Projet ROLL2RAIL: New dependable rolling stock for a more sustainable, intelligent and comfortable rail transport in Europe. Deliverable D2.2 - Characterisation of the Railway Environment for Radio Transmission Ronald Raulefs, Stephan Sand, Eneko Echeverria, Imanol Baz, Ion Ansoategui, Thomas Jost, Andreas Lehner, Stephan Pfletschinger, Paul Unterhuber, Marion Berbineau, et al.

To cite this version:

Ronald Raulefs, Stephan Sand, Eneko Echeverria, Imanol Baz, Ion Ansoategui, et al.. Projet ROLL2RAIL: New dependable rolling stock for a more sustainable, intelligent and comfortable rail transport in Europe. Deliverable D2.2 - Characterisation of the Railway Environment for Radio Trans- mission. [Research Report] IFSTTAR - Institut Français des Sciences et Technologies des Transports, de l’Aménagement et des Réseaux. 2016, 186p. ￿hal-01664165￿

HAL Id: hal-01664165 https://hal.archives-ouvertes.fr/hal-01664165 Submitted on 14 Dec 2017

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés.

Contract No. H2020 – 636032

NEW DEPENDABLE ROLLING STOCK FOR A MORE SUSTAINABLE, INTELLIGENT AND COMFORTABLE RAIL TRANSPORT IN EUROPE

D2.2 – Characterisation of the Railway Environment for Radio Transmission

Due date of deliverable: 30/09/2016

Actual submission date: 10/12/2016

Leader/Responsible of this Deliverable: Ronald Raulefs (DLR), Stephan Sand (DLR)

Reviewed: Y

Project funded from the European Union’s Horizon 2020 research and innovation programme

Dissemination Level

PU Public x

CO Confidential, restricted under conditions set out in Model Grant Agreement

CI Classified, information as referred to in Commission Decision 2001/844/EC

Start date of project: 01/05/2015 Duration: 30 months

R2R-T2.2-D-DZR-006-18 Page 1 of 186 10/12/2016

Contract No. H2020 – 636032

Document status

Revision Date Description

1 2015-09-11 First issue, with contributions from IK, outline of planned contributions from IFSTTAR and DLR

2 2016-02-10 Added contributions on Chapters 2, 3 and DLR contributions in Sections 4.1 and 4.4

3 2016-02-16 Added contributions from IFSTTAR in Sections 2.2.1, 2.2.3, 2.2.4 and 3.3.1

4 2016-03-03 Updated Report Contributors, edited Section 2.2.2 and References

5 2016-03-04 Updated Executive Summary and Introduction, Added contribution from IK in Section 4.2

6 2016-03-17 Added contributions from CAF Section 3.2, IK Section 4.2.7, shifted parts of IFSTTAR contribution to Section 4.3

7 2016-03-24 Edited Section 3.2 with updates from CAF, updated Abbreviations and Acronyms, revised Chapters 2, 3 and 4, added Conclusions

8 2016-05-24 Adaption of the document structure on the DDP and included sections about measurements in the field (section 5), and channel models (section 6).

9 2016-05-25 Adaption of Section 4.1 and measurement report of DLR-TI high speed measurements added in Section 5.1

10 2016-07-22 Adaption of Section 4.2

Added contribution from DLR on Chapter 6

11 2016-09-01 Added contributions from IK on Chapter 5 and 6 Adaption of Section 4.3 and added contributions from IFSTTAR on Chapter 5 DLR updated Section 6.1

12 2016-09-15 Added contributions from IFSTTAR on Chapter 5 and 6 General review of Chapter 4, 5 and 6

13 2016-09-30 Peris Eulalia: UNIFE First check

14 2016-10-20 Included review comments (Ronald)

15 2016-11-18 Final comments considered

16 2016-11-21 Removal of Section 4.4 about ITG-G5 measurements.

17 2016-12-10 Editorial modifications to TMT (TALGO) Comments during 2nd round of TMT Review. Version generated by WP2 Leader.

R2R-T2.2-D-DZR-006-18 Page 2 of 186 10/12/2016

Contract No. H2020 – 636032

REPORT CONTRIBUTORS

Name Company Details of Contribution

Eneko Echeverria, CAF Section 3.2 Sources of EM Noise Imanol Baz, Ion Ansoategui

Thomas Jost, DLR Editor of deliverable, Executive Summary, Andreas Lehner, Chapters 1, 2, 3, 4.1, 4.4, 5.1, 6.1 and Stephan Pfletschinger Conclusions (editor till March 2016), Ronald Raulefs (editor up from Mai 2016), Stephan Sand (editor), Paul Unterhuber

Marion Berbineau, IFSTTAR Sections 2.2, 3.3.1, 4.3, 5.3 and 6.3 Kun Yang

Iñaki Val, IK4-IKERLAN Sections 3.2, 3.3.3, 3.4.2, 4.2, 5.2 and 6.2 Aitor Arriola

Pierre Emmanuel Reb ALSTOM Host of 60GHz Measurements

Juan Moreno, MdM Hosts of Metro Measurements Juan Pablo García

Francesco Romano, Maurizio TrenItalia Hosts of High-Speed Measurements D'Atri, Fabrizio Tavano

R2R-T2.2-D-DZR-006-18 Page 3 of 186 10/12/2016

Contract No. H2020 – 636032

EXECUTIVE SUMMARY The Roll2Rail project aims to develop key technologies and to remove already identified blocking points for radical innovation in the field of railway vehicles, as part of a longer term strategy to revolutionize the rolling stock for the future. The results will contribute to the increase of the operational reliability and to the reduction of the life cycle costs. This project started in May 2015 and it is supported by the Horizon 2020 program of the European Commission. Roll2Rail is one of the lighthouse projects of Shift2Rail and will contribute to Innovation Program 1. At the end of the project all the results will be further developed, leading to demonstration in real vehicles or relevant environments in Shift2Rail.

Going into detail, this Roll2Rail project covers different rolling stock topics such as Traction (WP1), TCMS (WP2), Car-Body-Shell (WP3), Running-Gear (WP4), Brakes (WP5), Vehicle Interiors (WP6) and transversal activities such as Noise (WP7) and Energy Management (WP8).

In that context, WP2 work package’s concrete goal is to make research on technologies and architectures to allow new generation of train communication systems based on Wireless Transmission for Train Control and Monitoring System (TCMS), functions and Infotainment, CCTV applications, thus reducing or even completely eliminating, on board communication cables and simplifying the train coupling procedure.

This deliverable describes the motivation and methods for the characterization of the railway environment for radio transmission. After a thorough review of existing measurements and channel models in the railway domain, the gaps for a detailed characterization of the radio environment are identified and based on this analysis, a common methodology is defined. Since channel measurements require a considerable technical effort and experience, the methodology for channel sounding is described in detail. For the preparation and the execution of the measurement campaigns, a strong effort both in terms of equipment as well as trained staff is required. For this reason, the measurement campaign planning is outlined in detail in Chapter 4. Chapter 5 describes the actual three different measurement campaigns carried out. Two measurement campaigns were carried out in the field, i.e. the high-speed line and the urban/metro measurements. These included measurements with one or two trains in the operational environment. The third measurement campaign conducted stationary measurements in a depot for the 60 GHz measurements and a regional train. Finally, Chapter 6 describes the post processing and analysis of the channel measurements. For the high speed line measurements, the large scale and small scale fading parameters are estimated in Section 6.1 for the inter-vehicle channel measurements and the parameters for a tap-delay line channel model are provided. Similarly, Section 6.2 provides the large scale and small scale fading parameters of a channel model for the Metro for inter-consist, intra-consist, and train-to-train scenarios as well as parameters for a tap-delay line channel model for the first two scenarios. Finally, Section 6.3 analyses and estimates parameters for a regional train at 60 GHz carrier frequency for various measurement scenarios such as intra-vehicle and outdoor-to-roof.

To conclude, the results of Chapter 6 provide the basis for channel models which are then used in Task 2.7 to simulate and evaluate selected wireless technologies and in channel emulations in Task 2.8 to validate suitable radio technologies for wireless TCMS.

R2R-T2.2-D-DZR-006-18 Page 4 of 186 10/12/2016

Contract No. H2020 – 636032

ABBREVIATIONS AND ACRONYMS

4G 4th Generation mobile communication

5G 5th Generation mobile communication

AC Alternating Current

AP Access Point

C2C Consist-to-Consist

C2I Car-to-Infrastructure

CBTC Communication-Based Train Control

CCTV Closed-Circuit TeleVision

CIR Channel Impulse Response

CR Cognitive Radio

D2D Device-to-Device

DC Direct Current

DCS Digital Cellular System

DRU Digital Receiving Unit

ECN Ethernet Consist Network

EIRP Equivalent Isotropically Radiated Power

EM ElectroMagnetic

EMC ElectroMagnetic Compatibility

EMI ElectroMagnetic Interference

ERTMS European Rail Traffic Management System

ETB Ethernet Train Backbone

ETCS European Train Control System

FDTD Finite Difference Time Domain

FSPL Free Space Path Loss

GPRS General Packet Radio Service

GPS Global Positioning System

R2R-T2.2-D-DZR-006-18 Page 5 of 186 10/12/2016

Contract No. H2020 – 636032

GSCM Geometry-based stochastic channel model

GSM-R Global System for Mobile Communications – Railway

HSR High-Speed Railway

IEEE Institute of Electrical and Electronic Engineers

IMU Inertial Measurement Unit

IP Internet Protocol

ISM Industrial, Scientific, Medical

ITS Intelligent Transportation Systems

ITS-G5 European profile standard for the physical and medium access control layer of Intelligent Transport Systems operating in the 5 GHz frequency band

LOS Line-Of-Sight

LNA Low noise amplifier

LTE Long Term Evolution

MCSSS Multi-Carrier Spread Spectrum Signal

MIMO Multiple Input Multiple Output

MoM Method of Moments

MPC Multi path component

MUX Multiplexer

MVB Multifunction Vehicle Bus

OFDM Orthogonal Frequency-Division Multiplexing

PDF Probability Density Function

PDP Power Delay Profile

PTS Portable Transmitting Station

QP Quasi-Peak

RCAS Railway Collision Avoidance System

RFID Radio Frequency Identification

RFT Radio Frequency Tuner

R2R-T2.2-D-DZR-006-18 Page 6 of 186 10/12/2016

Contract No. H2020 – 636032

RSL Received Signal Level

RTE Real time Ethernet

RX Receiver

SCM Stochastic Channel Models

SISO Single Input Single Output

T2G Train-to-Ground

T2T Train-to-train

TCMS Train Control and Monitoring System

TCN Train Communication Network

TETRA Terrestrial Trunked Radio

TWC Train to Wayside Communication

TX Transmitter

UHF Ultra high frequency

UMTS Universal Mobile Telecommunications System

UPS Uninterruptible power supply

UWB Ultra-Wide-Band

VETAG VEhicle TAGging system

VNA Vector Network Analyzer

VSA Vector Signal Analyser

Wi-Fi Wireless Fideltity

WiMAX Worldwide Interoperability for Microwave Access

WSSUS Wide-Sense Stationary Uncorrelated Scattering

WTB Wire Train Bus

R2R-T2.2-D-DZR-006-18 Page 7 of 186 10/12/2016

Contract No. H2020 – 636032

TABLE OF CONTENTS Report Contributors ...... 3 Executive Summary ...... 4 Abbreviations and Acronyms ...... 5 Table of Contents...... 8 List of Figures ...... 11 List of Tables ...... 16 1. Introduction ...... 18 2. State-of-the-Art on Measurements and Models in Railways ...... 19 2.1 Current and Future Railway Communication Systems ...... 19 2.1.1 Inside Train Communication ...... 19 2.1.2 Outside Train Communication ...... 19 2.1.3 Current Wireless Systems ...... 20 2.1.4 Possible Directions for Future Wireless Systems ...... 20 2.2 Existing Channel Models and Measurements ...... 22 2.2.1 Types of Channel Models ...... 22 2.2.2 Measurements inside Vehicle ...... 24 2.2.3 Measurements outside Vehicle ...... 26 2.2.4 Identified Gaps of Existing Measurements and Models ...... 27 3. Common Methodology Definition ...... 31 3.1 Requirements for Measurement Scenarios ...... 31 3.2 Sources of EM Noise...... 31 3.2.1 Regulations and Standards ...... 32 3.2.2 EMI Environmental Characterisation ...... 32 3.2.3 Internal EMI sources ...... 32 3.2.4 External EMI sources ...... 35 3.2.5 Summary and additional information ...... 35 3.2.6 Equipment Location ...... 38 3.3 Available Measurement Equipment ...... 39 3.3.1 IK4-IKERLAN has the following equipment for channel measurements: ...... 39 3.3.2 DLR equipment ...... 41 3.4 Measurement Experience Know-How in Railways ...... 42 3.4.1 IFSTTAR...... 42 3.4.2 DLR ...... 42 3.4.3 IK4-Ikerlan Experience ...... 43

R2R-T2.2-D-DZR-006-18 Page 8 of 186 10/12/2016

Contract No. H2020 – 636032

3.5 Post Processing, Parameter Estimation and Modelling Capabilities of Measurements ...... 46 3.5.1 Channel Sounding Methodology ...... 46 3.5.2 Post Processing ...... 48 4. Measurement Scenario Description ...... 55 4.1 Measurement Campaign 1: High-Speed Trains in Italy ...... 55 4.1.1 Overview of Environments ...... 55 4.1.2 Manoeuvres ...... 56 4.1.3 Scenarios...... 61 4.1.4 Schedule...... 62 4.1.5 Measurement Equipment ...... 63 4.1.6 Trajectory of Measurements ...... 68 4.1.7 Documentation ...... 69 4.1.8 Detailed Schedule for Measurements ...... 70 4.2 Measurement Campaign 2: Metro in Madrid ...... 73 4.2.1 Measurement Procedure ...... 73 4.2.2 Measurement Schedule ...... 75 4.3 Complementary Measurement Campaign: 60 GHz for T2T Communications ...... 81 4.3.1 Motivation for the 60 GHz channel measurement ...... 81 4.3.2 Available Measurement Equipment ...... 81 4.3.3 Measurement Scenario Description ...... 87 4.3.4 Intra-Vehicle Communication (Metro and HSR) ...... 88 4.3.5 Intra-Consist Communication (Metro and HSR) ...... 88 4.3.6 Inter-Consist Communication (Metro and HSR) ...... 89 5. Measurements in the field ...... 90 5.1 High speed line – AC (TRENITALIA) ...... 91 5.1.1 Night 1 ...... 91 5.1.2 Night 2 ...... 93 5.1.3 Night 3 ...... 93 5.1.4 Night 4 ...... 94 5.2 Urban / metro (METRO DE MADRID) ...... 96 5.2.1 Night 1 ...... 97 5.2.2 Night 2 ...... 99 5.2.3 Night 3 ...... 100 5.2.4 Night 4 ...... 101 5.2.5 Summary ...... 101 5.3 Regional Trains (Alstom) ...... 102

R2R-T2.2-D-DZR-006-18 Page 9 of 186 10/12/2016

Contract No. H2020 – 636032

6. Channel Models ...... 103 6.1 HSR Inter-Vehicle Channel Model ...... 104 6.1.1 Data Post Processing ...... 104 6.1.2 Channel Characteristics ...... 104 6.1.3 Tapped delay line model ...... 107 6.2 Metro Channel Model ...... 118 6.2.1 Data Post Processing ...... 118 6.2.2 Inter-consist Model ...... 118 6.2.3 Intra-consist Model (Continuous Train) ...... 133 6.2.4 Intra-consist Model (Non-Continuous Train) ...... 138 6.2.5 Train-to-Train Model ...... 141 6.3 Regional Train: 60 GHz Channel Model ...... 150 6.3.1 Intra-Vehicle Communication (Regional trains) ...... 150 6.3.2 Inter Car Scenario ...... 152 6.3.3 Penetration through the train window ...... 155 6.3.4 Outdoor to Driver’s room ...... 159 6.3.5 Outdoor to Driver (Centre) ...... 163 6.3.6 Outdoor to Outdoor (O2O) ...... 166 6.3.7 Train to Infrastructure...... 170 6.3.8 Outdoor to Roof ...... 173 7. Conclusions ...... 178 8. References ...... 180

R2R-T2.2-D-DZR-006-18 Page 10 of 186 10/12/2016

Contract No. H2020 – 636032

LIST OF FIGURES Figure 1: Future train wireless communications ...... 21 Figure 2: Diffraction phenomena around car: electromagnetic simulation model (left); simulated electric field (right) ...... 26 Figure 3: Simulated and measured power-loss results for 3-meter setup with vertical polarization at 2.45 GHz ...... 27 Figure 4: Different lateral cuts of tunnels ...... 28 Figure 5: Intra-vehicle (left) and inter-vehicle (right); note that the vertical gray lines depict the boundaries between rail vehicles...... 29 Figure 6: Continuous (left) and non-continuous (right) trains ...... 29 Figure 7: Inter-consist: (a) centre; (b) edge ...... 29 Figure 8: Train-to-train: inside (left), roof (right) ...... 30 Figure 9: Vehicle Example A (Tram) ...... 38 Figure 10: Vehicle Example B (Tram) ...... 38 Figure 11: Vehicle Example C (Metro Madrid) ...... 38 Figure 12: VNA-based measurement setup for static channels: VNA (left); UWB antennas (right) 39 Figure 13: VSA-based measurement setup: transmitter (left); receiver (right) ...... 40 Figure 14: RUSK Channel Sounder in railway environment ...... 40 Figure 15: MEDAV RUSK block diagram ...... 41 Figure 16: Broadband multicarrier spread spectrum signal (MCSSS) magnitude in time (left) and frequency domain (right) ...... 41 Figure 17: Antennas on wheel and bogie for channel measurements ...... 43 Figure 18: Base station antenna located inside the train ...... 44 Figure 19: Access Point locations ...... 45 Figure 20: Sensor node locations under the car ...... 45 Figure 21: Sensor node located on gearbox (S2) ...... 46 Figure 22: Sensor node locations inside the car ...... 46 Figure 23: Principle of channel sounding ...... 47 Figure 24: Waveform generation at transmitter side ...... 48 Figure 25: Radiation pattern of Mobile Mark MGRM-UMB-1C antenna at 2.45 GHz ...... 49 Figure 26: Frequency response of Mobile Mark MGRM-UMB-1C antenna from 2 to 3 GHz ...... 49 Figure 27: Channel power gain along the frequency and space ...... 50 Figure 28: Small scale fading extraction ...... 51 Figure 29: Small scale envelope distribution ...... 51 Figure 30: Averaged power delay profile ...... 52 Figure 31: Spatial autocovariance ...... 53 Figure 32: Temporal power variations ...... 53

R2R-T2.2-D-DZR-006-18 Page 11 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 33: Doppler spectrum ...... 54 Figure 34: Manoeuvres ...... 57 Figure 35: Flow traffic of two trains ...... 57 Figure 36: Timeline for overtake manoeuvre ...... 58 Figure 37: Opposing traffic of two trains ...... 59 Figure 38: Time line for counter manoeuvre ...... 59 Figure 39: One train pass a second train ...... 60 Figure 40: Consist-to-Consist (Bombardier) ...... 61 Figure 41: Front to End measurement ...... 61 Figure 42: Receiver (left) and transmitter (right) of the channel sounding equipment ...... 63 Figure 43: Omni-directional (red marked) antenna on train 28, coach 3 ...... 64 Figure 44: Installation of the directional antenna ...... 65 Figure 45: Tetra antenna on locomotive ...... 66 Figure 46: Vehicle-to-Vehicle setup ...... 67 Figure 47: Train-to-Train with omni-directional antennas ...... 67 Figure 48: Train-to-Train with directional antennas ...... 67 Figure 49: Track map ...... 68 Figure 50: Frecciarossa ETR 500 ...... 69 Figure 51 : Channel Sounder ...... 73 Figure 52: Mobile Mark MGRM-WHF Wideband Monopole Antenna...... 74 Figure 53: Monopole Array antenna ...... 74 Figure 54: Inter-Consist measurements ...... 77 Figure 55: Train-to-Train Measurements ...... 78 Figure 56: Intra-Consist measurements on Continuous train ...... 79 Figure 57: Intra-Consist measurements on Non-Continuous train ...... 80 Figure 58: Photos of the mmw measurement setup – VµBIQ ...... 81 Figure 59: mmw measurement setup – VµBIQ ...... 82 Figure 60: Scheme of the equipment ...... 82 Figure 61: Characteristics of the mmw antenna ...... 82 Figure 62: Real and imaginary part of the chirp signal ...... 84 Figure 63: Back-to-Back measurement setup without RF module ...... 85 Figure 64: Channel Impulse Response from B2B measurement without RF module ...... 86 Figure 65: (a) waveguide attenuator (b) attenuation loss measurement ...... 86 Figure 66: mmw measurement setup ...... 87 Figure 67: Intra-vehicle links ...... 88 Figure 68: Intra-Consist links ...... 88

R2R-T2.2-D-DZR-006-18 Page 12 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 69: Inter-Consist links ...... 89 Figure 70: Measurement setup for inter-vehicle measurements ...... 91 Figure 71: Measurement setup for inside ITS-G5 measurements ...... 92 Figure 72: Intra vehicle/consist measurement ...... 92 Figure 73: Measurement setup for outside ITS-G5 measurements ...... 93 Figure 74: Measurement setup for T2T omni-directional measurements ...... 94 Figure 75: Measurement setup for T2T directional measurements ...... 95 Figure 76: Measured metro lines and stations ...... 96 Figure 77: Omnidirectional antennas for inter-consist measurements ...... 97 Figure 78: Inter-consist measurement setup ...... 97 Figure 79: Intra-consist measurement setup ...... 98 Figure 80: Omnidirectional antennas for intra-consist measurements ...... 98 Figure 81: Antenna array for intra-vehicle measurements ...... 99 Figure 82: Path loss measurement ...... 99 Figure 83: Train-to-train measurement setup ...... 100 Figure 84: Glass characterization setup ...... 100 Figure 85: Monopole antennas on a Series 5000 train ...... 101 Figure 86: View of the train for measurements in Valenciennes ...... 102 Figure 87 : The time-variant impulse response and its frequency-domain equivalents [89] ...... 104 Figure 88: Normalized PDP for 1. Night: scenario 1 (slow), scenario 2 (fast) ...... 106 Figure 89: PDF of Normalized PDP for 1. Night ...... 106 Figure 90 : Tapped delay line model ...... 107 Figure 91: 3 m measurement, distribution fitting for different taps ...... 114 Figure 92: 29 m measurement, distribution fitting for different taps ...... 116 Figure 93: Narrowband Results for Inter-Consist link in Tunnel (Line 10) ...... 118 Figure 94: Glass characterization: losses (blue graph); phase deviation (yellow graph) ...... 119 Figure 95: Channel taps vs distance for Inter-Consist link in Tunnel (Line 10) ...... 120 Figure 96: Probability Density Function of channel taps for Inter-Consist link in Tunnel (Line 10) 120 Figure 97: Channel taps vs distance for Inter-Consist link in Tunnel (Line 11) ...... 121 Figure 98: Probability Density Function of channel taps for Inter-Consist link in Tunnel (Line 11) 121 Figure 99: Narrowband Results for Inter-Consist link in Station (Line 10) ...... 122 Figure 100: Channel taps vs distance for Inter-Consist link in Station (Line 10) ...... 122 Figure 101: Probability Density Function of channel taps for Inter-Consist link in Station (Line 10) ...... 123 Figure 102: Channel taps vs distance for Inter-Consist link in Station (Line 11) ...... 123 Figure 103: Probability Density Function of channel taps for Inter-Consist link in Station (Line 11) ...... 124

R2R-T2.2-D-DZR-006-18 Page 13 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 104: Channel taps vs distance for Inter-Consist link in Station (Line 6) ...... 124 Figure 105: Probability Density Function of channel taps for Inter-Consist link in Station (Line 6) 125 Figure 106: Narrowband Results for Inter-Consist link in Open Air (Line 10) ...... 125 Figure 107: Channel taps vs distance for Inter-Consist link in Open Air (Line 10) ...... 126 Figure 108: Probability Density Function of channel taps for Inter-Consist link in Open Air (Line 10) ...... 126 Figure 109: Probability Density Function with External Antennas ...... 127 Figure 110: Probability Density Function with External and Internal Antenna ...... 128 Figure 111: Power Delay Profile (PDP) for Inter-Consist link in Line 10 ...... 129 Figure 112: Power Delay Profile (PDP) for Inter-Consist link in Line 11 ...... 130 Figure 113: Power Delay Profile (PDP) for Inter-Consist link in Line 6 ...... 132 Figure 114: PDP vs angle of arrival for Intra-Consist link in Tunnel (Line 11) ...... 133 Figure 115: PDP vs angle of arrival for Intra-Consist link in Station (Line 11) ...... 134 Figure 116: Path loss model for the Intra-Consist link (continuous train) ...... 134 Figure 117: Probability Density Function of Intra-Consist link in station (continuous train) ...... 135 Figure 118: Impact of people movement on Intra-Consist link (continuous train) ...... 136 Figure 119: Probability Density Function of Intra-Consist link (continuous train) with people movement ...... 136 Figure 120: PDP vs angle of arrival for Intra-Consist link in Open Air (Line 11) ...... 137 Figure 121: Path loss model for the Intra-Consist link in a non-continuous train ...... 138 Figure 122: Probability Density Function of Intra-Consist link in station (non-continuous train) ... 139 Figure 123: Channel taps vs distance for Intra-Consist link (non-continuous train) in Station (Line 6) ...... 140 Figure 124: Probability Density Function of channel taps for Intra-Consist link (non-continuous train) in Station (Line 6) ...... 140 Figure 125: Averaged Narrowband Results for Train-to-Train link in tunnel...... 142 Figure 126: Path loss model for Train-to-Train link in tunnel with outdoor antennas ...... 142 Figure 127: Averaged Narrowband Results for Train-to-Train link in station ...... 143 Figure 128: Path loss model for Train-to-Train link in station with outdoor antennas ...... 144 Figure 129: Averaged Narrowband Results for Train-to-Train link in open air ...... 145 Figure 130: Path loss model for Train-to-Train link in open air with outdoor antennas ...... 145 Figure 131: Comparison of averaged Narrowband Results for Train-to-Train links: outdoor receiving antenna (upper); indoor receiving antenna (bottom) ...... 146 Figure 132: Probability Density Function for Train-to-Train links with outdoor receiving antenna 147 Figure 133: Probability Density Function for Train-to-Train links with indoor receiving antenna... 148 Figure 134: Comparison of path loss models for Train-to-Train links with outdoor antennas ...... 149 Figure 135: Photos of the measurements (a) to (d) ...... 151 Figure 136: Angular power distribution for the two positions of the receiver ...... 151

R2R-T2.2-D-DZR-006-18 Page 14 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 137: Angular power distribution for the two positons of the receiver ...... 152 Figure 138: Inter car configuration ...... 153 Figure 139: Angular power distribution at P1 with the door open and close ...... 153 Figure 140: Angular power distribution at P2 with the door open and close ...... 153 Figure 141: Angular power distribution at P3 with the door open and close ...... 154 Figure 142: Sum-up of the angular power distribution for the 3 positions ...... 154 Figure 143 (a) and (b): illustration of the measurements configuration for penetration through the window ...... 156 Figure 144: Angular power distribution for the five positions of Figure 143 (a) ...... 157 Figure 145: Received signal level in dBm for each angle at each of the 5 positions ...... 158 Figure 146: Evolution of received signal level for Tx-Rx distance and for each angle ...... 159 Figure 147: (a) scheme of the measurements- (b) photography of the scenario ...... 160 Figure 148: Angular power distribution for (-22°, 0°, 22°) for each of the 7 positions outside the cabin as indicated on scheme Figure 147 (a) ...... 161 Figure 149: Received signal level for each position and each angle ...... 162 Figure 150: Received signal level versus Tx-Rx distance for each angle (-22°, 0°, +22°) ...... 163 Figure 151: Angular power distribution for (-22°, 0°, 22°) for each of the 7 positions outside the cabin as indicated on scheme Figure 147 (a), on the centre ...... 164 Figure 152: Received signal level for each position and each angle (-22°, 0°, +22°) ...... 165 Figure 153: Received signal level versus distance for each angle (-22°, 0°, +22°) ...... 166 Figure 154: Outdoor-Outdoor measurements to simulate transmission between consist...... 167 Figure 155: Outdoor-Outdoor measurements – Angular power distribution at the 7 positions ..... 168 Figure 156: Summary of received power level for the 7 positions and for [-22°, 0°, +22°]...... 168 Figure 157: Received power level versus the distance for [-22°, 0°, +22°] ...... 169 Figure 158: Transmitter on the roof of the train ...... 170 Figure 159: Angular power distribution at the 5 positions ...... 171 Figure 160: Received signal level in dBm versus angles in degree for the 5 positions ...... 172 Figure 161: Received power level versus the distance ...... 173 Figure 162: Illustration of the Outdoor-to-roof scenario ...... 173 Figure 163: Illustration of measurement scenario for Roof to Roof configuration ...... 174 Figure 164: Angular power distribution for the seven positions...... 175 Figure 165: Received signal in dBm for the 3 angles (-22°, 0°, +22°) and for the 7 positions ...... 176 Figure 166: Received signal level in dBm versus Tx-RX distance for each angle (-22°, 0°, +22°) 177

R2R-T2.2-D-DZR-006-18 Page 15 of 186 10/12/2016

Contract No. H2020 – 636032

LIST OF TABLES Table 1: Railway environment ...... 28 Table 2: European Standard EN50121-3-2 vs 2006 ...... 33 Table 3: European Standard EN50121-3-2 vs 2015 ...... 33 Table 4 Intentional EMI Sources ...... 36 Table 5: Overview of measurement equipment ...... 42 Table 6: Railway environment ...... 55 Table 7: Combination of possible environments (finally selected ones are marked red) ...... 56 Table 8: Combination of tracks and manoeuvre ...... 56 Table 9: Priorities for different scenarios ...... 61 Table 10: Time table for all four days ...... 70 Table 11: Basic data ...... 70 Table 12: Manoeuvres day 1 ...... 71 Table 13: Manoeuvres Day 2 ...... 71 Table 14: Manoeuvres Day 3 ...... 72 Table 15: Manoeuvres Day 4 ...... 72 Table 16: Features of Channel Sounder ...... 74 Table 17: Measurement plan ...... 76 Table 18: Inter-Consist measurement environments ...... 77 Table 19: Train-to-Train measurement environments ...... 78 Table 20: Intra-Consist measurement environments ...... 79 Table 21: Intra-Consist measurement environments ...... 80 Table 22: VµBIQ development system at 60 GHz with 2 GHz bandwidth ...... 81 Table 23: Characteristics of the recording system ...... 84 Table 24: Characteristics of the scope set up ...... 84 Table 25: Channel sounder's parameters ...... 87 Table 26: Summary of metro measurements ...... 101 Table 27: Path loss calculation results ...... 105 Table 28: Tap fitting parameters 3 m ...... 113 Table 29: Tap fitting parameters 29 m ...... 117 Table 30: Glass characterization results ...... 119 Table 31: Average attenuation in different environments...... 127 Table 32: Statistical parameters of external link ...... 127 Table 33: Statistical parameters for external-to-internal link ...... 128 Table 34: Multipath components for Inter-Consist link in Line 10 ...... 129 Table 35: RMS spread and Mean Delay for Inter-Consist link in Line 10 ...... 129

R2R-T2.2-D-DZR-006-18 Page 16 of 186 10/12/2016

Contract No. H2020 – 636032

Table 36: Statistical parameters of channel taps for Inter-Consist link in Line 10 ...... 130 Table 37: Multipath components for Inter-Consist link in Line 11 ...... 131 Table 38: RMS spread and Mean Delay for Inter-Consist link in Line 11 ...... 131 Table 39: Statistical parameters of channel taps for Inter-Consist link in Line 11 ...... 131 Table 40: Multipath components for Inter-Consist link in Line 6 ...... 132 Table 41: RMS spread and Mean Delay for Inter-Consist link in Line 6 ...... 132 Table 42: Statistical parameters of channel taps for Inter-Consist link in Line 6 ...... 133 Table 43: Statistical parameters of Intra-Consist link in station (continuous train) ...... 135 Table 44: Statistical parameters of Intra-Consist link (continuous train) with people movement .. 136 Table 45: RMS spread and Mean Delay for Intra-Consist link (continuous train) in Line 11 ...... 137 Table 46: Multipath components for Intra-Consist link (continuous train) in Line 11 ...... 137 Table 47: Statistical parameters of Intra-Consist link in station (non-continuous train) ...... 139 Table 48: RMS spread and Mean Delay for Intra-Consist link (non-continuous train) in Line 6.... 141 Table 49: Statistical parameters of channel taps for Intra-Consist link (non-continuous train) in Line 6 ...... 141 Table 50: Multipath components for Intra-Consist link (non-continuous train) in Line 6 ...... 141 Table 51: Statistical parameters for Train-to-Train links with outdoor receiving antenna ...... 147 Table 52: Statistical parameters for Train-to-Train links with indoor receiving antenna ...... 148 Table 53: Path loss model parameters ...... 149 Table 54: Characteristics of the measurements ...... 150 Table 55: Mean delay (ns) versus angle position ...... 152 Table 56: RMS delay (ns) versus angle position ...... 152 Table 57: Mean delay (ns) versus angle position ...... 155 Table 58: RMS delay (ns) versus angle position ...... 155 Table 59: Mean delay (ns) versus angle position ...... 158 Table 60: RMS delay (ns) versus angle position ...... 158 Table 61: Mean delay (ns) versus angle position ...... 162 Table 62: RMS delay (ns) versus angle position ...... 162 Table 63: Mean delay (ns) versus angle position ...... 165 Table 64: RMS delay (ns) versus angle position ...... 165 Table 65: Mean delay (ns) versus angle position ...... 169 Table 66: RMS delay (ns) versus angle position ...... 169 Table 67: Mean delay (ns) versus angle position ...... 172 Table 68: RMS delay (ns) versus angle position ...... 172 Table 69: Mean delay (ns) versus angle position ...... 176 Table 70: RMS delay (ns) versus angle position ...... 176

R2R-T2.2-D-DZR-006-18 Page 17 of 186 10/12/2016

Contract No. H2020 – 636032

1. INTRODUCTION This Roll2Rail project covers different rolling stock topics such as Traction (WP1), TCMS (WP2), Car-Body-Shell (WP3), Running-Gear (WP4), Brakes (WP5), Vehicle Interiors (WP6) and transversal activities such as Noise (WP7) and Energy Management (WP8).

In that context, WP2 work package’s concrete goal is to make research on technologies and architectures to allow new generation of train communication systems based on Wireless Transmission for Train Control and Monitoring System (TCMS), functions and Infotainment, CCTV applications, thus reducing or even completely eliminating, on board communication cables and simplifying the train coupling procedure.

The goal of this deliverable (D2.2 within WP2) is to study and characterise the railway environment in order to elaborate models for later simulation of radio communications technology for railways. Most relevant scenarios, where reflections, radio collisions or EMI may play a role, will be analysed. Therefore metro and mainline tunnels and stations, different power supply (i.e. AC, DC) and signalling systems will be considered. The objective of this deliverable is the faithful characterization of the railway environment for radio transmission. In order to avoid duplication of work load and results, Chapter 2 gives a thorough overview of existing channel measurements and channel models in the railway domain and identifies the missing pieces which are required for a complete characterization of the radio environment. Chapter 3 defines a common methodology which will be applied in the different measurement campaigns and reports relevant aspects and experiences from already finalized measurements. In Chapter 4, the measurement campaigns are described in quite some detail.

Chapter 5 describes the actual three different measurement campaigns carried out, i.e. the high- speed line measurements and the urban/metro measurements in the field as well as 60 GHz measurements for a regional train in a depot. Finally, Chapter 6 reports the post processing and analysis of the channel measurements as well as parameters for channel models considering large scale and small scale channel characteristics. These channel models are the basis for the simulations in Task 2.7 to evaluate selected wireless technologies and in channel emulators in Task 2.8 to validate suitable radio technologies for wireless TCMS.

R2R-T2.2-D-DZR-006-18 Page 18 of 186 10/12/2016

Contract No. H2020 – 636032

2. STATE-OF-THE-ART ON MEASUREMENTS AND MODELS IN RAILWAYS

2.1 CURRENT AND FUTURE RAILWAY COMMUNICATION SYSTEMS

2.1.1 Inside Train Communication On-board communication networks were installed aboard trains since the end of the 80s to reduce the cable beams used to transfer information between different devices. Multiplexing digital information technics over a serial cable have tried to replace most of the classical point-to-point copper lines or so called train lines. Wired communication networks were standardized for on- board railway applications in the end of the 90s (standard [1] and [2] by defining Wire Train Bus/Multifunction Vehicle Bus (WTB/MVB) networks for TCMS applications). In [3] a survey of railway embedded network solutions is presented.

Standard technologies such as WorldFIP, CANOpen, LonWorks, Profibus or Train Communication Network (TCN) are deployed either for metro or trains. Since the 2000s, manufacturers considered the Real time Ethernet technologies (RTE) by adding new standards to IEC 61375 standard series, such as Ethernet Train Backbone or Ethernet Consist Network (ETB/ECN). In addition to the control-command functions offered by the classical fieldbuses technologies, RTE provides Internet Protocol (IP) traffic.

In recent years, power line communication technology for communications inside vehicles in the field of aerospace and automotive industries experience important developments. This is also true for public transport vehicles.

2.1.2 Outside Train Communication Communication between vehicles in the public transport sector covers several applications. The first one, often referred to as the concept of carrier pigeon, consists in providing information on the fly between two vehicles. The use case is often a disabled vehicle, out of range of a communication network, that will transmit information to another vehicle passing nearby [4], [5]. Another use case, still in the research labs, is virtual coupling of two vehicles (car trains, subways, trams) this means no mechanical coupling but only wireless coupling for Train-to-Train (T2T) communications. Indeed, in the case of mechanical coupling of two vehicles (subways, trams, trains, High Speed Railway (HSR)), it is also necessary to interconnect high-speed networks embedded in both vehicles. This is now carried out mechanically using specific connectors that deteriorate quickly under hard railway operation conditions. One of the candidate technologies is UWB (Ultra Wide Band) such as the IEEE 802.15.4a standard. A survey of possible UWB applications in the railway domain is given in [6]. The UWB links are deemed more robust to frequency selective fading [7]. New technologies in the 60 GHz range like IEEE 802.11ad and machine-to-machine type communication systems as being defined in 4G and 5G may also be considered.

R2R-T2.2-D-DZR-006-18 Page 19 of 186 10/12/2016

Contract No. H2020 – 636032

2.1.3 Current Wireless Systems Apart from legacy systems (usually analog) that started its development in the early 80s, wireless systems in railways is a trend that is still on its first decade of life. There are three types of systems: first, those based on open standards, like Terrestrial Trunked Radio (TETRA), Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), IEEE 802.11 family of standards; second, slight modifications on some layer, but already based on open standards (e.g. Global System for Mobile Communications – Railway (GSM-R)); and, finally, proprietarily developed technologies for railways, e.g. Eurobalise.

Within the first type, there are many success cases, not only in the train-to-wayside field, but also in the sensor network field, with sensors all over the train, linked by a Bluetooth or Zigbee network. Within the vehicles, another common solution is to develop a wireless local area network to provide passengers’ access to the Internet. In this case, the link to the ground is provided by a mobile operator using Long Term Evolution (LTE), Universal Mobile Telecommunications System (UMTS) or GPRS. Moreover, ticket validation equipment based on Low Frequency (125-135 kHz) and High Frequency (13.56 MHz) RFID bands should be considered, and also vehicle tagging based on Ultra High Frequency RFID solutions: 865-869 MHz (EU), 902-928 MHz (USA), 952-955 MHz (Japan).

GSM-R is the most famous system of the second type because it is based on GSM standard phase 2+ [8], but with major modifications to fulfil many railway-based needs, like functional or regional addressing and many more. In subways, it is very likely to have some implementations of IEEE 802.11g with optimizations at certain levels to improve its performance, e.g. mobility aspects, like the TEBATREN solution already in service in Metro de Madrid and in some others. In the signalling field, the use of IEEE802.11-based radio for the Communication-Based Train Control (CBTC) solutions exists since several years [9]. Examples are systems from Bombardier and Dimetronic (now part of Siemens) and also Alstom’s Urbalis System.

Moreover, proprietary solutions have also its niche in the market. Traincom by Telefunken (recently acquired by Siemens) is a good example with a great acceptance in the railway sector. It is now in service in many places around the world, but the driverless line of Barcelona Metro is perhaps the most famous implementation of this solution. Another case is the Israeli company Radwin in Moscow Metro, launched last summer.

Further, new developments are LTE for CBTC applications for metro and tramways [10] and the railway proprietary European Train Control System (ETCS)-EUROBALISE application standardized at 27.095 MHz for tele-powering of the trackside beacon and 4.234 MHz for the communications itself [11].

2.1.4 Possible Directions for Future Wireless Systems The future wireless systems for railways need to address many issues like costs, interoperability and spectrum allocation. Depending on the point of view, actual technology is very expensive and sometimes it is not interoperable at all. GSM-R is a possible exception in terms of interoperability only. However, open standards like 3GPP LTE imply also heavy costs and possible dependence on mobile operators, which is something unlikely to be accepted by railway operators apart from other disadvantages.

R2R-T2.2-D-DZR-006-18 Page 20 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 1: Future train wireless communications Despite of all these issues, one of the aims of several research groups in Asia and Europe and projects all over the world, including Roll2Rail, is to study feasible wireless communication technologies for both T2G, T2T, and inside train communications (cf. Figure 1). At the current stage of WP2, it is not possible to jump to the conclusions, but some technologies are more feasible than others. It is not a secret that 3GPP LTE has introduced some functionality on its last release that targets the railway sector, like mobile relays, public safety issues, Device-to-Device (D2D) communications, etc. On the other hand, for Car-to-Car (C2C) communications, IEEE 802.11p is planned to be deployed in the smart cars field in the near future. So, it could be an option for T2T communications if high data rates are not required. Other solutions based on UWB technology or millimetre wave solutions in the range of 60 GHz are foreseen.

Furthermore, spectrum allocation is always a challenge. Industrial Scientific and Medical (ISM) bands at 2.4 and 5 GHz are always a possibility, but also imply potential problems, not only for security. Also there is some discussion on the possibility of using the Intelligent Transportation System (ITS) band at 5.9 GHz for urban rail systems. Facing the problem from the business perspective, partnerships with mobile operators to deploy mobile networks and also provide some non-safety services to operators and stakeholders is possible but implies some regulatory challenges that should be addressed, too.

Moreover, it is also important to account for the ongoing work on cognitive radio. The concept of cognitive radio was highlighted as an attractive solution to the problem of congestion of the radio spectrum occupied by licensed users [12], [13]. Cognitive Radio (CR) is a radio or a system capable of analysing its electromagnetic environment and able to dynamically and independently operational radio parameters to modify the operation of the system, i.e. throughput, interference cancellation, the interoperability, access to other radio networks. This field of research is very active at European and international level. For instance, the French project CORRIDOR (COgnitive Radio for Railway through Dynamic and Opportunistic spectrum Reuse) is paving the way for the development of Cognitive Radio technologies for railway applications. The project objectives were to design, develop and evaluate fundamental bricks of a CR system adapted to the requirements and constraints of HSR, e.g. high speed, electromagnetic interference, poor coverage of systems in rural area, etc.. More details and publications can be found in [14].

R2R-T2.2-D-DZR-006-18 Page 21 of 186 10/12/2016

Contract No. H2020 – 636032

To summarize this section, the main focus for future train wireless communications as depicted in Figure 1 needs to be on providing reliable and real-time data links with the required data rate for safety critical applications while providing best-effort high data rate links to other applications. In a first step research should focus on removing cabling and connectors that suffer from mechanical and environmental stress: Consist-to-consist autocoupler, car-to-car cabling, and bogie-to-carbody cabling. In conjunction with new railway applications such as virtual coupling, future train wireless communications will enable the railway operators to reduce downtime of trains and increase efficiency and safety of the railway system.

2.2 EXISTING CHANNEL MODELS AND MEASUREMENTS

2.2.1 Types of Channel Models One can distinguish three main categories of channel modelling approaches for performance evaluation that are used within the context of vehicle to vehicle channels: deterministic, geometrical-stochastic and stochastic.

Deterministic channel models Deterministic channel models characterize the C2C or car-to-infrastructure (C2I) channels in a completely deterministic way. They may be based on rigorous solving of Maxwell’s equations like Method of Moments (MoM) and Finite Difference Time Domain (FDTD) or on asymptotic approaches introducing the concept of “rays”, i.e. ray tracing. [15] and [16] present a new and original methodology to model electromagnetic wave propagation and to find antenna specifications and positions in tunnels of arbitrary cross-section bases on the use of transmission line matrix.

The ray tracing approach that is computationally-wise able to simulate much bigger scenarios compared to MoM or FDTD is the most widely used technique in characterizing channels in a deterministic manner. The main idea of ray tracing is to regenerate ‘rays’ from the transmitter to the receiver, taking into consideration reflections and diffractions occurring due to objects in the environment. Therefore, it is required to define the shape, properties and location of all the involved objects in the transmission environment. Hence, the detailed description of the environment and the following intensive computations are time and effort consuming, and the extracted Channel Impulse Response (CIR) cannot be easily generalized to different scenarios. In [17] ray tracing was applied to the car-to-car channel. The implementation was divided into three distinguishable parts. The first part was the modelling of the road traffic, i.e. other cars. To characterize the channel’s Doppler shift and Doppler spread accurately, the dynamic behaviour of the transmitting and receiving vehicles as well as the adjacent vehicles have to be characterized in a realistic manner. Authors considered microscopic modelling to simulate each vehicle separately, resulting in a realistic instantaneous position and velocity for each vehicle at each snapshot in the simulation. The Wiedemann traffic model was implemented to describe road traffic for different scenarios. The second part was the modelling of the environment, where a stochastic model was chosen to draws the environmental surroundings such as buildings, parked vehicles and trees. The third and last part is the modelling of the wave propagation. Here, ray tracing was used to simulate the multipath propagation from the transmitter to the receiver. A carrier frequency of 5.2 GHz was used in the simulation. Results were validated against measurement data obtained using a RUSK ATM vector channel sounder operating at the same carrier frequency. The comparison between

R2R-T2.2-D-DZR-006-18 Page 22 of 186 10/12/2016

Contract No. H2020 – 636032

measurement and simulation showed that the proposed deterministic channel model fits well to measurement results and real-life scenarios. Ray tracing for T2G environments was investigated in the mm-wave band in [18], where a HSR environment was considered including a transmitter mounted on the top of a fence while the receiver antenna is located on the top of the train. For the scenario, the ray tracing simulator RapLap was used to simulate the wireless propagation channel. Results, such as received power, delay spread and angular spread, were provided. A dynamic channel model was also proposed using obtained statistical parameters from the deterministic simulation method. A comparison to measurement data was omitted. In [19] the authors focused on the propagation channel inside a train wagon using ray tracing based on the EM CUBE software simulator from Emag Technologies Inc. [20].

It is important to notice that ray tracing techniques work only for plane surfaces. For curved surfaces the image theory which is used in ray tracing approach is no more valid and a ray launching approach associated with for example analytical description of the surfaces should be consider. Detailed are given in [21] and [22]. [22] focuses on radio wave propagation modelling for metro tunnels. The case of arched and curved tunnels is treated. Theoretical results obtained are compared with experimental results in various tunnels.

Geometry-based stochastic channel models (GSCM) In GSCM, scatterers, representing propagation paths are distributed in a virtual geometrical environment. Using the geometrical relations between the transmitter, a scatterer and the receiver, the delay and angle of arrival of different propagation paths can be calculated according to a simplified ray tracing procedure. Diffuse multipath contributions may be included by considering clusters of points as scatterers. GSCM can be easily adapted to different scenarios by changing the geometrical distribution of the scatterers, making a good compromise between complexity and accuracy. Furthermore, GSCM can be easily adapted to non-stationary environments based on the geometrical relations, making them a good candidate to describe T2T and T2G channels. In recent years, GSCM gained a lot of research interest in the C2C domain; their flexibility makes it tempting to extensively develop them in the T2T and T2G domains. Authors in [23] adopted a GSCM to characterize the MIMO channel in the C2C domain. Distributed scatterers were divided into three different groups. The first group represents paths occurring due to wave interactions with moving objects, i.e. other vehicles. A second group represents propagation paths from static objects like road signs or other structures next to the road or in the middle between both traffic lanes. The last group describes diffuse components originating from trees, buildings or walls. Each group is given different stochastic and deterministic properties (such as geometrical density, path-loss exponent and reference power). The model was validated by comparing simulations to measurements performed with the RUSK LUND channel sounder in both rural motorway and highway environments. Detailed discussion on vehicular channel characterization was presented in [24] and [25], where the authors discussed C2C channel and GSCM in detail. A GSCM for HSR was developed in [26] where a non-stationary wideband MIMO channel model was proposed that includes the Line-Of-Sight (LOS) component and propagation paths occurring due to one-time scattering.

R2R-T2.2-D-DZR-006-18 Page 23 of 186 10/12/2016

Contract No. H2020 – 636032

Stochastic channel models (SCM) SCM tend to describe the propagation channel based on stochastics without considering the underlying geometrical relations. In this type of channel models a certain structure (such as tapped delay line structure, Saleh-Valenzuela structure or finite state Markov chains structure) is assumed and modelled by random processes. Stochastic parameters for the characterization depend on pre-defined generalized environments (such as rural or highway environments) as well as different assumptions. To parameterize a model, intensive measurements in different scenarios have to be performed. Channel model parameters are then tuned to fit the results of the measurements. A tapped delay-Doppler profile model was proposed in [27] and [28] for the C2C channels. This model was also adopted by the IEEE 802.11p standards group for the development of its system. However, the Wide-Sense Stationary Uncorrelated Scattering (WSSUS) assumption in the channel model does not reflect the non-stationarity of the channel impulse responses reported in the measurements for the C2C environments. Authors in [29] provided parameters for C2C channel models based on the tapped-delay structure. Three model types were provided, where two are non-WSSUS based and the third one was based on the WSSUS assumption. Parameters were given for channel bandwidths of 5 and 10 MHz using measurements performed by the authors in both highway and urban scenarios. These measurements were taken at different times and under different traffic conditions. The model was updated by results in [30] using the same type of channel model to characterize propagation for channel bandwidths of 1, 20, 33.33 and 50 MHz, where the same data from the measurement campaign in [29] was used. In [31], the Rayleigh fading channel was modelled using a finite-state Markov structure as an evolution of the two-state Markov channel known as the Gilbert-Elliot channel [32] [33]. Finite-state Markov models were later developed to model tunnel channels [34] and the fast time-varying C2G channels [35]. Measurements of T2G in viaduct environment and agricultural environment were performed in [36] and [37]. In [36] an evaluation and development of LTE technology for HSR communications was presented. In [37], authors proposed a two-dimensional Ricean K-factor channel model based on the measurement results.

2.2.2 Measurements inside Vehicle An important issue for wireless networks in railway environments is the presence of a severe multipath due to the reflections in the metallic train structures; in order to tackle this, wireless propagation features need to be obtained through channel measurements and the results need to be carefully considered for the selection, design and deployment of the physical layer of the wireless network.

Concerning the characterization of the wireless propagation channel for railway environments, literature about this topic usually includes research papers about the propagation in different scenarios and at different frequency bands [38] [39]. The effect of certain structures like viaducts and terrain cuttings (canyons) onto a train-to-ground communication link (GSM-R) has been taken into account in [40] [41] [42] [43]. Results show that the classical models for propagation loss are not accurate for attenuation prediction.

All these works are related to the propagation study of the train-to-ground link, but wireless network applications require the characterization of the on board propagation. In [44] [45], intra-vehicle and inter-vehicle communication links are analysed using the 2 and 5 GHz bands. Measurement results show that the transmitted signal can re-enter vehicles through windows and that its contribution to inter-vehicle propagation is more relevant than the line-of-sight (LOS) signal. On the other hand,

R2R-T2.2-D-DZR-006-18 Page 24 of 186 10/12/2016

Contract No. H2020 – 636032

wideband propagation analysis with a 2.35 GHz channel sounder using planar and omnidirectional antennas placed at different locations is done in [46]. The performance of both antennas is compared, obtaining a larger path-loss in the case of the planar antenna, and showing a similar performance regarding delay spread.

The authors in [47] analyse the path loss and delay spread for both inter-vehicle and intra-vehicle communication links. All the measurements are done for a 100 MHz signal bandwidth using a VNA for the 2.45 GHz band, where results show that path loss is slightly smaller than free space. In [48], the authors study the wave propagation around the train, focusing on the wireless sensor network application, and taking into account that the wireless devices need to be installed under the train, e.g. bogie monitoring. They use a narrowband approach at 434 MHz band for wireless channel characterization by means of a signal generator and a spectrum analyser. Results show how the path loss increases for the case where the antennas are located below the train, in contrast to the case where the antennas are inside the train.

Apart from reviewing inside vehicle channel models for trains, we also summarize the literature on similar environments, i.e. bus and airplane cabins. Whereas only [49] measured channels inside a bus, [50], [51], [52], [53], [54], [55], [56], [57], [58], [59] measured channels inside airplanes. All measurements have been conducted for stationary vehicles. Further, [49], [54], [55] consider the influence of passengers on the channel measurements. For the large wide-body cabin of an A380 airplane with two aisles, [54] shows a reduced delay spread for UWB measurements between 3-8 GHz. Clearly, these results are not directly applicable to the train environment as passenger trains usually are single aisle. In contrast [49] and [55] conclude that human movement causes larger delay spreads at 5 GHz for a measurement bandwidth of 50 MHz inside a bus or medium-sized airplane with a single aisle. The airplane cabin measurements of [50], [51], [53], [54], [56], [57] show consistently a path loss exponent between two and three for medium to large aircrafts depending on the measurement setup, carrier frequency, and bandwidth. As an exception to these findings, the estimated path loss exponents are below one in the cargo bay of a military airplane [58], [59]. Here, the different behavior stems from the metallic cabin containing no equipment that absorbs the UWB signals. Similar behavior could be expected inside an empty cargo compartment of a train vehicle.

When considering the bandwidth of the measurements, we can distinguish narrow band, wide band and UWB measurements. For the UWB measurements [51], [54], [57], [58], [59], the findings vary greatly depending on the exact measurement setup, often requiring a large number of multipath components to model the channel characteristics accurately, e.g. [58]. The wide band models exhibit for carrier frequencies around 2 or 5 GHz frequency selective fading [49], [52], [55] [56] and contain both Ricean and Rayleigh fading paths. A 3-path tapped delay line model can be sufficient to describe the in-cabin channel behavior [49] [55]. For the narrow band measurements, the resulting channel model is a flat fading model [53]. The influence of signals propagating outside the cabin and then re-entering it are not considered, apart from the environment description in [49].

Ray tracing simulators can also be used for characterizing in-vehicle propagation [60] [61]; however, results are highly dependent on the accuracy of the simulation model.

R2R-T2.2-D-DZR-006-18 Page 25 of 186 10/12/2016

Contract No. H2020 – 636032

2.2.3 Measurements outside Vehicle Several works exist related to the characterization of train to ground wireless propagation [38], [39]. This is an active field of research these last years. The effect of structures like viaducts and terrain cuttings (canyons) onto a train-to-ground communication link with GSM-R has been analysed in [40], [41], [42], [62]. Overall results show that the classical models for propagation loss are not accurate for attenuation prediction.

While the research of the T2G channel is comprehensive, the propagation channel of the T2T is hardly described in literature. A measurement and analysis of a T2T channel was done in [63] considering the use of TETRA. Authors in [64] describe a channel model for a direct T2T link at 400°MHz centre frequency based on known communication models.

In order to characterize the propagation environment for inter-vehicle communications, a wireless channel measurement campaign was carried out in [65]. This campaign was done at La Sagra maintenance facilities in Mocejón, Spain. For these measurements the train was positioned in an open field in order to obtain the multipath coming from the train itself and not from reflections on surrounding objects. It was observed that in absence of reflections outside the train, the radio communication between the inside of the car and the bogie occurs due to the diffraction of the signal on the window edge and on the bottom edge of the car. Electromagnetic simulations were carried out with CST Microwave Studio [66] that confirms this behaviour (see Figure 2). As an example, Figure 3 shows the simulated and measured power-loss results at 2.45 GHz between two vertically-polarized antennas placed with a separation of 3 meters following the setup described in Figure 2. A good agreement is observed between simulations and measurements; the main difference is that in simulations higher power rays are received after the first 20 meters, which are due to the perfect conducting structures used in the car for obtaining reasonable simulation times, which make the car more reflective than in the real situation. On the other hand, it can be noted that the losses of -75 dB for the first ray are a combination of 50 dB free-space losses (i.e. 3 meters at 2.45 GHz) plus 25 dB of diffraction losses at the window edge.

Figure 2: Diffraction phenomena around car: electromagnetic simulation model (left); simulated electric field (right) It was also observed that when the base station is inside the car and the sensor nodes are under the car, coherence bandwidth 퐵푐 varies from 1.5 to 4 MHz at 2.45 GHz, and it decreases as the distance between the base station and the sensor node becomes larger. When both the base station and sensor nodes are under the car, 퐵푐 varies from 15 to 25 MHz. On the other hand, the delay spread is larger when the base station is inside the car (13.3 ns to 100 ns) than when is it under it (8.7 ns to 13.3 ns), due to the stronger multipath. Finally, the attenuation of the links, when the base station is inside the car, goes from 85 to 88 dB at 2.45 GHz, while it varies from 34 to 53 dB when it is under the car. Cross-polarization of the antennas has an influence when the LOS path prevails, and causes an attenuation increase of 10-15 dB.

R2R-T2.2-D-DZR-006-18 Page 26 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 3: Simulated and measured power-loss results for 3-meter setup with vertical polarization at 2.45 GHz From these results it was concluded that positioning the base station under the vehicle provided a more stable link, with lower delay spread and higher 퐵푐. However, it must be noted that wireless sensor networks operating with a bandwidth lower than the 퐵푐 will suffer from flat fading, and therefore spatial antenna diversity and/or frequency agility will need to be implemented to overcome this issue.

2.2.4 Identified Gaps of Existing Measurements and Models The summary of existing channel measurements and models shows that important aspects for modern wireless T2T applications are missing. Most of the measurements and the resulting channel models refer to T2G. The small amount of investigations on T2T aspects are limited on either low train speed or lower frequency bands. In several publications cellular channel models are used for C2C communications. Apart from the points mentioned in Section 2.1.4, the following gaps have been identified.

Railway Environment The railway environment differs depending on the type of track significantly to a road environment. Furthermore, the shape of cars on a street differs much more severely compared to the profile of trains for wave propagation. On the other side, most of the artificial and non-artificial obstacles that are found next to roads (see Table 1) may occur also next to railways.

One of the most challenging and diverse environments for communications in railway traffic are tunnels. Especially underground trains but also HSR and commuter trains are operating in tunnels. The shape and the material of the tunnel is heavily influencing the propagation: tunnels excavated with boring machines (i.e. smooth walls, with almost no changes on their cross-section – see Figure 4a), man-made tunnels (i.e. frequent changes on tunnel section, walls made of bricks, etc.), one-track tunnels (Figure 4b), stations (both pit-shaped and tunnel-shaped, Figure 4c and Figure 4d respectively). Some tunnels are also made with metal (some case in new-York and London). In this case the propagation rules inside these types of tunnels are completely different.

R2R-T2.2-D-DZR-006-18 Page 27 of 186 10/12/2016

Contract No. H2020 – 636032

Table 1: Railway environment Urban Area Sub-Urban Rural Special Curves scenarios Tunnels Bridge / Viaduct Cuttings Cross Bridge Noise Barrier Catenary Obstacles Signalling System Roof Building Vegetation (tree) Open field

(a) (b) (c) (d)

Figure 4: Different lateral cuts of tunnels

High Total Velocity Previous measurements reported in underground or general railway environments have been mostly focused on the T2G links. Future measurements shall focus on intra-vehicle, inter-vehicle, inter-consist and T2T links from mid velocities in subway networks up to high total velocities for HSR.

The Intra-vehicle scenario investigates the wireless link between different elements inside a single vehicle (see Figure 5 left). These will be LOS links, mostly affected by the internal structure of the vehicle and passengers. Here several scenarios shall be measured, such as person presence/movement, narrow/wide train gauge, as well as vehicles with/without corridor in-between (i.e. continuous and non-continuous trains – see Figure 6).

R2R-T2.2-D-DZR-006-18 Page 28 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 5: Intra-vehicle (left) and inter-vehicle (right); note that the vertical gray lines depict the boundaries between rail vehicles

Figure 6: Continuous (left) and non-continuous (right) trains

(a)

(b) Figure 7: Inter-consist: (a) centre; (b) edge

R2R-T2.2-D-DZR-006-18 Page 29 of 186 10/12/2016

Contract No. H2020 – 636032

Inter-vehicle scenarios include wireless links that go beyond a single vehicle, involving both the next vehicle and the exterior of the vehicle (both roof and bogie) – see Figure 5 right. In this case the environment inside (see Figure 6) and outside of the train in combination with the velocity is influencing the propagation channel. Hence, it will be necessary to measure and model different propagation environments with different speeds.

An inter-consist connection establishes a wireless link between one consist to another one, with the antennas located on the roof of the train. The antenna position may vary between the centre and the edge of the cars (see Figure 7). In case of omni-directional antennas, high Doppler shifts resulting from the high speed of the train in combination with reflected or scattered multipath components located in the surrounding environment may cause the main influence on the received signal. If directional antennas are used curve radii and train vibrations need to be considered for the beam-width of the employed antennas.

High Relative Velocity The channel between two moving trains regarding the T2T communications as well as the interference between adjacent trains using the same wireless technology is hardly discussed in literature. The scenarios differ from two trains stopped and located in parallel on depot (i.e. continuous interference) to two trains driving next to each other on parallel tracks in the same or opposite direction. The antennas can be placed both inside and on the roofs of the vehicles. Highly interesting is the influence of the Doppler shift on the receiver side. Due to possible relative speeds of 600 km/h and above, high Doppler shifts may occur.

Figure 8: Train-to-train: inside (left), roof (right)

R2R-T2.2-D-DZR-006-18 Page 30 of 186 10/12/2016

Contract No. H2020 – 636032

3. COMMON METHODOLOGY DEFINITION

3.1 REQUIREMENTS FOR MEASUREMENT SCENARIOS The objective of the channel measurement campaigns is the characterization of the physical channel between the transmitting and the receiving antenna for a given frequency band. The outcome of these measurements is generic in the sense that it does not depend on the selected transmission technology but only on the frequency band and the chosen scenario. This approach, which is known as channel sounding, requires a careful preparation of the measurement campaign and involves a considerable amount of professional measurement equipment. In order to obtain good measurements of the channel itself, the absence of any interfering transmitters is highly desirable. The obtained measured data with channel sounding can be used as direct input for simulations or it can be applied to derive one or multiple channel models. The impact of interference can be included in simulations, based also on channel models derived from the channel sounding measurements.

In other words, measurements based on channel sounding only depend on antenna locations and the relevant environment (scatterers, reflecting material and movement) but not on transmit parameters like data rates, packet sizes of modulation formats.

A simpler approach for channel measurements applies a selected wireless technology and directly provides results like packet error rates, delay, data rates for this technology. This has been the method for the measurement with ITS-G5 equipment reported in Section Error! Reference source not found.. The obtained results are naturally specific for the equipment deployed during the measurements, including transmitter, receiver, antennas and cabling.

3.2 SOURCES OF EM NOISE The railway environment increasingly makes use of more innovative applications and seeks to develop new technologies. This requires using more sophisticated equipment for each application. In an environment where different kind of equipment and applications can be found (power electronics, communications links, sensing applications…), all the electric and electronic equipment is required to be electromagnetically compatible, in such a way that the operation of one equipment does not disturb other equipment. This is known as EMC (Electromagnetic Compatibility), and is defined as the ability of an electronic equipment/system to operate as designed in its electromagnetic environment without affecting other equipment/system or being affected by the interferences from other equipment/system. Thus the design of an equipment/system is required to comply with the emission and immunity requirements of the environment in which it is intended to run. Compliance with these requirements is usually demonstrated by testing using the harmonized standards.

For the development of a Wireless TCMS, the effect of the electromagnetic interferences of the environment should be considered for wireless communications. The Electromagnetic Interference, EMI, is defined as the unwanted EM (electromagnetic) energy, which may or may not affect the victim equipment. Disturbances may be produced by either intentional or spurious sources from equipment, or by natural causes (e.g. lightning, electrostatic discharge).

R2R-T2.2-D-DZR-006-18 Page 31 of 186 10/12/2016

Contract No. H2020 – 636032

3.2.1 Regulations and Standards EMC is regulated by the 2004/108/EC directive and the new directive 2014/30/EU, which is applicable from 20 April 2016. According to these EMC directives, for railway environment, the European harmonized standards series EN 50121 is to be used in order to show that the essential EMC requirements are fulfilled. Part 3-2 of EN 50121 corresponds to apparatus on board the rolling stock and sets the requirements for emission and immunity for all equipment on board the rolling stock [67].

Additionally to this, within the rolling stock there are also equipment/systems (e.g. radios) covered by the R&TTE 1999/5/EC directive and the new directive 2014/53/EU which is applicable from 13 June 2016. This directive calls for ETSI (European Telecommunications Standards Institute) to set the corresponding EMC requirements.

3.2.2 EMI Environmental Characterisation For the environmental characterization, a classification of the different EMI sources should be done depending on the nature of the source. Internal EMI sources to the rolling stock cover all EMI sources, both intentional and unintentional, from equipment installed on board the rolling stock and destined to be used in railway applications. External EMI sources to the rolling stock cover all sources from equipment not destined for railway but which may affect the electromagnetic compatibility of the victim equipment.

Mainly there are four ways where EMI can couple into systems. These coupling mechanisms are radiated, conductive, capacitive and inductive. For this activity we will focus in radiated emissions as a source EMI.

3.2.3 Internal EMI sources Unintentional EMI sources

Within this category are included the equipment on board the rolling stock which do not have an electromagnetic radiating functionality (antenna) and their radiated emissions correspond to a spurious emission.

The requirements for radiated emissions for the on board apparatus (non-intentional) of rolling stock are defined in EN 50121-3-2 [67]. Under this classification are considered equipment as:

 Traction system  Auxiliary converter  Brake system  Access doors  Heating, ventilation, and air conditioning (Cabin and passengers)  …

R2R-T2.2-D-DZR-006-18 Page 32 of 186 10/12/2016

Contract No. H2020 – 636032

The EN 50121-3-2 standard sets the following requirements for the radiated emissions:

Table 2: European Standard EN50121-3-2 vs 2006 EN 50121-3-2:2006

Frequency Limit Ref. standard

30 MHz – 230 MHz 40 dBµV/m QP EN 55011 These limits are defined for a measurement distance of 10 m. The limits are increased 230 MHz – 1 GHz 47 dBµV/m QP EN 55011 by 10 dB for a measurement distance of 3 m.

In March 2015, the new version of EN 50121-3-2 was published and which includes additional requirements for the radiated emissions [67]. The new requirements cover the gap between 1 GHz to 6 GHz which is currently in use for several applications.

Table 3: European Standard EN50121-3-2 vs 2015 EN 50121-3-2:2015

Frequency Limit Ref. standard 30 MHz – 230 MHz 40 dBµV/m QP Measurement instrumentation defined in Clause 4of EN55016-1-1:2010. Measuring antennas defined in 4.4 of: EN 55016-1-4:2010. Measuring site described in Measurement distance 10 m. 230 MHz – 1 GHz 47 dBµV/m QP Clause 5 of EN 55016-1- 4:2010. Measurement method defined in 7.2 of EN 55016-3- 2:2010. 76 dBµV/m Peak Measurement instrumentation 1 GHz – 3 GHz defined in Clause 5 and 6 of 56 dBµV/m Avg EN55016-1-1:2010. Measuring antennas defined in 4.5of: EN 55016-1-4:2010. Measuring site described in Measurement distance 3 m. 80 dBµV/m Peak 3 GHz- 6 GHz Clause 8 of EN 55016-1- 60 dBµV/m Avg 4:2010. Measurement method defined in 7.3 of EN 55016-3- 2:2010.

R2R-T2.2-D-DZR-006-18 Page 33 of 186 10/12/2016

Contract No. H2020 – 636032

Intentional EMI sources

Within this category are included the equipment on board the rolling stock which include electromagnetic radiation as part of their functionality. Radio systems, fleet management systems, passenger information systems, train-to-wayside communication systems, ticket validation systems and signalling systems are included within this category. Mainly, the technologies in use by these equipment are mobile phone networks (GSM, UMTS, LTE), Wi-Fi, RFID and radio technologies (analogue radios and TETRA).

More detailed information is presented below for the different equipment/system:

ERTMS. The European rail traffic management system has two basic components, ETCS, the European Train Control system, and GSM-R, a radio system for providing voice and data communication.

 GSM-R is an international communication system for railway applications based on GSM and EIRENE specification. GSM-R is standardized in 900 MHz or 1800 MHz GSM frequency bands to be used around the world, but depending on the country GSM-R can use specific frequency bands.  ETCS – Eurobalise – BTM is standardized at 27.095 MHz for tele-powering the track balise and 4.234 MHz for the communication. Details for this communications are defined in the different UNISIG subsets. CBTC (Communications-Based Train Control)

CBTC is used in the open ISM radio band 2.4 GHz and 5.8 GHz for Wi-Fi. This is specified in IEEE standard “1474.1-2004 – IEEE Standard for Communications-Based Control (CBTC) Performance and Functional Requirements”.

Radio

 Analogue radio: depending on the country and train operator different radio systems are used.  TETRA (Terrestrial Trunked Radio) radio system. Ticket Validation Systems

Ticketing systems use RFID technology. The RFID frequency bands are LF (125-135 kHz), HF (13.56 MHz) and UHF (860-960 MHz). Mainly LF and HF devices are used for ticket validation systems and are catalogued as short-range devices.

Fleet Management Systems – Train to wayside communication (TWC)

Technologies used are mobile phone networks for the train to wayside communications. Also there are systems using inductive loops for vehicle detection (VETAG) that work between 10-100 kHz.

Passenger Information Systems

The information is received from an automatic vehicle location system and from control systems. Technologies used are GPS (reception) and mobile data phone network (GPRS, 3G or LTE) or Wi- Fi for data reception.

R2R-T2.2-D-DZR-006-18 Page 34 of 186 10/12/2016

Contract No. H2020 – 636032

3.2.4 External EMI sources As informative, external EMI sources to the rolling stock include sources like:

 Intentional emitters: radio, TV or mobile phone networks.

 Natural EMI: lightning and electrostatic discharge.

 Illegal jamming.

3.2.5 Summary and additional information In the following table a summary of the identified EMI sources is presented including additional information of working frequencies and emission power

R2R-T2.2-D-DZR-006-18 Page 35 of 186 10/12/2016

Contract No. H2020 – 636032

Table 4 Intentional EMI Sources

EQUIPMENT TECHNOLOGY FREQUENCY EMISSION POWER

Tx Range: 876 to 915 MHz GSMR 8 W Rx Range: 921 to 960 MHz Magnetic flux between antenna and ERTMS Power transmission frequency: 27.095 MHz ± 5 kHz (acc. balise Eurobalise UNISIG SUBSET-036) 7.7 nVs < Φ < 300 nVs (acc. Data transmission frequency: FC=4.234 MHz ± 175 kHz UNISIG SUBSET-036 - section 5.2.2.6)

Tx Range: 876 to 915 MHz GSMR 8 W Rx Range: 921 to 960 MHz

RADIO

Analogue radios. According Tx Range: 456.95 MHz to 458.65 MHz HF power output: 8 W - 10 W to country. Rx Range: 466.95 MHz to 468.65 MHz Emergency systems: 380-385 MHz, 390-395 MHz TETRA 10 W Civil systems: 410-470 MHz, 870-921 MHz GPRS: 900 + 1800 MHz for Spain. There can be different GPRS 2 W Fault values for other countries. Management 3G: 900 + 2100 MHz for Spain. However differences for System 3G 1 W other countries.

PIS GPRS: 900 MHz + 1800 MHz for Spain. There can be GPRS 2 W Passenger different values for other countries.

R2R-T2.2-T-DZR-006-17 Page 36 of 186 10/12/2016

Contract No. H2020 – 636032

EQUIPMENT TECHNOLOGY FREQUENCY EMISSION POWER

information 3G: 900 MHz + 2100 MHz for Spain. There can be different 3G 1 W system values for other countries. 4G/LTE: 800 MHz, 1800 MHz + 2600 MHz. for Spain. There 4G 1 W can be different values for other countries.

WIFI 2.4 GHz – 5 GHz 500 mW

CBTC Communication WIFI 2.8 GHz or 4.8 GHz Based Train Control RFID RFID Proximity cards 125 KHz Same as CBTC or ATO proprietary. VETAG – Vehicle ( TAG type) 865-869 MHz (EU), 902-928 MHz (US), 952-955 MHz (JPN)

Tagging UHF RFID

R2R-T2.2-T-DZR-006-17 Page 37 of 186 10/12/2016

Contract No. H2020 – 636032

3.2.6 Equipment Location Following two examples of different vehicles are presented with the on board location of the EMI sources from the main equipment.

Non-intentional equipment fulfilling EN 50121-3-2 are placed in the different cabinets along the train (passenger area or cabin), on roof or underframe depending on the equipment and the type of train.

RFID readers are placed near doors as this technology is used mainly for ticket validation application.

Wi-Fi and Radio/Mobile Technology antennas are placed on roof, mainly in the cabin, but also can be found in other places of the vehicle roof. Generally, the preferred location for these antennas is away from pantograph location because the sparks in the pantograph – catenary interface are the biggest interference source for this kind of equipment.

Figure 9: Vehicle Example A (Tram)

Figure 10: Vehicle Example B (Tram)

Figure 11: Vehicle Example C (Metro Madrid)

R2R-T2.2-T-DZR-006-18 Page 38 of 186 10/12/2016

Contract No. H2020 – 636032

3.3 AVAILABLE MEASUREMENT EQUIPMENT

3.3.1 IK4-IKERLAN has the following equipment for channel measurements:  For static measurements, a Keysight E5071A network analyser (frequency range up to 8.5 GHz) equipped with Electro-Metrics EM-6865 omnidirectional UWB (2-18 GHz) antennas is used (see Figure 12). This setup has the advantage of allowing a perfect synchronization between transmitter and receiver units, as they are both housed in the same equipment; conversely, having both units together limits the measurement distance, as long cables need to be run to transmitter and receiver antennas. The maximum 0 dBm output power of the VNA also limits the measurement distance. This setup has been used for indoor measurements with medium transmitter-receiver distances (e.g. office environments).

Figure 12: VNA-based measurement setup for static channels: VNA (left); UWB antennas (right)  For static and dynamic measurements, a setup made of a Keysight E4438C signal generator as transmitter and a 89600 Vector Signal Analyser (VSA) as receiver is used (see Figure 13). This setup also uses Electro-Metrics EM-6865 antennas. A multicarrier OFDM signal is programmed in the generator, and it is equipped with an external power amplifier (MiniCircuits TVA-11-422) in order to reach 1 W output power; on the other hand, a downconversion stage is added before the VSA in order to fit the signal into a 70 MHz Intermediate Frequency. Furthermore, in order to allow the synchronization of the oscillators in transmission and reception without running cables between them, Rubidium oscillators (Stanford Research Systems PRS10) with GPS receivers (Acutime GG Smart Antenna by Trimble) are used. This setup has been used for medium-long distance indoor and outdoor measurements.

R2R-T2.2-T-DZR-006-18 Page 39 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 13: VSA-based measurement setup: transmitter (left); receiver (right)

We have also used a Medav RUSK channel sounder for channel measurement campaigns [68] (Figure 14). This sounder can measure RF channels up to 6 GHz with a bandwidth of 240 MHz, a minimum period of 1.6 microseconds, and a maximum output power of 2W.

Figure 14: RUSK Channel Sounder in railway environment

Medav RUSK supports multiple transmit and receive antenna configurations. It measures the channel response matrix between transmitting and receiving antennas sequentially by switching between different transmitter-receiver antenna pairs (i.e. the sounder uses only one physical transmitter and receiver channel each time). In order to accomplish synchronous switching, rubidium reference oscillators are used at both the transmitter and the receiver. A block diagram of RUSK is shown in Figure 15.

R2R-T2.2-T-DZR-006-18 Page 40 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 15: MEDAV RUSK block diagram

For channel excitation, RUSK uses a Multi-Carrier Spread Spectrum Signal (MCSSS) with an almost rectangular shape in the frequency domain. This approach allows concentration of the transmitted signal energy in the band of interest (see Figure 16).

Figure 16: Broadband multicarrier spread spectrum signal (MCSSS) magnitude in time (left) and frequency domain (right)

3.3.2 DLR equipment DLR has the following equipment at their disposal:

 Rusk Channel Sounder: The most accurate measurement results would be available by using the Rusk Channel Sounder. For further channel models these data would be the best choice. This measurement system implicates the highest technical effort.

 Signal Generator + Signal Analyzer: This setting is a slimmed channel sounder. The data would be not as accurate as with the Rusk Channel Sounder and the effort would be negligible smaller.

 IEEE 802.11p Cohda MK5: The Cohda MK5 boards are working with the IEEE 802.11p standard and the measurement results would represent the error rates of a communication between two boards. This data is not useable to generate a channel model. Nevertheless, the effort is very small and these measurements could be done in parallel to the channel sounding.

R2R-T2.2-T-DZR-006-18 Page 41 of 186 10/12/2016

Contract No. H2020 – 636032

3.4 MEASUREMENT EXPERIENCE KNOW-HOW IN RAILWAYS

3.4.1 IFSTTAR IFSTTAR has ample experience in channel measurements for a wide range of scenarios for railway and automotive communications in the 2.45 GHz and 5.8 GHz bands, in the 10 GHz and 60 GHz band for anti-collision radar applications. The measurements concerns V2I configurations in outdoor or in tunnels and also V2V configurations [69], [70], [5]. IFSTTAR have also experience on EMC measurements in general [71].

IFSTTAR has also experience in measurements for GSM-R and also for GSM-R deployment along high speed line in the context of works realized for CERTIFER the notify body in France.

Here after the various types of equipment already available.

Table 5: Overview of measurement equipment equipment frequency BW nb spectrum analyser PXA 9020 20 Hz to 8.4 GHz 160 MHz 1 spectrum analyser MXA 9020 20 Hz to 8.4 GHz 160 MHz 1 Generator R&S 300 kHz to 6.4 GHz 18 MHz 2 Generator Anritsu 68397C 20 MHz to 67 GHz ~20MHz 1 VNA HP 8720 D 50MHz to 20 GHz 1 VNA HP 8753S 200 Khz to 6 GHz 1 Tektronix Scope TDS6124 12 GHz 40 Gs 1 Tektronix AWG 7102 10 GS/s 2

3.4.2 DLR DLR has ample experience in channel measurements for a wide range of scenarios, including

 Maritime communications in the 5-6 GHz band, with measurements on the Baltic Sea [72]

 Aeronautical communications [73] [74] [75]

 Mobile radio outdoor-to-indoor for positioning [76] [77]

 Airport channel measurements for AeroMACS [78]

 Mobile vehicle-to-vehicle channel measurements [79]

 Modelling of MIMO satellite channel with measurements from a zeppelin [80] [81]

In the railway domain, measurements have been performed to characterize the train-to-train channel in the 70 cm UHF band [82] and for the TETRA system [63]. These measurements formed the basis for the development of the Railway Collision Avoidance System (RCAS) [83], which takes advantage of direct train-to-train communication to increase the safety.

R2R-T2.2-T-DZR-006-18 Page 42 of 186 10/12/2016

Contract No. H2020 – 636032

3.4.3 IK4-Ikerlan Experience IK4-Ikerlan has experience executing wireless channel measurement campaigns in two particular scenarios for railway applications.

La Sagra maintenance facilities A wireless channel measurement campaign has been carried out on a high speed train located at La Sagra maintenance facilities in Mocejón, Spain. The goal was the characterization of the propagation environment where a wireless sensor network will be operating.

Figure 17: Antennas on wheel and bogie for channel measurements

For these measurements the train has been positioned in an open field in order to obtain the multipath coming from the train itself and not from reflections on surrounding objects. Omnidirectional magnetic-mount monopole antennas (Mobile Mark MGRM-UMB-1C [84]]) have been placed in six transmitting positions on bogie and wheels (see example in Figure 17), while receiving antennas have been placed inside and under the car in order to simulate two location options for the base station (see Figure 18). For each measurement, a transmitter antenna and a receiving antenna have been connected to two ports of a Vector Network Analyzer (VNA); the VNA has been configured with a bandwidth of 200 MHz, and a central frequency of 2.45 GHz for obtaining the channel response in the ISM band of 2.45 GHz (2.4-2.5 GHz). Static measurements have been done with this setup, so that the Power Delay Profile (PDP) and Coherence Bandwidth (Bc) of the different wireless links have been obtained.

R2R-T2.2-T-DZR-006-18 Page 43 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 18: Base station antenna located inside the train

CAF facilities A wireless channel measurement campaign has been carried out on a metro train located at CAF facilities in Beasain, Spain. The goal was the characterization of the propagation environment where a wireless sensor network will be operating. The equipment used in this campaign was RUSK, where the measurements were performed at a centre frequency of 2.53 GHz with a signal bandwidth of 100 MHz, spanned by 161 subcarriers. The transmitted signal was repeated with a period of 1.6 microseconds. At each spatial position, a block of ten snapshots was recorded for enhancement of the measurement signal-to-noise ratio through coherent averaging.

The measurements were done in the interior of the train, where two different scenarios have been considered, according to the localization of the sensor nodes either on the bogie or inside the car. For both measurement groups, the localization of the access points (AP) or data recollections has been the same, where eight different locations (APx) have been taken into account for these measurements. The AP3 and AP4 were located under the car, whereas the rest of the APs were installed inside the car. Figure 19 shows the access point placements, where Sx points the bogie where all the sensor nodes were located (see Figure 20).

R2R-T2.2-T-DZR-006-18 Page 44 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 19: Access Point locations

A rail guide has been used in order to get several channel snapshots along the cross and inline orientation of the train, therefore it can be seen how the channel changes with small displacements due to the train vibration, and the relative movement between the car and the bogie. The propagation channel has been measured at regular spatial intervals of 1 millimeter along a distance of approximately 1 meter. The rail was moving at a speed of 5.2 centimeter/second, therefore the signal is recorded at regular time intervals of 19 milliseconds.

Figure 20: Sensor node locations under the car

In addition to the measurements done in a static placement and using a rail guide, passenger movements have also been taken into account. The sensor nodes and the access point were located in some specific locations, while several passengers start moving around the car. The measurement equipment registers the channel variations due to the passenger’s movement. The channel was measured at regular time interval of 1 millisecond during approximately 60 seconds, where the antenna height was 90 centimeter.

R2R-T2.2-T-DZR-006-18 Page 45 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 21: Sensor node located on gearbox (S2)

As stated above, the sensor nodes have been placed under and inside the car. Figure 20 depicts the location of six different node sensors on the bogie of the train, where Figure 21 shows an example of the antenna located on the gearbox. On the other hand, the location of another six sensor nodes inside the car is shown in the Figure 22.

Figure 22: Sensor node locations inside the car

3.5 POST PROCESSING, PARAMETER ESTIMATION AND MODELLING CAPABILITIES OF MEASUREMENTS

3.5.1 Channel Sounding Methodology A channel sounder is a system consisting of a transmitter and a receiver that is designed to measure the properties of the wireless propagation channel. One can distinguish channel sounders working in the time domain or in the frequency domain [85]. In the following description, we will describe the principle operation of a channel sounder, using the Medav RUSK channel sounder owned by the German Aerospace Center (DLR) as an example.

R2R-T2.2-T-DZR-006-18 Page 46 of 186 10/12/2016

Contract No. H2020 – 636032

PA BP waveform s(t) generation in baseband Wireless channel Rubidium local frequency oscillator normal BP LNA LP y(t) A AGC D storage

Rubidium local frequency oscillator normal

Figure 23: Principle of channel sounding In a practical measurement setup as depicted in Figure 23, the channel sounder transmitter generates a wide-band signal. The bandwidth of the transmit signal is limited by the sampling 1 frequency 푓s = . In order to avoid aliasing effects and to allow for a moderate roll-off of the 푇s transmit Nyquist filter 푔(푡) (cf. Figure 24), the discrete-time signal 푠[푛] is generated as a band- limited signal in an Orthogonal Frequency Division Multiplex (OFDM)-like fashion, i.e. a multi-tone ̃ ̃ ̃ ̃ signal [86]. First, the symbols 퐗 = [푋0, 푋1, … , 푋푁c−1] are generated with constant amplitude and a phase chosen such that the peak-to-average power ratio of the signal 푠[푛] is minimized. These 푁c < 푁 symbols are zero-padded to yield the vector ̃ 퐗 = [ퟎ(푵−푵퐜)/ퟐ, 퐗, ퟎ(푵−푵퐜)/ퟐ] (1)

where ퟎ푛 denotes a vector with 푛 zero entries. The vector 퐗 is transformed by an 푁-point inverse Fast Fourier Transformation (FFT) to yield 푥 = [푥0, 푥1, … , 푥푁−1] with

ퟐ흅 푵−ퟏ 풋 풏풎 풙풏 = ∑풎=ퟎ 푿풎풆 푵 (2) and this signal is parallel-serial converted to

풔[풏] = 풙[퐦퐨퐝(풏, 푵)] for ∈ ℤ . (3)

The continuous-time baseband signal in Figure 24 is then

∞ 풔(풕) = ∑풏=−∞ 풔[풏] ⋅ 품(풕 − 풏푻퐬) . (4)

Note that the same symbols 퐗̃, 퐗, 퐱 are transmitted in every FFT-block and therefore the signal 푠[푛] is periodic. We can therefore consider 푠[푛] as an OFDM-like signal, for which the cyclic prefix is as long as the FFT-size and hence the CIR may be as long as 푁푇s without causing inter-symbol interference. Since the discrete-time signal 푠[푛] is already bandlimited due to the zero subcarriers in 퐗, there are only mild requirements on the transmit filter 푔(푡).

R2R-T2.2-T-DZR-006-18 Page 47 of 186 10/12/2016

Contract No. H2020 – 636032

[0] x[0] X … x[1] P

… s[n] s(t) N- … g(t) IFFT S … x[N-1] X[N-1]

Figure 24: Waveform generation at transmitter side The maximum bandwidth of the transmit signal is strictly limited by the transmit filter, which operates at the sampling frequency 푓푠 = 1⁄푇s which for the available equipment is given by 푓s = 320 MHz [87]. However, since some bandwidth has to be reserved for the filter roll-off and the “OFDM-like signal” 푠[푛] is not strictly bandlimited, the effective bandwidth has to be selected significantly smaller. In the case of the Medav RUSK channel sounder owned by DLR, the maximum effective bandwidth is 120 MHz.

Given the sampling frequency of 푓s = 320 MHz, the sampling interval is 푇푠 = 3.125 ns. To obtain long OFDM-like symbols and hence to allow for a long CIR, 푁 = 8192 is selected, which leads to an OFDM-like symbol time of 푇OFDM = 푁푇s = 25.6 μs. This corresponds to a maximum excess length of a multipath component of 푑max = 푇OFDM ⋅ 푐0 = 7680 m. The subcarrier spacing is fixed by the choice of 푁 and 푓s and is given by

1 1 Δf = = = 39 063 Hz . (5) TOFDM NTs

Out of the 푁 = 8192 subcarriers, only 푁푐 = 3072 are non-zero, and these define the actual bandwidth of the transmit signal 푠(푡), which is given by 퐵 = 푁푐 ⋅ Δ푓 = 120 MHz.

3.5.2 Post Processing The data obtained from the measurement campaign is processed in order to get characteristic parameters of the wireless channel. This information is used to model its statistical behaviour in simulation / emulation, and improve the physical layer design of the wireless communication system. Let’s summarize the signal post-processing steps:

 Antenna and cable effect calibration. The response of the antenna and cables used during the measurement campaign must be suppressed from measured signal in order to get the channel response between the transmitter and receiver antenna. Figure 25 shows the radiation pattern of a particular antenna measured in a semi-anechoic chamber, whereas the Figure 26 shows the frequency response of the same antenna for a defined frequency band.

R2R-T2.2-T-DZR-006-18 Page 48 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 25: Radiation pattern of Mobile Mark MGRM-UMB-1C antenna at 2.45 GHz

Figure 26: Frequency response of Mobile Mark MGRM-UMB-1C antenna from 2 to 3 GHz  Small scale fading. The total channel gain power experienced by the received signal (see Figure 27) is composed of the following contributions: path loss, large-scale fading and small-scale fading.

퐺푑퐵(푑) = 푃푅푥,푑퐵 − 푃푇푥,푑퐵 = 퐺0,푑퐵 − 푛10 log(푑) + 푋푆푆퐹,푑퐵 + 푋퐿푆퐹,푑퐵 (6)

R2R-T2.2-T-DZR-006-18 Page 49 of 186 10/12/2016

Contract No. H2020 – 636032

where 푑 is the Tx to Rx distance, 퐺0,푑퐵 is the path power gain at the reference distance 푑0, and 푛 is the path loss exponent. 푋푆푆퐹,푑퐵 is the small-scale fading contribution (random signal variations due to multipath observed over one small scale area), and 푋퐿푆퐹,푑퐵 is the large-scale fading contribution (random variation in local average of received power observed over a spatial extent of multiple small scale areas.

200 -40

180 -50 160

140 -60 120

100 -70

80

Frequency samples Frequency -80 60

40 -90

20

-100 20 40 60 80 100 Spatial samples Figure 27: Channel power gain along the frequency and space

For small scale fading analysis, these contributions must be extracted for each measurement (see Figure 28), obtaining the normalized channel frequency response

푑푛 1 퐻푛표푟푚 = √ 퐻 (7) 퐺0 푋퐿푆퐹

where 퐻 is the measured channel frequency response, and 퐻푛표푟푚 is the small scale fading contribution.

R2R-T2.2-T-DZR-006-18 Page 50 of 186 10/12/2016

Contract No. H2020 – 636032

-43 Channel gain Pathloss estimation -44

-45

-46

-47

-48

Channel gain (dB) gain Channel -49

-50

-51

-52

10 20 30 40 50 60 70 80 90 100 Spatial samples Figure 28: Small scale fading extraction

 Small scale fading statistics. From small-scale fading contribution it can be obtained the envelope statistical distribution. The most common distributions are Rayleigh and Rice (see Figure 29).

H_norm 1.2 Rice

1

0.8

0.6 Density

0.4

0.2

0 0 0.5 1 1.5 2 2.5 Data Figure 29: Small scale envelope distribution

 Power delay profile. For each channel measurement in a small scale area we obtain an instantaneous power delay profile, which is defined as the square magnitude of the impulse response ℎ(휏)

R2R-T2.2-T-DZR-006-18 Page 51 of 186 10/12/2016

Contract No. H2020 – 636032

푃퐷푃(휏) = |ℎ(휏)|2 (8)

For each of the transmitter-receiver spatial combination, the total of PDPs are averaged the averaged PDP (see Figure 30) as 퐴푃퐷푃(휏) = 퐸{푃퐷푃(휏)} (9)

where 퐸{∙} denotes the expectation over the measurement points. 1.2

1

0.8

0.6 Power

0.4

0.2

0 -2 0 2 4 6 8 Delay samples Figure 30: Averaged power delay profile

 Delay spread. Based on the APDP we can define the rms delay spread, which is the second central moment of the APDP

2 ∫ 퐴푃퐷푃(휏)휏2푑휏 ∫ 퐴푃퐷푃(휏)휏푑휏 푆(휏) = √ − ( ) (10) ∫ 퐴푃퐷푃(휏)푑휏 ∫ 퐴푃퐷푃(휏)푑휏

while the coherence bandwidth is related to the inverse of the delay spread.

 Envelope autocovariance. The autocovariance of the envelope is commonly used to characterize the spatial selectivity of the channel, and therefore the possibility of using spatial diversity techniques, such as antenna diversity. The Figure 31 shows the autocovariance curve of a channel, which states that the decorrelation distance is nearly 30 spatial samples. In this example, each sample equals one millimeter, therefore it means that two channel responses separated this distance will be decorrelated.

R2R-T2.2-T-DZR-006-18 Page 52 of 186 10/12/2016

Contract No. H2020 – 636032

1

0.9

0.8

0.7

0.6

0.5

0.4 Autocovariance

0.3

0.2

0.1

0 0 20 40 60 80 100 120 Spatial samples Figure 31: Spatial autocovariance

 Dynamic analysis. The temporal variations of the channel due to movement of the wireless devices or the environment may cause signal fading (see Figure 32), mainly because the multipath effect.

Figure 32: Temporal power variations

The temporal correlation of small-scale fading, within a small-scale fading area, is a measure of how fast the channel changes in time, and is often characterized in terms of

R2R-T2.2-T-DZR-006-18 Page 53 of 186 10/12/2016

Contract No. H2020 – 636032

Doppler spectrum of the received signal (see Figure 33). The spectrum width is conventionally expressed in terms of the root-mean-square Doppler-spread.

Figure 33: Doppler spectrum

On the other side, coherence time is the time duration over which the channel response is considered to be not varying. This parameter can be obtained by means of the temporal correlation of the channel. Is it also useful to know that the Doppler maximum frequency is related to the inverse of the coherence time.

R2R-T2.2-T-DZR-006-18 Page 54 of 186 10/12/2016

Contract No. H2020 – 636032

4. MEASUREMENT SCENARIO DESCRIPTION

4.1 MEASUREMENT CAMPAIGN 1: HIGH-SPEED TRAINS IN ITALY The main objective for this measurement campaign is to characterize the channel for moving trains, both for communications within the same train (vehicle-to-vehicle), as well as for train-to- train communication. This is important for evaluating the interference coming from trains on adjacent tracks. At the same time, these measurements will be useful for future train-to-train communication. The measurements are planned to be conducted in a way that the outcome will not depend on the communication technology but will be general enough to derive channel models for more than one application.

4.1.1 Overview of Environments Terrestrial propagation is highly affected by the environment. The channel characteristic is one of the key points for radio communication, also in case of wireless railway communication. On a track from A to B, a train can pass different scenarios as mentioned in Table 6; combinations of those scenarios (see Table 7) are defining the environment for the propagation path. Most of the given scenarios are common for T2G as well as for T2T communications.

Table 6: Railway environment

Urban Area Sub-Urban Rural Curves Special Scenarios Tunnels Bridge / Viaduct Cuttings Cross Bridge Noise Barrier Catenary Obstacles Signalling System

Roof Building Vegetation (tree) Open field High density Railway Station Shunting Yard

R2R-T2.2-T-DZR-006-18 Page 55 of 186 10/12/2016

Contract No. H2020 – 636032

Table 7: Combination of possible environments (finally selected ones are marked red) Curve Bridge Buildings Vegetation Roof Catenary Noise Signaling Open Barrier System field

Urban x x x x x x x

Sub-Urban x x x x x x x

Rural x x x x x x x

Tunnel x x x

Viaduct x x x x x x x

Cutting x x x x x

Railway Station x x x x x x

Shunting Yard x x x x x x

From experience on past railway measurements, the following aspects should be considered:

 Pylons/masts that support the catenary can have a significant impact (all metallic and periodically installed). They are very close from the antennas and can be seen as reflectors with a high speed.  The tunnel case is specific but it can be described as an urban canyon with a roof. Indeed, with a rectangular cross-section, tunnel walls can be seen as buildings along the track.

4.1.2 Manoeuvres The possible movement of trains is listed in Table 8. Within the communication range of a T2T system, trains can drive these different manoeuvres depending on the amount of tracks.

Table 8: Combination of tracks and manoeuvre

# Tracks Flow Contra Flow Overtake Opposing Traffic Pass (v2=0) Coupling

1 x x x

Two 2 x x x x x x

3+ x x x x x x

R2R-T2.2-T-DZR-006-18 Page 56 of 186 10/12/2016

Contract No. H2020 – 636032

R2R

2 Trains 1 Train

Vehicle - flow opposing pass Vehicle

"virtual" Vehicle- Blank overtake coupled - Vehicle

var d catch up Front to End

var  escape

Figure 34: Manoeuvres

Flow of two trains In this scenario, two trains drive in the same direction on parallel tracks (cf. Figure 35). One manoeuvre could be set up as virtual coupled trains. The variables are the common velocity and the distance between the two trains. The relative velocity should be zero. The other possibility would be an overtake manoeuvre as described in Figure 36.

x -x 2 -x 1 0 x1 x2

acceleration Channel measurements deceleration

Figure 35: Flow traffic of two trains

R2R-T2.2-T-DZR-006-18 Page 57 of 186 10/12/2016

Contract No. H2020 – 636032

Both trains start in same direction

Starting measurement, trains

have reached the desired speed

Overtaking

Falling back

Repeat manoeuvre

x -x 2 -x 1 0 x1 x2

Figure 36: Timeline for overtake manoeuvre

Opposing of two trains The goal of the scenario opposing traffic is to measure two trains passing each other with high relative speed (see Figure 37). For this manoeuvre both trains are stopped at the same distance from the foreseen crossing point, at which they should cross with constant velocity in opposite directions, as depicted below in Figure 38.

R2R-T2.2-T-DZR-006-18 Page 58 of 186 10/12/2016

Contract No. H2020 – 636032

x -x 2 -x 1 0 x1 x2

acceleration Channel measurements acceleration

Figure 37: Opposing traffic of two trains

Both trains start in opposite directions

Trains have reached

desired speed

Crossing at desired location

Trains decelerate

Trains stop

x -x 2 -x 1 0 x 1 x 2

Figure 38: Time line for counter manoeuvre

R2R-T2.2-T-DZR-006-18 Page 59 of 186 10/12/2016

Contract No. H2020 – 636032

Both trains start to move at the same time and should reach at the points ±푥1 their defined velocity, which is then kept constant until passing the points ±1 on the other side of the crossing point at 푥 = 0. The maximum measurement distance is 10 km and for having a constant Doppler frequency during the measurement, both trains should keep their velocity constant in the interval [−푥1, 푥1]. With a measurement distance of 10 km, 푥1 = 5 km.

For the fast counter-direction manoeuvre, 푥2 = 12 km, while for slow one, 푥2 ≈ 5 km is sufficient.

One train pass another In this manoeuvre one moving train passes with a certain velocity another train standing still (yellow train in Figure 39 with v2=0).

x -x 2 -x 1 0 x 1 x 2

acceleration Channel measurements deceleration

Figure 39: One train pass a second train

Inter-Vehicle measurements The inter-vehicle measurements can scope 3 different placements of the antennas (red dots). For consist to consist measurements, the antennas can be placed as shown in Figure 40 a). In this way we meet the proposal of T2.5 and use directional antennas right on the edge of the vehicle, but the antenna cable is not mounted on the roof itself. Another possibility would be to mount the antennas on the bottom of the train next to the chassis.

The possibility to install antennas in the middle of the vehicle Figure 40 b) needs to be investigated concerning security aspects regarding the antenna feeding cable in case of a catenary emergency. To extend the distance between the antennas, a blank consist could be between the two “active” consists. In case of a Front-to-End channel characterization Figure 41 shows the possible measurement setup.

a) Antennas next to each other

R2R-T2.2-T-DZR-006-18 Page 60 of 186 10/12/2016

Contract No. H2020 – 636032

b) Antennas in the middle of the consist Figure 40: Consist-to-Consist (Bombardier)

Train End Vehicle Vehicle Train Front

Figure 41: Front to End measurement

For all measurements it is mandatory to install GPS antennas or use existing GPS antennas to track the measurements.

4.1.3 Scenarios Out of the favourite environments in Table 7 (red marked) and the manoeuvres shown in Figure 34 a list of the most interesting scenarios was established (Table 9). The measurements are weighted by 1 to 4 for each manoeuvre.

Table 9: Priorities for different scenarios Urban Sub-Urban Rural Tunnel Railway station Noise-Barrier Buildings Vegetation Open field Buildings Roof Flow 2 2 3 4 1 Opposing 2 2 3 4 1 Pass 2 3 4 1 1 Inter-Vehicle 2 4 3 1 1

1… highest 2… mid-high 3… mid-low 4… low priority

This results in a tight measurement schedule for 4 days (nights).

R2R-T2.2-T-DZR-006-18 Page 61 of 186 10/12/2016

Contract No. H2020 – 636032

4.1.4 Schedule

Day 1: One train – Inter-Vehicle measurements Different antennas (directional and omnidirectional) should be mounted in different positions (see Figure 41) on the train.

Priority:

1. Railway station 2. Urban 3. Tunnel

50 km/h vmax

Day 2: Two trains – Omni-directional antennas Priority:

1. Tunnel 2. Noise barrier 3. Vegetation 4. Open field Flow (virtually coupled) -> Opposing -> Opposing -> Flow (catch up, escape) Day 3: Two trains – Directional antennas The antennas will be mounted inside the train at the train driver cabin. This is only possible, if the radio characteristics of the front window are known. Therefore, a datasheet or measurements (TRI provides a window, DLR measure the radio characteristics) are necessary.

Priority: 1. Tunnel 2. Noise barrier 3. Vegetation 4. Open field Flow (virtually coupled) -> Opposing -> Flow (catch up, escape) Day 4: Two trains Additional measurements, e.g. to repeat some scenarios from previous days or modified scenarios based on lessons learnt.

R2R-T2.2-T-DZR-006-18 Page 62 of 186 10/12/2016

Contract No. H2020 – 636032

4.1.5 Measurement Equipment Channel Sounder

The Rusk channel sounder is the primary measuring equipment. This system consists of a transmitter and a receiver and is rather bulky, as can be seen in Figure 42. Both parts are mounted in a 19” rack of approximate dimension 60 cm x 90 cm x 135 cm and a weight of over 90 kg.

Transmitter and receiver have to be connected by a cable of at most 100 m length for synchronization before the measurements could be started (in the morning). They also have to be connected after the measurements (in the evening) to ensure that synchronization is not lost. Due to the dimension of the equipment, they first have to be placed in the trains and then they have to be connected for ca. 1 hour for synchronization.

IMU

UPS Power amplifier 24 V PTS 10 MHz Rubidium Clock R&S Generator Screen 10 MHz Rubidium Clock RFT 24 V DRU UPS

Figure 42: Receiver (left) and transmitter (right) of the channel sounding equipment

The racks for the channel sounder (see Figure 42) need to be placed as close as possible to the measurement antennas to minimize the antenna cable length. The racks need to be fixed in the train with tension belts. The actual position depends on the possibilities of fixing, the power supply and the antenna cables. The channel sounder is represented as cubes (TX in yellow, RX in orange) in the Figure 46, Figure 47 and Figure 48.

R2R-T2.2-T-DZR-006-18 Page 63 of 186 10/12/2016

Contract No. H2020 – 636032

Antennas

For the measurements following antennas are used:

a) Huber+Suhner SWA-0859/360/4/0/DFRX30_2 This omni-directional antenna fits best in case of RF-characteristics and is used for the major measurements. It was already partially installed on the train roof (cf. Figure 43). Further installations were necessary.

 Frequency bands: GPS, GSM, UMTS, LTE, Wi-Fi and WiMAX  Usage: 5.2 GHz, 3.5 GHz, GPS  Red dot ( active, inactive), used on day 1 and 2

Figure 43: Omni-directional (red marked) antenna on train 28, coach 3

b) Huber+Suhner SPA-2456/75/9/0/DF_1 This directional antenna is used for measurements in front of the train. Therefore the antenna is installed in the nose of the train as shown in Figure 44 bottom. When the lid of the nose is closed, the antenna (marked with a red ring) is in an upright position right in front of the coupler.

 Frequency band: Wi-Fi  Usage: 5.2 GHz  Red arrow ( ), only used on day 3

R2R-T2.2-T-DZR-006-18 Page 64 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 44: Installation of the directional antenna

c) Huber+Suhner SWA-0459/360/4/25/DFRX30 This multiband omni-directional antenna will be used for subsystems. It would also offer the measurement frequencies, but the quality of the radiation pattern is not fitting for measurements. The antenna is mounted on the top of the locomotive above the driver’s cabin (see Figure 45, black antenna).

 Frequency bands: Tetra, DVB-T, GPS, GSM, UMTS, LTE, Wi-Fi and WiMAX  Usage: Tetra, GPS  Violet square ( ),used for every measurement

R2R-T2.2-T-DZR-006-18 Page 65 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 45: Tetra antenna on locomotive

d) TEDAP TDBLD Triband GSM-DCS-GPS Antenna This antenna is fixed installed on each train and used for train communications and GPS information. For the measurement the GPS signal is used if needed.

 Frequency bands: GSM, DCS, GPS  Usage: GPS  Green star ( ), will be used for every measurement

Additional sensors Several additional sensors were mounted in the train: a) Next to the channel sounder or on the floor of the locomotive an IMU was placed to record the acceleration and turn rate of the train. In Figure 42 the IMU is placed on the top of the channel sounder for illustration.

b) For a detailed distance recording between the trains a distance-radar was mounted in the nose of the train next to the directional antennas.

Installation

For the different measurement scenarios different antennas is fixed mounted on the train and stay there after finishing the measurements. All the other equipment is replaced afterwards. The blue lines represent the antenna cables.

a) Day 1 Vehicle-to-Vehicle measurement For the V2V measurement 3 SWA-0859 antennas will be used (see Figure 46). The TX- channel sounder rack and antenna are placed in/on car 3. One receiving antenna is placed on car 2, in a distance of approximately 4 m from TX-antenna and the second RX-antenna is on car 1 approximately 30 m away from TX. The sounder rack is placed as close as possible to the antenna in car 1 due to link budget reasons. For measurements, the two RX-antennas are time-multiplexed which leads to a slightly longer measurement period time. The advantage of this method is that one environment is measured with both antenna settings at one run with a negligible time shift in comparison to the slowly changing environment.

R2R-T2.2-T-DZR-006-18 Page 66 of 186 10/12/2016

Contract No. H2020 – 636032

Loc A 1 2 3 10 11 Loc B

Figure 46: Vehicle-to-Vehicle setup

b) Day 2 Train-to-Train omni-directional 2 SWA-0859 antennas is used, one antenna on each train as shown in Figure 47. The location is the same on both trains and as for the measurement on day 1 (V2V).

Loc A 1 2 3 10 11 Loc B

Loc A 1 2 3 10 11 Loc B

Figure 47: Train-to-Train with omni-directional antennas

c) Day 3 Train-to-Train directional For this setup, one SPA-2456 antenna is installed on each train in front of the locomotive. Two positions are possible:

 Behind the front nose at the bottom of the locomotive. This front fairing is made of plastic and offers enough space behind. Advantage would be the position itself and the plastic in the beam direction of the antenna. A higher effort of installation needs to be considered.  Behind the front glass of the drivers cab. In this way, the antenna would be mounted inside the drivers cab. An advantage would be the easy installation. The biggest disadvantage is the unknown RF-specification of the window. This could ruin the whole measurement. As GPS reference the SWA-0459 antenna should be mounted on top in front of the locomotive.

The rack needs to be placed in the coach next to the locomotive as shown in Figure 48, because next to the antenna and in the drivers cab there is no space for the measurement equipment.

Loc A 1 2 3 10 11 Loc B

Loc A 1 2 3 10 11 Loc B

Figure 48: Train-to-Train with directional antennas

R2R-T2.2-T-DZR-006-18 Page 67 of 186 10/12/2016

Contract No. H2020 – 636032

4.1.6 Trajectory of Measurements The measurements take place on the track between the station Roma Termini and Napoli Centrale as shown with the red line in Figure 49. On this track all interesting scenarios are covered in different sections. The track is doubled tracked and trains are operating with speeds up to 300 km/h.

Figure 49: Track map The train Frecciarossa ETR 500 (see Figure 50) is used for measurements. These trains are composed of 11 couches for a total of 574 seats:

 Coach 1: Executive  Coach 2-3-4: Business  Coach 5 : Business + Bistro  Coach 6: Premium  Coach 7-8-9-10-11: Standard

Each coach is 26 m and the locomotives at each side are 22 m long.

R2R-T2.2-T-DZR-006-18 Page 68 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 50: Frecciarossa ETR 500

4.1.7 Documentation

Measurement data The raw data is recorded on each device itself and is downloaded after each run by the corresponding staff. A copy of the files and one backup copy is stored separately.

Protocols Every procedure should follow prescribed protocols. Incidents and additional steps need to be recorded.

Audio-Visual The environment during the measurement runs should be recorded with video systems. Due to the night conditions it is difficult to record the surroundings of the track.

All the processes of preliminary work (e.g. installations), the measurement itself and parts of the post processing should be documented with videos and pictures. A documentation video would be useful for final reporting and promotion. This video can be found in: https://www.youtube.com/watch?v=ndzAO-PuNTY&t=48s.

R2R-T2.2-T-DZR-006-18 Page 69 of 186 10/12/2016

Contract No. H2020 – 636032

4.1.8 Detailed Schedule for Measurements Table 10: Time table for all four days Time Location Description

21:00 – 24:00 Depot in Naples Loading of measurement equipment into trains, set-up of equipment, calibration of transmitter and receiver

0:00 – 5:00 Railway track between Channel measurements Naples and Rome

5:00 – 7:00 Depot in Naples Second calibration of transmitter and receiver, unloading of measurement equipment

Table 11: Basic data Distance Rome – Naples 220 km

Travelling time at maximum speed 67 min

Acceleration (moderate) 0.5 m/s2

Acceleration (fast) 1 m/s2

Distance and time to reach a speed of 300 km/h with moderate acceleration:

푣2 푑 = ≈ 7 푘푚 푎푐,1 2푎

푇ac,1 ≈ 3 min

The distance to reach a speed of 30 km/h is, again with moderate acceleration,

푣2 푑 = ≈ 70 m ac,2 2푎

푇ac,2 ≈ 17 s

R2R-T2.2-T-DZR-006-18 Page 70 of 186 10/12/2016

Contract No. H2020 – 636032

Day 1 One train, the transmitter and receiver are located in the same train but in different cars.

Table 12: Manoeuvres day 1 Velocity Duration Description

1 v1 = 50 km/h 4 h Travel with constant speed from Naples in direction to Rome till PK 36, covers all possible scenarios

2 Maximum 45 min Travel back to depot at maximum allowed speed

Day 2, 3 and 4 Two trains, the transmitter and receiver are located in different trains. The measurements are done on

 Day 2 with directional antennas,  Day 3 + 4 with omni-directional antennas,

Table 13: Manoeuvres Day 2 Velocity Duration Surrounding Description

1 v1 = 50 km/h 2h 40 min Urban, Passing, same direction, slow sub-urban, Train 1 drives at constant speed, while train 2 v2 = 40, …, tunnels, slowly catches up train 1 until same height, then 60 km/h open field, train 2 slows down until it is 7 km behind train 1 forest This manoeuvre is repeated several times till PK 75

2 v1 = 300 km/h 20 min Forest Counter-direction, fast Train 1 stops at PK 75, Train 2 at PK 36, Train 2 v 2 = -300 km/h changes direction and then both trains accelerate to max speed, crossing at ~ PK 61

3 v1 = 270 km/h 35 min Passing, same direction, fast Train 1 changes direction at PK 36, accelerates to v2 = 240, …, max speed, Train 2 starts at PK 61 and accelerates 300 km/h to 250 km/h

4 v1 = 30 km/h 25 min Urban, Counter-direction, slow sub-urban Train 2 stops at PK 215 and changes direction, v 2 = - 30 km/h Train 1 stops at PK 203, both trains accelerate to 30 km/h, crossing at ~ PK 209

R2R-T2.2-T-DZR-006-18 Page 71 of 186 10/12/2016

Contract No. H2020 – 636032

Table 14: Manoeuvres Day 3 Velocity Duration Surrounding Description

1 v1 = 50 km/h 2h Urban, Passing, same direction, slow sub-urban, Train 1 drives at constant speed, while train 2 v2 = 40, …, tunnels, slowly overtakes train 1 until 5 km, then train 2 60 km/h open field, slows down until it is 5 km behind train 1 forest This manoeuvre is repeated several times till PK 75

2 v1 = 300 km/h 8 min Forest Counter-direction, fast Train 1 stops at PK 75, Train 2 at PK 36, Train 2 v 2 = -300 km/h changes direction and then both trains accelerate to max speed, crossing at ~ PK 61

3 v1 = 270 km/h 40 min Passing, same direction, fast Train 1 changes direction at PK 36, accelerates to v2 = 240, …, max speed, Train 2 starts at PK 61 and accelerates 300 km/h to 250 km/h

Table 15: Manoeuvres Day 4 Velocity Duration Surrounding Description

1 v1 = 50 km/h 2h Urban, Passing, same direction, slow sub-urban, Train 1 drives at constant speed, while train 2 v2 = 40, …, tunnels, slowly overtakes train 1 until 5 km, then train 2 60 km/h open field, slows down until it is 5 km behind train 1 forest This manoeuvre is repeated several times till PK 75

2 v1 = 270 km/h 40 min Passing, same direction, fast Train 1 changes direction at PK 36, accelerates to v2 = 240, …, max speed, Train 2 starts at PK 61 and accelerates 300 km/h to 250 km/h

3 v1 = 30 km/h 21 min Urban, Counter-direction, slow sub-urban Train 2 stops at PK 215 and changes direction, v 2 = - 30 km/h Train 1 stops at PK 203, both trains accelerate to 30 km/h, crossing at ~ PK 209

R2R-T2.2-T-DZR-006-18 Page 72 of 186 10/12/2016

Contract No. H2020 – 636032

4.2 MEASUREMENT CAMPAIGN 2: METRO IN MADRID The main objective of this measurement campaign is to characterize the channel in underground trains, which is a completely different scenario compared to both regular and high speed trains. The measurements will be conducted in such a way that the outcome will not depend on the communication technology but will be general enough to derive channel models for more than one application. The outcome of the measurements will be delay spread and transfer functions for each of the measured wireless links. The measurements are scheduled for 30th May - 3rd June 2016.

A representative subset of wireless links and scenarios will be selected taking into account the inputs coming from Tasks T2.1 (requirements), T2.5 (architecture), and the documentation provided by Metro de Madrid for Task T2.2. They are described in the following sections.

4.2.1 Measurement Procedure The following measurement equipment will be used: a channel sounder for channel characterization, and a Vector Network Analyzer (VNA) with horn antennas for material characterization. The channel sounder is depicted in the next figure, and its main features are summarized in Table 16.

Figure 51 : Channel Sounder

R2R-T2.2-T-DZR-006-18 Page 73 of 186 10/12/2016

Contract No. H2020 – 636032

Table 16: Features of Channel Sounder Transmitter Frequency range 500-6010 (4 bands) MHz Output power 42 dBm IF bandwidth 1/10/30/100 MHz Modulation Pulse/external 10 ns/ LTE Receiver Frequency range 400-7000 (4 bands) MHz IF dual conversion 860/160 MHz Noise figure 3 dB IF bandwidth 5/10/20/100 MHz Demodulation Logarithmic detector /LTE Dynamic range 90 dB

For the channel sounder, the following configuration will be used in the measurements:

 Frequency: 5.2 GHz  Bandwidth: 100 MHz  Output power: 42 dBm For each measurement, two antennas will be connected to the channel sounder:

a) At one end, a wideband monopole antenna (Mobile Mark MGRM-WHF) will be placed (see Figure 52). This is a magnetic-mount antenna with 3-meter coaxial cable and SMA connector.

Figure 52: Mobile Mark MGRM-WHF Wideband Monopole Antenna b) At the other end, a 12-monopole array will be used (Figure 53).

Figure 53: Monopole Array antenna

R2R-T2.2-T-DZR-006-18 Page 74 of 186 10/12/2016

Contract No. H2020 – 636032

This antenna array allows measuring the delay profile for different directions. This way the angle of arrival of each path can be found, and the Doppler spectrum that fits best can be estimated (e.g. Jakes, flat, asymmetrical Jakes, etc).

Once the antennas are placed on the train, the train will be moved to different environments (tunnels, stations, open air), and 20-30 different delay profiles will be measured in each environment moving the train at approximately 1 km/h. Each of these delay profiles will consist of 12 measurements taken with the antenna array. On the other hand, in order to check the consistency of the results, for each type of environment the measurements will be repeated in 3 different locations of the same kind.

4.2.2 Measurement Schedule The measurement plan is summarized in Table 17, and detailed in the next sections. The measurements will be divided in an afternoon shift (15h to 20h) and a night shift (1h to 6h), where the measurements will be done in mainline from 2h to 5h.

Additionally, during the measurement campaign audio-visual material will be generated, pictures and videos. A final video has been released and it is public in the following link: https://www.youtube.com/watch?v=spSeptrzbQg&t=1s.

R2R-T2.2-T-DZR-006-18 Page 75 of 186 10/12/2016

Contract No. H2020 – 636032

Table 17: Measurement plan

Number of Locations Date Time Link Location Train Type of Environment per Environment Equipment MONDAY 30/05/2016 Afternoon (15h-20h) Inter-consist (preparation) Cuatro Vientos Depot Continuous (S8000) - - Night (1h-6h) / Measurements: 2h-5h Inter-consist (measurements - line) Cuatro Vientos Depot -> Line 10 or Line 11 Tunnel (wide) - Straight 3 Channel Sounder 2 1 Tunnel (wide) - Curve 3 Tunnel (narrow) - Straight 3 R T Tunnel (narrow) - Curve 3 Station (pit-shaped) 3 Station (tunnel-shaped) 3 Open Air 3 TUESDAY 31/05/2016 Afternoon (15h-20h) Inter-consist (measurements - depot) Cuatro Vientos Depot Continuous (S8000) Multipath from nearby train Channel Sounder Train-to-Train (preparation) Cuatro Vientos Depot - - Night (1h-6h) / Measurements: 2h-5h Train-to-Train (measurements - line) Cuatro Vientos Depot -> Line 10 or Line 11 Tunnel (wide) - Straight 3 Channel Sounder 1 Station (pit-shaped) 3

T Open Air 3

2

3 R

WEDNESDAY 01/06/2016 Afternoon (15h-20h) Intra-consist (preparation) Cuatro Vientos Depot Continuous (S8000) - - Night (1h-6h) / Measurements: 2h-5h Intra-consist (mesurements - line) Cuatro Vientos Depot -> Line 10 or Line 11 Tunnel (wide) - Straight 3 Channel Sounder Tunnel (narrow) - Straight 3 1 2

T R Station (pit-shaped) 3 Open Air 3 THURSDAY 02/06/2016 Afternoon (15h-20h) Intra-consist (preparation) Cuatro Vientos Depot -> Laguna Depot Non-Continuous (S5000) - - Night (1h-6h) / Measurements: 2h-5h Intra-consist (mesurements - line) Laguna Depot -> Line 6 Tunnel (wide) - Straight 3 Channel Sounder Tunnel (narrow) - Straight 3 1 2

T R Station (pit-shaped) 3 Open Air 3 FRIDAY 03/06/2016 Afternoon (15h-20h) Glass characterization Laguna Depot VNA+Horn Antennas Night (1h-6h) / Measurements: 2h-5h Backup night for pending measurements

R2R-T2.2-T-DZR-006-18 Page 76 of 186 18/11/2016

Contract No. H2020 – 636032

Monday 30th May The measurements on Monday will focus on Inter-Consist measurements on wide gauge continuous train (Series 8000). The preparation of the channel sounder and the antennas will be done in the afternoon at Cuatro Vientos depot. The following picture shows the location of transmitter and receiver antennas on the continuous train.

55.04 m

18.1 m

2 1

R T

2.8 m

T TX Unit

R RX Unit Inter-ConsistLink

n TransmitterAntenna RF Cable TX 2.43 m

m Receiver Antenna RF Cable RX

Figure 54: Inter-Consist measurements

The measurements will be done at night on Line 10 or Line 11, depending on availability; Line 10 has diverse environments, and includes tunnel, station and open air, while Line 11 includes only tunnel and station. The following environments will be measured:

Table 18: Inter-Consist measurement environments Type of Environment Number of Locations per Environment

Tunnel (wide) - Straight 3 Tunnel (wide) - Curve 3 Tunnel (narrow) - Straight 3 Tunnel (narrow) - Curve 3

Station (pit-shaped) 3

Station (tunnel-shaped) 3

Open Air 3

R2R-T2.2-T-DZR-006-18 Page 77 of 186 10/12/2016

Contract No. H2020 – 636032

Tuesday 31st May On Tuesday afternoon, first the effect of multipath in inter-consist links due to train-crossing will be measured at Cuatro Vientos depot. After that, the Train-to-Train measurements will be prepared. The aim of the train-to-train scenario is to simulate either the interference of adjacent trains that use the same wireless TCMS technology, or a vehicle-to-vehicle communication. The following picture shows the location of transmitter and receiver antennas for these measurements:

1

Train 1 T

2

Train 2 3 R

T TX Unit

R RX Unit Train-to-Train Link

n TransmitterAntenna RF Cable TX m Receiver Antenna RF Cable RX Figure 55: Train-to-Train Measurements

The measurements will be done at night on Line 10 or Line 11, at the following environments:

Table 19: Train-to-Train measurement environments

Number of Locations Type of Environment per Environment Tunnel (wide) - Straight 3 Station (pit-shaped) 3 Open Air 3

R2R-T2.2-T-DZR-006-18 Page 78 of 186 10/12/2016

Contract No. H2020 – 636032

Wednesday 1st June This day will focus on Intra-Consist measurements. The preparation of the setup will be done in the afternoon. The following picture shows the location of transmitter and receiver antennas.

55.04 m

18.1 m

1 2

T R

2.8 m

T TX Unit Intra-ConsistLink R RX Unit

n TransmitterAntenna RF Cable TX 2.43 m

m Receiver Antenna RF Cable RX

Figure 56: Intra-Consist measurements on Continuous train

The measurements will be done at night in Line 10 or Line 11, at the following environments:

Table 20: Intra-Consist measurement environments

Number of Locations Type of Environment per Environment Tunnel (wide) - Straight 3 Tunnel (narrow) - Straight 3 Station (pit-shaped) 3 Open Air 3

R2R-T2.2-T-DZR-006-18 Page 79 of 186 10/12/2016

Contract No. H2020 – 636032

Thursday 2nd June The measurements on Thursday will be focused on Intra-Consist links in wide gauge non- continuous trains (Series 5000), in order to check the impact of the separation between cars in this type of links. The preparation will be done in the afternoon: first, both transmitter and receiver units of the channel sounder will be mounted on a continuous train in Cuatro Vientos depot, and it will be moved to Laguna depot; once there, the channel sounder and the antennas will be placed on a non-continuous train for intra-consist measurements. The following picture shows the location of transmitter and receiver antennas.

36.02 m

17.4 m

1 2

T R

2.8 m

T RUSK TX Unit Intra-ConsistLink R RUSK RX Unit

n TransmitterAntenna RF Cable TX 2.42 m

m Receiver Antenna RF Cable RX

Figure 57: Intra-Consist measurements on Non-Continuous train

The measurements will be done at night in Line 6, at the following locations:

Table 21: Intra-Consist measurement environments

Number of Locations Type of Environment per Environment Tunnel (wide) - Straight 3 Tunnel (narrow) - Straight 3 Station (pit-shaped) 3 Open Air 3

Friday 3rd June On Friday afternoon, the measurement equipment will be dismounted and packed. In parallel, glass characterization measurements will be done with Vector Network Analyzer (VNA) and horn antennas.

Friday night has also been planned as backup, in case delays from previous nights do not allow finalizing some of the measurements.

R2R-T2.2-T-DZR-006-18 Page 80 of 186 10/12/2016

Contract No. H2020 – 636032

4.3 COMPLEMENTARY MEASUREMENT CAMPAIGN: 60 GHZ FOR T2T COMMUNICATIONS

4.3.1 Motivation for the 60 GHz channel measurement Millimeter-wave (mmw) radio technology is considered to be suitable for such short-coverage radio links with high data-rate due to high propagation loss and large available bandwidth. ITU-R recommends four carrier frequencies around 60 GHz including 58.32 GHz, 60.48 GHz, 62.64 GHz and 64.80 GHz with a bandwidth of 2.16 GHz, which promotes the corresponding standardization like IEEE 802.15.3c and IEEE 802.11ad aiming at high data-rate within short ranges. Besides, mmw technology regarded as one of 5G potential candidates has attracted plenty of interest from both academic and industrial sectors. It is certainly worthwhile to explore the radio channel properties at 60 GHz for the metro and high-speed train environments, which fits for both application requirements and industrial interests.

4.3.2 Available Measurement Equipment

Channel sounder IFSTTAR will use the VµBIQ equipment described here after. This equipment allows to obtain information on the channel at 60 GHz in a SISO configuration.

Table 22: VµBIQ development system at 60 GHz with 2 GHz bandwidth Number of equipment frequency BW equipement

Vbiq Mmw RF module 64.5 GHz Up to 1.2 GHz 1

Tektronix Scope TDS6124 C 12 GHz 40 Gs 1

Mmw antenna 60 GHz 2

Tektronix AWG 7102 10 GS/s 1

Figure 58: Photos of the mmw measurement setup – VµBIQ

R2R-T2.2-T-DZR-006-18 Page 81 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 59: mmw measurement setup – VµBIQ

Vbiq Tx module Vµbiq Rx module

AWG Oscilloscope

Figure 60: Scheme of the equipment The channel sounder consists of a transmitter sending a known signal, and a receiver which analyses the received signal, and stores information to files. The supported bandwidth is up to 1.2 GHz, while RF modules are built to be operated at 60 GHz. The maximum measurable delay is 1.33 s. The channel sounder is capable of conducting measurement of power and delay domain. The transmitter and the receiver are synchronized by using a RF cable to set a Marker. The channel sounder is modular and flexible, and can easily be configured for different measurement scenarios. The detailed description will be given in the following content.

Figure 61: Characteristics of the mmw antenna

R2R-T2.2-T-DZR-006-18 Page 82 of 186 10/12/2016

Contract No. H2020 – 636032

Two identical horn antennas with 15 dBi gain are utilized in the system. The antenna beam width is 22o, which is the same as our motion control sweep step. The detailed information can be found in the above Figure 61.

Signal waveform The default transmitter signal is a repetitive chirp signal, where the amplitude and the phase of signal are continuous at the transition from one chirp to another. The phase continuity is achieved by rounding the wanted chirp bandwidth and center frequency to give a whole number of signal periods during the duration of one chirp. Input to the chirp generation routine is:

 fs - Sample frequency

 fc - Centre frequency

 B - Chirp bandwidth

 A - Amplitude

 N - Number of samples

The chirp is defined by:

2 i  2cos    c  1mod  NiikikAc  1,....,1,0, (11)

Where:

 B  f c   2 N  5.0   f   s  k  N (12) and

 B   N  5.0  2 f s  kc  2 N (13)

  is the floor function. The intention of the modulo 1 -function in (11) is to avoid taking modulo 2 of large numbers in the cosine function, since modulo 1 can be implemented by using the floor function. The chirp is generated directly on Arbitrary Wave Generator (AWG), which could be any frequency as long as the chirp fits within half the sampling frequency. In our case, the AWG sampling rate is set to be 2.5 Gs/s, which is more than two times of bandwidth (1.2 GHz). The maximum samples supported by the AWG in one chirp duration are 3333 due to the memory size, which results in the maximum delay time of 1.33 µs.

R2R-T2.2-T-DZR-006-18 Page 83 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 62: Real and imaginary part of the chirp signal

Data recording system The transmitter (Tx) keeps on sending the chirp signal continuous. The main challenge is to record these measurement data from scope to disk, even though the scope can demonstrate the received signal on the screen. GPIB port introduced by Tektronix is used to record the data from scope to laptop disk by using IO control program and Matlab. The scope setups are shown in the following Table 23:

Table 23: Characteristics of the recording system 10 div per measurement 1 µs/div

Sample rate 5 GHz

Number of continuous chirps 7

10 s measurement continuous data can be transferred to the laptop local disk at one time due to the memory size limitation, which includes 7 continuous chirps. The next 7 chirps will be recorded in the next row of our file. 20 rows will be written in one file and the next file will be written repeatedly. All the chirp data has been converted to channel impulse response (CIR) after match filtering before restored.

The Scope setups are also recorded for each measurement campaign, which can be for the measurement campaign description and data analysis.

Table 24: Characteristics of the scope set up AcquisitionStartDelay 0

AcquistionTime 1.000000e-05

Driver tektronix

R2R-T2.2-T-DZR-006-18 Page 84 of 186 10/12/2016

Contract No. H2020 – 636032

DriverDetectionMode auto

Resource GPIB0::27::INSTR

SingleSweepMode off

Status open

Timeout 10

TriggerMode auto

TriggerLevel 5.000000e-01

TriggerSlope rising

TriggerSource CH4

WaveformLength 100000

Back-to-back measurement In order to verify the system accuracy and calibrate the absolute power level, a Back-to-Back (B2B) measurement is conducted. First, we connect the AWG and Oscilloscope directly to verify the baseband system and chirp signal (Figure 63). The complex impulse response measurement result is demonstrated in the Figure 64, from which it can be found that the chirp signal shows a good autocorrelation property and the signal-to-noise ratio (SNR) is over 50 dB.

Figure 63: Back-to-Back measurement setup without RF module

R2R-T2.2-T-DZR-006-18 Page 85 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 64: Channel Impulse Response from B2B measurement without RF module The next step is to verify our measurement solution with RF module designed by Vµbiq. The Tx and Rx modules are connected directly with a waveguide attenuator (Figure 65 (a)) and its attenuation loss is measured to be 31 dB by using a spectrum analyzer shown in in Figure 65 (b).

Figure 65: (a) waveguide attenuator (b) attenuation loss measurement

IFSTTAR is trying to provide the following equipment for channel measurements. The availability will depend on the budget needed to rent the equipment.

For 60 GHz, in order to obtain accurate channel information like received signal level, power-delay profile and Doppler spectrum, a dedicated mmw channel sounder comprising a transmitter (TX) and a receiver (RX) is needed, which can "sound" the channel by sending predefined signals and obtain the channel properties based on the post signal processing on the received signal. The sounder which was designed to operate at 60 GHz and emits a repetitive frequency chirp signal with 368 MHz bandwidth, which leads to a delay resolution of about 3 ns. The maximum transmit power is 16dBm and the antenna gain is 7.5 dBi, yielding an EIRP of 23.5 dBm. The antenna is integrated with the transceiver (AiP) and the opening angle is in the order of 60 degrees. There is a tradeoff between the maximum Doppler band and the maximum delay spread, which can be configured due to the requirements of the scenarios. Since this sounder is still under development, the current working mode is SISO and synchronization system is realized by using a 25 m connecting cable. Therefore, the maximum distance between transmitter and receiver is about 25 meters. However, both MIMO and synchronization system can be updated based on customer's request. The channel sounder’s parameters are summarized in details:

R2R-T2.2-T-DZR-006-18 Page 86 of 186 10/12/2016

Contract No. H2020 – 636032

Table 25: Channel sounder's parameters Carrier frequency Tunable between [57, 64] GHz

Bandwidth 368 MHz

Transmitting power 16 dBm

integrated antenna with 7.5 dBi antenna gain, opening angle is antenna info in order of 60 degrees

Delay resolution About 3 ns

Maximum delay span 2.8 microsec

Maximum Doppler shift span In the order of 1-3 kHz

Maximum MIMO antenna element 2×2

Figure 66: mmw measurement setup

4.3.3 Measurement Scenario Description It needs to be pointed out that the mmw radio technology is feasible only for the short-range radio link due to the high propagation loss. Besides, the mmw sounder system with integrated antennas is fragile. By taking the feasibility into consideration, the related measurement campaigns should be performed for short-distances (maximum 25 m) with limited mobility. As a result, the following scenarios are proposed.

R2R-T2.2-T-DZR-006-18 Page 87 of 186 10/12/2016

Contract No. H2020 – 636032

4.3.4 Intra-Vehicle Communication (Metro and HSR) Links: This scenario will consist of measuring the wireless link between different elements inside a single vehicle. These will be Line-Of-Sight (LOS) links, mostly affected by the internal structure of the vehicle.

Figure 67: Intra-vehicle links

Measurement locations: It can be measured on depot.

Variations: 1. Person presence (a single metro car can handle up to 200 people; for measurements, a group of 10-20 people will be available). 2. Person movement. 3. Train gauge (narrow and wide), as well as vehicles with/without corridor between (i.e. continuous new trains, and non-continuous old trains). 4. Impact of material properties (this could be measured on old trains – it will be put as a low priority action).

4.3.5 Intra-Consist Communication (Metro and HSR) Links: This scenario will include wireless link that go beyond a single vehicle, involving both the next vehicle and the exterior of the vehicle (both roof and bogie).

Figure 68: Intra-Consist links

R2R-T2.2-T-DZR-006-18 Page 88 of 186 10/12/2016

Contract No. H2020 – 636032

Measurement locations: It would be feasible to keep the vehicle still for different propagation environments (different kinds of tunnels, stations, as well as outside – open field). Although different environments can affect the communication link, especially for vehicles without a corridor between cars, it could be difficult to mount the sounder outside the train due to integrated antenna systems. However, if a firmed and secured mounting solution can be proposed, it may be feasible for the measurement campaigns with mobility.

Note: additional measurements will be made on roof and bogie in order to characterize EMI noise coming from catenary and engine (DC old engines and AC new engines). These measurements will be done with a wideband antenna and a spectrum analyzer, and will be used as an input for task T2.7.

Variations: 1. Vehicles with/without corridor between them. 2. Vehicles with steel/aluminium frames. 3. Person movement 4. Train gauge 5. Propagation scenario: o Bends will be available on depot o Open air scenarios will be available on a test track (To be confirmed) o Several types of stations and tunnels can be available on a test track (To be confirmed) 4.3.6 Inter-Consist Communication (Metro and HSR) Links: It will include wireless links that go from one consist to another one, locating the antennas on the roof of the train (both at the centre and at the edges – see Figure 69).

Figure 69: Inter-Consist links

If mm-wave measurement equipment is available, the channel at mm-waves will be characterized for the edge configuration.

Measurement locations: The mmw channel sounding is suggested to be performed when the vehicle is still. However, if a firmed and secured mounting solution can be proposed, it may be feasible for the measurement campaigns with mobility.

R2R-T2.2-T-DZR-006-18 Page 89 of 186 10/12/2016

Contract No. H2020 – 636032

5. MEASUREMENTS IN THE FIELD This chapter reports the actual carried out measurements and deviations from the planned measurements for the high speed line in Section 5.1, for urban/Metro line in Section 5.2, and for the 60 GHz mmWave radio technology on regional trains in Section 5.3. The planned measurements are detailed correspondingly in Sections 4.1, 4.2 and 4.3.

R2R-T2.2-T-DZR-006-18 Page 90 of 186 10/12/2016

Contract No. H2020 – 636032

5.1 HIGH SPEED LINE – AC (TRENITALIA)

5.1.1 Night 1 Inter-Vehicle Measurement

One train, the transmitter and receiver of the channel sounder were located in the same train but on different cars (see Figure 46). The installation of the equipment had to be done at Napoli Centrale. At 23:15 Train 28 arrived at the station on platform 18. At 23:30 the measurement start was shifted to 00:45. Due to the delay, phase 1 was shortened to PK 60. From PK 60 to PK 36 the speed was 100 km/h. Phase 2 was done as planned. Within these measurements a multiplexer was used. Two antennas were measured at the receiving side in parallel, i.e. measurements were performed in Single-Input Multiple-Output (SIMO) mode.

Figure 70: Measurement setup for inter-vehicle measurements

Please note: The power amplifier with 37 dBm has been used!

General configuration: Bandwidth 120 MHz Center frequency 5.2 GHz Excess delay 12.8 μs Antennas MUX 1 => 2 1 => 5 Time grid 2.048 ms TX power 37 dBm (EIRP  20 dBm) Antennas omni-directional (8 dBi gain) TX, car 1 RX1 => MUX channel 2, car 2 RX2 => MUX channel 5, car 3

R2R-T2.2-T-DZR-006-18 Page 91 of 186 10/12/2016

Contract No. H2020 – 636032

Intra-Consist Measurements

The intra-consist measurements were done inside the train from one coach to another with the ITS-G5 system (cf. Figure 71). In this case, the intra-consist and intra vehicle measurements could be performed in parallel to the inter-vehicle measurements. A data set of 1.5 hours was collected.

Intra-Vehicle Measurements

The smallest measured distance with ITS-G5 was within one vehicle. This measurement represents an intra-vehicle link, Figure 71 shows the setup. The GPS fix is used for the timestamp of the transmitted and received data.

Cohda Cohda Tx Rx

Figure 71: Measurement setup for inside ITS-G5 measurements

Figure 72: Intra vehicle/consist measurement

R2R-T2.2-T-DZR-006-18 Page 92 of 186 10/12/2016

Contract No. H2020 – 636032

5.1.2 Night 2 Two trains: The transmitter and receiver were located in different trains. The measurements were done with directional antennas. For the calibration measurements, the two trains needed to be placed in a way that TX and RX were on the same level next to each other to minimize the length of additional cables. This manoeuvre resulted in a total length of necessary railway track of around 640 m. This length was not available at the station Napoli Centrale or the Depot of Naples. Therefore, the calibration had to be done on the high speed line.

During this calibration measurement strange calibration values occurred. Using the same Rubidium frequency normal, the phase spectrum was time variant. Additionally, the phase was symmetric around the IF instead of linear over the transfer function. The reason might be a different electrical ground between either trains or problems with the inverter from DC to AC used inside the train. A later check of the channel sounder at the hotel could not verify the problems during the measurement campaign. Due to these calibration problems on the track, the channel sounder measurements were cancelled. In addition the ITS-G5 measurement equipment was set up for T2T measurements with omni-directional antennas as shown in Figure 73. A reduced and adapted amount of manoeuvres (Table 13, phase 1 and 3) were complete with ITS-G5 equipment and RCAS measurements from 3:00 to 5:30.

Cohda Cohda Tx Rx

Figure 73: Measurement setup for outside ITS-G5 measurements

5.1.3 Night 3 Two trains: The transmitter and receiver were located in different trains. The measurements were done with omni-directional antennas (see Figure 74). For an easier realization of the calibration measurement, Train 07 was turned around. In this case, Locomotive B of Train 07 was located next to Locomotive A of Train 28. Hence, the calibration measurement was performed in the Station Napoli Centrale with a platform between the trains. Within this setting, no problems with the phase in the calibration arose. All manoeuvres were performed as planned in Table 14. The measurement setup is shown in Figure 74 and the settings are explained below.

R2R-T2.2-T-DZR-006-18 Page 93 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 74: Measurement setup for T2T omni-directional measurements

Please note: The high power amplifier with 46 dBm has been used!

General configuration: Bandwidth 120MHz Center frequency 5.2GHz Excess delay 12.8 μs Antennas 1 => 1 Time grid 1.024ms TX power 37 dBm (set) using 46 dBm amplifier (EIRP  33 dBm) Antennas omni-directional (8 dBi gain) 5.1.4 Night 4 Two trains: The transmitter and receiver were located in different trains. The planned measurements setting were using omni-directional antennas. Because of the cancelled channel sounder measurements in night 2, in night 4 directional antennas were used. The calibration measurement was performed in the same way as in night 2. Within the calibration, similar phase problems were recorded as in night 2, even with powering all the equipment from one train. Nevertheless, the measurement run was performed with the setup shown in Figure 75 and started with this calibration on time. All manoeuvres were performed as planned listed in Table 15.

R2R-T2.2-T-DZR-006-18 Page 94 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 75: Measurement setup for T2T directional measurements

Please note: The high power amplifier with 46 dBm has been used!

General configuration:

Bandwidth 120MHz Center frequency 5.2GHz Excess delay 12.8 μs Antennas 1 => 1 Time grid 1.024ms TX power 37 dBm using 46 dBm amplifier (EIRP  34 dBm) Antennas directional (9 dBi gain)

R2R-T2.2-T-DZR-006-18 Page 95 of 186 10/12/2016

Contract No. H2020 – 636032

5.2 URBAN / METRO (METRO DE MADRID) Measurements were carried out in Line 6, Line 10 and Line 11 of Metro de Madrid. The specific lines and stations measured are highlighted in Figure 76 (L6 in grey, L10 in blue, and L11 in green).

Figure 76: Measured metro lines and stations Both wideband and narrowband measurements were carried out. Wideband measurements were done using a channel sounder with the following configuration:

Transmitter Power (dBm) 30 Carrier Frequency (MHz) 2600 Bandwidth (MHz) 80 Modulation Pulses Demodulation Logarithmic Detector

Narrowband measurements where done transmitting an unmodulated carrier frequency with a signal generator, and using as a receiver a National Instruments USRP 2901 Software Defined Radio equipment, which has two receiver ports. The following configuration was used for these narrowband measurements:

Transmitter Power (dBm) 19 Carrier Frequency (MHz) 2600 Bandwidth (MHz) Continuous wave

R2R-T2.2-T-DZR-006-18 Page 96 of 186 10/12/2016

Contract No. H2020 – 636032

5.2.1 Night 1 The preparation of the antennas and measurement equipment was done in the afternoon at Cuatro Vientos depot of Metro de Madrid. Two Mobile-Mark MGRM-WHF antennas were placed on the roof of a Series 8000 continuous train, at the edges of two cars, and spaced 1 meter apart (see Figure 77).

Figure 77: Omnidirectional antennas for inter-consist measurements These antennas were connected to the transmitter and receiver equipment, placed inside the cars; an additional antenna was also placed on a tripod inside the train, in order to check the roof-to- indoor link (see Figure 78).

2 1

3

R T

Figure 78: Inter-consist measurement setup At night, the train was moved to Line 10, and both wideband and narrowband measurements were done between Cuatro Vientos and Tribunal stations. This line has tunnels, stations and also open air environments.

After doing the inter-consist measurements, two MGRM-WHF antennas were placed inside the train at a distance of 15m from each other (5m and 10m from the connection of two cars, respectively) for intra-consist measurements. The measurement setup is shown in Figure 79 and the antennas in Figure 80. With this setup, wideband measurements were carried out.

R2R-T2.2-T-DZR-006-18 Page 97 of 186 10/12/2016

Contract No. H2020 – 636032

1 2

T R

Figure 79: Intra-consist measurement setup

Figure 80: Omnidirectional antennas for intra-consist measurements After completing these measurements, the wideband transmitter was checked and an unexpected behaviour was observed, so it was decided to repeat the wideband measurements in Line 10; they were rescheduled for Night 3.

R2R-T2.2-T-DZR-006-18 Page 98 of 186 10/12/2016

Contract No. H2020 – 636032

5.2.2 Night 2 The second night, measurements were done in Line 11 between La Fortuna and Pan Bendito stations. This line covers tunnel and station environments, but open air was also measured on the way from the depot to the first station (La Fortuna). First, wideband intra-consist measurements were done using a six-antenna array (Figure 81). Then, narrowband intra-consist measurements were done using two monopole antennas, as in Figure 80. Finally, with the monopole antennas which were already installed on the roof, wideband and narrowband inter-consist measurements were also taken.

Figure 81: Antenna array for intra-vehicle measurements After returning to the depot, narrowband intra-consist measurements were done to check the impact of people movement on intra-consist links. The path loss along the train was also measured moving one of the indoor antennas in narrowband configuration (Figure 82).

Figure 82: Path loss measurement

R2R-T2.2-T-DZR-006-18 Page 99 of 186 10/12/2016

Contract No. H2020 – 636032

5.2.3 Night 3 In preparation for the train-to-train measurements to be done at night, in the afternoon of the third day monopole antennas were installed on a second train, both on the roof and inside the train (Figure 83).

1

T

2

3 R

Figure 83: Train-to-train measurement setup Additionally, the properties of the glass of the trains were measured using a measurement setup with horn antennas (Figure 84).

Figure 84: Glass characterization setup At night, narrowband train-to-train measurements were done in Line 10. For these measurements, one of the trains was kept at a fixed location, while the other train crossed it at the next track. One tunnel, one station and one open air locations were used for this. Additionally, before starting the train-to-train measurements, wideband and narrowband inter-consist measurements were taken while moving to the first location in Line 10.

R2R-T2.2-T-DZR-006-18 Page 100 of 186 10/12/2016

Contract No. H2020 – 636032

5.2.4 Night 4 During the afternoon of the fourth day, monopole antennas were placed in a Series 5000 train (non-continuous train) at Laguna depot (Figure 85). At night, the train was moved into Line 6, and narrowband and wideband measurements were taken at Laguna station, including path loss measurements inside the train.

Figure 85: Monopole antennas on a Series 5000 train

5.2.5 Summary Table 26 summarizes the measurements taken in Metro de Madrid.

Table 26: Summary of metro measurements WIDEBAND NARROWBAND Schedule Line Train Link T S OA T S OA Night 1 L10 S8000 Inter-consist (roof, - - - X X X (continuous) roof-to-inside) Night 2 L11 S8000 Intra-consist X X X - - - (continuous) (inside, array) Intra-consist (inside) - - - X X - Inter-consist (roof) X X X X X - Intra-consist - - - - X - (inside, people+path loss) Night 3 L10 S8000 Train-to-Train - - - X X X (continuous) Inter-consist (roof) X X X X X X Night 4 L6 S5000 Inter-consist (roof) - X - - X - (non-continuous) Intra-consist - X - - X - (inside, path loss) T: tunnel; S: station; OA: open air

R2R-T2.2-T-DZR-006-18 Page 101 of 186 10/12/2016

Contract No. H2020 – 636032

5.3 REGIONAL TRAINS (ALSTOM) The 60 GHz mmw radio technology is usable only for the short-range radio link due to the high propagation loss. Besides, the mmw sounder system with integrated antennas is fragile. By taking the feasibility into consideration, the related measurement campaigns should be performed for short-distances (maximum 25 m) with limited mobility. As a result, the following scenarios are proposed.

Thanks to ALSTOM, we had access to the regional trains in Alstom’s factory in Valenciennes in a garage. Under these conditions, only some configurations can be measured.

Figure 86: View of the train for measurements in Valenciennes Following measurement scenarios were performed:

 Intra-Vehicle Communication  Inter Car Scenario  Penetration through the train window  Outdoor to Driver’s room  Outdoor to Driver (Centre)  Outdoor to Outdoor (O2O)  Train to Infrastructure  Outdoor to Roof

More details to each scenario and the different setups are presented in Chapter 6.3.

R2R-T2.2-T-DZR-006-18 Page 102 of 186 10/12/2016

Contract No. H2020 – 636032

6. CHANNEL MODELS This chapter presents the post processing and analysis of the channel measurements described in Chapter 5. For the high speed line measurements, the large scale and small scale fading parameters are estimated in Section 6.1 for the inter-vehicle channel measurements and the parameters for a tap-delay line channel model are provided. Similarly, Section 6.2 provides the large scale and small scale fading parameters of a channel model for Metro inter-consist, intra- consist, and train-to-train scenarios as well as parameters for a tap-delay line channel model for the first two scenarios. Finally, Section 6.3 analyses and estimates parameters for a regional train at 60 GHz carrier frequency for various measurement scenarios such as intra-vehicle and outdoor- to-roof.

R2R-T2.2-T-DZR-006-18 Page 103 of 186 10/12/2016

Contract No. H2020 – 636032

6.1 HSR INTER-VEHICLE CHANNEL MODEL In this section a channel model for data communication between train wagons is presented. Based on the channel sounder data collected in the first night of the HSR-measurements in Italy, an inter- vehicle channel model was developed. A description of the measurement scenario can be found in Section 4.1.3 and the detailed measurement setup in Section 5.1.1.

6.1.1 Data Post Processing In the post processing, as described in Section 3.5, the calibration measurements need to be taken into account. One calibration measurement was done from the N-connector of the antenna cable to the N-connector of the MUX-LNA cable at the receiver side (see “calibration” in Figure 70). This calibration was done before the measurement run and is respected within the measurements automatically. The calibration measurements for the cables connecting MUX rx1 and rx2 with the antennas (see Figure 70) and the MUX itself were done before the measurement run, but need to be considered in the post processing. Therefore, the output of the channel sounder H_rawt(,t) is divided with the transfer function of the MUX H_muxt(,t) to cancel out the cable and MUX effects. The Fourier transformation of Ht(,t) gives the time-variant channel impulse response (CIR) h(,t). Any other equivalents of the CIR can be calculated as illustrated in Figure 87.

Owing to the SISO measurement setup, an angle of arrival cannot be derived. Hence, the antenna pattern cannot be taken into account and the antennas need to be considered as part of the channel. This is a common approach for channel sounder measurements.

time-variant impulse response

Doppler-variant time-variant impulse response transfer function

Doppler-variant transfer function

Figure 87 : The time-variant impulse response and its frequency-domain equivalents [89]

6.1.2 Channel Characteristics

Large scale fading The measurement antennas are placed on the roof of the train as for commercial train applications (see Figure 43). Due to the conditions of same height and short distance, shadowing effects cannot occur. The short range link is 3 m and the long range link is 29 m. The path loss results to a constant average PRx of the LOS link with minor variation.

R2R-T2.2-T-DZR-006-18 Page 104 of 186 10/12/2016

Contract No. H2020 – 636032

a) Path loss

The theoretical received power was calculated as in (14) before the measurement and the transmit power PTx was adjusted not to saturate the input of the channel sounder receiver.

4휋∙푟∙푓 푘 푃 = 푃 − 퐿 + 퐺 − 10 ∙ log ( ) + 퐺 − 퐿 푅푥 푇푥 푐푎푏푙푒 푇푥 푐 푅푥 푐푎푏푙푒 (14)

Table 27: Path loss calculation results

Mux 1 Mux 2

PTx dBm 37

Lcable dB 5.5

GTx dBi 8

GRx dBi 8 8

Lcable dB 17.8 4.3 r m 3 29 f GHz 5.2 c m/s 3*108 k 2 FSPL dB -56.31 -76.02

PRx dBm -26.6 -32.8

The cable losses Lcable were measured for every used cable and considered with the calibration measurements. The antenna gains GTx and GRx can be found in the datasheet of the used antennas. b) Power Delay Profile

The power delay profile (PDP) is calculated out of the CIR of one scenario normalized in two dimensions. First, the CIR is normalized in power with respect to the theoretical free space path loss (FSPL) and is indexed with F. Second, the delay 휏 is normalized w.r.t. the maximum such that τG = 0 s. For communication aspects the absolute delay is not important and therefore a normalization of the delay is not adding any error. To reduce the effect of small scale fading on the PDP analysis, the CIRs are averaged over distances in blocks of d = 40 ∙ λ; lG represents the blocks and Nl the number of snapshots in each block [90].

1 푡푙푔+1 2 푃휏 ,퐹(푙푔, 휏푛) = ∑ |ℎ휏 ,퐹(푡푘, 휏푛)| (15) 퐺 푁푙 푡푘=푡푙푔 퐺

Figure 88 shows the normalized PDP for the two measured link distances of 3 m and 29 m for a slow and fast manoeuvre during the first night. The probability density functions (PDF) of the PDP is presented in Figure 89 (a) and (b) for 3 m and in Figure 89 (c) and (d) for 29 m. All four figures show the strong LOS behaviour of the link, visible with the highest concentration at delay τ = 0 sec. The shift to -3 dB for the fast and -5 dB of the slow measurements is caused by the varying antenna characteristics.

R2R-T2.2-T-DZR-006-18 Page 105 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 88: Normalized PDP for 1. Night: scenario 1 (slow), scenario 2 (fast)

(a) 3 m slow (b) 3 m fast

(c) 29 m slow (d) 29 m fast

Figure 89: PDF of Normalized PDP for 1. Night

R2R-T2.2-T-DZR-006-18 Page 106 of 186 10/12/2016

Contract No. H2020 – 636032

Comparing the PDPs of the two different link distances, the PDP of the 29 m link shows strong MPC around 8 and 16 ns. As described in 4.1.8 and 5.1.1 the transmitter and receiver were mounted on the same train which is moving with a certain speed along the track in a changing environment. Hence, constant MPCs are caused by scatterers on the train itself. In Figure 46 the antenna positions are shown and in Figure 43 the roof of coach 3. The delays 휏1 = 8.3 푛푠 and 휏2 = 16.6 푛푠 are resulting in an additional distance of ∆푑1 = 2.5 푚 and ∆푑2 = 5 푚. These distances match with the positions of the additional antennas on coach 3.

Small scale fading The strongest variation of the received power is caused by multipath propagation. The railway environment and also the train cause reflections and scattering. For the analysis a sliding average window is used to cancel out the large scale fading. For the slow ride the window is 40 snapshots for the 3 m for the 29 m link. The small scale fading is analysed for each tap within a delay of 100 ns.

6.1.3 Tapped delay line model

Bandwidth The bandwidth of the measurement signal of the RUSK DLR Channel Sounder is 퐵 = 120 푀퐻푧. This results in a minimum spacing of the measurement samples of ∆휏 = 8.33 푛푠. The delay spacing of the measurements is the lower bound of the delay in the tapped delay line model. With the delay spacing of the taps the bandwidth of the model is set. To achieve a technology independent channel model the maximum bandwidth of 퐵 = 120 푀퐻푧 is used for the model.

Significant taps The PDP in Figure 88 shows a decrease in power of 20 dB or more within a delay of 100 ns. Samples within this range can be considered as significant for the channel model. W.r.t the chosen bandwidth the amount of significant taps is equivalent to the measurement samples within 100 ns.

Figure 90 : Tapped delay line model

Figure 90 shows a block diagram of the tapped delay line model. The total number of taps equals to 푛 = 13. The delay between the taps is 푇 = 8.33 푛푠.

R2R-T2.2-T-DZR-006-18 Page 107 of 186 10/12/2016

Contract No. H2020 – 636032

Tap model fitting parameters In the following tables and figures the parameters of the fitted channel model for the 3 m and 29 m link for low speed are presented. Three of the most common distributions were used to fit to the measurement data:

 Rician distribution: The Rician distribution is used for modelling multipath propagation with a strong dominant path (LOS). The shape fitting is adjusted with the mean  and the 2 variance 휎 ; 퐼0 is the zero-order modified Bessel function of the first kind [91].

x2+μ2 x∙μ x −( ) 푝 = I ( ) ∗ ∗ e 2σ2 푥 0 σ2 σ2 (16)

The Rician k-factor is defined as the ratio of the dominant to the scattering path for one multipath component (in the case of tapped delay line, for one tap) [90]:

2 푃푚푒푎푛 μ k = 10 ∙ 푙표푔10 ( ) = 10 ∙ log10 ( 2) (17) 푃푓푎푑 2∙σ

 Rayleigh distribution: For dense scattering propagation environments the Rayleigh distribution is used. The mean 휇 = 0 and the variance 휎2 is the tuning parameter for the fitting.

x2 x −( ) p = ∗ e 2휎2 x σ2 (18)

 Nakagami distribution: The Nakagami distribution is used to model scattering channels as well. With the shape parameter  and the scale parameter  this distribution was originally used for ionospheric links [92].

μ∙x2 휇 휇 푥(2휇−1) −( ) p = 2 ( ) ∗ ∗ e ω x 휔 Γ(μ) (19)

In Table 28 the fitting parameters for the 3 m link are listed. The normalized power out of the PDP is given for all tabs; the worst case (3m slow speed) is used. Caused by the short link distance, only the first tap, including the LOS link, needs to be modelled different to the rest of the taps. As shown in

Tap Norm. Power Fading k-factor Rician Rayleigh # dB distribution dB    1 -5.15 Rice 17.44 1 0.095 - 2 -16.08 3 --19-63 4 -21.93 5 -26.47 Rayleigh - - - 0.9 6 -28.14 7 -29.67 8 -31.93 9 -33.08 1

R2R-T2.2-T-DZR-006-18 Page 108 of 186 10/12/2016

Contract No. H2020 – 636032

10 -34.42 0.9 11 -33.44 1 12 -34.96 13 -36.03 0.9

(a) 3 m tap 1 (LOS)

(b) 3 m taps 2-8, 10, 13

R2R-T2.2-T-DZR-006-18 Page 109 of 186 10/12/2016

Contract No. H2020 – 636032

(c) 3 m taps 9, 11, 12 Figure 91 (a), tap one is Rician distributed with a k-factor of 17.44 dB. Tap two to thirteen are nearly similar and fit to two Rayleigh distributions as shown in

Tap Norm. Power Fading k-factor Rician Rayleigh # dB distribution dB    1 -5.15 Rice 17.44 1 0.095 - 2 -16.08 3 --19-63 4 -21.93 5 -26.47 Rayleigh - - - 0.9 6 -28.14 7 -29.67 8 -31.93 9 -33.08 1 10 -34.42 0.9 11 -33.44 1 12 -34.96 13 -36.03 0.9

R2R-T2.2-T-DZR-006-18 Page 110 of 186 10/12/2016

Contract No. H2020 – 636032

(a) 3 m tap 1 (LOS)

(b) 3 m taps 2-8, 10, 13

R2R-T2.2-T-DZR-006-18 Page 111 of 186 10/12/2016

Contract No. H2020 – 636032

(c) 3 m taps 9, 11, 12 Figure 91 (b) and (c). These results match with our observations during the measurement campaign.

R2R-T2.2-T-DZR-006-18 Page 112 of 186 10/12/2016

Contract No. H2020 – 636032

Table 28: Tap fitting parameters 3 m

Tap Norm. Power Fading k-factor Rician Rayleigh # dB distribution dB    1 -5.15 Rice 17.44 1 0.095 - 2 -16.08 3 --19-63 4 -21.93 5 -26.47 Rayleigh - - - 0.9 6 -28.14 7 -29.67 8 -31.93 9 -33.08 1 10 -34.42 0.9 11 -33.44 1 12 -34.96 13 -36.03 0.9

(a) 3 m tap 1 (LOS)

R2R-T2.2-T-DZR-006-18 Page 113 of 186 10/12/2016

Contract No. H2020 – 636032

(b) 3 m taps 2-8, 10, 13

(c) 3 m taps 9, 11, 12 Figure 91: 3 m measurement, distribution fitting for different taps

R2R-T2.2-T-DZR-006-18 Page 114 of 186 10/12/2016

Contract No. H2020 – 636032

(a) 29 m tap 1 (LOS)

(b) 29 m taps 2-3

R2R-T2.2-T-DZR-006-18 Page 115 of 186 10/12/2016

Contract No. H2020 – 636032

(c) 29 m taps 4-6

(d) 29 m taps 7-13 Figure 92: 29 m measurement, distribution fitting for different taps

R2R-T2.2-T-DZR-006-18 Page 116 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 92 shows the fitting of the 29 m link. The LOS link is modelled with a Rice distribution in tap one with a k-factor of 10.97 dB. As described before in the paragraph Power Delay Profile constant MPCs occur in this measurement. Tap two and three include these MPCs as shown in Figure 92 (b); the fitting can be realized with a Rayleigh distribution. In Figure 92 (c) and (d) the slightly different taps 4-6 and 7-13 are shown with a Rayleigh fitting. All parameters of all fittings for all taps are listed in Table 29.

Table 29: Tap fitting parameters 29 m

Tap Norm. Fading k-factor Rician Rayleigh Power distribution # dB dB    1 -6.00 Rice 10.97 1 0.2 - 2 -12.61 1 3 -15.65 4 -18.36 1.19 5 -19.80

6 -20.41 Rayleigh - - - 7 -22.06

8 -23.61

9 -25.48 1.25 10 -26.36 11 -27.44 12 -28.15 13 -28.94

For both models (3 m and 29 m link) the normalized power in Table 28 and Table 29 is describing the average received power normalized to the FSPL. Due to the nature of these short links on the roof of HSRs, shadowing effects or other effects described by large scale fading do not appear. The small scale effects are described by the fading distribution for each tap, in which each tap is normalized to its own maximum.

R2R-T2.2-T-DZR-006-18 Page 117 of 186 10/12/2016

Contract No. H2020 – 636032

6.2 METRO CHANNEL MODEL In this Section the channel models obtained for metro environment will be described, including Inter-consist, Intra-consist and Train-to-train links. For each type of link, separate models will be obtained for tunnel, station and open air environments. First of all, the post-processing of the data will be described.

6.2.1 Data Post Processing Regarding narrowband measurements, in order to model the channel it is necessary to analyze statistically the small-scale fading of the signal. This has been done by removing the large-scale fading of the channel, and checking which theoretical Probability Density Functions (PDF) fits best with the result. In order to obtain the large-scale fading of the channel, narrowband data have been averaged every 40 samples.

Regarding wideband measurement results, RMS Spread and Mean Delay have been calculated. Then, the Power Delay Profile (PDP) of the channel has been calculated, and an 8-tap channel model has been obtained, where each tap is spaced 12.5 ns, corresponding to the receiver bandwidth of 80 MHz.

6.2.2 Inter-consist Model

Tunnel Figure 93 shows the narrowband results for a tunnel in Line 10. Instantaneous and averaged data are presented for antennas placed on the roof of the train (Outdoor antenna); the reception of one antenna inside the train is also presented for comparison (Indoor antenna).

Figure 93: Narrowband Results for Inter-Consist link in Tunnel (Line 10) It can be observed that the signal is attenuated 30-35 dB when one of the antennas is placed inside the train, compared to the outdoor case. This is because in this situation there is no Line of Sight between the two antennas, and the signal from the antenna on the roof reaches the antenna

R2R-T2.2-T-DZR-006-18 Page 118 of 186 10/12/2016

Contract No. H2020 – 636032

in the train through diffraction on the train structure (e.g. window frames, car interconnection structure, etc) and also through reflections on the tunnel walls. It can also be observed that the amplitude of the small-scale fading is lower for the case of the outdoor antennas, and mainly slow signal variations are observed; this is due to the influence of the large Line Of Sight component and the stability of the surrounding reflections. On the contrary, for the case of the indoor antenna fast signal variations are observed.

In order to illustrate the effect of the windows of the car, the electrical properties of the glass have been obtained (see measurement setup in Figure 84). Figure 94 and Table 30 show the result of this characterization. It can be observed that the losses due to the glass are not significant (1.8 to 4.4 dB), and a phase variation is observed due to the permittivity of the glass.

Figure 94: Glass characterization: losses (blue graph); phase deviation (yellow graph)

Table 30: Glass characterization results Frequency (GHz) Attenuation (dB) Phase (degrees) 1.6 1.9 -54.4 2.3 1.8 -80.9 3 4.4 -105

Regarding wideband results, the evolution of each of the 8 taps of the channel model is presented in Figure 95 for a tunnel in Line 10, and in Figure 97 for a tunnel in Line 11, with outdoor antennas in both cases. It can be observed the main influence of the first tap (i.e. the LOS component). The corresponding Probability Density Functions (PDFs) for each tap are displayed in Figure 96 and Figure 98, respectively.

R2R-T2.2-T-DZR-006-18 Page 119 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 95: Channel taps vs distance for Inter-Consist link in Tunnel (Line 10)

Figure 96: Probability Density Function of channel taps for Inter-Consist link in Tunnel (Line 10)

R2R-T2.2-T-DZR-006-18 Page 120 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 97: Channel taps vs distance for Inter-Consist link in Tunnel (Line 11)

Figure 98: Probability Density Function of channel taps for Inter-Consist link in Tunnel (Line 11)

R2R-T2.2-T-DZR-006-18 Page 121 of 186 10/12/2016

Contract No. H2020 – 636032

Station Narrowband results for a station in Line 10 are shown in Figure 99.

Figure 99: Narrowband Results for Inter-Consist link in Station (Line 10) Regarding wideband results, the evolution of the taps and their PDFs are shown in Figure 100 and Figure 101 respectively, for a station in Line 10. Results are also presented for a station in Line 11 in Figure 102 and Figure 103, and for Line 6 in Figure 104 and Figure 105.

Figure 100: Channel taps vs distance for Inter-Consist link in Station (Line 10)

R2R-T2.2-T-DZR-006-18 Page 122 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 101: Probability Density Function of channel taps for Inter-Consist link in Station (Line 10)

Figure 102: Channel taps vs distance for Inter-Consist link in Station (Line 11)

R2R-T2.2-T-DZR-006-18 Page 123 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 103: Probability Density Function of channel taps for Inter-Consist link in Station (Line 11)

Figure 104: Channel taps vs distance for Inter-Consist link in Station (Line 6)

R2R-T2.2-T-DZR-006-18 Page 124 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 105: Probability Density Function of channel taps for Inter-Consist link in Station (Line 6)

Open Air Figure 106 shows the narrowband results in open air, measured in Line 10. Wideband results, also for Line 10, are summarized in Figure 107 and Figure 108.

Figure 106: Narrowband Results for Inter-Consist link in Open Air (Line 10)

R2R-T2.2-T-DZR-006-18 Page 125 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 107: Channel taps vs distance for Inter-Consist link in Open Air (Line 10)

Figure 108: Probability Density Function of channel taps for Inter-Consist link in Open Air (Line 10)

R2R-T2.2-T-DZR-006-18 Page 126 of 186 10/12/2016

Contract No. H2020 – 636032

Analysis Table 31 summarizes the attenuation of the different scenarios, obtained from narrowband measurements.

Table 31: Average attenuation in different environments Average attenuation with Average attenuation with External to internal external antennas external and internal antenna difference (dB) (dB) (dB) Station 40.49 75.12 34.63 Tunnel 37.52 73.19 35.67 Open Air 37.22 79.83 42.61

It can be observed that the value of attenuation is slightly higher at the station, when it should have been similar to tunnel or open air. However, the amount of data recorded at the station is lower, due to the lower distance covered (120m), and therefore the obtained average value is less accurate. On the other hand, it can be observed that there is a 34-35 dB difference between the outdoor and indoor receiver antennas in station and tunnel environments; this value rises up to 43 dB in open air, as there is less energy coming from external reflections.

The measured narrowband data with the external antennas fit statistically to Rice distribution, where the LOS component is substantially higher than the multipath components. The Probability Density Functions (PDF) and their parameters are shown in Figure 109 and Table 32.

Figure 109: Probability Density Function with External Antennas Table 32: Statistical parameters of external link PDF Rice K (dB) Standard Deviation Station 0.15 0.96 Tunnel 9.61 0.41 Open Air 16.41 0.20

R2R-T2.2-T-DZR-006-18 Page 127 of 186 10/12/2016

Contract No. H2020 – 636032

As can be observed, the link with external antennas shows clearly a Rician PDF, due to the excellent LOS in this type of link. The lower value of the K-Rice coefficient in the station environment is due to a higher multipath.

On the other hand, the link from the antenna in the roof to the antenna inside the train fits statistically to a Rayleigh distribution, as there is no LOS between the antennas (see Figure 110 and Table 33). The standard deviation is lower for open air and station, due to less reflections coming from the environment (the station is the opposite case, with a higher value of standard deviation due to a higher level of reflections and signal fading).

Figure 110: Probability Density Function with External and Internal Antenna Table 33: Statistical parameters for external-to-internal link Parameter Value PDF Rayleigh Standard Station 2.64 deviation Tunnel 1.62 Open Air 1.79

Regarding wideband measurements, Figure 111 shows the magnitude of each of the 8 multipath components (taps) with respect to the LOS component in Line 10, for the link with external antennas.

R2R-T2.2-T-DZR-006-18 Page 128 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 111: Power Delay Profile (PDP) for Inter-Consist link in Line 10 These values are summarized in Table 34.

Table 34: Multipath components for Inter-Consist link in Line 10 Number of Delay (ns) Attenuation (dB) TAP Station Tunnel Open Air 1: ray LoS 0 0 0 0 2 12.5 13.7 12.8 12.8 3 25 19.9 18.9 18.8 4 37.5 11.9 11.3 11.3 5 50 12.7 13.1 12.0 6 62.5 15.1 14.5 14.5 7 75 12.7 12.1 12.1 8 87.5 17.1 16.2 16.2

RMS spread and Mean Delay values have also been extracted from wideband measurements in Line 10, as well as the statistical distributions of the different taps (see Table 35 and Table 36).

Table 35: RMS spread and Mean Delay for Inter-Consist link in Line 10 RMS delay spread (ns) Mean delay (ns) Tunnel 62.43 36.04 Station 61.24 31.10 Open Air 61.59 34.56

R2R-T2.2-T-DZR-006-18 Page 129 of 186 10/12/2016

Contract No. H2020 – 636032

Table 36: Statistical parameters of channel taps for Inter-Consist link in Line 10 Number Station Tunnel Open Air of TAP K(dB) Standard K(dB) Standard K(dB) Standard deviation deviation deviation 1: ray 12.06 0.11 12.33 0.12 9.99 0.21 LoS 2 10.44 0.15 12.10 0.11 10.47 0.17 3 7.83 0.25 10.87 0.16 10.39 0.20 4 13.53 0.09 14.45 0.07 13.07 0.10 5 13.52 0.08 14.63 0.06 9.18 0.23 6 11.65 0.12 12.97 0.11 11.37 0.14 7 16.19 0.04 15.62 0.05 12.75 0.10 8 14.10 0.07 15.74 0.05 12.87 0.10

It can be observed that the RMS spread is very similar in all environments. However, there is a difference in the mean delay (up to 15% higher in tunnel compared to station).

Equivalent results have been obtained for Line 11 (Station and Tunnel) and Line 6 (Station). They are presented in Figure 112, Table 37, Table 38 and Table 39 for Line 11, and Figure 113, Table 40, Table 41 and Table 42 for Line 6.

Figure 112: Power Delay Profile (PDP) for Inter-Consist link in Line 11

R2R-T2.2-T-DZR-006-18 Page 130 of 186 10/12/2016

Contract No. H2020 – 636032

Table 37: Multipath components for Inter-Consist link in Line 11 Number Delay (ns) Attenuation (dB) of TAP Station Tunnel 1: ray 0 0 0 LoS 2 12.5 16.11 14.96 3 25 22.81 21.16 4 37.5 11.57 11.80 5 50 14.78 13.81 6 62.5 16.27 14.74 7 75 12.78 12.14 8 87.5 18.01 17.20

Table 38: RMS spread and Mean Delay for Inter-Consist link in Line 11 RMS spread (ns) Mean delay (ns) Tunnel 42.44 15.35 Station 40.00 18.50

Table 39: Statistical parameters of channel taps for Inter-Consist link in Line 11 Number of Station Tunnel TAP K(dB) Standard deviation K(dB) Standard deviation 1: ray LoS 11.42 0.12 8.51 0.23 2 10.77 0.16 7.57 0.26 3 7.90 0.32 6.40 0.34 4 8.20 0.23 7.75 0.29 5 9.21 0.18 6.98 0.31 6 11.09 0.14 9.83 0.18 7 14.50 0.07 10.71 0.15 8 13.18 0.08 11.91 0.12

R2R-T2.2-T-DZR-006-18 Page 131 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 113: Power Delay Profile (PDP) for Inter-Consist link in Line 6 Table 40: Multipath components for Inter-Consist link in Line 6 Number of TAP Delay (ns) Attenuation (dB) Station 1: ray LoS 0 0 2 12.5 12.53 3 25 21.01 4 37.5 9.24 5 50 8.40 6 62.5 15.32 7 75 15.25 8 87.5 17.17

Table 41: RMS spread and Mean Delay for Inter-Consist link in Line 6 RMS spread (ns) Mean delay (ns) Station 34.11 16.80

R2R-T2.2-T-DZR-006-18 Page 132 of 186 10/12/2016

Contract No. H2020 – 636032

Table 42: Statistical parameters of channel taps for Inter-Consist link in Line 6 Number Station of TAP K(dB) Standard Deviation 1: ray 10.70 0.17 LoS 2 10.91 0.18 3 7.79 0.23 4 12.30 0.11 5 8.32 0.21 6 11.46 0.13 7 9.71 0.17 8 12.36 0.12

6.2.3 Intra-consist Model (Continuous Train)

Tunnel Wideband measurements were done in Line 11 using a six-antenna array. This allows distinguishing the angle of arrival of the different groups of rays. Figure 114 shows the obtained response in a tunnel, where each antenna covers an angle of 60 degrees.

Figure 114: PDP vs angle of arrival for Intra-Consist link in Tunnel (Line 11) It can be observed that most of the energy comes from the antenna facing the transmitter antenna, which is located on the left side, and most multipath is received by the two antennas on its sides. On the other hand, if the transmitting antenna was fixed and the receiver array was moving, it would be possible to obtain the Doppler spectrum of the train with this measurement. In this case, both antennas were moving at the same time, and therefore it is not possible to separate the reflections due to the train from those due to the environment; the first ones do not produce Doppler, while the second ones do. However, the plots show that most of the energy in the intra- consist link comes from the reflections inside the train, what implies that there is little influence of the environment and therefore little influence of the Doppler spectrum.

R2R-T2.2-T-DZR-006-18 Page 133 of 186 10/12/2016

Contract No. H2020 – 636032

Station Figure 115 shows the results of the measurements with the array in a station environment.

Figure 115: PDP vs angle of arrival for Intra-Consist link in Station (Line 11) Afterwards, in order to obtain the path loss of the intra-consist link in a continuous train, the receiver antenna was moved along the train and narrowband measurements were taken. Figure 116 shows the average of the received signal and the path loss model.

Figure 116: Path loss model for the Intra-Consist link (continuous train) In order to represent this path loss evolution, a two-slope model has been chosen due to the different nature of the received signal near and far from the transmitter:

55.8 𝑖푓 푑 < 6 퐴(푑퐵) = { (20) 31.6 · log(푑(푚)) + 31.2 𝑖푓 푑 ≥ 6

R2R-T2.2-T-DZR-006-18 Page 134 of 186 10/12/2016

Contract No. H2020 – 636032

It can be observed that near the transmitter the path loss exponent (n) is very small (less than 0.5), and therefore the path loss can be estimated with a constant value. Afterwards, the path-loss exponent is 3.1, higher than the free-space value, due to the reflections in the various inner structures of the train (e.g. walls, vertical hand-hold bars, interconnections between cars, etc). A statistical modeling has also been done for these narrowband measurements, obtaining a fit to a Rician model, due to the high LOS component (see Figure 117 and Table 43).

Figure 117: Probability Density Function of Intra-Consist link in station (continuous train) Table 43: Statistical parameters of Intra-Consist link in station (continuous train) Parameter Value PDF Rician Standard deviation 0.54 K (dB) 1.55

The presence of people inside the train has also been measured at the depot. Figure 118 shows a narrowband measurement over time with random people movement (even blocking the LOS link). The statistical modeling is shown in Figure 119 and Table 44. Signal fading up to 30 dB is observed, as well as averaged variations of 10 dB. It is also observed that the impact of the environment is higher when the LOS is blocked; therefore, for this kind of situations the use of MIMO solutions or solutions with spatial diversity would be advisable.

R2R-T2.2-T-DZR-006-18 Page 135 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 118: Impact of people movement on Intra-Consist link (continuous train)

Figure 119: Probability Density Function of Intra-Consist link (continuous train) with people movement Table 44: Statistical parameters of Intra-Consist link (continuous train) with people movement Parameter Value PDF Rician Standard deviation 0.36 K (dB) 5.35

R2R-T2.2-T-DZR-006-18 Page 136 of 186 10/12/2016

Contract No. H2020 – 636032

Open Air Figure 120 shows the results of the measurements with the array in an open air environment.

Figure 120: PDP vs angle of arrival for Intra-Consist link in Open Air (Line 11)

Environments Comparison RMS spread and Mean Delay values have been extracted from the wideband measurements in Line 11, as well as the statistical distributions of the different taps (Table 45 and Table 46)

Table 45: RMS spread and Mean Delay for Intra-Consist link (continuous train) in Line 11 Parameter RMS delay spread (ns) Mean delay (ns) Station 66.56 58.9 Tunnel 66.9 53.5 Open Area 64.4 51.4

Table 46: Multipath components for Intra-Consist link (continuous train) in Line 11 Number of Delay (ns) Attenuation (dB) TAP Station Tunnel Open area 1: ray LoS 0 0 0 0 2 12.5 5.6 7 4.8 3 25 5.2 7.2 5.8 4 37.5 7.1 11.3 8.3 5 50 12.7 13.1 15.0 6 62.5 15.1 16.5 15.5 7 75 17.2 17.1 17.1 8 87.5 18 18.2 18.6

R2R-T2.2-T-DZR-006-18 Page 137 of 186 10/12/2016

Contract No. H2020 – 636032

6.2.4 Intra-consist Model (Non-Continuous Train) Narrowband measurements were done at Laguna station (Line 6) for obtaining the path loss inside a non-continuous train. Figure 121 shows the result.

Figure 121: Path loss model for the Intra-Consist link in a non-continuous train The equation for the path loss model is the following one:

퐴(푑퐵) = 22.37 · log(푑(푚)) + 40.5 (21)

The statistical modeling of these narrowband measurements is presented in Figure 122 and Table 47.

R2R-T2.2-T-DZR-006-18 Page 138 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 122: Probability Density Function of Intra-Consist link in station (non-continuous train) Table 47: Statistical parameters of Intra-Consist link in station (non-continuous train) Parameter Value PDF Rician Standard deviation 0.63 K (dB) 0.96

Wideband measurements were also carried out with transmitter and receiver antennas placed in separate cars in the non-continuous train. The evolution of the taps and their PDFs are shown in Figure 123 and Figure 124, respectively.

R2R-T2.2-T-DZR-006-18 Page 139 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 123: Channel taps vs distance for Intra-Consist link (non-continuous train) in Station (Line 6)

Figure 124: Probability Density Function of channel taps for Intra-Consist link (non- continuous train) in Station (Line 6) RMS spread and Mean Delay values have also been extracted from wideband measurements in Line 6, as well as the statistical distributions of the different taps (see Table 48 and Table 49). On the other hand, the magnitude of each of the 8 multipath components (taps) with respect to the LOS component is shown in Table 50.

R2R-T2.2-T-DZR-006-18 Page 140 of 186 10/12/2016

Contract No. H2020 – 636032

Table 48: RMS spread and Mean Delay for Intra-Consist link (non-continuous train) in Line 6 RMS spread (ns) Mean delay (ns) Station 75.78 46.03

Table 49: Statistical parameters of channel taps for Intra-Consist link (non-continuous train) in Line 6 Number of TAP K(dB) Standard Deviation 1: ray LoS 8.27 0.96 2 8.45 0.43 3 10.97 0.40 4 11.18 0.64 5 10.12 0.21 6 11.92 0.35 7 12.68 0.47 8 12.64 0.25

Table 50: Multipath components for Intra-Consist link (non-continuous train) in Line 6 Number of TAP Delay (ns) Attenuation (dB) Station 1: ray LoS 0 0 2 12.5 2.1 3 25 5.2 4 37.5 2.3 5 50 8.1 6 62.5 9.1 7 75 11.2 8 87.5 12.2

6.2.5 Train-to-Train Model In this Section the results of train-to-train measurements of Line 10 are presented, along with the proposed models.

Tunnel Figure 125 shows the averaged narrowband result for the train-to-train link in tunnel environment. The receiver train was stopped inside a tunnel, while the transmitter train crossed it.

R2R-T2.2-T-DZR-006-18 Page 141 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 125: Averaged Narrowband Results for Train-to-Train link in tunnel For the case of the outdoor antennas, a model is obtained for the path loss (see Figure 126).

Figure 126: Path loss model for Train-to-Train link in tunnel with outdoor antennas

R2R-T2.2-T-DZR-006-18 Page 142 of 186 10/12/2016

Contract No. H2020 – 636032

The equation for this path loss model is the following one:

퐴 (푑퐵) = 14.5 푙표푔 (푑(푚)) + 54.8 (22)

Station Figure 127 and Figure 128 show the averaged narrowband result and the path loss model in station environment. In this case the transmitter train was stopped inside the station, while the receiver train crossed it.

Figure 127: Averaged Narrowband Results for Train-to-Train link in station

R2R-T2.2-T-DZR-006-18 Page 143 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 128: Path loss model for Train-to-Train link in station with outdoor antennas The equation for the path loss model is the following one:

퐴 (푑퐵) = 16.4 푙표푔 (푑(푚)) + 47.8 (23)

Open Air The crossing in open air was done between Batán and Casa de Campo stations. The receiver train was stopped, while the transmitter train crossed it. Figure 129 shows the averaged narrowband result, while Figure 130 shows the path loss model.

R2R-T2.2-T-DZR-006-18 Page 144 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 129: Averaged Narrowband Results for Train-to-Train link in open air

Figure 130: Path loss model for Train-to-Train link in open air with outdoor antennas The equation for the path loss model is the following one:

퐴 (푑퐵) = 15.1 푙표푔 (푑(푚)) + 59.9 (24)

R2R-T2.2-T-DZR-006-18 Page 145 of 186 10/12/2016

Contract No. H2020 – 636032

Environments Comparison Figure 131 represents the normalized narrowband results for Train-to-Train links in the three environments. It can be observed that with the outdoor antennas the reception of the signal starts at a distance of approximately 250 m in station environment, while in open air it starts at 50 m, and in the tunnel at 100 m. This indicates that the station environment is more reflective than the open air, and the tunnel represents an intermediate case.

Figure 131: Comparison of averaged Narrowband Results for Train-to-Train links: outdoor receiving antenna (upper); indoor receiving antenna (bottom) In order to analyze the small-scale fading in Train-to-Train links, the averaged part of the signal has been removed to the recorded waveform, and it has been fitted with theoretical PDFs. Figure 132 and Table 51 show the results for the outdoor receiving antenna, while Figure 133 and Table 52 show the result for the indoor receiving antenna. These results indicate the presence of a strong LOS component, both for outdoor and indoor antennas.

R2R-T2.2-T-DZR-006-18 Page 146 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 132: Probability Density Function for Train-to-Train links with outdoor receiving antenna Table 51: Statistical parameters for Train-to-Train links with outdoor receiving antenna Parameter Value PDF Rician Standard deviation Station 0.36 Tunnel 0.29 Open Air 0.26 K (dB) Station 10.83 Tunnel 12.92 Open Air 13.79

R2R-T2.2-T-DZR-006-18 Page 147 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 133: Probability Density Function for Train-to-Train links with indoor receiving antenna Table 52: Statistical parameters for Train-to-Train links with indoor receiving antenna Parameter Value PDF Rician Standard deviation Station 0.31 Tunnel 0.25 Open Area 0.26 K (dB) Station 12.33 Tunnel 14.23 Open Area 14.00

Finally, Figure 134 shows a comparison of the path loss models in the different propagation environments for the case of outdoor antennas.

R2R-T2.2-T-DZR-006-18 Page 148 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 134: Comparison of path loss models for Train-to-Train links with outdoor antennas These path loss models follow this expression:

퐴 (푑퐵) = 10푛 푙표푔 (푑(푚)) + 푃0 (25)

where n is the path loss index, P0 is the path loss at 1m, and d is the distance. The values of these parameters for the different propagation environments are summarized in Table 53. It can be observed that in all cases the propagation index is lower than 2 (free space value), due to the reflections coming both from the environment (tunnel and station) and from the train structure itself.

Table 53: Path loss model parameters

Environment n P0 Tunnel 1.45 54.8 Station 1.64 47.8 Open Air 1.51 59.9

R2R-T2.2-T-DZR-006-18 Page 149 of 186 10/12/2016

Contract No. H2020 – 636032

6.3 REGIONAL TRAIN: 60 GHZ CHANNEL MODEL

6.3.1 Intra-Vehicle Communication (Regional trains) This scenario consists of measuring the wireless link between different elements inside a single vehicle. These will be Line-Of-Sight (LOS) links, mostly affected by the internal structure of the vehicle. As shown in Figure 135 (a), two typical positions are selected. The azimuth angles at the Rx antenna were swept in 22° steps from 0 to 360°, while the Tx antenna is fixed with the opening to the corridor shown in Figure 135 (b).

Table 54: Characteristics of the measurements Cabin Zone 1 Length (m) 2.89 Width (m) 2.15 Tx/Rx height (m) 1

Cabin Zone2 Rx to right-side wall (m) 0.4 Rx to the back of the wall (m) 0.8 Tx/Rx height (m) 1

(a) (b)

R2R-T2.2-T-DZR-006-18 Page 150 of 186 10/12/2016

Contract No. H2020 – 636032

(c) (d)

Figure 135: Photos of the measurements (a) to (d) (a): scheme of the scenarios, (b): view of the TX and Rx position on first floor. (c): view of Tx and Rx position on 2nd floor. (d): view of the stairs

Here after, the figures show the Angular Power distribution. 0° corresponds to the facing direction to the front wall. Measurement with TX and RX on two different floor were taken as illustrated in scheme Figure 135 a.

Figure 136: Angular power distribution for the two positions of the receiver

R2R-T2.2-T-DZR-006-18 Page 151 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 137: Angular power distribution for the two positons of the receiver

The high power region is observed between [-20°, +20°], [150°, 240°] degrees. The Receive signal level (RSL) at downstairs location is slightly higher than the upstairs. The following Table 55 and Table 56 summarize the values of Mean delay and RMS delay in ns for different angle positions.

Table 55: Mean delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 Upstair 4.2 4.1 15.6 26.0 22.0 21.8 26.0 24.8 54.2 21.8 16.1 16.7 14.9 10.9 21.4 13.2 4.1 Downstair 40.4 8.2 66.7 45.3 123.0 173.5 195.3 183.4 218.1 36.7 87.9 167.7 230.0 277.2 138.4 22.1 7.9

Table 56: RMS delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 Upstair 1.3 1.2 13.6 13.2 8.0 9.1 11.0 9.4 155.9 12.8 5.3 6.0 6.9 5.3 14.8 12.2 1.3 Downstair 195.6 7.7 255.7 167.8 325.9 323.6 378.1 311.8 438.9 13.9 258.8 300.5 444.5 505.7 337.6 5.1 8.2

The tables show that the mean delay and RMS delay at downstairs location is much larger than at upstairs location. It shows more reflection paths at downstairs location.

6.3.2 Inter Car Scenario The scenarios are illustrated by the scheme on Figure 138 and shown in the photography. This scenario includes wireless link between the cars. It has been performed when the door is closed and open. As previously, the receiver is swept over 360° for each position. The received power versus angular positon is given on the following figures.

R2R-T2.2-T-DZR-006-18 Page 152 of 186 10/12/2016

Contract No. H2020 – 636032

(a) (b)

Figure 138: Inter car configuration Figure 139 to Figure 141 illustrate the angular power distribution at the three different positions when the door is open or close.

Position1 (P1) - Door open Position1 (P1) - Door close

Figure 139: Angular power distribution at P1 with the door open and close

Position2 (P2) – Door open Position2 (P2) – Door close

Figure 140: Angular power distribution at P2 with the door open and close

R2R-T2.2-T-DZR-006-18 Page 153 of 186 10/12/2016

Contract No. H2020 – 636032

Position3 (P3) – Door open Position3 (P3) – Door close

Figure 141: Angular power distribution at P3 with the door open and close We can observe that when the door is closed, the received signal level is concentrating on the direction of the obstructed line of sight (OLOS) between the Tx and the Rx. Beamforming or directional antenna pattern is needed in this case. At the same location, the metal door can cause about 10 dB attenuation loss. The RSL in the direction of OLOS is about 10 dB higher than the rest of directions.

The power region for P1 is roughly uniform except the region of [240°, 260°] degrees. The angular power distribution of P2 and P3 are consistent. It can be concluded the angular power distribution between upstairs and downstairs at similar locations are consistent for the inter Car scenario when the door is open. At the same locations the angular distribution for door open case and door close case are consistent except the dedicated power in the OLOS direction.

Figure 142: Sum-up of the angular power distribution for the 3 positions The following Table 57 and Table 58 summarize the values of Mean delay and RMS delay in ns for different angle positions.

R2R-T2.2-T-DZR-006-18 Page 154 of 186 10/12/2016

Contract No. H2020 – 636032

Table 57: Mean delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 P1 31,4 192,9 18,4 21,3 8,3 11,4 11,4 22,1 21,7 29,9 9,8 98,4 302,8 22,1 26,8 30,7 202,9 P1 door close 15,3 7,6 7,6 8,1 59,5 249,7 23,2 246,5 34,9 132,7 16,9 581,2 636,2 499,7 490,0 313,8 289,3 P2 7,7 7,6 8,9 29,1 25,3 27,5 16,4 27,8 33,5 21,8 12,2 21,3 20,3 20,5 114,0 8,1 7,5 P2 door close 8,8 16,3 36,4 627,4 652,8 660,2 651,7 607,9 52,2 407,2 21,5 652,4 493,9 551,3 379,5 11,4 9,0 P3 1061,3 894,6 1073,0 754,4 744,0 459,0 133,1 257,3 110,6 74,1 63,9 76,9 374,1 655,5 634,3 326,3 611,3 P3 door close 11,7 11,3 12,2 21,8 20,0 350,3 631,3 301,2 25,4 30,9 604,3 648,4 651,1 613,0 20,4 162,7 11,2

Table 58: RMS delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 P1 1.4 398.9 52.0 127.0 5.8 7.6 4.8 11.9 6.3 11.1 7.1 303.7 517.1 10.5 10.8 16.4 414.4 P1 door close 12.6 1.1 1.6 1.2 145.8 320.9 74.8 422.4 51.9 343.7 6.6 427.4 403.7 431.4 452.3 473.9 453.5 P2 1.4 1.0 2.7 20.1 7.0 7.8 5.6 11.7 6.5 12.0 1.2 11.2 12.2 10.9 328.2 1.8 1.3 P2 door close 1.2 10.2 132.2 402.6 402.1 394.8 402.7 428.8 132.9 450.3 8.9 399.1 448.3 448.4 440.4 6.7 1.5 P3 525.0 611.7 517.3 650.8 654.9 622.3 386.9 511.2 317.4 241.2 255.1 287.0 591.8 662.2 661.1 565.0 655.3 P3 door close 6.6 4.3 3.5 23.4 13.8 412.8 405.2 395.7 2.5 8.4 410.7 395.6 396.4 408.8 7.5 324.6 6.3

In the Table 57 and Table 58 the mean delay and RMS delay at downstairs location is much larger than at upstairs location. It shows more reflection paths at downstairs location. In addition, the RMS and mean delay value turn out to be small when the RX antenna point at the TX while the RMS and mean delay value is large at the other angles due to high mount of reflections when the door of the driver car is closed.

6.3.3 Penetration through the train window In this paragraph we present the results obtain by measuring the penetration losses through the train window as illustrated on Figure 143 (a) and (b). The Tx is located inside the train close to the train window as illustrated on the photography Figure 143 (b). The distance from the Tx and the Rx to the train window are 1.08 m and 0.84 m, respectively. The Tx and Rx antenna height are the same, which is 2.3 m. The Rx location spacing is set to be 2 m. There is a metallic grid between Train 2 and the test train. The distance from the train 1 and the metallic grid to the test train are 5.55 m and 1 m, respectively. The angle 0 represents that Tx and Rx are facing each other. The Receiver takes several positions as shown on scheme Figure 143 (a). For each position we rotate the receiver over 360°. The received power versus angular positon is given on the following Figure 144 (position 1 to position 5). Figure 145 gives a synthetic view of the results. Figure 146 gives the evolution of the received signal level in dBm versus the Tx-Rx distance.

R2R-T2.2-T-DZR-006-18 Page 155 of 186 10/12/2016

Contract No. H2020 – 636032

(a) (b)

Figure 143 (a) and (b): illustration of the measurements configuration for penetration through the window

R2R-T2.2-T-DZR-006-18 Page 156 of 186 10/12/2016

Contract No. H2020 – 636032

Angular power distribution at position 1 Angular power distribution at position 2

Angular power distribution at position 3 Angular power distribution at position 4

Angular power distribution at position 5

Figure 144: Angular power distribution for the five positions of Figure 143 (a)

R2R-T2.2-T-DZR-006-18 Page 157 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 145: Received signal level in dBm for each angle at each of the 5 positions In both Figure 145 and Figure 146, the received signal level is higher at the direction that the RX is pointing to the TX. Strong reflections are found at the angle between [210°, 230°] degrees due to the surroundings.

The following Table 59 and Table 60 summarize the values of Mean delay and RMS delay in ns for different angle positions.

Table 59: Mean delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 P1 6.3 6.4 6.3 32.1 17.1 11.7 8.8 8.4 11.0 8.1 8.5 9.4 13.5 6.3 6.4 6.4 6.2 P2 17.4 14.1 7.8 9.9 24.2 21.3 9.9 10.2 10.2 12.7 22.5 21.0 17.1 23.1 24.0 23.8 21.3 P3 12.3 10.2 5.8 5.6 5.5 8.1 9.0 7.1 7.3 8.1 9.1 12.7 21.7 39.9 22.4 18.0 12.9 P4 10.9 9.1 8.0 7.9 7.8 4.6 4.9 13.7 8.2 8.8 10.8 41.3 26.9 46.9 262.3 30.6 13.6 P5 527.3 22.2 17.8 9.2 12.7 11.0 13.7 39.0 240.7 19.3 25.5 34.3 81.0 331.5 185.9 529.6 231.1

Table 60: RMS delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 P1 1.3 1.1 1.1 101.7 3.9 3.2 1.5 1.0 8.7 1.3 0.9 1.2 7.0 1.2 0.8 0.8 1.3 P2 6.6 3.2 0.8 5.7 12.3 13.3 1.0 1.1 1.3 8.7 18.2 8.5 1.7 14.4 8.7 4.6 8.7 P3 2.9 3.1 1.0 1.3 1.5 3.2 4.3 1.4 1.0 1.6 1.5 1.4 8.7 21.9 6.1 2.5 2.3 P4 4.9 1.3 0.9 1.3 1.5 1.5 1.5 8.8 2.3 2.1 1.5 29.0 18.7 21.1 389.8 39.4 6.2 P5 456.3 14.9 10.0 1.2 9.5 2.6 7.9 62.1 449.5 9.0 15.3 22.0 165.1 421.4 265.6 470.7 400.0

R2R-T2.2-T-DZR-006-18 Page 158 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 146: Evolution of received signal level for Tx-Rx distance and for each angle The mean delay and RMS delay become larger gradually with the increase of the Tx-Rx distance, which means that more radio waves are reflected accordingly.

6.3.4 Outdoor to Driver’s room Virtual coupling is considered to be feasible in the further railway. To support it, it is important to investigate the radio channel for this application. One of these scenarios is to mount the system antenna inside the car. This Outdoor-to-Driver (O2D) scenario is designed to study this case.

In this scenario, the transmitter is located inside the driver cabin as illustrated in the photography Figure 147 (b). The receiver takes several positions outside the cabin (cf. scheme on Figure 147 (a)). Both the Tx and the Rx antenna height are 2.6 m. 7 Rx locations are measured with a location spacing of 3 m. The distance from location to the side-line of the trail is 0.4 m. The distance between the first location to the Tx is set to be 5 m. The Rx antenna is swept between [-22°, +22°] since the vibration of the train during the trip won’t be more that this range. Here, 0° corresponds to the facing direction to the Tx.

R2R-T2.2-T-DZR-006-18 Page 159 of 186 10/12/2016

Contract No. H2020 – 636032

(a) (b)

Figure 147: (a) scheme of the measurements- (b) photography of the scenario The following Figure 148 (a) to (f) illustrate the angular power distribution for each of the position of the transmitter (-22°, 0°, +22°).

Angular power distribution at position 1 Angular power distribution at position 2

R2R-T2.2-T-DZR-006-18 Page 160 of 186 10/12/2016

Contract No. H2020 – 636032

Angular power distribution at position 3 Angular power distribution at position 4

Angular power distribution at position 5 Angular power distribution at position 6

Angular power distribution at position 7

Figure 148: Angular power distribution for (-22°, 0°, 22°) for each of the 7 positions outside the cabin as indicated on scheme Figure 147 (a) The following Figure 149 presents a summary of the previous curves. The Table 61 and Table 62, summarize the values observed for the mean delay and the RMS delay.

R2R-T2.2-T-DZR-006-18 Page 161 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 149: Received signal level for each position and each angle We observe that from position P1 to P5 (up to 17 m), the highest RSLs are received at the angle of 0 degree. When the Tx-Rx distance is more than 17 m, the highest RSLs are received at the angel of -220. However, the difference between the largest RSL and the smallest RSL are within 5 dB.

Table 61: Mean delay (ns) versus angle position Degree -22 0 22 P1 8,6 8,5 8,4 P2 7,6 8,2 7,7 P3 15,8 7,7 7,7 P4 9,7 7,6 7,7 P5 501,6 8,4 8,0 P6 555,9 7,1 7,3

P7 467,8 8,4 16,5 Table 62: RMS delay (ns) versus angle position Degree -22 0 22 P1 1.1 1.0 1.0 P2 1.4 1.3 1.2 P3 60.6 0.8 1.1 P4 19.8 1.3 1.1 P5 431.4 1.1 1.1 P6 424.2 1.1 1.0 P7 451.5 1.2 14.2

The Table 62 and Table 63 show that the RMS delay and the Mean delay at large Tx-Rx distance (Position P5, P6 and P7) are much larger at the angle of -22°, which means large reflections from surrounds are incident at these setups.

R2R-T2.2-T-DZR-006-18 Page 162 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 150: Received signal level versus Tx-Rx distance for each angle (-22°, 0°, +22°) Figure 150 shows that the RSL difference between different angles decrease with the increase of the Tx-Rx distance, which means the antenna directivity in 60 GHz band at short Tx-Rx distances is sensitive. A larger antenna pattern can be needed for this case.

6.3.5 Outdoor to Driver (Centre) This scenario is designed for the virtual coupling as the previous one, in which the transmitter is located inside the driver cabin. The only difference is that Rx is located in the middle of the track (see Figure 147 a). Similarly, both the Tx and the Rx antenna height are 2.6 m. 7 Rx locations are measured with a location spacing of 3 m. The distance between the first location to the Tx is set to be 5 m. The Rx antenna is swept between [-22°, +22°] since the vibration of the train during the trip will not be more than this range. Figure 151 illustrates the different received power values for the seven positions. Figure 152 is a summary of Figure 151. The mean delay and RMS delay for each position and each angle is given in Table 63 and Table 64. Figure 153 presents the received signal level versus distance for the angle of [-22°, 0°, +22°].

Angular power distribution at position 1 Angular power distribution at position 2

R2R-T2.2-T-DZR-006-18 Page 163 of 186 10/12/2016

Contract No. H2020 – 636032

Angular power distribution at position 3 Angular power distribution at position 4

Angular power distribution at position 5 Angular power distribution at position 6

Angular power distribution at position 7

Figure 151: Angular power distribution for (-22°, 0°, 22°) for each of the 7 positions outside the cabin as indicated on scheme Figure 147 (a), on the centre

R2R-T2.2-T-DZR-006-18 Page 164 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 152: Received signal level for each position and each angle (-22°, 0°, +22°) Figure 152 shows thatthe highest RSLs are all received at the angle of 0 degree. The difference between the largest RSL and the smallest RSL are within 12 dB. It can be concluded that a dedicated antenna pattern pointing to the Tx is efficient for this scenario.

Table 63: Mean delay (ns) versus angle position Degree -22 0 22 P1 6.6 6.8 6.8 P2 7.0 6.9 6.9 P3 7.1 7.1 6.5 P4 7.3 7.1 7.1 P5 410.9 7.7 7.8 P6 402.6 8.0 7.6 P7 297.8 7.8 7.8 Table 64: RMS delay (ns) versus angle position Degree -22 0 22 P1 0.9 1.0 1.3 P2 1.3 1.6 0.9 P3 1.7 1.3 1.5 P4 1.2 1.1 0.8 P5 451.7 0.9 0.8 P6 470.3 1.0 0.9 P7 424.5 1.3 1.5

In the above two tables we observe that the RMS delay and the Mean delay at large Tx-Rx distance (Position P5, P6 and P7) are much larger at the angle of -22°, which is consistent with the previous scenario.

R2R-T2.2-T-DZR-006-18 Page 165 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 153: Received signal level versus distance for each angle (-22°, 0°, +22°) Similarly, the above Figure 153 shows that the RSL difference between different angles decrease with the increase of the Tx-Rx distance. It will be more efficient to point antenna beam towards the Tx than any other angle at all the locations.

6.3.6 Outdoor to Outdoor (O2O) It is feasible that Tx and the Rx are outside the train near the mechanical coupling system (see photography on Figure 154 (b) in order to simulate a future transmission in the context of virtual coupling. In this scenario, the receiver is located at several positions on the same longitudinal axis as indicated on the scheme Figure 154 (a). Both the Tx and the Rx antenna height are 1.14 m. 7 Rx locations are measured with a location spacing of 3 m. The distance from location to the side- line of the trail is 0.4 m. The distance between the first location to the Tx is set to be 4 m. The Rx antenna is swept between [-22°, +22°].

R2R-T2.2-T-DZR-006-18 Page 166 of 186 10/12/2016

Contract No. H2020 – 636032

(a) (b)

Figure 154: Outdoor-Outdoor measurements to simulate transmission between consist

Angular power distribution at position 1 Angular power distribution at position 2

Angular power distribution at position 3 Angular power distribution at position 4

R2R-T2.2-T-DZR-006-18 Page 167 of 186 10/12/2016

Contract No. H2020 – 636032

Angular power distribution at position 5 Angular power distribution at position 6

Angular power distribution at position 7

Figure 155: Outdoor-Outdoor measurements – Angular power distribution at the 7 positions

Figure 156: Summary of received power level for the 7 positions and for [-22°, 0°, +22°] We can observe on Figure 156 that most of the peak RSLs (except P4) are all received at the angle of 0 degree. The difference between the largest RSL and the smallest RSL are within 5 dB. It can be concluded that a dedicated antenna pattern pointing to the Tx is efficient for this scenario.

R2R-T2.2-T-DZR-006-18 Page 168 of 186 10/12/2016

Contract No. H2020 – 636032

Table 65: Mean delay (ns) versus angle position Degree -22 0 22 P1 7.8 7.9 7.9 P2 7.9 8.1 7.9 P3 8.5 8.5 8.3 P4 9.2 9.4 8.9 P5 7.2 7.1 7.1 P6 8.6 8.2 8.1 P7 8.3 8.3 8.7 Table 66: RMS delay (ns) versus angle position Degree -22 0 22 P1 1.2 1.2 1.1 P2 1.0 1.1 1.4 P3 0.8 1.2 1.3 P4 1.2 1.3 0.7 P5 0.8 1.3 1.3 P6 1.5 1.0 0.7 P7 1.4 1.0 1.4

The above Table 65 and Table 66 show that the RMS delay and the Mean delay is small and consistent, which shows that the RMS delay and the Mean delay is uncorrelated with the Tx-Rx distance. It can also be concluded few reflections from environment are found.

Figure 157: Received power level versus the distance for [-22°, 0°, +22°] We can observe on the above Figure 157 that there is a big fading at P4 for all the angles. It will be more efficient that the antenna pattern is pointing to Tx than the other angles at all the locations.

R2R-T2.2-T-DZR-006-18 Page 169 of 186 10/12/2016

Contract No. H2020 – 636032

6.3.7 Train to Infrastructure As described in [93], two kinds of links between the access points (APs)/ transceivers of the train and the infrastructures of fixed networks are required. These links have following requirements: high data rates, low latencies, bidirectional streams, less than 100 ms latencies and 98–99 percent availability. It will be important to exploit the feasibility of mmw system to fulfil these requirements. Therefore, in this configuration, the transmitter is located on the roof of the train as illustrated on the photography given on Figure 158 (b). The receiver is situated on the ground and takes different positions as shown on the scheme given in Figure 158 (a).

Both the Tx and the Rx antenna height are 4.5 m. The azimuth angles at the Rx antenna were swept in 22° steps from 0 to 360°, while the Tx antenna is fixed on the roof of the train. All the location and surrounding information is given in detail in the Figure 158 (a). Figure 159 gives the angular power distribution at the 5 different positions. Then in Figure 160 we present the received power level for each angle for the 5 positions. Table 67 and Table 68 give the values of the mean delay and RMS delay for each angle at each position. We remind that for each transmitter position, the antenna is rotated overs 360°. 0° corresponds to the case where Tx and Rx are in LOS (same axis for the antennas).

(a) (b)

Figure 158: Transmitter on the roof of the train

R2R-T2.2-T-DZR-006-18 Page 170 of 186 10/12/2016

Contract No. H2020 – 636032

Angular power distribution at position 1 Angular power distribution at position 2

Angular power distribution at position 3 Angular power distribution at position 4

Angular power distribution at position 5

Figure 159: Angular power distribution at the 5 positions

R2R-T2.2-T-DZR-006-18 Page 171 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 160: Received signal level in dBm versus angles in degree for the 5 positions We can observe on Figure 159 and Figure 160 that the RSL is concentrating on the direction of the line of sight (LOS) between the Tx and the Rx except the location 5, where the metallic pole blocks the LOS of the radio link. At location 5, strong reflections are found at the angles ranging in [330°, 350°]. Beamforming or directional antenna pattern is needed under the LOS conditions.

Table 67: Mean delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 p1 7.8 7.7 10.3 71.0 291.7 459.3 240.5 286.2 217.9 229.7 117.1 62.1 88.6 103.6 28.5 13.7 7.8 p2 8.6 8.4 14.1 423.6 237.2 372.9 190.8 57.2 65.9 62.6 119.5 249.0 169.8 168.6 18.6 27.7 107.3 p3 8.0 13.4 15.3 597.6 538.4 292.1 180.3 77.2 46.8 48.2 69.6 427.0 275.5 7.6 7.7 8.0 8.1 p4 446.2 549.1 603.2 614.6 614.4 358.7 257.5 250.5 62.9 29.3 49.6 106.0 244.7 292.0 139.9 384.5 731.0 p5 37.1 35.7 164.1 444.9 593.8 218.4 263.1 249.0 183.8 160.4 147.7 18.3 16.9 8.6 7.6 7.8 7.7

Table 68: RMS delay (ns) versus angle position Degree 0 22 44 66 88 110 132 154 176 198 220 242 264 286 308 330 352 P1 1.3 1.2 10.0 147.5 416.6 455.1 168.4 265.3 27.1 117.1 104.8 24.3 65.8 137.0 0.8 9.7 1.1 P2 1.0 0.8 12.6 374.2 229.6 346.6 203.7 1.7 20.1 39.5 98.2 288.3 277.3 310.6 1.0 27.4 246.2 P3 1.1 5.9 6.3 398.9 378.2 164.4 125.3 69.4 0.8 1.3 54.5 404.2 314.0 1.2 1.1 1.0 1.4 P4 463.4 397.8 395.2 395.2 392.7 251.2 31.4 92.6 90.3 62.4 75.2 100.7 272.3 313.4 280.1 469.1 560.7 P5 94.6 87.9 173.8 402.6 390.6 90.7 78.9 90.1 97.5 100.9 221.2 19.4 17.8 1.6 0.9 1.1 0.9

Table 67 and Table 68 show that the RMS delay and the Mean delay turn out to be large if the angles are different from the LOS directions. It can also be found that at P4 location (NLOS), the mean delay and RMS delay are much larger than the rest of positons, which proves that close metallic surrounding can greatly change the radio propagation in the 60 GHz band. In addition, large reflections are found in this factory.

R2R-T2.2-T-DZR-006-18 Page 172 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 161: Received power level versus the distance for [0°, 22°, 44°,66°,88°,110°,132°,154°,176°,198°,220°,242°,264°,286°,308°,330°,352°] Figure 161 shows that the largest RSL difference between different angles is within 15 dB and uncorrelated with the Tx-Rx distance.

6.3.8 Outdoor to Roof

Figure 162: Illustration of the Outdoor-to-roof scenario In this scenario, the transmitter is located on the roof of the train while the receiver is situated at the same height to simulate the virtual coupling when both Tx and Rx antennas are installed on the roof of the two cars. It needs to be mentioned that in our measurement campaign, there is only one car in the station which can be used in our measurement campaign. However, the Outdoor-to-Roof (O2R) still provide good information for the virtual coupling. The transmitter is located on the roof of the train as illustrated on the photography given on Figure 163 (b). The receiver is situated on the ground and takes different positions as shown on the scheme given in Figure 163 (a). Both the Tx

R2R-T2.2-T-DZR-006-18 Page 173 of 186 10/12/2016

Contract No. H2020 – 636032

and the Rx antenna height are 4.5 m. The azimuth angles at the Rx antenna were swept in 220 steps from 00 to 3600, while the Tx antenna is fixed on the roof the train. All the location and surrounding information is given in detail in the Figure 163 (a).

Figure 164 gives the angular power values for 3 angles (-22°, 0°, +22°) for each transmitter position in green (Figure 163 (a)). Figure 165 summarises the results regarding the received signal level in dBm for each angle and each position. Table 69 and Table 70 give the mean value for the delay and table give the RMS delay for each position and each angle.

(a) (b)

Figure 163: Illustration of measurement scenario for Roof to Roof configuration

Angular power distribution for position 1 Angular power distribution for position 2

R2R-T2.2-T-DZR-006-18 Page 174 of 186 10/12/2016

Contract No. H2020 – 636032

Angular power distribution for position 3 Angular power distribution for position 4

Angular power distribution for position 5 Angular power distribution for position 6

Angular power distribution for position 7

Figure 164: Angular power distribution for the seven positions

R2R-T2.2-T-DZR-006-18 Page 175 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 165: Received signal in dBm for the 3 angles (-22°, 0°, +22°) and for the 7 positions Figure 165 shows that the highest RSLs is measured at the angles of 0° and 22°. The difference between the largest RSL and the smallest RSL are within 4 dB, which is ignorable.

Table 69: Mean delay (ns) versus angle position Degree -22 0 22 P1 8.4 8.5 8.5 P2 7.1 7.2 7.0 P3 7.1 7.3 7.4 P4 8.1 8.4 8.4 P5 6.8 6.7 7.1 P6 8.2 8.4 8.1 P7 8.5 8.7 8.6 Table 70: RMS delay (ns) versus angle position Degree -22 0 22 P1 1.0 1.4 1.1 P2 1.0 1.0 1.1 P3 1.0 1.0 1.3 P4 1.2 1.4 1.3 P5 1.0 1.1 1.4 P6 1.0 1.0 1.2 P7 1.3 1.2 1.5

Similar with the O2O scenario, the two Table 69 and Table 70 highlight that the RMS delay and the Mean delay are small and consistent, which shows that the RMS delay and the Mean delay is uncorrelated with the Tx-Rx distance. It can also be concluded that few reflections from environment are found.

R2R-T2.2-T-DZR-006-18 Page 176 of 186 10/12/2016

Contract No. H2020 – 636032

Figure 166: Received signal level in dBm versus Tx-RX distance for each angle (-22°, 0°, +22°) In Figure 166 we observe that the RSL difference at the angle of 0° and 22° is small. However, the RSL difference between these two angles and -22o increase with the increase of the Tx-Rx distance.

R2R-T2.2-T-DZR-006-18 Page 177 of 186 10/12/2016

Contract No. H2020 – 636032

7. CONCLUSIONS The Roll2Rail project covers different rolling stock topics. The goal of WP2 is to make research on technologies and architectures to provide fundamentals for a train communication systems based on Wireless Transmission for Train Control and Monitoring System (TCMS).

This deliverable describes the motivation and methods for the characterization of the railway environment for radio transmission. An overview of existing measurements and channel models in the railway domain is given and the gaps for a detailed characterization of the radio environment are identified. Since channel measurements require a considerable technical effort and experience, the methodology for channel sounding is described in some detail. The measurements preparations are described in detail in Chapter 4. For the HSR measurements the RUSK DLR Channel Sounder was used, dynamic wideband (120 MHz) measurements at 5.2 GHz were performed. In the METRO environment a pulse-based Channel Sounder was used and static as well as dynamic measurements with wideband (80 MHz) and narrowband settings at 2.6 GHz were performed. Table 27 gives an overview of the different settings and measurements. The use of a narrower bandwidth compared to the HSR RUSK measurements means that the obtained characterization has lower temporal accuracy for multipath characterization (12.5 ns vs 8.33 ns). The mmWave measurements were performed with IFSTTAR’s 60 GHz Channel Sounder. Static wideband measurements at 60 GHz were performed in a production hall.

A strong effort both in terms of equipment as well as trained staff is required to execute measurement campaigns. Chapter 5 describes the actual three different measurement campaigns carried out in the field. These include measurements with one or two trains for the high-speed line and the urban/metro measurements in the operational environment as well as stationary measurements in a depot for the 60 GHz measurements and a regional train. A HSR or Metro is not a laboratory, therefore difficulties like the need of extra-long cables, railway certified equipment (antennas, etc.), or the transport and installation of fragile measurement equipment occur. A common problem is the power supply; these vehicles are not designed to support measurement equipment that is energy consuming and sensitive to power interference (EMC). In addition, mobility support and real-time data recording are always challenges for the channel measurements. For outdoor scenarios such as virtual coupling, it is challenging to perform measurements with vehicle vibrations and weather changes, which are quite sensitive in the mmWave Bands.

Further measurements are needed for investigations on ultra-reliable, MIMO and mmWave wireless communications for railways. These may include additional measurement data for different scenarios and environments, trains and carrier frequencies as well as different applications next to wireless TCMS.

Chapter 6 reports the post processing and analysis of the channel measurements. For the high speed line measurements, the large scale and small scale fading parameters are estimated in Section 6.1 for the inter-vehicle channel measurements and the parameters for a tapped delay line channel model are provided. Similar, Section 6.2 provides the large scale and small scale fading parameters for a Metro channel model for inter-consist, intra-consist, and train-to-train scenarios as well as parameters for a tapped delay line channel model for the first two scenarios. Finally, Section 6.3 analyses and estimates parameters for a regional train at 60 GHz carrier frequency for various measurements scenarios ranging from intra-vehicle to outdoor-to-roof.

R2R-T2.2-T-DZR-006-18 Page 178 of 186 10/12/2016

Contract No. H2020 – 636032

To conclude, the results of Chapter 6 provide the basis for the channel models which are used in simulations in Task 2.7 to evaluate selected wireless technologies and in channel emulators in Task 2.8 to validate suitable radio technologies for wireless TCMS.

In case of HSR and Metro measurements the tapped delay line channel models of roof top measurements will be used as input for Physical Layer simulations in T2.7. As discussed in T2.7 meeting, a mapping-table of channel models with channel parameters for different scenarios will be evaluated in different Physical Layer link simulations to obtain system performance parameters in terms of delay time and packet error rate, which can be further evaluated as inputs in higher-layer system simulations.

The results of Chapter 6 show that the HSR, the METRO and the mmWave measurements have been successful to provide the necessary generic tapped delay line channel models for evaluating wireless TCMS. Further analysis of the measurement data beyond Roll2Rail will confirm this and enable new applications such as virtual coupling in addition to wireless TCMS.

R2R-T2.2-T-DZR-006-18 Page 179 of 186 10/12/2016

Contract No. H2020 – 636032

8. REFERENCES

[1] IEEE-SA Standards Board, IEEE STd 1473-1999 Standard for Communications Protocol Aboard Trains, IEEE, 1999.

[2] IEC, IEC 61375-1Electric railway equipment – Train bus, IEC, 1999.

[3] M. Wahl, “A Survey of railway embedded network solutions. Towards the use of Industrial Ethernet technologies,” Les collections de I'INRETS, p. 61, March 2010.

[4] D. Van Den Abeele, M. Wahl and M. Berbineau, “Procédé de transfert de données d’alerte entre un véhicule ferroviaire en panne et un centre de contrôle, dispositif associé,” ALTSTOM, 2010.

[5] Y. E Hilali, C. Tatkeu, P. Deloof, L. Sakkila, A. Rivenq and J. M. Rauvaen, “Enhanced high data rate communication system using embedded cooperative radar for intelligent transports systems,” Transport research Part C, pp. 429-439, 2010.

[6] “IGR-D-WIF-007-06 – Ultra Wide Band Application report,” INTEGRAIL report.

[7] M. Di Benedetto and G. Giancola, “Understanding UWB radio fundamentals,” Prentice Hall, 2004.

[8] J. Cellmer, "Le réseau GSM-R de RFF," in REE, pp. 45-52.

[9] M. Fitzmaurice, “Use of 2.4 GHz frequency band for Communications Based Train Control data communications systems,” in Rail Conference, Atlanta, 2006.

[10] N. Gresset and H. Bonneville, “Fair Preemption for Joint Delay Constrained and Best Effort Traffic Scheduling in Wireless Networks,” in Communication Technologies for Vehicles, vol. Volume 9066 of the series Lecture Notes in Computer Science, Springer, pp. 141-152.

[11] UNISIG, “ERTMS/ETCS - FFFIS for Eurobalise,” 2012.

[12] I. Mitola and G. J. Maguire, “Cognitive radio: making software radios more personal,” IEEE Personal Communications, pp. 13-18, August 1999.

[13] J. Palicot, De la radio logicielle a la radio intelligente, Hermes, 2010.

[14] IFSTTAR, “Cognitive Radio for Railway through Dynamic and Opportunistic spectrum Reuse,” 2012. [Online]. Available: http://corridor.ifsttar.fr/. [Accessed 2015].

[15] J. Avella Castiblanco, M. Berbineau, M. Ney, I. Massy und F. Gallée, „Time-domain methods to model electromagnetic wave propagation in confined environments,“ in Wave Propagation, Academy Publish, 2012.

[16] J. Avella Castiblanco, D. Seetharamdoo, M. Berbineau, M. Ney und F. Gallée, „Electromagnetic Wave Propagation Modeling for Finding Antenna Specifications and Positions in Tunnels of Arbitrary Cross-

R2R-T2.2-T-DZR-006-18 Page 180 of 186 10/12/2016

Contract No. H2020 – 636032

section, in Wave Propagation Theories and Applications,“ Y. Zheng, Hrsg., InTech, 2013, pp. 261-288.

[17] J. Maurer, T. Fugen, T. Schafer and W. Wiesbeck, “A new inter-vehicle communications (IVC) channel model,” in IEEE Veh. Technol. Conf. 2004 Fall, 2004.

[18] Y. Chang, M. Furukawa, H. Suzuki and K. Fukawa, “Propagation Analysis with Ray Tracing Method for High Speed Trains Environment at 60 GHz,” in IEEE VTC Spring, 2015.

[19] E. Arruti, M. Mendicute and J. Thompson, “A Novel WINNER Based Model for Wireless Communications inside Train Carriages,” in European Modelling Symposium (EMS), 2013.

[20] “Emag Technologies Inc.,” [Online]. Available: http://www.emagtech.com/products/emagware/emcube.

[21] D. Didascalou, Ray optical wave propagation modelling in arbitrarily shaped tunnels, PhD Dissertation, University of Karlsruhe, 2000.

[22] E. Masson, P. Combeau, Y. Cocheril, L. Aveneau, M. Berbineau und R. Vauzelle, „Ray Launching Modeling in Curved Tunnels with Rectangular or Non Rectangular Section,“ in Wave Propagation Theories and Applications, Y. Zheng, Hrsg., 2013.

[23] J. Karedal, F. Tufvesson, N. Czink, A. Paier, C. Dumard, T. Zemen, C. Mecklenbräuker and A. Molisch, “A Geometry-Based Stoachastic MIMO Model for Vehicle-to-Vehicle Communications,” IEEE Transactions on Wireless Communications, vol. 8, no. 7, pp. 3646-3657, July 2009.

[24] C. Mecklenbräuker, A. Molisch, J. Karedal, F. Tufvesson, A. Paier, L. Bernado, T. Zemen, O. Klemp and N. Czink, “Vehicular Channel Characterization and Its Implications for Wireless System Design and Performance,” Proceedings of the IEEE, vol. 99, no. 7, pp. 1189-1212, July 2011.

[25] X. Cheng, Q. Yao, M. Wen, C.-X. Wang, L.-Y. Song and B.-L. Jiao, “Wideband Channel Modeling and Intercarrier Interference Cancellation for Vehicle-to-Vehicle Communication Systems,” IEEE Journal on Selected Areas in Communications, vol. 31, no. 9, pp. 434-448, September 2013.

[26] A. Ghazal, C.-X. Wang, H. Haas, M. Beach, X. Lu, D. Yuan and X. Ge, “A Non-Stationary MIMO Channel Model for High-Speed Train Communication Systems,” in IEEE VTC Spring, 2012.

[27] G. Acosta-Marum and M. A. Ingram, “Six time- and frequency-selective empirical channel models for vehicular wireless LANs,” in IEEE VTC Fall, 2007.

[28] G. Acosta-Marum and M. Ingram, “Doubly selective vehicle-to-vehicle channel measurements and modeling at 5.9 GHz,” in Int. Symp. Wireless Personal Multimedia Commun., 2006.

[29] I. Sen and D. W.Matolak, “Vehicle–vehicle channel models for the 5-GHz band,” IEEE Trans. Intell. Transp. Syst., vol. 9, no. 2, pp. 235-245, 2008.

[30] Q. Wu, D. Matolak and I. Sen, “5-GHz-Band Vehicle-to-Vehicle Channels: Models for Multiple Values of

R2R-T2.2-T-DZR-006-18 Page 181 of 186 10/12/2016

Contract No. H2020 – 636032

Channel Bandwidth,” IEEE Transactions on Vehicular Technology, vol. 59, no. 5, pp. 2620-2625, 2010.

[31] H. S. Wang and N. Moayeri, “Finite-state Markov channel-a useful model for radio communication channels,” IEEE Transactions on Vehicular Technology, vol. 44, no. 1, pp. 163-171, 1995.

[32] E. N. Gilbert, “Capacity of a burst-noise channel,” Bell Syst. Tech. J., vol. 39, pp. 1253-1265, 1960.

[33] E. 0. Elliott, “Estimates of error rates for codes on burst-noise channels,” Bell Syst. Tech. J., vol. 42, pp. 1977-1997, 1963.

[34] H. Wang, F. Yu, L. Zhu, T. Tang and B. Ning, “Finite-state Markov modeling of tunnel channels in communication-based train control (CBTC) systems,” in IEEE ICC, 2013.

[35] S. Lin, Y. Li, Y. Li, B. Ai and Z. Zhong, “Finite-state Markov channel modeling for vehicle-to- infrastructure communications,” in IEEE 6th International Symposium on Wireless Vehicular Communications (WiVeC), 2014.

[36] Z. Min, W. Muqing, S. Yanzhi, Y. Deshui, D. Shiping, Z. Panfeng, Z. Xiangbing and G. Shuyun, “Analysis and Modeling for Train-Ground Wireless Wideband Channel of LTE on High-Speed Railway,” in IEEE VTC Spring, 2013.

[37] Z. Min, W. Muqing, H. Qian, D. Shiping, Z. Panfeng, Z. Xiangbing and G. Shuyun, “A two-dimensional integrated channel model for train-ground communications,” in International Conference on Telecommunications (ICT), 2014.

[38] T. Zhou, C. Tao, L. Liu, J. Qiu and R. Sun, “High-speed railway channel measurements and characterizations: A review,” Journal of Modern Transportation, vol. 20, pp. 199-205, 2012.

[39] L. Liu, C. Tao, H. Chen, T. Zhou, R. Sun and J. Qiu, “Survey of wireless channel measurement and characterization for high-speed railway scenarios,” Tongxin Xuebao/Journal on Communications, vol. 35, pp. 115-127, 2014.

[40] L. Liu, C. Tao, J. Qiu, H. Chen, L. Yu, W. Dong and Y. Yuan, “Position-based modeling for wireless channel on high-speed railway under a viaduct at 2.35 GHz,” IEEE Journal on Selected Areas in Communications, pp. 834-845, 2012.

[41] Y. Zhang, Z. He, W. Zhang, L. Xiao and S. Zhou, “Measurement-Based Delay and Doppler Characterizations for High-Speed Railway Hilly Scenario,” International Journal of Antennas and Propagation, p. 8, 2014.

[42] L. Tian, J. Zhang and C. Pan, “Small scale fading characteristics of wideband radio channel in the U- shape cutting of high-speed railway,” in IEEE Vehicular Technology Conference, 2013.

[43] Z. Z. B. A. J. D. Ruisi He, „An Empirical Path Loss Model and Fading Analysis for High-Speed Railway Viaduct Scenarios,“ IEEE Antennas and Wireless Propagation Letters, Bd. 10, pp. 808-812, 2011.

R2R-T2.2-T-DZR-006-18 Page 182 of 186 10/12/2016

Contract No. H2020 – 636032

[44] N. Kita, T. Ito, S. Yokoyama, T. Ming-Chien, Y. Sagawa, M. Ogasawara and M. Nakatsugawa, “Experimental study of propagation characteristics for wireless communications in high-speed train cars,” in 3rd European Conference on Antennas and Propagation (EuCSAP), 2009.

[45] T. Ito, N. Kita, W. Yamada, T. Ming-Chien, Y. Sagawa, M. Ogasawara, M. Nakatsugawa and T. Sugiyama, “Study of propagation model and fading characteristics for wireless relay system between long-haul train cars,” in 2011, 5th European Conference on Antennas and Propagation (EuCAP).

[46] D. Weihui, L. Guanyi, Y. Li, D. Haiyu and Y. Jianhua, “Channel properties of indoor part for high-speed train based on wideband channel measurement,” in 5th International ICST Conference on Communications and Networking in China (CHINACOM), 2010.

[47] A. Mariscotti, A. Marrese, N. Pasquino and R. S. L. Moriello, “Characterization of the radio propagation channel aboard trains for optimal coverage at 2.45 GHz,” in IEEE International Workshop on Measurements and Networking Proceedings (M&N), 2013.

[48] M. Gruden, A. Westman, J. Platbardis, P. Hallbjorner and A. Rydberg, “Reliability experiments for wireless sensor networks in train environment,” in European Wireless Technology Conference (EuWIT), 2009..

[49] D. Matolak and A. Chandrasekaran, “5 GHz Intra-Vehicle Channel Characterization,” in IEEE Vehicular Technology Conference (VTC Fall) 2012, Quebec City, QC, Canada, 2012.

[50] K. Chetcuti, C. Debono, R. Farrugia and S. Bruillot, “Wireless propagation modelling inside a business jet,” in IEEE EUROCON 2009, St.-Petersburg, Russia, 2009.

[51] S. Chiu, J. Chuang and D. Michelson, “Characterization of UWB Channel Impulse Responses Within the Passenger Cabin of a Boeing 737-200 Aircraft,” IEEE Transactions on Antennas and Propagation, vol. 58, no. 3, pp. 935-945, 2010.

[52] N. Diaz and J. Esquitino, “Wideband channel characterization for wireless communications inside a short haul aircraft,” in IEEE 59th Vehicular Technology Conference (VTC Spring) 2004, Milan, Italy, 2004.

[53] R. Errico and L. Rudant, “UHF radio channel characterization for Wireless Sensor Networks within an aircraft,” in 5th European Conference on Antennas and Propagation (EUCAP), Rome, Italy, 2011.

[54] M. Jacob, K. L. Chee, I. Schmidt, J. Schuur, W. Fischer, M. Schirrmacher and T. Kurner, “Influence of passengers on the uwb propagation channel within a large wide-bodied aircraft,” in 3rd European Conference on Antennas and Propagation (EuCAP) 2009, Berlin, Germany, 2009.

[55] D. Matolak and A. Chandrasekaran, “Aircraft intra-vehicular channel characterization in the 5 GHz band,” in Integrated Communications, Navigation and Surveillance Conference (ICNS) 2008, Bethesda, MD, USA, 2008.

[56] N. Moraitis and P. Constantinou, “Radio Channel Measurements and Characterization inside Aircrafts for In-Cabin Wireless Networks,” in IEEE 68th Vehicular Technology Conference (VTC-Fall) 2008,

R2R-T2.2-T-DZR-006-18 Page 183 of 186 10/12/2016

Contract No. H2020 – 636032

Calgary, BC, Canada, 2008.

[57] I. Schmidt, J. Jemai, R. Piesiewicz, R. Geise, M. Schwark, T. Kurner, M. Schirrmacher and P. Thielker, “UWB propagation channels within an aircraft and an office building environment,” in IEEE Antennas and Propagation Society International Symposium (AP-S) 2008, San Diego, CA, USA, 2008.

[58] C. Spiliotopoulos and A. Kanatas, “Path-Loss and Time-Dispersion Parameters of UWB Signals in a Military Airplane,” IEEE Antennas and Wireless Propagation Letters, vol. 8, pp. 790-793, 2009.

[59] C. Spiliotopoulos, A. Kanatas and G. Efthymoglou, “UWB channel parameters in a C130 airplane,” in IEEE 20th International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) 2009, Tokyo, Japan, 2009.

[60] E. Aguirre, P. Lopez-Iturri, L. Azpilicueta and F. Falcone, “Characterization of wireless channel response in in-vehicle environments,” in 14th Mediterranean Microwave Symposium (MMS), 2014.

[61] M. Shirafune, T. Hikage, T. Nojima, M. Sasaki, W. Yamada and T. Sugiyama, “Propagation characteristic estimations of 2 GHz inter-car wireless links in high-speed train cars in a railway tunnel,” in International Symposium on Antennas and Propagation (ISAP), 2014.

[62] H. Ruisi, Z. Zhandui, B. Ai and D. Jianwen, “An empirical path loss model and fading analysis for high- speed railway viaduct scenarios,” IEEE Antennas and Wireless Propagation Letters, vol. 10, pp. 808- 812, 2011.

[63] A. Lehner, C. Rico García and T. Strang, “On the Performance of TETRA DMO Short Data Service in Railway VANETs,” Springer Science+Busniss Media, 17 May 2012.

[64] C. Rico García, A. Lehner, T. Strang and K. Frank, “Channel Model for Train to Train Communication using the 400 Mhz Band,” IEEE 67th Vehicular Technology Conference (VTC-Spring) 2008, Singapore, 2008.

[65] I. Val, A. Arriola, C. Cruces, R. Torrego, E. Gomez and X. Arizkorreta, "Time-synchronized Wireless Sensor Network for structural helath monitoring applications in railway environments," in IEEE World Conference on Factory Communictation Systems (WFCS), 2015.

[66] "CST Microwave Studio," [Online]. Available: www.cst.com.

[67] European Standard EN 50121-3-2, „Railway applications – Electromagnetic Compatibility, Part 3-2: Rolling stock - Apparatus,“ 2015.

[68] Medav, „RUSK Multidimensional Channel Sounder,“ [Online]. Available: http://www.medav.de/fileadmin/redaktion/documents/English/medav_rusk_mimo.pdf.

[69] E. C. Y. B. M. H. V. R. A. Masson, „MIMO Channel Characterization in Subway Tunnel for Train-to- Wayside applications,“ in 12th International Conference on ITS Telecommunications (ITST) 2012, TAIPEI, TAIWAN, 2012.

R2R-T2.2-T-DZR-006-18 Page 184 of 186 10/12/2016

Contract No. H2020 – 636032

[70] F. B. A. A. G. D. N. R. K. R. B. C. Y. P. H. S. E. Kaltenberger, „Broadband wireless channel measurements for high speed trains,“ in 2015 IEEE International Conference on Communications (ICC), London, UK, 2015.

[71] S. Dudoyer, V. Deniau, R. Adriano, N. Ben Slimen, J. Rioult, B. Meyniel und M. Berbineau, „Study of the Susceptibility of the GSM-R Communications to the Electromagnetic Interferences of the Rail Environment,“ IEEE Transactions on Electromagnetic Compatibility, Bd. 3, pp. 667-676, 2012.

[72] W. Wang, G. Hoerack, T. Jost, R. Raulefs, M. Walter und U.-C. Fiebig, „Propagation Channel at 5.2 GHz in Baltic Sea with Focus on Scattering Phenomena,“ in EuCap, Lisbon, Portugal, 2015.

[73] M. Schwinzerl, T. Jost, F. Pérez-Fontán, M. Schönhuber, W. Wang, M. Walter, T. Pelzmann, G. Obertaxer, N. Floury und R. Prieto-Cerdeira, „Wide-Band Characterization of Antennae Plus Aircraft Platform Patterns in L- and Ka-Band,“ in EuCap, Lisbon, Portugal, 2015.

[74] M. Walter und M. Schnell, „The Doppler-Delay Characteristic of the Aeronautical Scatter Channel,“ in IEEE Vehicular Technology Conference (VTC), San Francisco, 2011.

[75] N. Schneckenburger, D. Shutin, T. Jost, T. Thiasiriphet, M. Walter, A. Filip und M. Schnell, „An Aeronautical Air-Ground Channel Investigation for APNT Applications,“ in ION GNSS, Tampa, USA, 2014.

[76] W. Wang, T. Jost und U.-C. Fiebig, „A Comparison of Outdoor-to-Indoor Wideband Propagation at S- Band and C-Band for Ranging,“ IEEE Transactions on Vehicular Technology, Bd. 64, Nr. 10, pp. 4411- 4421, 2015.

[77] W. Wang, T. Jost, U.-C. Fiebig und W. Koch, „A time variant outdoor-to-indoor channel model for mobile radio based navigation applications,“ International Journal of Antennas and Propagation, 2015.

[78] S. Gligorevic, R. Zierhut, T. Jost und W. Wang, „Airport Channel Measurements at 5.2GHz,“ in EuCAP, Berlin, 2009.

[79] M. Walter, U.-C. Fiebig und A. Zajic, „Experimental Verification of the Non-Stationary Statistical Model for V2V Scatter Channels,“ in IEEE Vehicular Technology Conference (VTC), Fall, Vancouver, Canada, 2014.

[80] A. Lehner und A. Steingass, „Spatial Dynamic Wideband Modeling of the MIMO Satellite-to-Earth Channel,“ International Journal of Antennas and Propagation, 2014.

[81] A. Lehner, Multipath Channel Modelling for Satellite Navigation Systems, Erlangen: PhD Thesis, Universität Erlangen-Nürnberg, 2007.

[82] A. Lehner, C. Rico García, T. Strang und O. Heirich, „Measurement and Analysis of the Direct Train to Train Propagation Channel in the 70 cm UHF-Band,“ in 3rd International Workshop on Communication Technologies for Vehicles, Nets4Cars/Nets4Trains, Oberpfaffenhofen, 2011.

R2R-T2.2-T-DZR-006-18 Page 185 of 186 10/12/2016

Contract No. H2020 – 636032

[83] DLR, „RCAS - Railway Collision Avoidance System,“ [Online]. Available: http://www.collision- avoidance.org/.

[84] Mobile Mark, „Mobile Mark MGRM-UMB-1C Magnet Mount Antenna,“ [Online]. Available: http://www.mobilemark.com/images/spec%20sheets/pg%2018%20RM3-9001900-spec.pdf.

[85] S. Salous, Radio propagation measurements and channel modelling, J. Wiley, 2013.

[86] R. S. Thomä, M. Landmann, A. Richter and U. Trautwein, “Mutlidimensional high-resolution sounding measurement,” in Smart Antennas, EURASIP Book Seris on SP&C, 2005, pp. 241-270.

[87] Medav, “RUSK Channelsounder,” [Online]. Available: http://www.channelsounder.de.

[88] P. Unterhuber, A. Lehner und F. de Ponte Müller, „Measurement and Analysis of ITS-G5 in Railway Environments,“ in 10th International Workshop on Communication Technologies for Vehicles, Nets4Cars Nets4Trains Nets4Aircraft, Donostia - San Sebastian, 2016.

[89] P. Unterhuber, S. Pfletschinger, S. Sand, M. Soliman, T. Jost, A. Arriola, I. Val, C. Cruces, J. Moreno, J. P. García-Nieto, C. Rodríguez, M. Berbineau, E. Echeverría und I. Baz, „A Survey of Channel Measurements and Models for Current and Future Railway Communication Systems,“ Hindawi Mobile Information Systems, 2016.

[90] W. Wei, Channel Measurement and Modeling for Mobile Radio Based Positioning, vol. 34, Aachen: Shaker, 2014.

[91] S. O. Rice, “Mathematical Analysis of Randon Noise,” Bell System Technical Journal, vol. 24, pp. 46-156, 1945.

[92] M. Nakagami, "The m-Distribution—A General Formula of Intensity Distribution of Rapid Fading," Statistical Methods in Radio Wave Propagation, pp. 3-36, 1960.

[93] B. Ai, K. Guan, M. Rupp, T. Kurner, X. Cheng, X.-F. Yin, Q. Wang, G.-Y. Ma, Y. Li, L. Xiong and J.-W. Ding, "Future railway services-oriented mobile communications network," IEEE Communications Magazine, vol. Issue: 10, no. Volume: 53, pp. 78-85, 2015.

[94] J. Karedal, F. Tufvesson, N. Czink, A. Paier, C. Dumard, T. Zemen, C. Mecklenbräuker und A. Molisch, „A Geometry-Based Stoachastic MIMO Model for Vehicle-to-Vehicle Communications,“ IEEE Transactions on Wireless Communications, Bd. 8, Nr. 7, pp. 3646-3657, July 2009.

[95] “Roll2Rail,” 2015. [Online]. Available: http://www.roll2rail.eu/.

R2R-T2.2-T-DZR-006-18 Page 186 of 186 10/12/2016