the single element
in Fig. 4.2bb); for space reasons only the azimuth cut of the measured patterns (as polar plots) for the Tx
array is shown in Fig. 4.5. Note that the Rx array shows a similar pattern and thus is not presented here.
85

(a) (b)
Figure 4.2: (a) Single Antenna Element top/side view, (b) HFSS Simulated Antenna Pattern
Figure 4.3: Array Design - Top/side view
86

Figure 4.4: Antenna Radiation Pattern of a single antenna element on the array
Figure 4.5: Tx Antenna Patterns
87

4.3.2 RF Chain Design
The sounding signal is located in a band of width 60 MHz around a center frequency of 820 MHz. This
band is close enough to the 700 MHz public safety band to provide essentially identical channel characteristics, but suffered from significantly less interference at our measurement locations. We also note that
this bandwidth is similar to, or larger, than what is expected to be used for public safety applications below 1 GHz.∗ The sounding signal is a custom-designed multi-tone sequence, similar to the Zhadoff-Chu
sequences used in LTE, but with modifications that reduce the PAPR of filtered and oversampled signals
[40]. We employ 60 subcarriers, leading to 1 MHz subcarrier spacing, corresponding to maximum 300m
propagation distance that can be unambiguously determined; this is anticipated to be more than enough
for all the measured scenarios. Distance resolution, corresponding to the 60 MHz bandwidth, is 5 m. Note
that the waveform of a sounding signal of a channel sounder need not be the same as the waveform used
in the system to be explored based on the measured channels
This sounding signal is pre-computed in MATLAB, loaded into a software defined radio (SDR), namely
a National Instruments USRP 2954, and transmitted in bursts. Each burst consists of 100 snapshots, where
each snapshot in turn repeats the sounding sequence 256 times, thus allowing the measurement of a 16×16
MIMO setup. A dead time is inserted in the transmit signal between each repetition, to allow the switch to
turn to the new position and settle. The output of the SDR is upconverted to the carrier frequency, filtered,
and sent to the elements of the transmit array via a fast electronic switch. A set of 16 training sequences is
transmitted from each TX antenna element (allowing the RX to switch through all of its antenna elements);
then the electronic switch connects to the next TX antenna element.
The signal is transmitted over the propagation channel, which distorts it and adds AWGN (Additive
White Gaussian Noise). To obtain the signals at the different receive antenna elements, these elements are
connected, sequentially, via an electronic switch, to a downconversion chain, such that each repetition of
∗Note that it is always possible to derive a model for narrower bandwidth from a wideband model through simple filtering,
while the converse step is not possible.
88

a training sequence is received by a different RX antenna element.† The downconverted signal is sampled
with 80 Msamples per second in I and Q branch, and the samples are streamed via a USB-3 interface to a
harddisk, stored for later postprocessing.
4.3.3 Timing
Tx and Rx frequency and timing are synchronized via GPS-disciplined Rubidium (Rb) clocks. The basic
triggering and handshaking is achieved via control signals on a wireless bridge, namely the Digi XPress
Eth Bridge, operating at a frequency of 910 MHz. Note that this frequency falls outside of our sharp RF
filter, blocking any potential interference or amplifiers’ saturation that could be caused by the wireless
bridges.
4.4 Measurement Campaign
The measurements were performed at the University Park Campus of the University of Southern California, in Los Angeles, CA, USA. The indoor units were placed in the Hughes Aircraft Building (EEB), a
5-storey building with steel-concrete construction. The windows are regular (non-energy-saving) windows, thus providing relatively small attenuation of signals at 820 MHz. There were two TX (outdoor)
positions: one directly in front of the building in which the RX (indoor device) was located, and one where
there is another 5-storey steel-concrete building (Tutor Hall RTH) in between, see Fig. 4.6.
The Hughes Aircraft Building consists of small offices arranged around an oblong corridor, see Fig. 4.7.
The interior walls are drywall, which has relatively low attenuation, though structures like elevators and
pipes from bathrooms provide higher attenuation. As mentioned above, we distinguish between LoS and
NLoS situations depending on whether the TX has LoS to the facade of the building or not. We furthermore
distinguish between LoS and “deep” LoS depending on whether the RX is in the rooms on the front side
†Note that before downconversion, the signal needs to be filtered by a sharp RF filter to reduce the strong adjacent-band
interference from GSM signals.
89

Figure 4.6: Map of the measurement scenario showing the TX positions for LOS (red contour) and NLOS
(blue contour) scenarios
(with respect to the TX) of the building, or the rear side.‡ The 3D Euclidean distances between TX and RX
range from 25 to 40 m for LoS, 40 to 50 m for deep LoS, and 70 to 90 m for NLoS.
Overall, we measured the channels for 420 location combinations, namely 140 per scenario. Measurements were done for the indoor unit moving across 4 floors, 35 positions per floor per scenario (1 meter
sampling distance per floor).
4.5 Processing and Evaluations
Using the measurement setup described in Section II, we obtained a collection of time-domain raw data that
was stored by the Rx SDR onto the harddisk. We first use a pre-computed indexing file based on the switching sequences on both the Tx and the Rx, the waveform properties such as number of points and sampling
‡Note that some papers call NLoS what we call deep LoS.
90

Figure 4.7: Floorplan of the building in which the RXs were placed. Positions marked in blue were used
for the LOS and the NLOS measurements, whereas positions marked in red were used for the "deep LOS"
measurements
rate, in addition to any hardware-related constraints such as registers’ reset times. This indexing operation divides the received raw data into a received waveform per Tx-Rx antenna elements combination,
where a Fast Fourier Transform (FFT) operation is used to get the frequency domain response from these
measurements. Each measurement is now represented as a three-dimensional matrix Hmeas(fk, nr, nt),
i.e., the transfer function at the fk-th frequency point within the 60 MHz bandwidth (790-850 MHz), using
the nr-th antenna on the Rx and the nt-th antenna on the Tx, respectively. The dimensions of Hmeas are
thus N × NRx × NTx where N is the number of frequency points (60), and NRx = 16 and NTx = 16 are
the number of antennas on Rx and Tx sides respectively .
In order to eliminate the effects of the system transfer function (the transfer function of the transceivers
without the switches and antenna arrays), we perform a back-to-back (B2B) calibration. This involves first
a measurement over a wired cable + attenuator channel, HSys+Cable+Attn. We then use a calibrated Vector
Network Analyser (VNA), Agilent 8720ET, to measure HCable+Attn, the isolated response of the cable and
91

the attenuator. Given those 2 measurements, we are able to obtain the B2B calibration of our system by
simply dividing HSys+Cable+Attn by HCable+Attn:
HB2B(fk) = HSys+Cable+Attn(fk)
HCable+Attn(fk)
(4.1)
We can then obtain the calibrated per-element channel transfer function H(fk, nr, nt) by dividing the
raw data with the B2B calibration data as
H(fk, nr, nt) = Hmeas(fk, nr, nt)
HB2B(fk)
(4.2)
Since each antenna element has a directional characteristic, we henceforth call this transfer function the
"directional" transfer function.
From this, the directional power delay profile (PDP) is computed as
Pdirec(τ, nr, nt) = |F −1
f
{H(f, nr, nt}|2
(4.3)
where F
−1
f
is the inverse fast Fourier transform (IFFT) with respect to f. Finally, we apply noise thresholding and delay gating, similar to [8], which is
P(τ, nr, nt) =



Pdirec(τ, nr, nt) if (τ ≤ τgate) ∧ (Pdirec(τ, nr, nt) ≥ Pλ)
0 otherwise
(4.4)
where τgate is the delay gating value selected to avoid using long delay points and points with "wraparound" effect of the IFFT, and Pλ is the noise threshold to not count delay bins with noise which could
particularly distort delay spread and angular spreads [gomez2023noise]. For our current measurements,
τgate is set to 250m excess runlength, while Pλ is selected to be 6 dB above the average noise level of the
92

PDP. The noise floor is calculated as the power averaged among the first 25% of the delay bins, when sorted
in ascending order of powers. The average noise level is then computed as the noise floor + 5.10dB.
Additionally, we analyze the channel behavior from an "omni-directional" and "Max-dir" perspectives,
where the "Max-dir" PDP is selected as the antenna-pair direction with the highest power, as can be seen
in Equation 4.5,
PmaxDir(τ ) = P(τ, nˆr, nˆt), (4.5)
where the indices nˆr and nˆt are defined as [ˆnr, nˆt] =nr,nt
P
τ P(τ, nr, nt). "Omni" is generated in an
approach similar to the one described in [17], i.e. by reconstructing the omni-directional pattern from the
full double-directional capture by selecting the maximum power component (the direction of the highest
contribution) per delay bin
Pomni(τ ) = max
nr,nt
P(τ, nr, nt) (4.6)
To analyze the channel in all parametric domains, i.e., propagation delay, τ , angle of arrival in azimuth
and elevation, ϕA and θA, and angle of departure in azimuth and elevation, ϕD and θD, the 5-dimensional
beamforming spectrum can be expressed as
BF5D(τ, ϕA, θA, ϕD, θD) ≜
P
k
P
nr
P
nt
|Hˆ (fk, nr, nt)b
∗
nr
(ϕA, θA; fk)b
∗
nt
(ϕD, θD; fk)e
j2πfkτ
|
2
P
f
P
nr
|bnr
(ϕA, θA; fk)|
2 P
nt
|bnt
(ϕD, θD; fk)|
2
, (4.7)
where Hˆ (f, nr, nt) is defined as the processed channel transfer function after the delay gating and noise
thresholding, P(τ, nr, nt) = |F −1
f
{Hˆ (f, nr, nt)}|2
, and bnr
(ϕA, θA; fk) and bnt
(ϕD, θD; fk) are the calibrated antenna pattern of Rx and Tx array, respectively. The operation (·)
∗
stands for the complex conjugate. The beamforming spectrum is maximized when the MPCs fall onto the parameter grids.
93

Since the elevation beamwidth is relatively large, it is possible to suffer from aliasing in elevation.
Thus, only azimuth angles are considered in the following analysis. The 3-D beamforming spectrum is
then defined as
BF3D(τ, ϕA, ϕD) = X
θA
X
θD
BF5D(τ, ϕA, θA, ϕD, θD), (4.8)
where the contributions from all elevations are combined.
Using the directional and omni-directional PDPs described in the previous section, we proceed to compute several parameters that are commonly used to characterize propagation channels, such as path loss,
RMS delay spread, power distribution across MPCs, and angular spreads.
4.5.1 Path loss
We compute path loss as the sum of the powers in each delay bin in the PDP:
P L =
X
τ
Pomni(τ ) (4.9)
4.5.2 RMS Delay spread
To calculate the RMS delay spread, we start out by computing the normalized first-order moment, the
mean delay, that is given by
T =
P
τ Pomni(τ )τ
P
τ Pomni(τ )
(4.10)
94

The square root of the normalized second-order central moment is known as rms delay spread and is
defined by
Sτ =
sP
τ Pomni(τ )τ
2
P
τ Pomni(τ )
− T2 (4.11)
Note that long-delayed samples with small power can have a disproportionate impact on the delay
spread; for this reason, noise thresholding and delay gating employed as explained in the previous section
are especially important in the assessment of delay spread.
4.5.3 Angular spread
Angular spreads are another key channel parameter that gives a quantitative description of the angular
dispersion. This can be modelled for both the Tx and the Rx, in both azimuth and elevation directions. We
use Fleury’s definition [39]
Sϕ =
sR
|exp(jϕ) − µϕ|
2AP S(ϕ)dϕ
R
AP S(ϕ)dϕ
(4.12)
with
µϕ =
R
exp(jϕ)AP S(ϕ)dϕ
R
AP S(ϕ)dϕ
(4.13)
where AP S(ϕ) is the Angular Power Spectrum averaging from the generated beamforming spectrum,
taking AP S(ϕA) = P
τ
P
θA
P
ϕD
P
θD
BF5D(τ, ϕA, θA, ϕD, θD) as an example. The angular spread
following this definition will be unit-less, ranging between 0 and 1, where 0 corresponds to no spread in
the channel and 1 corresponds to power coming uniformly from all directions. It is important to perform
the noise thresholding and the delay gating methods described in the previous section before computing
95

the APS, in order to suppress accumulation of noise in the direction in which no significant MPCs occur
[gomez2023noise].
For our measurements, and given the 3D MIMO nature of our sounder, we are able to evaluate the
APSs and hence the angular spreads at both the Tx (Angles of Departure) and the Rx (Angles of Arrival),
and for both azimuth and elevation directions.
4.5.4 Power distribution over MPC
The contribution from each MPC is extracted from the 3D beamforming spectrum. The location of a certain
MPC with i as its index, (τi
, ϕA,i, ϕD,i), can be acquired by finding the local maximum from the spectrum
BF3D(τ, ϕA, ϕD).
The delay resolution is 5 meters, and the azimuthal 3dB beamwidth of the array element is approximately 60deg. The channel observation allows the collection of signals from a wide range in spatial and
temporal, which is not effective to distinguish the LOS component. Finer grid size helps in localizing LOS
with better accuracy in parameters, resulting in better estimation of power. Therefore, in the beamforming
generation, the grids in the delay domain are ten multiples of frequency points, resulting in 0.5 meter delay
resolution. Similarly, higher angular resolution can be acquired by the beamforming from the correlation
between the channel observation and the well-calibrated antenna pattern, for example, 5 degree spacing
in our analysis. Note, however, that such a denser sampling does not change the fundamental resolution
limits induced by the finite aperture or bandwidth.
However, it is still possible that the MPCs are located between two gridpoints. Thus, the consideration
of multiple surrounding grids is a solution to overcome the aforementioned problem. In this paper, for
each MPC candidate, we select a lattice centered at (τi
, ϕA,i, ϕD,i). In the delay domain, 3 delay gridpoints
centered at τi are included. In the angular domain, grids within 20 degree on each side are considered
96

to contribute to the same MPC. The lattice denoted by its center location is expressed as (τi
, ϕA,i, ϕD,i),
where the bold Greek letter refers to the vector of multiple grids in each parametric domain.
We then define a parameter κ5D as
κ5D =
BF3D(τ1, ϕA,1, ϕD,1)
PN
i=2 BF3D(τi
, ϕA,i, ϕD,i)
(4.14)
where i refers to the location of the i
th local maximum of the 3D Beamforming spectrum BF3D, ordered
by magnitude, so that the lattice centered at (τ1, ϕA,1, ϕD,1) signifies the location of the largest local
maximum. Note that - due to the reliance on local maxima - this parameter is not necessarily identical to
the Rice factor.
4.6 Results
Following the evaluation procedure described in the previous section, we evaluate all 420 measurement
positions that were recorded in our campaign, varying across 3 different scenarios (LoS, DeepLoS, NLoS)
and across the 4 floors of EEB. The results obtained are summarized below.
4.6.1 Power delay profiles
We first show three sample PDPs, one in a LoS scenario, one in a DeepLoS scenario, and one in a NLoS
scenario. All these cases are on the 2nd floor of the building, and towards the same direction relative to
the outdoor Tx. For the LoS PDP in Fig. 4.8, apart from a strong peak that at the correct LoS distance
(33m), we observe a very slow exponential decay; we conjecture that this is due to the Rx being indoor and
thus the corridors are acting as a waveguide, so that MPCs are coming from all directions and with a lot
of delay dispersion. In the deepLoS scenario shown in Fig. 4.9, we observe a similar characterstic with the
slow exponential decay and the clear peak at the (quasi)LOS distance; however we also see a few MPCs
97

Figure 4.8: Sample PDP for LOS scenario, 2nd floor position 4, dT x−Rx = 33m
Figure 4.9: Sample PDP for DeepLOS scenario, 2nd floor position 4, dT x−Rx = 41m
with significant powers arriving after the LoS peak. We observe about 10 dB power loss for the DeepLoS
compared to the LoS scenario.
The NLoS scenarios generally show a stronger delay dispersion of the MPCs when compared to the
LoS case, since dominant LoS components are missing and therefore, the difference in power between the
various MPCs is less. This is clear in Fig. 4.10 where we observe multiple clusters that are of comparable
power to the shortest-delay path. From our geometrical analysis, these MPCs are mainly due to diffraction
around the edges of the shadowing building, and reflections from surrounding buildings.
98

Figure 4.10: Sample PDP for NLOS scenario, 2nd floor position 4, dT x−Rx = 75m
4.6.2 Angular power spectrum
In terms of APSs for LoS, DeepLos, and NLoS scenarios, we show sample APSs for the same positions that
we plotted the PDPs for in the previous section. All these cases are on the 2nd floor of the building, and
towards the same direction relative to the outdoor Tx. For the direction of departure in the LoS case seen
in Fig. 4.11, we observe a high concentration of power pointing directly in the LoS direction (i.e. the angle
for which the Tx and Rx are facing each others), in both azimuth and elevation. For the arrival angles,
however, we see a much less concentrated power, with significant contributions from MPCs coming from
different directions. We note that the corridors Azimuthal Direction-of-Arrival with respect to the receiver
antenna array are around 1 and 4.5 rad. This goes again to confirm the "waveguiding" effect caused by
the corridors. The DeepLoS scenario seen in Fig. 4.12 shows a similar pattern for the departure, which
makes sense since we are radiating into the same direction. The arrival however shows concentration of
the power coming from the direction of corridors connecting the front and the back of the building.
In the NLoS scenario, Fig. 4.13 shows a change in the APS of departure where the power is now divided
into multiple clusters all heading towards and around the shadowing building, and reception of the highestpower component comes from those directions as well. This signifies that the direct path going through
99

Figure 4.11: Sample APS for LOS scenario, 2nd floor position 4, dT x−Rx = 33m
Figure 4.12: Sample APS for DeepLOS scenario, 2nd floor position 4, dT x−Rx = 41m
the blocking building was sufficiently attenuated to the point that another MPC with longer delay has
arrived with less attenuation. We still also see some other MPCs coming from different directions as seen
in the previous cases as well. Fig. 4.14 shows the side-by-side comparison of the MPCs in the APSs and
the environment structure. One cluster of MPCs (shown in red color) is penetrating into the shadowing
building (RTH), and two other clusters (yellow and dark blue) diffract around the edges of the building.
Depending on the measurement positions, there are also noticable reflections caused by the Gerontology
building across McClintock Avenue.
100

Figure 4.13: Sample APS for NLOS scenario, 2nd floor position 4, dT x−Rx = 75m
Figure 4.14: Side-by-side sample view of the main propagation processes involved in NLOS positions
101

PL (dB) LOS DeepLOS NLOS
Floor Mean Std deviation Mean Std deviation Mean Std deviation
2 -65.45 9.68 -79.89 2.62 -99.31 6.65
3 -66.41 9.63 -81.23 2.44 -96.14 6.61
4 -71.84 3.39 -87.25 2.72 -100.63 3.77
5 -73.09 2.58 -89.56 2.46 -99.66 4.12
Table 4.1: Pathloss mean and standard deviation for all floors across the different scenarios
4.6.3 Path loss
We proceed with the analysis of the statistics (over the ensemble of locations) of the channel parameters,
starting with path loss. Table 4.1 shows the average and standard deviation of the pathloss across the 3
different scenarios and for all 4 floors.
For the LoS and DeepLoS scenarios, we observe the expected trend of increasing path loss when moving
to higher floors. This is due both to the increased distance between TX and RX, and to the fact that when
the Rx is located on higher floors, the actual LoS is strongly attenuated by the various ceilings it has to
penetrate on its way from the street level to the Rx location - whether on the front of the building (LoS)
or deep into the corridors (DeepLoS). However, this trend does not hold for the NLoS cases, i.e., there is
no clear decreasing trend of power. This is partly due to the flatter elevation angle at which the MPCs are
incident on the EEB building (both because of the larger Euclidean distance between TX and RX, and the
fact that strong paths take a detour around the blocking building). Furthermore, in this scenario there is
rich scattering, with the MPCs coming from many different directions. Moving up in floors might affect
the direct path power, but might lead to new MPC with significant power that might compensate for that
power loss, as observed in the sample NLoS APS from the previous section. In addition, there is about 15
dB loss when going from LOS to DeepLOS positions (propagation through the corridors to the other side of
the building), and about 35 dB loss when going from LOS to NLOS positions (propagation through/around
the shadowing building).
102

Figure 4.15: Parameters definition for the Pathloss modelling
Overall, path loss lies in the range of -60 to -80dB for LoS, -75 to -95dB for DeepLoS, and -90 to -105dB
for NLoS scenarios. Basic link budgets show that communication are possible for all of these situations for
typical handset transmit powers of 20dBm.
For the modelling of the pathloss, we follow a similar approach as the 3GPP and IMT models for
outdoor-to-indoor propagation [3gppportal , 128]
P L = P Lb + P Ltw + P Lin (4.15)
where P Lb is the "basic pathloss" which represents the loss in the outdoor scenario, P Ltw is the penetration loss into the building in which the Rx is, and P Lin is the loss due to the indoor propagation. We show
the parameters for our measurement scenarios in Fig 4.15.
In our analysis, we assume the Free-space Path-loss (FSPL) model for the outdoor propagation, i.e.
P Lb = 20log10(d3D−out) + 20log10(f) + 20log10(
4π
c
) − GTx − GRx (4.16)
where d3D−out is the outdoor 3D distance as shown in Fig 4.15, f corresponds to the operating frequency,
c is the speed of light in vacuum, and GT x and GRx are the antenna gains for Tx and Rx respectively.
103

(a) (b) (c)
Figure 4.16: (a) 3D scatter plot for LOS pathloss modelling, (b) Projection of model onto angular axis, (c)
Project of model onto distance axis
We then model P Lo2i
, the outdoor to indoor propagation from the intercept point of the building into
the indoor unit as
P Lo2i = P Ltw + P Lin = α + β · θi + γ · d3D−in + ϵ (4.17)
where α, β and γ are the parameters of the linear model, and ϵ is the random variation of the data with
respect to its mean, that is commonly modeled as a zero-mean normal distribution ϵ ∼ N (0, σ), where
σ represents the standard deviation of the distribution. These parameters can be obtained by maximum
likelihood estimation (MLE). We separately extract the parameters α, β , γ and σ for the ensemble of LOS
and DeepLOS measurement points. The scatter plots for the LOS and DeepLOS models can be seen in
Figs. 4.16 and 4.17 respectively. Details of the parameters and their confidence intervals are given in Table
4.2.The linear model fits the data nicely, as evidenced by the relatively small confidence intervals for all of
the parameters. The average indoor pathloss proportionality constant γ is 0.5 for LOS points and 0.67 for
DeepLOS points, which is on a similar order as other models such as 3GPP and IMT.
The shadowing fitting results are shown in Fig. 4.18 for both LOS and DeepLOS scenarios, and the
parameters are listed in Table 4.3. The standard deviation of about 4 dB; inspection of Fig. 4.18 shows
104

(a) (b) (c)
Figure 4.17: (a) 3D scatter plot for DeepLOS pathloss modelling, (b) Projection of model onto angular axis,
(c) Project of model onto distance axis
(a) (b)
Figure 4.18: Shadowing models for (a) LOS, (b) DeepLOS
that the lognormal model that is commonly used for the deviations from the linear fit model provides an
excelled approximation to the measured distribution.
The NLOS points exhibit additional attenuation compared to the LOS position and the FSPL, due to
the impact of blockage, and the extra losses involved in the additional propagation mechanisms such as
reflection and diffraction in the various scenarios. Comparing the same Rx positions in both scenarios, and
Parameter Linear model parameters estimated with 95% CI
α αmin,95% αmax,95% β βmin,95% βmax,95% γ γmin,95% γmax,95%
P LLOS 0.2 -0.5 0.5 0.19 0.1 0.3 0.5 0.2 0.8
P LDeepLOS 1.6 -0.6 3.5 0.39 0.27 0.52 0.67 0.28 1.05
Table 4.2: P Lo2i
linear modelling parameters and the corresponding confidence intervals
105

Parameter Statistical model parameters estimated with 95% CI
µ µmin,95% µmax,95% σ σmin,95% σmax,95%
ϵLOS 0.003 -0.01 0.01 4.54 4.23 4.85
ϵDeepLOS 0.002 -0.01 0.01 3.75 3.58 3.89
Table 4.3: Shadowing parameters
Figure 4.19: Excess NLOS pathloss
accounting for the difference in FSPL depending on the Tx-Rx distance, the excess PL, P LExcess|NLOS is
computed as
P LExcess|NLOS,i = F SP LNLOS,i − F SP LLOS,i (4.18)
where i refers to the i
th Rx position. Figure 4.19 shows the CDF of the excess pathloss for NLOS positions compared to LOS positions for all 140 Rx measurement points. The average excess loss is about 11
dB. Further parametric description of the NLOS pathloss was unsuccessful due to the limitation of our
measurement campaign, which featured a narrow range for the incidence angle (10 degrees) and a narrow range for the indoor propagation distance (0.5 meters). Attempts to fit within such a narrow range
made the confidence intervals of the extracted parameters very large; furthermore extrapolation beyond
the measured range of distances and angles should be avoided in any case.
106

RMS Delay spread (ns) LOS DeepLOS NLOS
Floor Mean Std deviation Mean Std deviation Mean Std deviation
2 79.11 2.61 88.62 2.91 124.24 2.17
3 84.51 2.71 98.90 2.49 113.69 2.91
4 89.46 2.92 109.94 2.77 117.46 2.52
5 90.22 2.82 114.01 2.86 115.73 2.67
Table 4.4: RMS delay spread mean and standard deviation for all floors across the different scenarios
4.6.4 RMS delay spread
The RMS delay spread ranges between 75 and 95ns for LoS scenarios, 85 to 115 ns for DeepLoS scenarios,
and 110 to 130ns for NLoS scenarios. The mean and standard deviation for all measurement positions are
shown in table 4.4.
For the LoS and DeepLoS scenarios, we observe an increasing RMS delay spread while we are moving
into higher floors. This is due to the decrease in power of the LoS path and thus more MPCs come into play
with comparable powers. Such a trend does not hold for the NLoS scenarios because for those scenarios,
the particularities of the channel near the Rx and the outside reflected paths that are taken for different
positions have a larger influence than the floor height.
Overall, RMS delay spread is thus in the range of 75 to 130 ns . Thus, for a measurement bandwidth
of our sounding campaign (60 MHz) where the resolvable delay bin is 17 ns, there is ample delay diversity
available in the channel.
The analysis of the CDF of the RMS delay spreads in dBs in Fig 4.20 shows that the lognormal distribution (i.e. Gaussian on a dB scale) provides an excellent fit. We model the two LOS scenarios (LOS and
DeepLOS) together. The standard deviation observed for both the LOS group and the NLOS scenario are
similar (0.8 to 1 dB).
The delay spread shows a linear relationship with respect to the distance based on our analysis of the
various measurement positions in the LOS/DeepLOS scenarios. This is shown in Fig 4.21a where the delay
107

(a) (b)
Figure 4.20: RMS Delay Spread models for (a) LOS/DeepLOS, (b) NLOS
(a) (b)
Figure 4.21: RMS Delay Spread vs distance for (a) LOS/DeepLOS, (b) NLOS
spread increases with increasing propagation distance. However, such a conclusion cannot be made about
the NLOS positions, see Fig 4.21b.
4.6.5 Angular spread
Tables 4.5, 4.6, 4.7, and 4.8 show the means and standard deviations for the spreads of Azimuth of Arrival
(AoA), Azimuth of Departure (AoD), Elevation of Arrival (EoA) and Elevation of Departure (EoD), for
all the measured scenarios and across all 4 floors. Note that we provide results for Full and restricted
cases in the tables of azimuthal spreads (AoA and AoD). For "Full", we are computing the angular spreads
108

AoA LOS DeepLOS NLOS
Full Restricted Full Restricted Full Restricted
Floor Mean Std Mean Std Mean Std Mean Std Mean Std Mean Std
2 0.685 0.032 0.696 0.026 0.748 0.012 0.796 0.017 0.751 0.013 0.799 0.008
3 0.738 0.015 0.749 0.02 0.782 0.021 0.798 0.023 0.832 0.014 0.849 0.003
4 0.786 0.022 0.796 0.017 0.795 0.021 0.794 0.034 0.882 0.028 0.892 0.025
5 0.831 0.015 0.844 0.011 0.782 0.023 0.790 0.028 0.926 0.037 0.949 0.034
Table 4.5: Azimuth of Arrival spread mean and standard deviation for all floors across the different scenarios
AoD LOS DeepLOS NLOS
Full Restricted Full Restricted Full Restricted
Floor Mean Std Mean Std Mean Std Mean Std Mean Std Mean Std
2 0.298 0.015 0.292 0.016 0.301 0.013 0.297 0.016 0.389 0.049 0.394 0.051
3 0.491 0.025 0.292 0.024 0.592 0.026 0.299 0.014 0.401 0.061 0.391 0.044
4 0.302 0.031 0.290 0.028 0.300 0.016 0.289 0.012 0.431 0.038 0.397 0.034
5 0.521 0.038 0.298 0.031 0.689 0.034 0.289 0.023 0.398 0.046 0.391 0.041
Table 4.6: Azimuth of Departure spread mean and standard deviation for all floors across the different
scenarios
from the APS generated by the full Azimuthal direction of the 5D beamforme, while for "Restricted", we
restrict the angular range when computing the APS from the 5D beamformer result to only the components
going towards the building in which the Rx is located. This distinction was done because a significant
floor-dependent reflection was observed from the building behind the Tx in the LOS/DeepLOS scenarios,
introducing a bias in the azimuthal spreads results.
Starting with the AoA results, we observe an increase in the AoA spread with the increase in height
of Rx for the LoS scenario. No such trend is visible for the DeepLoS and NLoS however. Remarkably, the
Elevation of Arrival LOS DeepLOS NLOS
Floor Mean Std deviation Mean Std deviation Mean Std deviation
2 0.608 0.027 0.576 0.019 0.569 0.021
3 0.625 0.017 0.573 0.023 0.562 0.016
4 0.585 0.018 0.558 0.024 0.538 0.021
5 0.564 0.022 0.540 0.022 0.529 0.024
Table 4.7: Elevation of Arrival spread mean and standard deviation for all floors across the different scenarios
109

Elevation of Departure LOS DeepLOS NLOS
Floor Mean Std deviation Mean Std deviation Mean Std deviation
2 0.677 0.025 0.664 0.018 0.478 0.010
3 0.666 0.031 0.648 0.020 0.468 0.009
4 0.591 0.026 0.571 0.024 0.451 0.009
5 0.497 0.027 0.483 0.019 0.464 0.016
Table 4.8: Elevation of Departure spread mean and standard deviation for all floors across the different
scenarios
spread in AoA is very significant, even in LoS positions. This is due to the Rx being indoor and in closeproximity to a lot of reflectors, so that MPCs impinge from all directions. This trend is also seen for the
EoA spreads, which conforms with the previous analysis that the MPCs are being reflected both vertically
and horizontally inside the building in which the Rx unit is located.
The Tx side shows significantly lower spreads for both AoD and EoD compared to the Rx side. This is
due to the Tx being outdoors, where significant clusters of MPCs only go into the directions of surrounding
buildings. The biggest contributor for the LOS cases turns out to be the RTH building behind the Tx; the
effect of that building can be seen more clearly by comparing the azimuthal spreads for both the Restricted
and Full cases described previously. Note however that due to the physical properties of the scenario,
the back reflection was much more significant for the 3rd and 5th floor positions. In addition, for the
elevation of departure, in certain scenarios, significant contributions are coming from a ground-reflection,
thus increasing the EoD spread.
Fig. 4.22 shows the CDFs for the 4 angular spreads in the Restricted case, while Fig. 4.23 shows the
CDFs for the azimuthal spreads in the Full case. We notice the irregular shape of CDF in the latter case
caused by the floor-dependent back-reflection.
4.6.6 Power distribution of MPCs
In this section, we distinguish between the Max-Dir and the Omni PDP as was described in Sec. III. we
expect the values of κ for the directional cases to be larger than the omni-directional ones because of
110

(a) (b)
(c) (d)
Figure 4.22: CDF for the angular spreads computed from Restricted Beamformer across the 3 measurement
scenarios, (a) AoA, (b) AoD, (c) EoA, (d) EoD
(a) (b)
Figure 4.23: CDF for the angular spreads computed from Full Beamformer across the 3 measurement scenarios, (a) AoA, (b) AoD
111

5D Kappa (dB) Floor LOS DeepLOS NLOS
Max Dir
2 8.4275 1.1544 -0.3950
3 7.6622 1.9468 -0.8183
4 7.5037 2.2030 -1.1815
5 7.8218 1.9936 -1.2313
Omni
2 3.7290 -0.3374 -2.7218
3 3.4453 -0.1119 -2.8078
4 4.0429 0.4493 -3.2160
5 3.8968 -0.0979 -3.2752
Table 4.9: 5D Kappa for all floors across the different scenarios
(a) (b) (c)
Figure 4.24: CDF for the κ across the 3 measurement scenarios for both Omni and Max Dir (a) LOS, (b)
DeepLOS, (c) NLOS
the large number of MPCs seen in the APSs and PDPs for the latter case. This is indeed borne out by
the measurements: for the LoS case, the directional mean of κ is 8dB, while the omni-directional mean is
around 3.7dB. For the NLoS measurements, the average κ values are considerably lower, around -0.8dB for
the Max Dir and -3dB for the omni case. The DeepLoS scenarios are somewhere in between - about 2dB
average κ for Max Dir and 0dB for omni-directional.
Fig 4.24 shows the CDFs of the κ5D parameter for both Omni and Max Dir and for all measurement
scenarios.
Fig. 4.25 shows the scatter plots of κ5D vs distance, for both Omni and Max-dir cases. Generally,
κ5D decreases with increasing distance between Tx and Rx. This is expected because with the increasing
distance, the reflected/diffracted MPCs have powers that are closer to the strongest component. This trend
is seen for both Omni and Max-dir evaluated PDPs.
112

(a) (b)
Figure 4.25: κ5D vs distance for (a) Max Dir, (b) Omni
4.7 Sidelink Simulations
4.7.1 Sidelink background and Diversity methods
LTE Sidelink is a mechanism that LTE has set up to provide means for devices, called user equipments
(UEs) to communicate between each other without going through base stations (eNB). Such UE-UE communication is especially important for first responders when the UEs do not have an available connection
to the eNB, either because they are out of range, or the infrastructure has been damaged; e.g., at the scene
of a natural or man-made disaster.§
Any Sidelink system consists of three main functionalities: Synchronization, discovery, and communications. This motivates the structure of the logical, transport, and physical channels¶
that are specific to Sidelink, where some channels handle the transmission of control information (Sidelink Broadcast
Control CHannel SBCCH, Sidelink Shared CHannel SL-SCH, Physical Sidelink Control CHannel PSCCH),
some channels for handling the payload of the Sidelink communications (SL-SCH, Physical Sidelink Shared
CHannel PSSCH), and some channels are more specific for either the Sidelink Synchronization (Sidelink
§Besides the PSO applications discussed here, Sidelink has a variety of other applications, including Vehicle to Vehicle (V2V)
communications, Vehicle to Infrastructure (V2I) communication and Vehicle to Pedestrian (V2P) communication for improved
transportation safety, Internet of things (IOT) applications such as industrial automation, healthcare, out of coverage monitoring
of forest fires and many more.
¶
in this section only, ”channels" is a set of bits for transmission of specific information, while later sections employ ”channels"
as short for "propagation channels".
113

Broadcast CHannel SL-BCH, Sidelink synchronization signal SLSS) or for the discovery process (Sidelink
Discovery CHannel SL-DCH, Physical Sidelink Discovery CHannel PSDCH).
Generally, Sidelink can be viewed as a "transmit-centric" standard, where each UE wishing to communicate will need to send out information in a broadcast mode to a group of users. A communication
between UE A and UE B for example would be a transmit operation from UE A (while UE B is receiving),
followed by a transmit operation from UE B (while UE A is receiving). All tranmissions would occur in
the uplink band or timeslots, depending if it’s FDD or TDD. Consequently, the transmission format closely
hews to the uplink format, where DFT-spread OFDM is used for modulation.
In this section, we propose a completely standards-compliant performance improvement approach for
LTE-Sidelink by exploiting well-established antenna diversity techniques. This approach consists of an
Applique that sits on top of the standard, where the transmit signal is received at several antenna elements,
and the signals from these antennas are then further processed to exploit the desirable difference in the
channel seen by each of these antenna elements. we consider the two main ways of exploiting the signals
from the multiple diversity branches:
• Selection diversity, where the “best” signal copy is selected, while the other copies are discarded. Our
selection criteria is based on the power received per antenna element, averaged across all frequency
points. In addition, in order to show both the best and the worst case scenarios, we show in this
paper the results for Selection Best and Selection Worst techniques, selecting the highest and the
lowest power correspondingly.
• Combining diversity, where all the copies of the signal are combined. In this form of diversity, we
further distinguish two forms of combining techniques: (i) equal gain combining (EGC) gives equal
weights to all antenna elements, irrespective of the channel characteristics seen by each of them, (ii)
Maximum Ratio Combining (MRC) weights the different antenna elements by a weight proportional
114

to the magnitude of the channel seen by each of these antennas. The different weighted links are
then constructively added to generate the combined signal.
In the rest of this section, we will provide simulation results for Selection Best, Selection Worst, Maximum Ratio Combining, and Equal Gain Combining when employed on the receiver side. The importance
of this Applique lies in the fact that these diversity methods can be implemented without any change in the
standard. For the Rx, the selection can be done in RF (requiring only a power sensor), while the combining
could be done in RF (using adaptive phase shifters and adjustable-gain amplifiers, obtaining the required
channel state information (CSI) is difficult with a single RF chain, since the standard does not foresee specific measures to repeat training signals; thus reception by multiple RF chains and combining in baseband
is preferable. Also in this case, the output of the combiner provides an ”effective" channel with improved
SNR and robustness.
For the Tx, we consider antenna selection only. This is motivated by the fact that for Sidelink, reciprocity generally holds, since the transmission from device A to device B are in the same band as from
device B to device A, and thus an RF sensor can determine the best antenna at device A when it is in Rx
mode; the system then uses the same antenna for the Tx case.
4.7.2 Simulation Procedure
In order to study the performance of our applique and the different diversity techniques on D2D channels,
we have used and extended the MATLAB LTE Toolbox. This toolbox is an add-on component of MATLAB,
which provides standard-compliant functions and applications for the design and development of LTE systems, including Sidelink. This toolbox can be used to support the performance assessment of physical layer
development, verify standards development, and generate standard-compliant LTE waveforms. Using the
MATLAB LTE toolbox, we can configure, simulate and measure communication links, and investigate the
115

performance of different Sidelink physical channels. In our implementation, we study the relationship between Block Error Rate (BLER) and the Signal-to-Noise Ratio (SNR), where we simulate the four diversity
techniques discussed previously (MRC, EGC, Selection Best, Selection Worst) for each of the SNR points.
BLER in here refers to the ratio of the number of erroneous blocks to the total number of blocks received
over the applied wireless channel. In order to evaluate the channel’s performance properly, we implement
the key communication modules in both transmitter and receiver, including waveform generation, channel coding/decoding, resource mapping/demapping, QPSK modulation/demodulation, DFT-based OFDM
modulation/demodulation, in addition to our diversity combining module. We vary the SNR of the transmitted signal by changing the power of the noise, and we derive the BLER once the transport block is
recovered after passing through the transmission and reception process chains. We store the BLER data
for each of the different diversity combining techniques and we overlay the plot of the BLER vs SNR waterfall curves for all four of these techniques. We also compute the SISO channel performance by evaluating
the channel response assuming omni-directional elements on both the Tx and the Rx. We then compute
the SNR gains achieved by employing the different diversity techniques compared to the SISO channel.
This simulation procedure was repeated for all 420 measurement positions that were recorded in our
campaign, varying across 3 different scenarios (LoS, DeepLoS, NLoS) and across the 4 floors of EEB.
4.7.3 Sidelink sample simulation
Running the simulation procedure previously described, we show the BLER output curves for the different
combining techniques of the Applique, for one sample measurement position (EEB 2nd Floor, LoS scenario)
in Fig 4.26. We observe a gain for the selection of about 5 dB compared to the SISO case, and an additional
11 dB for the combining techniques (MRC and EGC) compared to the Best Selection method. This sample
result shows the possible SNR gains that can be obtained by employing the diversity techniques of our
applique.
116

Figure 4.26: Sample BLER vs SNR Sidelink result for one measurement position
More importantly, an interesting observation from these figures is that the BLER vs SNR waterfall
curves are significantly steep, where we move from 100% BLER to 0% within a range of less than 5 dB
SNR. This is important because it means that the curves are mostly parallel, and the introduced SNR shift
in between the different curves can be uniquely identified. This indicates that the SNR improvements for
the different diversity techniques can be obtained from the channel function directly, without having to
go through the time-consuming Sidelink BLER simulations, basically establishing a one-to-one mapping
between the BLER curves and the corresponding SNR gains at this particular location. To verify this, we
show in Figures 4.27 the cumulative distribution functions (cdfs), taken over the ensemble of measurement
points) of the SNR gains for both LOS and NLOS scenarios, once using the sidelink simulation BLER curves
output, and once using the channel transfer function directly. We see a really close correspondence in
between the 2 methods, seeing a maximum of 0.5 dB SNR difference. Following this observations, for the
rest of the paper, we will be providing all the SNR distribution results computed directly from the channel
transfer function. Note that the above simulations use a bandwidth of 20 MHz.
117

(a) (b)
Figure 4.27: CDF of SNR gains evaluated from Sidelink Simulations vs directly from the channel transfer
function, for LOS and NLOS scenarios. Evaluations done for second floor of EEB only.
4.7.4 Diversity gains
Running the simulation procedure previously discussed, we show the average SNR gains for the different
diversity techniques in comparison to the omni-omni SISO channel in table 4.10, and the CDFs for the SNR
gains can be seen in Fig 4.28 for the three scenarios. These are the gains observed for a bandwidth of 20
MHz, i.e., wideband operation. We observe a gain of around 5dB for the simple Selection diversity methods
that can be implemented in RF by the use of only a power sensor. We observe another 7dB of SNR gain
between MRC and the Selection best method, which speaks to the richness of the environment in MPCs,
which conforms again with our sample APS plots shown in the previous section. Another interesting
observation from the table is that MRC and EGC are performing very similar, with about 0.6-1 dB difference
on average. This suggests that if combining is done in RF, the extra achievable gain obtained by MRC might
not be worth the trouble of requiring adjustable gain amplifiers on all the RF chains.
It is worth noting that the pathloss is completely different across these scenarios. We describe the
pathloss distribution in more details in [10]. In this paper, we are only showing the diversity gains because
the pathloss is irrelevant when we are showing the relative gains vs the SISO channel, since pathloss affects
all schemes the same way.
118

Table 4.10: Mean SNR gain for different diversity techniques
LOS DeepLOS NLOS
MRC vs SISO 12.41 dB 12.46 dB 12.48 dB
EGC vs SISO 11.81 dB 11.65 dB 11.59 dB
Selection Best vs SISO 5.06 dB 5.12 dB 5.18 dB
Selection Worst vs SISO -4.46 dB -8.63 dB -10.6 dB
(a) (b) (c)
Figure 4.28: CDF for the SNR gains for LOS, DeepLOS and NLOS scenarios. Ensembles consist of all
measurement points (all floors).
4.7.5 Comparison to standard channel models
Assuming an omni-directional antenna on the Tx, and following the 3GPP 36.873 3D O2I LOS and NLOS
models simulated in Quadriga, we have generated the double-directional channel impulse response for a
1x16 Single-Input Multiple-Output (SIMO) channel employing two antenna array configurations on the
Rx: (i) a 16-antennas Uniform Linear Array (ULA) with 0.5 λ spacing, and (ii) our constructed 16-elements
cylindrical antenna array (described in Sec. II.A). We simulate the possible SNR gains from the Rx diversity
techniques from the channel transfer function directly for those 2 cases and compare it to the achievable
gains as evaluated from our measurement campaign.
We show in Fig 4.29 the comparison between those three cases and the performance of our diversity
applique on each of those models. In order not to overload the figures, we only show the comparison of
SISO and MRC for the three different configurations (ULA-3GPP model, CYL-3GPP model, CYL-measured);
we note here that MRC and EGC give essentially the same results. Furthermore, we show here the results
119

(a) (b)
Figure 4.29: SNR gain distribution for the different diversity techniques for 3GPP and Measured channel,
LOS and NLOS scenarios
for narrowband operation, which will demonstrate in a more pronounced way the different slopes of the
cdfs. We firstly observe sigificant differences between the ULA and the cylindrical antenna configurations,
but also the diversity gain difference for the cylindrical array between 3GPP model and our measurements
is several dB.
120

Bibliography
[1] url: https://telcomaglobal.com/p/5g-key-performance-indicators.
[2] url: https://www.ntia.doc.gov/files/ntia/publications/2003-allochrt.pdf.
[3] url: https://kevinbird61.github.io/simulation-wireless-802.11/.
[4] url: https://www.ansys.com/products/electronics/ansys-hfss.
[5] 3GPP Technical Specification 38.211, "NR; Physical channels and modulation. url: https://portal.
3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3213.
[6] Taimoor Abbas, Johan Karedal, Fredrik Tufvesson, Alexander Paier, Laura Bernadó, and
Andreas F Molisch. “Directional analysis of vehicle-to-vehicle propagation channels”. In: 2011
IEEE 73rd Vehicular Technology Conference (VTC Spring). IEEE. 2011, pp. 1–5.
[7] Naveed A Abbasi et al. “THz band channel measurements and statistical modeling for urban
microcellular environments”. In: IEEE Transactions on Wireless Communications (2023). early
access.
[8] Naveed A Abbasi, Jorge Gomez-Ponce, Revanth Kondaveti, Shahid M Shaikbepari, Shreyas Rao,
Shadi Abu-Surra, Gary Xu, Charlie Zhang, and Andreas F Molisch. “Thz band channel
measurements and statistical modeling for urban d2d environments”. In: IEEE Transactions on
Wireless Communications (2022).
[9] Oussama Akhdar, David Carsenat, Cyril Decroze, Moctar Mouhamadou, and Thierry Monédière.
“Direction of arrival measurements for outdoor-to-indoor channel characterization”. In: 2009 3rd
European Conference on Antennas and Propagation. IEEE. 2009, pp. 2280–2282.
[10] Hammoud et Al. “Propagation Channel Measurements for Indoor-to-Outdoor Communications
for Device-to-Device Public Safety Applications”. In: In Preparation ().
[11] Bardia Alavi, Nayef Alsindi, and Kaveh Pahlavan. “UWB channel measurements for accurate
indoor localization”. In: MILCOM 2006-2006 IEEE Military Communications conference. IEEE. 2006,
pp. 1–7.
121

[12] Hussein Ali Ameen, AK Mahamad, S Saon, D Md Nor, and K Ghazi. “A review on vehicle to
vehicle communication system applications”. In: Indonesian Journal of Electrical Engineering and
Computer Science 18.1 (2020), pp. 188–198.
[13] Leonardo Babun, Ali Ihsan Yürekli, and Ismail Güvenç. “Multi-hop and D2D communications for
extending coverage in public safety scenarios”. In: 2015 IEEE 40th local computer networks
conference workshops (LCN Workshops). IEEE. 2015, pp. 912–919.
[14] Celalettin Umit Bas et al. “A real-time millimeter-wave phased array MIMO channel sounder”. In:
2017 IEEE 86th Vehicular Technology Conference (VTC-Fall). IEEE. 2017, pp. 1–6.
[15] Celalettin Umit Bas. “Real-time channel sounder designs for millimeter-wave and ultra-wideband
communications”. PhD thesis. University of Southern California, 2018.
[16] Celalettin Umit Bas et al. “Real-time millimeter-wave MIMO channel sounder for dynamic
directional measurements”. In: IEEE Transactions on Vehicular Technology 68.9 (2019),
pp. 8775–8789.
[17] Celalettin Umit Bas, Rui Wang, Seun Sangodoyin, Sooyoung Hur, Kuyeon Whang, Jeongho Park,
Jianzhong Zhang, and Andreas F Molisch. “28 GHz microcell measurement campaign for
residential environment”. In: GLOBECOM 2017-2017 IEEE Global Communications Conference.
IEEE. 2017, pp. 1–6.
[18] Laura Bernado, Thomas Zemen, Fredrik Tufvesson, Andreas F Molisch, and
Christoph F Mecklenbräuker. “Delay and Doppler spreads of nonstationary vehicular channels
for safety-relevant scenarios”. In: IEEE Transactions on Vehicular Technology 63.1 (2013), pp. 82–93.
[19] Laura Bernadó et al. “The (in-) validity of the WSSUS assumption in vehicular radio channels”. In:
2012 IEEE 23rd International Symposium on Personal, Indoor and Mobile Radio
Communications-(PIMRC). IEEE. 2012, pp. 1757–1762.
[20] Vinnu Bhardwaj and Chandra R Murthy. “On optimal routing and power allocation for D2D
communications”. In: 2015 IEEE International Conference on Acoustics, Speech and Signal Processing
(ICASSP). IEEE. 2015, pp. 3063–3067.
[21] Mate Boban, Diego Dupleich, Naveed Iqbal, Jian Luo, Christian Schneider, Robert Müller,
Ziming Yu, David Steer, Tommi Jämsä, Jian Li, et al. “Multi-band vehicle-to-vehicle channel
characterization in the presence of vehicle blockage”. In: IEEE access 7 (2019), pp. 9724–9735.
[22] Collin Brady and Sumit Roy. “A robust algorithm for LTE D2D (sidelink) discovery for public
safety communications”. In: IEEE Systems Journal (2021).
[23] Gabriel Brown and Heavy Reading. “Exploring the potential of mmWave for 5G mobile access”.
In: Qualcomm White Paper (2016).
[24] Robert JC Bultitude, Yvo LC de Jong, JA Pugh, Sana Salous, and Khawar Khokhar. “Measurement
and modelling of emergency vehicle-to-indoor 4.9 GHz radio channels and prediction of IEEE
802.16 performance for public safety applications”. In: 2007 IEEE 65th Vehicular Technology
Conference-VTC2007-Spring. IEEE. 2007, pp. 397–401.
122

[25] Xuesong Cai et al. “A switched array sounder for dynamic millimeter-wave channel
characterization: Design, implementation and measurements”. In: IEEE Transactions on Antennas
and Propagation (2023). early access.
[26] Derek Caudill et al. “Omnidirectional Channel Sounder With Phased-ArrayAntennas for 5G
Mobile Communications”. In: IEEE Transactions on Microwave Theory and Techniques 67.7 (2019),
pp. 2936–2945.
[27] Derek Caudill et al. “Real-Time mmWave Channel Sounding Through Switched Beamforming
With 3-D Dual-Polarized Phased-Array Antennas”. In: IEEE Transactions on Microwave Theory
and Techniques 69.11 (2021), pp. 5021–5032. doi: 10.1109/TMTT.2021.3104278.
[28] Jianfeng Chen, Yan Wen, and Yuhao Zhang. “Propagation Channel Characterization of an
Indoor-to-Outdoor Scenario at 868 MHz”. In: 2019 IEEE International Symposium on Antennas and
Propagation and USNC-URSI Radio Science Meeting (APSURSI). IEEE. 2019, pp. 1270–1271.
[29] Aditya Chopra, Andrew Thornburg, Ojas Kanhere, Saeed S Ghassemzadeh, Milap Majmundar,
and Theodore S Rappaport. “A real-time millimeter wave V2V channel sounder”. In: 2022 IEEE
Wireless Communications and Networking Conference (WCNC). IEEE. 2022, pp. 2607–2612.
[30] Aditya Chopra, Andrew Thornburg, Ojas Kanhere, Abbas Termos, Saeed S Ghassemzadeh, and
Theodore S Rappaport. “Real-time millimeter wave omnidirectional channel sounder using
phased array antennas”. In: GLOBECOM 2020-2020 IEEE Global Communications Conference. IEEE.
2020, pp. 1–7.
[31] Brecht De Beelde et al. “V-band channel modeling, throughput measurements, and coverage
prediction for indoor residential environments”. In: Electronics 11.4 (2022), p. 659.
[32] Albert Demba and Dietmar PF Möller. “Vehicle-to-vehicle communication technology”. In: 2018
IEEE international conference on electro/information technology (EIT). IEEE. 2018, pp. 0459–0464.
[33] Tobias Doeker et al. “Channel measurements and modeling for low terahertz communications in
an aircraft cabin”. In: IEEE Transactions on Antennas and Propagation 70.11 (2022),
pp. 10903–10916.
[34] Jinfeng Du et al. “Directional Measurements in Urban Street Canyons From Macro Rooftop Sites
at 28 GHz for 90% Outdoor Coverage”. In: IEEE Transactions on Antennas and Propagation 69.6
(2021), pp. 3459–3469. doi: 10.1109/TAP.2020.3044398.
[35] Diego Dupleich et al. “Multi-Band Propagation and Radio Channel Characterization in Street
Canyon Scenarios for 5G and Beyond”. In: IEEE Access 7 (2019), pp. 160385–160396.
[36] Fatih Erden et al. “28 GHz mmWave Channel Measurements and Modeling in a Library
Environment”. In: 2020 IEEE Radio and Wireless Symposium (RWS). 2020, pp. 52–55. doi:
10.1109/RWS45077.2020.9050106.
[37] Carlos A Fernandes et al. “Review of 20 Years of Research on Microwave and Millimeter-wave
Lenses at" Instituto de Telecomunicaço es"”. In: IEEE Antennas and Propagation Magazine 57.1
(2015), pp. 249–268.
123

[38] Jeffrey A Fessler and Alfred O Hero. “Space-alternating generalized expectation-maximization
algorithm”. In: IEEE Transactions on signal processing 42.10 (1994), pp. 2664–2677.
[39] Bernard H Fleury. “First-and second-order characterization of direction dispersion and space
selectivity in the radio channel”. In: IEEE Transactions on Information Theory 46.6 (2000),
pp. 2027–2044.
[40] Mathias Friese. “Multitone signals with low crest factor”. In: IEEE Transactions on
Communications 45.10 (1997), pp. 1338–1344.
[41] Anirban Ghosh, Aniruddha Chandra, Tomas Mikulasek, Ales Prokes, Jaroslaw Wojtun,
Jan M Kelner, and Cezary Ziolkowski. “Vehicle-to-Vehicle Path Loss Modeling at Millimeter-Wave
Band for Crossing Cars”. In: IEEE Antennas and Wireless Propagation Letters 22.9 (2023),
pp. 2125–2129.
[42] Jorge Gomez-Ponce et al. “Impact of noisy measurements with Fourier-based evaluation on
condensed channel parameters”. In: IEEE Transactions on Wireless Communications (2023).
[43] Herbert Groll et al. “Scanning Aperture Antennas with Spherical Shells”. In: 2019 International
Conference on Electromagnetics in Advanced Applications (ICEAA). IEEE. 2019, pp. 1220–1223.
[44] Qiongyang Guo et al. “Measurement of millimeter-wave 3D MIMO channel in large indoor
environment”. In: 2021 15th European Conference on Antennas and Propagation (EuCAP). IEEE.
2021, pp. 1–5.
[45] Carl Gustafson et al. “On mm-wave multipath clustering and channel modeling”. In: IEEE
Transactions on Antennas and Propagation 62.3 (2013), pp. 1445–1455.
[46] S Hamid, A Al-Dweik, M Mirahmadi, K Mubarak, and A Shami. “Indoor-to-outdoor channel
characterization for modeling and prediction of interference in next generation wireless
networks”. In: 2015 9th European Conference on Antennas and Propagation (EuCAP). IEEE. 2015,
pp. 1–5.
[47] Hussein Hammoud et al. “A Novel Low-Cost Channel Sounder for Double-Directionally Resolved
Measurements in the MmWave band”. In: IEEE International Conference on Communications.
(2024), under review.
[48] Hussein Hammoud et al. “A Novel Low-Cost Channel Sounder for Double-Directionally Resolved
Measurements in the MmWave band”. In: IEEE T. Wireless Comm. (2024), under review.
[49] Katsuyuki Haneda et al. “A statistical spatio-temporal radio channel model for large indoor
environments at 60 and 70 GHz”. In: IEEE Transactions on Antennas and Propagation 63.6 (2015),
pp. 2694–2704.
[50] Ruisi He, Christian Schneider, Bo Ai, Gongpu Wang, Zhangdui Zhong, Diego A Dupleich,
Reiner S Thomae, Mate Boban, Jian Luo, and Yunyong Zhang. “Propagation channels of 5G
millimeter-wave vehicle-to-vehicle communications: Recent advances and future challenges”. In:
IEEE vehicular technology magazine 15.1 (2019), pp. 16–26.
124

[51] Johannes Hejselbaek et al. “Channel sounding system for mm-wave bands and characterization of
indoor propagation at 28 GHz”. In: International journal of wireless information networks 24.3
(2017), pp. 204–216.
[52] Joseph Hoellinger, Gloria Makhoul, Raffaele D’Errico, and Thierry Marsault. “V2V Dynamic
Channel Characterization in 5G mmWave Band”. In: 2022 International Symposium on Antennas
and Propagation (ISAP). IEEE. 2022, pp. 525–526.
[53] Markus Hofer, David Lóschenbrand, Stefan Zelenbaba, Anja Dakić, Benjamin Rainer, and
Thomas Zemen. “Wireless 3ghz and 30 ghz vehicle-to-vehicle measurements in an urban street
scenario”. In: 2022 IEEE 96th Vehicular Technology Conference (VTC2022-Fall). IEEE. 2022, pp. 1–5.
[54] JA Högbom. “Aperture synthesis with a non-regular distribution of interferometer baselines”. In:
Astronomy and Astrophysics Supplement, Vol. 15, p. 417 15 (1974), p. 417.
[55] Jie Huang et al. “Multi-frequency mmWave massive MIMO channel measurements and
characterization for 5G wireless communication systems”. In: IEEE Journal on Selected Areas in
Communications 35.7 (2017), pp. 1591–1605.
[56] Sooyoung Hur et al. “Synchronous channel sounder using horn antenna and indoor
measurements on 28 GHz”. In: 2014 IEEE International Black Sea Conference on Communications
and Networking (BlackSeaCom). IEEE. 2014, pp. 83–87.
[57] Furqan Jameel, Zara Hamid, Farhana Jabeen, Sherali Zeadally, and Muhammad Awais Javed. “A
survey of device-to-device communications: Research issues and challenges”. In: IEEE
Communications Surveys & Tutorials 20.3 (2018), pp. 2133–2168.
[58] Yufeng Jiang and Yongquan Zhang. “Measurement-based analysis of indoor-to-outdoor
propagation characteristics in the sub-1 GHz band”. In: Journal of Communications 15.3 (2020),
pp. 222–230.
[59] Ojas Kanhere, Aditya Chopra, Andrew Thornburg, Theodore S Rappaport, and
Saeed S Ghassemzadeh. “Performance impact analysis of beam switching in millimeter wave
vehicular communications”. In: 2021 IEEE 93rd Vehicular Technology Conference (VTC2021-Spring).
IEEE. 2021, pp. 1–7.
[60] Udit Narayana Kar and Debarshi Kumar Sanyal. “A critical review of 3GPP standardization of
device-to-device communication in cellular networks”. In: SN Computer Science 1.1 (2020),
pp. 1–18.
[61] Margot Karam, William Turney, Kevin L Baum, Philippe J Sartori, Laddie Malek, and
Isselmou Ould-Dellahy. “Outdoor-indoor propagation measurements and link performance in the
VHF/UHF bands”. In: 2008 IEEE 68th Vehicular Technology Conference. IEEE. 2008, pp. 1–5.
[62] Myung-Don Kim et al. “Directional delay spread characteristics based on indoor channel
measurements at 28 GHz”. In: 2015 IEEE 26th Annual International Symposium on Personal, Indoor,
and Mobile Radio Communications (PIMRC). IEEE. 2015, pp. 403–407.
125

[63] Martin Koch. “Terahertz communications: A 2020 vision”. In: Terahertz frequency detection and
identification of materials and objects. Springer, 2007, pp. 325–338.
[64] Ashutosh Kumar, Prabhu Kaul, and S. K. Aggarwal. “Measurement and Analysis of
Indoor-to-Outdoor Channel Characteristics in Sub-1 GHz Frequency Band for IoT Applications”.
In: IEEE Access 9 (2021), pp. 27966–27977.
[65] Dave Laurenson and Peter Grant. “A review of radio channel sounding techniques”. In: 2006 14th
European Signal Processing Conference. IEEE. 2006, pp. 1–5.
[66] Juyul Lee et al. “Omnidirectional Millimeter-Wave Propagation Characteristics of Corridor
Environments based on Measurements at 28, 38, 71 and 82 GHz”. In: 2022 16th European
Conference on Antennas and Propagation (EuCAP). 2022, pp. 1–5. doi:
10.23919/EuCAP53622.2022.9769436.
[67] Yixuan Li, Guoyue Zhang, Yan Liu, Li Qian, and Qinghua Chen. “Analysis of the
Indoor-to-Outdoor Propagation Channel at 700 MHz for IoT Applications”. In: Sensors 20.23
(2020), p. 6948.
[68] Qifan Liu, Jingjing Wang, Yuan Fang, Zhongyuan Li, Mingming Li, and Xiaohua Li. “Wideband
indoor-to-outdoor channel measurement and modeling for the IoT application”. In: IEEE Internet
of Things Journal 8.10 (2021), pp. 8512–8522.
[69] G Lovnes et al. “Channel sounding measurements at 59 GHz in city streets”. In: 5th IEEE
International Symposium on Personal, Indoor and Mobile Radio Communications, Wireless
Networks-Catching the Mobile Future. Vol. 2. IEEE. 1994, pp. 496–500.
[70] Lu Lu, Geoffrey Ye Li, A Lee Swindlehurst, Alexei Ashikhmin, and Rui Zhang. “An overview of
massive MIMO: Benefits and challenges”. In: IEEE journal of selected topics in signal processing 8.5
(2014), pp. 742–758.
[71] Yejian Lyu, Pekka Kyösti, and Wei Fan. “Sub-terahertz channel sounder: Review and future
challenges”. In: China Communications 20.6 (2023), pp. 26–48.
[72] George R. MacCartney et al. “A Flexible Millimeter-Wave Channel Sounder With Absolute
Timing”. In: IEEE Journal on Selected Areas in Communications 35.6 (2017), pp. 1402–1418. doi:
10.1109/JSAC.2017.2687838.
[73] Alexander Maltsev et al. “Experimental investigations of 60 GHz WLAN systems in office
environment”. In: IEEE Journal on Selected Areas in Communications 27.8 (2009), pp. 1488–1499.
[74] Konstantinos Manolakis and Wen Xu. “Time synchronization for multi-link D2D/V2X
communication”. In: 2016 IEEE 84th Vehicular Technology Conference (VTC-Fall). IEEE. 2016,
pp. 1–6.
[75] Thomas L Marzetta. “Massive MIMO: an introduction”. In: Bell Labs Technical Journal 20 (2015),
pp. 11–22.
126

[76] Ali Masood, Muhammad Mahtab Alam, Yannick Le Moullec, Luca Reggiani, Davide Scazzoli,
Maurizio Magarini, and Rizwan Ahmad. “ProSe direct discovery: experimental characterization
and context-aware heuristic approach to extend public safety networks lifetime”. In: IEEE Access 9
(2021), pp. 130055–130071.
[77] David W Matolak. “Modeling the vehicle-to-vehicle propagation channel: A review”. In: Radio
Science 49.9 (2014), pp. 721–736.
[78] David W Matolak, Kate A Remley, Christopher Holloway, and Camillo Gentile.
“Outdoor-to-indoor channel dispersion and power-delay profile models for the 700-MHz and
4.9-GHz bands”. In: IEEE Antennas and Wireless Propagation Letters 15 (2015), pp. 441–443.
[79] Allan Wainaina Mbugua et al. “Millimeter wave multi-user performance evaluation based on
measured channels with virtual antenna array channel sounder”. In: Ieee Access 6 (2018),
pp. 12318–12326.
[80] Christoph F Mecklenbrauker, Andreas F Molisch, Johan Karedal, Fredrik Tufvesson,
Alexander Paier, Laura Bernadó, Thomas Zemen, Oliver Klemp, and Nicolai Czink. “Vehicular
channel characterization and its implications for wireless system design and performance”. In:
Proceedings of the IEEE 99.7 (2011), pp. 1189–1212.
[81] Millimeter Wave Products Inc. Waveguide Polarization Switch: Frequency 50 GHz - 75 GHz, Cross
Polarization -20 dB, 90 Degrees or RH/LH Polarization, 30 mA. Millimeter Wave Products Inc.
Accessed: 2023-11-08. 2023. url:
https://www.miwv.com/waveguide-polarization-switch-frequency-50-ghz-75-ghz/.
[82] A. F. Molisch et al. Spinning directional antenna in centimeter and millimeter wave bands. Patent
WO 2022155493 A1. Available at: https://patents.google.com/patent/WO2022155493A1/en. Sept.
2022.
[83] Andreas F Molisch et al. “Measurement of the capacity of MIMO systems in frequency-selective
channels”. In: IEEE VTS 53rd Vehicular Technology Conference, Spring 2001. Proceedings (Cat. No.
01CH37202). Vol. 1. IEEE. 2001, pp. 1–6.
[84] Andreas F Molisch. Wireless Communications: From Fundamentals to Beyond 5G. John Wiley &
Sons, 2023.
[85] Andreas F Molisch, Thomas Choi, Naveed Abbasi, Francois Rottenberg, and Jianzhong Zhang.
“Millimeter-Wave Channels”. In: Wiley 5G Ref: The Essential 5G Reference Online (2019), pp. 1–46.
[86] Andreas F Molisch, Fredrik Tufvesson, Johan Karedal, and Christoph F Mecklenbrauker. “A
survey on vehicle-to-vehicle propagation channels”. In: IEEE Wireless Communications 16.6
(2009), pp. 12–22.
[87] Sangmi Moon, Hun Choe, Myeonghun Chu, Cheolwoo You, Huaping Liu, Jeong-Ho Kim,
Jihyung Kim, Dae Jin Kim, and Intae Hwang. “DFT-based channel estimation scheme for sidelink
in D2D communication”. In: Wireless Personal Communications 97.1 (2017), pp. 1197–1215.
127

[88] Hung-Anh Nguyen et al. “Instantaneous direction of arrival measurements in mobile radio
channels using virtual circular array antennas”. In: 2016 IEEE Globecom Workshops (GC Wkshps).
IEEE. 2016, pp. 1–7.
[89] Sinh LH Nguyen et al. “Comparing radio propagation channels between 28 and 140 GHz bands in
a shopping mall”. In: (2018), pp. 1–5.
[90] Shuai Nie et al. “72 GHz millimeter wave indoor measurements for wireless and backhaul
communications”. In: 2013 IEEE 24th Annual International Symposium on Personal, Indoor, and
Mobile Radio Communications (PIMRC). IEEE. 2013, pp. 2429–2433.
[91] Jesper Odum Nielsen et al. “A channel sounder for massive MIMO and mmWave channels”. In:
IEEE Communications Magazine 56.12 (2018), pp. 67–73.
[92] Nikolaos Ntetsikas et al. “60 GHz Outdoor to Indoor (O2I) Propagation Measurements in a
University Campus”. In: 2022 IEEE 23rd International Workshop on Signal Processing Advances in
Wireless Communication (SPAWC). IEEE. 2022, pp. 1–5.
[93] Peter B Papazian, Camillo Gentile, Kate A Remley, Jelena Senic, and Nada Golmie. “A radio
channel sounder for mobile millimeter-wave communications: System implementation and
measurement assessment”. In: IEEE Transactions on Microwave Theory and Techniques 64.9 (2016),
pp. 2924–2932.
[94] Theodore S Rappaport et al. “38 GHz and 60 GHz angle-dependent propagation for cellular &
peer-to-peer wireless communications”. In: 2012 IEEE International Conference on
Communications (ICC). IEEE. 2012, pp. 4568–4573.
[95] Theodore S Rappaport, Shu Sun, Rimma Mayzus, Hang Zhao, Yaniv Azar, Kevin Wang,
George N Wong, Jocelyn K Schulz, Mathew Samimi, and Felix Gutierrez. “Millimeter wave mobile
communications for 5G cellular: It will work!” In: IEEE access 1 (2013), pp. 335–349.
[96] Olivier Renaudin. “Experimental channel characterization for vehicle-to-vehicle communication
systems.” PhD thesis. Catholic University of Louvain, Louvain-la-Neuve, Belgium, 2013.
[97] Olivier Renaudin, Veli-Matti Kolmonen, Pertti Vainikainen, and Claude Oestges. “Wideband
measurement-based modeling of inter-vehicle channels in the 5-GHz band”. In: IEEE Transactions
on Vehicular Technology 62.8 (2013), pp. 3531–3540.
[98] Dennis M Rose, Thomas Jansen, and Thomas Kürner. “Indoor to outdoor propagation—Measuring
and modeling of femto cells in LTE networks at 800 and 2600 MHz”. In: 2011 IEEE GLOBECOM
Workshops (GC Wkshps). IEEE. 2011, pp. 203–207.
[99] Shahid Saeed, Rizwan Ali, Khurram Shahzad, and Muhammad Hussain. “Measurement-based
analysis of indoor-to-outdoor radio channel in sub-1 GHz band for smart grid communication
systems”. In: IET Communications 15.5 (2021), pp. 1–9.
[100] Manuel Garcıa Sánchez, Mónica Portela Táboas, and Edgar Lemos Cid. “Millimeter wave radio
channel characterization for 5G vehicle-to-vehicle communications”. In: Measurement 95 (2017),
pp. 223–229.
128

[101] Louis L Scharf. “Statistical signal processing: applications to beamforming, detection, and
estimation”. In: Medical Imaging 2004: Ultrasonic Imaging and Signal Processing. Vol. 5373. SPIE.
2004, pp. 1–17.
[102] Ralph Schmidt. “Multiple emitter location and signal parameter estimation”. In: IEEE Transactions
on Antennas and Propagation 34.3 (1986), pp. 1–5.
[103] Robert Schneider, Dirk Didascalou, and Werner Wiesbeck. “Impact of road surfaces on
millimeter-wave propagation”. In: IEEE Transactions on Vehicular Technology 49.4 (2000),
pp. 1314–1320.
[104] Mansoor Shafi, Andreas F Molisch, Peter J Smith, Thomas Haustein, Peiying Zhu,
Prasan De Silva, Fredrik Tufvesson, Anass Benjebbour, and Gerhard Wunder. “5G: A tutorial
overview of standards, trials, challenges, deployment, and practice”. In: IEEE journal on selected
areas in communications 35.6 (2017), pp. 1201–1221.
[105] Sapna Shukla, Varsha Khare, Shubhanshi Garg, and Paramanand Sharma. “Comparative Study of
1G, 2G, 3G and 4G”. In: J. Eng. Comput. Appl. Sci 2.4 (2013), pp. 55–63.
[106] Rohit Singh, William Lehr, Douglas Sicker, and Kazi Mohammed Saidul Huq. “Beyond 5G: the
role of THz spectrum”. In: Tprc47: The 47th research conference on communication, information and
internet policy. 2019.
[107] Christopher Slezak et al. “Empirical Effects of Dynamic Human-Body Blockage in 60 GHz
Communications”. In: IEEE Communications Magazine 56.12 (2018), pp. 60–66. doi:
10.1109/MCOM.2018.1800232.
[108] Peter FM Smulders et al. “Frequency-domain measurement of the millimeter wave indoor radio
channel”. In: IEEE Transactions on Instrumentation and Measurement 44.6 (1995), pp. 1017–1022.
[109] Ruoyu Sun et al. “Design and calibration of a double-directional 60 GHz channel sounder for
multipath component tracking”. In: 2017 11th European Conference on Antennas and Propagation
(EUCAP). 2017, pp. 3336–3340. doi: 10.23919/EuCAP.2017.7928270.
[110] Harsh Tataria et al. “27.5-29.5 GHz switched array sounder for dynamic channel characterization:
Design, implementation and measurements”. In: arXiv preprint arXiv:2105.10712 (2021).
[111] Cihan Tepedelenlioglu, Ali Abdi, and Georgios B Giannakis. “The Ricean K factor: estimation and
performance analysis”. In: IEEE Transactions on Wireless Communications 2.4 (2003), pp. 799–810.
[112] Third Generation Partnership Project. url: https://www.3gpp.org/.
[113] R.S. Thoma, D. Hampicke, A. Richter, G. Sommerkorn, A. Schneider, U. Trautwein, and
W. Wirnitzer. “Identification of time-variant directional mobile radio channels”. In: IEEE
Transactions on Instrumentation and Measurement 49.2 (2000), pp. 357–364. doi: 10.1109/19.843078.
[114] Cheng-Xiang Wang, Xiang Cheng, and David I Laurenson. “Vehicle-to-vehicle channel modeling
and measurements: recent advances and future challenges”. In: IEEE communications magazine
47.11 (2009), pp. 96–103.
129

[115] Hui Wang, Josè Rodriguez-Pineiro, Xuefeng Yin, Haowen Wang, and Ziming Yu. “Comparison
analysis of 2.4 GHz and mm-wave V2V channel modelling based on measurements”. In: 2019 13th
European Conference on Antennas and Propagation (EuCAP). IEEE. 2019, pp. 1–5.
[116] Hui Wang, Xuefeng Yin, Xuesong Cai, Haowen Wang, Ziming Yu, and Juyul Lee. “Fading
characterization of 73 GHz millimeter-wave V2V channel based on real measurements”. In:
Communication Technologies for Vehicles: 13th International Workshop,
Nets4Cars/Nets4Trains/Nets4Aircraft 2018, Madrid, Spain, May 17-18, 2018, Proceedings 13.
Springer. 2018, pp. 159–168.
[117] Hui Wang, Xuefeng Yin, José Rodrıguez-Piñeiro, Juyul Lee, and Myung-Don Kim. “Shadowing
and multipath-fading statistics at 2.4 GHz and 39 GHz in vehicle-to-vehicle scenarios”. In: 2020
14th European Conference on Antennas and Propagation (EuCAP). IEEE. 2020, pp. 1–5.
[118] Jian Wang, Jian Wang, and Richard Rouil. BLER performance evaluation of LTE device-to-device
communications. US Department of Commerce, National Institute of Standards and Technology,
2016.
[119] Rui Wang et al. “A real-time MIMO channel sounder for vehicle-to-vehicle propagation channel
at 5.9 GHz”. In: 2017 IEEE International Conference on Communications (ICC). IEEE. 2017.
[120] Rui Wang, Olivier Renaudin, C Umit Bas, Seun Sangodoyin, and Andreas F Molisch.
“High-resolution parameter estimation for time-varying double directional V2V channel”. In:
IEEE Transactions on Wireless Communications 16.11 (2017), pp. 7264–7275.
[121] Wei Wang, Thomas Jost, and Uwe-Carsten Fiebig. “A comparison of outdoor-to-indoor wideband
propagation at S-band and C-band for ranging”. In: IEEE Transactions on Vehicular Technology
64.10 (2014), pp. 4411–4421.
[122] T. Ward. Ad Hoc Networks: Current Status and Future Trends. States Academic Press, 2022.
[123] Zhu Wen et al. “mmWave MIMO channel sounding for 5G”. In: 1st International Conference on 5G
for Ubiquitous Connectivity. IEEE. 2014, pp. 192–197.
[124] Yunchou Xing et al. “Propagation Measurements and Path Loss Models for sub-THz in Urban
Microcells”. In: (2021).
[125] Hao Xu et al. “Measurements and models for 38-GHz point-to-multipoint radiowave
propagation”. In: IEEE Journal on Selected Areas in Communications 18.3 (2000), pp. 310–321.
[126] Atsushi Yamamoto, Koichi Ogawa, Tetsuo Horimatsu, Akihito Kato, and Masayuki Fujise.
“Path-loss prediction models for intervehicle communication at 60 GHz”. In: IEEE Transactions on
vehicular technology 57.1 (2008), pp. 65–78.
[127] William F Young, Christopher L Holloway, Galen Koepke, Dennis Camell, Yann Becquet, and
Kate A Remley. “Radio-wave propagation into large building structures—Part 1: CW signal
attenuation and variability”. In: IEEE Transactions on Antennas and Propagation 58.4 (2010),
pp. 1279–1289.
130

[128] Jianhua Zhang. “IMT-2020 Channel Model”. In: Wiley 5G Ref: The Essential 5G Reference Online
(2019), pp. 1–18.
[129] Peize Zhang et al. “Indoor small-scale spatiotemporal propagation characteristics at multiple
millimeter-wave bands”. In: IEEE Antennas and Wireless Propagation Letters 17.12 (2018),
pp. 2250–2254.
[130] Peize Zhang et al. “Millimeter-wave space-time propagation characteristics in urban macrocell
scenarios”. In: ICC 2019-2019 IEEE International Conference on Communications (ICC). IEEE. 2019,
pp. 1–6.
[131] Erich Zochmann et al. “Measured delay and Doppler profiles of overtaking vehicles at 60 GHz”.
In: 12th European Conference on Antennas and Propagation (EuCAP 2018). 2018, pp. 1–5. doi:
10.1049/cp.2018.0470.
[132] Erich Zöchmann, Markus Hofer, Martin Lerch, Jiri Blumenstein, Seun Sangodoyin, Herbert Groll,
Stefan Pratschner, Sebastian Caban, D LÖschenbr, Laura Bernado, et al. “Statistical evaluation of
delay and Doppler spread in 60 GHz vehicle-to-vehicle channels during overtaking”. In: 2018
IEEE-APS Topical Conference on Antennas and Propagation in Wireless Communications (APWC).
IEEE. 2018, pp. 1–4.
[133] Erich Zöchmann, Markus Hofer, Martin Lerch, Stefan Pratschner, Laura Bernadó,
Jiri Blumenstein, Sebastian Caban, Seun Sangodoyin, Herbert Groll, Thomas Zemen, et al.
“Position-specific statistics of 60 GHz vehicular channels during overtaking”. In: Ieee Access 7
(2019), pp. 14216–14232.
131

Abstract (if available)

Abstract To ensure optimal performance of any wireless system, a comprehensive understanding of the wireless propagation channel is essential. An accurate channel model enables communication systems to mitigate negative impacts or leverage positive attributes. Propagation channel measurements, or channel sounding, are the most precise method for obtaining the true characteristics of a given environment. This dissertation introduces an innovative channel sounder design for high-frequency communications (mm-wave/terahertz) and demonstrates its application in measuring and modeling double-directional dynamic vehicular channels. Given the unique propagation characteristics at these frequencies, it is expected that future millimeter-wave systems will primarily use beam-forming antenna arrays to counteract the higher path loss. Thus, understanding the angular spectrum and its temporal evolution is critical for the effective design of these systems. While this information can be gathered using full, switched, or phased antenna arrays, the cost and availability of such arrays and the necessary electronic components are often prohibitive. To address this, we developed and constructed a mechanical structure named ReRoMA (Redirecting Rotating Mirror Arrangement), capable of acquiring directionally-resolved measurements in dynamic environments. ReRoMA can capture high-angular-resolution snapshots of the environment in approximately one second, significantly faster than the traditional method of using rotating-horn antennas. Additionally, we present findings from a vehicle-to-vehicle measurement campaign using a ReRoMA-based channel sounder operating at 60 GHz. This investigation covers various aspects of the channel at these frequencies, including moving scatterers, blocking objects, foliage, and statistical evaluations of propagation parameters such as path loss, delay spread, angular spreads, power distribution across multi-path components, and stationarity region—all of which are crucial for wireless system design and testing at these frequencies.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Channel sounding for next-generation wireless communication systems

PDF

Real-time channel sounder designs for millimeter-wave and ultra-wideband communications

PDF

Signal processing for channel sounding: parameter estimation and calibration

PDF

Multidimensional characterization of propagation channels for next-generation wireless and localization systems

PDF

mmWave dynamic channel measurements for localization and communications

PDF

Design and analysis of reduced complexity transceivers for massive MIMO and UWB systems

PDF

Exploiting side information for link setup and maintenance in next generation wireless networks

PDF

Design and analysis of large scale antenna systems

PDF

Propagation channel characterization and interference mitigation strategies for ultrawideband systems

PDF

Communicating over outage-limited multiple-antenna and cooperative wireless channels

PDF

Hybrid beamforming for massive MIMO

PDF

Spatial modes for optical communications and spatiotemporal beams

PDF

High capacity optical and RF/mm-wave communications utilizing orbital angular momentum multiplexing

PDF

Using beams carrying orbital angular momentum for communications and remote sensing

PDF

Channel state information feedback, prediction and scheduling for the downlink of MIMO-OFDM wireless systems

PDF

Orbital angular momentum based spatially multiplexed optical and millimeter-wave communications

PDF

Enhancing collaboration on the edge: communication, scheduling and learning

PDF

Multichannel data collection for throughput maximization in wireless sensor networks

PDF

Improving spectrum efficiency of 802.11ax networks

PDF

Elements of next-generation wireless video systems: millimeter-wave and device-to-device algorithms

Asset Metadata

Creator Hammoud, Hussein (author)

Core Title Double-directional channel sounding for next generation wireless communications

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Degree Conferral Date 2024-08

Publication Date 08/02/2024

Defense Date 07/23/2024

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag Antennas,beamforming,channel measurements,channel modelling,D2D,MIMO,mmwave,V2V,wireless communications

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Molisch, Andreas F. (committee chair), Wang, Yue (committee member), Willner, Alan E. (committee member)

Creator Email hhammoud@usc.edu,sounamd1994@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113998T54

Unique identifier UC113998T54

Identifier etd-HammoudHus-13326.pdf (filename)

Legacy Identifier etd-HammoudHus-13326

Document Type Dissertation

Format theses (aat)

Rights Hammoud, Hussein

Internet Media Type application/pdf

Type texts

Source 20240805-usctheses-batch-1192 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu