16.03.2015 Views

Final report on link level and system level channel models - Winner

Final report on link level and system level channel models - Winner

Final report on link level and system level channel models - Winner

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

WINNER D5.4 v. 1.4<br />

IST-2003-507581 WINNER<br />

D5.4 v. 1.4<br />

<str<strong>on</strong>g>Final</str<strong>on</strong>g> Report <strong>on</strong> Link Level <strong>and</strong> System Level Channel Models<br />

Date of Delivery to the CEC: Nov. 18 th , 2005<br />

Author(s):<br />

Participant(s):<br />

Workpackage:<br />

Daniel S. Baum, Hassan El-Sallabi, Tommi Jämsä, Juha Meinilä, Pekka<br />

Kyösti, Xi<strong>on</strong>gwen Zhao, Daniela Laselva, Jukka-Pekka Nuutinen, Lassi<br />

Hentilä, Pertti Vainikainen, Jarmo Kivinen, Lasse Vuokko, Per<br />

Zetterberg, Mats Bengtss<strong>on</strong>, Kai Yu, Niklas Jaldén, Terhi Rautiainen,<br />

Kimmo Kalliola, Marko Milojevic, Christian Schneider, Jan Hansen.<br />

EBIT, EBITT, ETHZ, HUT, KTH, NOK, TUI<br />

WP5 – Channel Modelling<br />

Estimated pers<strong>on</strong> m<strong>on</strong>ths: 66<br />

Security:<br />

Public<br />

Nature:<br />

R<br />

Versi<strong>on</strong>: 1.4<br />

Total number of pages: 167<br />

Abstract: This document presents WINNER <strong>channel</strong> <strong>models</strong>. The <strong>channel</strong> <strong>models</strong> cover WINNER<br />

propagati<strong>on</strong> scenarios for indoor, urban macro-cell <strong>and</strong> micro-cell, stati<strong>on</strong>ary feeder, suburban macro-cell,<br />

<strong>and</strong> rural macro-cell. Both geometric-based stochastic <strong>channel</strong> model <strong>and</strong> reduced-variability (clustered<br />

delay-line) <strong>models</strong> are presented. The <strong>channel</strong> <strong>models</strong> are mainly based <strong>on</strong> measurement data.<br />

Keyword list: Channel modelling, propagati<strong>on</strong> scenarios, wideb<strong>and</strong>, <strong>channel</strong> sounder, cluster delay<br />

domain, angle domain, measurements, delay spread, ray, angle-spread, arrival, departure<br />

Disclaimer:<br />

Page 1 (167)


WINNER D5.4 v. 1.4<br />

Executive Summary<br />

This deliverable presents WINNER <strong>channel</strong> <strong>models</strong> for <strong>link</strong> <strong>level</strong> <strong>and</strong> <strong>system</strong> <strong>level</strong> simulati<strong>on</strong>s of short<br />

range <strong>and</strong> wide area wireless communicati<strong>on</strong> <strong>system</strong>s. The developed <strong>channel</strong> <strong>models</strong> follow guidelines<br />

stated in WINNER deliverable D5.2. The <strong>models</strong> are antenna independent, i.e., different antenna<br />

c<strong>on</strong>figurati<strong>on</strong>s <strong>and</strong> different element patterns can be inserted. The covered propagati<strong>on</strong> scenarios are<br />

indoor small office, urban micro-cell, indoor, stati<strong>on</strong>ary feeder, suburban macro-cell, urban macro-cell,<br />

<strong>and</strong> rural macro-cell. The generic WINNER <strong>channel</strong> model follows a geometric-based stochastic <strong>channel</strong><br />

modelling approach, which allows creating of virtually unlimited double directi<strong>on</strong>al radio <strong>channel</strong> model.<br />

Clustered delay line <strong>models</strong> have also been created for calibrati<strong>on</strong> <strong>and</strong> comparis<strong>on</strong> of different<br />

simulati<strong>on</strong>s. The developed <strong>models</strong> are based <strong>on</strong> both literature <strong>and</strong> extensive measurement campaigns<br />

that have been carried out within the WINNER project.<br />

Page 2 (167)


WINNER D5.4 v. 1.4<br />

Authors<br />

Partner Name Ph<strong>on</strong>e / Fax / e-mail<br />

ETHZ Daniel S. Baum Ph<strong>on</strong>e: +41 44 632 2791<br />

Fax: +41 44 632 1209<br />

E-mail: dsbaum@nari.ee.ethz.ch<br />

HUT Hassan El-Sallabi Ph<strong>on</strong>e: +358 9 451 5960<br />

Fax: +358 9 451 2152<br />

E-mail: hsallabi@cc.hut.fi<br />

EBIT Tommi Jämsä Ph<strong>on</strong>e: +358 40 344 2000<br />

Fax: +358 8 551 4344<br />

E-mail: tommi.jamsa@elektrobit.com<br />

EBIT Juha Meinilä Ph<strong>on</strong>e: +358 40 344 2000<br />

Fax: +358 8 551 4344<br />

E-mail: juha.meinila@elektrobit.com<br />

EBIT Pekka Kyösti Ph<strong>on</strong>e: +358 40 344 2000<br />

Fax: +358 8 551 4344<br />

E-mail: pekka.kyosti@elektrobit.com<br />

EBIT Xi<strong>on</strong>gwen Zhao Ph<strong>on</strong>e: +358 40 344 2000<br />

Fax: +358 9 5121233<br />

E-mail: xi<strong>on</strong>gwen.zhao@elektrobit.com<br />

EBIT Daniela Laselva Ph<strong>on</strong>e: +358 40 344 2000<br />

Fax: +358 8 551 4344<br />

E-mail: daniela.laselva@elektrobit.com<br />

EBIT Jukka-Pekka Nuutinen Ph<strong>on</strong>e: +358 40 344 2000<br />

Fax: +358 8 551 4344<br />

E-mail: jukka-pekka.nuutinen@elektrobit.com<br />

Page 3 (167)


WINNER D5.4 v. 1.4<br />

Partner Name Ph<strong>on</strong>e / Fax / e-mail<br />

EBIT Lassi Hentilä Ph<strong>on</strong>e: +358 40 344 2000<br />

Fax: +358 8 551 4344<br />

E-mail: lassi.hentila@elektrobit.com<br />

HUT Pertti Vainikainen Ph<strong>on</strong>e: +358 9 451 2251<br />

Fax: +358 9 451 2152<br />

E-mail: pertti.vainikainen@tkk.fi<br />

HUT Jarmo Kivinen Ph<strong>on</strong>e: +358 9 451 2242<br />

Fax: +358 9 451 2152<br />

E-mail: jarmo.kivinen@tkk.fi<br />

HUT Lasse Vuokko Ph<strong>on</strong>e: +358 9 451 6064<br />

Fax: +358 9 451 2152<br />

E-mail: lasse.vuokko@tkk.fi<br />

KTH Per Zetterberg Ph<strong>on</strong>e: +46 8 7907785<br />

Fax:<br />

E-mail: per.zetterberg@s3.kth.se<br />

KTH Mats Bengtss<strong>on</strong> Ph<strong>on</strong>e: +46 8 7908463<br />

Fax:<br />

E-mail: mats.bengtss<strong>on</strong>@s3.kth.se<br />

KTH Niklas Jaldén Ph<strong>on</strong>e: +46 8 7908415<br />

Fax:<br />

E-mail: niklasj@s3.kth.se<br />

NOK Terhi Rautiainen Ph<strong>on</strong>e: +358 50 4837218<br />

Fax: +358 7180 36857<br />

E-mail: terhi.rautiainen@nokia.com<br />

Page 4 (167)


WINNER D5.4 v. 1.4<br />

Partner Name Ph<strong>on</strong>e / Fax / e-mail<br />

NOK Kimmo Kalliola Ph<strong>on</strong>e: +358 50 4837226<br />

Fax: +358 7180 36857<br />

E-mail: kimmo.kalliola@nokia.com<br />

TUI Marko Milojevic Ph<strong>on</strong>e: +49 3677 69 2615<br />

Fax: +49 3677 69 1195<br />

E-mail: marko.milojevic@tu-ilmenau.de<br />

TUI Christian Schneider Ph<strong>on</strong>e: +49 3677 69 1157<br />

Fax: +49 3677 69 1113<br />

E-mail: christian.schneider@tu-ilmenau.de<br />

ETHZ Jan Hansen Ph<strong>on</strong>e : +41 44 632 0290<br />

Fax: +41 44 632 1209<br />

E-mail: hansen@nari.ee.ethz.ch<br />

Page 5 (167)


WINNER D5.4 v. 1.4<br />

Table of C<strong>on</strong>tents<br />

Part I................................................................................................................. 11<br />

1. Introducti<strong>on</strong> ............................................................................................... 12<br />

2. WINNER Scenarios.................................................................................... 14<br />

2.1 Scenario definiti<strong>on</strong>s.............................................................................................................. 14<br />

2.1.1 Scenario A1: Indoor small office .................................................................................. 14<br />

2.1.2 Scenario B1: Urban micro-cell ..................................................................................... 15<br />

2.1.3 Scenario B3: Indoor hotspot ......................................................................................... 15<br />

2.1.4 Scenario B5: Stati<strong>on</strong>ary feeder ..................................................................................... 15<br />

2.1.5 Scenario C1: Suburban macro-cell................................................................................ 16<br />

2.1.6 Scenario C2: Urban macro-cell..................................................................................... 16<br />

2.1.7 Scenario D1: Rural macro-cell...................................................................................... 17<br />

3. WINNER Channel Models ......................................................................... 17<br />

3.1 Generic model...................................................................................................................... 17<br />

3.1.1 Large-scale parameters................................................................................................. 17<br />

3.1.2 Average power of ZDSC c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> their delays .................................................... 22<br />

3.1.3 Directi<strong>on</strong>al distributi<strong>on</strong>s of ZDSCs............................................................................... 23<br />

3.1.4 Antenna gain................................................................................................................ 25<br />

3.1.5 Path-loss <strong>models</strong>.......................................................................................................... 26<br />

3.1.6 Probability of line of sight............................................................................................ 26<br />

3.1.7 Generati<strong>on</strong> of <strong>channel</strong> coefficients................................................................................ 27<br />

3.2 Reduced variability “clustered delay line” model .................................................................. 29<br />

3.2.1 Scenario A1................................................................................................................. 30<br />

3.2.2 Scenario B1 ................................................................................................................. 31<br />

3.2.3 Scenario B3 ................................................................................................................. 32<br />

3.2.4 Scenario B5 ................................................................................................................. 33<br />

3.2.5 Scenario C1 ................................................................................................................. 37<br />

3.2.6 Scenario C2 ................................................................................................................. 38<br />

3.2.7 Scenario D1................................................................................................................. 39<br />

Part II................................................................................................................ 41<br />

4. Modelling Approaches.............................................................................. 42<br />

4.1 Generic <strong>channel</strong> modelling approach .................................................................................... 42<br />

4.1.1 Distincti<strong>on</strong> between <strong>channel</strong> <strong>models</strong> for <strong>link</strong>-<strong>level</strong> <strong>and</strong> <strong>system</strong>-<strong>level</strong> simulati<strong>on</strong>............. 42<br />

4.1.2 Comparis<strong>on</strong> between deterministic <strong>and</strong> stochastic <strong>channel</strong> modeling............................. 42<br />

4.1.3 Interference modeling .................................................................................................. 43<br />

4.1.4 Framework .................................................................................................................. 43<br />

4.2 Stati<strong>on</strong>ary-feeder scenarios B5 ............................................................................................. 51<br />

4.2.1 B5a LOS stati<strong>on</strong>ary feeder: rooftop-to-rooftop.............................................................. 51<br />

4.2.2 B5b LOS stati<strong>on</strong>ary feeder: street-<strong>level</strong> to street-<strong>level</strong>................................................... 52<br />

4.2.3 B5c hotspot LOS stati<strong>on</strong>ary-feeder: below rooftop to street-<strong>level</strong>.................................. 52<br />

4.2.4 B5d hotspot NLOS stati<strong>on</strong>ary feeder: rooftop to street-<strong>level</strong>.......................................... 52<br />

4.3 Coefficient generati<strong>on</strong> approaches........................................................................................ 53<br />

4.3.1 Stati<strong>on</strong>ary stochastic .................................................................................................... 53<br />

Page 6 (167)


WINNER D5.4 v. 1.4<br />

4.3.2 Sum-of-Sinusoids......................................................................................................... 54<br />

4.3.3 Problem details ............................................................................................................ 54<br />

4.3.4 Comparis<strong>on</strong> ................................................................................................................. 55<br />

4.3.5 Kr<strong>on</strong>ecker correlati<strong>on</strong>................................................................................................... 55<br />

5. Measurements <strong>and</strong> Literature Review ..................................................... 56<br />

5.1 Measurement <strong>system</strong>s .......................................................................................................... 56<br />

5.1.1 Principle of <strong>channel</strong> sounding....................................................................................... 56<br />

5.1.2 Channel sounders employed ......................................................................................... 56<br />

5.2 Measurement campaigns ...................................................................................................... 61<br />

5.2.1 Scenario A1................................................................................................................. 61<br />

5.2.2 Scenario B1 ................................................................................................................. 62<br />

5.2.3 Scenario B3 ................................................................................................................. 62<br />

5.2.4 Scenario C1 ................................................................................................................. 63<br />

5.2.5 Scenario C2 ................................................................................................................. 63<br />

5.2.6 Scenario D1................................................................................................................. 64<br />

5.2.7 Measurement summary ................................................................................................ 65<br />

5.3 Descripti<strong>on</strong> of key references ............................................................................................... 67<br />

5.4 Results of analysis items ...................................................................................................... 67<br />

5.4.1 Path-loss <strong>and</strong> shadow fading ........................................................................................ 67<br />

5.4.2 LOS probability ........................................................................................................... 73<br />

5.4.3 DS <strong>and</strong> maximum excess-delay distributi<strong>on</strong>.................................................................. 74<br />

5.4.4 Azimuth AS at BS <strong>and</strong> MS........................................................................................... 79<br />

5.4.5 Distributi<strong>on</strong> of the azimuth angles of the multipath comp<strong>on</strong>ents.................................... 83<br />

5.4.6 Angle proporti<strong>on</strong>ality factor ......................................................................................... 85<br />

5.4.7 Modelling of PDP ........................................................................................................ 87<br />

5.4.8 Number of ZDSC......................................................................................................... 91<br />

5.4.9 Distributi<strong>on</strong> of ZDSC delays ........................................................................................ 93<br />

5.4.10 Delay proporti<strong>on</strong>ality factor ......................................................................................... 96<br />

5.4.11 Ricean K-factor............................................................................................................ 98<br />

5.4.12 Cross-polarizati<strong>on</strong> ratio (XPR) ................................................................................... 101<br />

5.4.13 Large-scale parameter analysis item ........................................................................... 105<br />

5.5 Literature review................................................................................................................ 111<br />

5.5.1 Scenario A1............................................................................................................... 111<br />

5.5.2 Scenario B3 ............................................................................................................... 115<br />

5.5.3 Scenario B5 ............................................................................................................... 116<br />

5.5.4 Scenario C1 ............................................................................................................... 120<br />

5.5.5 Scenario C2 ............................................................................................................... 122<br />

5.5.6 Scenario D1............................................................................................................... 125<br />

5.6 Interpretati<strong>on</strong> of results ...................................................................................................... 127<br />

5.6.1 Path-loss.................................................................................................................... 127<br />

5.6.2 Power-delay profile.................................................................................................... 131<br />

5.6.3 Delay spread.............................................................................................................. 131<br />

5.6.4 K-factor ..................................................................................................................... 132<br />

5.6.5 Cross-polarizati<strong>on</strong> discriminati<strong>on</strong> (XPR) .................................................................... 132<br />

5.6.6 Doppler ..................................................................................................................... 132<br />

5.6.7 Angle-spread.............................................................................................................. 132<br />

Page 7 (167)


WINNER D5.4 v. 1.4<br />

5.6.8 Antenna gain.............................................................................................................. 132<br />

5.6.9 Frequency dependence of the propagati<strong>on</strong> parameters................................................. 133<br />

6. Channel Model Implementati<strong>on</strong> ............................................................. 135<br />

6.1 Overview for implementing the model................................................................................ 135<br />

6.1.1 Time sampling <strong>and</strong> interpolati<strong>on</strong> ................................................................................ 135<br />

6.1.2 Coordinate <strong>system</strong>...................................................................................................... 135<br />

6.1.3 Generati<strong>on</strong> of correlated large-scale parameters.......................................................... 137<br />

6.2 Interfaces ........................................................................................................................... 138<br />

6.2.1 Example input parameters.......................................................................................... 138<br />

6.2.2 Example output parameters ........................................................................................ 141<br />

6.3 Guidelines <strong>and</strong> examples <strong>on</strong> performing <strong>system</strong>-<strong>level</strong> simulati<strong>on</strong>s....................................... 142<br />

6.3.1 H<strong>and</strong>over ................................................................................................................... 142<br />

6.3.2 Interference................................................................................................................ 143<br />

6.3.3 Multi-cell <strong>and</strong> multi-user............................................................................................ 143<br />

6.3.4 Multihop <strong>and</strong> relaying................................................................................................ 144<br />

7. Test <strong>and</strong> Verificati<strong>on</strong> of the Channel Model <strong>and</strong> Its Implementati<strong>on</strong> .. 145<br />

7.1 Test cases........................................................................................................................... 145<br />

7.1.1 General test cases....................................................................................................... 145<br />

7.1.2 Input/output parameters.............................................................................................. 145<br />

7.1.3 Validati<strong>on</strong> of computati<strong>on</strong>.......................................................................................... 146<br />

8. References............................................................................................... 148<br />

9. Appendix.................................................................................................. 153<br />

9.1 Other scenarios .................................................................................................................. 153<br />

9.1.1 Scenario definiti<strong>on</strong>s.................................................................................................... 153<br />

9.2 Measurement campaigns for other scenarios ....................................................................... 153<br />

9.2.1 Scenario “high mobility short range hot spot”............................................................. 153<br />

9.2.2 Urban ad-hoc peer-to-peer.......................................................................................... 154<br />

9.3 Measurement results for other scenarios ............................................................................. 154<br />

9.3.1 Scenario C2: typical urban macro-cell - KTH campaign.............................................. 154<br />

9.3.2 Scenario “high mobility short range hot spot”............................................................. 156<br />

9.4 Literature review for other scenarios................................................................................... 167<br />

9.4.1 Scenario “high mobility short range hot spot”............................................................. 167<br />

Page 8 (167)


WINNER D5.4 v. 1.4<br />

List of Acr<strong>on</strong>yms <strong>and</strong> Abbreviati<strong>on</strong>s<br />

3GPP<br />

3 rd Generati<strong>on</strong> Partnership Project<br />

3GPP2 3 rd Generati<strong>on</strong> Partnership Project 2<br />

ACF<br />

ADC<br />

AoA<br />

AoD<br />

APP<br />

APS<br />

AS<br />

AWGN<br />

B3G<br />

BER<br />

BRAN<br />

BS<br />

BW<br />

C/I<br />

CDL<br />

CW<br />

D 3 SF<br />

DoA<br />

DoD<br />

DS<br />

EBIT<br />

EBITT<br />

ESPRIT<br />

ETHZ<br />

ETSI<br />

FDD<br />

FIR<br />

FS<br />

GPS<br />

HIPERLAN<br />

HUT<br />

IR<br />

ISIS<br />

KTH<br />

LNS<br />

LOS<br />

MCSSS<br />

METRA<br />

MIMO<br />

MPC<br />

MS<br />

MUSIC<br />

Auto-Correlati<strong>on</strong> Functi<strong>on</strong><br />

Analog-to-Digital C<strong>on</strong>verter<br />

Angle of Arrival<br />

Angle of Departure<br />

A Posteriori Probability<br />

Angle Power Spectrum<br />

Azimuth Spread<br />

Additive White Gaussian Noise<br />

Bey<strong>on</strong>d 3G<br />

Bit Error Rate<br />

Broadb<strong>and</strong> Radio Access Networks<br />

Base Stati<strong>on</strong><br />

B<strong>and</strong>width<br />

Carrier to Interference ratio<br />

Clustered Delay Line<br />

C<strong>on</strong>tinuous Wave<br />

Double-Directi<strong>on</strong>al Delay-Spread Functi<strong>on</strong><br />

Directi<strong>on</strong> of Arrival<br />

Directi<strong>on</strong> of Departure<br />

Delay Spread<br />

Elektrobit Ltd<br />

Elektrobit Testing Ltd<br />

Estimati<strong>on</strong> of Signal Parameters via Rotati<strong>on</strong>al Invariance Techniques<br />

Eidgenössische Technische Hochschule Zürich (Swiss Federal Institute of Technology<br />

Zurich)<br />

European Telecommunicati<strong>on</strong>s St<strong>and</strong>ards Institute<br />

Frequency Divisi<strong>on</strong> Duplex<br />

Finite Impulse Resp<strong>on</strong>se<br />

Fixed Stati<strong>on</strong><br />

Global Positi<strong>on</strong>ing System<br />

High Performance Local Area Network<br />

Helsinki University of Technology (TKK)<br />

Impulse Resp<strong>on</strong>se<br />

Initializati<strong>on</strong> <strong>and</strong> Search Improved SAGE<br />

Kungliga Tekniska Högskolan (Royal Institute of Technology in Stockholm)<br />

Log-Normal Shadowing<br />

Line-of-Sight<br />

Multi-Carrier Spread Spectrum Signal<br />

Multi-Element Transmit <strong>and</strong> Receive Antennas (European IST project)<br />

Multiple-Input Multiple-Output<br />

Multi-Path Comp<strong>on</strong>ent<br />

Mobile Stati<strong>on</strong><br />

Multiple Signal Classificati<strong>on</strong><br />

Page 9 (167)


WINNER D5.4 v. 1.4<br />

NACM<br />

NLOS<br />

NOK<br />

OFDM<br />

OLOS<br />

PAS<br />

PD 3 S<br />

PDP<br />

RMS<br />

PN<br />

RIMAX<br />

RF<br />

RX<br />

SAGE<br />

SCM<br />

SCME<br />

SF<br />

SIMO<br />

SoS<br />

SW<br />

TDL<br />

TUI<br />

TX<br />

WINNER<br />

WPx<br />

XPR<br />

XPR H<br />

XPR V<br />

ZDSC<br />

ZDSC_A<br />

ZDSC_D<br />

No Auto-Correlati<strong>on</strong> Mode<br />

N<strong>on</strong> Line-of-Sight<br />

Nokia<br />

Orthog<strong>on</strong>al Frequency-Divisi<strong>on</strong> Multiplexing<br />

Obstructed Line-of-Sight<br />

Power Azimuth Spectrum<br />

Power Double-Directi<strong>on</strong>al Delay-Spectrum<br />

Power-Delay Profile<br />

Root Mean Square<br />

Pseudo Noise<br />

maximum likelihood parameter estimati<strong>on</strong> framework for joint superresoluti<strong>on</strong> estimati<strong>on</strong><br />

of both specular <strong>and</strong> dense multipath comp<strong>on</strong>ents<br />

Radio Frequency<br />

Receiver<br />

Space-Alternating Generalized Expectati<strong>on</strong>-maximizati<strong>on</strong><br />

Spatial Channel Model<br />

Spatial Channel Model Extended<br />

Shadow Fading<br />

Single-Input Multiple-Output<br />

Sum of Sinusoids<br />

Software<br />

Tapped Delay-Line<br />

Technische Universität Ilmenau<br />

Transmitter<br />

Wireless World Initiative New Radio<br />

Work Package x of WINNER project<br />

Cross-Polarisati<strong>on</strong> Ratio<br />

Horiz<strong>on</strong>tal Polarisati<strong>on</strong> XPR<br />

Vertical Polarisati<strong>on</strong> XPR<br />

Zero Delay-Spread Cluster<br />

Zero Delay-Spread Cluster of Arrival<br />

Zero Delay-Spread Cluster of Departure<br />

Page 10 (167)


WINNER D5.4 v. 1.4<br />

PART I<br />

The deliverable D5.4 is divided into two major parts. This first part is the<br />

relatively short main part <strong>and</strong> c<strong>on</strong>tains the essence of the deliverable,<br />

specifically the <strong>channel</strong> model definiti<strong>on</strong>.<br />

Page 11 (167)


WINNER D5.4 v. 1.4<br />

1. Introducti<strong>on</strong><br />

WINNER project is aiming at a Bey<strong>on</strong>d-3G (B3G) radio <strong>system</strong> using a frequency b<strong>and</strong>width of 100<br />

MHz for <strong>on</strong>e radio c<strong>on</strong>necti<strong>on</strong> <strong>and</strong> a radio frequency lying most probably somewhere between 2 <strong>and</strong> 6<br />

GHz in spectrum. The research c<strong>on</strong>cerning the suitability of certain communicati<strong>on</strong> parameters, like<br />

modulati<strong>on</strong>, coding, symbol rate, MIMO antenna utilisati<strong>on</strong> etc., is performed through extensive<br />

simulati<strong>on</strong>s. The simulati<strong>on</strong> results depend str<strong>on</strong>gly <strong>on</strong> the radio <strong>channel</strong>. Hence, the radio <strong>channel</strong> is a<br />

crucial part of the simulati<strong>on</strong>. On <strong>on</strong>e h<strong>and</strong>, it is very important to use a very accurate <strong>and</strong> realistic<br />

<strong>channel</strong> model in the simulati<strong>on</strong> to enable reliable simulati<strong>on</strong> results. On the other h<strong>and</strong>, the complexity<br />

of the simulati<strong>on</strong> should be kept low. Therefore, the research challenge is to create a <strong>channel</strong> model which<br />

is realistic enough <strong>and</strong> simple.<br />

WINNER Work Package 5 (WP5) is focused <strong>on</strong> multi-dimensi<strong>on</strong>al radio <strong>channel</strong> modelling. Totally six<br />

partners are involved in WP5, namely Elektrobit (EBIT, in year 2004, <strong>and</strong> Elektrobit Testing EBITT in<br />

year 2005), Helsinki University of Technology (HUT), Nokia (NOK), Royal Institute of Technology in<br />

Stockholm (KTH), Swiss Federal Institute of Technology Zurich (ETHZ), <strong>and</strong> Technical University of<br />

Ilmenau (TUI). Up to now, the situati<strong>on</strong> is such that there are no widely accepted <strong>channel</strong> <strong>models</strong><br />

available which are suitable for WINNER <strong>system</strong> parameters. Therefore, WINNER WP5 has to create<br />

new <strong>channel</strong> <strong>models</strong> needed in the project. For the initial purposes, WP5 selected <strong>and</strong> recommended two<br />

existing <strong>channel</strong> <strong>models</strong>, which are called initial <strong>channel</strong> <strong>models</strong> [D5.1]. The <strong>models</strong> are 3GPP/3GPP2<br />

Spatial Channel Model (SCM) [3GPP SCM] for outdoor simulati<strong>on</strong>s <strong>and</strong> IEEE 802.11n MIMO model<br />

[802.11n] for indoor simulati<strong>on</strong>s. Because the SCM model was not suitable for WINNER simulati<strong>on</strong>s as<br />

such, WP5 performed some modificati<strong>on</strong>s <strong>and</strong> implemented the extended SCM model (SCME) [SCME].<br />

However, in spite of these modificati<strong>on</strong>s, the initial <strong>channel</strong> <strong>models</strong> were not good enough for the<br />

advanced simulati<strong>on</strong>s. C<strong>on</strong>sequently new WINNER <strong>models</strong> are needed.<br />

The WINNER <strong>channel</strong> <strong>models</strong> were implemented in two steps. In the first step, <strong>channel</strong> <strong>models</strong> for the<br />

most urgently needed propagati<strong>on</strong> scenarios with a limited number of parameters were created.<br />

Propagati<strong>on</strong> scenario means here the propagati<strong>on</strong> envir<strong>on</strong>ment <strong>and</strong> certain propagati<strong>on</strong> related parameters<br />

specified to meaningful values. The main difference between different propagati<strong>on</strong> scenarios exists due to<br />

the diverse envir<strong>on</strong>ments. Channel model parameters were defined for five propagati<strong>on</strong> scenarios<br />

(prioritised scenarios) according to [D7.2], namely indoor small office (A1), urban micro-cell (B1),<br />

stati<strong>on</strong>ary feeder (B5), urban macro-cell (C2), <strong>and</strong> rural macro-cell (D1). These <strong>models</strong> are described in<br />

the deliverable D5.3 [D5.3]. In the sec<strong>on</strong>d step the <strong>channel</strong> <strong>models</strong> were upgraded so that more<br />

parameters are included in the <strong>models</strong>. Two more scenarios – indoor (B3) <strong>and</strong> suburban (C1) – are also<br />

included based <strong>on</strong> the feedback from other work packages. The <strong>channel</strong> <strong>models</strong> created in the first step,<br />

<strong>and</strong> updated in the sec<strong>on</strong>d step, are described in this deliverable, D5.4.<br />

In this deliverable, we describe a generic <strong>channel</strong> model framework that is subsequently used as a basis<br />

for the <strong>channel</strong> <strong>models</strong> of all scenarios, except B5. Furthermore, we present clustered delay line (CDL)<br />

<strong>models</strong> for calibrati<strong>on</strong> <strong>and</strong> comparis<strong>on</strong> simulati<strong>on</strong>s. The generic modelling approach allows the creati<strong>on</strong><br />

of virtually unlimited double directi<strong>on</strong>al radio <strong>channel</strong> realizati<strong>on</strong>s. The generic <strong>channel</strong> model is a raybased<br />

multi-<strong>link</strong> model that is antenna independent, scalable <strong>and</strong> capable of modelling <strong>channel</strong>s for<br />

MIMO c<strong>on</strong>necti<strong>on</strong>s. The <strong>models</strong> are based <strong>on</strong> the existing literature <strong>and</strong> the parameters extracted from<br />

eleven measurement campaigns performed by the WP5. The selecti<strong>on</strong> of the model parameters is based<br />

both <strong>on</strong> the measurements <strong>and</strong> informati<strong>on</strong> found in the literature. The measurements were performed by<br />

five partners, namely EBIT/EBITT, HUT, KTH, NOK, <strong>and</strong> TUI. Different <strong>channel</strong> sounders, most of<br />

them capable of measurements at 2 <strong>and</strong> 5 GHz frequency ranges <strong>and</strong> 100 MHz b<strong>and</strong>width, were used.<br />

Measurement results were analyzed using beam-forming <strong>and</strong> super-resoluti<strong>on</strong> methods. The analyzed<br />

items, e.g. path loss, shadow fading characteristics, power delay profiles, delay spreads, angle-spreads,<br />

<strong>and</strong> cross-polarisati<strong>on</strong> ratio (XPR), were analyzed for the scenarios of interest.<br />

In WP5 <strong>on</strong>e activity has been the implementati<strong>on</strong> of the 3GPP/3GPP2 SCM <strong>channel</strong> model. The model<br />

was implemented in software by the WP5. Later, its extensi<strong>on</strong> to 5 GHz frequency range <strong>and</strong> 100 MHz<br />

b<strong>and</strong>width [SCME] was implemented. The extensi<strong>on</strong> work has been published in [BGS+05].<br />

We have compiled a set of requirements from various documents, specifically the WP2 Channel Model<br />

Requirements, the WP5 Deliverable D5.2 [D5.2], WP7 deliverable D7.2 [D7.2], <strong>and</strong> <str<strong>on</strong>g>report</str<strong>on</strong>g>ed<br />

shortcomings of the <strong>channel</strong> <strong>models</strong> selected for initial usage [D5.1]. The main requirements are proper<br />

characterisati<strong>on</strong> of spatial properties for MIMO support, large set of possible <strong>channel</strong>s as well as some<br />

limited r<strong>and</strong>omness <strong>channel</strong>s, c<strong>on</strong>sistency in time, frequency <strong>and</strong> space, e.g. inherent <strong>link</strong> between angle<br />

spectrum <strong>and</strong> Doppler spectrum, time-variability of bulk parameters, <strong>and</strong> extended polarisati<strong>on</strong> support.<br />

The document is organized in a way to provide best readability. Its overall c<strong>on</strong>tent is divided into 2 major<br />

parts. The first part is relatively short <strong>and</strong> c<strong>on</strong>tains the core informati<strong>on</strong> provided in this deliverable. Part I<br />

begins with an introducti<strong>on</strong>, background informati<strong>on</strong> c<strong>on</strong>cerning our approach, <strong>and</strong> the requirements <strong>on</strong><br />

Page 12 (167)


WINNER D5.4 v. 1.4<br />

the <strong>models</strong> defined within the WINNER project. It is followed by related scenario definiti<strong>on</strong>s. The major<br />

<strong>and</strong> last chapter of part I c<strong>on</strong>tains the brief but comprehensive definiti<strong>on</strong> of the <strong>channel</strong> <strong>models</strong>. Part II<br />

provides more elaborate background informati<strong>on</strong> <strong>on</strong> model development. It c<strong>on</strong>tains detailed discussi<strong>on</strong><br />

<strong>on</strong> our modelling approach, the underlying data of our <strong>models</strong> (measurements <strong>and</strong> literature review), <strong>and</strong><br />

the interpretati<strong>on</strong> thereof. Two more chapters are dedicated to the <strong>channel</strong> model implementati<strong>on</strong>, <strong>and</strong> the<br />

test <strong>and</strong> verificati<strong>on</strong> of the model. The document ends with references <strong>and</strong> further, so far unused results of<br />

this project.<br />

Page 13 (167)


WINNER D5.4 v. 1.4<br />

2. WINNER Scenarios<br />

These are the propagati<strong>on</strong> scenarios defined in WINNER. Scenarios marked in bold are prioritized<br />

scenarios that were modelled <strong>and</strong> implemented as the WINNER <strong>channel</strong> <strong>models</strong>.<br />

Table 2.1: Propagati<strong>on</strong> scenarios defined in WINNER.<br />

Scenario Definiti<strong>on</strong><br />

LOS/N<br />

LOS<br />

Mob. AP ht UE ht Distance<br />

range<br />

Note<br />

A1<br />

In building<br />

Indoor small<br />

office / residential<br />

LOS/<br />

NLOS<br />

0–5<br />

km/h<br />

2 m 1 m 3 - 100 m Deterministic room layout<br />

A2<br />

In building<br />

Indoor to outdoor NLOS 0–5<br />

km/h<br />

AP inside <strong>and</strong> coverage<br />

outside the building.<br />

B1<br />

Hotspot<br />

Typical urban<br />

micro-cell<br />

LOS/<br />

NLOS<br />

0–70<br />

km/h<br />

Below RT,<br />

e.g. 10 m<br />

1.5 m 20 - 400 m<br />

B2<br />

Hotspot<br />

B3<br />

Hotspot<br />

B4<br />

Hotspot<br />

Bad urban NLOS 0–70<br />

km/h<br />

Indoor LOS 0–5<br />

km/h<br />

Outdoor to indoor NLOS 0–5<br />

km/h<br />

Airport-type. Coverage in<br />

shopping hall with BTS<br />

outside.<br />

B5a<br />

Hotspot<br />

LOS stat. feeder,<br />

rooftop to rooftop<br />

LOS 0 km/h Above RT. Above<br />

RT.<br />

30m - 8 km<br />

B5b<br />

Hotspot<br />

LOS stat. feeder,<br />

street-<strong>level</strong> to<br />

street-<strong>level</strong><br />

LOS 0 km/h 2-5 m 2-5 m<br />

B5c<br />

B5d<br />

LOS stat. feeder,<br />

below-rooftop to<br />

street-<strong>level</strong><br />

NLOS stat. feeder,<br />

rooftop to street<strong>level</strong><br />

LOS 0 km/h As B1. As B1. As B1. As B1.<br />

NLOS 0km/h As C2. 1.5-10 m. As C2. As C2.<br />

C1<br />

Metropol<br />

Suburban<br />

LOS/<br />

NLOS<br />

0–70<br />

km/h<br />

35 - 3000 m<br />

C2<br />

Metropol<br />

Typical urban<br />

macro-cell<br />

LOS/<br />

NLOS<br />

0–70<br />

km/h<br />

Above RT,<br />

e.g. 32 m<br />

1.5 m 35 - 3000 m<br />

C3<br />

Metropol<br />

C4<br />

Metropol<br />

C5<br />

Metropol<br />

Bad urban NLOS 0–70<br />

km/h<br />

Outdoor to indoor NLOS 0–70<br />

km/h<br />

LOS feeder LOS 0 km/h<br />

D1<br />

Rural<br />

Rural macro-cell<br />

LOS/<br />

NLOS<br />

0–200<br />

km/h<br />

Above RT,<br />

e.g. 45 m<br />

1.5 m 35m - 10 km<br />

D2<br />

Rural<br />

LOS moving<br />

networks (feeder)<br />

LOS 0–300<br />

km/h<br />

2.1 Scenario definiti<strong>on</strong>s<br />

In the following subsecti<strong>on</strong>s, we present WP5 view to the envir<strong>on</strong>ments of the five prioritized scenarios.<br />

2.1.1 Scenario A1: Indoor small office<br />

Scenario A1 envir<strong>on</strong>ment is described in [D7.2]. This represents typical office envir<strong>on</strong>ment, where the<br />

area per floor is 5000 m 2 , number of floors is 3 <strong>and</strong> room dimensi<strong>on</strong>s are 10 m x 10 m x 3 m <strong>and</strong> the<br />

corridors have the dimensi<strong>on</strong>s 100 m x 5 m x 3 m. The A1 indoor office model is illustrated in Figure 2.1.<br />

Page 14 (167)


WINNER D5.4 v. 1.4<br />

Figure 2.1: Layout of the A1 indoor scenario.<br />

The measured envir<strong>on</strong>ment resembles this definiti<strong>on</strong>, but is not identical [WP5AR]. It is assumed that<br />

propagati<strong>on</strong> parameters can be deduced from these measurements.<br />

2.1.2 Scenario B1: Urban micro-cell<br />

This scenario is defined for envir<strong>on</strong>ment where both fixed stati<strong>on</strong> <strong>and</strong> mobile stati<strong>on</strong> antenna heights are<br />

below surrounding buildings <strong>and</strong> both are outdoors. This scenario covers both LOS <strong>and</strong> NLOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. The envir<strong>on</strong>ment is defined for Manhattan like grid. The envir<strong>on</strong>ment streets can<br />

be classified as a main street, where the fixed stati<strong>on</strong> is located, perpendicular streets <strong>and</strong> parallel streets.<br />

The scenario is defined for street distance from 20 m to 400 m. In this envir<strong>on</strong>ment, the radio propagati<strong>on</strong><br />

<strong>and</strong> cell shape are c<strong>on</strong>fined within the area defined by the surrounding buildings.<br />

2.1.3 Scenario B3: Indoor hotspot<br />

The scenario B3 is described in [D7.2] <strong>and</strong> represents a typical indoor hot spot applicati<strong>on</strong> with a wide<br />

coverage area but n<strong>on</strong>-ubiquitous <strong>and</strong> low mobility (0-5 km/h). In this scenario traffic of high density can<br />

be expected. Typically applicati<strong>on</strong> scenarios can be found in c<strong>on</strong>ference halls, factory halls, entrance halls<br />

of train stati<strong>on</strong>s <strong>and</strong> airports, where the indoor envir<strong>on</strong>ment is characterised by large distances. The<br />

dimensi<strong>on</strong>s of such large halls can range from 20 m x 20 m x 5 m up to more then 100 m in width <strong>and</strong><br />

length as well as 20 m in height. Both LOS <strong>and</strong> NLOS propagati<strong>on</strong> situati<strong>on</strong>s can be found in this<br />

scenario.<br />

2.1.4 Scenario B5: Stati<strong>on</strong>ary feeder<br />

The definiti<strong>on</strong> of this scenario is less well understood by WP5 than are the others. WP5 found that that<br />

NLOS cases are also of interest for the feeder applicati<strong>on</strong>s. We therefore discuss <strong>models</strong> for NLOS cases<br />

as well. The following different feeder scenarios have been studied:<br />

• B5a Hotspot LOS stati<strong>on</strong>ary feeder: rooftop-to-rooftop<br />

• B5b Hotspot LOS stati<strong>on</strong>ary feeder: street-<strong>level</strong>-to-street-<strong>level</strong>.<br />

• B5c Hotspot LOS stati<strong>on</strong>ary feeder: blow-rooftop-to-street-<strong>level</strong><br />

• B5d Hotspot NLOS stati<strong>on</strong>ary feeder: above-rooftop-to-street-<strong>level</strong>.<br />

The scenarios of B5a <strong>and</strong> B5b are discussed below:<br />

2.1.4.1 Scenario B5a: LOS stati<strong>on</strong>ary feeder: rooftop-to-rooftop<br />

Our underst<strong>and</strong>ing of this case is illustrated in Figure 2.2. Wireless feeder master-stati<strong>on</strong>, probably <strong>on</strong> an<br />

elevated building, is c<strong>on</strong>nected to <strong>on</strong>e or several wireless feeder peripheral stati<strong>on</strong>s. A hot-spot wireless<br />

access point is then c<strong>on</strong>nected to the peripheral. As indicated in the picture, a cable is needed to c<strong>on</strong>nect<br />

the roof-top wireless feeder peripheral antenna. Alternatively a wireless soluti<strong>on</strong> may be possible also for<br />

these hops but then requiring additi<strong>on</strong>al antennas <strong>and</strong> transceivers.<br />

Page 15 (167)


WINNER D5.4 v. 1.4<br />

Feeder-<strong>link</strong><br />

Masterstati<strong>on</strong><br />

Peripheral<br />

Cable<br />

Hot-spot<br />

Figure 2.2: Illustrati<strong>on</strong> of LOS stati<strong>on</strong>ary feeder: rooftop-to-rooftop.<br />

2.1.4.2 Scenario B5b: LOS stati<strong>on</strong>ary feeder: street-<strong>level</strong> to street-<strong>level</strong><br />

Our underst<strong>and</strong>ing of this case in indicated in Figure 2.3. Both ends of the <strong>link</strong> are located a few meters<br />

above ground <strong>and</strong> the model is aimed for 2-5 meter antenna heights. In many cases it may be possible to<br />

place the antennas high enough such that the first Fresnel z<strong>on</strong>e is clear <strong>and</strong> therefore free-space<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s apply.<br />

Hotspot<br />

<strong>and</strong> feederperipheral.<br />

Feeder-<strong>link</strong><br />

Hotspot<br />

<strong>and</strong> feedermaster.<br />

Figure 2.3: Illustrati<strong>on</strong> of wireless LOS feeder-<strong>link</strong>: street-<strong>level</strong>.<br />

2.1.5 Scenario C1: Suburban macro-cell<br />

The scenario C1 is defined for a suburban outdoor envir<strong>on</strong>ment, where the coverage is ubiquitous. In<br />

suburban macrocells base stati<strong>on</strong>s are located well above the rooftops to allow wide area coverage.<br />

Buildings are typically low residential detached houses with <strong>on</strong>e or two floors, or blocks of flats with a<br />

few floors. Occasi<strong>on</strong>al open areas such as parks or playgrouds between the houses make the envir<strong>on</strong>ment<br />

rather open. Streets have r<strong>and</strong>om orientati<strong>on</strong>s, <strong>and</strong> no urban-like regular strict grid structure is observed.<br />

Vegetati<strong>on</strong> is modest.<br />

2.1.6 Scenario C2: Urban macro-cell<br />

In typical urban macrocell, mobile stati<strong>on</strong> is at street <strong>level</strong> <strong>and</strong> fixed base stati<strong>on</strong> clearly above<br />

surrounding building heights. As for propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s, n<strong>on</strong>- or obstructed line-of-sight is a comm<strong>on</strong><br />

case, since street <strong>level</strong> is often reached by a single diffracti<strong>on</strong> over the rooftop. The building blocks can<br />

form either a regular Manhattan type of grid, or have more irregular locati<strong>on</strong>s. Typical building heights in<br />

urban envir<strong>on</strong>ments are over four storeys. Outdoor-to-indoor modelling is not part of typical urban<br />

macrocell scenario, but is a different scenario (Table 2.1, C4).<br />

Page 16 (167)


WINNER D5.4 v. 1.4<br />

2.1.7 Scenario D1: Rural macro-cell<br />

Scenario D1 is defined <strong>on</strong>ly through its size (100 km 2 ) <strong>and</strong> hexag<strong>on</strong>al cell lay-out in [D7.2].<br />

The rural envir<strong>on</strong>ment we measured is flat, c<strong>on</strong>sisting of mainly sparsely located houses al<strong>on</strong>g roads that<br />

lead trough fields <strong>and</strong> some small forests <strong>and</strong> a small village. This should be c<strong>on</strong>sidered when interpreting<br />

results based <strong>on</strong> our model.<br />

3. WINNER Channel Models<br />

This chapter describes WINNER MIMO <strong>channel</strong> <strong>models</strong> of seven propagati<strong>on</strong> scenarios for <strong>link</strong> <strong>level</strong> <strong>and</strong><br />

<strong>system</strong> <strong>level</strong> simulati<strong>on</strong>s. Link <strong>level</strong> is defined for a single communicati<strong>on</strong> <strong>link</strong>. System <strong>level</strong> is defined<br />

for multi communicati<strong>on</strong> <strong>link</strong>s <strong>and</strong> base stati<strong>on</strong>s. Five of these scenarios are the prioritized propagati<strong>on</strong><br />

scenarios defined in the WINNER project in [D7.2] for short range <strong>and</strong> wide area wireless<br />

communicati<strong>on</strong>s. The prioritized scenarios are: Scenario A1 for indoor small office envir<strong>on</strong>ments,<br />

Scenario B1 for microcell urban envir<strong>on</strong>ment, Scenario B5 for hotspot LOS stati<strong>on</strong>ary wireless feeder,<br />

Scenario C2 for Metropolitan ubiquitous coverage in macrocell urban envir<strong>on</strong>ment, Scenario D1 for<br />

macrocell rural envir<strong>on</strong>ment. The two additi<strong>on</strong>al scenarios are part of the WINNER <strong>channel</strong> model:<br />

Scenario B3 for indoor propagati<strong>on</strong> <strong>and</strong> Scenario C1 for macrocell suburban envir<strong>on</strong>ment.<br />

In this chapter, we provide descripti<strong>on</strong> of the generic <strong>channel</strong> model, which is based <strong>on</strong> the principles of<br />

the SCM [3GPP SCM], for scenario A1, B1, B3, C1, C2, <strong>and</strong> D1. We also present clustered delay line<br />

(CDL) <strong>models</strong> for the menti<strong>on</strong>ed six scenarios of interest to generic model, <strong>and</strong> stati<strong>on</strong>ary feeder <strong>models</strong><br />

for scenario B5. The generic <strong>channel</strong> model is a geometric-based stochastic <strong>channel</strong> model. The following<br />

subsecti<strong>on</strong>s describe the WINNER phase-I MIMO <strong>channel</strong> <strong>models</strong> at 5 GHz.<br />

3.1 Generic model<br />

We apply the framework of the generic <strong>channel</strong> modelling approach presented in Chapter 4 to WINNER<br />

scenarios A1, B1, B3, C1, C2, <strong>and</strong> D1. Scenario B5 is not c<strong>on</strong>sidered in the generic <strong>channel</strong> model since<br />

it is a stati<strong>on</strong>ary wireless feeder scenario, where transmitter <strong>and</strong> receiver ends are fixed. Scenario B5 is<br />

modelled separately as clustered (tapped) delay line model (CDL) in Secti<strong>on</strong> 3.2.4.<br />

The generic <strong>channel</strong> model generates a number of ZDSCs. Their delays <strong>and</strong> directi<strong>on</strong>al properties are<br />

extracted from statistical distributi<strong>on</strong>s that corresp<strong>on</strong>d to a specific scenario, which are obtained from<br />

measurement results or from literature. The number of ZDSCs varies from <strong>on</strong>e scenario to another.<br />

Indeed, the number of ZDSCs itself is a r<strong>and</strong>om variable. However, in order to reduce the complexity for<br />

simulati<strong>on</strong> purpose, it has been kept as a fixed parameter. The median of the number ZDSCs is selected.<br />

We fix the number of rays within each ZDSC to 10 rays that have same delays <strong>and</strong> powers <strong>and</strong> may differ<br />

in angles, either departure or arrival. The directi<strong>on</strong>al properties of each ZDSC may vary from <strong>on</strong>e<br />

scenario to another <strong>and</strong> from departure side to arrival side. The WINNER generic <strong>channel</strong> model is<br />

antenna independent. Hence, different antenna c<strong>on</strong>figurati<strong>on</strong>s can be supported. In later terminology, the<br />

down<strong>link</strong> is c<strong>on</strong>sidered, where the transmitter is the fixed stati<strong>on</strong> (BS) <strong>and</strong> the receiver is the mobile<br />

stati<strong>on</strong> (MS). However, the same <strong>models</strong> can also be used for up<strong>link</strong> simulati<strong>on</strong>s due to the reciprocity of<br />

the radio <strong>channel</strong>.<br />

3.1.1 Large-scale parameters<br />

The radio <strong>channel</strong> is in general not stati<strong>on</strong>ary. Nevertheless, over short periods of time <strong>and</strong> space, <strong>channel</strong><br />

parameters experience small variati<strong>on</strong>s, <strong>and</strong> the assumpti<strong>on</strong> of short-term stati<strong>on</strong>arity is often a very good<br />

approximati<strong>on</strong>. The parameters characterizing our <strong>channel</strong> model are called bulk parameters. The time<br />

durati<strong>on</strong>s, over which these bulk parameters are c<strong>on</strong>stant, are termed <strong>channel</strong> segments a.k.a. drops in the<br />

nomenclature of the SCM. Over time <strong>and</strong> space, bulk parameters change <strong>and</strong> we characterize this<br />

variability statistically.<br />

There are a large number of bulk parameters. Bulk parameters include detailed or low-<strong>level</strong> bulk<br />

parameters such as number of paths, path powers, path angles at both <strong>link</strong> ends, path elevati<strong>on</strong>s at both<br />

<strong>link</strong> ends, <strong>and</strong> path delays. To characterize the <strong>channel</strong> with fewer parameters, higher <strong>level</strong>, e.g. sec<strong>on</strong>dorder,<br />

statistics are extracted <strong>on</strong> a per-segment basis, which we denote large-scale or dispersi<strong>on</strong> metric<br />

parameters. Large-scale parameters characterize the distributi<strong>on</strong>s of <strong>and</strong> between previously menti<strong>on</strong>ed<br />

low-<strong>level</strong> bulk parameters. Because realisati<strong>on</strong>s of large-scale parameters are drawn <strong>on</strong>ly <strong>on</strong>ce per<br />

<strong>channel</strong> segment, they are bulk parameters themselves. The following large-scale parameters are<br />

c<strong>on</strong>sidered:<br />

• Shadowing. The log-normal shadowing (LNS) value is the comm<strong>on</strong> shadowing across (i.e., for<br />

all) clusters. The variability across clusters around the LNS is given by an additive (in logdomain)<br />

Gaussian distributi<strong>on</strong> with a fixed st<strong>and</strong>ard deviati<strong>on</strong> of 3 dB.<br />

Page 17 (167)


WINNER D5.4 v. 1.4<br />

• Cross-polarizati<strong>on</strong> ratio (XPR). No distincti<strong>on</strong> is made between clusters <strong>and</strong> segments in the<br />

current model. Therefore, the resulting variability of XPR is equivalent if evaluated across<br />

clusters or across segments.<br />

• Total angle-spread <strong>and</strong> delay-spread. These parameters characterize the power dispersi<strong>on</strong> in<br />

angle <strong>and</strong> delay domain across clusters. Note that this is a high-<strong>level</strong> characterizati<strong>on</strong>. The more<br />

detailed properties of angle <strong>and</strong> delay dispersi<strong>on</strong> are each defined by a set of two variables. This<br />

is firstly, a mean angle <strong>and</strong> a delay offset for each single cluster, <strong>and</strong> sec<strong>on</strong>dly, an angle-spread<br />

<strong>and</strong> a delay-spread for each cluster. Here,<br />

• The angle-spread per cluster <strong>and</strong> delay-spread per cluster values are c<strong>on</strong>stants.<br />

• The mean angle per cluster <strong>and</strong> the delay offset per cluster distributi<strong>on</strong>s are functi<strong>on</strong>s of<br />

the total (per segment) angle-spread <strong>and</strong> the total (per segment) delay-spread.<br />

Large-scale parameters of the <strong>channel</strong> have clear influence <strong>on</strong> the <strong>channel</strong> characteristics. This can be<br />

noticed in delay domain characteristics through the RMS delay spread <strong>and</strong> in the angle domain through<br />

the RMS angle-spread in departure <strong>and</strong> in arrival. The RMS delay spread has influence <strong>on</strong> power delay<br />

spectrum <strong>and</strong> <strong>on</strong> the probability density functi<strong>on</strong> (pdf) of path delays through the parameter r τ .. The<br />

statistical distributi<strong>on</strong>s that generate spatial properties of the ZDSCs are functi<strong>on</strong>s of RMS angle-spread<br />

through azimuth angle propati<strong>on</strong>ality factor ( r ϕ ) <strong>and</strong> RMS azimuth angle-spread ( σ ϕ ) in the arrival side,<br />

<strong>and</strong> through departure angle proporti<strong>on</strong>ality factor ( r φ ) <strong>and</strong> RMS departure angle-spread ( σ φ ) in the<br />

departure side. The dispersi<strong>on</strong> parameters σ ϕ <strong>and</strong> σ φ are sometimes correlated with log-normal<br />

shadowing (LNS), which is important for interference calculati<strong>on</strong>s, h<strong>and</strong>over algorithms, etc. For each set<br />

of RMS delay spread <strong>and</strong> RMS angle-spread departure, RMS angle-spread departure arrival <strong>and</strong> LNS<br />

within each <strong>channel</strong> segment, correlati<strong>on</strong> between them has to be c<strong>on</strong>sidered. These large-scale<br />

parameters are often <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in literature to have log-normal distributi<strong>on</strong>s.<br />

Our framework allows for any distributi<strong>on</strong> for the large-scale parameters <strong>and</strong> also introduces a modelling<br />

of the auto-correlati<strong>on</strong> over the service area. This is achieved by using scenario <strong>and</strong> parameter specific<br />

g ⋅ to transform the large-scale parameters into a domain where they can be treated as<br />

transformati<strong>on</strong>s ( )<br />

Gaussian. The mean, µ , cross-correlati<strong>on</strong> <strong>and</strong> auto-correlati<strong>on</strong> matrix R ( 0)<br />

are then defined in the<br />

transformed domain. The realizati<strong>on</strong>s of the large-scale parameters are then obtained as<br />

−1<br />

0.5<br />

−1<br />

0.5<br />

0.5<br />

5<br />

g R ? x, y + µ<br />

⋅<br />

R 0 is obtained as R ( 0) = EΛ<br />

0.<br />

( ( ) ), where g ( ) is the inverse transform, <strong>and</strong> ( )<br />

T<br />

from the eigen-decompositi<strong>on</strong> R( 0 ) = EΛE<br />

of R ( 0)<br />

. The auto-correlati<strong>on</strong> is achieved by generating m<br />

( m = 6 for A1, <strong>and</strong> m = 4 for all other scenarios) independent Gaussian r<strong>and</strong>om processes,<br />

? x, y = ξ1 x,<br />

y Kξ<br />

m<br />

x,<br />

y , each <strong>on</strong>e with mean zero <strong>and</strong> variance <strong>on</strong>e in the positi<strong>on</strong>s x, y where the<br />

( ) [ ( ) ( )] T<br />

mobiles are located. The auto-correlati<strong>on</strong> of the process c<br />

( x,<br />

y)<br />

2<br />

E{ ξ ( x y ) ξ ( x , y )} = exp( − r / λ ), where ( ) ( ) 2<br />

c<br />

1, 1 c 2 2<br />

∆<br />

c<br />

∆ r =<br />

no auto-correlati<strong>on</strong> mode (NACM) in which the parameters<br />

the r<strong>and</strong>om variable ?( x, y) [ ξ ( x,<br />

y) ( x y)<br />

] T<br />

1<br />

Kξ<br />

m<br />

,<br />

x<br />

1 − x0<br />

+ y1<br />

− y0<br />

m<br />

ξ is given by<br />

. However, we also define a<br />

λ , K ,λ are all set to zero, or equivalently,<br />

= , is r<strong>and</strong>omized independently for each locati<strong>on</strong>. The<br />

required parameters for generating the correlated large-scale parameters are thus the transformati<strong>on</strong><br />

~<br />

s ( x , y)<br />

= g( s( x,<br />

y)<br />

) (or actually its inverse), the mean µ <strong>and</strong> correlati<strong>on</strong> R ( 0)<br />

of the transformed largescale<br />

parameters, <strong>and</strong> the de-correlati<strong>on</strong> distance parameters λ , K 1<br />

,λm<br />

.<br />

This informati<strong>on</strong> is available in Table 3.1 to Table 3.5. Table 3.1 lists the distributi<strong>on</strong> functi<strong>on</strong> for each<br />

modelled parameter in each scenario. For normally distributed r<strong>and</strong>om variables the original <strong>and</strong><br />

transformed variable is identical, except for the delay-spread in scenario B3, where the transformati<strong>on</strong> is a<br />

9<br />

multiplicati<strong>on</strong> with a factor 10 (for numerical reas<strong>on</strong>s). For parameters of log-Gumbel <strong>and</strong> log-Logistic<br />

distributi<strong>on</strong>, the transformati<strong>on</strong> (<strong>and</strong> their inverse) are given by:<br />

~<br />

−<br />

s = g s = −Q<br />

1 F log s ,ν ,ς<br />

(3.1)<br />

<strong>and</strong><br />

( ) (<br />

Gumbel( 10<br />

( ) ))<br />

−1<br />

( ~ −1<br />

s = g s ) = exp log(10) F Q(<br />

~ s ),ν ,ς<br />

Gumbel<br />

−<br />

1<br />

( ( ))<br />

1 ( F Logistic 10<br />

)<br />

−1<br />

( log(10) F ( Q(<br />

~ s ),ν ,ς ))<br />

−<br />

( s) = −Q<br />

( log ( s)<br />

,ν ,ς )<br />

(3.2)<br />

~<br />

s = g<br />

, (3.3)<br />

−1<br />

s = g ( ~ s ) = exp<br />

Logistic<br />

−<br />

(3.4)<br />

Page 18 (167)


WINNER D5.4 v. 1.4<br />

respectively, where F<br />

Gumbel( x,ν ,ς ) <strong>and</strong> Logistic ( x,ν,ς )<br />

distributi<strong>on</strong>s defined in Secti<strong>on</strong> 5.4.3, <strong>and</strong> Q −1<br />

( x)<br />

variables i.e.<br />

Q<br />

F are the CDF of the Gumbel <strong>and</strong> Logistic<br />

1<br />

x<br />

∫<br />

−∞<br />

( x) = exp⎜<br />

⎟dt<br />

2π<br />

is the inverse of the CDF for Gaussian r<strong>and</strong>om<br />

⎛ − t<br />

⎜<br />

⎝ 2<br />

2<br />

⎞<br />

⎟<br />

⎠<br />

. (3.5)<br />

In Table 3.3, the so-called positi<strong>on</strong> ν <strong>and</strong> scale ς parameters for the distributi<strong>on</strong>s are listed, except for<br />

Scenario A1 (with 6 instead of 4 parameters) which is listed in Table 3.5. This means that if the largescale<br />

parameter c is log-Gumbel or log-Logistic, the transformed distributi<strong>on</strong> will have zero mean, <strong>and</strong><br />

unit variance, i.e., µ<br />

c<br />

= 0 <strong>and</strong> R c, c ( 0) = 1.<br />

This can be understood by noting that the mean <strong>and</strong> variance<br />

are taken into account already in the transformati<strong>on</strong>. For log-normal distributi<strong>on</strong>s, we use the<br />

transformati<strong>on</strong><br />

~<br />

= g( s) = log ( s)<br />

(3.6)<br />

s<br />

10<br />

s<br />

g<br />

=<br />

−1<br />

(<br />

~<br />

~ s<br />

s ) = 10<br />

with the excepti<strong>on</strong> of shadow-fading (or sometimes called log-normal shadowing, LNS) where we use<br />

~<br />

= g( s) = 10log ( s)<br />

(3.8)<br />

s<br />

10<br />

s =<br />

−<br />

g<br />

1<br />

(<br />

~ 0.1 s<br />

s ) = 10<br />

~<br />

in order to get the transformed shadow-fading in dB scale. For a log-normal distributed parameter c , the<br />

mean<br />

µ <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> ( 0)<br />

c<br />

R are the mean ν <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> ς listed in Table 3.3.<br />

c,c<br />

For normally distributed bulk parameters no transformati<strong>on</strong> is required (i.e. the transformed <strong>and</strong><br />

untransformed value are identical) <strong>and</strong> thus the mean <strong>and</strong><br />

in Table 3.3.<br />

µ <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> ( 0)<br />

c<br />

c,c<br />

(3.7)<br />

(3.9)<br />

R are listed<br />

In Table 3.5, the cross-correlati<strong>on</strong> between the transformed parameters are listed for scenario A1, <strong>and</strong> in<br />

Table 3.2 for the other scenarios. In teRMS of R ( 0)<br />

, the cross-correlati<strong>on</strong> between parameters r <strong>and</strong> c is<br />

given by<br />

c r , c<br />

r,<br />

r<br />

r,<br />

c<br />

( 0)<br />

( 0) R ( 0)<br />

R<br />

= . (3.10)<br />

R<br />

Thus by combining the cross-correlati<strong>on</strong> <strong>and</strong> variance informati<strong>on</strong>, the matrix R ( 0)<br />

can be derived. In<br />

Table 3.3, a correlati<strong>on</strong> distance ∆ is listed for each large-scale parameter. The correlati<strong>on</strong> distance is<br />

based <strong>on</strong> fitting of a single exp<strong>on</strong>ential exp( − ∆r / ∆)<br />

to the auto-correlati<strong>on</strong> functi<strong>on</strong> of the transformed<br />

large-scale parameter. This value is based <strong>on</strong> measurements or literature or a combinati<strong>on</strong> thereof.<br />

However, since the true auto-correlati<strong>on</strong> actually follows the equati<strong>on</strong> (*) of Secti<strong>on</strong> 4.1.4.1.4, i.e.<br />

2<br />

E { ( x , y ) s( x y )} = R( ∆r)<br />

, ( ) ( ) 2<br />

R<br />

s<br />

1 1 2,<br />

2<br />

⎛<br />

⎜<br />

⎝<br />

⎛<br />

⎜<br />

⎝<br />

c,<br />

c<br />

∆ r = x<br />

(3.11)<br />

2 − x1<br />

+ y2<br />

− y1<br />

∆r<br />

⎞ ⎛ ∆r<br />

⎞⎞<br />

⎟<br />

K<br />

⎜ ⎟⎟<br />

(*). (3.12)<br />

λ1<br />

⎠ ⎝ λm<br />

⎠⎠<br />

0.5<br />

0.5,T<br />

( ∆r) = R ( 0) diag⎜exp⎜−<br />

⎟,<br />

,exp⎜−<br />

⎟⎟R<br />

( 0)<br />

. This means<br />

c,<br />

c , will be a mixture of the m exp<strong>on</strong>entials of (*). However,<br />

they are selected in a way that the results are roughly the same as the single exp<strong>on</strong>ential. The values of the<br />

“eigenvalue auto-correlati<strong>on</strong> distances” λ<br />

1,<br />

K ,λm<br />

are listed in Table 3.4. Note that there is no <strong>on</strong>e-to-<strong>on</strong>e<br />

mapping between any of the lambda parameters <strong>and</strong> any of the large-scale parameters. The correlati<strong>on</strong><br />

distance ∆ is included to allow a more easy interpretati<strong>on</strong> of the auto-regressive characteristics of the<br />

model.<br />

0.5<br />

T 0.5<br />

5<br />

where R ( 0)<br />

is obtained from the eigendecompositi<strong>on</strong> R( 0) = EΛE<br />

as R ( 0) = EΛ<br />

0.<br />

that each autocorrelati<strong>on</strong> functi<strong>on</strong>, R ( ∆r)<br />

The justificati<strong>on</strong> for the expressi<strong>on</strong> (*) is that it produces a model from which it is computati<strong>on</strong>ally simple<br />

to generate data, <strong>and</strong> which at the same time gives a fit to experimental auto-correlati<strong>on</strong> functi<strong>on</strong>s which<br />

is typically equally good as the single exp<strong>on</strong>ential modelling.<br />

The derivati<strong>on</strong> of some of parameters λ<br />

1,<br />

K ,λm<br />

for each scenario, <strong>and</strong> in some case also other<br />

parameters, are given in Secti<strong>on</strong> 5.4.13 below.<br />

Page 19 (167)


WINNER D5.4 v. 1.4<br />

The values <strong>and</strong> distributi<strong>on</strong>s were obtained from measurements at 5 GHz <strong>and</strong> from literature.<br />

In simulati<strong>on</strong>s which include both LOS <strong>and</strong> NLOS mobiles, the large-scale parameters of the LOS <strong>and</strong><br />

NLOS mobiles are modelled as independent, <strong>and</strong> thus they should be generated separately.<br />

Delayspread<br />

AoD<br />

spread<br />

AoA<br />

spread<br />

σ<br />

τ<br />

σ<br />

φ<br />

σ<br />

ϕ<br />

Table 3.1: Distributi<strong>on</strong> functi<strong>on</strong>s of large-scale parameters.<br />

A1 B1 B3 C1 C2 D1<br />

LOS NLOS LOS NLOS LOS NLOS LOS NLOS NLOS LOS NLOS<br />

LN LN Gumb Gumb N N LN LN LN LN LN<br />

LN LN Logist Gumb N N LN LN LN LN LN<br />

LN LN Logist Gumb N N LN LN LN LN LN<br />

Shadowing LN LN LN LN LN LN LN LN LN LN LN<br />

AoD<br />

Elevati<strong>on</strong><br />

spread σ<br />

θ<br />

AoA<br />

Elevati<strong>on</strong><br />

spread σ<br />

ϕ<br />

LN<br />

LN<br />

LN<br />

LN<br />

N<br />

LN<br />

Gumb<br />

Logist<br />

Normal (Gaussian)<br />

Log-normal, i.e., log10(Gauss)<br />

Log-Gumbel<br />

Log-Logistic<br />

Scenarios<br />

Table 3.2: Cross-correlati<strong>on</strong> between large-scale parameters.<br />

B1 B3 C1 C2 D1<br />

LOS NLOS LOS NLOS LOS NLOS NLOS LOS NLOS<br />

σ<br />

φ vs σ<br />

τ 0.50 0.18 0.17 0.13 -0.29 0.3 0.4 -0.07 -0.35<br />

Cross-Correlati<strong>on</strong>s<br />

σ<br />

ϕ vs σ<br />

τ 0.76 0.42 -0.2 0.49 0.78 0.7 0.6 0.21 0.12<br />

σ<br />

ϕ vs LNS -0.45 -0.40 -0.17 0.11 -0.16 -0.3 -0.3 -0.11 0.13<br />

σ<br />

φ vs LNS -0.50 0.01 -0.32 -0.18 0.36 -0.4 -0.6 -0.07 0.60<br />

σ<br />

τ vs LNS -0.41 -0.65 0.17 0.34 -0.71 -0.4 -0.4 -0.71 -0.51<br />

σ<br />

φ vs σ<br />

ϕ 0.37 0.07 0.19 0.28 -0.35 0.3 0.4 -0.49 -0.15<br />

Note: Sign of LNS has been defined so that positive LNS means more received power at MS than predicted by PL<br />

model.<br />

Scenarios<br />

Table 3.3: Distributi<strong>on</strong>s parameters of large-scale parameters.<br />

B1 B3 C1 C2 D1<br />

LOS NLOS LOS NLOS LOS NLOS LOS LOS NLOS<br />

ν -7.38 -7.09 26 45 -8.8 -7.26 -6.63 -7.8 -7.6<br />

σ τ<br />

ζ 0.24 0.11 8.2 6.9 0.49 0.33 0.32 0.57 0.48<br />

Page 20 (167)


WINNER D5.4 v. 1.4<br />

∆<br />

τ<br />

(m)<br />

6.0 5.0 4.5 1.82 64 40 40 64.2 36.3<br />

ν 0.40 1.24 26.4 38 1.14 0.53 0.93 1.22 0.96<br />

σ φ<br />

ζ 0.23 0.20 10.5 11.7 0.12 0.36 0.22 0.21 0.45<br />

∆<br />

φ 13.2 2.4 2.2 0.62 2.0 30 50 24.8 2.7<br />

σ ϕ<br />

LNS<br />

Notes:<br />

ν 1.4 1.6 13.1 9.5 1.61 1.67 1.72 1.52 1.52<br />

ζ 0.12 0.19 7.6 4.5 0.20 0.3 0.14 0.18 0.27<br />

∆<br />

ϕ<br />

(m)<br />

ζ<br />

(dB)<br />

∆<br />

LNS<br />

(m)<br />

1.6 3.2 0.83 0.61 18.2 30 50 3.5 15.1<br />

2.3 3.1 1.4 2.1<br />

4.0<br />

6.0<br />

8 8<br />

9.1 5.2 4.36 6.16 23.0 50 50 40 120<br />

1. Values for ∆ are merely provided for informati<strong>on</strong>. Values of λ (see table below) are used in<br />

coefficient generati<strong>on</strong>.<br />

2. Scenarios C1 LOS <strong>and</strong> D1 LOS c<strong>on</strong>tain two shadowing std. deviati<strong>on</strong>s; <strong>on</strong>e (top) for before <strong>and</strong><br />

<strong>on</strong>e (bottom) for after the path-loss breakpoint.<br />

Parameters:<br />

ν: Locati<strong>on</strong> parameter (i.e., mean in case of normal distributi<strong>on</strong>)<br />

ζ: Scale parameter (i.e., st<strong>and</strong>ard deviati<strong>on</strong> in case of normal distributi<strong>on</strong>)<br />

∆: Correlati<strong>on</strong> distance of normal variable<br />

3.5<br />

6.0<br />

8.0<br />

Table 3.4: Lambda parameters.<br />

A1 B1 B3 C1 C2 D1<br />

LOS NLOS LOS NLOS LOS NLOS LOS NLOS NLOS LOS NLOS<br />

λ<br />

1 (m) 2.0 3.5 2.0 5.0 4.5 7.0 40.0 44.0 50.0 3.0 15.0<br />

λ<br />

2 (m) 2.0 2.0 12.0 2.3 0.8 0.6 2.0 30.0 45.0 10.0 2.0<br />

λ<br />

3 (m) 2.0 3.0 3.0 3.0 4.0 1.8 35.0 30.0 40.0 60.0 15.0<br />

λ<br />

4 (m) 3.0 2.5 9.1 5.2 2.2 0.6 27.0 47.0 52.0 42.0 120.0<br />

λ<br />

5 (m) 3.5 5.0<br />

λ<br />

6 (m) 6.0 4.0<br />

Table 3.5: Distributi<strong>on</strong> parameters for A1 sub-scenarios.<br />

Scenario Correlati<strong>on</strong> coefficients St<strong>and</strong>ard<br />

deviati<strong>on</strong>s<br />

ς<br />

Means<br />

γ<br />

Decorrelati<strong>on</strong><br />

distance ∆<br />

(m)<br />

LOS 1,00 0,46 0,74 -0,68 0,49 0,63<br />

0,46 1,00 0,40 -0,05 0,77 0,38<br />

0.27<br />

0.31<br />

-7.40<br />

0.74<br />

7.00<br />

5.90<br />

Page 21 (167)


WINNER D5.4 v. 1.4<br />

0,74 0,40 1,00 -0,44 0,42 0,83<br />

-0,68 -0,05 -0,44 1,00 -0,11 -0,28<br />

0,49 0,77 0,42 -0,11 1,00 0,44<br />

0,63 0,38 0,83 -0,28 0,44 1,00<br />

0.26<br />

3.10<br />

0.20<br />

0.22<br />

1.52<br />

0.00<br />

0.88<br />

0.94<br />

2.30<br />

6.00<br />

1.30<br />

3.50<br />

NLOS 1,00 -0,10 0,31 -0,50 -0,61 -0,05<br />

-0,10 1,00 -0,26 -0,01 0,20 -0,14<br />

0,31 -0,26 1,00 -0,41 -0,28 -0,19<br />

-0,50 -0,01 -0,41 1,00 0,25 0,10<br />

-0,61 0,20 -0,28 0,25 1,00 0,45<br />

-0,05 -0,14 -0,19 0,10 0,45 1,00<br />

0.19<br />

0.23<br />

0.14<br />

3.50<br />

0.21<br />

0.17<br />

-7.60<br />

1.30<br />

1.57<br />

0.00<br />

1.06<br />

1.10<br />

Order of parameters: delay-spread, AoD azimuth-spread, AoA azimuth-spread, shadowing, AoD<br />

elevati<strong>on</strong>-spread, AoA elevati<strong>on</strong>-spread<br />

4.20<br />

4.90<br />

2.50<br />

3.40<br />

3.20<br />

2.60<br />

3.1.2 Average power of ZDSC c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> their delays<br />

The average power of every ZDSC is calculated in delay domain as explained in Chapter 4. Two<br />

functi<strong>on</strong>s are required to calculate the expected power of each ZDSC c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> their delays. They<br />

are the power delay spectrum <strong>and</strong> the probability density functi<strong>on</strong> of ZDSC delays. It is shown in<br />

Chapter 4, that for the case when both P ( τ ) <strong>and</strong> f ( τ ) are exp<strong>on</strong>ential, the expected power of ZDSC<br />

depends <strong>on</strong> the value of the parameter r τ <strong>and</strong> the RMS delay spread of the <strong>channel</strong> segmentσ . In order<br />

to make the average power of ZDSC varying from delay to delay <strong>and</strong> from <strong>on</strong>e <strong>channel</strong> segment to<br />

another in a similar manner that is usually seen in measurement results, the shadowing r<strong>and</strong>omizati<strong>on</strong><br />

effect <strong>on</strong> each ZDSC is modelled. Thus, the expected power of ZDSC of each segment is obtained as:<br />

'<br />

P n<br />

−ζ<br />

2 ⎛<br />

{ }<br />

( rτ<br />

−1) ⎞<br />

10<br />

α τ ∝ exp⎜−τ′<br />

⎟ 10<br />

= E<br />

⎜ rτ<br />

σ ⎟ ⋅ , (3.13)<br />

⎝<br />

τ ⎠<br />

where ζ<br />

n is an i.i.d. Gaussian r<strong>and</strong>om variable with zero mean <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> ζ , <strong>and</strong> the delay<br />

τ ′ is the normalized delay to the delay of the first arrival ZDSC. The normalized delay of the first arrival<br />

ZDSC is zero. For the case when the ZDSC delays have uniform distributi<strong>on</strong>, the expected power of<br />

ZDSC of each segment is obtained as:<br />

'<br />

P n<br />

−ξn<br />

2<br />

10<br />

{ α τ } ∝ exp( −τ<br />

' σ τ<br />

) ⋅10<br />

= E<br />

(3.14)<br />

The ZDSC delay distributi<strong>on</strong>s <strong>and</strong> power delay spectrum of different scenarios are presented in Table 3.6.<br />

For calculati<strong>on</strong> of <strong>channel</strong>s with cross-polarisati<strong>on</strong>, the cross-polarisati<strong>on</strong> ratio (XPR) for vertical to<br />

horiz<strong>on</strong>tal (XPR V ) <strong>and</strong> for horiz<strong>on</strong>tal to vertical (XPR H ) are needed (for definiti<strong>on</strong> see 5.4.12). The values<br />

XPR V <strong>and</strong> XPR V of different scenarios are given in Table 3.6.<br />

n<br />

τ<br />

Scenarios<br />

ZDSC<br />

Delay<br />

distributi<strong>on</strong><br />

Table 3.6: Formulae for calculating the ZDSC power c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> delay for the c<strong>on</strong>sidered<br />

scenarios <strong>and</strong> XPR V <strong>and</strong> XPR H .<br />

A1 B1 B3 C1 C2 D1<br />

LOS NLOS LOS NLOS LOS NLOS LOS NLOS NLOS LOS NLOS<br />

Exp Exp Exp Uniform<br />

(0,800ns)<br />

Exp<br />

(0,130ns)<br />

Exp<br />

(0,220ns)<br />

Exp Exp Exp Exp Exp<br />

rτ<br />

3.0 2.4 3.2 2.2 1.90 1.58 2.4 1.5 2.3 3.8 1.7<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

ϒ<br />

r r r<br />

τ<br />

τ<br />

ζ (dB) 3<br />

τ<br />

'<br />

−t 10<br />

P n ϒ στ<br />

− ξ<br />

n<br />

e 10 n<br />

1<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

rτ<br />

−1<br />

r r r r r r r<br />

τ<br />

XPR V µ 11.4 9.7 8.6 8 0.5 0.1 7.9 3.3 7.6 6.9 7.9<br />

τ<br />

τ<br />

τ<br />

τ<br />

Page 22 (167)<br />

τ<br />

τ


WINNER D5.4 v. 1.4<br />

(dB)<br />

XPR H<br />

(dB)<br />

σ<br />

µ<br />

σ<br />

3.4 3.5 1.8 1.8 1.07 0.69 3.3 2.5 3.4 2.3 3.5<br />

10.4 10.0 9.5 6.9<br />

3.4 3.1 2.3 2.8<br />

Notes:<br />

Not<br />

avail.<br />

Not<br />

avail.<br />

Not<br />

avail.<br />

Not<br />

avail.<br />

3.7 5.7 2.3 7.2 7.5<br />

2.5 2.9 0.2 2.8 4.0<br />

1. For scenario B3, XPR H values are not available. In the <strong>channel</strong> model implementati<strong>on</strong>, these<br />

values have been substituted by XPR V .<br />

2. Distributi<strong>on</strong> of XPR is log-normal, i.e., XPR = 10 X/10 , where X is Gaussian with st<strong>and</strong>ard<br />

deviati<strong>on</strong> σ <strong>and</strong> mean µ.<br />

Average powers of the ZDSC are normalized so that the total power of all ZDSCs is equal to <strong>on</strong>e. Then,<br />

the normalized power of the nth ZDSC is<br />

P<br />

'<br />

n<br />

n<br />

= Q<br />

P<br />

∑<br />

n=<br />

1<br />

P<br />

'<br />

n<br />

(3.15)<br />

where Q is the number of ZDSCs. For the case when LOS model is used, the power of the direct<br />

comp<strong>on</strong>ent is c<strong>on</strong>sidered in the normalizati<strong>on</strong> such that the ratio of the direct power to the scattered power<br />

is the K-factor.<br />

'<br />

n<br />

n<br />

=<br />

Q<br />

P<br />

P<br />

( K + 1)∑<br />

n=<br />

1<br />

The K-factor for LOS scenarios can be calculated as given in Table 3.7.<br />

P<br />

'<br />

n<br />

, (3.16)<br />

k<br />

P D<br />

= . (3.17)<br />

k +1<br />

Table 3.7: K factor formulae for LOS scenarios.<br />

Scenarios A1 B1 B3 C1 D1<br />

K [dB] 8.7 + 0.051*d 3+ 0.0142d 6 - 0.26*d 17.1 – 0.021*d 3.7 + 0.019*d<br />

Distance d is in m.<br />

It should be noted that when LOS comp<strong>on</strong>ent exists, the ZDSC will have 10+1 rays.<br />

3.1.3 Directi<strong>on</strong>al distributi<strong>on</strong>s of ZDSCs<br />

There are two types of angle informati<strong>on</strong> for each ZDSC. These are the mean angle <strong>and</strong> the offset angles<br />

of each ray from the mean within each cluster. The zero mean azimuth departure (azimuth arrival) angle<br />

is the transmitter (receiver) broadside directi<strong>on</strong>. The generated mean angle of departures <strong>and</strong> angle of<br />

arrivals are relative to the direct path between transmitter <strong>and</strong> receiver with respect to the broadside of<br />

transmitter or receiver, respectively. The angle definiti<strong>on</strong>s <strong>and</strong> references that are used in the generic<br />

<strong>channel</strong> model are the same as those presented in [3GPP SCM]. The relati<strong>on</strong> between the mean azimuthdeparture<br />

<strong>and</strong> the mean azimuth-arrival probability density functi<strong>on</strong>s of each ZDSC <strong>and</strong> their delays is<br />

generated through correlated large-scale parameters used in the corresp<strong>on</strong>ding density functi<strong>on</strong>s. We<br />

fixed the power azimuth spectrum of each ZDSC at the departure <strong>and</strong> the arrival sides assuming it as<br />

Laplacian. The RMS angle-spread of each ZDSC is fixed to <strong>on</strong>e value, which may be different in<br />

departure from arrival <strong>and</strong> may vary from scenario to scenario. However, the power azimuth spectrum<br />

(PAS) <strong>and</strong> the angle-spread values (AS ) of each ZDSC can be changed in the model if needed. The<br />

distributi<strong>on</strong> parameters of the mean angle of departure (AoD) <strong>and</strong> the mean angle of arrival (AoA) the<br />

ZDSCs may vary from <strong>on</strong>e scenario to another. The ZDSC azimuth-departure PAS <strong>and</strong> azimuth-arrival<br />

PAS are defined by 10 rays having predefined offset angles for Laplacian PAS from the mean angles of<br />

Page 23 (167)


WINNER D5.4 v. 1.4<br />

the ZDSC. The 10 rays are spaced in angle domain <strong>and</strong> have identical power. The power of each ray is<br />

P n /10, where P n is the average power of the nth ZDSC. The offset angle spacing depends <strong>on</strong> the value of<br />

AS of ZDSC_D for the departure side <strong>and</strong> value of AS of ZDSC_A for the arrival side. The AS<br />

<strong>and</strong><br />

φ<br />

AS are per ZDSC <strong>and</strong> are different from the<br />

ϕ<br />

φ<br />

ϕ<br />

σ <strong>and</strong>σ , respectively, which are the composite<br />

angle-spread involving all ZDSC. The rays of the ZDSC have r<strong>and</strong>om phases. The angle distributi<strong>on</strong>s <strong>and</strong><br />

power azimuth spectrum at the transmitter or receiver sides may vary from <strong>on</strong>e scenario to another. Table<br />

3.8 states the angle distributi<strong>on</strong>s of different scenarios that are defined in WINNER generic <strong>channel</strong><br />

model. The σ <strong>and</strong> σ influence in generati<strong>on</strong> of the ZDSC directi<strong>on</strong>al informati<strong>on</strong> through the<br />

φ<br />

ϕ<br />

parameters r<br />

φ <strong>and</strong>r ϕ , respectively. The locati<strong>on</strong> parameter for all distributi<strong>on</strong>s is zero <strong>and</strong> the scaling<br />

parameters are defined by r φ<br />

σ φ for departure angles <strong>and</strong> by r ϕ<br />

σ ϕ for arrival angles. The generati<strong>on</strong> of<br />

the angle offsets from the rays from cluster mean angle depends <strong>on</strong> AS<br />

φ for departure side <strong>and</strong> <strong>on</strong><br />

ASϕ<br />

arrival cluster. The offset angles are determined by multiplicati<strong>on</strong> of the angles spread per ZDSC<br />

with basis vector of offset angles (BO), e.g., AS<br />

φ *BO. Table 3.9 states the number of ZDSCs <strong>and</strong> the<br />

number of rays in each cluster as well as their AS<br />

φ <strong>and</strong> AS<br />

ϕ for the c<strong>on</strong>sidered scenarios. The angles<br />

of BO vector <strong>and</strong> the calculati<strong>on</strong> of the offset angles are presented in Table 3.10.<br />

ϕ<br />

φ<br />

Scenarios<br />

AoD<br />

distributi<strong>on</strong><br />

AoD<br />

scaling<br />

parameter<br />

AoA<br />

distributi<strong>on</strong><br />

AoA<br />

scaling<br />

parameter<br />

Table 3.8: Distributi<strong>on</strong>s of azimuth <strong>and</strong> departure angles.<br />

A1 B1 B3 C1 C2 D1<br />

LOS NLOS LOS NLOS LOS NLOS LOS NLOS NLOS LOS NLOS<br />

Wrapped Gaussian<br />

2.0σ φ 1.2σ φ 3.4σ φ 1.1σ φ 1.9σ φ 1.3σ φ 1.4σ ϕ 2.3σ φ 3.2σ φ 0.8σ ϕ 1.2σ<br />

φ<br />

Wrapped Gaussian<br />

1.7σ ϕ 2.1σ ϕ 3.6σ ϕ 3.6σ ϕ 1.5σ ϕ 1.6σ ϕ 1.8σ ϕ 1.8σ ϕ 3.2σ ϕ 2.2σ ϕ 1.3σ<br />

ϕ<br />

Scenarios<br />

Table 3.9: Number of ZDSCs <strong>and</strong> the number of rays in each cluster.<br />

A1 B1 B3 C1 C2 D1<br />

LOS NLOS LOS NLOS LOS NLOS LOS NLOS NLOS LOS NLOS<br />

Number of ZDSC 16 11 8 16 15 24 15 14 20 11 10<br />

Rays per ZDSC 10<br />

AS<br />

φ (deg) 5 5 3 10 4.7 5.5 5 2 2 1.5 1.5<br />

AS<br />

ϕ (deg) 5 5 18 22 5.4 12.5 5 10 15 3 3<br />

Table 3.10: Offset angles Rays within a ZDSC as a functi<strong>on</strong> of<br />

AS<br />

φ ,<br />

Ray number Basis vector offset angles (BO) Rays offset angles<br />

1,2 ± 0.0742<br />

3,4 ± 0.2532<br />

5,6 ± 0.4986<br />

7,8 ± 0.8913<br />

AS<br />

ϕ , AS ϕ , <strong>and</strong> AS θ .<br />

OA = AS X BO<br />

Page 24 (167)


WINNER D5.4 v. 1.4<br />

9,10 ± 1.9718<br />

OA : Offset angles, BO: Basis vector of offset angles<br />

Rays angle offsets of different values of<br />

angle-spread values of<br />

AS<br />

φ or<br />

AS<br />

φ or<br />

AS<br />

ϕ in Table 3.9 can be obtained by multiplying the<br />

AS<br />

ϕ times each angle of the basis vector offset angles (BO). For<br />

example AS<br />

φ = 5; the offset angles of the rays within a ZDSC is OA = 5 x BO = [ ±0.2226 ±0.7596<br />

±1.4960 ±2.6740 ±5.9154].<br />

Elevati<strong>on</strong> angle informati<strong>on</strong> are important for indoor envir<strong>on</strong>ments, i.e., Scenario A1 <strong>and</strong> B3. The generic<br />

model includes elevati<strong>on</strong> plane for these scenarios. The elevati<strong>on</strong> plane model parameters of these two<br />

scenarios are given in Table 3.11.<br />

Table 3.11: Elevati<strong>on</strong> plane model parameters.<br />

Scenarios<br />

A1<br />

B3<br />

LOS NLOS LOS NLOS<br />

ν (S-dB) 0.88 1.06<br />

σ θ<br />

ζ (S-dB) 0.20 0.21<br />

∆<br />

τ (m) 1.3 3.2<br />

ν (S-dB) 0.94 1.10<br />

σ ϑ<br />

Elevati<strong>on</strong> AoD distributi<strong>on</strong><br />

ζ (S-dB) 0.22 0.17<br />

∆<br />

τ (m) 3.5 2.6<br />

Wrapped<br />

Gaussian<br />

AoD scaling parameter 1.9 1.4<br />

Elevati<strong>on</strong> AoA distributi<strong>on</strong><br />

Wrapped<br />

Gaussian<br />

AS θ (deg) 3 3<br />

AS ϕ (deg) 3 3<br />

Cross-Correlati<strong>on</strong>s<br />

σ<br />

θ vs σ<br />

τ 0.46 -0.61<br />

σ<br />

ϑ vs σ<br />

τ 0.74 -0.05<br />

σ<br />

θ vs LNS -0.05 0.25<br />

σ<br />

ϑ vs LNS -0.44 0.11<br />

σ<br />

ϑ vs σ<br />

θ 0.44 0.45<br />

3.1.4 Antenna gain<br />

In principle the <strong>channel</strong> model is antenna independent at both fixed stati<strong>on</strong> <strong>and</strong> mobile stati<strong>on</strong>. Any 2D<br />

antenna c<strong>on</strong>figurati<strong>on</strong> <strong>and</strong> pattern can be embedded in the model. If elevati<strong>on</strong> plane parameters are<br />

included, 3D antenna geometries can be embedded. For example, <strong>on</strong>e can use the 3GPP antenna pattern<br />

in WINNER model. The antenna pattern that has been used in [3GPP SCM] at the BS is 3-sector antenna<br />

used for each sector. It is specified by:<br />

A<br />

( γ )<br />

⎡ ⎛ γ ⎞<br />

=−min⎢12<br />

⎜ ⎟<br />

⎢ ⎝γ<br />

3dB<br />

⎣ ⎠<br />

2<br />

, Am<br />

⎤<br />

o<br />

o<br />

⎥, where 180 < φ


WINNER D5.4 v. 1.4<br />

where γ is defined as the angle between the directi<strong>on</strong> of interest <strong>and</strong> the boresight of the antenna. The<br />

γ<br />

3dB is the 3dB beamwidth in degrees, <strong>and</strong> A m is the maximum attenuati<strong>on</strong>. For a 3 sector scenario γ<br />

3dB<br />

is 70 degrees, A m = 20dB. However, other antenna patterns can also be used, if needed.<br />

3.1.5 Path-loss <strong>models</strong><br />

Path-loss <strong>models</strong> at 5 GHz for c<strong>on</strong>sidered scenarios have been developed based <strong>on</strong> measurement results<br />

or from literature. The developed path <strong>models</strong> are presented in Table 3.12 including the shadow fading<br />

values. The path-loss <strong>models</strong> have the form as in (3.21), where d is the distance between transmitter <strong>and</strong><br />

receiver, the fitting parameter A includes the path-loss exp<strong>on</strong>ent parameter <strong>and</strong> parameter B is the<br />

intercept.<br />

( d ) B<br />

PL = Alog +<br />

(3.19)<br />

Table 3.12: Path-loss <strong>models</strong>.<br />

Scenario path loss [dB] shadow<br />

fading<br />

st<strong>and</strong>ard<br />

dev.<br />

applicability<br />

range<br />

A1<br />

B1<br />

B3<br />

LOS 18.7 log 10 (d[m]) + 46.8 σ = 3.1 dB 3 m < d < 100 m<br />

NLOS 36.8 log 10 (d[m]) + 38.8 σ = 3.5 dB 3 m < d < 100 m<br />

LOS 22.7 log 10 (d[m])+41.0 σ = 2.3dB 10 m < d < 650 m<br />

NLOS 0.096 d 1 [m] + 65 +<br />

σ = 3.1dB 10 m < d 1 < 550 m<br />

(28 – 0.024d 1 [m]) log 10 (d 2 [m])<br />

w/2 < d 2 < 450 m *)<br />

LOS 13.4 log 10 (d[m]) + 36.9 s = 1.4 dB 5 m < d < 29 m<br />

NLOS 3.2 log 10 (d[m]) + 55.5 s = 2.1 dB 5 m < d < 29 m<br />

C1<br />

s = 4.0 dB<br />

LOS 23.8 log 10 (d) + 41.6<br />

23.8 log 10 (d BP ) ****) +)<br />

40.0 log 10 (d/d BP ) + 41.6 + s = 6.0 dB,<br />

30 m < d < d BP<br />

d BP < d < 5 km<br />

NLOS 40.2 log 10 (d[m]) + 27.7 **) σ = 8 dB 50 m < d < 5 km<br />

C2 NLOS 35.0 log 10 (d[m]) +38.4 ***) σ = 8 dB 50 m < d < 5 km<br />

D1<br />

σ = 3.5dB<br />

LOS 21.5 log 10 (d[m]) + 44.6<br />

21.5 log 10 (d BP ) ****) +)<br />

40.0 log 10 (d/d BP ) + 44.6 + σ = 6.0dB<br />

30 m < d < d BP<br />

d BP < d < 10 km<br />

NLOS 25.1 log 10 (d[m]) + 55.8 σ = 8.0dB 30 m < d < 10 km<br />

*)<br />

w is LOS street width, d 1 is distance al<strong>on</strong>g main street, d 2 is distance al<strong>on</strong>g perpendicular street.<br />

**) Validity bey<strong>on</strong>d 1 km not c<strong>on</strong>firmed by measurement data.<br />

***)<br />

Validity bey<strong>on</strong>d 2 kms not c<strong>on</strong>firmed by measurement data.<br />

****)<br />

d BP is the break-point distance: d BP = 4 h BS h MS / ?, where h BS is antenna height at BS, h MS is<br />

antenna height at MS, <strong>and</strong> ? is the wavelength. Validity bey<strong>on</strong>d d BP not c<strong>on</strong>firmed by measurement<br />

data.<br />

+)<br />

BS antenna heights in the measurements: C1 LOS: 11.7 m, D1: 19 – 25 m.<br />

3.1.6 Probability of line of sight<br />

System <strong>level</strong> simulati<strong>on</strong>s require the probability of line of sight for c<strong>on</strong>sidered scenarios A1, B1, B3, C1,<br />

C2, <strong>and</strong> D1. They are given as follows:<br />

Page 26 (167)


WINNER D5.4 v. 1.4<br />

3.1.6.1 Scenario A1<br />

3.1.6.2 Scenario B1<br />

⎧<br />

1 d ≤ 2.5m<br />

⎪<br />

P = ⎨<br />

1 − 0.9( 1 − ( 1.24 − 0.61log )<br />

3<br />

) 13<br />

10( d) d > 2.5m<br />

⎪⎩<br />

⎧1 d ≤ 15m<br />

⎪<br />

P = ⎨ 3<br />

1 ( 1 ( 1.56 0.48log ( ))<br />

) 13<br />

⎪ − − −<br />

10<br />

d d > 15m<br />

⎩<br />

(3.20)<br />

(3.21)<br />

where<br />

d = d + d , <strong>and</strong> d 1 <strong>and</strong> d 2 are like in Table 2.9.<br />

2 2<br />

1 2<br />

3.1.6.3 Scenario B3<br />

For the big factory halls, airport <strong>and</strong> train stati<strong>on</strong>s:<br />

⎧1,<br />

d < 10m<br />

P LOS<br />

= ⎨<br />

(3.22)<br />

⎩exp(<br />

−(<br />

d −10) / 45)<br />

For big lecture hall or c<strong>on</strong>ference hall:<br />

⎪<br />

⎧ 1, d < 5m<br />

P LOS<br />

= ⎨ d − 5<br />

(3.23)<br />

1−<br />

,5 m < d < 40 m<br />

⎪⎩ 150<br />

3.1.6.4 Scenario C1<br />

d[m]<br />

P = exp( − )<br />

(3.24)<br />

500m<br />

3.1.6.5 Scenario C2<br />

For scenario C2, <strong>on</strong>ly NLOS is c<strong>on</strong>sidered. In this case P(LOS) = 0.<br />

3.1.6.6 Scenario D1<br />

3.1.7 Generati<strong>on</strong> of <strong>channel</strong> coefficients<br />

d[m]<br />

P = exp( − )<br />

(3.25)<br />

1000m<br />

The generati<strong>on</strong> of <strong>channel</strong> parameters is performed per <strong>channel</strong> segment. During each <strong>channel</strong> segment<br />

the AoAs <strong>and</strong> AoDs, <strong>and</strong> delays of each ZDSC are fixed while the <strong>channel</strong> goes through fast fading<br />

according to the virtual moti<strong>on</strong> of the MS, which has a velocity vector v. The assumed <strong>system</strong> has S<br />

antennas at the transmitter side <strong>and</strong> U antennas at the receiver side. The WINNER generic <strong>channel</strong> model<br />

is a geometric-based stochastic model. There are a large number of r<strong>and</strong>om variables that are incorporated<br />

in the modelling approach. Hence, many parameters must be fixed within the simulati<strong>on</strong> run to make the<br />

computati<strong>on</strong> time feasible. These parameters may differ from <strong>on</strong>e scenario to another. For instance the<br />

ZDSC angle-spread ( AS or AS ) is fixed for all departure <strong>and</strong> arrival ZDSCs but may have different<br />

φ<br />

ϕ<br />

angles. These parameters represent some of the characteristics of different scenarios.<br />

To obtain MIMO <strong>channel</strong> coefficients the following steps are followed:<br />

1) Select <strong>on</strong>e of the scenarios to be simulated: A1, B1, B3, C1, C2, or D1.<br />

2) Assign locati<strong>on</strong>s of transmitters (BS), receivers (MS), separating distance <strong>and</strong> their antenna<br />

orientati<strong>on</strong>s. The orientati<strong>on</strong> of MS antenna is drawn from iid uniform distributi<strong>on</strong> U(0 o ,360 o ).<br />

Assign velocity vector to each MS. Assign LOS situati<strong>on</strong> to each locati<strong>on</strong> according to the<br />

probability.<br />

3) Calculate the path loss associated with transmitter-receiver of every MS <strong>and</strong> every BS if needed.<br />

4) Generate the vector ?( x, y)<br />

in the points i<br />

yi<br />

−1<br />

0.5<br />

obtain the large-scale parameters as R ?( x, y)<br />

x , where MSs are located, see Secti<strong>on</strong> 6.1.3. Then<br />

( µ )<br />

g +<br />

, the parameters can be found in the<br />

Page 27 (167)


WINNER D5.4 v. 1.4<br />

0.5<br />

Tables of Secti<strong>on</strong> Error! Reference source not found.. Note that ( 0)<br />

0.5<br />

5<br />

R ( 0) = EΛ<br />

0.<br />

T<br />

from the eigen-decompositi<strong>on</strong> R( 0 ) = EΛE<br />

of R ( 0)<br />

.<br />

R shall be obtained as<br />

5) Based <strong>on</strong> the generated large-scale parameters:<br />

a. Generate the delays, azimuth AoD <strong>and</strong> azimuth AoA of each ZDSC through r<strong>and</strong>om<br />

variable generators of the corresp<strong>on</strong>ding probability density functi<strong>on</strong>s of the selected<br />

scenario. Generate elevati<strong>on</strong> AoD <strong>and</strong> AoA for indoor Scenarios.<br />

b. Order the delays <strong>and</strong> normalize them to the smallest delay.<br />

h<br />

c. Calculate the average power of each ZDSC within the <strong>channel</strong> segment as described in<br />

Secti<strong>on</strong> 3.1.2. Assign the power of each ray within the ZDSC as P n<br />

/ M , where M is<br />

the number of rays within ZDSC of a specific scenario, which is fixed to 10 rays.<br />

d. AoA <strong>and</strong> AoD are sorted in ascending order of absolute values. Shortest delays are are<br />

assosiated to AoA <strong>and</strong> AoD with smallest absolute values. Respectively, l<strong>on</strong>gest delay<br />

is assosiated to AoA <strong>and</strong> AoD with largest absolute value.<br />

e. Assign angle offset of rays in departure <strong>and</strong> arrival from predefined set of offset angles<br />

of the selected scenario <strong>and</strong> assign r<strong>and</strong>om phases from U(0 o ,360 o ) to the 10 rays of the<br />

ZDSC. Table 3.10 shows how to obtain offset angles of rays as a functi<strong>on</strong> of anglespreads.<br />

f. R<strong>and</strong>omly couple departure rays to arrival rays.<br />

g. Determine the AoD <strong>and</strong> AoA for all rays within each ZDSC with respect to the<br />

broadside of transmitter <strong>and</strong> receiver, respectively.<br />

h. Determine the antenna gain at transmitter G ( AoD ) <strong>and</strong> receiver ( )<br />

where n is the nth ZDSC <strong>and</strong> m is the mth ray within the nth cluster.<br />

i. Apply path loss <strong>and</strong> shadowing to each ray within all ZDSCs.<br />

t<br />

n<br />

G AoA ,<br />

j. With the knowledge of MS velocity vector, fast fading of each ZDSC can be calculated<br />

for every <strong>channel</strong> segment, while the bulk parameters <strong>and</strong> MS locati<strong>on</strong>s remained fixed.<br />

k. For linear array c<strong>on</strong>figurati<strong>on</strong>, the <strong>channel</strong> coefficient h ( t)<br />

u , s,<br />

n<br />

due to the nth ZDSC<br />

for each antenna pair, element s from transmitter <strong>and</strong> element u from receiver is given<br />

by:<br />

j⎡<br />

⎣kd<br />

s sin( φm, n ) +Φm,<br />

n⎤<br />

⎦<br />

Gt( φmn<br />

.<br />

) e<br />

⋅<br />

M ⎜<br />

⎟<br />

jkdu<br />

sin( ϕ m,<br />

n)<br />

usn , ,<br />

() =<br />

nσ<br />

SF∑ ⎜<br />

r( ϕmn<br />

,<br />

) ⋅ ⎟ (3.26)<br />

m = 1 ⎜<br />

⎟<br />

jk v cos( ϕ −θ<br />

) t<br />

h t P G e<br />

⎛<br />

⎜<br />

⎝<br />

e<br />

Assuming that cluster n=1 is the <strong>on</strong>e with normalized delay τ 1 =0 (i.e. the cluster with the<br />

lowest delay of all), an opti<strong>on</strong>al LOS comp<strong>on</strong>ent may be taken into account as follows:<br />

LOS<br />

u , s,<br />

n<br />

( t)<br />

=<br />

using<br />

where<br />

1<br />

h<br />

K + 1<br />

m,<br />

n<br />

v<br />

⎞<br />

⎟<br />

⎠<br />

[ kd sin( θ ) +Φ ]<br />

j<br />

⎛<br />

s BS LOS<br />

G<br />

⎞<br />

⎜<br />

s(<br />

θBS<br />

) e<br />

⋅<br />

σ<br />

⎟<br />

SFK<br />

j(<br />

kdu<br />

sin( θ MS ) +Φ LOS )<br />

+ δ ( n −1)<br />

⋅ ⎜ Gu<br />

( θMS<br />

e<br />

⋅⎟<br />

(3.27)<br />

K + 1<br />

⎜<br />

jk v cos( θ MS −θ<br />

v ) t<br />

⎟<br />

⎝<br />

e<br />

⎠<br />

u, s,<br />

n<br />

)<br />

⎧1 for n=<br />

0<br />

δ ( n)<br />

= ⎨<br />

⎩ 0 else<br />

φ<br />

n,m is the azimuth angle of departure of mth ray within the nth ZDSC.<br />

ϕ<br />

n,m is the azimuth angle of arrival of mth ray within the nth ZDSC.<br />

M is the number of rays within ZDSC, which is 10.<br />

G ( φ<br />

,m<br />

) is the antenna gain of transmitter (BS) for the mth ray within the nth ZDSC.<br />

t<br />

n<br />

r<br />

n<br />

Page 28 (167)


WINNER D5.4 v. 1.4<br />

G ( ϕ<br />

,m<br />

) is the antenna gain of receiver (MS) for the mth ray within the nth ZDSC.<br />

r<br />

d s<br />

d u<br />

k<br />

n<br />

is the distance between antenna elements of the linear array at transmitter.<br />

is the distance between antenna elements of the linear array at receiver.<br />

is the wave number.<br />

Φ<br />

n,m is the phase of the mth ray within the nth ZDSC.<br />

v<br />

is the speed of the MS.<br />

θ<br />

v<br />

is the angle of the MS velocity vector.<br />

If cross-polarisati<strong>on</strong> is c<strong>on</strong>sidered, additi<strong>on</strong>al cross polarized rays per each ZDSC are generated with<br />

same angle <strong>and</strong> delay informati<strong>on</strong> as those of the co-polarized rays described earlier but different r<strong>and</strong>om<br />

xy ,<br />

phases ( Φ ) from uniform distributi<strong>on</strong> U(0°,360°). The complex field pattern at transmitter <strong>and</strong><br />

mn ,<br />

receiver for vertical polarisati<strong>on</strong><br />

receiver<br />

h<br />

F<br />

t<br />

,<br />

h<br />

r<br />

v<br />

F<br />

t<br />

,<br />

v<br />

F<br />

r<br />

respectively, <strong>and</strong> for horiz<strong>on</strong>tal polarisati<strong>on</strong> for transmitter <strong>and</strong><br />

vh<br />

F , respectively. The cross-polarized amplitude from vertical to horiz<strong>on</strong>tal ( κ<br />

mn , ) or<br />

hv<br />

horiz<strong>on</strong>tal to vertical κ<br />

mn , for each ray within each ZDSC are calculated from their corresp<strong>on</strong>ding XPR<br />

values given by lognormal distributi<strong>on</strong> of parameters given in Table 3.6 for different scenarios. The κ is<br />

selected for every ray with each cluster from indpenent lognormal distributi<strong>on</strong> with parameters given in<br />

Table 3.3. It is assumed that the XPR r<strong>and</strong>om variables of different rays are independent from angles <strong>and</strong><br />

delays. The <strong>channel</strong> coefficient with polarizati<strong>on</strong> can be calculated as given in [3GPP SCM] as<br />

h () t = Pσ<br />

usn , ,<br />

n SF<br />

M<br />

∑<br />

m = 1<br />

( φ )<br />

( φ )<br />

( ϕ )<br />

( ϕ )<br />

T<br />

⎛ v<br />

v<br />

⎡<br />

vv vh vh<br />

F ⎤<br />

t m, n ⎡exp( j<br />

mn ,<br />

) κ<br />

m, n<br />

exp( j<br />

,<br />

) ⎤ ⎡F<br />

⎤ ⎞<br />

⎜<br />

Φ<br />

Φ<br />

mn r m,<br />

n<br />

⎢ ⎥ ⎢<br />

⎥ ⎢ ⎥⋅⎟<br />

⎜ h<br />

hv hv hh<br />

h<br />

⎢Ft mn ,<br />

⎥ ⎢κ<br />

mn ,<br />

exp( jΦ m, n) exp( jΦ mn ,<br />

) ⎥ ⎢Fr mn ,<br />

⎥ ⎟<br />

⎜⎣ ⎦ ⎣<br />

⎦ ⎣ ⎦ ⎟<br />

⎜<br />

⎟<br />

⎜<br />

⎟<br />

jkds sin( φmn , ) jkdu sin( ϕm, n ) jk v cos( ϕm,<br />

n−θv<br />

) t<br />

⎜<br />

e e ⋅ e<br />

⎟<br />

⎜<br />

⎟<br />

⎜<br />

⎟<br />

⎝<br />

⎠<br />

(3.28)<br />

For LOS ray (not cluster), the off-diag<strong>on</strong>al elements of the polarizati<strong>on</strong> matrix are zero by definiti<strong>on</strong>, i.e.,<br />

the XPR (given by the two κ) is infinity for the LOS ray.<br />

3.2 Reduced variability “clustered delay line” model<br />

This <strong>channel</strong> model is somehow different from the c<strong>on</strong>venti<strong>on</strong>al tapped delay line <strong>models</strong> in a sense that<br />

fading within each tap is generated by a sum of sinusoids i.e., the rays within the cluster of that tap.<br />

However, it is based <strong>on</strong> similar principles of the ZDSC <strong>channel</strong> modelling approach. Clustered delay line<br />

(CDL) model is composed of a number of separate delayed clusters. Each cluster has a number of<br />

multipath comp<strong>on</strong>ents (rays) that have the same known delay values but differ in known angle of<br />

departure <strong>and</strong> known angle of arrival. The cluster’s angle-spread may be different from that of BS to that<br />

of the MS. The offset angles of the rays depend <strong>on</strong> the angles spread at BS or MS <strong>and</strong> are calculated as<br />

shown in Table 3.10. The offset angles represent the Laplacian PAS of each ZDSC. The average power,<br />

mean AoA, mean AoD of clusters, angle-spread at BS <strong>and</strong> angle-spread at MS of each cluster in the CDL<br />

are extracted or estimated from measurement results at 5 GHz <strong>and</strong> chip frequency (f c ) of 100 MHz for<br />

Scenarios A1, C2 <strong>and</strong> D1, <strong>and</strong> f c =60 MHz for scenario B1 or obtained from literature as in Scenario B5.<br />

In the CDL model each ZDSC is composed of 10 rays with fixed offset angles <strong>and</strong> identical power. In the<br />

case of ZDSC where a ray of dominant power exists, the ZDSC has 10+1 rays. This dominant ray has a<br />

zero angle offset. The departure <strong>and</strong> arrival rays are coupled r<strong>and</strong>omly. The CDL table of all scenarios of<br />

interest are give below, where the ZDSC power <strong>and</strong> the power of each ray are tabulated. The CDL <strong>models</strong><br />

offer well-defined radio <strong>channel</strong>s with fixed parameters to obtain comparable simulati<strong>on</strong> results with<br />

relatively n<strong>on</strong>-complicated <strong>channel</strong> <strong>models</strong>.<br />

Page 29 (167)


WINNER D5.4 v. 1.4<br />

3.2.1 Scenario A1<br />

The number of different taps in the delay line has been selected according to the measurements data. The<br />

number selected is larger than the median value to represent an envir<strong>on</strong>ment that is more dem<strong>and</strong>ing than<br />

average <strong>and</strong> also to provide a reas<strong>on</strong>ably low frequency correlati<strong>on</strong>. The offset angles of each ray within<br />

every ZDSC are calculated as shown in Table 3.10. The CDL parameters of LOS <strong>and</strong> NLOS c<strong>on</strong>diti<strong>on</strong> are<br />

given in Table 3.13 <strong>and</strong> Table 3.14, respectively.<br />

3.2.1.1 LOS<br />

ZDSC<br />

#<br />

delay<br />

[ns]<br />

Table 3.13: Scenario A1: LOS Clustered delay line model, indoor envir<strong>on</strong>ment.<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 1 m/s,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0 0 12.7 -0.23 * -22.9 **<br />

2 5 -1.7 4.25 -61.8 -10.7<br />

3 10 -6.2 0.54 -42.8 -16.2<br />

4 15 -7.7 -4.55 -33.9 -17.7<br />

5 20 -9.3 0.78 -27.2 -19.3<br />

6 25 -12.1 -3.14 16.6 -22.1<br />

7 30 -12.7 3.74 41.2 -22.7<br />

8 35 -12.7 -2.83 -15.6 -22.7<br />

9 45 -14.7 2.01 2.54 -24.7<br />

10 55 -16.3 5.83 -89.1 -26.3<br />

11 65 -16.8 -10.9 40.9 -26.8<br />

-∞<br />

Number of rays /ZDSC = 10 +<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 5<br />

ZDSC AS at BS [º] = 5<br />

Composite AS at MS [º] = 32.5<br />

Composite AS at BS [º] = 5.1<br />

12 75 -18.4 -13.4 124.0<br />

-28.4<br />

*<br />

**<br />

+<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

3.2.1.2 NLOS<br />

ZDSC #<br />

Table 3.14: Scenario A1: NLOS Clustered delay line model, indoor envir<strong>on</strong>ment.<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 1 m/s,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0 0 0<br />

2 5 -0.9 17.1 19.4 -10.9<br />

3 10 -1.5 -2.09 33.2 -11.5<br />

4 15 -1.6 4.99 15.2 -11.6<br />

5 20 -2.0 -11.4 -20.7 -12.0<br />

6 25 -2.6 -22.9 71.2 -12.6<br />

7 30 -3.4 43.4 48.3 -13.4<br />

8 35 -4.5 -32.2 92.6 -14.5<br />

9 40 -5.5 -22.0 49.0 -15.5<br />

10 45 -5.5 52.4 43.4 -15.5<br />

11 50 -5.0 1.57 -66.1 -15.0<br />

-∞<br />

Number of rays /ZDSC = 10<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 5<br />

ZDSC AS at BS [º] = 5<br />

Composite AS at MS [º] = 39.1<br />

Composite AS at BS [º] = 23.2<br />

12 55 -4.7 37.8 -30.8<br />

-14.7<br />

Page 30 (167)


WINNER D5.4 v. 1.4<br />

13 65 -5.4 8.39 -59.0 -15.4<br />

14 75 -9.0 -27.5 14.1 -19.0<br />

15 85 -11.3 -43.9 -7.76 -21.3<br />

16 95 -12.5 13.7 -0.59 -22.5<br />

17 105 -13.6 63.8 -13.4 -23.6<br />

18 115 -15.1 2.31 3.29 -25.1<br />

19 125 -16.8 8.77 4.22 -26.8<br />

20 135 -18.7 31.1 15.9 -28.7<br />

3.2.2 Scenario B1<br />

The parameters of the CDL model have been extracted from measurements with chip frequency of 60<br />

MHz at frequency range of 5.3 GHz. The number of ZDSCs is selected to be close to the median <strong>and</strong><br />

provide low frequency correlati<strong>on</strong>. The CDL parameters of LOS <strong>and</strong> NLOS c<strong>on</strong>diti<strong>on</strong> are given in Table<br />

3.15 <strong>and</strong> Table 3.16, respectively.<br />

3.2.2.1 LOS<br />

Table 3.15: Scenario B1: LOS Clustered delay line model.<br />

ZDSC #<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 50 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0 0 16 -0.11 * -26.11 **<br />

2 10 -1.2 -22 -10 9 -1.72 -20.7<br />

3 25 -7.4 -12 20 3 -9.16 -22.16<br />

4 30 -7.4 -12 20 -17.4<br />

5 45 -8.4 -2 -123 -18.4<br />

6 65 -13.0 10 -31 -23<br />

7 85 -15.1 -4 161 -25.1<br />

8 105 -16.1 8 -7<br />

-26.1<br />

-∞<br />

Number of rays<br />

/ZDSC =10 +<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = =18<br />

ZDSC AS at BS [º] =3<br />

Composite AS at MS [º]<br />

=37.1<br />

Composite AS at BS [º] =<br />

5.6<br />

*<br />

**<br />

+<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

3.2.2.2 NLOS<br />

ZDSC<br />

#<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

Table 3.16: Scenario B1: NLOS Clustered delay line model.<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 50 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 -1.25 4 0 9 -1.8 * -20.8 **<br />

2 10 0 40 25 6 -1 * -17 *<br />

3 40 -0.38 -10. 29 -10.38<br />

4 60 -0.10 48. -31 -10.10<br />

5 85 -0.73 -36. 37 -10.73<br />

6 110 -0.63 -40 21 -10.63<br />

7 135 -1.78 -26 13<br />

-∞<br />

Number of rays /ZDSC<br />

= 10 +<br />

Ray Power [dB]<br />

-11.78<br />

ZDSC AS at MS [º]<br />

=22<br />

ZDSC AS at BS [º]<br />

=10<br />

Composite AS at MS<br />

[º] =36.4<br />

Composite AS at BS<br />

[º] =12.4<br />

Page 31 (167)


WINNER D5.4 v. 1.4<br />

8 165 -4.07 -28 117 -14.07<br />

9 190 -5.12 -12 21 -15.12<br />

10 220 -6.34 -14 1 -16.34<br />

11 245 -7.35 14 15 -17.35<br />

12 270 -8.86 8 9 -18.86<br />

13 300 -10.1 -24 19 -20.1<br />

14 325 -10.5 -14 1 -20.5<br />

15 350 -11.3 -22 -13 -21.3<br />

16 375 -12.6 2 11 -22.6<br />

17 405 -13.9 8 -1 -23.9<br />

18 430 -14.1 -2 43 -24.1<br />

19 460 -15.3 -10 33 -25.3<br />

20 485 -16.3 -54 -19 -26.3<br />

*<br />

**<br />

+<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

3.2.3 Scenario B3<br />

3.2.3.1 LOS<br />

ZDSC<br />

#<br />

Delay<br />

[ns]<br />

Power<br />

[dB]<br />

Table 3.17: Scenario B3: LOS Clustered delay line model.<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 1.5 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0 -0.9 26 -0.01 * -36 **<br />

2 5 -0.9 -4.0 1.3 7 -1.69 -18.69<br />

3 10 -1.8 -2.4 3.3 2 -3.92 -15.92<br />

4 15 -2.7 -0.9 1.9 -12.7<br />

5 20 -3.6 -2.5 10.5 -13.6<br />

6 25 -4.5 -0.4 14.9 -14.5<br />

7 30 -5.4 -15.1 4.6 -15.4<br />

8 40 -7.2 -3.9 3.3 -17.2<br />

9 50 -9.0 -5.1 5.8 -19.0<br />

10 60 -10.8 -2.0 1.4 -20.8<br />

11 70 -12.6 -23.6 7.6 -22.6<br />

12 80 -14.4 -16.3 4.5 -24.4<br />

13 90 -16.2 -8.2 6.2 -26.2<br />

- ∞<br />

Number of rays/ZDSC = 10<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 5.4<br />

ZDSC AS at BS [º] = 4.7<br />

Composite AS at MS [º] = 18.1<br />

Composite AS at BS [º] = 3.7<br />

14 100 -18.0 50.9 -12.9 -28.0<br />

15 110 -19.8 -4.9 0.5 -29.8<br />

16 120 -21.6 -41.2 20.9<br />

-31.6<br />

*<br />

**<br />

+<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

Page 32 (167)


WINNER D5.4 v. 1.4<br />

3.2.3.2 NLOS<br />

ZDSC<br />

#<br />

delay<br />

[ns]<br />

Table 3.18: Scenario B3: NLOS Clustered delay line model.<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

MS speed = 1.5 km/h.<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 -19,3 -1,3 -10<br />

2 5 -0.5 -14,3 0,8 -10.5<br />

3 10 -1.08 -12,5 1 -11.08<br />

4 15 -1.63 -2,9 -1,7 -11.63<br />

5 20 -2.17 -34,4 9,3 -12.17<br />

6 25 -2.76 -12,1 9,1 -12.76<br />

7 30 -3.26 -20,8 9,7 -13.26<br />

8 40 -4.35 -6,8 2,2 -14.35<br />

9 50 -5.43 -5,6 1,2 -15.43<br />

10 60 -6.52 1,0 0,9 -16.52<br />

11 70 -7.61 -19,1 6,6 -17.61<br />

12 80 -8.69 -24,9 4,8 -18.69<br />

13 90 -9.78 -14,3 2,2 -19.78<br />

14 100 -10.87 48,0 -16,6 -20.87<br />

15 110 -11.95 24,9 -3,3 -21.95<br />

16 120 -13.04 -23,3 19,7 -23.04<br />

17 140 -15.21 -37,2 36,7 -25.21<br />

18 160 -17.38 39,2 -3,9 -27.38<br />

19 180 -19.56 29,2 -0,9 -29.56<br />

20 200 -21.73 25,2 -5,1 -31.73<br />

21 210 -22.82 -3,5 -18,9 -32.82<br />

22 220 -23.91 -25,6 -8,3<br />

-33.91<br />

K-factor = -∞<br />

Number of rays/ZDSC = 10<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 12.5<br />

ZDSC AS at BS [º] = 5.5<br />

Composite AS at MS [º] = 18.7<br />

Composite AS at BS [º] = 3<br />

3.2.4 Scenario B5<br />

For the stati<strong>on</strong>ary feeder scenarios <strong>on</strong>ly CDL <strong>models</strong> have been created. The CDL <strong>models</strong> are based <strong>on</strong><br />

the parameters in the tables below which are derived from literature. Note that the CDL <strong>models</strong> <strong>on</strong>ly<br />

approximate the selected parameters. The motivati<strong>on</strong>s for each parameter can be found in Secti<strong>on</strong> xxx.<br />

Basically any antenna pattern can be used with the <strong>models</strong> However, for the B5 scenario at distances<br />

larger than 300 meters the 3 dB beamwidth γ of <strong>on</strong>e of the <strong>link</strong> ends should be smaller than 10<br />

3dB<br />

degrees while the other is smaller than 53 degrees. An example antenna patterns that can be used is:<br />

A<br />

( γ )<br />

⎡ ⎛ γ ⎞<br />

=−min⎢12<br />

⎜ ⎟<br />

⎢ ⎝γ<br />

3dB<br />

⎣ ⎠<br />

2<br />

, Am<br />

⎤<br />

o<br />

o<br />

⎥, where 180 < φ


WINNER D5.4 v. 1.4<br />

Table 3.19: Parameters selected for scenario B5a LOS stati<strong>on</strong>ary feeder: rooftop to rooftop.<br />

Parameter<br />

Value<br />

Path-loss (dB) Loss = .5 + 20log10( fc<br />

/ 2.5GHz) + 23.5log10( d)<br />

+ δ<br />

slow<br />

Shadow-fading<br />

Power-delay profile<br />

Delay-spread<br />

K-factor<br />

XPR<br />

Doppler<br />

Angle-spread of n<strong>on</strong>-direct comp<strong>on</strong>ents.<br />

36 ,<br />

30m< d 110dB), is given in Table 3.22, Table 3.23, <strong>and</strong> Table 3.24, respectively.<br />

Table 3.21: Parameters selected for scenario B5b LOS stati<strong>on</strong>ary feeder: street-<strong>level</strong> to street-<strong>level</strong>.<br />

Parameter<br />

Value<br />

Path-loss (dB) ( h − h )( h − h )<br />

r<br />

b<br />

= 4<br />

Loss<br />

Loss<br />

b<br />

0<br />

λ<br />

b<br />

0<br />

( r) = −20log( λ /( 4πr)<br />

) + σ free + δ free,<br />

r ≤ rb<br />

( r) = σ free − 20log10( λ /( 4πrb<br />

)) + 40log( r / rb ) + δ bey<strong>on</strong>d,<br />

r > rb<br />

Page 34 (167)


WINNER D5.4 v. 1.4<br />

Shadow-fading s free=3dB, r ≤ rb<br />

,<br />

Range definiti<strong>on</strong><br />

Power-delay profile<br />

Delay-spread<br />

σ bey<strong>on</strong>d =7dB, r > rb<br />

Range 1: Loss


WINNER D5.4 v. 1.4<br />

ZDSC<br />

#<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º] Freq. of<br />

<strong>on</strong>e<br />

scatterer<br />

mHz<br />

K-<br />

factor<br />

[dB]<br />

MS speed N/A<br />

1 0 -1.5 0.0 0.0 744 13.0 -1.8 * -24.7 **<br />

2 5 -10.2 -71.7 70.0 -5 -20.2<br />

3 30 -16.6 167.4 -27.5 -2872 -26.6<br />

4 45 -19.2 -143.2 106.4 434 -29.2<br />

5 75 -20.9 34.6 94.8 294 -30.9<br />

6 90 -20.6 -11.2 -94.0 118 -30.6<br />

7 105 -16.6 78.2 48.6 2576 -26.6<br />

8 140 -16.6 129.2 -96.6 400 -26.6<br />

9 210 -23.9 -113.2 41.7 71 -33.9<br />

10 210 -12.0 -13.5 -83.3 3069 -22.0<br />

11 250 -23.9 145.2 176.8 1153 -33.9<br />

12 270 -21.0 -172.0 93.7 -772 -31.0<br />

13 275 -17.7 93.7 -6.4 1298 -27.7<br />

14 475 -24.6 106.5 160.3 -343 -34.6<br />

15 595 -22.0 -67.0 -50.1 -7 -32.0<br />

-∞<br />

Number of rays/ZDSC = 10 +<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 2<br />

ZDSC AS at BS [º] = 2<br />

Composite AS at MS [º] =42.8<br />

Composite AS at BS [º] = 50.2<br />

16 690 -29.2 -95.1 -149.6 -186 -39.2<br />

17 855 -32.9 -2.0 161.5 -2288 -42.9<br />

18 880 -32.9 66.7 68.7 26 -42.9<br />

19 935 -28.0 160.1 41.6 -1342 -38.0<br />

20 1245 -29.6 -21.8 142.2 -61<br />

-39.6<br />

*<br />

**<br />

+<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

ZDSC<br />

#<br />

delay<br />

[ns]<br />

Table 3.24: Clustered delay-line model street-<strong>level</strong> to street-<strong>level</strong> range 3.<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º] Freq. of<br />

<strong>on</strong>e<br />

scatterer<br />

mHz<br />

K-<br />

factor<br />

[dB]<br />

MS speed N/A<br />

1 0 -2.6 0.0 0.0 744 10.0 -3.0 * -23.0 **<br />

2 10 -8.5 -71.7 70.0 -5 -18.5<br />

3 90 -14.8 167.4 -27.5 -2872 -24.8<br />

4 135 -17.5 -143.2 106.4 434 -27.5<br />

5 230 -19.2 34.6 94.8 295 -29.2<br />

6 275 -18.8 -11.2 -94.0 118 -28.8<br />

7 310 -14.9 78.2 48.6 2576 -24.9<br />

8 420 -14.9 129.2 -96.6 400 -24.9<br />

9 630 -22.1 -113.2 41.7 71 -32.1<br />

10 635 -10.3 -13.5 -83.3 3069 -20.3<br />

-∞<br />

Number of rays/ZDSC = 10 +<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 2<br />

ZDSC AS at BS [º] = 2<br />

Composite AS at MS [º] =52.3<br />

Composite AS at BS [º] = 61.42<br />

11 745 -22.2 145.2 176.8 1153<br />

-32.2<br />

Page 36 (167)


WINNER D5.4 v. 1.4<br />

12 815 -19.2 -172.0 93.7 -772 -29.2<br />

13 830 -16.0 93.7 -6.4 1298 -26.0<br />

14 1430 -22.9 106.5 160.3 -343 -32.9<br />

15 1790 -20.3 -67.0 -50.1 -7 -30.3<br />

16 2075 -27.4 -95.1 -149.6 -186 -37.4<br />

17 2570 -31.1 -2.0 161.5 -2287 -41.1<br />

18 2635 -31.2 66.7 68.7 26 -41.2<br />

19 2800 -26.3 160.1 41.6 -1342 -36.3<br />

20 3740 -27.8 -21.8 142.2 -61 -37.8<br />

*<br />

**<br />

+<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

3.2.5 Scenario C1<br />

3.2.5.1 LOS<br />

ZDSC<br />

#<br />

Table 3.25: Scenario C1: LOS Clustered delay line model, suburban envir<strong>on</strong>ment.<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-factor<br />

[dB]<br />

MS speed = 50 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0 0 10.0 -0.41* -20.4**<br />

2 5 -3.4 -15.9 -67.9 -13.4<br />

3 10 -2.7 9.34 -84.0 -12.7<br />

4 15 -2.6 -29.4 -51.2 -12.6<br />

5 20 -4.8 -6.32 -91.2 -14.8<br />

6 25 -7.5 -20.4 -94.5 -17.5<br />

7 30 -9.4 1.24 -22.8 -19.4<br />

8 35 -9.7 10.3 -17.2 -19.7<br />

9 40 -8.8 11.3 87.4 -18.8<br />

10 45 -8.9 5.12 -73.0 -18.9<br />

-∞<br />

11 50 -9.4 14.1 -120 -19.4<br />

12 70 -13.1 -18.9 -71.8 -23.1<br />

13 90 -14.2 2.84 2.87 -24.2<br />

14 110 -17.4 16.2 -128 -27.4<br />

15 130 -17.3 -14.2 20.5 -27.3<br />

16 150 -18.1 6.30 -48.2 -28.1<br />

17 170 -17.0 -4.64 52.0 -27.0<br />

18 190 -16.1 4.30 -43.5 -26.1<br />

19 210 -19.4 -0.79 11.9<br />

-29.4<br />

*<br />

**<br />

+<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

Number of rays /ZDSC = 10 +<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 5<br />

ZDSC AS at BS [º] = 5<br />

Composite AS at MS [º] = 45.8<br />

Composite AS at BS [º] = 14.2<br />

Page 37 (167)


WINNER D5.4 v. 1.4<br />

3.2.5.2 NLOS<br />

Clustered delay line model has an RMS delay spread of 62 ns, <strong>and</strong> composite angle-spread of 53 <strong>and</strong> 5<br />

degrees in MS <strong>and</strong> BS, respectively. No K-factor is introduced for NLOS.<br />

Table 3.26: Clustered delay-line model for Scenario C1 NLOS<br />

ZDSC #<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 50 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0 0 -10<br />

2 5 -0.6 4 35 -10.6<br />

3 15 -1.8 -2 60 -11.8<br />

4 25 -2.3 -6 -39 -12.3<br />

5 60 -7.8 1 -56 -17.8<br />

6 80 -14.0 -5 165 -24.0<br />

7 105 -12.9 -8 -69 -22.9<br />

8 120 -9.8 -10 -109 -29.8<br />

9 205 -19.5 12 75 -29.5<br />

10 240 -17.4 22 120 -27.4<br />

11 255 -15.1 -25 138 -25.1<br />

-∞<br />

Number of rays /ZDSC = 10<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 10<br />

ZDSC AS at BS [º] = 2<br />

Composite AS at MS [º] = 53<br />

Composite AS at BS [º] = 5<br />

12 350 -18.3 10 -177 -28.3<br />

13 380 -13.9 4 150 -23.9<br />

14 410 -19.9 -1 179<br />

-29.9<br />

3.2.6 Scenario C2<br />

Clustered delay line model has an RMS delay spread of 310 ns, <strong>and</strong> composite angle-spread of 53 <strong>and</strong> 8<br />

degrees in MS <strong>and</strong> BS, respectively. No K-factor is introduced. The parameters are given in Table 3.27.<br />

Table 3.27: Scenario C2: NLOS Clustered delay line model.<br />

ZDSC #<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 50 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 -0.5 0 0 -10.5<br />

2 5 0.0 4 4 -10.0<br />

3 135 -3.4 -3 7 -13.4<br />

4 160 -2.8 -4 10 -12.8<br />

5 215 -4.6 -7 21 -14.6<br />

6 260 -0.9 8 -45 -10.9<br />

7 385 -6.7 10 -75 -16.7<br />

8 400 -4.5 17 65 -14.5<br />

9 530 -9.0 -8 160 -19.0<br />

10 540 -7.8 -8 155 -17.8<br />

11 650 -7.4 -4 88 -17.4<br />

-∞<br />

Number of rays /ZDSC = 10<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 15<br />

ZDSC AS at BS [º] = 2<br />

Composite AS at MS [º] = 53<br />

Composite AS at BS [º] = 8<br />

12 670 -8.4 -7 80 -18.4<br />

13 720 -11.0 -9 -90 -21.0<br />

14 750 -9.0 -9 -105<br />

-19.0<br />

Page 38 (167)


WINNER D5.4 v. 1.4<br />

15 800 -5.1 12 8 -15.1<br />

16 945 -6.7 -17 45 -16.7<br />

17 1035 -12.1 19 50 -22.1<br />

18 1185 -13.2 12 -15 -23.2<br />

19 1390 -13.7 19 -25 -23.7<br />

20 1470 -19.8 21 100 -29.8<br />

3.2.7 Scenario D1<br />

3.2.7.1 LOS<br />

Table 3.28: Scenario D1: LOS Clustered delay line model, rural envir<strong>on</strong>ment.<br />

ZDSC #<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 120 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0.0 0.0 10.9 -0.34 * -21.2 **<br />

2 5 -3.4 45.7 -25.5 -13.4<br />

3 10 -11.4 -12.7 35.6 -21.4<br />

4 15 -16.4 20.7 54.0 -26.4<br />

5 25 -17.8 -9.6 25.0 -27.8<br />

6 35 -17.9 -24.8 136.9 -27.9<br />

7 45 -18.8 -9.8 -7.6 -28.8<br />

8 55 -19.3 -9.6 21.5 -29.3<br />

9 65 -19.5 -21.3 -96.5 -29.5<br />

10 75 -18.5 8.2 -26.5 -28.5<br />

11 85 -19.0 21.9 -92.7 -29.0<br />

-∞<br />

Number of rays /ZDSC = 10 +<br />

Ray Power [dB]<br />

ZDSC AS at MS [º] = 1.5<br />

ZDSC AS at BS [º] = 1.5<br />

Composite AS at MS [º] = 24<br />

Composite AS at BS [º] = 21.5<br />

12 95 -19.6 23.2 -5.0 -29.6<br />

13 160 -18.7 28.7 -64.5<br />

-28.7<br />

* Power of dominant ray,<br />

** Power of each other ray,<br />

+<br />

Clusters with high K-factor will have 11 rays.<br />

3.2.7.2 NLOS<br />

Table 3.29: Scenario D1: NLOS Clustered delay line model, rural envir<strong>on</strong>ment.<br />

ZDSC<br />

#<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 120 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 0 0 -10.0<br />

2 5 -1.0 -14.3 -27.1 -11.0<br />

3 10 -5.8 20.1 27.6 -15.8<br />

4 15 -8.2 21.6 -14.3 -18.2<br />

5 20 -9.0 14.5 15.9 -19.0<br />

6 25 -9.5 -13.9 -28.5 -19.5<br />

7 30 -10.1 7.06 23.7<br />

-∞<br />

Number of rays /ZDSC<br />

= 10 +<br />

Ray Power [dB]<br />

-20.1<br />

ZDSC AS at MS [º] =<br />

3<br />

ZDSC AS at BS [º] =<br />

1.5<br />

Composite AS at MS<br />

[º] = 17.9<br />

Composite AS at BS<br />

[º] = 22.4<br />

Page 39 (167)


WINNER D5.4 v. 1.4<br />

8 35 -10.6 -66.7 -50.4 -20.6<br />

9 40 -11.1 9.92 50.5 -21.1<br />

10 45 -11.6 -21.3 32.0 -21.6<br />

11 50 -12.0 -34.9 15.7 -22.0<br />

12 65 -13.1 -4.88 12.7 -23.1<br />

13 80 -13.8 19.1 -7.40 -23.8<br />

14 95 -15.3 11.6 -4.82 -25.3<br />

15 110 -16.4 9.8 0.16 -26.4<br />

16 125 -16.8 -13.3 31.6 -26.8<br />

17 140 -17.9 -14.2 3.62 -27.9<br />

18 155 -18.6 71.1 14.6 -28.6<br />

19 170 -18.6 -20.2 27.4 -28.6<br />

Page 40 (167)


WINNER D5.4 v. 1.4<br />

PART II<br />

This sec<strong>on</strong>d part c<strong>on</strong>tains more detailed informati<strong>on</strong> about our modelling<br />

approach, our measurements <strong>and</strong> literature analysis, <strong>and</strong> the <strong>channel</strong> model<br />

implementati<strong>on</strong>.<br />

Page 41 (167)


WINNER D5.4 v. 1.4<br />

4. Modelling Approaches<br />

In this secti<strong>on</strong>, we discuss the modelling <strong>and</strong> coefficient generati<strong>on</strong> approach of existing spatial <strong>channel</strong><br />

<strong>models</strong>. Then our selected approach is presented in detail.<br />

Apart from the generic, fully r<strong>and</strong>om <strong>channel</strong> model, we define a clustered delay-line model derived from<br />

our generic model by limiting the r<strong>and</strong>omness (fixing the value) of certain parameters. The reduced<br />

variability aids the comparability of results based <strong>on</strong> shorter simulati<strong>on</strong> times.<br />

4.1 Generic <strong>channel</strong> modelling approach<br />

The generic <strong>channel</strong> modelling approach has been followed earlier in COST259 <strong>and</strong> in 3GPP<br />

st<strong>and</strong>ardizati<strong>on</strong>. In COST259, the approach has been followed for directi<strong>on</strong>al antenna <strong>channel</strong> <strong>models</strong> for<br />

smart antennas wireless applicati<strong>on</strong>s. The COST259 was mainly for antenna array applicati<strong>on</strong> at <strong>on</strong>e end,<br />

usually the base stati<strong>on</strong> side. The 3GPP st<strong>and</strong>ardizati<strong>on</strong> <strong>channel</strong> model, known as the 3GPP/3GPP2<br />

spatial <strong>channel</strong> model (SCM), was developed for MIMO approaches in third generati<strong>on</strong> cellular <strong>system</strong>s.<br />

A generic <strong>channel</strong> modelling approach can be thought as a <strong>channel</strong> model framework that can be applied<br />

in different scenarios. Each scenario has scenario specific distributi<strong>on</strong>s <strong>and</strong> parameters. By changing the<br />

scenario specific distributi<strong>on</strong>s in angle <strong>and</strong> delay domains as well as the scenario specific parameters, we<br />

can have different <strong>channel</strong> <strong>models</strong> for different scenarios under the same framework of the <strong>channel</strong><br />

model.<br />

4.1.1 Distincti<strong>on</strong> between <strong>channel</strong> <strong>models</strong> for <strong>link</strong>-<strong>level</strong> <strong>and</strong> <strong>system</strong>-<strong>level</strong> simulati<strong>on</strong><br />

Workpackage 5, as defined in the Annex, is divided into a <strong>link</strong>-<strong>level</strong> <strong>and</strong> a <strong>system</strong>-<strong>level</strong> modelling effort<br />

with task 4 representing the former <strong>and</strong> task 5 the latter part. During the evoluti<strong>on</strong> of our work though, we<br />

found that we had to be very careful with such a divisi<strong>on</strong> because it turned out not to be inherently clear<br />

where to draw the line. To counter this problem, we c<strong>on</strong>sequently defined a set of properties for each of<br />

the two <strong>level</strong>s in our deliverable D5.2. It has turned out most practical to implement both the <strong>link</strong>-<strong>level</strong><br />

<strong>and</strong> <strong>system</strong>-<strong>level</strong> features in <strong>on</strong>e model. Here we underst<strong>and</strong> the <strong>system</strong>-<strong>level</strong> <strong>channel</strong> modelling as in the<br />

SCM model [3GPP SCM]. Then it is possible to emphasize either the <strong>system</strong>-<strong>level</strong> features or the <strong>link</strong><strong>level</strong><br />

features or both by selecting the parameters properly.<br />

Our c<strong>on</strong>clusi<strong>on</strong> is that it can be potentially dangerous to define a certain divisi<strong>on</strong> <strong>and</strong> separate the two<br />

<strong>channel</strong> <strong>models</strong>. C<strong>on</strong>sider for example a model where shadowing is c<strong>on</strong>sidered a higher <strong>level</strong> than delayor<br />

angle-spread <strong>and</strong> for this reas<strong>on</strong> treated independently. As a c<strong>on</strong>sequence, a likely c<strong>on</strong>clusi<strong>on</strong> drawn<br />

from low-<strong>level</strong> simulati<strong>on</strong>s is that angle- <strong>and</strong> delay-spread significantly improves capacity. However, if<br />

all three parameters were simulated corporately, including their cross-correlati<strong>on</strong>s, the soluti<strong>on</strong> might be<br />

completely opposite, specifically that the capacity loss from shadowing outweighs the gain from delay<strong>and</strong><br />

angle-spread.<br />

In summary, we favour <strong>channel</strong> <strong>models</strong> that c<strong>on</strong>tain both the <strong>link</strong>-<strong>level</strong> <strong>and</strong> the <strong>system</strong>-<strong>level</strong> features<br />

defined at the same time. Hence, it depends <strong>on</strong> the applicati<strong>on</strong>, which feature is switched <strong>on</strong> or off.<br />

4.1.2 Comparis<strong>on</strong> between deterministic <strong>and</strong> stochastic <strong>channel</strong> modeling<br />

Channel modeling can be broadly split into two areas that differ in the goal or applicati<strong>on</strong> <strong>and</strong> the type of<br />

underlying data.<br />

Deterministic <strong>channel</strong> modeling can be employed when detailed envir<strong>on</strong>ment data is available. Detailed<br />

envir<strong>on</strong>ment data means positi<strong>on</strong>, size <strong>and</strong> orientati<strong>on</strong> of man-made objects (houses, buildings, bridges,<br />

roads, etc.) as well as natural objects (foliage or dominant plants, rocks, ground properties, etc.). The<br />

basic idea is that if the propagati<strong>on</strong> envir<strong>on</strong>ment is known to a sufficient degree, wireless propagati<strong>on</strong> is a<br />

deterministic process that allows determining or predicting its characteristics at every point in space. It is<br />

also referred to as propagati<strong>on</strong> predicti<strong>on</strong> <strong>and</strong> is the type of modeling used for cell planning, i.e., the<br />

analysis of optimum locati<strong>on</strong>s for BS deployment <strong>and</strong> the predicti<strong>on</strong> of the resulting coverage, capacity,<br />

<strong>and</strong> data rates. In deterministic <strong>channel</strong> modeling, <strong>channel</strong> measurements are made in the same<br />

envir<strong>on</strong>ment for which detailed data is available <strong>and</strong> then used to optimize the match between predicti<strong>on</strong><br />

model <strong>and</strong> measurements.<br />

Stochastic <strong>channel</strong> modeling <strong>on</strong> the other h<strong>and</strong> is based <strong>on</strong> a stochastic view of the wireless <strong>channel</strong>.<br />

Measurements are made in a large variety of locati<strong>on</strong>s <strong>and</strong> envir<strong>on</strong>ments to obtain a data set with a good<br />

representati<strong>on</strong> of the underlying statistical properties. Influence parameters based <strong>on</strong> the envir<strong>on</strong>ment<br />

characteristics may be used to refine the statistical accuracy for similar envir<strong>on</strong>ments. As such,<br />

classificati<strong>on</strong> is an important tool to trade off accuracy versus universality of statements.<br />

What we aim for in WINNER is the predicti<strong>on</strong> of statistical behavior of the <strong>channel</strong>. Knowledge of<br />

statistical <strong>channel</strong> parameters allows making more general statements. Especially, they allow evaluating<br />

Page 42 (167)


h<br />

WINNER D5.4 v. 1.4<br />

the properties <strong>and</strong> usefulness of communicati<strong>on</strong> schemes in case of large-scale deployment. Hence, we<br />

follow the stochastic <strong>channel</strong> modeling approach in our analysis.<br />

4.1.3 Interference modeling<br />

Interference modelling is an applicati<strong>on</strong> subset of <strong>channel</strong> <strong>models</strong> that deserves additi<strong>on</strong>al c<strong>on</strong>siderati<strong>on</strong>.<br />

Basically, communicati<strong>on</strong> <strong>link</strong>s that c<strong>on</strong>tain interfering signals are to be treated just as any other <strong>link</strong>.<br />

However, in many of today’s communicati<strong>on</strong> <strong>system</strong>s these interfering signals are not treated <strong>and</strong><br />

processed in the same way as the desired signals <strong>and</strong> thus modelling the interfering <strong>link</strong>s with full<br />

accuracy is inefficient.<br />

A simplificati<strong>on</strong> of the <strong>channel</strong> modelling for the interference <strong>link</strong> is often possible but closely <strong>link</strong>ed<br />

with the communicati<strong>on</strong> architecture. This makes it difficult for a generalized treatment in the c<strong>on</strong>text of<br />

<strong>channel</strong> modelling. In the following we will thus c<strong>on</strong>strain ourselves to giving some possible ideas of how<br />

this can be realised. Note that these are all combined signal <strong>and</strong> <strong>channel</strong> <strong>models</strong>. The actual<br />

implementati<strong>on</strong> will have to be based <strong>on</strong> the computati<strong>on</strong>al gain from computati<strong>on</strong>al simplificati<strong>on</strong> versus<br />

the additi<strong>on</strong>al programming overhead.<br />

AWGN interference<br />

The simplest form of interference is modelled by additive white Gaussian noise. This is sufficient for<br />

basic C/I (carrier to interference ratio) evaluati<strong>on</strong>s when coupled with a path loss <strong>and</strong> shadowing model. It<br />

might be extended with e.g. <strong>on</strong>-off keying (to simulate the n<strong>on</strong>-stati<strong>on</strong>ary behaviour of actual transmit<br />

signals) or other techniques that are simple to implement.<br />

Filtered noise<br />

The possible wideb<strong>and</strong> behaviour of an interfering signal is not reflected in the AWGN model above. An<br />

implementati<strong>on</strong> using a complex SCM or WIM <strong>channel</strong>, however, might be unnecessarily complex as<br />

well because the high number of degrees of freedom does not become visible in the noise-like signal<br />

anyway. Thus we propose something al<strong>on</strong>g the lines of a simple, sample-spaced FIR filter with Rayleighfading<br />

coefficients.<br />

Prerecorded interference<br />

A large part of the time-c<strong>on</strong>suming process of generating the interfering signal is the modulati<strong>on</strong> <strong>and</strong><br />

filtering of the signal, which has to be d<strong>on</strong>e at chip frequency. Even if the interfering signal is detected<br />

<strong>and</strong> removed in the communicati<strong>on</strong> receiver (e.g., multi-user detecti<strong>on</strong> techniques) <strong>and</strong> thus rendering a<br />

PN generator too simple, a method of precomputing <strong>and</strong> replaying the signal might be viable. The<br />

repeating c<strong>on</strong>tent of the signal using this technique is typically not an issue as the c<strong>on</strong>tent of the interferer<br />

is discarded anyway.<br />

4.1.4 Framework<br />

MIMO <strong>channel</strong> characterizati<strong>on</strong>, which takes into account directi<strong>on</strong>al characteristics at the transmitter <strong>and</strong><br />

receiver sides, is widely known as double directi<strong>on</strong>al <strong>channel</strong> modelling. We separate the effective radio<br />

<strong>channel</strong> in effects from wave propagati<strong>on</strong> <strong>on</strong> <strong>on</strong>e h<strong>and</strong> <strong>and</strong> antenna resp<strong>on</strong>se <strong>on</strong> the other h<strong>and</strong> to develop<br />

antenna independent MIMO <strong>channel</strong> model. By using the far-field, narrowb<strong>and</strong>, discrete wave, <strong>and</strong><br />

geometric diffracti<strong>on</strong> assumpti<strong>on</strong>, the effect of the antennas can be reduced to the effect of field pattern<br />

<strong>and</strong> to a phase shift based <strong>on</strong> the angle of the impinging wave, its wavelength, <strong>and</strong> the geometry of the<br />

antennas. This means that any antenna c<strong>on</strong>figurati<strong>on</strong>, orientati<strong>on</strong>, <strong>and</strong> pattern of antenna elements at both<br />

ends can be inserted in the model. In multipath envir<strong>on</strong>ment, each ray can be described by its path delay<br />

(τ), azimuth departure angle (φ), elevati<strong>on</strong> departure angle (θ), azimuth arrival angle (ϕ ), elevati<strong>on</strong><br />

arrival angle (ϑ ) <strong>and</strong> complex amplitude (α ) of the wave <strong>and</strong> polarisati<strong>on</strong> informati<strong>on</strong> matrix. The<br />

framework of the generic <strong>channel</strong> model is for all scenarios where <strong>on</strong>e terminal is mobile while the other<br />

is fixed. It is based <strong>on</strong> principles of existing work presented in [3GPP SCM], [SV87], [Cor01], [GEYC],<br />

[PMF00], [Fle00], [AlPM02], <strong>and</strong> generalized to MIMO case with elevati<strong>on</strong> angles at both ends. The<br />

generic model of MIMO <strong>channel</strong> for n<strong>on</strong>-stati<strong>on</strong>ary envir<strong>on</strong>ment can be described by <strong>channel</strong> impulse<br />

resp<strong>on</strong>se with horiz<strong>on</strong>tal <strong>and</strong> vertical polarisati<strong>on</strong> between antenna element s at transmitter <strong>and</strong> antenna<br />

element u at receiver as:<br />

u,<br />

s<br />

L(<br />

t)<br />

Mn(<br />

t)<br />

( t;<br />

τ,<br />

φ,<br />

θ,<br />

ϕ,<br />

ϕ)<br />

=∑ ∑<br />

e<br />

v<br />

T,<br />

s<br />

( φn,<br />

m,<br />

θn<br />

, m)<br />

( φ , θ )<br />

vv<br />

n,<br />

m<br />

j k( φ ( t) , θ ( t )),<br />

xT<br />

, s j k( ϕ ( t) , ϑ ( t)<br />

),<br />

x<br />

T<br />

vv vh<br />

vh<br />

( jΦn,<br />

m<br />

) κn , m<br />

exp( jΦn , m<br />

)<br />

hv hh<br />

hh<br />

( jΦ<br />

) κ exp( jΦ<br />

)<br />

h<br />

hv<br />

h<br />

n= 1 m=<br />

1 ⎢ T,<br />

s n,<br />

m n,<br />

m ⎥ ⎢ n,<br />

m n,<br />

m n,<br />

m<br />

n,<br />

m ⎥⎢<br />

R,<br />

u n,<br />

m n,<br />

m<br />

n,<br />

m<br />

⎛<br />

⎜⎡F<br />

⎜⎢<br />

F<br />

⎝⎣<br />

n,<br />

m<br />

e<br />

n,<br />

m<br />

⎤ ⎡κ<br />

⎥ ⎢<br />

⎦ ⎣κ<br />

n,<br />

m<br />

R,<br />

u<br />

exp<br />

exp<br />

e<br />

j2πνn,<br />

mt<br />

δ<br />

( τ −τ<br />

) δ( φ −φ<br />

) δ( θ −θ<br />

) δ( ϕ −ϕ<br />

) δ( ϑ −ϑ<br />

)<br />

n<br />

n,<br />

m<br />

⎤⎡F<br />

⎥⎢<br />

⎦⎣F<br />

v<br />

R,<br />

u<br />

( ϕn , m,<br />

ϑn,<br />

m<br />

) ⎤<br />

⎥ •<br />

( ϕ , ϑ ) ⎥⎦<br />

n,<br />

m<br />

n,<br />

m<br />

n,<br />

m<br />

(4.1)<br />

Page 43 (167)


WINNER D5.4 v. 1.4<br />

where t is the time, the full polarimetric (2x2) transfer matrix, κ ( t)<br />

, includes the losses <strong>and</strong><br />

depolarisati<strong>on</strong> of all physical processes (reflecti<strong>on</strong>, diffracti<strong>on</strong>, scattering, etc) of each multipath<br />

r<br />

r<br />

comp<strong>on</strong>ents, xT,<br />

s is the positi<strong>on</strong> of the antenna element s of transmit antenna array, x Ru , is the positi<strong>on</strong><br />

of the antenna element u of the receive antenna array u, <strong>and</strong> ν<br />

l is the Doppler comp<strong>on</strong>ent. Note that all<br />

parameters are in general time variant, which is not shown for simpler presentati<strong>on</strong>.<br />

4.1.4.1 Inter-segment modeling<br />

The radio <strong>channel</strong> is in general not stati<strong>on</strong>ary. Nevertheless, over short periods of time <strong>and</strong> space, <strong>channel</strong><br />

parameters vary very little, <strong>and</strong> the assumpti<strong>on</strong> of short-term stati<strong>on</strong>arity is often a very good<br />

approximati<strong>on</strong>. The parameters characterizing our <strong>channel</strong> model are called bulk parameters. The time<br />

durati<strong>on</strong>s, over which these bulk parameters are c<strong>on</strong>stant, are denoted <strong>channel</strong> segment a.k.a. drops in the<br />

nomenclature of the SCM. Over time <strong>and</strong> space, bulk parameters change <strong>and</strong> we characterize this<br />

variability statistically.<br />

For simulati<strong>on</strong> purposes, the first goal typically is to experience the range of variability of the <strong>channel</strong><br />

rather than the medium-term evoluti<strong>on</strong> behaviour. Thus, the initial focus is mainly <strong>on</strong> the joint<br />

distributi<strong>on</strong>s of bulk parameters. Between <strong>channel</strong> segments, i.e. realizati<strong>on</strong>s of these r<strong>and</strong>om variables,<br />

independence is assumed. The physical interpretati<strong>on</strong> is that <strong>channel</strong> segments are relatively short <strong>channel</strong><br />

observati<strong>on</strong> periods that are significantly separated from each other in time or space.<br />

Each term of the matrix <strong>channel</strong> is a sum of L multipath comp<strong>on</strong>ents that can be described by the time<br />

htτφθϕϑ , , , , , , given as:<br />

varying double-directi<strong>on</strong>al delay spread functi<strong>on</strong> (D 3 SF), ( )<br />

L<br />

ht (, τφθϕϑ , , , , ) = ∑ αn() t δφ ( −φn, θ −θn, ϕ −ϕn, ϑ−ϑn, τ −τn)<br />

. (4.2)<br />

n=<br />

1<br />

The instantaneous power double-directi<strong>on</strong>al-delay spectrum (PD 3 S) can be written as:<br />

L<br />

2<br />

PI(, t τφθϕϑ , , , , ) = ∑ αn() t δφ ( −φn, θ −θn, ϕ −ϕn, ϑ−ϑn, τ −τn)<br />

, (4.3)<br />

n=<br />

1<br />

where ⋅ is the absolute value of the argument. The per <strong>channel</strong> segment (local) average PD 3 S, which<br />

represents <strong>channel</strong> characteristics per <strong>channel</strong> segment, can be defined as:<br />

P<br />

( τφθϕϑ , , , , ) = E { P ( t, τφθϕϑ , , , , )}<br />

s t I<br />

L<br />

∑<br />

= p δφ ( −φ , θ −θ , ϕ −ϕ , ϑ−ϑ ) δτ ( −τ<br />

)<br />

n=<br />

1<br />

n n n n n n<br />

In (4.4), the variables {L, p n , φ n , θ n , ϕ n , ϑ n , τ n } are the bulk parameters introduced above. The L paths are<br />

characterized by the orientati<strong>on</strong> of the last-bounce scatterer as seen from transmitter <strong>and</strong> receiver, as well<br />

as the total delay. This approach stems from superresoluti<strong>on</strong> parameter estimati<strong>on</strong> techniques (e.g.,<br />

MUSIC, ESPRIT, SAGE) which decompose a measured <strong>channel</strong> resp<strong>on</strong>se based <strong>on</strong> the above model.<br />

The D 3 SF characterizes the dispersive behaviour of the mobile radio <strong>channel</strong> in delay domain <strong>and</strong><br />

directi<strong>on</strong> domain seen either at transmitter or receiver sides. The equati<strong>on</strong>s are valid regardless of which<br />

terminal is the transmitter, either BS or MS <strong>and</strong> which terminal is the receiver either MS or BS. All<br />

parameters in (4.4) are r<strong>and</strong>om variables, since scatterers’ locati<strong>on</strong>s change with movement of the MS.<br />

Hence, the D 3 SF is a r<strong>and</strong>om process, which is described by joint distributi<strong>on</strong> of its r<strong>and</strong>om variables. The<br />

statistics of multipath comp<strong>on</strong>ents amplitudes, delays, <strong>and</strong> azimuth <strong>and</strong> elevati<strong>on</strong> angles at both ends are<br />

generally not separable. Hence, they have to be described in joint probability density functi<strong>on</strong>s (pdf).<br />

However, multidimensi<strong>on</strong>al joint pdf is not tractable mathematically. Therefore, simplificati<strong>on</strong>s are<br />

needed for simulati<strong>on</strong> purposes. As a result, <strong>on</strong>ly partial dependencies of distributi<strong>on</strong>s of different<br />

parameters are usually assumed.<br />

One of the most comm<strong>on</strong> assumpti<strong>on</strong>s is uncorrelated scattering (US). We assume independence of all<br />

parameters for different paths, i.e., different n. Therefore, (4.4) is characterized by the joint distributi<strong>on</strong><br />

f( pn, φn, θn, ϕn, ϑn, τ<br />

n)<br />

, which is independent of n.<br />

The expectati<strong>on</strong> in (4.4) is over short periods of time, where <strong>channel</strong> parameters vary <strong>on</strong>ly slightly, <strong>and</strong><br />

the assumpti<strong>on</strong> of short-term stati<strong>on</strong>arity is valid. The over segments (global) average PD 3 S is obtained<br />

by taking the expectati<strong>on</strong> of the per <strong>channel</strong> segment (local) average PD 3 S over all bulk parameters<br />

(4.4)<br />

Page 44 (167)


WINNER D5.4 v. 1.4<br />

{ }<br />

( τφθϕϑ , , , , ) ( τφθϕϑ , , , , )<br />

P = E P s<br />

(4.5)<br />

The average PD 3 S in (4.4) is the average per <strong>channel</strong> segment, which may differ from <strong>on</strong>e segment to<br />

another. In order to relate (4.4) to the average PD 3 S over segments (4.5), all bulk parameters except the<br />

r<strong>and</strong>om powers can be pulled out of the expectati<strong>on</strong> to arrive at:<br />

P ( τφθϕϑ , , , , )<br />

{ (, , , , )| ,..., , ,..., , ,..., , ,..., , ,..., }<br />

= ∫ E P τφθϕϑ φ φ θ θ ϕ ϕ ϑ ϑ τ τ<br />

s 1 L 1 L 1 L 1 L 1 L<br />

L<br />

∏<br />

i=<br />

1<br />

f( φ, θ , ϕ , ϑ, τ ) dφdθdϕ dϑdτ<br />

i i i i i i i i i i<br />

(4.6)<br />

L<br />

= ∑ E{ pi<br />

| τφθϕϑ , , , , } f( τφθϕϑ , , , , )<br />

i=<br />

1<br />

{ τφθϕϑ}<br />

= LE p| , , , , f(, τφθϕϑ , , , ).<br />

From (4.6), the over segments (global) average PD 3 S (i.e., ( , , , , )<br />

P τφθϕϑ ) is equivalent to the<br />

c<strong>on</strong>diti<strong>on</strong>al expected power of the multipath comp<strong>on</strong>ents multiplied by the joint double-directi<strong>on</strong>al-delay<br />

probability density functi<strong>on</strong>.<br />

The power spectrum in each dimensi<strong>on</strong> is obtained by integrati<strong>on</strong> over other dimensi<strong>on</strong>s. Thus, power<br />

delay spectrum P ( τ ) , power azimuth-departure-angle spectrum P ( φ)<br />

, power elevati<strong>on</strong>-departure-angle<br />

spectrum P ( θ ) , power azimuth-arrival-angle spectrum P ( ϕ)<br />

, power elevati<strong>on</strong>-arrival-angle spectrum<br />

P can be derived as:<br />

( ϑ)<br />

( ) ( , , , , )<br />

P τ P τφθϕϑ dφdθdϕdϑ<br />

= ∫∫∫∫<br />

(4.7)<br />

( ) ( , , , , )<br />

P φ P τφθϕϑ dτdθdϕdϑ<br />

= ∫∫∫∫<br />

(4.8)<br />

( ) ( , , , , )<br />

P θ P τφθϕϑ dφdτdϕdϑ<br />

= ∫∫∫∫<br />

(4.9)<br />

( ) ( , , , , )<br />

P ϕ P τφθϕϑ dφdθdτdϑ<br />

= ∫∫∫∫<br />

(4.10)<br />

( ) ( , , , , )<br />

P ϑ P τφθϕϑ dφdθdϕdτ<br />

= ∫∫∫∫<br />

. (4.11)<br />

The corresp<strong>on</strong>ding marginal probability density functi<strong>on</strong>s (pdf) of parameters of each domain can be<br />

derived from:<br />

( ) ( , , , , )<br />

f τ f τφθϕϑ dφdθdϕdϑ<br />

= ∫∫∫∫<br />

(4.12)<br />

( ) ( , , , , )<br />

f φ f τφθϕϑ dd τ θdϕdϑ<br />

= ∫∫∫∫<br />

(4.13)<br />

( ) ( , , , , )<br />

f θ f τφθϕϑ dφdτdϕdϑ<br />

= ∫∫∫∫<br />

(4.14)<br />

( ) ( , , , , )<br />

f ϕ f τφθϕϑ dφdθdτdϑ<br />

= ∫∫∫∫<br />

(4.15)<br />

( ) ( , , , , )<br />

f ϑ f τφθϕϑ dφdθdϕdτ<br />

where f ( τ ), f ( φ)<br />

, f ( θ ), f ( ϕ)<br />

, f ( ϑ)<br />

, <strong>and</strong> ( , , , , )<br />

= ∫∫∫∫<br />

, (4.16)<br />

f τφθϕϑ are the pdf of path delays, the pdf of<br />

azimuth departure angles, the pdf of the elevati<strong>on</strong> departure angles, the pdf of the azimuth arrival angles,<br />

the pdf of the elevati<strong>on</strong> arrival angles, <strong>and</strong> the joint double-directi<strong>on</strong>al-delay probability density functi<strong>on</strong><br />

of argument parameters, respectively. Similarly, the P ( τ ) , P ( φ)<br />

, P ( θ ) , P ( ϕ)<br />

, P ( ϕ)<br />

can be expressed<br />

as:<br />

( ) { }<br />

P τ = LE p| τ f( τ)<br />

(4.17)<br />

Page 45 (167)


WINNER D5.4 v. 1.4<br />

( ) { }<br />

P φ = LE p| φ f( φ)<br />

(4.18)<br />

( ) { }<br />

P θ = LE p| θ f( θ)<br />

(4.19)<br />

( ) { }<br />

P ϕ = LE p| ϕ f( ϕ)<br />

(4.20)<br />

( ) { }<br />

where E{ p|<br />

τ }, E{ p|<br />

φ } , E{ p|<br />

θ }, E{ p|<br />

ϕ } , <strong>and</strong> { | }<br />

P ϑ = LE p| ϑ f( ϑ)<br />

, (4.21)<br />

E p ϑ are the expected power of the<br />

multipath comp<strong>on</strong>ents c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> their delays, azimuth departure angle, elevati<strong>on</strong> departure angle,<br />

azimuth arrival angle, elevati<strong>on</strong> arrival angle, respectively.<br />

4.1.4.1.1 Expected power c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> delay<br />

The estimated expected power of multipath comp<strong>on</strong>ents c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> delays can be obtained from<br />

(4.17) as:<br />

{ } ( )<br />

E p| τ ∝ P τ / f( τ)<br />

. (4.22)<br />

In order to make the c<strong>on</strong>cept of the generic <strong>channel</strong> model approach clear, we can think of the case when<br />

both P ( τ ) <strong>and</strong> f ( τ ) are exp<strong>on</strong>ential decaying functi<strong>on</strong>s. The <strong>on</strong>e-side exp<strong>on</strong>ential decaying functi<strong>on</strong><br />

P τ is expressed as:<br />

that describes the ( )<br />

where<br />

P<br />

( τ )<br />

( −τ<br />

σ ),<br />

⎧ exp<br />

τ<br />

for τ > 0<br />

⎪<br />

∝ ⎨<br />

⎪<br />

⎩0,<br />

otherwise<br />

(4.23)<br />

σ<br />

τ is the RMS delay spread. The exp<strong>on</strong>ential functi<strong>on</strong> that describes the probability density<br />

f τ is expressed as:<br />

f τ ∝ exp −τ<br />

~<br />

, (4.24)<br />

functi<strong>on</strong> of the delays ( )<br />

where<br />

σ ~<br />

τ<br />

( ) ( )<br />

is st<strong>and</strong>ard deviati<strong>on</strong> of the path delays. Hence, the expected power c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> delay<br />

(4.25) can be written as:<br />

Now, let us define a parameter r τ as follows:<br />

use (4.26) in (4.25), we get:<br />

σ τ<br />

⎛ σ%<br />

τ<br />

−σ<br />

⎞<br />

τ<br />

Pn<br />

= E{ p| τ}<br />

∝exp<br />

⎜−τ ⎟. (4.25)<br />

⎝ σσ %<br />

τ τ ⎠<br />

σ ~<br />

τ<br />

r<br />

τ<br />

=<br />

(4.26)<br />

στ<br />

⎛ rτ<br />

−1⎞<br />

Pn<br />

= E{ p| τ}<br />

∝exp⎜−τ<br />

⎟<br />

⎝ rτσ<br />

τ ⎠ . (4.27)<br />

Thus, the expected power of multipath comp<strong>on</strong>ents c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> delay depends <strong>on</strong> the RMS delay<br />

spread <strong>and</strong> the parameter that describes the ratio between the path delays st<strong>and</strong>ard deviati<strong>on</strong> <strong>and</strong> the RMS<br />

delay spread.<br />

For the case when P ( τ ) is exp<strong>on</strong>ential as in (4.24) <strong>and</strong> the f ( τ ) is uniform U ( 0,<br />

τ )<br />

power c<strong>on</strong>diti<strong>on</strong>ed <strong>on</strong> delay (4.22) can be written as:<br />

Pn<br />

4.1.4.1.2 The power azimuth-delay spectrum<br />

max<br />

, the expected<br />

⎛ τ ⎞<br />

= E{ p| τ}<br />

∝exp⎜−<br />

⎟<br />

⎝ στ<br />

⎠ , τ ≤ τ<br />

max<br />

(4.28)<br />

We will focus our discussi<strong>on</strong> <strong>on</strong> azimuth angles at both transmitter <strong>and</strong> receiver. Now, we call the double-<br />

P φ , ϕ,<br />

τ <strong>and</strong> its corresp<strong>on</strong>ding<br />

directi<strong>on</strong>al-delay spectrum as the double-azimuth-delay spectrum, i.e., ( )<br />

Page 46 (167)


WINNER D5.4 v. 1.4<br />

pdf as the double-azimuth-delay probability density functi<strong>on</strong> f ( φ , ϕ,<br />

τ ). The joint functi<strong>on</strong>s f ( φ , ϕ,<br />

τ )<br />

<strong>and</strong> P ( φ , ϕ,<br />

τ ) are mathematically intractable as it is a joint distributi<strong>on</strong> of n<strong>on</strong>-Gaussian r<strong>and</strong>om<br />

variables. Hence, we can study the power azimuth-angle-delay spectrum, P ( φ , ϕ,<br />

τ ). Under the<br />

assumpti<strong>on</strong> that the power spectrum functi<strong>on</strong> of <strong>on</strong>e domain is independent of partial informati<strong>on</strong> in other<br />

domains, if ∆τ<br />

is small enough such that the RMS delay spread of multipath comp<strong>on</strong>ents within ∆ τ is<br />

very small <strong>and</strong> close to zero, while they are separated in azimuth-departure angle domain, we call P ~ ( φ)<br />

as the zero-delay-spread cluster of departure (ZDSC_D). This defines cluster characteristics of multipath<br />

comp<strong>on</strong>ents that are separated in azimuth angle of departure domain but have almost same delays.<br />

Similarly with the power azimuth-arrival-angle-delay spectrum such that having the same argument about<br />

having ∆ τ small enough such that the RMS delay spread of multipath comp<strong>on</strong>ents within ∆ τ is very<br />

small <strong>and</strong> close to zero while they are separated in azimuth-arrival angles domain, we call P ~ ( ϕ)<br />

as the<br />

zero-delay-spread cluster of arrival (ZDSC_A). This defines cluster characteristics of multipath<br />

comp<strong>on</strong>ents that are separated in azimuth-angle of arrival domain <strong>and</strong> have almost same delay. With the<br />

argument discussed above we can say that:<br />

4.1.4.1.3 Large-scale parameters<br />

P<br />

( φ ϕ,<br />

τ ) ∝ P( φ) P( ϕ) P( τ )<br />

P is a functi<strong>on</strong> of the RMS angle-<br />

It is known that P ( φ)<br />

is a functi<strong>on</strong> of the RMS angle-spreadσ φ , ( ϕ)<br />

spreadσ , <strong>and</strong> ( τ )<br />

, (4.29)<br />

ϕ P is a functi<strong>on</strong> of the RMS delay spreadσ τ . For each power departure-azimuthdelay<br />

spectrum <strong>and</strong> power arrival-azimuth-delay spectrum that represents a specific <strong>channel</strong> segment, the<br />

sets ( σ , σ , σ ) are c<strong>on</strong>sidered fixed but they change from <strong>channel</strong> segment to another with movement<br />

φ<br />

τ<br />

ϕ<br />

of the MS. Hence, these sets can be c<strong>on</strong>sidered as r<strong>and</strong>om variables. Therefore, they can be described by<br />

f σ , σ , σ . The marginal probability density functi<strong>on</strong> of each<br />

joint probability density functi<strong>on</strong> as ( )<br />

dispersi<strong>on</strong> metric can be obtained as:<br />

f<br />

f<br />

f<br />

ϕ<br />

φ<br />

τ<br />

( στ<br />

) = ∫∫ f ( σ<br />

ϕ<br />

σ<br />

φ<br />

, στ<br />

) dσφdσ<br />

ϕ<br />

( σ<br />

φ<br />

) = ∫∫ f ( σ<br />

ϕ<br />

σ<br />

φ<br />

, στ<br />

) dστdσ<br />

ϕ<br />

( σ<br />

ϕ<br />

) = ∫∫ f ( σϕ<br />

σ<br />

φ<br />

, στ<br />

) dστdσ<br />

ϕ<br />

, (4.30)<br />

, (4.31)<br />

, (4.32)<br />

In literature the distributi<strong>on</strong>s of these parameters are usually <str<strong>on</strong>g>report</str<strong>on</strong>g>ed as lognormal for some of outdoor<br />

scenarios. To represent the <strong>channel</strong> characteristics, the sets ( σ , σ , σ ) must be selected r<strong>and</strong>omly<br />

while c<strong>on</strong>sidering the correlati<strong>on</strong> between them to represent their <strong>channel</strong> segment.<br />

4.1.4.1.4 Bulk parameter cross-correlati<strong>on</strong><br />

Generati<strong>on</strong> of multipath comp<strong>on</strong>ent characteristics in teRMS of rays parameters, i.e., delays, angle of<br />

departures <strong>and</strong> angle of arrivals are drawn from r<strong>and</strong>om number generators specified by probability<br />

density functi<strong>on</strong>s of the corresp<strong>on</strong>ding parameters by combining the Gaussian distributi<strong>on</strong> with the<br />

transformati<strong>on</strong> functi<strong>on</strong> g ( x)<br />

, see Secti<strong>on</strong> 4.1.4.2 below. These distributi<strong>on</strong>s are functi<strong>on</strong>s of dispersi<strong>on</strong><br />

metrics that are discussed in Secti<strong>on</strong> 3.1.1 earlier. These dispersi<strong>on</strong> metrics might be correlated with each<br />

other, with lognormal shadowing <strong>and</strong> cross-polarisati<strong>on</strong> ratio. Thus, correlati<strong>on</strong> has to be c<strong>on</strong>sidered in<br />

generati<strong>on</strong> of dispersi<strong>on</strong> metric <strong>and</strong> shadowing. For each <strong>link</strong>, the correlati<strong>on</strong>s between all large-scale<br />

parameters are taken into account. In additi<strong>on</strong>, the correlati<strong>on</strong> of these parameters between two MS <strong>and</strong><br />

<strong>on</strong>e BS (or <strong>on</strong>e MS at two points in time) are modelled by c<strong>on</strong>sidering the auto-correlati<strong>on</strong> properties of<br />

the large-scale parameters. However, the cross-correlati<strong>on</strong> in the <strong>link</strong>s between <strong>on</strong>e MS <strong>and</strong> two BS are<br />

set to zero in this model based <strong>on</strong> the discussi<strong>on</strong> in Secti<strong>on</strong> 4.1.4.2.3.<br />

4.1.4.1.5 Azimuth angle distributi<strong>on</strong>s of ZDSC<br />

The mean departure angle of ZDSC_D <strong>and</strong> mean arrival angle of ZDSC_A can be located anywhere<br />

within the azimuth-departure-angle domain or azimuth-arrival-angle domain. The departure (arrival)<br />

angles of the rays within the ZDSC_D (ZDSC_A) are generated to satisfy certain angle-spreads within the<br />

cluster. In order to reduce the complexity of the <strong>channel</strong> model the same angle-spreads of all ZDSC is<br />

assumed. These angle-spreads may vary from scenario to another. In order to minimize the model<br />

complexity, the angle spacing between rays within the cluster is c<strong>on</strong>sidered fixed to satisfy a specific<br />

angle-spread. The azimuth angles spacing of rays is predefined as an offset from a mean angle of the<br />

φ<br />

τ<br />

ϕ<br />

Page 47 (167)


WINNER D5.4 v. 1.4<br />

cluster (ZDSC). These mean angles of the ZDSCs are generated by r<strong>and</strong>om generators of defined<br />

probability density functi<strong>on</strong>s. The probability density functi<strong>on</strong>s of azimuth angles of ZDSC of either<br />

departure or arrival are denoted as f ( φ)<br />

<strong>and</strong> f ( ϕ)<br />

, respectively, are independent of their delays. When<br />

the pdf of ZDSC_A <strong>and</strong> ZDSC_D are zero mean truncated Gaussian, they can be written as<br />

where Ψ =<br />

<strong>and</strong><br />

where Ω =<br />

Here<br />

σ ~ φ <strong>and</strong><br />

1<br />

2πσ~<br />

1<br />

2πσ~<br />

ϕ<br />

φ<br />

π<br />

∫<br />

−π<br />

π<br />

∫<br />

−π<br />

⎛<br />

2<br />

⎜<br />

ϕ<br />

exp<br />

−<br />

⎝ 2 σ ~<br />

2<br />

⎛ φ<br />

exp<br />

⎜ −<br />

⎝ 2 σ ~<br />

⎛<br />

2<br />

( ) = 1<br />

⎞<br />

⎜<br />

ϕ<br />

f ϕ<br />

exp − ⎟<br />

2 ~ Ψ<br />

⎝ 2 σ ~ 2<br />

(4.33)<br />

πσ<br />

ϕ ϕ ⎠<br />

2<br />

ϕ<br />

⎞<br />

⎟<br />

dϕ<br />

,<br />

⎠<br />

⎛<br />

2<br />

( ) = 1<br />

⎞<br />

⎜<br />

φ<br />

f φ<br />

exp − ⎟<br />

2 ~ Ω<br />

⎝ 2 σ ~ 2<br />

(4.34)<br />

πσ<br />

φ φ ⎠<br />

2<br />

⎞<br />

⎟dφ<br />

⎠<br />

σ ~ ϕ are st<strong>and</strong>ard deviati<strong>on</strong>s <strong>and</strong> are related to the RMS angle-spreads σ<br />

φ <strong>and</strong><br />

σ<br />

ϕ represent the composite RMS angle-<br />

the parameter r φ <strong>and</strong> r ϕ , respectively. Note that the<br />

spreads not per cluster angle-spreads (AS).<br />

4.1.4.1.6 Impulse resp<strong>on</strong>se of ZDSC<br />

σ<br />

φ <strong>and</strong><br />

σ<br />

ϕ through<br />

With very wide b<strong>and</strong>width, fading of comp<strong>on</strong>ent of certain delays is due to interference between<br />

multipath comp<strong>on</strong>ents that arrive in clusters having same or very close delays but differ in angle of<br />

arrivals <strong>and</strong>/or angle of departures. This has been discussed previously <strong>and</strong> represents the c<strong>on</strong>cept of<br />

ZDSC. Having the c<strong>on</strong>cept of ZDSC, the functi<strong>on</strong> (D 3 SF) can be written as:<br />

N M<br />

, , , , , , , ,<br />

( τφθϕϑ) = ∑∑ αnm , ( t) δ ( φ−φnm ,<br />

θ −θnm ,<br />

ϕ −ϕnm ,<br />

ϑ−ϑnm , ) δ ( τ −τn)<br />

ht<br />

n= 1 m=<br />

1<br />

(4.35)<br />

where N is number of ZDSCs, <strong>and</strong> M is the number of rays within the cluster. Here in (4.37), we assume<br />

same number of rays in each ZDSC. The spacing in angle domain between rays around mean angle of the<br />

cluster is determined to satisfy certain angle-spread of certain power azimuth spectrum. The power<br />

divisi<strong>on</strong> between rays of total cluster power could be dependent of angle of arrival (departure) or same<br />

power in all rays can also be assumed. For the case when equal power between rays is assumed, the<br />

angles are separated based <strong>on</strong> certain PAS. One widely used PAS is the Laplacian power spectrum, the<br />

power of each ray is P n<br />

M , where P<br />

n is the power of the nth cluster, the departure or arrival angles are<br />

spaced n<strong>on</strong>-uniformly to based <strong>on</strong> Laplacian PAS.<br />

4.1.4.2 Correlati<strong>on</strong> of large-scale parameters between <strong>link</strong>s<br />

In the generic WINNER model, large-scale parameters give a higher-<strong>level</strong> characterizati<strong>on</strong> of the<br />

propagati<strong>on</strong> <strong>channel</strong>. These parameters are treated as r<strong>and</strong>om variables <strong>on</strong> a <strong>channel</strong> segment basis. They<br />

are r<strong>and</strong>omized in a first step – <strong>and</strong> <strong>on</strong>ly there-after - are the detailed parameters of the <strong>channel</strong> model<br />

being r<strong>and</strong>omized using these large-scale parameters as input.<br />

The following large-scale parameters may be c<strong>on</strong>sidered (currently <strong>on</strong>ly the first 6 are actually used)<br />

1. Delay-spread<br />

2. AoD angle-spread<br />

3. AoA angle-spread<br />

4. Shadow fading<br />

5. AoD elevati<strong>on</strong> spread.<br />

6. AoA elevati<strong>on</strong> spread.<br />

7. Cross polarisati<strong>on</strong> ratio 1.<br />

8. Cross polarisati<strong>on</strong> ratio 2.<br />

Page 48 (167)


WINNER D5.4 v. 1.4<br />

Depending <strong>on</strong> the measurement capabilities <strong>and</strong> the scenario requirements some of the parameters may be<br />

neglected. Let us denote the vector of these elements s ( x, y)<br />

where we think of s( x, y)<br />

as a stochastic<br />

s x, y be m .We call<br />

multivariate process where x <strong>and</strong> y is the positi<strong>on</strong> of the user. Let the size of ( )<br />

s ( x, y)<br />

“the large-scale parameter vector” or “large-scale vector” for short.<br />

The angles in ( x, y)<br />

unit-less (i.e. not in dB). The value of s ( x, y)<br />

in two positi<strong>on</strong>s s ( ) <strong>and</strong> s(<br />

)<br />

s are taken to be degrees, the delay-spread in sec<strong>on</strong>ds, <strong>and</strong> the remaining <strong>on</strong>es are<br />

x , y x , y 1 1<br />

2 2 can be used to<br />

represent two users or the same user which is at the two positi<strong>on</strong>s at two different time instants. When<br />

1<br />

2<br />

two base-stati<strong>on</strong> sites are involved, two different vectors s ( x, y)<br />

<strong>and</strong> s ( x, y)<br />

characterize the <strong>link</strong> to the<br />

two sites <strong>and</strong> the positi<strong>on</strong> x, y , respectively.<br />

In the first three secti<strong>on</strong>s following, we further elaborate this model in teRMS of distributi<strong>on</strong>, autocorrelati<strong>on</strong>,<br />

<strong>and</strong> inter-base-stati<strong>on</strong> correlati<strong>on</strong>. In the fourth secti<strong>on</strong> we describe how an evolving <strong>channel</strong><br />

can be generated based <strong>on</strong> the model.<br />

The model obtained here is similar to the ideas in [Alg02]. Here however, we c<strong>on</strong>sider <strong>system</strong> <strong>level</strong><br />

variables with different correlati<strong>on</strong> distances as well as n<strong>on</strong> log-normal variables. We also analyze crosscorrelati<strong>on</strong><br />

functi<strong>on</strong>s an issue which is completely overlooked in [Alg02]. Furthermore, we discuss<br />

extensi<strong>on</strong> to include inter-site correlati<strong>on</strong>s in Secti<strong>on</strong> 4.1.4.2.3 below.<br />

4.1.4.2.1 Distributi<strong>on</strong>s<br />

Based <strong>on</strong> measurements <strong>and</strong> literature, we have found transfoRMS g ( s)<br />

for each element of ( x, y)<br />

~<br />

that the transformed vector ( x , y) g( s( x,<br />

y)<br />

)<br />

s such<br />

s = is a vector of Gaussian r<strong>and</strong>om variables for each scenario.<br />

We assume that the elements of this vector are jointly Gaussian. We have also found the inverse<br />

transform, such that we can easily generate samples of the distributi<strong>on</strong> by generating a vector of r<strong>and</strong>om<br />

−<br />

s x , y = g<br />

1 ~<br />

s x,<br />

y .<br />

variables <strong>and</strong> then perform the inverse i.e. ( ) ( ( ))<br />

4.1.4.2.2 Auto-correlati<strong>on</strong>s<br />

In order to facilitate generati<strong>on</strong> of multiple realizati<strong>on</strong>s of ~ s ( x, y)<br />

in several<br />

positi<strong>on</strong>s x = x , y = y 1 1 , x = x , y = y<br />

~ 2 2 , … need to know the mean <strong>and</strong> the auto-correlati<strong>on</strong> of the vector<br />

s ( x, y)<br />

i.e.<br />

= {<br />

~ s ( x, y)<br />

} , R( r ) = Ε (<br />

~<br />

s( x , y ) − µ )( ~<br />

s( x y ) − µ )<br />

µ Ε<br />

{ }<br />

T<br />

∆<br />

1 1<br />

0,<br />

0<br />

2<br />

, where ( ) ( ) 2<br />

∆ r = x<br />

, (4.36)<br />

1 − x0<br />

+ y1<br />

− y0<br />

where we have assumed that the autocorrelati<strong>on</strong> is <strong>on</strong>ly a functi<strong>on</strong> of the distance between any two points.<br />

The correlati<strong>on</strong> matrix c<strong>on</strong>tains the cross-correlati<strong>on</strong> functi<strong>on</strong>s between all element variables. In order to<br />

arrive at a model which is usable also when simulating cases involving many positi<strong>on</strong>s we have impose a<br />

structure <strong>on</strong> R ( ∆r)<br />

namely<br />

R<br />

⎛<br />

⎜<br />

⎝<br />

⎛<br />

⎜<br />

⎝<br />

∆r<br />

⎞ ⎛ ∆r<br />

⎞⎞<br />

⎟<br />

K<br />

⎜ ⎟⎟<br />

(*) (4.37)<br />

λ1<br />

⎠ ⎝ λm<br />

⎠⎠<br />

0.5<br />

0.5,T<br />

( ∆r) = R ( 0) diag⎜exp⎜−<br />

⎟,<br />

,exp⎜−<br />

⎟⎟R<br />

( 0)<br />

0.5<br />

T 0.5<br />

5<br />

where R ( 0)<br />

is obtained from the eigendecompositi<strong>on</strong> R( 0) = EΛE<br />

as R ( 0) = EΛ<br />

0. .<br />

Using a model with this structure we can simulate realizati<strong>on</strong>s of ( x, y)<br />

m independent Gaussian r<strong>and</strong>om processes, ?( x, y) [ ξ ( x,<br />

y) ( x y)<br />

] T<br />

1<br />

Kξ<br />

m<br />

,<br />

variance <strong>on</strong>e. The autocorrelati<strong>on</strong> of process ξc ( x,<br />

y)<br />

is given by exp( − ∆r / λ c<br />

)<br />

dimensi<strong>on</strong>al “maps” can be performed with a filtering operati<strong>on</strong>. Following the generati<strong>on</strong> of ( x, y)<br />

transformed large-scale vector is obtained as<br />

~ s by first generating<br />

= , each <strong>on</strong>e with mean zero,<br />

0.<br />

( x,<br />

y) = R ( ∆r) ?( x y) + µ<br />

~ 5<br />

s ,<br />

. Generating such two<br />

? the<br />

. (4.38)<br />

Thus we have in Secti<strong>on</strong> fitted model parameters µ <strong>and</strong>λ , K,λ<br />

to our measurements.<br />

Note that the resulting effective autocorrelati<strong>on</strong> functi<strong>on</strong> for each large-scale parameter is not exp<strong>on</strong>ential,<br />

as comm<strong>on</strong>ly found in literature, but rather a sum of weighted exp<strong>on</strong>ential functi<strong>on</strong>s.<br />

4.1.4.2.3 Multi-site cross-correlati<strong>on</strong>s<br />

The justificati<strong>on</strong> for introducing the cross-correlati<strong>on</strong>s of the large-scale parameters is that the <strong>link</strong>s<br />

between a pair of base-stati<strong>on</strong>s <strong>and</strong> a mobile-stati<strong>on</strong> is that 1) there may be many comm<strong>on</strong> scatterers in<br />

the close proximity of the MS 2) comm<strong>on</strong> shadowing objects of two mobiles located close in angle <strong>and</strong> 3)<br />

1<br />

m<br />

Page 49 (167)


WINNER D5.4 v. 1.4<br />

cell sub-areas with different local propagati<strong>on</strong> characteristics as indicated in Figure 4.1, Figure 4.2 <strong>and</strong><br />

Figure 4.3, respectively. In the example of Figure 4.2 the correlati<strong>on</strong> could arise from the obstructi<strong>on</strong> of<br />

the same building while in Figure 4.3 the correlati<strong>on</strong> arises from the fact that the local envir<strong>on</strong>ment of the<br />

mobile stati<strong>on</strong> is the same for both base-stati<strong>on</strong>s. In the examples of Figure 4.1<strong>and</strong> Figure 4.2 it is<br />

reas<strong>on</strong>able that the correlati<strong>on</strong> would increase when the angle β between the two base-stati<strong>on</strong>s seen at the<br />

mobile stati<strong>on</strong>, see Figure 4.4. Such a dependence correlati<strong>on</strong> has been observed in [Maw92] where the<br />

shadow-fading is investigated.<br />

X<br />

BS<br />

MS<br />

X<br />

BS<br />

Figure 4.1: Links to two base-stati<strong>on</strong>s with comm<strong>on</strong> scatterers.<br />

BS<br />

BS<br />

MS<br />

Figure 4.2: Shadowing by the same object in the two <strong>link</strong>s of two different base-stati<strong>on</strong>s.<br />

BS<br />

A<br />

B<br />

C<br />

BS<br />

Figure 4.3: Two base-stati<strong>on</strong>s with three local areas A, B <strong>and</strong> C which are characterized as A) open<br />

green-field, B) wooded area C) Built-up area.<br />

BS1<br />

r 1 r 2<br />

B<br />

BS2<br />

h 1<br />

h<br />

2<br />

MS<br />

Figure 4.4: The geometry of two base-stati<strong>on</strong>s <strong>and</strong> a mobile-stati<strong>on</strong>.<br />

Page 50 (167)


WINNER D5.4 v. 1.4<br />

A measurement campaign has been c<strong>on</strong>ducted in a metropolitan typical urban macro-cell (scenario C2),<br />

with simultaneous measurements of the three <strong>link</strong>s between a single mobile-stati<strong>on</strong>, <strong>and</strong> three sector<br />

antenna arrays, <strong>on</strong>e of them located just above average rooftop <strong>level</strong> <strong>and</strong> the other two sectors well above<br />

roof-top <strong>level</strong>, see Secti<strong>on</strong> 5.2.5. No correlati<strong>on</strong> was found between the two sites in the measurements.<br />

However, this is believed to be due to measurement problems <strong>and</strong> we believe the true correlati<strong>on</strong> between<br />

sectors of the same cell is full. The geometry of the measurements where such that the angle β between<br />

the base-stati<strong>on</strong>s was typically large as illustrated in Figure 4.4. In additi<strong>on</strong> the base-stati<strong>on</strong> heights were<br />

different. These measurements show that for metropolitan typical urban envir<strong>on</strong>ments (scenario C2)<br />

under the c<strong>on</strong>diti<strong>on</strong>s stated the correlati<strong>on</strong> is zero.<br />

In the future it is possible to develop <strong>models</strong> the correlati<strong>on</strong> is situati<strong>on</strong>s where it exists. It is likely that it<br />

<strong>on</strong>ly exist in a few situati<strong>on</strong>s when β is small <strong>and</strong> the distances r 1 <strong>and</strong> r2<br />

are close. This situati<strong>on</strong> would<br />

actually be preferable from the point of view of complexity of the <strong>channel</strong> generati<strong>on</strong> in simulati<strong>on</strong>s. A<br />

computati<strong>on</strong>ally efficient way of introducing such correlati<strong>on</strong>s is described by the following two<br />

equati<strong>on</strong>s<br />

<strong>and</strong><br />

( ) µ<br />

0.5<br />

( x,<br />

y) = ( ∆r) ? ( x,<br />

y) + ( x y)<br />

~ 1<br />

1<br />

s R ξ , +<br />

(4.39)<br />

( ) µ<br />

0.5<br />

( x,<br />

y) = ( ∆r) ? ( x,<br />

y) + ( x y)<br />

~ 2<br />

2<br />

s R ξ , + , respectively, (4.40)<br />

1<br />

2<br />

where the r<strong>and</strong>om processes ξ ( x, y)<br />

, ? ( x, y)<br />

<strong>and</strong> ( x, y)<br />

~1 ~2<br />

s ( x,<br />

y)<br />

<strong>and</strong> s ( x,<br />

y)<br />

is introduced by the comm<strong>on</strong> r<strong>and</strong>om process ξ ( x, y)<br />

T<br />

{?<br />

( x y) ? ( x,<br />

y)<br />

} = diag( c , K,<br />

)<br />

<strong>and</strong><br />

,<br />

1<br />

? are independent <strong>and</strong> the correlati<strong>on</strong> between<br />

c m<br />

.Assuming<br />

E (4.41)<br />

1 1,T<br />

2 2,T<br />

{?<br />

( x y) ? ( x,<br />

y)<br />

} E ? ( x,<br />

y) ? ( x,<br />

y)<br />

{ } = diag( 1- c , K,<br />

− )<br />

E = 1 , (4.42)<br />

this produces a cross-correlati<strong>on</strong> of the form<br />

E<br />

,<br />

1<br />

{ ~ 1 2, T<br />

0.5<br />

0.5, T<br />

s ( x,<br />

y) ~ s ( x,<br />

y)<br />

} = ( ∆r) diag( c , , c ) R ( ∆r)<br />

R K . (4.43)<br />

A key questi<strong>on</strong> here is whether this correlati<strong>on</strong> structure <strong>models</strong> the true cross-correlati<strong>on</strong> accurately<br />

enough or not. In the multi-cell measurements available now correlati<strong>on</strong> was found between sites, see<br />

Secti<strong>on</strong>s 5.2.5.1 <strong>and</strong> 9.3.1. For this reas<strong>on</strong> the <strong>system</strong> <strong>level</strong> vectors of different sites will be modeled as<br />

independent.<br />

In the WINNER model as well as in SCM the shadow-fading is correlated with a correlati<strong>on</strong> coefficient<br />

of 0.5. The <strong>on</strong>ly reference known to the authors where inter-site correlati<strong>on</strong> has been studied is [Maw92].<br />

In this paper the cross-correlati<strong>on</strong> between the log-normal-fading is approximated as 0.9<br />

− θ / 200 where<br />

θ is the angle between the two sites <strong>and</strong> seen from the MS. The model is based <strong>on</strong> measurements two<br />

base-stati<strong>on</strong>s at 45 <strong>and</strong> 90 meter height, the distance to the two base-stati<strong>on</strong>s are 2-38 km <strong>and</strong> the<br />

frequencies are in the 154-922 MHz range. These parameters are substantially different from the<br />

measurements made in the WINNER project <strong>and</strong> we therefore chose to put neglect them.<br />

4.2 Stati<strong>on</strong>ary-feeder scenarios B5<br />

It should be noted that for the other prioritized scenarios WP5 have measurement data <strong>and</strong> improved<br />

modelling of them is an <strong>on</strong>going effort while for the feeder scenarios this is not the case. In any case WP5<br />

feels that the interim <strong>channel</strong> <strong>models</strong> are sufficiently good as a starting point for simulati<strong>on</strong>s <strong>and</strong> further<br />

discussi<strong>on</strong>s. In stati<strong>on</strong>ary feeder both terminals are fixed, the <strong>channel</strong> modelling approach is clustered<br />

delay-line modelling approach. Clustered delay line <strong>models</strong> for the first two sub-scenarios (B5a <strong>and</strong> B5b),<br />

defined in Subsecti<strong>on</strong> 1.3.2.1.3, are defined in Secti<strong>on</strong> 4 based <strong>on</strong> the analysis below <strong>and</strong> we describe how<br />

<strong>models</strong> for the B5c <strong>and</strong> B5d sub-scenarios could be obtained by modifying scenario B1 <strong>and</strong> C2,<br />

respectively.<br />

4.2.1 B5a LOS stati<strong>on</strong>ary feeder: rooftop-to-rooftop<br />

The propagati<strong>on</strong> scenario in the LOS measurements in [PT00] <strong>and</strong> [SCK05] is the most similar to ours,<br />

although in [SCK05] the distance is very short. Therefore, we will base the propagati<strong>on</strong> model mostly <strong>on</strong><br />

[PT00] for the parameters that are available from in that paper. Remaining parameters will be derived<br />

from the other papers. Only a tapped delay-line model is provided. This scenario probably can be<br />

characterized by a str<strong>on</strong>g line of sight <strong>and</strong> single-bounce reflecti<strong>on</strong> as indicated in Figure 4.5. If the LOS<br />

path is slightly obstructed, the influence of multi-path-related parameters will be str<strong>on</strong>g, <strong>and</strong> far-away<br />

1<br />

m<br />

c m<br />

Page 51 (167)


WINNER D5.4 v. 1.4<br />

reflecti<strong>on</strong>s can be expected due to the free-space c<strong>on</strong>diti<strong>on</strong>s from/to the reflectors. Due to lack of<br />

measurements we will use fixed angle-spread, delay-spread <strong>and</strong> XPR-value. L<strong>on</strong>g <strong>and</strong> short-term fading<br />

will be used however. Thus <strong>on</strong>ly a tapped delay-line model will be provided for this scenario. Note that<br />

due to the single-bounce nature of the propagati<strong>on</strong>, directive antennas are very effective in reducing<br />

delay-spread <strong>and</strong> other impacts of multi-path, see e.g. the delay-spread reducti<strong>on</strong> in [PT00]. We use the<br />

RMS-delay-spread value of 40 ns. This is the largest value observed in a measurement campaign which<br />

utilized antennas with 53 degrees <strong>and</strong> 10 degree opening angles in the <strong>link</strong>-ends, see [PT00]. This is also<br />

close to the median RMS-delay-spread with basically omni-directi<strong>on</strong>al antennas measured in [OBL+02]<br />

but somewhat larger than in [SCK05] but the measurements in [SCK05] are at very short range.<br />

Therefore, the model should be understood such that it is applicable using omni-directi<strong>on</strong>al antennas for<br />

up to 300meters distance, while beam-widths comparable or narrower that the aforementi<strong>on</strong>ed 10/53<br />

degrees should be used at larger distances.<br />

X<br />

X<br />

Figure 4.5: Single Bounce Reflecti<strong>on</strong> Model<br />

4.2.2 B5b LOS stati<strong>on</strong>ary feeder: street-<strong>level</strong> to street-<strong>level</strong><br />

The measurement campaigns listed in the literature review are performed at very different frequencies: all<br />

the way from 2 to 10GHz. However, in papers e.g. [Bal02], [SBA+02] the results for different carried<br />

frequencies are very similar. Therefore we chose to disregard the difference in frequency for this interim<br />

<strong>channel</strong> model. The principle adopted for the WINNER <strong>models</strong> allows for various correlati<strong>on</strong>s between<br />

different parameters such as angle-spread, shadow-fading <strong>and</strong> delay-spread. We will use <strong>on</strong>e such<br />

dependence namely that in [MKA02], which dependence is between path loss <strong>and</strong> delay-spread. For B5b<br />

however, <strong>on</strong>ly Clustered-delay-line <strong>models</strong> will be provided (CDL) <strong>and</strong> the dependence between path loss<br />

<strong>and</strong> delay-spread is h<strong>and</strong>led by selecting <strong>on</strong>e of three different CDL <strong>models</strong>.<br />

Our underst<strong>and</strong>ing of the scenario is that both the transmitter <strong>and</strong> receivers have many scatterers in their<br />

close vicinity similar as theorized in [Sva02]. In additi<strong>on</strong> there can also be l<strong>on</strong>g echoes from the ends of<br />

the street. However, there is a line-of-sight ray between the transmitter <strong>and</strong> receiver. When this path is<br />

str<strong>on</strong>g the c<strong>on</strong>tributi<strong>on</strong> from all the scatters is small <strong>and</strong> therefore also all the different foRMS of<br />

dispersi<strong>on</strong>. However, bey<strong>on</strong>d the breakpoint distance the scatterers start to play an important role. Based<br />

<strong>on</strong> the BS <strong>and</strong> MS height of most references we assume that model is valid for 2-5 meter access point<br />

heights. A clustered delay-line model with the properties given below is defined. The parameters <strong>and</strong> their<br />

motivati<strong>on</strong>s are as follows.<br />

4.2.3 B5c hotspot LOS stati<strong>on</strong>ary-feeder: below rooftop to street-<strong>level</strong>.<br />

This can be modelled identical to the LOS versi<strong>on</strong> of the B1 model except that support for the Doppler<br />

spectrum of stati<strong>on</strong>ary cases has to be introduced. How to support higher feeder peripheral stati<strong>on</strong><br />

antennas than typical mobile-stati<strong>on</strong> heights has not been c<strong>on</strong>sidered yet.<br />

We propose the introducti<strong>on</strong> of individual Doppler frequencies similar to the model in [TPE02]. The<br />

Doppler frequency will not be a functi<strong>on</strong> of the AoA at the receiver since the <strong>channel</strong> variati<strong>on</strong> is not due<br />

to temporal variati<strong>on</strong>s of the <strong>channel</strong> in fixed applicati<strong>on</strong>s. We select the Doppler model of [Erc01],<br />

which has somewhat higher Doppler spread than [DGM+03] probably due to the influence of traffic.<br />

4.2.4 B5d hotspot NLOS stati<strong>on</strong>ary feeder: rooftop to street-<strong>level</strong>.<br />

This model is based <strong>on</strong> C2 model except the Doppler spectrum <strong>and</strong> an additi<strong>on</strong>al term in the path-loss<br />

model. The Doppler spectrum can be h<strong>and</strong>led as in B5c. To support higher heights of the feeder<br />

peripheral stati<strong>on</strong>s than in the C2 model, a compensati<strong>on</strong> term is introduced. We have investigated the<br />

term based <strong>on</strong> the Cost 231 Walfish-Ikegami, Walfish-Bert<strong>on</strong>i <strong>and</strong> Hata-<strong>models</strong> [Cost231], [MBX94],<br />

[Hat80] for the scenario depicted in Figure 3-2. We have set the parameters to w =30meter, x=w/2,<br />

h = h +10<br />

b B . The results for h<br />

B<br />

=12, 18 <strong>and</strong> 24 meter are shown in Figure 4.7. As a comprise between<br />

these curves we propose a gain from using a higher MS antenna than 0.1meter as<br />

Page 52 (167)


WINNER D5.4 v. 1.4<br />

Heigh_Gain dB<br />

= 0.7h m<br />

(4.44)<br />

where the gain is in dB <strong>and</strong> the height h<br />

m is in meters. The Doppler modelling is made identical to that of<br />

B5c.<br />

Figure 4.6: Illustrati<strong>on</strong> of the c<strong>on</strong>sidered scenario in B5 NLOS stati<strong>on</strong>ary feeder: rooftop to street<strong>level</strong>.<br />

Figure 4.7: Compensati<strong>on</strong> term.<br />

4.3 Coefficient generati<strong>on</strong> approaches<br />

We have selected two st<strong>and</strong>ardized spatial <strong>channel</strong> <strong>models</strong> as <strong>channel</strong> <strong>models</strong> for initial usage [D5.1],<br />

namely the 3GPP SCM <strong>and</strong> the IEEE 802.11n model. The former model is <strong>on</strong>ly for outdoor <strong>and</strong> the latter<br />

is <strong>on</strong>ly for indoor scenarios. Interestingly, these st<strong>and</strong>ards also represent two comm<strong>on</strong> but different<br />

approaches to coefficient generati<strong>on</strong>. In the following, we will evaluate the individual advantages <strong>and</strong><br />

disadvantages of these approaches. We also examine the issue of creating a model that is not restricted to<br />

Kr<strong>on</strong>ecker type spatial correlati<strong>on</strong>.<br />

4.3.1 Stati<strong>on</strong>ary stochastic<br />

Example: ETSI BRAN HIPERLAN/2, IEEE 802.11n<br />

Advantages:<br />

• Efficient coefficient generati<strong>on</strong> by correlati<strong>on</strong> of r<strong>and</strong>om variables.<br />

Disadvantages:<br />

• Calculati<strong>on</strong> of spatial autocorrelati<strong>on</strong> functi<strong>on</strong>s requires numerical integrati<strong>on</strong>.<br />

• Two-dimensi<strong>on</strong>al filtering across antenna elements <strong>and</strong> time required.<br />

Page 53 (167)


WINNER D5.4 v. 1.4<br />

• Model approach is by default stati<strong>on</strong>ary. Modelling of n<strong>on</strong>-stati<strong>on</strong>ary effects requires extensi<strong>on</strong>.<br />

4.3.2 Sum-of-Sinusoids<br />

Example: Jakes’ fading generator, 3GPP SCM<br />

Advantages:<br />

• Correlati<strong>on</strong> across antenna elements <strong>and</strong> time created implicitly.<br />

• N<strong>on</strong>-stati<strong>on</strong>ary processes potentially easier to integrate.<br />

Disadvantages:<br />

• Requires a large number of sinusoids for realistic modelling <strong>and</strong> thus computati<strong>on</strong>ally expensive.<br />

• Resulting Doppler spectra are peaky (with number of peaks less or equal to number of<br />

sinusoids).<br />

• The SOS approach builds <strong>on</strong> the assumpti<strong>on</strong> that any <strong>channel</strong> resp<strong>on</strong>se can be separated into a<br />

sum of reflectors represented as Dirac-functi<strong>on</strong>s in time <strong>and</strong> space.<br />

4.3.3 Problem details<br />

Not all of the above points might be obvious to the reader. In the following, some of the advantages <strong>and</strong><br />

disadvantages are thus explained in more details.<br />

4.3.3.1 Stochastic approach<br />

Filtering process. Essentially, the <strong>channel</strong> resp<strong>on</strong>se is correlated across space <strong>on</strong>ly. This correlati<strong>on</strong> is<br />

characterized by the spatial ACF, which is calculated by numerical integrati<strong>on</strong> from the APS. This spatial<br />

ACF is then mapped into two dimensi<strong>on</strong>s; the correlati<strong>on</strong> between signals at antenna elements depending<br />

<strong>on</strong> inter-element spacing, <strong>and</strong> the correlati<strong>on</strong> between signals in time depending <strong>on</strong> the movement of the<br />

mobile. This yields a two-dimensi<strong>on</strong>al kernel which is then used for filtering uncorrelated Gaussian<br />

samples.<br />

N<strong>on</strong>-stati<strong>on</strong>ary effects. The model is stati<strong>on</strong>ary by default. While the incorporati<strong>on</strong> of certain timevariable<br />

parameters is straightforward, e.g. Ricean K-factor, other n<strong>on</strong>-stati<strong>on</strong>ary effects, i.e. timeevoluti<strong>on</strong><br />

of Doppler spectrum or angle parameters, requires c<strong>on</strong>tinuous re-calculati<strong>on</strong> of the filter kernel<br />

<strong>and</strong> is thus computati<strong>on</strong>ally expensive.<br />

4.3.3.2 SOS approach<br />

Number of sinusoids required. In the SOS framework, fading is ensured by defining the positi<strong>on</strong> of the<br />

sinusoids in delay <strong>and</strong> angle in such a way that a minimum number of sinusoids always falls within the<br />

resoluti<strong>on</strong> capabilities of the observati<strong>on</strong> <strong>system</strong>. This minimum number between 4 <strong>and</strong> 8 [Gald04]<br />

ensures the observati<strong>on</strong> of a close to Rayleigh distributi<strong>on</strong>. If the number of sinusoids drops below this<br />

minimum amount, the observer will first see unusual distributi<strong>on</strong>s <strong>and</strong> finally identify single, discrete<br />

scatterers, both of which is typically not a desired effect. Note that str<strong>on</strong>g discrete scatterers, typically<br />

associated with LOS scenarios, are implemented as an opti<strong>on</strong>al additi<strong>on</strong>al comp<strong>on</strong>ent (SCM secti<strong>on</strong> "Line<br />

of sight") because the power of a single sinusoid is by definiti<strong>on</strong> fixed <strong>and</strong> small (e.g. 1/20 of a tap in<br />

SCM).<br />

In the following, we illustrate this point with an example. In the SOS framework, an APS is generated by<br />

changing the spacing of the sinusoids (because each sinusoid is defined to have equal power). A typical<br />

APS <str<strong>on</strong>g>report</str<strong>on</strong>g>ed for outdoor scenarios is a Laplacian functi<strong>on</strong> with a log-normal distributed AS [AIP02].<br />

Next, we determine how to distribute the sinusoids. The lowest density (sinusoids per degree) occurs at<br />

the outer ends of the Laplacian functi<strong>on</strong> <strong>and</strong> for large AS values (we pick 30 degree according to the<br />

reference). Assuming we want to be accurate to -20dB from the peak of the APS, then the angle range is<br />

roughly ±100 degrees from the centre. Let's say the maximum resoluti<strong>on</strong> of our observer is 10 degree (for<br />

example by using a highly directive antenna). In these 10 degrees we want to have a minimum of say 5<br />

sinusoids, i.e. a density of 5/10. The total required number of sinusoids then can be derived as 2345 per<br />

delay-tap.<br />

There are two ways to decrease this high number of sinusoids. One is the introducti<strong>on</strong> of variability in<br />

power of the sinusoids. The distributi<strong>on</strong> of taps can then be different to the APS <strong>and</strong>, with respect to the<br />

previous example, might be more uniform than Laplacian. Hence, in the limiting case of uniform<br />

distributi<strong>on</strong> of taps, the power at each sinusoid would vary according to the Laplace functi<strong>on</strong>, <strong>and</strong> the<br />

resulting number of sinusoids then would be 101.<br />

The sec<strong>on</strong>d approach for reducing the number of sinusoids is to assume that the AS at each path is not<br />

equal to the total AS (over all paths) but smaller (like in SCM). Following again the example from above<br />

Page 54 (167)


WINNER D5.4 v. 1.4<br />

<strong>and</strong> choosing an AS of 3 degrees, the required number of sinusoids is 239 for equal amplitude sinusoids<br />

<strong>and</strong> 11 for sinusoids with varying amplitude.<br />

Peaky Doppler spectrum. Due to the fixed velocity, each scatterer can be attributed a single Doppler<br />

frequency comp<strong>on</strong>ent. The resulting Doppler spectrum is simply the additi<strong>on</strong> of these comp<strong>on</strong>ents, which<br />

are limited in number by the amount of sinusoids per path. Ways to mitigate this are to use a high number<br />

of sinusoids, to shorten the time-durati<strong>on</strong> of the drops (frequency resoluti<strong>on</strong> decreases), or introduce<br />

instati<strong>on</strong>arity of the velocity.<br />

Discrete Scatterer Framework. The assumpti<strong>on</strong> that any <strong>channel</strong> resp<strong>on</strong>se can be separated into a sum<br />

of reflectors represented as Dirac-functi<strong>on</strong>s in time <strong>and</strong> space implies infinite accuracy. The measurement<br />

<strong>system</strong> to obtain these parameters would need infinite b<strong>and</strong>width, antennas, <strong>and</strong> power. The<br />

parameterizati<strong>on</strong> is thus problematic from an exact physical interpretati<strong>on</strong> point of view.<br />

4.3.4 Comparis<strong>on</strong><br />

It is important to note that both approaches c<strong>on</strong>verge to each other with increasing number of sinusoids<br />

for SOS <strong>and</strong> decreasing length of stati<strong>on</strong>ary segments for the stochastic approach. Hence, the essential<br />

questi<strong>on</strong> is which model is preferable for reas<strong>on</strong>able assumpti<strong>on</strong>s about the <strong>channel</strong> <strong>and</strong> implementati<strong>on</strong><br />

parameters. Channels that can not even be assumed short-term stati<strong>on</strong>ary will be more difficult to<br />

implement (with equal computati<strong>on</strong>al complexity) as a stochastic model than with SOS. On the other<br />

h<strong>and</strong>, <strong>channel</strong>s with large AS per tap will be more difficult to implement (with equal computati<strong>on</strong>al<br />

complexity) as SOS than with a stochastic model.<br />

The WINNER Channel Model follows the SOS approach. It is seen as flexible framework <strong>and</strong> it enables<br />

more easily advanced future modelling features like time evoluti<strong>on</strong> of <strong>channel</strong> model parameters.<br />

4.3.5 Kr<strong>on</strong>ecker correlati<strong>on</strong><br />

Many stochastic MIMO <strong>channel</strong> <strong>models</strong> apply what is called the Kr<strong>on</strong>ecker assumpti<strong>on</strong> for the antenna<br />

correlati<strong>on</strong> matrices. This assumpti<strong>on</strong> states that the correlati<strong>on</strong> matrix, obtained as C = E{ vec(H)<br />

vec(H) H }, can be written as a Kr<strong>on</strong>ecker product, i.e. C = C Rx ⊗ C Tx , where C Rx <strong>and</strong> C Tx are receive <strong>and</strong><br />

transmit correlati<strong>on</strong> matrices, respectively. The Kr<strong>on</strong>ecker property is useful in many ways; most<br />

importantly it significantly reduces the number of model parameters, <strong>and</strong> it greatly simplifies the<br />

analytical treatment (such as for capacity evaluati<strong>on</strong>). It implies that the joint transmit <strong>and</strong> receive APS<br />

functi<strong>on</strong> can be written as a product of two independent APS at transmitter <strong>and</strong> receiver.<br />

Other publicati<strong>on</strong>s [HOHB02] based <strong>on</strong> measurement results have made a point that this assumpti<strong>on</strong><br />

could not be verified empirically in all scenarios evaluated. In reacti<strong>on</strong> to that, researchers have tried to<br />

come up with new methods that represent a compromise between the abstracti<strong>on</strong> <strong>and</strong> simplificati<strong>on</strong> of the<br />

Kr<strong>on</strong>ecker assumpti<strong>on</strong> <strong>and</strong> the rather complex case with no assumpti<strong>on</strong>s at all. In [OHWW03], an<br />

approach is presented where the c<strong>on</strong>diti<strong>on</strong> of a separable APS is alleviated into the c<strong>on</strong>diti<strong>on</strong> of<br />

independent eigenbasis of receiver to the transmit weights, <strong>and</strong> vice versa.<br />

Our preliminary analysis shows that, while both arguments certainly have significance, it is in practice<br />

important to carefully examine the underlying basis that the correlati<strong>on</strong> matrix is computed <strong>on</strong>. We start<br />

with the most detailed model. In case of a wideb<strong>and</strong> <strong>system</strong>, the <strong>channel</strong> is represented as a tapped delay<br />

line, i.e. H(τ) = H 1 δ(τ - τ 1 ) + … + H n δ(τ - τ n ). Furthermore, each delay-tap matrix can be split into a<br />

sum of c<strong>on</strong>tributi<strong>on</strong>s from different angle clusters, i.e. H i = H i1 + … + H im . We can now argue that with<br />

sufficient splitting <strong>and</strong> thus subdivisi<strong>on</strong> of the delay-angle domain, we can always reach a point such that<br />

all the smallest parts H ij have a separable APS <strong>and</strong> thus a Kr<strong>on</strong>ecker correlati<strong>on</strong> matrix. Any <strong>system</strong> with<br />

a resoluti<strong>on</strong> capability less than that will observe <strong>on</strong>ly linear combinati<strong>on</strong>s of H ij which might well not<br />

hold up to the Kr<strong>on</strong>ecker assumpti<strong>on</strong>. Thus we can always define a <strong>channel</strong> model based <strong>on</strong> Kr<strong>on</strong>ecker<br />

correlated comp<strong>on</strong>ents, while a <strong>system</strong> employing this model might not observe such properties.<br />

In summary this means that we can build a <strong>channel</strong> model by defining a set of clusters (in delay-angle<br />

domain) al<strong>on</strong>g with their independent APS at transmitter <strong>and</strong> receiver.<br />

Page 55 (167)


WINNER D5.4 v. 1.4<br />

5. Measurements <strong>and</strong> Literature Review<br />

Our <strong>models</strong> are based <strong>on</strong> 3 pillars, namely existing (spatial) <strong>channel</strong> <strong>models</strong>, new publicati<strong>on</strong>s regarding<br />

all kind of <strong>channel</strong> modelling aspects, <strong>and</strong> finally measurements c<strong>on</strong>ducted within the work of WINNER.<br />

5.1 Measurement <strong>system</strong>s<br />

5.1.1 Principle of <strong>channel</strong> sounding<br />

The operati<strong>on</strong> principle of a <strong>channel</strong> sounder is to transmit a known signal using <strong>on</strong>e antenna in <strong>on</strong>e place<br />

<strong>and</strong> to receive it using another antenna in another place. The operati<strong>on</strong> is thus very similar to that of a<br />

vector network analyzer. The key difference is that the transmitter <strong>and</strong> receiver are separate units. For this<br />

reas<strong>on</strong>, both the receiver <strong>and</strong> transmitter must be phase locked to accurate frequency st<strong>and</strong>ards (typically<br />

Rubidium clocks) in order to maintain phase coherence.<br />

In simplest form the <strong>channel</strong> can be sounded by generating a CW RF signal with a signal generator at the<br />

transmitter, <strong>and</strong> mixing it down at the receiver using another generator tuned to the same frequency. The<br />

voltage at the mixer output gives the narrowb<strong>and</strong> radio <strong>channel</strong> as a functi<strong>on</strong> of time. In practice<br />

amplifiers <strong>and</strong> filters are also needed in the <strong>system</strong>. The advantage of this type of <strong>channel</strong> sounder is that<br />

it can be easily built using st<strong>and</strong>ard laboratory equipment. However, the drawback is that the wideb<strong>and</strong><br />

<strong>channel</strong> properties can’t be measured. KTH used this type of <strong>channel</strong> sounder in their WINNER<br />

measurements.<br />

In order to solve the wideb<strong>and</strong> properties of the radio <strong>channel</strong>, also the sounding itself needs to be d<strong>on</strong>e<br />

using a wideb<strong>and</strong> signal. The wideb<strong>and</strong> signal can be generated either using direct signal spreading (used<br />

e.g. in PropSound <strong>and</strong> HUT sounders) or OFDM-type of transmissi<strong>on</strong> (used in RUSK sounder).<br />

Naturally, also the receiver needs to be wideb<strong>and</strong>. The radio <strong>channel</strong> is estimated either in time or<br />

frequency domain by cross-correlating the received signal with a replica of the original transmitted signal.<br />

Typically this is d<strong>on</strong>e numerically after sampling the wideb<strong>and</strong> signal at the output of a vector<br />

demodulator. The result is either the complex impulse resp<strong>on</strong>se or frequency transfer functi<strong>on</strong> of the<br />

<strong>channel</strong> (these two are c<strong>on</strong>nected by a Fourier transform). In additi<strong>on</strong>, the effects of the transmitter <strong>and</strong><br />

receiver need to be removed from the result using e.g. inverse filtering methods.<br />

MIMO <strong>channel</strong> sounding requires transmissi<strong>on</strong> <strong>and</strong> recepti<strong>on</strong> using multiple antennas. In all wideb<strong>and</strong><br />

sounder <strong>system</strong>s used in WINNER measurements this was achieved with a single transmitter – receiver<br />

pair through time multiplexing with synchr<strong>on</strong>ous antenna switches in the transmitter <strong>and</strong> receiver. Using<br />

electr<strong>on</strong>ic RF switches the switching can be made fast enough to capture essentially the same radio<br />

<strong>channel</strong> with all antennas even in mobile measurements. In KTH measurement <strong>system</strong> the MIMO<br />

measurement was d<strong>on</strong>e using four parallel transmitters tuned at slightly different frequencies <strong>and</strong> four<br />

parallel receivers each receiving all transmitted frequency t<strong>on</strong>es.<br />

5.1.2 Channel sounders employed<br />

Four different radio <strong>channel</strong> measurement <strong>system</strong>s were used in the measurement campaigns. The results<br />

obtained with the HUT sounder in campaigns outside WINNER are used as background data in the<br />

<strong>channel</strong> model. The measurement <strong>system</strong>s are listed in Table 5.1.<br />

Partner<br />

Table 5.1: Measurement <strong>system</strong>s used in <strong>channel</strong> measurements.<br />

Measurement<br />

<strong>system</strong> type<br />

Manufacturer<br />

Hyper<strong>link</strong><br />

EBIT Propsound Elektrobit http://www.propsim.com/<br />

NOK<br />

Propsound<br />

+ HUT antennas<br />

Elektrobit<br />

http://www.propsim.com/<br />

TUI RUSK Medav http://www.<strong>channel</strong>sounder.de/<br />

KTH KTH specific N/A N/A<br />

HUT HUT specific N/A http://www.tkk.fi/Units/Radio/research/<br />

rf_applicati<strong>on</strong>s_in_mobile_communicati<strong>on</strong>/radio_<strong>channel</strong>/<br />

radio_<strong>channel</strong>_sounder.htm<br />

The PropSound TM <strong>and</strong> RUSK <strong>channel</strong> sounders are commercial wideb<strong>and</strong> radio <strong>channel</strong> measurements<br />

<strong>system</strong>s, while the HUT wideb<strong>and</strong> <strong>channel</strong> sounder is mainly self-made. The narrowb<strong>and</strong> measurement<br />

Page 56 (167)


WINNER D5.4 v. 1.4<br />

<strong>system</strong> used by KTH is self-made based <strong>on</strong> commercial laboratory equipment. All measurements <strong>system</strong>s<br />

are described in more detail below. For further informati<strong>on</strong> the reader is advised to c<strong>on</strong>sult the web pages<br />

under the <strong>link</strong>s given in Table 5.1.<br />

5.1.2.1 PropSound <strong>channel</strong> sounder<br />

Figure 5.1 presents a schematic diagram of PropSound <strong>channel</strong> sounder. The sounder <strong>system</strong> c<strong>on</strong>sists of<br />

separate transmitter <strong>and</strong> receiver parts, which both are c<strong>on</strong>trolled through a pers<strong>on</strong>al computer. Both TX<br />

<strong>and</strong> RX are able to c<strong>on</strong>trol RF switches, which are synchr<strong>on</strong>ized in order to make time-multiplexed<br />

MIMO measurements.<br />

Figure 5.1: Block diagram of PropSound multidimensi<strong>on</strong>al sounder.<br />

NOK <strong>and</strong> EBIT both have a PropSound <strong>channel</strong> sounder. However, the <strong>system</strong>s have somewhat different<br />

characteristics. The key features of the EBIT PropSound <strong>channel</strong> sounder are listed in Table 5.2, while the<br />

NOK PropSound features are listed in Table 5.3.<br />

Table 5.2: Key features of EBIT PropSound <strong>channel</strong> sounder.<br />

Feature Value Note<br />

Frequency range<br />

1.7 GHz – 2.7 GHz<br />

Depending <strong>on</strong> RF unit<br />

Max. number of transmitter <strong>channel</strong>s<br />

(Tx-antennas)<br />

Max. number of receiver <strong>channel</strong>s (Rxantennas)<br />

Maximum RF power in antenna input<br />

(after TX switch)<br />

Chip rate<br />

Sequence length (defines maximum<br />

excess delay)<br />

5.1 GHz – 5.9 GHz<br />

64 Limited by the switch<br />

64 Limited by the switch<br />

26 dBm<br />

Up to 100 Mchips/s<br />

31 – 4096<br />

Propagati<strong>on</strong> delay resoluti<strong>on</strong> 10 ... 10000 ns ( = 1 / chip rate) With ISIS resoluti<strong>on</strong><br />

significantly better<br />

Max. meas. data storage rate<br />

27 Mb/sec<br />

Table 5.3: Key features of NOK PropSound sounder.<br />

Feature Value Note<br />

Frequency range 1.8−2.1GHz, 2.1−2.5GHz, depending <strong>on</strong> RF unit<br />

Page 57 (167)


WINNER D5.4 v. 1.4<br />

Max. number of transmitter <strong>channel</strong>s<br />

(Tx-antennas)<br />

Max. number of receiver <strong>channel</strong>s (Rxantennas)<br />

Maximum RF power in antenna input<br />

(after TX switch)<br />

Max. zero-to-zero b<strong>and</strong>width<br />

Chip rate<br />

Sequence length (defines maximum<br />

excess delay)<br />

Sampling frequency<br />

Propagati<strong>on</strong> delay resoluti<strong>on</strong><br />

5.15−5.35 GHz, 5.725−5.875 GHz<br />

64 limited by NOK switch<br />

64 limited by NOK switch<br />

+24 dBm +35dBm with external<br />

power amplifier<br />

200 MHz<br />

0.5 ... 100 MHz<br />

31 ... 4096 chips<br />

1 ... 2000 MHz<br />

10 ... 10000 ns ( = 1 / chip rate)<br />

RF sensitivity -87 dBm (@100 MHz<br />

b<strong>and</strong>width)<br />

Max. meas. data storage rate<br />

2 x 20 Mbytes/s<br />

5.1.2.2 Medav <strong>channel</strong> sounder<br />

Figure 5.2 shows the principal structure of the Medav RUSK Channel Sounder [Medav]. Furthermore,<br />

Table 5.4 summarizes the technical key features of the sounder setup, which were used during the MIMO<br />

measurements.<br />

Mobile<br />

Transmitter<br />

RF down c<strong>on</strong>verter<br />

Receiver<br />

Digital demodulator<br />

Display<br />

Arbitrary<br />

waveform<br />

generator<br />

Local<br />

Oscillator<br />

Mux<br />

Tx<br />

array<br />

Rx array Mux<br />

~<br />

Local<br />

Oscillator<br />

~<br />

A<br />

D<br />

8 bit<br />

640 MHz<br />

PC<br />

hard disc<br />

array<br />

Rubidium<br />

frequency<br />

reference<br />

Positi<strong>on</strong> (GPS, wheel sensors)<br />

Rubidium<br />

frequency<br />

reference<br />

Positi<strong>on</strong> (data telemetry)<br />

Figure 5.2: Block diagram of the RUSK Channel Sounder from Medav.<br />

Table 5.4: Key features of the Medav RUSK <strong>channel</strong> sounder.<br />

Feature Value Note<br />

Frequency range 1.2 ….2.x GHz; 5…6 GHz customized<br />

Max. number of transmitter <strong>channel</strong>s<br />

(Tx-antennas)<br />

Max. number of receiver <strong>channel</strong>s (Rxantennas)<br />

Maximum RF power in antenna input<br />

(after TX switch)<br />

Test signal<br />

Sequence length (defines maximum<br />

excess delay)<br />

Up to 2 16 <strong>channel</strong>s, e.g.,<br />

also depending <strong>on</strong><br />

16 transmit <strong>and</strong><br />

switches<br />

64 receive antennas also depending <strong>on</strong><br />

switches<br />

2 W also depending <strong>on</strong><br />

switches<br />

Multi Carrier Spread Spectrum<br />

Signal (MCSSS)<br />

256 – 8192 spectral lines depending <strong>on</strong> IR length<br />

Page 58 (167)


WINNER D5.4 v. 1.4<br />

Sampling frequency 320 MHz at Tx <strong>and</strong> Rx<br />

Propagati<strong>on</strong> delay resoluti<strong>on</strong><br />

4.17 ns (1/b<strong>and</strong>width)<br />

Impulse resp<strong>on</strong>se length 0.8 µs – 25.6 µs adjustable to the<br />

envir<strong>on</strong>ment<br />

RF sensitivity<br />

Max. meas. data storage rate<br />

-90 dBm<br />

2x160 Mbytes/s<br />

The RUSK Channel Sounder uses an excitati<strong>on</strong> signal c<strong>on</strong>cept, which is known as the “periodic multisine<br />

signal”. This approach is well known from frequency domain <strong>system</strong> identificati<strong>on</strong> in measurement<br />

engineering. In communicati<strong>on</strong> engineering teRMS this signal may be called a multicarrier spread<br />

spectrum signal (MCSSS). Regarding the overall spectral shape, the main advantage of multicarrier<br />

spread spectrum signal (MCSSS) is its “brickwall-type” shape which allows c<strong>on</strong>centrating the signal<br />

energy exactly to the b<strong>and</strong> of interest. This can even be multiple b<strong>and</strong>s when spectral magnitudes are set<br />

to zero. One example applicati<strong>on</strong> is FDD sounding, which means that the sounder simultaneously excites<br />

both the up- <strong>and</strong> the down-<strong>link</strong> b<strong>and</strong>. Figure 5.3 presents the MCSSS in time (top row, left) <strong>and</strong> frequency<br />

domain (top row, right).<br />

Figure 5.3: Broadb<strong>and</strong> multicarrier spread spectrum signal (MCSSS) magnitude in the time <strong>and</strong><br />

frequency domain (top row) <strong>and</strong> estimated CIR <strong>and</strong> received signal spectrum (bottom row).<br />

In case of multipath transmissi<strong>on</strong>, the power spectrum of the received signal is shaped by frequency<br />

selective fading as shown for example in Figure 5.3 (bottom row, right). Furthermore the impulse<br />

resp<strong>on</strong>se (bottom row, left), which would result from inverse Fourier transform of frequency resp<strong>on</strong>se, is<br />

shown in the same figure.<br />

A MIMO <strong>channel</strong> sounder measures the <strong>channel</strong> resp<strong>on</strong>se matrix between all M Tx antennas at the transmit<br />

side <strong>and</strong> all M Rx antennas at the receiver side. This could be carried out by applying a parallel multiple<br />

<strong>channel</strong> transmitter <strong>and</strong> receiver. However, true parallel <strong>system</strong>s not <strong>on</strong>ly are extremely expensive. They<br />

are also inflexible (when c<strong>on</strong>sidering changing the number of antenna <strong>channel</strong>s) <strong>and</strong> susceptible to phase<br />

drift errors. Also parallel operati<strong>on</strong> of the transmitter <strong>channel</strong>s would cause specific problems since the<br />

M Tx transmitted signals have to be separated at the receiver. A much more suitable sounder architecture is<br />

based <strong>on</strong> switched antenna access. A switched antenna sounder c<strong>on</strong>tains <strong>on</strong>ly <strong>on</strong>e physical transmitter <strong>and</strong><br />

receiver <strong>channel</strong>. Only the antennas <strong>and</strong> the switching <strong>channel</strong>s are parallel. This reduces the sensitivity to<br />

<strong>channel</strong> imbalance.<br />

Figure 5.4 shows the switching time frame of a sequential MIMO sounder using antenna arrays at both<br />

sides of the <strong>link</strong>, which is applied in the RUSK MIMO Channel Sounder. Any rectangle block in the<br />

Page 59 (167)


WINNER D5.4 v. 1.4<br />

figure represents <strong>on</strong>e period of the transmit/receive signal. Synchr<strong>on</strong>ous switching at the Rx <strong>and</strong> Tx is<br />

required in order to clearly assign the received signal periods to any input-output combinati<strong>on</strong> of the<br />

<strong>channel</strong> matrix. Timing <strong>and</strong> switching frame synchr<strong>on</strong>izati<strong>on</strong> is established during an initial<br />

synchr<strong>on</strong>izati<strong>on</strong> process prior to measurement data recording <strong>and</strong> must be maintained over the complete<br />

measurement time even in the case of remote operati<strong>on</strong> of Tx <strong>and</strong> Rx. This is accomplished by rubidium<br />

reference oscillators at both Rx <strong>and</strong> Tx.<br />

M Tx<br />

M Rx<br />

Tx<br />

Rx<br />

Tx switching sequence<br />

Rx switching sequence<br />

Tx<br />

<strong>channel</strong><br />

Rx<br />

1<br />

2<br />

3<br />

1<br />

2<br />

3<br />

4<br />

t p = T x Signal period = τ max = maximum path excess delay<br />

time<br />

t s = τ max 2 M Tx M Rx = total snapshot time durati<strong>on</strong><br />

Figure 5.4: MIMO sounder switching time frame.<br />

5.1.2.3 KTH measurement <strong>system</strong><br />

The KTH measurement <strong>system</strong> is based <strong>on</strong> four parallel transmitters tuned to slightly different<br />

frequencies. Figure 5.5 presents a block diagram of the transmitter.<br />

∑<br />

Reference<br />

Antenna<br />

TX module 1<br />

TX module 2<br />

splitter<br />

splitter<br />

TX antenna 1<br />

TX antenna 2<br />

Laptop PC<br />

TX module 3<br />

splitter<br />

TX antenna 3<br />

TX module 4<br />

splitter<br />

TX antenna 4<br />

Figure 5.5: Block diagram of the transmitter chains.<br />

Each base stati<strong>on</strong> c<strong>on</strong>tains four parallel receiver chains as shown in Figure 5.6. The combined RF signal<br />

from each vertical array is input to its corresp<strong>on</strong>ding RX module. The output signal of the RX module is<br />

fed to an analog-to-digital c<strong>on</strong>verter (ADC). The ADC collects 40000 samples from each signal chain per<br />

sec<strong>on</strong>d <strong>and</strong> writes the sampled data to the hard disk of a computer. These raw data samples are processed<br />

<strong>and</strong> analyzed later by software in an off-line fashi<strong>on</strong>.<br />

Page 60 (167)


WINNER D5.4 v. 1.4<br />

RX 1<br />

RX module 1<br />

RX 2<br />

RX module 2<br />

RX 3<br />

ADC<br />

PC<br />

RX module 3<br />

RX 4<br />

RX module 4<br />

Figure 5.6: Block diagram of the receiver chains.<br />

The frequencies are first estimated per segments of several sec<strong>on</strong>ds following which <strong>channel</strong> estimati<strong>on</strong> is<br />

performed. Both the transmitter <strong>and</strong> both base-stati<strong>on</strong> are c<strong>on</strong>nected to GPS. The transmitter is <strong>on</strong>-off<br />

keyed in order to facilitate fine temporal synchr<strong>on</strong>izati<strong>on</strong> at the base-stati<strong>on</strong>s.<br />

5.1.2.4 HUT <strong>channel</strong> sounder<br />

The working principle of the HUT <strong>channel</strong> sounder is almost identical to the Elektrobit PropSound<br />

<strong>channel</strong> sounder. The TX unit is a st<strong>and</strong>al<strong>on</strong>e unit synchr<strong>on</strong>ized with the RX unit in order to synchr<strong>on</strong>ize<br />

the <strong>channel</strong> switching in both TX <strong>and</strong> RX end before performing time-multiplexed MIMO measurements.<br />

The RX unit is used at the mobile end <strong>and</strong> is c<strong>on</strong>trolled through a pers<strong>on</strong>al computer. The sounder has 2,<br />

5 <strong>and</strong> 60 GHz frequency ranges. The key features of the HUT 5 GHz <strong>channel</strong> sounder, whose results are<br />

used in WINNER, are presented in Table 5.5.<br />

Table 5.5: Key features of HUT 5 GHz <strong>channel</strong> sounder.<br />

Feature Value Note<br />

Frequency range<br />

5.25– 5.35 GHz<br />

Max. number of transmitter <strong>channel</strong>s<br />

(Tx-antennas)<br />

Max. number of receiver <strong>channel</strong>s (Rxantennas)<br />

Maximum RF power in antenna input<br />

(after TX switch)<br />

Chip rate<br />

Sequence length (defines maximum<br />

excess delay)<br />

Propagati<strong>on</strong> delay resoluti<strong>on</strong><br />

Max. meas. data storage rate<br />

32 Limited by the switch<br />

32 Limited by the switch<br />

4 W<br />

Up to 60 Mchips/s<br />

255<br />

17 ns ( = 1 / chip rate)<br />

2*20 Mb/sec<br />

5.2 Measurement campaigns<br />

Several measurement campaigns have been designed <strong>and</strong> performed in WINNER WP5 during 2004 to<br />

obtain parameters for the WINNER <strong>channel</strong> <strong>models</strong>. The following secti<strong>on</strong>s give overviews of the<br />

campaigns.<br />

5.2.1 Scenario A1<br />

5.2.1.1 EBIT campaign<br />

Measurements c<strong>on</strong>ducted during 2004 for A1 were performed at two centre-frequencies, 2.45 <strong>and</strong> 5.25<br />

GHz at Elektrobit premises in Oulu, Finl<strong>and</strong>. The measurement results were included in the deliverable<br />

D5.3 [D5.3].<br />

Page 61 (167)


WINNER D5.4 v. 1.4<br />

In 2005 a new series of measurements was performed at two locati<strong>on</strong>s, Oulu University main building<br />

<strong>and</strong> Oulu University wing building Tietotalo. Two different buildings were measured at 5.25 GHz with<br />

100 MHz b<strong>and</strong>width. In the two buildings, more than 8 BSs were chosen with many different routes.<br />

Tietotalo is a typical office envir<strong>on</strong>ment, the corridors of which are narrow with widths around 1.8<br />

meters. In the university main building, the corridors have different width, the widest is around 3.5<br />

meters. In the room measurements at the university main building, the room size is very close to 10 m by<br />

10 m, as in the definiti<strong>on</strong> of the scenario A1. In Tietotalo the sizes of the measured rooms were<br />

comparable to 10 m by 10 m. The largest combined sets of IRs for deriving <strong>channel</strong> <strong>models</strong> <strong>and</strong><br />

parameters are over 55000.<br />

The indoor envir<strong>on</strong>ments here are divided into the following 4 cases:<br />

(1) Corridor-corridor LOS (c-c LOS): both BS <strong>and</strong> MS were placed at the corridors.<br />

(2) Room-corridor <strong>and</strong> corridor-room NLOS (r-c NLOS): BS in a room, MS in an adjacent corridor<br />

vice versa.<br />

(3) Room-room LOS/OLOS (r-r LOS/OLOS): both BS <strong>and</strong> MS in the rooms.<br />

(4) Corridor-corridor NLOS (c-c NLOS)<br />

In the analysis, 100 IRs for a drop are used (about 1.4 meters), if no window length is menti<strong>on</strong>ed.<br />

The following definiti<strong>on</strong>s are used: paths = peaks = ZDSC, Noise cut <strong>level</strong>s used in the analysis: For<br />

LOS: 28 ~ 35 dB. For NLOS: 15 dB ~ 30 dB.<br />

5.2.2 Scenario B1<br />

5.2.2.1 HUT background campaign<br />

These measurements were performed outside the WINNER project by HUT <strong>and</strong> they were used as<br />

background data for Scenario B1. The sounder was the HUT sounder with centre-frequency 5.3 GHz <strong>and</strong><br />

b<strong>and</strong>width (chip rate) 60 MHz. The measurements in the micro-cell envir<strong>on</strong>ment included both LOS <strong>and</strong><br />

NLOS routes in a fairly regular rectangle street grid with 4-5-storey buildings. In the measurements the<br />

fixed Tx antenna was a 16-element dual-polarised planar patch antenna array <strong>and</strong> the mobile Rx antenna a<br />

semispherical antenna with 15 dual-polarized patch elements. The antennas are shown in Figure 5.9.<br />

5.2.3 Scenario B3<br />

5.2.3.1 TUI campaign<br />

These measurements were d<strong>on</strong>e with the RUSK ATM MIMO sounder by Medav [Medav]. The centrefrequency<br />

in the measurements was 5.2 GHz <strong>and</strong> the b<strong>and</strong>width was 120 MHz. The indoor envir<strong>on</strong>ment<br />

c<strong>on</strong>sisted of a university lecture hall (c<strong>on</strong>ference hall) <strong>and</strong> the entrance area (foyer) next to it. The foyer is<br />

characterized by a 2 floor open envir<strong>on</strong>ment, with dimensi<strong>on</strong> of 15m x 30m x 8m. The c<strong>on</strong>ference hall is<br />

a typical lecture hall envir<strong>on</strong>ment with slowly elevated sitting rows; the dimensi<strong>on</strong> is 30m x 35m x 15m.<br />

In all measurement cases the Tx was moving <strong>on</strong> the same track, whereby the positi<strong>on</strong> of the Rx was<br />

changed in the hall. Within the c<strong>on</strong>ference hall scenario two different main setups where used, namely<br />

LOS <strong>and</strong> NLOS/OLOS, where in the latter the LOS was obstructed by an absorber mat.<br />

The pictures of high-resoluti<strong>on</strong> antennas used in TUI measurement campaign are shown in Figure 5.7.<br />

Figure 5.7: High-resoluti<strong>on</strong> antennas used in TUI campaigns. Left: 16-element uniform circular<br />

array, right: dual-polarized 8-element uniform linear array.<br />

Page 62 (167)


WINNER D5.4 v. 1.4<br />

5.2.4 Scenario C1<br />

5.2.4.1 EBIT campaign<br />

Measurements were c<strong>on</strong>ducted during 2004 for the suburban scenario C1 at the centre-frequency 5.25<br />

GHz. Measurements by Elektrobit were performed in Heinäpää relatively near to Oulu centre in an area,<br />

where the houses are lower than in the centre of the town, with some parking lots, parks <strong>and</strong> trees al<strong>on</strong>g<br />

the streets in between the houses. The height of the houses varied typically from 3 to 6 stories. Due to the<br />

measurement routes mainly LOS c<strong>on</strong>diti<strong>on</strong>s were encountered. It should be noted that this envir<strong>on</strong>ment<br />

was clearly different from the envir<strong>on</strong>ment, where Nokia performed their measurements for C1 scenario.<br />

5.2.4.2 NOK <strong>and</strong> HUT campaign<br />

Measurements were d<strong>on</strong>e using the NRC/RAD/EDM radio <strong>channel</strong> sounder (Elektrobit PropSound). The<br />

centre-frequency used in the measurements was 5.3 GHz, <strong>and</strong> the chip rate was 100 MHz. The<br />

measurement campaign was targeted to support the following WINNER <strong>channel</strong> model development<br />

purposes:<br />

• Creati<strong>on</strong> of a path-loss model<br />

• MIMO performance evaluati<strong>on</strong><br />

• Doppler spectra characterizati<strong>on</strong><br />

• Directi<strong>on</strong>-of-departure (DoD) <strong>and</strong> directi<strong>on</strong>-of-arrival (DoA) analysis<br />

Altogether three base stati<strong>on</strong> sites (some of them with two sectors) were measured in two different types<br />

of suburban envir<strong>on</strong>ments in Helsinki, Finl<strong>and</strong>. The first envir<strong>on</strong>ment was a residential area occupied<br />

mostly with low, <strong>on</strong>e or two floor single family or terraced houses with occasi<strong>on</strong>al open areas, such as<br />

playgrounds or gardens, in between. The sec<strong>on</strong>d envir<strong>on</strong>ment was rather loosely built residential area<br />

with higher 3-4 floor high apartment buildings. Open areas, play grounds <strong>and</strong> small forest secti<strong>on</strong>s were<br />

in between the apartment buildings, as well as occasi<strong>on</strong>al lower 1-2 floor buildings.<br />

A truck-operated crane was used to lift up the transmitter. The receiving antenna (user mockup) was<br />

located <strong>on</strong> top of the van, <strong>and</strong> the receiver sounder unit was inside the van. In this case several c<strong>on</strong>tinuous<br />

routes of lengths 50…800 meters each were measured. For DoA/DoD measurements the spherical array<br />

<strong>and</strong> the sounder receiver unit were set to a battery-powered trolley to allow slow enough moving speeds.<br />

Due to huge amounts of data in this case shorter routes of 10-20 meters were measured in dozens of<br />

different locati<strong>on</strong>s within the sector. Some antennas used in Nokia <strong>and</strong> HUT campaign are shown in<br />

Figure 5.8.<br />

Figure 5.8: Left: spherical antenna array (HUT Radio Laboratory) used in the mobile end of the<br />

radio <strong>link</strong> fro DOA characterizati<strong>on</strong>. Right: Planar dual-polarized array (HUT Radio Laboratory)<br />

used as a base stati<strong>on</strong> antenna.<br />

5.2.5 Scenario C2<br />

5.2.5.1 KTH measurements<br />

A measurement campaign was c<strong>on</strong>ducted in the Vasastan area of Stockholm city. The area can be<br />

characterized as a typical European urban with mostly six to eight stories high st<strong>on</strong>e buildings <strong>and</strong><br />

occasi<strong>on</strong>al higher buildings <strong>and</strong> church towers. The measurements were d<strong>on</strong>e in up<strong>link</strong> directi<strong>on</strong> with a<br />

mobile-stati<strong>on</strong> transmitting four separate CWs <strong>on</strong> four separate antennas with a frequency separati<strong>on</strong> of<br />

Page 63 (167)


WINNER D5.4 v. 1.4<br />

approximately <strong>on</strong>e kilohertz <strong>and</strong> a nominal centre-frequency of 1766.6MHz. The four MS antennas are<br />

slanted patch with a half-power beam-width of 80-degree, which point in four different directi<strong>on</strong>s offset<br />

90-degrees from each other, see Figure 5.9. The CW of the transmit antennas are not frequency locked,<br />

resulting in an unknown phase-offset in the four vector <strong>channel</strong>s between the mobile-stati<strong>on</strong> <strong>and</strong> <strong>on</strong>e<br />

base-stati<strong>on</strong> will result. The signals transmitted by the four antenna at MS are received by two basestati<strong>on</strong>s,<br />

down-c<strong>on</strong>verted to complex I&Q base-b<strong>and</strong> <strong>and</strong> saved <strong>on</strong> a hard-disc. One of the base stati<strong>on</strong>s<br />

(Kårhuset-A) has a four-element antenna array <strong>and</strong> is placed <strong>on</strong> a roof barely above the average building<br />

height in its sector of coverage. The other base-stati<strong>on</strong> has two four antenna arrays c<strong>on</strong>nected (Vanadis-B<br />

<strong>and</strong> Vanadis-C) <strong>and</strong> is placed <strong>on</strong> a roof some ten meters above the average building height. Views from<br />

basestati<strong>on</strong> are shown in Figure 5.10. The two arrays are located <strong>on</strong> different edges of the same building<br />

some 20-meters from each other <strong>and</strong> offset 120-degrees in pointing directi<strong>on</strong>. In fr<strong>on</strong>t of Vanadis-B are<br />

some trees, which may have an impact <strong>on</strong> the propagati<strong>on</strong>. The distance between the two sites is 900<br />

meters <strong>and</strong> the measurements were d<strong>on</strong>e in with mobile located in the area between the two BSs. The<br />

path-loss slope from Kårhuset was estimated to be around 40-45dB/dec while it is 25-30dB/dec from<br />

Vanadis. The total measurement route covers about 15km of mobile trajectory.<br />

Figure 5.9: Left: MS antennas, right: BS antenna array.<br />

Figure 5.10: Left: View from “Kårhuset-A” basestati<strong>on</strong>. Right: View from “Vanadis-B” BS.<br />

5.2.6 Scenario D1<br />

5.2.6.1 EBIT campaign<br />

Measurements c<strong>on</strong>ducted during 2004 for D1 were performed at two centre-frequencies, 2.45 <strong>and</strong> 5.25<br />

GHz in Tyrnävä, a small village near Oulu, <strong>and</strong> its surroundings. The measurement results were included<br />

in the deliverable D5.3 [D5.3].<br />

In the year 2005 three new BS locati<strong>on</strong>s were measured at 5.25 GHz with 100 MHz b<strong>and</strong>width. In<br />

additi<strong>on</strong>, PL measurements were c<strong>on</strong>ducted with a smaller b<strong>and</strong>-width 10 MHz to increase the sensitivity<br />

of the receiver equipment. This arrangement allowed us to obtain path losses up to 10 km distance.<br />

At the same locati<strong>on</strong>s some measurements at 2.45 GHz were performed to investigate the effect of centrefrequency<br />

<strong>on</strong> the path loss <strong>and</strong> other <strong>channel</strong> parameters. For practical reas<strong>on</strong>s the measurements at 2.45<br />

Page 64 (167)


WINNER D5.4 v. 1.4<br />

GHz were c<strong>on</strong>ducted in a smaller scope with fewer routes. However, the routes used were the same as the<br />

routes used for the 5.25 GHz measurements.<br />

Map of the measurement envir<strong>on</strong>ment <strong>and</strong> the BS locati<strong>on</strong>s are shown in Figure 5.11.<br />

Figure 5.11 Base stati<strong>on</strong> locati<strong>on</strong>s for the Tyrnävä measurements in the summer 2005.<br />

5.2.7 Measurement summary<br />

Table 5.6 presents a summary of WINNER measurement campaigns <strong>and</strong> shows which scenarios are<br />

supported by measurement data.<br />

Page 65 (167)


WINNER D5.4 v. 1.4<br />

Table 5.6: Summary of WINNER <strong>channel</strong> measurements.<br />

Corresp<strong>on</strong>ding test<br />

scenario<br />

Partner Measurement type Locati<strong>on</strong> Envir<strong>on</strong>ment descripti<strong>on</strong> Center<br />

frequency<br />

A1 (Indoor) EBIT Indoor (office<br />

building)<br />

Oulu, Finl<strong>and</strong> Typical office, cubicles, corridors 5.25 GHz 100 MHz (chip<br />

rate)<br />

B<strong>and</strong>width TX power BTS<br />

height<br />

MS<br />

height<br />

MS speed Max MS-BS<br />

distances<br />

# of BTS<br />

antennas<br />

# of MS<br />

antennas<br />

BTS antenna type MS antenna<br />

type<br />

23 dBm 2.1 m 0.9 m ~1 m/s < 100 m 32, 1 or 16 52, 1 or 18 Dual-polarized 4x4<br />

UPA, dipole<br />

Dual-polarized<br />

2x9 ODA, dipole<br />

A1 (Indoor) EBIT Indoor (office<br />

building)<br />

Oulu, Finl<strong>and</strong> Typical office, cubicles, corridors 2.45 GHz 100 MHz (chip<br />

rate)<br />

23 dBm 2.1 m 0.9 m ~1 m/s < 100 m 32, 1 or 16 56, 1 or 18 Dual-polarized 4x4<br />

UPA, dipole<br />

Dual-polarized<br />

3x8 ODA, dipole<br />

B1 LOS (urban) HUT Microcell LOS Helsinki,<br />

Finl<strong>and</strong><br />

B1 NLOS (urban) HUT Microcell NLOS Helsinki,<br />

Finl<strong>and</strong><br />

B1 / D1<br />

(suburban/rural)<br />

Microcellular city center<br />

measurements with LOS<br />

Microcellular city center<br />

measurements with NLOS<br />

TUI Outdoor, Hot spot Ulm, Germany suburban/rural area with highway<br />

bridge, Hot spot (public access),<br />

B3 (Indoor) TUI Indoor (large<br />

lecture hall <strong>and</strong> )<br />

Ilmenau,<br />

Germany<br />

Car to Bridge<br />

5.3 GHz 60 MHz (chip<br />

rate)<br />

5.3 GHz 60 MHz (chip<br />

rate)<br />

36 dBm at<br />

antenna<br />

input<br />

36 dBm at<br />

antenna<br />

input<br />

5.2 GHz 120 MHz 27 dBm at<br />

TX switch<br />

Auditorium of TUI, large indoor hall 5.2 GHz 120 MHz 27 dBm at<br />

TX switch<br />

10 m 1.6 m 0.2 m/s 400 m 16 dualpolarized<br />

10 m 1.6 m 0.2 m/s 180 m 16 dualpolarized<br />

15 dual-polarized<br />

=> <strong>channel</strong>s<br />

32x30<br />

15 dual-polarized<br />

=> <strong>channel</strong>s<br />

32x30<br />

4x4 planar array<br />

with +/-45<br />

polarizati<strong>on</strong>s<br />

4x4 planar array<br />

with +/-45<br />

polarizati<strong>on</strong>s<br />

~5 m 2.10 m up to ~ 10 m/s 100-150m 8x2 16 ULA of 8 dualpolarized<br />

(V/H)<br />

patch elements<br />

3.95 m <strong>and</strong><br />

(3 m +<br />

3.65 m)<br />

1.10 m fixed positi<strong>on</strong>s,<br />

~1m/s UCA16<br />

at a track<br />

30m 8x2 16 ULA of 8 dualpolarized<br />

(V/H)<br />

patch elements<br />

spherical with<br />

H/V polarizati<strong>on</strong>s<br />

spherical with<br />

H/V polarizati<strong>on</strong>s<br />

UCA of 16,<br />

vertical polar.<br />

UCA of 16,<br />

vertical polar.<br />

C1 metropol<br />

(suburban)<br />

C1 metropol<br />

(suburban)<br />

EBIT Suburban macro Oulu, Finl<strong>and</strong> Suburban residential: max. 2-<br />

storey houses, BTS antenna above<br />

rooftop <strong>level</strong>, outdoors LOS<br />

obstructed mainly <strong>on</strong>ly by<br />

NOK &<br />

HUT<br />

Suburban macro Helsinki,<br />

Finl<strong>and</strong><br />

(Paloheinä,<br />

Munkkivuori)<br />

C2 metropol (urban) KTH Typical (European)<br />

Urban.<br />

vegetati<strong>on</strong><br />

Suburban residential: max. 2-store<br />

houses, BTS antenna above<br />

rooftop <strong>level</strong>, outdoors LOS<br />

obstructed mainly <strong>on</strong>ly by<br />

vegetati<strong>on</strong><br />

Stockholm City Urban macrocell. DCS1800<br />

up<strong>link</strong><br />

5.25 GHz 100 MHz (chip<br />

rate)<br />

5.3 GHz 100 MHz (chip<br />

rate)<br />

23 dBm 11.7m / 7.6<br />

m<br />

39 dBm (at<br />

antenna<br />

input)<br />

CW 18 dBm /<br />

antenna<br />

~25 m 2 m (<strong>on</strong><br />

top of a<br />

van)<br />

some 10 m<br />

above<br />

rooftop &<br />

just above<br />

rooftop<br />

1.8 m ~3 m/s 1 km 32, 1 or 16 52, 1 or 18 Dual-polarized 4x4<br />

UPA, dipole<br />

~1 m/s<br />

(pedestrian,<br />

with trolley);<br />

~10 m/s<br />

(vehicular, with<br />

car)<br />

< 1.1-1.3 km,<br />

depending <strong>on</strong><br />

the setup<br />

16 elements<br />

(each 2<br />

polarizati<strong>on</strong>s)<br />

1.8m 0-15m/s 1.2km 4/array (3<br />

arrays)<br />

4 (terminal mock<br />

up) or 14<br />

(spherical array)<br />

=> <strong>channel</strong>s: 8x4,<br />

4x2 (for MIMO,<br />

PL, Doppler),<br />

32x28 (for<br />

DoD/DoD)<br />

Dual-polarized (+/-<br />

45) planar array<br />

4 ULA with -45pol.<br />

slanted 4-stacked<br />

patch elements.<br />

0.55lambda<br />

spacing.<br />

Dual-polarized<br />

2x9 ODA, dipole<br />

terminal mock-up<br />

or spherical array<br />

(with trolley for<br />

DoA/DoD)<br />

Four -45polarized<br />

patches <strong>on</strong> the<br />

sides of a cube.<br />

C4 metropol (urban)<br />

outdoor-to-indoor<br />

KTH Typical (European)<br />

Urban.<br />

Stockholm City Urban macrocell. DCS1800<br />

up<strong>link</strong><br />

CW 18 dBm /<br />

antenna<br />

some 10 m<br />

above<br />

rooftop &<br />

just above<br />

rooftop<br />

1.8m ˜ 0.9m/s 4/array (3<br />

arrays)<br />

4 ULA with -45pol.<br />

slanted 4-stacked<br />

patch elements.<br />

0.55lambda<br />

spacing.<br />

Four -45polarized<br />

patches <strong>on</strong> the<br />

sides of a cube.<br />

D1 (rural) EBIT Wide area rural Tyrnävä,<br />

Finl<strong>and</strong><br />

Countryside, very flat, BTS<br />

antennas at mast, mainly LOS<br />

5.25 GHz 100 MHz (chip<br />

rate)<br />

23 dBm 17.6 m 1.7 m ~3 m/s 1.7 km (at<br />

maximum)<br />

32, 1 or 16 52, 1 or 18 Dual-polarized 4x4<br />

UPA, dipole<br />

Dual-polarized<br />

2x9 ODA, dipole<br />

D1 (rural) EBIT Wide area rural Tyrnävä,<br />

Finl<strong>and</strong><br />

Countryside, very flat, BTS<br />

antennas at mast, mainly LOS<br />

2.45 GHz 100 MHz (chip<br />

rate)<br />

23 dBm 17.6 m 1.7 m ~3 m/s 1.7 km (at<br />

maximum)<br />

32, 1 or 16 56, 1 or 18 Dual-polarized 4x4<br />

UPA, dipole<br />

Dual-polarized<br />

3x8 ODA, dipole<br />

NA NOK &<br />

HUT<br />

Urban ad hoc Helsinki,<br />

Finl<strong>and</strong><br />

(Ruoholahti)<br />

Urban outdoor, 5-6 store houses,<br />

indoor: metro stati<strong>on</strong>, supermarket<br />

NA HUT Indoor ad hoc Espoo, Finl<strong>and</strong> University office buildings, both<br />

older <strong>and</strong> modern<br />

Explanati<strong>on</strong>s:<br />

ULA = Uniform linear array<br />

UPA = Uniform planar array<br />

ODA = Omni directi<strong>on</strong>al array<br />

5.3 GHz 100 MHz (chip<br />

rate)<br />

5.3 GHz 60 MHz (chip<br />

rate)<br />

39 dBm (at<br />

antenna<br />

input)<br />

36 dBm at<br />

antenna<br />

input<br />

1.5 m 1.5 m ~0.83 m/s < 100 m 15 elements<br />

(each 2<br />

polarizati<strong>on</strong>s)<br />

1.6 m 1.6 m 0.2 m/s <br />

15 dual-polarized<br />

=> <strong>channel</strong>s<br />

30x30<br />

Spherical with +/-<br />

45 polarizati<strong>on</strong>s<br />

spherical with +/-45<br />

polarizati<strong>on</strong>s<br />

Spherical with<br />

H/V polarizati<strong>on</strong>s<br />

spherical with<br />

H/V polarizati<strong>on</strong>s<br />

Other informati<strong>on</strong><br />

Separate measurement setups<br />

for pathloss, MIMO<br />

characterizati<strong>on</strong>, Doppler <strong>and</strong><br />

DoA/DoD.<br />

Separate measurement setups<br />

for pathloss, MIMO<br />

characterizati<strong>on</strong>, Doppler <strong>and</strong><br />

DoA/DoD.<br />

Background data of HUT<br />

Background data of HUT<br />

MIMO meas. for DoA <strong>and</strong> DoD<br />

MIMO meas. for DoA <strong>and</strong> DoD<br />

Separate measurement setups<br />

for pathloss, MIMO<br />

characterizati<strong>on</strong>, Doppler <strong>and</strong><br />

DoA/DoD.<br />

Separate measurement setups<br />

for pathloss, MIMO<br />

characterizati<strong>on</strong>, Doppler <strong>and</strong><br />

DoA/DoD. Planar-to-spherical<br />

array measurements with trolley<br />

<strong>on</strong>ly for DoA/DoD.<br />

Signal received from three<br />

sectors with 4-element arrays<br />

distributed <strong>on</strong> two base-stati<strong>on</strong><br />

sites. Four closely spaced CW<br />

frequencies.<br />

Signal received from three<br />

sectors with 4-element arrays<br />

distributed <strong>on</strong> two base-stati<strong>on</strong><br />

sites. Four closely spaced CW<br />

frequencies.<br />

Separate measurement setups<br />

for pathloss, MIMO<br />

characterizati<strong>on</strong>, Doppler <strong>and</strong><br />

DoA/DoD.<br />

Separate measurement setups<br />

for pathloss, MIMO<br />

characterizati<strong>on</strong>, Doppler <strong>and</strong><br />

DoA/DoD.<br />

Mobile peer-to-peer type of<br />

measurement, mainly for<br />

DoA/DoD<br />

Mobile peer-to-peer type of<br />

measurement<br />

Page 66 (167)


WINNER D5.4 v. 1.4<br />

5.3 Descripti<strong>on</strong> of key references<br />

5.4 Results of analysis items<br />

A list of analysis items was made, see [WP5AI]. The analysis items were divided into to categories<br />

namely priority 1 <strong>and</strong> priority 2. In the following two subsecti<strong>on</strong>s, selected set of analysis results from the<br />

campaigns are listed. Additi<strong>on</strong>al results are found in the appendices <strong>and</strong> in the measurement <str<strong>on</strong>g>report</str<strong>on</strong>g>s<br />

[WP5AR].<br />

The measurements data that has been used for extracti<strong>on</strong> of <strong>channel</strong> parameters for scenario B1 are not<br />

WINNER measurement data but can be c<strong>on</strong>sidered as background data from Helsinki University of<br />

Technology, Finl<strong>and</strong>. Sounder frequency is 5.3 GHz with 60 MHz chip rate <strong>and</strong> 120 MHz sampling rate<br />

for each I <strong>and</strong> Q comp<strong>on</strong>ent of the signal. A cutting threshold of 20 dB from the peaks of PDPs is adopted<br />

in the measurement results shown below. However, when there are good propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s like LOS<br />

cases, a higher threshold is used. It is said that propagati<strong>on</strong> is LOS if an unobstructed path between the<br />

locati<strong>on</strong> of the transmitter <strong>and</strong> the locati<strong>on</strong> of the receiver exists.<br />

5.4.1 Path-loss <strong>and</strong> shadow fading<br />

Path-loss (PL) <strong>and</strong> shadow fading (SF) are c<strong>on</strong>sidered to be parameters of the highest priority in <strong>channel</strong><br />

modeling. Path-loss is loss of signal power between transmitter <strong>and</strong> receiver end. SF is the variance of the<br />

PL.<br />

The noise threshold was selected to be -20 dB from the impulse resp<strong>on</strong>se (IR) peak, <strong>and</strong> the IR samples<br />

below that are removed. Data within a small area is averaged in order to remove the effect of fast fading.<br />

The averaging window depended <strong>on</strong> the envir<strong>on</strong>ment <strong>and</strong> the centre-frequency. Spatial averaging is d<strong>on</strong>e<br />

by combining the wideb<strong>and</strong> MIMO matrices in power.<br />

PL at a certain snapshot is calculated from the calibrated IRs as a wideb<strong>and</strong> path loss<br />

PL = −10log<br />

( ∑ h(<br />

)<br />

10<br />

τ i<br />

2<br />

τ<br />

i<br />

) + GT<br />

+ GR<br />

(5.1)<br />

where G T <strong>and</strong> G R are antenna gains at the transmitter <strong>and</strong> at the receiver, respectively.<br />

The path-loss model is derived using linear regressi<strong>on</strong> (LMSE) of the scatter plot of the PL vs. distance<br />

between the transmitter <strong>and</strong> the receiver: PL(d) = A log 10 (d) + B = 10 n log 10 (d) + B, where B is PL<br />

intercept, <strong>and</strong> n is the PL exp<strong>on</strong>ent.<br />

5.4.1.1 Scenario A1<br />

5.4.1.1.1 Measurements in corridor - corridor LOS c<strong>on</strong>diti<strong>on</strong>s<br />

The measurements were c<strong>on</strong>ducted in corridors, the width of which was ranging from 1.5 to 3.5, <strong>and</strong> so<br />

that the BS <strong>and</strong> MS were in the same corridor with a LOS between them. The path-loss curve for the 5.25<br />

GHz centre-frequency in corridor-corridor LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong> can be seen in the Figure 5.12.<br />

Page 67 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.12. Indoor path loss in corridor – corridor LOS c<strong>on</strong>diti<strong>on</strong>s.<br />

The equati<strong>on</strong> for the path loss can now be expressed as<br />

PL(d) = 46.8 + 18.7 log 10 (d), s = 3.1 dB ( 5.2)<br />

where d is the distance <strong>and</strong> s is the st<strong>and</strong>ard deviati<strong>on</strong> of the shadow fading. The equati<strong>on</strong> is valid from 1<br />

m to 200 m.<br />

Very similar results were obtained in the previous measurement campaign [D5.3] for LOS c<strong>on</strong>diti<strong>on</strong>s, but<br />

for limited range. Other similar results, with the path-loss exp<strong>on</strong>ent less than 2, have been discussed in<br />

several references cited in Secti<strong>on</strong> 5.5.<br />

Figure 5.13. Shadow fading distributi<strong>on</strong> in an indoor corridor – corridor LOS envir<strong>on</strong>ment.<br />

5.4.1.1.2 Measurements in room - corridor NLOS c<strong>on</strong>diti<strong>on</strong>s<br />

The path-loss curve for the 5.25 GHz centre-frequency in room – corridor (corridor – room) LOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s can be seen in Figure 5.14. The measurements were performed so that either the<br />

BS was in the corridor <strong>and</strong> the MS in the room al<strong>on</strong>g the corridor or vice versa. The wooden doors from<br />

the corridor to the rooms were closed. It was measured that the attenuati<strong>on</strong> of the door was 4 dB.<br />

Figure 5.14. Indoor path loss in room – corridor LOS c<strong>on</strong>diti<strong>on</strong>s.<br />

Page 68 (167)


WINNER D5.4 v. 1.4<br />

The equati<strong>on</strong> for the path loss can now be expressed as:<br />

PL (d) = 38.8 + 36.8 log 10 (d), s = 3.5 dB (5.3)<br />

where d is the distance <strong>and</strong> s is the st<strong>and</strong>ard deviati<strong>on</strong> of the shadow fading.<br />

The equati<strong>on</strong> is valid from 3 m to 50 m. It is assumed that it can be used until 100 m, although this has<br />

not been verified by measurements. From 1 m to 3 m free-space loss formula should be used.<br />

Quite natural assumpti<strong>on</strong> is that most part of the path between the MS <strong>and</strong> BS the signal propagates in the<br />

corridor. Therefore it is slightly surprising that the path loss is much steeper in this case than in the<br />

corridor – corridor propagati<strong>on</strong>. The reas<strong>on</strong> must be in the mechanism by which the wave couples from<br />

the corridor to the room, or vice versa.<br />

5.4.1.2 Scenario B1<br />

The received power is calculated by summing over all antennas <strong>and</strong> delay bins in power in each<br />

measurement point. The calibrati<strong>on</strong> is d<strong>on</strong>e based <strong>on</strong> back to back measurement in anechoic chamber.<br />

Antenna gains are not included in the received power. The fitting of the parameters is d<strong>on</strong>e using least<br />

square error method. Path-loss model for LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong> is presented in Figure 5.15 with<br />

measurements. Equati<strong>on</strong> (5.4) is the path-loss model for LOS case <strong>and</strong> Equati<strong>on</strong> (5.5) is for NLOS case.<br />

The defined distance in both equati<strong>on</strong>s is the direct distance between transmitter <strong>and</strong> receiver terminals.<br />

Figure 5.15: Path-loss modelling in LOS c<strong>on</strong>diti<strong>on</strong>.<br />

LOS path-loss model:<br />

PL = 41.0 + 22.7⋅<br />

log10 (d), σ = 2.3dB<br />

(5.4)<br />

NLOS path-loss model:<br />

where<br />

( d , d ) PL 10n<br />

( d )<br />

PL = +<br />

, σ = 3.1 dB (5.5)<br />

1 2 0<br />

log10<br />

2<br />

PL = .096d<br />

65 <strong>and</strong> n = −0 .0024 d + 1<br />

2. 8<br />

(5.6)<br />

0<br />

0<br />

1<br />

+<br />

where d 1 <strong>and</strong> d 2 are shown in Figure 5.16. The model in Equati<strong>on</strong> (5.5) is developed in WINNER project<br />

based <strong>on</strong> measurement data fitted in [ZRKV04].<br />

Page 69 (167)


WINNER D5.4 v. 1.4<br />

o<br />

d 1<br />

d 2<br />

d<br />

MS<br />

BS<br />

Figure 5.16: Layout of regular street grids.<br />

5.4.1.3 Scenario B3<br />

The path loss shown in Figure 5.17 was calculated at the centre-frequency of 5.2 GHz <strong>and</strong> b<strong>and</strong>width of<br />

100 MHz (measured 120 MHz) in LOS <strong>and</strong> NLOS/OLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

-45<br />

-45<br />

-50<br />

-50<br />

PL [dB]<br />

PL [dB]<br />

-55<br />

-55<br />

-60<br />

-60<br />

10 1<br />

d [m]<br />

-65<br />

10 1<br />

d [m]<br />

(a)<br />

(b)<br />

Figure 5.17: Path-loss under LOS <strong>and</strong> NLOS/OLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>.<br />

Within the scenario B3 under LOS c<strong>on</strong>diti<strong>on</strong> the equati<strong>on</strong> for the path loss was to be found as:<br />

PL = 36.9 + 13.4 log 10 (d) with s = 1.4 dB for LOS <strong>and</strong><br />

PL = 55.5 + 3.2 log 10 (d) with s = 2.1 dB for NLOS/OLOS,<br />

where d is the distance <strong>and</strong> s is the st<strong>and</strong>ard deviati<strong>on</strong> of the shadow fading.<br />

In Figure 5.18, distributi<strong>on</strong> of the shadow fading <strong>and</strong> SF versus distance are shown.<br />

0.16<br />

0.16<br />

0.14<br />

0.14<br />

0.12<br />

0.12<br />

0.1<br />

0.1<br />

PDF<br />

0.08<br />

PDF<br />

0.08<br />

0.06<br />

0.06<br />

0.04<br />

0.04<br />

0.02<br />

0.02<br />

0<br />

-4 -2 0 2 4<br />

SF [dB]<br />

0<br />

-10 -5 0 5 10<br />

SF [dB]<br />

(a)<br />

(b)<br />

Page 70 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.18: Shadow Fading distributi<strong>on</strong> in LOS (a) <strong>and</strong> NLOS (b) envir<strong>on</strong>ment.<br />

5.4.1.4 Scenario C1<br />

5.4.1.4.1 Measurements in LOS c<strong>on</strong>diti<strong>on</strong>s<br />

The path loss is shown in the figure below for the centre-frequency 5.25 GHz in LOS propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong>s. The measurements have been c<strong>on</strong>ducted in a 100 MHz b<strong>and</strong>width.<br />

Figure 5.19: Path-loss in a suburban envir<strong>on</strong>ment <strong>and</strong> LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Path-loss formula is<br />

PL(d) = 41.6 + 23.8 log 10 (d), s = 4 dB. (5.7)<br />

where d is the distance <strong>and</strong> s is the st<strong>and</strong>ard deviati<strong>on</strong> of the shadow fading.<br />

The equati<strong>on</strong> is valid from 25 m to the break-point value, see Secti<strong>on</strong> 5.6.1. From 1 m to 25 m, free-space<br />

loss formula should be used.<br />

5.4.1.5 Scenario D1<br />

5.4.1.5.1 Measurements in LOS c<strong>on</strong>diti<strong>on</strong>s<br />

In the D1 rural LOS scenario the measurements were c<strong>on</strong>ducted both in the year 2004 [D5.3] <strong>and</strong> in the<br />

year 2005. In the campaign of the year 2005, the LOS measurement was c<strong>on</strong>ducted in slightly different<br />

routes than in the previous year. Measurements were performed for three different BS locati<strong>on</strong>s, each<br />

having several measurement routes. The results are shown in the Figure 5.20.<br />

Page 71 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.20: Rural LOS path loss at 5.25 GHz.<br />

Now the equati<strong>on</strong> for the rural LOS path loss was<br />

PL(d) = 44.6 + 21.5 log 10 (d), s = 4.2 dB. (5.8)<br />

where d is the distance <strong>and</strong> s is the st<strong>and</strong>ard deviati<strong>on</strong> of the shadow fading.<br />

The equati<strong>on</strong> is valid from 30 m to the break-point value, see Secti<strong>on</strong> 5.6.1. From 1 m to 30 m, the freespace<br />

loss formula should be used.<br />

This is nearly equal to the results achieved in the campaign of the previous year. The final path-loss<br />

model for D1 LOS envir<strong>on</strong>ment as well as other path-loss <strong>models</strong> are discussed in Secti<strong>on</strong> 5.6.1.<br />

Path-loss was also investigated for l<strong>on</strong>ger distances in a separate measurement, where the base stati<strong>on</strong><br />

antenna heights were higher, 19 – 25m, <strong>and</strong> a narrower b<strong>and</strong>width was used to achieve better sensitivity.<br />

At the same time path losses for rural LOS <strong>and</strong> NLOS c<strong>on</strong>diti<strong>on</strong>s were investigated. The path-loss result<br />

for the l<strong>on</strong>gest route is shown in the Figure 5.21. The NLOS c<strong>on</strong>diti<strong>on</strong> was defined so that the path loss<br />

exceeded the free-space path loss by 10 dB or more. Three l<strong>on</strong>g routes of this kind were measured <strong>and</strong><br />

averaged to obtain the NLOS <strong>and</strong> over-all path-loss equati<strong>on</strong>s. LOS results were calculated from the<br />

short-range measurements. However, the l<strong>on</strong>g-distance measurements show clearly that very l<strong>on</strong>g LOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s are possible in a flat envir<strong>on</strong>ment like the <strong>on</strong>e near Tyrnävä.<br />

Page 72 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.21: Path-loss in rural scenario <strong>on</strong> the route 3.1.<br />

The average corrected path-loss formula for the over-all path loss in the measurements was<br />

PL(d) = 50.4 + 25.8 log 10 (d), s = 8.4 dB (5.9)<br />

The average corrected path-loss formula for the NLOS path loss in the measurements was<br />

PL(d) = 55.8 + 25.1 log 10 (d), s = 6.7 dB (5.10)<br />

where d is the distance <strong>and</strong> s is the st<strong>and</strong>ard deviati<strong>on</strong> of the shadow fading. Correcti<strong>on</strong> means that the<br />

cutting of the high values was estimated <strong>and</strong> compensated.<br />

It should be noted that the definiti<strong>on</strong> of NLOS was performed according to the power difference of 10 dB<br />

from the free-space loss. Another note is that the noise-floor cuts the weakest signals, so that the highest<br />

path losses were cut as well. It can be assumed, however, that the effect of this limiting is relatively small.<br />

The rural NLOS measurement results were obtained from three routes, which means quite a limited set of<br />

measurements. Therefore the model has been compared with literature <strong>and</strong> adjusted appropriately in<br />

Secti<strong>on</strong> 5.6.1.<br />

5.4.2 LOS probability<br />

LOS probability is the probability that the LOS propagati<strong>on</strong> between the transmitter <strong>and</strong> the receiver<br />

exists.<br />

5.4.2.1 Scenario A1<br />

The probability of line-of-sight (LOS) propagati<strong>on</strong> vs. distance is a functi<strong>on</strong> we denote the p LOS functi<strong>on</strong>.<br />

For scenario A1, this characteristic can be derived analytically because the geometry of the scenario is<br />

known exactly.<br />

A simple ad-hoc fit of the derived p LOS functi<strong>on</strong> is given as:<br />

where x = 1 - log 10 (d / 2.5) / log 10 (100 / 2.5).<br />

5.4.2.2 Scenario B3<br />

p LOS (d) = 1 – (1 – x 3 ) 1/3 * (1 – 5 / 50), (5.11)<br />

In [LUI99] measurement results for the big factory hall envir<strong>on</strong>ment are presented. Length, width <strong>and</strong><br />

height are 90, 30 <strong>and</strong> 10 m respectively. BS height was 8 m <strong>and</strong> MS height was 1.5 m. Average<br />

probability of LOS was 0.5. Up to 10 m distance in such a big halls there is almost always LOS (if Rx is<br />

placed right). Therefore, we propose for the big factory halls, airport <strong>and</strong> train stati<strong>on</strong>s:<br />

⎧1,<br />

d < 10m<br />

P LOS<br />

= ⎨<br />

(5.12)<br />

⎩exp(<br />

−(<br />

d −10) / 45)<br />

where d is in meters. The figure below shows this functi<strong>on</strong> of the distance.<br />

Page 73 (167)


WINNER D5.4 v. 1.4<br />

1<br />

Probability of LOS<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

20 40 60 80 100<br />

distance [m]<br />

Figure 5.22: Probability of LOS in factory halls, airports, <strong>and</strong> train stati<strong>on</strong>s.<br />

In the measurement campaign c<strong>on</strong>ducted at the TUI, measurements of the big lecture hall (c<strong>on</strong>ference<br />

hall) were performed. Probability of LOS was 100%. For the NLOS measurements obstructi<strong>on</strong>s were<br />

artificially made. The str<strong>on</strong>gest reas<strong>on</strong> not to have LOS would be a pers<strong>on</strong> between Tx <strong>and</strong> Rx. Since the<br />

measurements were performed when the hall was empty, the presence of some pers<strong>on</strong>s should be taken<br />

into account. If Rx is placed at the ceiling (or some other appropriate place) LOS will be almost<br />

guarantied. We propose:<br />

⎪<br />

⎧ 1, d < 5m<br />

P LOS<br />

= ⎨ d − 5<br />

1−<br />

,5 m < d < 40 m<br />

⎪⎩ 150<br />

where d is in meters. This functi<strong>on</strong> is presented in the figure below.<br />

(5.13)<br />

1<br />

Probability of LOS<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

5 10 15 20 25 30 35 40<br />

distance [m]<br />

Figure 5.23: Probability of LOS in lecture or c<strong>on</strong>ference halls.<br />

5.4.2.3 Scenario D1<br />

Probability of LOS in the D1 scenario is proposed to be modeled with an exp<strong>on</strong>ential functi<strong>on</strong><br />

1 d<br />

P(<br />

LOS)<br />

= exp( − )<br />

(5.14)<br />

d<br />

0<br />

d 0<br />

where d is the distance between the BS <strong>and</strong> the MS <strong>and</strong> d 0 is a c<strong>on</strong>stant defining the steepness of the<br />

exp<strong>on</strong>ential decay.<br />

Default value for d 0<br />

is proposed to be 1 km. The reas<strong>on</strong> for proposing this model is the following: It is<br />

very near the model for LOS probability defined in [3GPP SCM] at small distances. In additi<strong>on</strong> it does<br />

not go to zero at the cell boundary, so that it can be used in the <strong>system</strong>-<strong>level</strong> modelling of interference.<br />

5.4.3 DS <strong>and</strong> maximum excess-delay distributi<strong>on</strong><br />

RMS delay spread is the square root of the sec<strong>on</strong>d central moment of the PDP normalized to the total<br />

power. Max excess delay is the maximum delay after the first peak in PDP.<br />

Page 74 (167)


WINNER D5.4 v. 1.4<br />

The measured distributi<strong>on</strong>s have been tested against the following theoretic distributi<strong>on</strong>s to find out the<br />

best fit:<br />

1) Log-normal distributi<strong>on</strong><br />

f<br />

( x − v )<br />

2<br />

1 2σ<br />

( x)<br />

=<br />

2πσ<br />

e<br />

2<br />

( x − v)<br />

, F(<br />

x)<br />

= 1−<br />

Q(<br />

)<br />

(5.15)<br />

σ<br />

2) Logistic PDF <strong>and</strong> CDF<br />

x−v<br />

ζ<br />

e<br />

f ( x)<br />

=<br />

⎛<br />

ζ ⎜1<br />

+ e<br />

⎝<br />

x−v<br />

ζ<br />

⎞<br />

⎟<br />

⎠<br />

2<br />

,<br />

F(<br />

x)<br />

= 1−<br />

where ν is the locati<strong>on</strong> parameter <strong>and</strong> scale parameter ζ > 0.<br />

1<br />

1+<br />

e<br />

x−v<br />

ζ<br />

(5.16)<br />

3) Gumbel (log of Weibull distributi<strong>on</strong>) PDF <strong>and</strong> CDF<br />

1 ⎡ x − v x − v ⎤<br />

⎛ x − v ⎞<br />

f ( x)<br />

= exp⎢−<br />

− exp( )<br />

ζ<br />

⎥ , F ( x)<br />

= exp⎜−<br />

exp( − ) ⎟ (5.17)<br />

⎣ ζ ζ ⎦<br />

⎝ ζ ⎠<br />

where ν is the locati<strong>on</strong> parameter <strong>and</strong> scale parameter ζ > 0.<br />

5.4.3.1 Scenario A1<br />

The distributi<strong>on</strong> of the RMS-delay spread was investigated. The 10, 50 <strong>and</strong> 90 % values for the<br />

Cumulative Distributi<strong>on</strong> Functi<strong>on</strong>s of RMS-delay spread are given below for the 5.25 GHz centrefrequency<br />

<strong>and</strong> different LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Table 5.7: RMS delay spreads for A1 indoor scenario.<br />

RMS-DS (ns)<br />

Corri.-Corri. Corri.-Room Room-Room<br />

LOS NLOS NLOS LOS(OLOS)<br />

10% 15.0 13.4 13.5 9.4<br />

Percentile<br />

50% 38.0 25.2 25.1 14.2<br />

90% 75.7 48.6 40.6 18.9<br />

mean 43.0 28.5 26.5 14.2<br />

The distributi<strong>on</strong>s of the maximum excess delay were also calculated for the different envir<strong>on</strong>ments. The<br />

10, 50 <strong>and</strong> 90 % values for the Cumulative Distributi<strong>on</strong> Functi<strong>on</strong>s of the maximum excess delay are given<br />

below for the 5.25 GHz centre-frequency <strong>and</strong> different LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Table 5.8: Maximum excess delays for the A1 indoor scenario.<br />

Corri.-Corri. Corri.-Room Room-Room<br />

Maximum excess delay range LOS NLOS NLOS LOS(OLOS)<br />

(ns)<br />

UO main building 10% 247.1 54.1 97.3 72.5<br />

widest corridor 50% 265.0 107.5 185.0 95.0<br />

3.5 m<br />

90% 624.4 249.7 255.0 130.0<br />

room size (10*10m) mean 349.0 135.0 181.8 98.1<br />

Page 75 (167)


WINNER D5.4 v. 1.4<br />

Cumulative Distributi<strong>on</strong> Functi<strong>on</strong>s of the RMS-delay spread are given in the Figure 5.24 a <strong>and</strong> b below<br />

for the 5.25 GHz centre-frequency <strong>and</strong> c-c LOS <strong>and</strong> r-c NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. Best fit is<br />

achieved with the log-normal distributi<strong>on</strong>.<br />

a<br />

Figure 5.24: a) CDF of the A1 indoor (corridor – corridor) LOS envir<strong>on</strong>ment, fitting to normal,<br />

Gumbel <strong>and</strong> logistic distributi<strong>on</strong>s shown. b) CDF of the A1 indoor (room – corridor) NLOS<br />

envir<strong>on</strong>ment, fitting to normal distributi<strong>on</strong> shown.<br />

b<br />

5.4.3.2 Scenario B1<br />

The mean RMS delay spread for LOS <strong>and</strong> NLOS cases has been calculated for large number of <strong>channel</strong><br />

segments. The mean value of each <strong>channel</strong> segment has been calculated for data collected of every 10λ,<br />

where about five <strong>channel</strong> impulse resp<strong>on</strong>ses has been measured per wavelength. Fitting the log of the<br />

measured RMS delay spread with different distributi<strong>on</strong>s is shown in Figure 5.25. For NLOS case, the<br />

lognormal distributi<strong>on</strong> is not very close to measurement data as well as the log-logistic distributi<strong>on</strong>s. For<br />

LOS case, the Gumbel distributi<strong>on</strong> follows most of the points of the measurement CDF. The closest<br />

distributi<strong>on</strong> is the Gumbel distributi<strong>on</strong>, which is a special case of the Fisher-Tippett Distributi<strong>on</strong>. It is<br />

particularly c<strong>on</strong>venient for extreme values data. It may be used as an alternative to the normal distributi<strong>on</strong><br />

in the case of skewed empirical data.<br />

(a) LOS<br />

Figure 5.25: RMS delay spread in Scenario B1.<br />

(b) NLOS<br />

Figure 5.26 shows the maximum excess delay for both LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. Table 5.9<br />

presents mean <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong>s of both RMS delay spread <strong>and</strong> the ZDSC excess delays for both<br />

LOS <strong>and</strong> NLOS cases.<br />

Page 76 (167)


WINNER D5.4 v. 1.4<br />

Cumulative Probability<br />

1<br />

0.5<br />

Empirical CDF<br />

0<br />

0 500 1000 1500<br />

τ, ns<br />

Cumulative Probability<br />

1<br />

0.5<br />

Empirical CDF<br />

0<br />

0 500 1000 1500<br />

τ, ns<br />

(a) LOS<br />

Figure 5.26: ZDSC excess delay.<br />

(b) NLOS<br />

Table 5.9: Mean <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> of the RMS delay spread <strong>and</strong> maximum excess delay.<br />

Propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong> LOS NLOS<br />

Centre frequencies (GHz) 5.25 5.25<br />

RMS delay spread Mean (ns) 37 74<br />

Std (ns) 25 25<br />

Max excess delay Mean (ns) 327 567<br />

Std (ns) 234 130<br />

5.4.3.3 Scenario B3<br />

The mean RMS delay spread for LOS <strong>and</strong> NLOS cases has been calculated for large number of <strong>channel</strong><br />

segments. The CDF <strong>and</strong> percentiles are presented in Figure 5.27 <strong>and</strong> Table 5.10, respectively. The mean<br />

value of each <strong>channel</strong> segment has been calculated for data set, where ten <strong>channel</strong> impulse resp<strong>on</strong>ses have<br />

been measured. The CDF <strong>and</strong> percentiles of maximum excess delay for both LOS <strong>and</strong> NLOS cases are<br />

shown in Figure 5.28 <strong>and</strong> Table 5.11 respectively.<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 20 40 60 80<br />

RMS delay spread [ns]<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 20 40 60 80<br />

RMS delay spread [ns]<br />

(a) LOS<br />

(b) NLOS/OLOS<br />

Figure 5.27: RMS delay spread, scenario B3.<br />

Page 77 (167)


WINNER D5.4 v. 1.4<br />

Table 5.10: Percentiles RMS delay spread [ns].<br />

RMS delay spread (ns)<br />

LOS<br />

NLOS<br />

10% 13.2 22.3<br />

Percentile<br />

50% 23.7 37.7<br />

90% 34.6 48.9<br />

mean 23.5 36.7<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 100 200 300 400<br />

Max excess delay [ns]<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 100 200 300 400<br />

Max excess delay [ns]<br />

(a) LOS<br />

(b) NLOS/OLOS<br />

Figure 5.28: Maximum excess delay, scenario B3.<br />

Table 5.11: Percentiles of maximum excess delay [ns].<br />

Maximum excess delay (ns)<br />

LOS<br />

NLOS<br />

10% 83.4 116.7<br />

Percentile<br />

50% 125.0 175.1<br />

90% 175.0 249.9<br />

mean 129.3 186.1<br />

5.4.3.4 Scenario C1<br />

5.4.3.4.1 Measurements in LOS c<strong>on</strong>diti<strong>on</strong>s<br />

The distributi<strong>on</strong> of the RMS-delay spread in C1 suburban scenario was investigated. The 10, 50 <strong>and</strong> 90 %<br />

values for the Cumulative Distributi<strong>on</strong> Functi<strong>on</strong>s of the distributi<strong>on</strong> of the RMS-delay spread are given<br />

below for the 5.25 GHz centre-frequency <strong>and</strong> LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

RDS (ns) 10% 50% 90% mean<br />

Suburban envir<strong>on</strong>ment 9 59 175 84<br />

Table 5.12: Percentiles of the RMS-delay spread in suburban envir<strong>on</strong>ment.<br />

5.4.3.5 Scenario D1<br />

5.4.3.5.1 Ordinary delays<br />

The The 10, 50 <strong>and</strong> 90 % percentiles of the measured RMS-delay spread are shown in the Table 5.13 for<br />

5.25 GHz in 100 MHz b<strong>and</strong>width in LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. In [5.3] it was <str<strong>on</strong>g>report</str<strong>on</strong>g>ed that<br />

Page 78 (167)


WINNER D5.4 v. 1.4<br />

the behaviour is very similar at 2.45 <strong>and</strong> 5.25 GHz centre-frequencies. The maximum excess delays were<br />

found to be roughly two to three times higher than the RMS-delay spreads.<br />

Table 5.13: Percentiles of the RMS-delay spread in a rural envir<strong>on</strong>ment.<br />

Rms delay spread (ns) LOS NLOS<br />

10% 2.5 4.3<br />

Percentile<br />

50% 15.4 37.1<br />

90% 84.4 89.5<br />

mean 36.8 42.1<br />

5.4.3.5.2 Excepti<strong>on</strong>ally l<strong>on</strong>g delays<br />

The RMS delay spread as fucti<strong>on</strong> of distance al<strong>on</strong>g the measurement route was discussed in [5.3]. One<br />

example is shown in Figure 5.29. It can be seen that near 740 m from the start of the measurement route<br />

there is an abrupt rise of the RMS delay spread. The delay spread jumps there from some tens of<br />

nanosec<strong>on</strong>ds up to 800 ns for a short interval, about 25 m. The reas<strong>on</strong> is obviously a reflecti<strong>on</strong> from a<br />

nearby radio mast.<br />

900<br />

800<br />

700<br />

Dealay spread (ns)<br />

600<br />

500<br />

400<br />

300<br />

200<br />

100<br />

0<br />

0 100 200 300 400 500 600 700 800 900<br />

Distance (m )<br />

Figure 5.29: RMS delay spread as a functi<strong>on</strong> of distance.<br />

It should be noted that this kind of reflectors, e.g. radio masts <strong>and</strong> supporting pillars of power lines, are<br />

quite comm<strong>on</strong> in our rural envir<strong>on</strong>ments. However, the probability of reflecti<strong>on</strong>s was not possible to be<br />

estimated in our current campaign.<br />

This kind of excepti<strong>on</strong>ally delayed paths can not be modelled with the primary model with exp<strong>on</strong>entially<br />

distributed delay spreads. They have to be modelled as far clusters [SCM]. However, at the current model<br />

this kind of excepti<strong>on</strong>al phenomen<strong>on</strong> has been neglected.<br />

5.4.4 Azimuth AS at BS <strong>and</strong> MS<br />

Azimuth angle-spread is calculated like described in [3GPP SCM] from DoA <strong>and</strong> path power values. It is<br />

known as circular angle-spread. Here it is calculated at both BS <strong>and</strong> MS <strong>link</strong> end.<br />

5.4.4.1 Scenario A1<br />

The cumulative distributi<strong>on</strong> functi<strong>on</strong>s of the azimuth spreads at 5.25 GHz are shown in Figure 5.30 for<br />

LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. The percentiles for the CDF functi<strong>on</strong>s for the angle-spreads are<br />

shown in the table Table 5.14 below.<br />

Table 5.14: Percentiles of the RMS azimuth spread.<br />

Combined Corri.-Corri. LOS Corri.-Room NLOS<br />

Tietotalo & Main building Azim. Elev. Azim. Elev.<br />

Page 79 (167)


WINNER D5.4 v. 1.4<br />

BS,<br />

MS,<br />

σ φ<br />

σ ϕ<br />

10% 2.7 4.5 10.2 6.3<br />

50% 4.8 7.6 21.5 10.9<br />

90% 16.2 13.6 39.5 21.7<br />

mean 7.0 8.4 23.0 12.9<br />

10% 14.5 4.1 24.6 7.5<br />

50% 36.6 10.1 37.4 12.1<br />

90% 67.0 15.4 56.2 21.9<br />

mean 38.8 9.9 39.3 13.5<br />

(a) LOS (Corri-Corri)<br />

(b) NLOS (Corri.-Room)<br />

Figure 5.30: Example CDFs of Azimuth spreads at BS <strong>and</strong> MS.<br />

5.4.4.2 Scenario B1<br />

Figure 5.31 shows RMS azimuth angle-spread at the MS <strong>and</strong> for LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Figure 5.32 presents RMS azimuth angle-spread at the BS for both LOS <strong>and</strong> NLOS propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong>s. We have not made any statistical fitting comparis<strong>on</strong> based <strong>on</strong> some well known techniques<br />

like KS test. However, based <strong>on</strong> Figure 5.31 the lognormal distributi<strong>on</strong> assumpti<strong>on</strong> is not as good as the<br />

log-logistic distributi<strong>on</strong> in fitting the RMS azimuth angle-spread at the MS for the case of LOS, while<br />

Gumbel distributi<strong>on</strong> has better fitting for the NLOS case. Distributi<strong>on</strong>s fitting to the RMS azimuth spread<br />

at the BS can be seen in Figure 5.32. Again for the LOS case, the log-logistic distributi<strong>on</strong> has better<br />

fitting to measurements than the lognormal distributi<strong>on</strong>. And again for NLOS case, the Gumbel<br />

distributi<strong>on</strong> has better fitting for the NLOS case.<br />

Page 80 (167)


WINNER D5.4 v. 1.4<br />

(a) LOS<br />

(b) NLOS<br />

Figure 5.31: RMS azimuth angle-spread at the MS.<br />

(a) LOS<br />

Figure 5.32: RMS azimuth angle-spread at the BS.<br />

(b) NLOS<br />

5.4.4.3 Scenario B3<br />

The cumulative distributi<strong>on</strong> functi<strong>on</strong>s of the RMS angle-spreads at 5.20 GHz (120 MHz b<strong>and</strong>width) are<br />

shown respectively in Figure 5.33 <strong>and</strong> Figure 5.34 for LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s of the B3<br />

scenario. The RMS angle-spread is calculated using the circular angle-spread formula [3GPP SCM]. No<br />

statistical fitting comparis<strong>on</strong> based <strong>on</strong> some well known techniques like KS test is applied. The<br />

percentiles for the CDF functi<strong>on</strong>s for the angle-spreads are shown in the table below.<br />

Table 5.15: Percentiles of the RMS azimuth spread.<br />

Link end BS MS<br />

Propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong><br />

Percentile<br />

(degrees)<br />

LOS NLOS LOS NLOS<br />

10 2 5 20 14<br />

50 9 15 63 40<br />

90 18 31 100 96<br />

Prob(angular spread @BS < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 10 20 30 40 50 60 70<br />

angular spread @BS [deg]<br />

(a)<br />

Prob(angular spread @MS < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 10 20 30 40 50 60 70<br />

angular spread @MS [deg]<br />

(b)<br />

Figure 5.33: RMS angle-spreads at (a) BS (AoA) <strong>and</strong> (b) MS (AoD) for the B3 scenario under LOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>.<br />

Page 81 (167)


WINNER D5.4 v. 1.4<br />

Prob(angular spread @BS < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 10 20 30 40 50 60 70<br />

angular spread @BS [deg]<br />

(a)<br />

Prob(angular spread @MS < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 10 20 30 40 50 60 70<br />

angular spread @MS [deg]<br />

(b)<br />

Figure 5.34: RMS angle-spreads at (a) BS (AoA) <strong>and</strong> (b) MS (AoD) for the B3 scenario under<br />

NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>.<br />

5.4.4.4 Scenario D1<br />

Measured angle-spread cumulative distributi<strong>on</strong> functi<strong>on</strong>s at MS <strong>and</strong> BS at 5.25 GHz are shown in Figure<br />

5.35. The percentiles for the azimuth spreads at BS <strong>and</strong> at MS are shown in the Table 5.16.<br />

Table 5.16: Percentiles of the RMS azimuth spread.<br />

Rural Tyrnävä LOS NLOS<br />

BS,<br />

MS,<br />

σ φ<br />

σ ϕ<br />

10% 10.2 5.6<br />

50% 21.9 18.0<br />

90% 36.2 34.3<br />

mean 21.7 19.5<br />

10% 8.3 6.0<br />

50% 20.3 22.3<br />

90% 37.5 36.4<br />

mean 22.4 21.9<br />

(a) LOS<br />

(b) NLOS<br />

Figure 5.35: CDF of the RMS azimuth angle-spreads at BS <strong>and</strong> MS in LOS (a) <strong>and</strong> NLOS (b)<br />

c<strong>on</strong>diti<strong>on</strong>s.<br />

Page 82 (167)


WINNER D5.4 v. 1.4<br />

5.4.5 Distributi<strong>on</strong> of the azimuth angles of the multipath comp<strong>on</strong>ents<br />

The distributi<strong>on</strong> of the azimuth angle of arrivals <strong>and</strong> angle of departure for both LOS <strong>and</strong> NLOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s are calculated. Those results are based <strong>on</strong> superresoluti<strong>on</strong> path parameter<br />

estimati<strong>on</strong>s or the beamform method. CDFs are presented as well as characteristic parameters as 10%,<br />

50%, 90% <strong>and</strong> mean of the distributi<strong>on</strong> are extracted.<br />

5.4.5.1 Scenario B1<br />

The azimuth angle of arrivals at the MS (receive) <strong>and</strong> angle of departure from the BS (transmitter) for<br />

both LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s has been extracted from measurement data using<br />

beamforming techniques. It was noted that for LOS c<strong>on</strong>diti<strong>on</strong>s there are two propagati<strong>on</strong> mechanisms that<br />

take place for signals arrive the MS. The forward propagati<strong>on</strong>, i.e., direct propagati<strong>on</strong> directi<strong>on</strong> from BS<br />

to the MS, <strong>and</strong> the backscattering for signals that travel bey<strong>on</strong>d the MS <strong>and</strong> scatter back to the MS from<br />

the opposite directi<strong>on</strong>. These two sources of arriving signals at the MS make the modelling of the azimuth<br />

arrivals to be different. Figure 5.36 shows the cumulative probability distributi<strong>on</strong> functi<strong>on</strong> of arrival<br />

angles at the MS from both directi<strong>on</strong>s. The logistic <strong>and</strong> normal distributi<strong>on</strong>s are closer in fitting with<br />

measurement data from that with Laplacian distributi<strong>on</strong> for signals that arrive due to backscattering. In<br />

the forward propagati<strong>on</strong> case, the normal distributi<strong>on</strong> is not in good fit with measurements data. Laplacian<br />

distributi<strong>on</strong> is not in an excellent fitting with measurements but closer. For angle of departure from the<br />

BS to both LOS <strong>and</strong> NLOS cases the logistic <strong>and</strong> Laplacian are in better agreement with measurement<br />

data compared to normal distributi<strong>on</strong> as can be seen in Figure 5.37.<br />

(a) DoA due to Backscattering propagati<strong>on</strong><br />

(b) DoA from forward propagati<strong>on</strong>.<br />

Figure 5.36: Azimuth directi<strong>on</strong> of arrivals of multipath comp<strong>on</strong>ents in LOS cases.<br />

(a) DoD LOS case<br />

(b) DoD NLOS case.<br />

Figure 5.37: Azimuth directi<strong>on</strong> of departure of multipath comp<strong>on</strong>ents in LOS <strong>and</strong> NLOS cases.<br />

Page 83 (167)


WINNER D5.4 v. 1.4<br />

5.4.5.2 Scenario B3<br />

The cumulative distributi<strong>on</strong> functi<strong>on</strong>s of the AoAs <strong>and</strong> AoDs for the multipath comp<strong>on</strong>ents at 5.20 GHz<br />

(120 MHz b<strong>and</strong>width) are shown in Figure 5.38 <strong>and</strong> Figure 5.39 for LOS <strong>and</strong> NLOS propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong>s. The estimati<strong>on</strong> results for the AoAs <strong>and</strong> the AoDs are based <strong>on</strong> the superresoluti<strong>on</strong> algorithm<br />

RIMAX [RIMAX]. No statistical fitting comparis<strong>on</strong> based <strong>on</strong> some well known techniques like KS test is<br />

applied. The percentiles for the CDF functi<strong>on</strong>s for the AoAs <strong>and</strong> AoDs are shown in the table below.<br />

Here results are shown for the LOS <strong>and</strong> NLOS propagati<strong>on</strong> case.<br />

Table 5.17: Percentiles of the distributi<strong>on</strong> of azimuth.<br />

Link end BS MS<br />

Propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong><br />

Percentile<br />

(degrees)<br />

LOS NLOS LOS NLOS<br />

10 -49.1 -41.8 -107.3 -125.5<br />

50 -1.8 -1.8 -5.5 -5.5<br />

90 38.2 38.2 110.9 114.5<br />

mean -1.3 -0.4 -0.1 -3.5<br />

1<br />

1<br />

Prob(angle @BS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Prob(angle @MS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-150 -100 -50 0 50 100 150<br />

angle @BS [deg]<br />

0<br />

-150 -100 -50 0 50 100 150<br />

angle @MS [deg]<br />

(a)<br />

(b)<br />

Figure 5.38: CDFs of azimuth angles at (a) BS (AoA) <strong>and</strong> (b) MS (AoD) for the B3 scenario under<br />

LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>.<br />

1<br />

1<br />

Prob(angle @BS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Prob(angle @MS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-150 -100 -50 0 50 100 150<br />

angle @BS [deg]<br />

0<br />

-150 -100 -50 0 50 100 150<br />

angle @MS [deg]<br />

(a)<br />

(b)<br />

Figure 5.39: CDFs of azimuth angles at (a) BS (AoA) <strong>and</strong> (b) MS (AoD) for the B3 scenario under<br />

NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>.<br />

Page 84 (167)


WINNER D5.4 v. 1.4<br />

5.4.6 Angle proporti<strong>on</strong>ality factor<br />

The angle proporti<strong>on</strong>ality factor (r AS ) is defined as the ratio between the st<strong>and</strong>ard deviati<strong>on</strong> of the azimuth<br />

angles of the multipath comp<strong>on</strong>ents <strong>and</strong> the RMS azimuth spread. This parameter is needed in <strong>channel</strong><br />

model.<br />

5.4.6.1 Scenario A1<br />

The angle proporti<strong>on</strong>ality factor is shown in the Table 5.18 below for an indoor (A1) envir<strong>on</strong>ment for the<br />

different LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Table 5.18: The percentiles for the CDF of the angle proporti<strong>on</strong>ality factor.<br />

Combined<br />

Corri.-Corri. Corri.-Room Room-Room<br />

Tietotalo & Main building LOS NLOS NLOS LOS (OLOS)<br />

BS,<br />

MS,<br />

r φ<br />

r ϕ<br />

10% 0.98 0.00 0.93 1.38<br />

50% 1.40 0.78 1.11 1.72<br />

90% 1.99 1.72 1.63 2.58<br />

mean 1.45 0.99 1.22 1.90<br />

10% 1.04 0.00 1.01 0.85<br />

50% 1.44 0.81 1.31 1.26<br />

90% 1.74 2.13 1.70 1.59<br />

mean 1.41 1.64 1.33 1.25<br />

5.4.6.2 Scenario B1<br />

The r AS has been extracted for signals arrive at (or depart from) the MS both in LOS <strong>and</strong> NLOS. Figure<br />

5.40 shows the results for MS terminal in both LOS <strong>and</strong> NLOS. The corresp<strong>on</strong>ding results for the BS are<br />

shown in Figure 5.41.<br />

The percentiles for the CDF of the angle proporti<strong>on</strong>ality factor in scenario B1 at the MS <strong>and</strong> at the BS are<br />

shown in Table 5.19 <strong>and</strong> Table 5.20, respectively.<br />

1<br />

Empirical CDF<br />

1<br />

Empirical CDF<br />

CDF<br />

0.5<br />

CDF<br />

0.5<br />

0<br />

0 5 10 15<br />

0<br />

0 2 4 6 8 10<br />

r AS<br />

r AS<br />

(a) LOS case<br />

(b) NLOS<br />

Figure 5.40: Angle proporti<strong>on</strong>ality factor at the MS.<br />

Table 5.19: The percentiles for the CDF of the angle proporti<strong>on</strong>ality factor in scenario B1 at the<br />

MS.<br />

Link end<br />

MS<br />

Propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong> LOS NLOS<br />

Page 85 (167)


WINNER D5.4 v. 1.4<br />

r AS<br />

10% 2 2<br />

50% 4 3.5<br />

90% 8 6<br />

1<br />

Empirical CDF<br />

1<br />

Empirical CDF<br />

CDF<br />

0.5<br />

CDF<br />

0.5<br />

0<br />

0 5 10 15<br />

0<br />

0 2 4 6<br />

r AS<br />

r AS<br />

(a) LOS case<br />

(b) NLOS<br />

Figure 5.41: Angle proporti<strong>on</strong>ality factor at the BS.<br />

Table 5.20: The percentiles for the CDF of the angle proporti<strong>on</strong>ality factor in scenario B1 at the BS.<br />

Link end<br />

BS<br />

Propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong> LOS NLOS<br />

10% 2 0.9<br />

r AS<br />

50% 4 1.1<br />

90% 7 2.2<br />

5.4.6.3 Scenario B3<br />

The angle proporti<strong>on</strong>ality factor (r AS ) has been extracted for signals arrive at (or depart from) the MS both<br />

in LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s for B3 scenario. Figure 5.42 shows the results at the MS <strong>and</strong><br />

BS for LOS. The corresp<strong>on</strong>ding results for the NLOS are shown in Figure 5.43. The percentiles for the<br />

CDF of the angle proporti<strong>on</strong>ality factor in B3 scenario are shown in Table 5.21 for LOS <strong>and</strong> NLOS.<br />

1<br />

1<br />

Prob(r-factor @BS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Prob(r-factor @BS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 2 4 6 8 10<br />

r-factor @BS<br />

0<br />

0 2 4 6 8 10<br />

r-factor @BS<br />

(a)<br />

(b)<br />

Figure 5.42: Angle proporti<strong>on</strong>ality factor at the (a) BS <strong>and</strong> (b) MS under LOS.<br />

Page 86 (167)


WINNER D5.4 v. 1.4<br />

1<br />

1<br />

Prob(r-factor @BS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Prob(r-factor @MS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 2 4 6 8 10<br />

r-factor @BS<br />

0<br />

0 2 4 6 8 10<br />

r-factor @MS<br />

(a)<br />

(b)<br />

Figure 5.43: Angle proporti<strong>on</strong>ality factor at the (a) BS <strong>and</strong> (b) MS under NLOS.<br />

Table 5.21: The percentiles for the CDF of the angle proporti<strong>on</strong>ality factor in scenario B3 at the BS<br />

<strong>and</strong> the MS, LOS <strong>and</strong> NLOS.<br />

Link end BS MS<br />

Propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong><br />

Percentile<br />

(degrees)<br />

LOS NLOS LOS NLOS<br />

10 0.6 0.7 1.0 0.8<br />

50 1.2 1.2 1.5 1.1<br />

90 2.3 2.7 2.8 1.8<br />

mean 1.5 1.6 1.9 1.3<br />

5.4.6.4 Scenario D1<br />

The angle proporti<strong>on</strong>ality factor for a rural (D1) envir<strong>on</strong>ment is shown in the Table 5.22 below for the<br />

LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Table 5.22: The percentiles for the CDF of the angle proporti<strong>on</strong>ality factor.<br />

Rural Tyrnävä LOS NLOS<br />

BS,<br />

MS,<br />

r φ<br />

r ϕ<br />

10% 0.00 0.00<br />

50% 0.74 0.62<br />

90% 1.58 2.85<br />

mean 1.17 2.11<br />

10% 0.00 0.00<br />

50% 3.66 2.32<br />

90% 10.4 13.1<br />

mean 6.70 8.84<br />

5.4.7 Modelling of PDP<br />

Power Delay Profile (PDP) is the distributi<strong>on</strong> of the power of the multipath comp<strong>on</strong>ents versus delay<br />

time. Power delay profiles for LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s have been fitted to the exp<strong>on</strong>ential<br />

functi<strong>on</strong><br />

P<br />

−bτ<br />

( τ ) e<br />

= (5.18)<br />

where τ is the excess delay <strong>and</strong> b is a time c<strong>on</strong>stant. Excess delay is difference between delays of the<br />

multipath comp<strong>on</strong>ents <strong>and</strong> the delay of the first multipath comp<strong>on</strong>ent.<br />

Page 87 (167)


WINNER D5.4 v. 1.4<br />

5.4.7.1 Scenario A1<br />

Power delay profile at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in an indoor envir<strong>on</strong>ment is<br />

shown in the figure Figure 5.44 for LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. Power delay profiles for LOS<br />

<strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s have been fitted to the exp<strong>on</strong>ential functi<strong>on</strong><br />

where τ is the excess delay <strong>and</strong> b is a time c<strong>on</strong>stant.<br />

P<br />

−bτ<br />

( τ ) e<br />

= (5.19)<br />

The results are grouped in the following way: corridor to corridor (c-c) LOS, corridor to room/room to<br />

corridor (r-c) NLOS, room to room (r-r) LOS <strong>and</strong> corridor to corridor (c-c) NLOS. This grouping adapts<br />

the results more precisely to the defined A1 scenario. The results for them are shown in the table below.<br />

Table 5.23: Time c<strong>on</strong>stants for PDPs (MHz).<br />

c-c LOS c-r NLOS r-r LOS c-c NLOS<br />

50 30 69 32<br />

a<br />

Figure 5.44: Power delay profile at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in an A1<br />

indoor envir<strong>on</strong>ment for corridor – corridor LOS <strong>and</strong> room – corridor NLOS propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong>s.<br />

b<br />

The peak at 250 ns delay in the Figure 5.44 a represents a reflecti<strong>on</strong> from the corridor end. If desired, it<br />

could be introduced in the model depending <strong>on</strong> the positi<strong>on</strong> of the BS. In our model we will neglect it,<br />

because of the low <strong>level</strong> of it.<br />

5.4.7.2 Scenario B1<br />

Mean measured power delay profiles (PDP) of LOS <strong>and</strong> NLOS averaged over all corresp<strong>on</strong>ding routes<br />

are shown in Figure 5.45. They are modelled <strong>and</strong> shown to be exp<strong>on</strong>ential decaying functi<strong>on</strong>.<br />

Page 88 (167)


WINNER D5.4 v. 1.4<br />

P [dB]<br />

0<br />

-2<br />

-4<br />

-6<br />

-8<br />

-10<br />

-12<br />

-14<br />

-16<br />

-18<br />

-20<br />

0 0.2 0.4 0.6 0.8 1 1.2 1.4<br />

τ [s]<br />

x 10 -7<br />

(a) LOS<br />

P [dB]<br />

0<br />

-2<br />

-4<br />

-6<br />

-8<br />

-10<br />

-12<br />

-14<br />

-16<br />

-18<br />

-20<br />

0 1 2 3 4 5 6 7<br />

τ [s]<br />

x 10 -7<br />

(b) NLOS<br />

Figure 5.45: PDP of LOS <strong>and</strong> NLOS c<strong>on</strong>diti<strong>on</strong>s.<br />

5.4.7.3 Scenario B3<br />

Power delay profiles at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.2 GHz centre-frequency in an indoor envir<strong>on</strong>ment are<br />

shown in the figure below for LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s, <strong>and</strong> the 10, 50 <strong>and</strong> 90 percentiles<br />

are shown in the table.<br />

The value for b was <str<strong>on</strong>g>report</str<strong>on</strong>g>ed to be 35.7 MHz for LOS <strong>and</strong> 21.9 MHz for NLOS c<strong>on</strong>diti<strong>on</strong>s.<br />

Table 5.24: Time c<strong>on</strong>stants for PDPs (MHz).<br />

Time c<strong>on</strong>stant<br />

[MHz]<br />

LOS<br />

NLOS<br />

35.7 21.9<br />

0<br />

0<br />

-5<br />

-5<br />

Power (dB)<br />

-10<br />

-15<br />

Power (dB)<br />

-10<br />

-15<br />

-20<br />

-20<br />

-25<br />

0 50 100 150<br />

Excess delay [ns]<br />

(a) LOS<br />

-25<br />

0 50 100 150 200 250<br />

Excess delay [ns]<br />

(b) NLOS/OLOS<br />

Figure 5.46: Modeling of power delay profile (PDP) an indoor envir<strong>on</strong>ment (large) LOS (a) <strong>and</strong><br />

NLOS/OLOS (b) propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

5.4.7.4 Scenario C1<br />

Power delay profile at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in a suburban envir<strong>on</strong>ment is<br />

shown in the Figure 5.47 for LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. The profile has been fitted to the exp<strong>on</strong>ential<br />

functi<strong>on</strong><br />

P<br />

−bτ<br />

( τ ) e<br />

= (5.20)<br />

Page 89 (167)


WINNER D5.4 v. 1.4<br />

where τ is the excess delay <strong>and</strong> b is a time c<strong>on</strong>stant. For the suburban LOS envir<strong>on</strong>ment the c<strong>on</strong>stant b is<br />

40 MHz.<br />

Figure 5.47: Power delay profile in the C1 (suburban) LOS envir<strong>on</strong>ment with fitting to exp<strong>on</strong>ential<br />

model.<br />

5.4.7.5 Scenario D1<br />

Power delay profiles for the rural envir<strong>on</strong>ment were investigated in 2004 <strong>and</strong> <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [D5.3] for LOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. In our current campaign both LOS <strong>and</strong> NLOS c<strong>on</strong>diti<strong>on</strong>s were investigated. The<br />

results of the current campaign are shown in the figure Figure 5.48. The results in [D5.3] are comparable,<br />

but less detailed.<br />

Mean PDP profiles at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in a rural envir<strong>on</strong>ment are<br />

shown in the following figures.<br />

Figure 5.48: PDP profile in LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Power delay profile for LOS c<strong>on</strong>diti<strong>on</strong>s has been fitted to two segments with an exp<strong>on</strong>ential functi<strong>on</strong><br />

P<br />

−bτ<br />

( τ ) e<br />

= (5.21)<br />

where τ is the excess delay <strong>and</strong> b is the time-c<strong>on</strong>stant. Here the c<strong>on</strong>stants b 1 is 220 MHz for the first<br />

segment <strong>and</strong> b 2 is 15.6 MHz for the sec<strong>on</strong>d <strong>on</strong>e.<br />

For the D1 rural NLOS c<strong>on</strong>diti<strong>on</strong>s the PDP has been investigated in the current measurement campaign.<br />

The measured PDP can be seen in the Figure 5.49 below with dual slope <strong>and</strong> single slope fitting.<br />

Page 90 (167)


WINNER D5.4 v. 1.4<br />

a<br />

Figure 5.49: PDP profile in NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. a) dual slope model, b) single slope<br />

model.<br />

b<br />

Here the c<strong>on</strong>stants b 1 <strong>and</strong> b 2 for the dual slope fitting are 130 MHz <strong>and</strong> 16.4 MHz, respectively. In spite<br />

of the fact that the best fit can be obtained with the dual slope profile we will use a single slope profile in<br />

the model for simplicity. Then the line fitted to the profile will have the time-c<strong>on</strong>stant b = 60 MHz.<br />

5.4.8 Number of ZDSC<br />

This sub-secti<strong>on</strong> presents number of clusters that has been extracted from measurements. The extracted<br />

clusters are based <strong>on</strong> definiti<strong>on</strong> that used in the <strong>channel</strong> model as clusters with zero delay spread. In other<br />

words, the c<strong>on</strong>sidered clustering is in angle domain. These clusters are called zero-delay-spread clusters<br />

(ZDSC). Detailed discussi<strong>on</strong> about ZDSC is given in Chapter 4.<br />

5.4.8.1 Scenario A1<br />

The distributi<strong>on</strong> of the number of clusters was investigated in an A1 indoor envir<strong>on</strong>ment. The results are<br />

shown below as the 10, 50 <strong>and</strong> 90 % percentiles of the distributi<strong>on</strong>.<br />

Table 5.25: Percentiles for the number of paths in A1 indoor scenario <strong>and</strong> different propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong>s.<br />

Number of paths<br />

Corri.-Corri. Corri.-Room Room-Room<br />

LOS NLOS NLOS LOS(OLOS)<br />

10% 8.0 5.0 4.0 5.0<br />

50% 13.0 8.0 8.0 7.0<br />

90% 19.0 15.0 14.0 9.0<br />

mean 13.0 9.0 9.0 7.0<br />

5.4.8.2 Scenario B1<br />

Figure 5.50 shows the cumulative probability of the ZDSC in both LOS <strong>and</strong> NLOS c<strong>on</strong>diti<strong>on</strong>s. Table 5.26<br />

lists the 10, 50, 90 percentiles of the empirical cumulative probability of the extracted number of ZDSCs.<br />

Page 91 (167)


WINNER D5.4 v. 1.4<br />

1<br />

Empirical CDF<br />

1<br />

Empirical CDF<br />

0.8<br />

0.8<br />

CDF<br />

0.6<br />

0.4<br />

CDF<br />

0.6<br />

0.4<br />

0.2<br />

0.2<br />

0<br />

0 5 10 15<br />

Number of ZDSC<br />

(a) LOS<br />

0<br />

5 10 15 20 25<br />

Number of ZDSC<br />

(b) NLOS<br />

Figure 5.50: Number of ZDSC.<br />

Table 5.26: Number of extracted ZDSC.<br />

Percentile 10 50 90<br />

LOS 2 6 10<br />

NLOS 11 14 17<br />

5.4.8.3 Scenario B3<br />

The results for the number of the ZDSCs shown in are again calculated by resampling the data with 100<br />

MHz sampling rate. Table 5.27 presents the 10, 50 <strong>and</strong> 90 percentiles of the cumulative distributi<strong>on</strong> of the<br />

number of ZDSCs.<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 4 8 12 16 20 24 28 32 36<br />

Number of ZDSC<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 4 8 12 16 20 24 28 32 36<br />

Number of ZDSC<br />

(a) LOS<br />

(b) NLOS/OLOS<br />

Figure 5.51: Number of ZDSCs.<br />

Table 5.27: Number of ZDSCs.<br />

Number of ZDSC<br />

LOS<br />

NLOS<br />

Percentile<br />

10% 8 14<br />

50% 13 22<br />

Page 92 (167)


WINNER D5.4 v. 1.4<br />

90% 17 27<br />

mean 13 22<br />

5.4.8.4 Scenario C1<br />

The distributi<strong>on</strong> of the number of clusters was investigated in a suburban C1 LOS envir<strong>on</strong>ment. The<br />

results are shown in the Table 5.28 below as the 10, 50 <strong>and</strong> 90 % percentiles of the distributi<strong>on</strong>.<br />

Table 5.28: Percentiles of the CDF of the distributi<strong>on</strong> of the number of clusters in a suburban (C1)<br />

LOS envir<strong>on</strong>ment.<br />

No. ZDSC 10% 50% 90% mean<br />

Suburban macro 3 8 22 8<br />

5.4.8.5 Scenario D1<br />

The distributi<strong>on</strong> of the number of clusters was investigated in the measurement campaign. The results are<br />

shown in the Table 5.29 as the 10, 50 <strong>and</strong> 90 % percentiles of the distributi<strong>on</strong>. The results differ from the<br />

results <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [D5.3]. The reas<strong>on</strong> is the larger number of routes measured in the latter campaign.<br />

Percentiles of the number of paths in rural envir<strong>on</strong>ment are shown in the<br />

Table 5.29: Percentiles of the number of paths in rural envir<strong>on</strong>ment.<br />

Number of paths LOS NLOS<br />

Percentile<br />

10% 1.0 1.0<br />

50% 5.0 6.0<br />

90% 17.0 14.0<br />

mean 7.4 6.7<br />

5.4.9 Distributi<strong>on</strong> of ZDSC delays<br />

Each ZDSC has a number of multipath comp<strong>on</strong>ents that differs in angle of arrivals or angle of departures<br />

but they have very close delays, i.e., multipath comp<strong>on</strong>ents that have differential delays within a chip<br />

durati<strong>on</strong> are c<strong>on</strong>sidered as <strong>on</strong>e ZDSC. Since the measurements <strong>system</strong> does not provide absolute delay,<br />

the differential delay of the ZDSCs are extracted from measurements <strong>and</strong> for both LOS <strong>and</strong> NLOS<br />

c<strong>on</strong>diti<strong>on</strong>s.<br />

5.4.9.1 Scenario A1<br />

The percentiles of the distributi<strong>on</strong> of the path delays are shown in the Table 5.30 below. The distributi<strong>on</strong><br />

can be fitted to an exp<strong>on</strong>ential distributi<strong>on</strong>, see Figure 5.52.<br />

Table 5.30: The 10, 50 <strong>and</strong> 90 % percentiles for the cumulative distributi<strong>on</strong> functi<strong>on</strong> of the path<br />

delays for an indoor envir<strong>on</strong>ment at 5.25 GHz, <strong>and</strong> different propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Path delays<br />

Corri.-Corri. Corri.-Room Room-Room<br />

LOS NLOS NLOS LOS(OLOS)<br />

Percentile<br />

10% 11.5 0.0 0.0 0.0<br />

50% 132.5 72.5 82.5 57.5<br />

90% 376.3 217.0 220.0 122.5<br />

mean 174.5 102.8 100.6 61.7<br />

Page 93 (167)


WINNER D5.4 v. 1.4<br />

a<br />

Figure 5.52: a) Distributi<strong>on</strong>s of the path delays for the different sub-scenarios.<br />

b<br />

5.4.9.2 Scenario B1<br />

Figure 5.53 shows the empirical probability density functi<strong>on</strong> of the differential delays of the ZDSCs in<br />

LOS <strong>and</strong> NLOS cases. It can be seen that the distributi<strong>on</strong> of ZDSC delays follows exp<strong>on</strong>ential shape for<br />

LOS case <strong>and</strong> follows uniform distributi<strong>on</strong> shape in NLOS up to about 400 ns <strong>and</strong> after 400 ns it has<br />

exp<strong>on</strong>ential shape.<br />

(a) LOS<br />

(b) NLOS<br />

Figure 5.53: Empirical probability density functi<strong>on</strong>s of the ZDSC delays.<br />

5.4.9.3 Scenario B3<br />

In Figure 5.54 distributi<strong>on</strong>s of the ZDSC delays for the scenario B3 for LOS <strong>and</strong> NLOS envir<strong>on</strong>ments are<br />

presented. As expected, in NLOS case probability of ZDSC with higher delays is higher as in the LOS<br />

case.<br />

Page 94 (167)


WINNER D5.4 v. 1.4<br />

0.35<br />

PDF<br />

0.3<br />

0.25<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

0<br />

0 20 40 60 80 100<br />

ZDSC delays [ns]<br />

PDF<br />

0.25<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

0<br />

0 20 40 60 80 100<br />

ZDSC delays [ns]<br />

(a) LOS<br />

(b) NLOS<br />

Figure 5.54: Distributi<strong>on</strong>s of the ZDSC delays for the different sub-scenarios, a) LOS, b) NLOS.<br />

5.4.9.4 Scenario C1<br />

The percentiles of the distributi<strong>on</strong> of the path delays is shown in the Table 5.31.<br />

Table 5.31: Percentiles of the distributi<strong>on</strong> of the path delays in a suburban C1 LOS scenario.<br />

Path delay (ns) 10% 50% 90% mean<br />

Suburban macro 9.0 175 1525 528<br />

Also now the distributi<strong>on</strong> of the path delays fits well with the exp<strong>on</strong>ential distributi<strong>on</strong>.<br />

5.4.9.5 Scenario D1<br />

The percentiles of the CDF of the path delays are shown in Table 5.32. The measured probability density<br />

functi<strong>on</strong>s of the path delays are shown in the Figure 5.55. The distributi<strong>on</strong>s can be fitted to an exp<strong>on</strong>ential<br />

distributi<strong>on</strong> as can be seen in the figure.<br />

Table 5.32: The 10, 50 <strong>and</strong> 90 % percentiles for the cumulative distributi<strong>on</strong> functi<strong>on</strong> of the path<br />

delays for an outdoor LOS <strong>and</strong> NLOS envir<strong>on</strong>ments at 5.25 GHz.<br />

Path delay (ns)<br />

Percentile<br />

LOS<br />

NLOS<br />

10% 0 0<br />

50% 100 80<br />

90% 403 294<br />

mean 165 124<br />

The time c<strong>on</strong>stants are 140 ns for LOS <strong>and</strong> 110 ns for NLOS c<strong>on</strong>diti<strong>on</strong>s.<br />

Page 95 (167)


WINNER D5.4 v. 1.4<br />

a<br />

Figure 5.55: a) Distributi<strong>on</strong>s of the path delays for the different sub-scenarios, a) LOS, b) NLOS.<br />

b<br />

5.4.10 Delay proporti<strong>on</strong>ality factor<br />

The delay proporti<strong>on</strong>ality factor (r DS ) is defined as the ratio between the st<strong>and</strong>ard deviati<strong>on</strong> of the delays<br />

of the multipath comp<strong>on</strong>ents <strong>and</strong> RMS delay spread.<br />

5.4.10.1 Scenario A1<br />

The delay proporti<strong>on</strong>ality factor in an A1 indoor envir<strong>on</strong>ment was calculated. The percentiles for the CDF<br />

of the delay proporti<strong>on</strong>ality factor are shown in the Table 5.33 below.<br />

Table 5.33: The 10, 50 <strong>and</strong> 90 % percentiles for the cumulative distributi<strong>on</strong> functi<strong>on</strong> of the delay<br />

proporti<strong>on</strong>ality factor in an indoor envir<strong>on</strong>ment.<br />

Delay proporti<strong>on</strong>ality factor:<br />

r τ<br />

Corri.-Corri. Corri.-Room Room-Room<br />

LOS NLOS NLOS LOS(OLOS)<br />

Percentile<br />

10% 1.9 1.5 1.7 2.4<br />

50% 3.0 2.2 2.4 3.2<br />

90% 7.5 3.9 3.2 4.2<br />

mean 3.9 2.5 2.4 3.2<br />

5.4.10.2 Scenario B1<br />

Figure 5.56 shows the empirical cumulative distributi<strong>on</strong> functi<strong>on</strong> of the proporti<strong>on</strong>ality factor both in<br />

LOS <strong>and</strong> NLOS. The median values are used as a fixed parameter in <strong>channel</strong> modelling part.<br />

Page 96 (167)


WINNER D5.4 v. 1.4<br />

(a) LOS<br />

(b) NLOS<br />

Figure 5.56: Delay proporti<strong>on</strong>ality factor r DS .<br />

5.4.10.3 Scenario B3<br />

Figure 5.57 shows the empirical cumulative distributi<strong>on</strong> functi<strong>on</strong> of the delay proporti<strong>on</strong>ality factor in<br />

scenario B3 for both LOS <strong>and</strong> NLOS case. Percentiles of delay proporti<strong>on</strong>ality factor are given in the<br />

Table 5.34.<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 1 2 3 4 5<br />

r ds<br />

(a) LOS<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 1 2 3 4 5<br />

r ds<br />

(b) NLOS/OLOS<br />

Figure 5.57: Delay proporti<strong>on</strong>ality factor r DS .<br />

Table 5.34: Percentiles of delay proporti<strong>on</strong>ality factor.<br />

Delay proporti<strong>on</strong>ality factor<br />

LOS<br />

NLOS<br />

10% 1.27 1.19<br />

Percentile<br />

50% 1.80 1.58<br />

90% 2.59 1.93<br />

mean 1.90 1.58<br />

5.4.10.4 Scenario C1<br />

The percentiles for the CDF of the delay proporti<strong>on</strong>ality factor are shown in the Table 5.35 below.<br />

Page 97 (167)


WINNER D5.4 v. 1.4<br />

Table 5.35: The 10, 50 <strong>and</strong> 90 % percentiles for the cumulative distributi<strong>on</strong> functi<strong>on</strong> of the delay<br />

proporti<strong>on</strong>ality factor in a rural envir<strong>on</strong>ment.<br />

Delay proporti<strong>on</strong>ality factor:<br />

Percentile<br />

r τ<br />

LOS<br />

NLOS<br />

10% 2.0 1.2<br />

50% 3.8 1.7<br />

90% 8.5 2.9<br />

mean 4.7 1.9<br />

5.4.10.5 Scenario D1<br />

The percentiles for the CDF of the delay proporti<strong>on</strong>ality factor for scenario D1 are presented in the Table<br />

5.36.<br />

Table 5.36: The 10, 50 <strong>and</strong> 90 % percentiles for the cumulative distributi<strong>on</strong> functi<strong>on</strong> of the delay<br />

proporti<strong>on</strong>ality factor in a rural envir<strong>on</strong>ment.<br />

delay proporti<strong>on</strong>al factor:<br />

Percentile<br />

r τ<br />

LOS NLOS<br />

10% 2.0 1.2<br />

50% 3.8 1.7<br />

90% 8.5 2.9<br />

mean 4.7 1.9<br />

5.4.11 Ricean K-factor<br />

Narowb<strong>and</strong> Ricean K factor in the LOS regi<strong>on</strong>s has been analysed. Ricean K-factor is the ratio of power<br />

of the direct LOS comp<strong>on</strong>ent to the total power of the diffused n<strong>on</strong>-line-of-sight comp<strong>on</strong>ents.<br />

5.4.11.1 Scenario A1<br />

The narrow-b<strong>and</strong> Ricean K-factor as a functi<strong>on</strong> of distance at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centrefrequency<br />

in an indoor envir<strong>on</strong>ment is c<strong>on</strong>sidered in the corridor-corridor LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

CDF of the K-factor is shown in the Figure 5.58. The fitting of the CDF with normal distributi<strong>on</strong> is<br />

shown.<br />

Figure 5.58: a) Distributi<strong>on</strong> of the K-factor in A1 indoor scenario at 5.25 GHz centre-frequency.<br />

K-factor in the A1 indoor scenario <strong>and</strong> LOS corridor to corridor propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s at 5.25 GHz<br />

centre-frequency is shown in the figure Figure 5.59 as functi<strong>on</strong> of the BS – MS distance.<br />

Page 98 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.59: K-factor as functi<strong>on</strong> of distance in an A1 indoor scenario at 5.25 GHz centrefrequency.<br />

It can be seen that in this measurement the K-factor increases from 9 to 17 dB, when the distance<br />

increases from 0 to 150 m. Formula for the K-factor is<br />

K = 8.7 + 0.051 d, (5.22)<br />

where d is the distance between the BS <strong>and</strong> the MS.<br />

Table 5.37: The 10, 50 <strong>and</strong> 90 % percentiles for the cumulative distributi<strong>on</strong> functi<strong>on</strong> of the<br />

narrowb<strong>and</strong> K-factor (dB) for an indoor LOS envir<strong>on</strong>ment at 5.25 GHz.<br />

Narrowb<strong>and</strong> Ricean K-factor<br />

Percentile<br />

Corri.-Corri.<br />

LOS<br />

10% 3.0<br />

50% 12.7<br />

90% 18.4<br />

mean 11.5<br />

5.4.11.2 Scenario B1<br />

The narrowb<strong>and</strong> K-factor as functi<strong>on</strong> of distance (D) has been estimated from measurements at 5 GHz for<br />

LOS cases. The fitting linear equati<strong>on</strong> as a functi<strong>on</strong> of distance needed l<strong>on</strong>g distance measurements data.<br />

The slope (0.0142) of the following fitting equati<strong>on</strong> (5.25) is based <strong>on</strong> measurement data obtained from<br />

Nokia. The narrowb<strong>and</strong> K-factor is given in dB as<br />

K = 0 .0142D<br />

+ 3<br />

(5.23)<br />

The K-factor increases with distance in urban microcell. Near the transmitter there are several modes<br />

(multiple reflecti<strong>on</strong>s from walls), which cause low K-factor. When the distance is increased, the number<br />

of modes decrease (high-order modes have high attenuati<strong>on</strong>) <strong>and</strong> K-factor increases.<br />

5.4.11.3 Scenario B3<br />

The Ricean K factor for scenario B3 as a functi<strong>on</strong> of the distance <strong>and</strong> the CDF of it are shown in Figure<br />

5.60. The K factor decreases fast with distance inrease.<br />

Page 99 (167)


WINNER D5.4 v. 1.4<br />

K factor [dB]<br />

20<br />

15<br />

10<br />

5<br />

0<br />

-5<br />

measurement data based<br />

linear fitting<br />

K[dB] = 6-0.26*d[m]<br />

-10<br />

0 5 10 15 20 25 30<br />

distance [m]<br />

(a)<br />

CDF<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

measurement<br />

based data<br />

Percentiles:<br />

10%: -2 dB<br />

50%: 1 dB<br />

90%: 4.9 dB<br />

0<br />

-10 -5 0 5 10<br />

K factor [dB]<br />

(b)<br />

Figure 5.60: Scenario B3, LOS: (a) Ricean K factor as a functi<strong>on</strong> of distance, (b) CDF of the Ricean<br />

K factor.<br />

5.4.11.4 Scenario C1<br />

The CDF percentiles of the K-factor in a suburban LOS envir<strong>on</strong>ment are given in the Table 5.38.<br />

Table 5.38: Percentiles of the CDF of the Ricean K-factor in a C1 LOS envir<strong>on</strong>ment.<br />

Percentile 10% 50% 90% mean<br />

K-factor (dB) 2.6 10.0 20.7 10.9<br />

The Ricean K-factor as a functi<strong>on</strong> of distance in a suburban LOS envir<strong>on</strong>ment is shown in the Figure<br />

5.61.<br />

Figure 5.61: Ricean K-factoe as functi<strong>on</strong> of distance in a suburban envir<strong>on</strong>ment.<br />

The equati<strong>on</strong> for the K-factor can be expressed as<br />

where d is the distance between the BS <strong>and</strong> the MS.<br />

5.4.11.5 Scenario D1<br />

K = 17.1 – 0.0205 d (5.24)<br />

The percentiles of the cumulative distributi<strong>on</strong> functi<strong>on</strong> (CDF) of the Ricean K-factor at 100 MHz<br />

b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in a rural LOS envir<strong>on</strong>ment can be found in Table 5.39.<br />

Page 100 (167)


WINNER D5.4 v. 1.4<br />

Table 5.39: The 10, 50 <strong>and</strong> 90 % percentiles for the cumulative distributi<strong>on</strong> functi<strong>on</strong> of the<br />

narrowb<strong>and</strong> K-factor (dB) for a rural LOS envir<strong>on</strong>ment at 5.25 GHz.<br />

Percentile<br />

K-factor (dB)<br />

10% -0.6<br />

50% 10.9<br />

90% 20.0<br />

mean 10.1<br />

The cumulative distributi<strong>on</strong> functi<strong>on</strong> (CDF) of the Ricean K-factor at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz<br />

centre-frequency in a rural LOS envir<strong>on</strong>ment is shown in the figure below. It can be seen that the<br />

measured results fit quite well in log-normal distributi<strong>on</strong>. The parameters of the distributi<strong>on</strong> are: mean<br />

10.1 dB <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> 8.0 dB.<br />

K-factor in the D1 rural scenario <strong>and</strong> LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s at 5.25 GHz centre-frequency is shown<br />

in the Figure 5.62 as functi<strong>on</strong> of the BS – MS distance.<br />

Figure 5.62: Ricean K-factor as functi<strong>on</strong> of distance in a D1 rural envir<strong>on</strong>ment.<br />

In the rural LOS, we also noticed that K increases with increasing distance for the scenario D1. The<br />

formula is<br />

where d is the distance between the BS <strong>and</strong> the MS.<br />

5.4.12 Cross-polarizati<strong>on</strong> ratio (XPR)<br />

K = 3.7 + 0.019 d (5.25)<br />

The cross-polarizati<strong>on</strong> ratio XPR V is defined as the ratio of power received from vertical to vertical<br />

polarizati<strong>on</strong> to the power received from vertical to horiz<strong>on</strong>tal polarizati<strong>on</strong> as<br />

P<br />

VV<br />

XPR<br />

V<br />

= <strong>and</strong><br />

PVH<br />

P<br />

HH<br />

XPR<br />

H<br />

= (5.26)<br />

PHV<br />

Respectively, XPR H is defined as the power ratio between HH <strong>and</strong> HV comp<strong>on</strong>ents. The XPR values are<br />

extracted from the estimated propagati<strong>on</strong> paths using the str<strong>on</strong>gest path (LOS) <strong>and</strong> the reflected paths<br />

(scattering).<br />

5.4.12.1 Scenario A1<br />

The CDF percentile values of the XPR at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in an<br />

indoor envir<strong>on</strong>ment is shown in the Table 5.40.<br />

Table 5.40: Percentiles of the cross-polarizati<strong>on</strong> ratio.<br />

Page 101 (167)


WINNER D5.4 v. 1.4<br />

A1 indoor<br />

direct path scattered paths<br />

(LOS)<br />

(NLOS)<br />

10% 13.4 7.1<br />

XPR V 50% 23.2 11.2<br />

90% 31.3 15.8<br />

mean / std 22.6 / 7.7 11.4 / 3.4<br />

10% 12.3 6.2<br />

XPR H<br />

50% 18.3 10.2<br />

90% 25.3 15.1<br />

mean / std 18.7 / 5.8 10.4 / 3.4<br />

Figure 5.63: CDFs of the XPR V <strong>and</strong> XPR H in an A1 indoor envir<strong>on</strong>ment.<br />

5.4.12.2 Scenario B1<br />

The CDF of the cross-polarizati<strong>on</strong> ratio (XPR) at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in<br />

a rural envir<strong>on</strong>ment is shown in the Figure 5.64 <strong>and</strong> Figure 5.65 for LOS <strong>and</strong> NLOS envir<strong>on</strong>ments,<br />

respectively. Also 10, 50 <strong>and</strong> 90% percentiles are shown in Table 5.41 <strong>and</strong> Table 5.42.<br />

Table 5.41: Cross-polarizati<strong>on</strong> ratio in LOS.<br />

P V/H [dB] XPR V [dB] XPR H [dB]<br />

Mean -1.0 8.6 9.5<br />

Median -1.2 8.7 9.8<br />

STD 2.1 1.8 2.3<br />

Page 102 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.64: Distributi<strong>on</strong> of XPR in LOS microcell: left VP <strong>and</strong> right HP.<br />

Table 5.42: Cross-polarizati<strong>on</strong> ratio in NLOS.<br />

P V/H [dB] XPR V [dB] XPR H [dB]<br />

Mean 0.4 8.0 6.9<br />

Median 0.5 7.9 6.8<br />

STD 2.5 1.8 2.8<br />

Figure 5.65: Distributi<strong>on</strong> of XPR in NLOS microcell: left VP <strong>and</strong> right HP.<br />

5.4.12.3 Scenario B3<br />

Prob(XPD)<br />

0.1<br />

0.08<br />

0.06<br />

0.04<br />

0.02<br />

Prob(XPD < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-4 -2 0 2 4<br />

XPD [dB]<br />

(a)<br />

0<br />

-4 -2 0 2 4<br />

XPD [dB]<br />

(b)<br />

Page 103 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.66: XPR 1 under LOS with (a) as PDF <strong>and</strong> (b) as CDF.<br />

Prob(XPD)<br />

0.1<br />

0.08<br />

0.06<br />

0.04<br />

0.02<br />

Prob(XPD < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-4 -2 0 2 4<br />

XPD [dB]<br />

(a)<br />

0<br />

-4 -2 0 2 4<br />

XPD [dB]<br />

(b)<br />

Figure 5.67: XPR 1 under NLOS with (a) as PDF <strong>and</strong> (b) as CDF.<br />

Table 5.43: Percentiles of the cross-polarizati<strong>on</strong> ratio XPR 1 .<br />

Propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong><br />

Percentile<br />

(degrees)<br />

LOS<br />

NLOS<br />

10 -1.2 -0.7<br />

50 0.7 0.1<br />

90 1.6 1.1<br />

mean 0.5 0.1<br />

The st<strong>and</strong>ard deviati<strong>on</strong> for the XPR 1 under LOS was found to be 1.07 dB <strong>and</strong> for NLOS 0.69 dB.<br />

5.4.12.4 Scenario C1<br />

5.4.12.4.1 LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s<br />

The CDF of the XPR values at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in a suburban<br />

envir<strong>on</strong>ment is shown in the Figure 5.68.<br />

Figure 5.68: CDF’s of the XPR V <strong>and</strong> XPR H for 5.25 GHz in suburban envir<strong>on</strong>ment.<br />

Page 104 (167)


WINNER D5.4 v. 1.4<br />

5.4.12.5 Scenario D1<br />

The CDF of the cross-polarizati<strong>on</strong> ratio (XPR) at 100 MHz b<strong>and</strong>width <strong>and</strong> 5.25 GHz centre-frequency in<br />

a rural envir<strong>on</strong>ment is shown in the Figure 5.69. The corresp<strong>on</strong>ding percentiles are listed in the Table<br />

5.44.<br />

Table 5.44: Percentiles of the cross-polarizati<strong>on</strong> ratios in a D1 rural envir<strong>on</strong>ment.<br />

D1 rural<br />

direct path<br />

(LOS)<br />

scattered paths<br />

(NLOS)<br />

10% 1.7 3.7<br />

XPR V 50% 12.2 6.3<br />

90% 20.7 9.2<br />

mean / std 11.7 / 7.8 6.4 / 2.2<br />

XPR H<br />

10% 3.2 3.2<br />

50% 13.5 6.1<br />

90% 23.3 9.1<br />

mean / std 13.2 6.1 / 2.3<br />

Figure 5.69: CDFs for the XPR V <strong>and</strong> XPR H for 5.25 GHz.<br />

5.4.13 Large-scale parameter analysis item<br />

In Secti<strong>on</strong> 3.1, the model for the so-called large-scale parameters are introduced. In the following<br />

subsecti<strong>on</strong>s the required parameters are estimated for scenario A1 LOS/NLOS, B1 LOS/NLOS, B3<br />

LOS/NOS, C1 LOS/NLOS, C2 NLOS <strong>and</strong> D1 LOS/NLOS. In all cases except A1, the vector of bulk<br />

parameters s( x, y)<br />

has four dimensi<strong>on</strong>s corresp<strong>on</strong>ding to the delay-spread, AoD spread, AoA spread <strong>and</strong><br />

log-normal shadowing. In A1, it has the additi<strong>on</strong>al dimensi<strong>on</strong> of AoD elevati<strong>on</strong> spread <strong>and</strong> AoA elevati<strong>on</strong><br />

spread. The required parameters are the vector of transformati<strong>on</strong> functi<strong>on</strong>s ~ s ( x , y)<br />

= g( s( x,<br />

y)<br />

), which<br />

s x, y into a vector ~ s ( x, y)<br />

of four Gaussian r<strong>and</strong>om variables. The mean<br />

transfoRMS the bulk vector ( )<br />

µ <strong>and</strong> correlati<strong>on</strong> R( 0)<br />

of the transformed r<strong>and</strong>om variable, <strong>and</strong> the decorrelati<strong>on</strong> distance parameters<br />

λ , K λ determine the variati<strong>on</strong> of the large-scale vector over the cell area through the equati<strong>on</strong>s<br />

1<br />

,<br />

4<br />

2<br />

E { ( x , y ) s( x y )} = R( ∆r)<br />

( ) ( ) 2<br />

R<br />

s<br />

1 1 2,<br />

2<br />

⎛<br />

⎜<br />

⎝<br />

⎛<br />

⎜<br />

⎝<br />

∆ r = x<br />

(5.27)<br />

2 − x1<br />

+ y2<br />

− y1<br />

∆r<br />

⎞ ⎛ ∆r<br />

⎞⎞<br />

⎟<br />

K<br />

⎜ ⎟⎟<br />

, (*) (5.28)<br />

λ1<br />

⎠ ⎝ λm<br />

⎠⎠<br />

0.5<br />

0.5,T<br />

( ∆r) = R ( 0) diag⎜exp⎜−<br />

⎟,<br />

,exp⎜−<br />

⎟⎟R<br />

( 0)<br />

0.5<br />

T 0.5<br />

5<br />

where R ( 0)<br />

is obtained from the eigendecompositi<strong>on</strong> R( 0) = EΛE<br />

as ( 0) = EΛ<br />

0.<br />

R .<br />

Page 105 (167)


WINNER D5.4 v. 1.4<br />

The parameters for scenario A1, B3 <strong>and</strong> D1 have been obtained from the WINNER measurements<br />

described in Secti<strong>on</strong> 5.2.1, 5.2.3 <strong>and</strong> 5.2.6. For scenario B1, reference data from TKK outside the<br />

WINNER project is used. A combinati<strong>on</strong> of the measurements described in Secti<strong>on</strong> 5.2.4 <strong>and</strong> literature,<br />

Secti<strong>on</strong> 5.5.4, is used for scenario C1. For scenario C2, results from the reference measurement data<br />

measured by Nokia outside the WINNER project <strong>and</strong> literature (see Secti<strong>on</strong> 5.5.5) is used.<br />

For scenario A1 LOS/NLOS, C1 LOS/NLOS, C2 NLOS, D1 LOS/NLOS <strong>and</strong> Bridge2Car, all parameters<br />

are modelled as log-normal, which means that the transformati<strong>on</strong> is simply “log(x)” <strong>and</strong> the inverse<br />

“10^x”. One excepti<strong>on</strong> is the shadow fading, where the transformati<strong>on</strong> is “10log(x)”, so that the<br />

transformed variable is in dB. In scenario B1 LOS, the delay-spread is log-Gumbel, while the AoD <strong>and</strong><br />

AoA spread is log-Logistic. In B1 NLOS, the delay-spread, AoD <strong>and</strong> AoA spread are all log-Gumbel.<br />

The transformati<strong>on</strong>s for these cases are described in Secti<strong>on</strong> 3.1.1. In scenario B3, the delay-spread, AoD<br />

spread, <strong>and</strong> AoA spread are modelled as normal (thus the transformed <strong>and</strong> untransformed variables are<br />

the same). In all cases, the shadow fading is modelled as log-normal.<br />

The mean µ <strong>and</strong> covariance matrix R ( 0)<br />

have been obtained for each scenario <strong>and</strong> are listed in the tables<br />

of Secti<strong>on</strong> 3.1.1. Also listed in Secti<strong>on</strong> 3.1.1 for each scenario <strong>and</strong> parameter is a decorrelati<strong>on</strong> distance<br />

∆ . This distance has been obtained by fitting a single exp<strong>on</strong>ential exp( − ∆r / ∆)<br />

to the auto-correlati<strong>on</strong><br />

functi<strong>on</strong>s. The expressi<strong>on</strong> (*) however, mixes all of the decorrelati<strong>on</strong> distances λ<br />

1,<br />

K,λm<br />

so that we need<br />

to do a joint fit of all four auto-correlati<strong>on</strong> functi<strong>on</strong>s. Below, we have plotted the exp<strong>on</strong>ential<br />

exp( − ∆r / ∆)<br />

together with the auto-correlati<strong>on</strong> obtained from (*). The decorrelati<strong>on</strong> distances<br />

λ<br />

1,<br />

K,λ m have been manually optimized <strong>and</strong> their values are listed in Table 3.4.<br />

5.4.13.1 A1 LOS<br />

Figure 5.70: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements in Scenario A1 LOS.<br />

5.4.13.2 A1 NLOS<br />

Page 106 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.71: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements in Scenario A1 NLOS.<br />

5.4.13.3 B1 LOS<br />

Figure 5.72: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements in Scenario B1 LOS.<br />

Page 107 (167)


WINNER D5.4 v. 1.4<br />

5.4.13.4 B1 NLOS<br />

Figure 5.73: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements in Scenario B1 NLOS.<br />

5.4.13.5 B3 LOS<br />

Figure 5.74: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements in Scenario B3 LOS.<br />

5.4.13.6 B3 NLOS<br />

Page 108 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.75: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements in Scenario B3 LOS.<br />

5.4.13.7 C1 LOS<br />

Figure 5.76: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements <strong>and</strong> literature for Scenario C1<br />

LOS.<br />

5.4.13.8 C1 NLOS<br />

Page 109 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.77: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements <strong>and</strong> literature for Scenario C1<br />

NLOS.<br />

5.4.13.9 D1 LOS<br />

Figure 5.78: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements from Scenario D1 LOS.<br />

5.4.13.10 D1 NLOS<br />

Page 110 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.79: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of Table 3.4<br />

<strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements from Scenario D1 NLOS.<br />

5.5 Literature review<br />

5.5.1 Scenario A1<br />

5.5.1.1 Path-loss<br />

In [WHL94], the indoor measurements were performed using network analyzer at 2 <strong>and</strong> 5 <strong>and</strong> 17 GHz,<br />

the RF b<strong>and</strong>width was 500 MHz. The envir<strong>on</strong>ment of the measurement locati<strong>on</strong>s was composed of a<br />

corridor of length 21.7 m, width 2 m <strong>and</strong> height 3 m <strong>and</strong> a room with dimensi<strong>on</strong>s 7 x 8 x 2.8 m. Both<br />

antennas were mounted <strong>on</strong> st<strong>and</strong>s at a height of 1.8 m.<br />

It was found that in LOS cases, the path-loss exp<strong>on</strong>ents are 1.5, 1.7, <strong>and</strong> 1.6 respectively at the three<br />

frequency b<strong>and</strong>s. There is almost no difference <strong>and</strong> <strong>on</strong>e cannot find how the path-loss exp<strong>on</strong>ents change<br />

with frequencies. However, for OLOS cases, the path-loss exp<strong>on</strong>ents were increased with the centrefrequencies.<br />

In [SG00], the measurements were performed at 5.2 GHz (RF BW was not clear) <strong>and</strong> mainly investigated<br />

the path-loss <strong>models</strong> for indoor envir<strong>on</strong>ments including the same floor corridor-corridor (LOS), corridorroom<br />

(NLOS), <strong>and</strong> room-room (NLOS) measurements <strong>and</strong> also different floor path-loss measurements.<br />

Antenna type: two patch antennas or two dipole antennas were applied.<br />

1. Same floor measurement results:<br />

office<br />

corrcorr.<br />

(LOS)<br />

corr.-<br />

room<br />

(NLOS)<br />

(NLOS)<br />

school<br />

roomroom<br />

corr.-<br />

room<br />

(NLOS)<br />

roomroom<br />

n 1.3 3.1 4.1 5.0 7.0<br />

(NLOS)<br />

PL 0 (dB) 47.4 46.1 47.9 30.6 11.3<br />

σ (dB) 2.2 2.9 2.7 2.1 3.9<br />

2. Cross floor measurement results:<br />

Horiz<strong>on</strong>tally polarized antennas:<br />

Case 1: The transmitter Tx was located at the 6th floor, then the receiver Rx was moving in a corridor at<br />

5th floor.<br />

Case 2: The Tx was at 6th floor, <strong>and</strong> the Rx was moving in vertical line between floors 0 <strong>and</strong> 5.<br />

Page 111 (167)


WINNER D5.4 v. 1.4<br />

Case 1 Case 2<br />

n 2.4 5.6<br />

PL 0 (dB) 76.4 69.5<br />

σ (dB) 1.8 0.69<br />

The transmissi<strong>on</strong> loss due to <strong>on</strong>e floor is about 30 dB in the office building. The floor losses are not<br />

increased linearly as in Keenan-Motley model. One experiment was performed at both 5 GHz <strong>and</strong> 900<br />

MHz to determine the dependence of loss <strong>on</strong> frequencies. However, no significant difference was seen,<br />

except for the expected difference in free-space loss.<br />

In reference [YMI+04], different kind of path-loss <strong>models</strong> were obtained based <strong>on</strong> measurements<br />

performed at 5.3 GHz with RF BW 30 MHz. They can be found in the following table.<br />

5.5.1.2 Rms delay spread<br />

In [1] the mean RMS delay spread decreases with centre-frequencies as shown in the table below.<br />

The difference between frequency ranges 2-2.5 GHz <strong>and</strong> 5-5.5 GHz in the table above is quite big, the<br />

ratio is 1/3 … ½. From 5-5.5 GHz to 17-17.5 GHz the difference is much smaller, if anything.<br />

In [YMI+04], [PLN+99], [OTTH01], <strong>and</strong> [YTL02] Yacoub, D.; Teich, W.; Lindner, J., „Capacity of<br />

Vehicle-Bridge MIMO Channels”, TD(02)118, COST 273, 5th Management Committee<br />

Meeting, Lisb<strong>on</strong> / Portugal, Sep. 19-20, 2002<br />

[ZKVS02] the indoor RMS delay spread statistic values were summarized in the table<br />

Page 112 (167)


WINNER D5.4 v. 1.4<br />

Class F [GHz] Distance (m) σ τ [ns] Value<br />

given<br />

LOS<br />

NLOS<br />

Method<br />

5.3 3-100 20-120 CDF 90% WCS [2]<br />

2.25 1-15 34.5-<br />

49.0<br />

5.25 1-15 14.4-<br />

15.7<br />

Ref<br />

mean VNA [1]<br />

mean VNA [1]<br />

5.3 5-200 30-180 CDF90% WCS [2]<br />

2.25 1-15 34.5-49 mean VNA [1]<br />

5.25 1-15 14-15.7 mean VNA [1]<br />

WCS: wideb<strong>and</strong> <strong>channel</strong> sounder.<br />

5.5.1.3 Angle-spreads<br />

Reference [DRX98] was c<strong>on</strong>sidering the tap <strong>and</strong> cluster angle-spreads of indoor WLAN <strong>channel</strong>s by<br />

using frequency domain measurements. The measured data with 400 MHz BW (5.0-5.4 GHz) were<br />

employed. FD (freq. domain)-SAGE was applied.<br />

1. Cluster <strong>and</strong> cluster AS: a cluster was based <strong>on</strong> the observati<strong>on</strong> that multipath comp<strong>on</strong>ents<br />

(MPCs) arrive in groups. AS means RMS angle-spread.<br />

2. Average tap AS: To find a tap AS for the <strong>channel</strong> with a specific delay resoluti<strong>on</strong> 1 f c , all the<br />

MPCs were collected for every 1 f c <strong>and</strong> put them in the same delay bin. For each individual<br />

tap, the instantaneous tap AS was calculated.<br />

Average tap AS <strong>and</strong> cluster AS<br />

Average tap AS with different b<strong>and</strong>widths<br />

It is interesting to notice that the mean tap AS <strong>and</strong> cluster AS have some difference, but small. The tap<br />

AS changes with RF b<strong>and</strong>width, but the difference is quite small.<br />

In [Xia96] mean azimuth spread <strong>on</strong> MS side at 5 GHz is 60.67° <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> is 14.26°. In the<br />

measurement setup BS antenna height was 3 m <strong>and</strong> MS antenna height 1.2 m.<br />

Measurement results from the COST 273 acti<strong>on</strong> for indoor office envir<strong>on</strong>ment are collected in [BBK+02].<br />

The following table c<strong>on</strong>tains informati<strong>on</strong> about azimuth, elevati<strong>on</strong> <strong>and</strong> delay spreads as well as about the<br />

number of identified clusters:<br />

Page 113 (167)


WINNER D5.4 v. 1.4<br />

5.5.1.4 Spatio-temporal correlati<strong>on</strong> properties<br />

Reference [EGT+99] is about the spatio-temporal correlati<strong>on</strong> properties for 5.2 GHz indoor propagati<strong>on</strong><br />

envir<strong>on</strong>ments. The RF b<strong>and</strong>width is 120 MHz. The definiti<strong>on</strong>s of the RMS delay spread <strong>and</strong> RMS<br />

azimuth angle-spread are the same as in D5.3. The heights of both antennas were 1.8, <strong>and</strong> 2.5 m,<br />

respectively.<br />

The measurements were performed in a large room (OFF), <strong>and</strong> entrance foyer (FOY), <strong>and</strong> two corridors<br />

(corr1 was a new building, <strong>and</strong> corr2 was an old building). The linear relati<strong>on</strong>ship between RMS AS <strong>and</strong><br />

DS was found in OFF-LOS case,<br />

τ 0.84*<br />

φ −1.6<br />

(ns) (5.29)<br />

RMS<br />

=<br />

RMS<br />

In some other cases, the relati<strong>on</strong>ships cannot be fitted into linear. However, the spatio-temporal<br />

correlati<strong>on</strong> coefficient can be found in the following table.<br />

It can be seen that in all LOS cases, DS <strong>and</strong> AS have good correlati<strong>on</strong>s, however, in all NLOS cases, the<br />

correlati<strong>on</strong> coefficients are quite small.<br />

In reference [MRA93], the cluster AoAs were found to follow Gaussian distributi<strong>on</strong>, <strong>and</strong> the cluster timeof-arrivals<br />

(TOA) were found to be exp<strong>on</strong>entially distributed.<br />

Page 114 (167)


WINNER D5.4 v. 1.4<br />

5.5.2 Scenario B3<br />

5.5.2.1 Reference data<br />

The measurement data for the large indoor scenarios were gathered with partly support by the WINNER<br />

project within a new lecture hall at the Technische Universität Ilmenau (TUI /.Germany). Measurement<br />

b<strong>and</strong>width <strong>and</strong> centre-frequency were selected to be 120 MHz <strong>and</strong> 5.2 GHz. The BS was mounted at the<br />

height of ~3.8m, whereby the transmit antenna was fixed <strong>on</strong> an automatic track with a length of 2 m<br />

between the tiers of the lecture hall. Different positi<strong>on</strong>s for the transmitter <strong>and</strong> receiver where measured.<br />

During the measurement LOS was dominating the propagati<strong>on</strong> characteristics. Furthermore<br />

measurements were performed where the LOS was obstructed.<br />

5.5.2.2 Publicati<strong>on</strong>s<br />

Large indoor envir<strong>on</strong>ment type MIMO measurements are quite rare. Most of them are not dealing with<br />

the measurement analysis items we are discussing here, but with some phenomena like distributi<strong>on</strong> of the<br />

eigenvalues etc.<br />

In [KHK+01] the indoor picocell SIMO measurements were performed in the transit hall of Helsinki<br />

airport using a spherical array of 32 dual-polarized antenna elements at 2.15 GHz b<strong>and</strong>. B<strong>and</strong>width was<br />

30 MHz. The centre of the array was at height of 1.7 m above ground <strong>level</strong>. The omnidirecti<strong>on</strong>al BS<br />

antenna was elevated at 4.6 m above the floor <strong>level</strong>, <strong>and</strong> located so that the visibility over the hall was<br />

good. The BS-MS distance varied within interval (10, 150) m.<br />

The porti<strong>on</strong> of line-of-sight (LOS) measurements was significant, of the order of 40 %.<br />

The measurement results indicate that the instantaneous cross polarizati<strong>on</strong> power ratio (XPR) is lognormally<br />

distributed. The median, mean <strong>and</strong> the st<strong>and</strong>ard deviati<strong>on</strong> of the XPR were found to be 8.8, 8.7<br />

<strong>and</strong> 5.2 dB respectively.<br />

The median, mean <strong>and</strong> the st<strong>and</strong>ard deviati<strong>on</strong> of the elevati<strong>on</strong> angle are respectively 4.1°/1.9°, 6.0°/3.2°,<br />

<strong>and</strong> 10.3°/12.2° (VP/HP).<br />

In [JXP01], large hall measurements at the Helsinki airport were performed at 5.3 GHz b<strong>and</strong>. B<strong>and</strong>width<br />

was 30 MHz. The TX antenna was at 4.55m <strong>and</strong> the RX antenna was at 1.55 m hight. Distances of up to<br />

200 m were covered.<br />

Figure 5.80: Helsinki airport hall setup<br />

For the LOS case at distances 8-100 m path-loss exp<strong>on</strong>ent <strong>and</strong> mean square error of the path loss are<br />

respectively n = 1.3 <strong>and</strong> STD = 2.0 dB.<br />

For the NLOS case at distances 35 - 200 m path-loss exp<strong>on</strong>ent <strong>and</strong> mean square error of the path loss are<br />

respectively n = 1.9 <strong>and</strong> STD = 2.7 dB.<br />

Mean K factor was 1 dB.<br />

Delay spread had values of 120 ns for the LOS case <strong>and</strong> 180 ns for the NLOS case (90% CDF).<br />

Max excess delay was around 600 ns in the NLOS case <strong>and</strong> 240 ns in the LOS case.<br />

Spatial correlati<strong>on</strong> functi<strong>on</strong> in NLOS situati<strong>on</strong> is shown in the Figure 5.81 <strong>and</strong> frequency correlati<strong>on</strong><br />

functi<strong>on</strong> in NLOS situati<strong>on</strong> is shown in the Figure 5.87, both for distances less than 30 m.<br />

Page 115 (167)


WINNER D5.4 v. 1.4<br />

Figure 5.81: Spatial correlati<strong>on</strong> functi<strong>on</strong> for<br />

NLOS case<br />

Figure 5.82: Frequency correlati<strong>on</strong> functi<strong>on</strong>s for<br />

NLOS case<br />

In the [MET_99] Doppler power spectrums of two big hall measurements are presented in Figure 5.88<br />

<strong>and</strong> Figure 5.89. In the envir<strong>on</strong>ment named Novi3, MS was at the height 1.69 m <strong>and</strong> BS at the 2.34m<br />

height. At the Aalborg internati<strong>on</strong>al airport MS was at the height 1.69 m <strong>and</strong> BS at the 2.53m.<br />

Figure 5.83: Averadge empirical Doppler power<br />

spectrum: Novi3 - recepti<strong>on</strong> hall<br />

Figure 5.84: Averadge empirical Doppler power<br />

spectrum: Aalborg internati<strong>on</strong>al airport<br />

In [LUI99] the results of the COST 259 are presented. Big hall envir<strong>on</strong>ment is named General<br />

Factory/Hall (GFH). Length, width <strong>and</strong> height are 90, 30 <strong>and</strong> 10 m respectively. BS height was 8 m <strong>and</strong><br />

MS height was 1.5 m.<br />

Probability of LOS was 0.5. Narrowb<strong>and</strong> path-loss exp<strong>on</strong>ent was found to be 2.2. Inter cluster delay<br />

spread is 360 ns. Mean XPR is 6 dB. XPR spread is 6 dB.<br />

5.5.3 Scenario B5<br />

Note also that for the feeder scenarios we do not have any data <strong>and</strong> therefore the modelling is based<br />

entirely <strong>on</strong> the literature study. Secti<strong>on</strong> 5.5.3.1 below addresses the Doppler spectrum for fixed<br />

applicati<strong>on</strong>s, while Secti<strong>on</strong> 5.5.3.2 <strong>and</strong> 5.5.3.3 reviews publicati<strong>on</strong>s <strong>on</strong> basic parameters for the “rooftop<br />

to rooftop” <strong>and</strong> “street <strong>level</strong>” to “street-<strong>level</strong>” scenarios.<br />

5.5.3.1 Doppler for stati<strong>on</strong>ary scenarios<br />

In comm<strong>on</strong> for the feeder scenarios studied here is the assumpti<strong>on</strong> that the positi<strong>on</strong> of the transmitter <strong>and</strong><br />

receiver are fixed. In mobile-communicati<strong>on</strong>s temporal variati<strong>on</strong>s are modelled by using a moving<br />

transmitter travel through an envir<strong>on</strong>ment of fixed scatterers. In fixed applicati<strong>on</strong>s the temporal variati<strong>on</strong>s<br />

are induced by the movements of the scatterers. In [TPE02] a theoretical model is built where the change<br />

of phase of scatter between time t <strong>and</strong> t+?t is given by<br />

Page 116 (167)


WINNER D5.4 v. 1.4<br />

f<br />

( γ ) cos( ϕ )<br />

c<br />

4π<br />

∆ t cos<br />

p p<br />

, (5.30)<br />

c<br />

where ϕ p is the angle between the directi<strong>on</strong> of scatterer movement <strong>and</strong> the directi<strong>on</strong> orthog<strong>on</strong>al to the<br />

reflecting surface <strong>and</strong> γ p the reflecti<strong>on</strong> angle. By proper selecti<strong>on</strong> of these angles different Doppler<br />

spectrums may be achieved. The results in [DGM+03] show a very narrow spectrum of <strong>on</strong>ly some 0.07<br />

Hz. In [Erc01] a much higher b<strong>and</strong>width of 5-6 Hz is proposed. We suspect that the higher b<strong>and</strong>width in<br />

[Erc01] is a worst case to account for influence from traffic.<br />

5.5.3.2 Scenario B5a - rooftop-to-rooftop<br />

The references [OBL+02], [PT00], [Dug99], [SDD00], [SCK05] treat scenarios similar to the described<br />

scenarios. In paper [OBL+02] a model based <strong>on</strong> measurements of rooftop-to-rooftop propagati<strong>on</strong> in a<br />

residential scenario at 5 GHz is presented. The transmitter antenna is a dipole <strong>and</strong> the receiver omnidirecti<strong>on</strong>al.<br />

Distances in the range 30-330 meters have been c<strong>on</strong>sidered <strong>and</strong> the LOS is sometimes<br />

obstructed by trees. A path-loss model (isotropic in dB) is derived from the measurements<br />

Loss = 46 .9 + 28log10( d)<br />

+ δ , 30m< d


WINNER D5.4 v. 1.4<br />

Figure 5.85: Comparis<strong>on</strong> of the path-loss <strong>models</strong> of [OBL+02], [PT00], free-space <strong>and</strong> a path-loss<br />

model we obtain from the results in [Dug99].<br />

In [SKE05], roof-top to roof-top MIMO measurements at 5.2 GHz are presented. Four different <strong>link</strong>s with<br />

distances of 210, 55, 180, 116 meter have been measured all with clear LOS. Measurement results include<br />

Doppler, K-factor, delay-spread, power-delay-profile, frequency correlati<strong>on</strong> <strong>and</strong> plots DoA/DoD superresoluti<strong>on</strong><br />

results from two out of the four <strong>link</strong>s. Doppler spreads of around 1 Hz at the 10 dB <strong>level</strong>. This<br />

spectrum seems to be identical in the measurements for all delay comp<strong>on</strong>ents. The K-factors measured<br />

are in the range 9.6 to 17.5 dB. The measured power delay profiles seem to be similar to a direct<br />

comp<strong>on</strong>ent plus exp<strong>on</strong>ential decay with some r<strong>and</strong>omizati<strong>on</strong>. Mean delay-spreads are in the range 6-30<br />

ns. The super-resoluti<strong>on</strong> plots show many comp<strong>on</strong>ents but most of them are very weak. A reas<strong>on</strong>able<br />

guess using <strong>on</strong>ly the plots is a power-weighted RMS delay-spread of 2 degrees.<br />

5.5.3.3 Scenario B5b - street-<strong>level</strong>-to-street-<strong>level</strong><br />

A classical two ray model with ground reflecti<strong>on</strong> results in a so-called breakpoint distance located at a<br />

distance r<br />

b given by<br />

λ<br />

h h<br />

r 4<br />

b m<br />

b<br />

= , (5.34)<br />

where hb<br />

<strong>and</strong> hm<br />

are the heights of the two ends of the <strong>link</strong>, respectively. When the distance between the<br />

two antennas is smaller than r b almost free-space path-loss is experienced. This has been observed in a<br />

number of studies [SBA+02], [OTH00], [SMI+00], [MKA02], [FBR+94] but due to reflecti<strong>on</strong>s from cars<br />

<strong>and</strong> other objects during traffic the actual breakpoint occurs at<br />

4 * (h b – h 0 ) * (h m – h 0 ) / λ (5.35)<br />

where h0<br />

is an effective ground height of typically 1.2-1.6 meters. At 5GHz we thus need 3.5 to 4 meter<br />

high antennas to achieve 380 meter free-space propagati<strong>on</strong>.<br />

In [SBA+02] <strong>and</strong> [Bal02] path loss <strong>and</strong> delay-spread measurements at 1.9 GHz <strong>and</strong> 5.8 GHz are<br />

performed in a scenario similar to what is c<strong>on</strong>sidered here. The transmitter antenna is bic<strong>on</strong>ical <strong>and</strong><br />

mounted six meters above ground in two different locati<strong>on</strong>s. The receiver antenna is omni-directi<strong>on</strong>al<br />

mounted <strong>on</strong> a minivan at 1.7 meters height <strong>and</strong> is mobile. The fading patterns at 1.9 GHz <strong>and</strong> 5.8 GHz are<br />

said to be “remarkably similar” although the measurements were not carried out simultaneously at the two<br />

frequencies. No obvious difference LOS <strong>and</strong> NLOS streets were found in teRMS of the difference in path<br />

loss between the two frequencies. The distributi<strong>on</strong> of the difference between 1.9 GHz <strong>and</strong> 5.9 GHz path<br />

loss (in dB) is said to be modelled well by a Gaussian distributi<strong>on</strong> with st<strong>and</strong>ard deviati<strong>on</strong> 4 dB for both<br />

Page 118 (167)


WINNER D5.4 v. 1.4<br />

transmitter locati<strong>on</strong>s. The mean of the difference in <strong>on</strong>e locati<strong>on</strong> was 12 dB <strong>and</strong> for the other 7 dB. In the<br />

paper the path-loss model of [SMI+00] is found to fit the 1.9 GHz measurements <strong>on</strong> LOS streets. This<br />

model is given by<br />

PL<br />

LOS<br />

= e<br />

−sr<br />

λ<br />

π<br />

⎛<br />

⎜<br />

⎝ 4<br />

2<br />

⎞<br />

⎟<br />

⎠<br />

1<br />

e<br />

r<br />

t<br />

1<br />

+ R e<br />

r<br />

− jkrt − jkr n<br />

m<br />

2<br />

(5.36)<br />

where rt<br />

is the line-of-sight path-length, R is the reflecti<strong>on</strong> coefficient of the road surface, <strong>and</strong> s is the<br />

visibility factor. The variable r rm is the distance via reflecti<strong>on</strong> which is described as<br />

r<br />

(( h − h ) + ( h − h )) 2<br />

2<br />

rm = r +<br />

(5.37)<br />

b 0 m 0<br />

where hb<br />

<strong>and</strong> hm<br />

are the base- <strong>and</strong> mobile-stati<strong>on</strong> heights <strong>and</strong> h0<br />

is an effective surface height which is<br />

different from zero due to reflecti<strong>on</strong>s from cars <strong>and</strong> other obstacles. A best fit to the eighteen LOS streets<br />

was found to be h<br />

0 = 1.2m <strong>and</strong> s = 0.001. The RMS-error from this model in the eighteen LOS streets is<br />

listed in a table. We notice that the breakpoint distance which is based <strong>on</strong> the clearance of the first Fresnel<br />

z<strong>on</strong>e with the parameters of the paper appears at such a short distance as 60 meters. The path-loss curve is<br />

similar to a fourth-order slope bey<strong>on</strong>d the breakpoint. We calculate an average the RMS error to be 7.1<br />

dB from this data. The RMS-delay spread is <str<strong>on</strong>g>report</str<strong>on</strong>g>ed to be 15-20% lower at 5.9 GHz than at 1.9 GHz.<br />

From inspecti<strong>on</strong> of the plots in the paper it appears that for LOS cases the delay-spread is 100-150 ns<br />

quite independently of the frequency.<br />

The paper [SMI+00] presents measurements with the transmitter at a height of 4 meters <strong>and</strong> the receiver<br />

at 2.7 meters in a Japanese residential area at 3.5 GHz. The height of the buildings is <strong>on</strong> average eight<br />

meters <strong>and</strong> is therefore higher than the antennas. If the ground <strong>level</strong>, h<br />

0 , is set to zero then breakpoint<br />

distance appears at 678 meters. The measurements up to 460 meters c<strong>on</strong>firmed that free-space<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s existed. Delay-spreads never exceeded 200 ns for the LOS measurements. The<br />

plotted power-delay profiles for LOS case seemed to show approximately the form of an exp<strong>on</strong>ential<br />

decay plus a direct path.<br />

The paper [MKA02] studies the impact of the traffic intensity in an urban area <strong>on</strong> the effective ground<br />

<strong>level</strong>. In the paper the base-stati<strong>on</strong> height is 4meters <strong>and</strong> the mobile-stati<strong>on</strong> height 1.6 meter or 2.7 meter.<br />

Measurements are d<strong>on</strong>e at 3.35, 8.45 <strong>and</strong> 15.75 GHz. The effective ground <strong>level</strong> is estimated to about 0.5<br />

meter during night-time <strong>and</strong> 1.4meter during daytime. The paper also presents RMS-delay-spread values<br />

versus path loss during night-time. From these figures we deduce that it is less than 200 ns at midnight<br />

before the breakpoint distance. Bey<strong>on</strong>d the breakpoint a 3.6 to 4.6 path-loss slope is observed. The<br />

st<strong>and</strong>ard deviati<strong>on</strong> around the mean value seems to be about ±5 dB before the breakpoint <strong>and</strong> ±10 dB<br />

after the breakpoint. A formula for the delay-spread is fitted to the data as<br />

s<br />

[ ns] exp( β )<br />

= , (5.38)<br />

where β is 0.050 during day-time <strong>and</strong> 0.049 during night time. The sample-points used for the fitting of<br />

this formula c<strong>on</strong>tain measurements at 3.35, 8.45 <strong>and</strong> 15.75 GHz. The paper does not state any variati<strong>on</strong><br />

with frequency.<br />

In [FBR+94] micro-cell measurements at 1900 MHz are analyzed for path loss <strong>and</strong> delay-spread with an<br />

MS height of 1.7meter <strong>and</strong> base-stati<strong>on</strong> heights of 3.7, 8.5, <strong>and</strong> 13.3. A path-loss model is fitted where a<br />

free-space propagati<strong>on</strong> law is used up to the breakpoint <strong>and</strong> a 3rd or 4th order law is recommended<br />

bey<strong>on</strong>d that point, shadow fading estimates are in the range 7-8 dB. No effective street-<strong>level</strong> modelling is<br />

used – maybe measurements were d<strong>on</strong>e when there was no traffic? An exp<strong>on</strong>ential dependence between<br />

the path loss <strong>and</strong> delay-spread as in previous reference is also found – however this time the model<br />

c<strong>on</strong>siders the maximum delay-spread. A visual inspecti<strong>on</strong> of the viewgraph of the paper seems to c<strong>on</strong>firm<br />

that typical delay-spreads obey the formula of [MKA02] given above.<br />

In [FDS+94] measurements were d<strong>on</strong>e with the transmitter at 4meter height <strong>and</strong> the receiver <strong>on</strong> the top of<br />

a Van at 2.5 meters. The measurements were d<strong>on</strong>e <strong>on</strong> Southampt<strong>on</strong> University Campus at 1.8 GHz. The<br />

results for LOS show a K-factor between 1 <strong>and</strong> 30 at range of up to the breakpoint. In c<strong>on</strong>trast for the<br />

NLOS measurements the K-factor is between 0 <strong>and</strong> 2.<br />

In [KVV05] polarizati<strong>on</strong> is analyzed in various urban scenarios. The <strong>on</strong>e, most similar to what is<br />

c<strong>on</strong>sidered here, is the urban micro-cell LOS case although the base-stati<strong>on</strong> is c<strong>on</strong>siderably more elevated<br />

than what we are c<strong>on</strong>sidering here <strong>and</strong> the mobile-stati<strong>on</strong> is less (BS height 10 meters, MS height 1.6<br />

meter in the measurements). For this scenario an XPR of around 9 dB is obtained.<br />

In [MIS01] directi<strong>on</strong>al measurements in an urban area with rotating antennas at 8.45 GHz are presented.<br />

The base-stati<strong>on</strong> height is four or eight meters while the mobile-stati<strong>on</strong> height is 3.0 meters. The angle-<br />

PL dB<br />

Page 119 (167)


WINNER D5.4 v. 1.4<br />

spread of the main arriving wave is found to be less than 1.5 degrees in all LOS scenarios. The weaker<br />

paths in the paper seem to be virtually uniform distributed over the entire azimuth.<br />

5.5.4 Scenario C1<br />

5.5.4.1 Scenario definiti<strong>on</strong><br />

In suburban macrocells base stati<strong>on</strong>s are located well above the rooftops to allow wide area coverage.<br />

Buildings are typically low residential houses with <strong>on</strong>e or few floors. Occasi<strong>on</strong>al open areas such as parks<br />

or playgrouds between the houses make the envir<strong>on</strong>ment rather open. Streets have r<strong>and</strong>om orientati<strong>on</strong>s,<br />

<strong>and</strong> no urban-like regular strict grid structure is observed. Vegetati<strong>on</strong> is modest.<br />

5.5.4.2 Reference data<br />

WINNER suburban macrocellular measurements were made in a typical Finnish suburban residential area<br />

with rather wide streets. Buildings in the area are mainly <strong>on</strong>e- or two-storey single ore detached houses.<br />

There are open areas between the buildings, such as playgrounds, parks or small forest areas. BS height in<br />

the measurements was ~25 meters, which is well above the surrounding buildings, <strong>and</strong> at or above the<br />

height of the highest neighbouring trees. Only close to BS there were clear unobstructed LOS areas, <strong>and</strong><br />

further away the MS-BS c<strong>on</strong>necti<strong>on</strong> was obstructed mainly by trees. Deep NLOS c<strong>on</strong>diti<strong>on</strong>s were<br />

achieved at l<strong>on</strong>g MS-BS distances. Maximum measured MS-BS distances were ~1100 meters.<br />

5.5.4.3 Path loss<br />

Propagati<strong>on</strong> studies at 5 GHz frequency b<strong>and</strong> in indoor domestic, office <strong>and</strong> commercial envir<strong>on</strong>ments<br />

have been frequently <str<strong>on</strong>g>report</str<strong>on</strong>g>ed, but wideb<strong>and</strong> outdoor studies at 5 GHz are not as<br />

numerous. In [YTL02] Yacoub, D.; Teich, W.; Lindner, J., „Capacity of Vehicle-<br />

Bridge MIMO Channels”, TD(02)118, COST 273, 5th Management Committee Meeting,<br />

Lisb<strong>on</strong> / Portugal, Sep. 19-20, 2002<br />

[ZKVS02] results for urban, suburban <strong>and</strong> rural envir<strong>on</strong>ments have been <str<strong>on</strong>g>report</str<strong>on</strong>g>ed. In this case the<br />

maximum mobile (MS) to base stati<strong>on</strong> (BS) distances were limited up to 300 meters, which for outdoor<br />

cellular <strong>channel</strong> modeling is not fully representative. Path loss <strong>models</strong> around 5 GHz in residential areas<br />

<strong>and</strong> with BS heights less than 10 meters are <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [SG00] <strong>and</strong> [DRX98]<br />

More studies around 2 GHz frequency have been made. In [EGT+99] results for extensive macrocellular<br />

suburban measurements in US have been <str<strong>on</strong>g>report</str<strong>on</strong>g>ed with BS antenna heights 12…79 meters. Maximum<br />

MS-BS distances up to 8 kms were measured in variety of envir<strong>on</strong>ments c<strong>on</strong>taining both hilly <strong>and</strong> flat<br />

terrains as well as light <strong>and</strong> moderate-to-heavy tree densities. Shadow fading st<strong>and</strong>ard deviati<strong>on</strong> was<br />

found to be in range 5-16 dB, <strong>and</strong> path-loss exp<strong>on</strong>ent was always found to be greater than two. Path loss<br />

exp<strong>on</strong>ent was found to have a str<strong>on</strong>g dependency <strong>on</strong> the BS antenna height <strong>and</strong> the terrain type: the<br />

higher the BS antenna height the smaller the path-loss exp<strong>on</strong>ent.<br />

In [MRA93] <strong>and</strong> [MEJ91] radio propagati<strong>on</strong> differences between 900 <strong>and</strong> 1800 MHz centre-frequencies<br />

have been compared in different envir<strong>on</strong>ments. In both the studies it was found that path-loss values at<br />

900 <strong>and</strong> 1800 MHz were highly correlated, <strong>and</strong> there was no significant difference in distance dependent<br />

behaviour. Theoretical free-space path-loss difference between 900 <strong>and</strong> 1800 MHz is 6 dB, <strong>and</strong> in flat<br />

open areas a value very close to it, 5.7 dB, was obtained [MRA93]. In suburban areas, however, this<br />

difference was increased to 9.3 dB [MRA93], which was explained to be due to higher vegetati<strong>on</strong> in<br />

suburban envir<strong>on</strong>ments, which attenuate 1800 MHz signals more than 900 MHz signals. In [MEJ91] PL<br />

differences between 900 <strong>and</strong> 1800 MHz frequencies were found to be 9…11 dB, i.e. higher than<br />

theoretical free-space loss. In [MRA93] shadow fading st<strong>and</strong>ard deviati<strong>on</strong> was found to be approximately<br />

1 dB higher at 1800 MHz, which agrees with results from Okumura [OOKF68].<br />

In [Xia96] it has been theoretically obtained that for BS antennas above surrounding rooftop <strong>level</strong>s in<br />

suburban <strong>and</strong> urban macrocells, the path loss increases by 38 dB per decade, <strong>and</strong> with frequency by 21 dB<br />

per decade. Path loss decreases with the bases stati<strong>on</strong> antenna height, with respect to the average rooftop<br />

<strong>level</strong>, by 18 dB per decade.<br />

Ref. [BBK+02] presents wideb<strong>and</strong> <strong>channel</strong> measurements at 3.7 GHz <strong>and</strong> 20 MHz b<strong>and</strong>widh in moderate<br />

density macrocellular suburban setting outside Illinois. Reported path-loss exp<strong>on</strong>ents are between 2.9 <strong>and</strong><br />

3.4, <strong>and</strong> shadow fading st<strong>and</strong>ard deviati<strong>on</strong>s vary between 5-10 dB. Maximum measured MS-BS distances<br />

were ~ 6 kms.<br />

Effect of vegetati<strong>on</strong> has been studied in [BBK+04] <strong>and</strong> it was shown that tree foliage creates an excess<br />

path-loss of between 3 <strong>and</strong> 7 dBs.<br />

The suburban measurements in the WINNER project were made in a typical Finnish residential area with<br />

a reas<strong>on</strong>ably flat terrain, open streets <strong>and</strong> modest vegetati<strong>on</strong>. Measurements were d<strong>on</strong>e during summer<br />

time, with all trees having their full foliage. Buildings are mainly <strong>on</strong>e- or two-floor single or detached<br />

Page 120 (167)


WINNER D5.4 v. 1.4<br />

houses surrounded by gardens or small yards. There are open areas between buildings, such as<br />

playgrounds, parks or small forest areas. Base stati<strong>on</strong> height during the measurements was ~20…25<br />

meters, which is well above the surrounding buildings, <strong>and</strong> at the height of the tallest nearby trees. MS<br />

height <strong>on</strong> top of a van was ~2 meters. Only close to BS there were clear unobstructed line-of-sight (LOS)<br />

areas, <strong>and</strong> further away MS-BS c<strong>on</strong>necti<strong>on</strong> was obstructed mainly by trees. Deep n<strong>on</strong>-line-of-sight<br />

(NLOS) propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s were achieved at l<strong>on</strong>g MS-BS distances.<br />

Two different BS sites, <strong>on</strong>e of them with two sectors, were chosen, so altogether data from three different<br />

BS sectors were collected for data analysis during different measurement runs. Maximum measured BS-<br />

MS distances were ~1100 meters, <strong>and</strong> individual route length varied between 100…900 meters. More<br />

detailed descripti<strong>on</strong>s <strong>on</strong> measurements can be found in [RKJ05].<br />

The path-loss model for suburban macrocellular envir<strong>on</strong>ment obtained from WINNER measurements<br />

reads as<br />

PL = A + 10 n log10 ( d )<br />

(5.39)<br />

with n = 4.02, <strong>and</strong> A = 27.7. It is valid in distance ranges 50…1000 meters. Similar PL exp<strong>on</strong>ent values<br />

for flat macrocellular suburban envir<strong>on</strong>ment with moderate of high tree density have <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in<br />

[EGT+99] around 2 GHz centre-frequency, <strong>and</strong> PL exp<strong>on</strong>ent values around 2.0-3.3 for LOS <strong>and</strong> 3.5-5.9<br />

for NLOS can be found in [SG00], <strong>and</strong> [DRX98]. However, in some of these cases the BS height, which<br />

is known to have effect <strong>on</strong> the PL behaviour, is lower than in our measurements. In our WINNER<br />

measurement we found the shadow fading comp<strong>on</strong>ent is log-normally distributed with st<strong>and</strong>ard deviati<strong>on</strong><br />

of 6.1 dB.<br />

COST231-Hata path-loss model [Cost231] for suburban macrocells is written as<br />

d<br />

PL = ( 44.9 − 6.55log<br />

10<br />

( hBS<br />

)) log<br />

10<br />

( ) + 45.5 + (35.46 −1.1h<br />

MS<br />

) log<br />

10<br />

( f<br />

c<br />

) −13.82 log<br />

10<br />

( hBS<br />

) + 0. 7h<br />

1000<br />

(5.40)<br />

In above all the distances <strong>and</strong> heights are given in meters, <strong>and</strong> centre-frequency fc is given in MHz. With<br />

h BS = 20 m, h MS = 2 m, <strong>and</strong> f c = 2000 MHz the model becomes<br />

PL = 29.6<br />

+ 36.4 log<br />

10<br />

( d )<br />

(5.41)<br />

Path loss difference due to theoretical free-space propagati<strong>on</strong> is 20*log 10 (5.3/2) = 8.5 dB. In the following<br />

figure 2 GHz COST231-Hata model with free-space path-loss correcti<strong>on</strong> <strong>and</strong> suburban macrocellular<br />

path-loss model obtained from WINNER measurements are shown. We see the difference between<br />

measurements <strong>and</strong> COST231 model is small.<br />

MS<br />

Figure 5.86: Comparis<strong>on</strong> of COST231-Hata model with free-space correcti<strong>on</strong> <strong>and</strong> path loss<br />

obtained from suburban macrocellular PL measurements in Helsinki. The measurement curve is<br />

shown <strong>on</strong>ly up to 1 km, which was the maximum measured range. COST231-Hata model is defined<br />

for distances greater than 1000 m.<br />

5.5.4.4 RMS delay spread<br />

Delay spreads around 2 GHz centre-frequency <strong>and</strong> 5 MHz b<strong>and</strong>width have been <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [AlPM02].<br />

For suburban envir<strong>on</strong>ment with BS height of 12 meters <strong>and</strong> no direct LOS between MS <strong>and</strong> BS <str<strong>on</strong>g>report</str<strong>on</strong>g>ed<br />

Page 121 (167)


WINNER D5.4 v. 1.4<br />

delay spread values typically vary between 200…800 ns, the median being around 350 ns. Log-normal<br />

distributi<strong>on</strong> was found to give a good fit to the measured delay spread distributi<strong>on</strong>. In typical urban<br />

envir<strong>on</strong>ments delay spreads were found to decrease with increasing BS antenna height.<br />

In [WHL+93], RMS delay spread distributi<strong>on</strong>s were compared in different envir<strong>on</strong>ments at 900 MHz <strong>and</strong><br />

1900 MHz centre-frequencies. With both frequencies the used chip rate was 10 MHz, <strong>and</strong> data was<br />

recorded simultaneously with both the frequencies. It was seen that propagati<strong>on</strong> behaviour in teRMS of<br />

RMS delay spread was very similar with both the centre-frequencies in semirural, suburban <strong>and</strong> urban<br />

cells.<br />

Delay spread characteristics for 3.7 GHz centre-frequency with 20 MHz b<strong>and</strong>width are given.<br />

Measurements were made in suburban areas outside Chicago, where also some distant high-rise buildings<br />

were in the envir<strong>on</strong>ment. BS height was 49 meters, <strong>and</strong> MS was installed at 2.7 meters above the ground.<br />

15 dB dynamics criteri<strong>on</strong> from the max peak power was used in calculating delay spreads. Median delay<br />

spread values for LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s were 240 <strong>and</strong> 360 ns, respectively. The<br />

combined delay spread was found to be 300 ns. As for number of rays, defined as local maxima of<br />

(instantaneous) power delay profiles, 90 percentile value of the cdf for LOS, NLOS <strong>and</strong> combined data<br />

were 3, 8 <strong>and</strong> 7 rays, respectively.<br />

In our WINNER measurements typical delay spreads were of the order of 13…125 ns, which are<br />

c<strong>on</strong>siderably smaller values than <str<strong>on</strong>g>report</str<strong>on</strong>g>ed by [AlPM02]. One reas<strong>on</strong>s for the difference is the higher BS<br />

antenna positi<strong>on</strong>. Rms delay spreads have often been <str<strong>on</strong>g>report</str<strong>on</strong>g>ed to show log-normal distributi<strong>on</strong>, as<br />

summarized for example in [GEYC]. However, instead of Gaussian, we have fitted a gumbel distributi<strong>on</strong><br />

to log10(DS), which shows a better match.<br />

5.5.4.5 Angle-spreads<br />

Azimuth spreads at BS have been given in [Pa03] for rural <strong>and</strong> suburban envir<strong>on</strong>ments at 2 GHz centrefrequency<br />

<strong>and</strong> 10 MHz b<strong>and</strong>width. The mean angle-spread of 2 degrees was found, with st<strong>and</strong>ard<br />

deviati<strong>on</strong> of 2 degrees. In urban areas they have been <str<strong>on</strong>g>report</str<strong>on</strong>g>ed to be 14 degrees <strong>and</strong> 5 degrees,<br />

respectively. The difference between the geometrical directi<strong>on</strong> of the mobile <strong>and</strong> the directi<strong>on</strong> of<br />

maximum received power was modeled as Gaussian. The st<strong>and</strong>ard deviati<strong>on</strong> is about 16 degrees near the<br />

base stati<strong>on</strong> <strong>and</strong> decreased to 8 degrees far away in urban envir<strong>on</strong>ment. In rural <strong>and</strong> suburban is much<br />

smaller, 2.7 degrees when distance is below 2 kms <strong>and</strong> above this decreases to 1.7 degrees. Mobile<br />

azimuth spreads were <str<strong>on</strong>g>report</str<strong>on</strong>g>ed as 35 degrees in suburban, <strong>and</strong> 20 degrees in rural envir<strong>on</strong>ment.<br />

5.5.5 Scenario C2<br />

5.5.5.1 Scenario definiti<strong>on</strong><br />

In urban macro cells base stati<strong>on</strong>s are located above roof tops to allow wide area coverage. Typical<br />

buildings comprise several floors (> 4) <strong>and</strong> street grids often form reguiar grid. Vegetati<strong>on</strong> is modest if<br />

any, <strong>and</strong> streets are occupied with pedestrian <strong>and</strong> vehicular traffic.<br />

5.5.5.2 Reference data<br />

Macrocellular data measured outside WINNER-project in Helsinki city centre at 5.3 GHz centrefrequency<br />

<strong>and</strong> 100 MHz chip rate was used for C2 parameter extracti<strong>on</strong>. The BS height in the<br />

measurements was ~40 meters, which is above the nearby surrounding buildings. Typical building height<br />

in the area was 4-7 stories. The measured data c<strong>on</strong>sists mostly NLOS or OLOS routes, but also some LOS<br />

secti<strong>on</strong>s in vicinity of the BS.<br />

5.5.5.3 Path loss<br />

COST231-Hata path-loss model [Cost231] is valid in frequency range from 1500-2000 MHz, BS-MS<br />

distances > 1 km, BS heights 30-200 meters <strong>and</strong> MS heights 1-10 meters. For urban macrocells the model<br />

is written as<br />

d<br />

PL = ( 44.9 − 6.55log<br />

10<br />

( hBS<br />

)) log<br />

10<br />

( ) + 48.5 + (35.46 −1.1h<br />

MS<br />

)log<br />

10<br />

( f<br />

c<br />

) −13.82 log<br />

10<br />

( hBS<br />

) + 0. 7h<br />

1000<br />

(5.42)<br />

In above all the distances <strong>and</strong> heights are given in meters, <strong>and</strong> centre-frequency f c is given in MHz. With<br />

h BS = 35 m, h MS = 2 m, <strong>and</strong> f c = 2000 MHz the model becomes<br />

PL = 31.0<br />

+ 34.8log<br />

10<br />

( d )<br />

(5.43)<br />

The path-loss model obtained from 5.3 GHz macrocellular NLOS measurements in Helsinki city centre is as follows:<br />

PL = 53.5<br />

+ 28.3log<br />

10<br />

( d),<br />

σ = 5.7<br />

(5.44)<br />

MS<br />

Page 122 (167)


WINNER D5.4 v. 1.4<br />

It is valid in distance ranges 100…2000 meters. BS <strong>and</strong> MS heights were 35 <strong>and</strong> 2 meters, respectively.<br />

COST231-Hata model has not been originally designed to 5 GHz frequency range. Path loss difference<br />

due to theoretical free-space propagati<strong>on</strong> is 20*log 10 (5.3/2) = 8.5 dB. In the following figure 2 GHz<br />

COST231-Hata model with free-space path-loss correcti<strong>on</strong> <strong>and</strong> macrocellular NLOS path-loss model<br />

obtained from Helsinki measurements are shown.<br />

Figure 5.87: Comparis<strong>on</strong> of COST231-Hata model with free-space correcti<strong>on</strong> <strong>and</strong> path loss<br />

obtained from urban macrocellular PL measurements in Helsinki. The measurement curve is<br />

shown <strong>on</strong>ly up to 2 kms, which was the maximum measured range.<br />

It is seen that due to different PL exp<strong>on</strong>ents the differences between PL predicti<strong>on</strong>s increase with<br />

increasing MS-BS distance. At 2 kms the difference is ~7 dB. The few <str<strong>on</strong>g>report</str<strong>on</strong>g>ed outdoor PL<br />

measurements around 5 GHz frequency range show that typical PL exp<strong>on</strong>ents in urban LOS areas are<br />

close to free-space propagati<strong>on</strong> exp<strong>on</strong>ent 2 [YMI+04], [YTL02] Yacoub, D.; Teich, W.;<br />

Lindner, J., „Capacity of Vehicle-Bridge MIMO Channels”, TD(02)118, COST 273, 5th<br />

Management Committee Meeting, Lisb<strong>on</strong> / Portugal, Sep. 19-20, 2002<br />

[ZKVS02], [SG00], <strong>and</strong> for NLOS the <str<strong>on</strong>g>report</str<strong>on</strong>g>ed values vary between 3.5 <strong>and</strong> 5.8 [YMI+04], [YTL02]<br />

Yacoub, D.; Teich, W.; Lindner, J., „Capacity of Vehicle-Bridge MIMO Channels”,<br />

TD(02)118, COST 273, 5th Management Committee Meeting, Lisb<strong>on</strong> / Portugal, Sep. 19-20,<br />

2002<br />

[ZKVS02], [SG00]. In these cases the maximum MS-BS distances have been < 1000 meters.<br />

Path losses <strong>and</strong> delay spreads between 430 <strong>and</strong> 5750 MHz frequencies have been compared in [Pa05].<br />

Data was collected simultaneously at the same measurement points in multiple envir<strong>on</strong>ments, <strong>and</strong> the<br />

chip rate at each centre-frequency was 100 MHz. Measurements were made in urban envir<strong>on</strong>ment in<br />

Denver, which covered a combinati<strong>on</strong> of urban high-rise, urban low-rise <strong>and</strong> line-of sight propagati<strong>on</strong><br />

paths. BS was installed <strong>on</strong> top of a five floor building at 17 meters, <strong>and</strong> maximum measured distances in<br />

the case were up to 5 km. It was observed that line-of sight near the BS (100…300 meters) the path-loss<br />

exp<strong>on</strong>ent was close to 2. In regi<strong>on</strong>s where radio paths become obstructed the path-loss exp<strong>on</strong>ents were<br />

increasing, <strong>and</strong> they also showed frequency dependency: path-loss exp<strong>on</strong>ent increased from 4.3 to 5.4<br />

between 430 <strong>and</strong> 5750 MHz. The shadow fading was normally distributed, <strong>and</strong> ranged between 3 <strong>and</strong> 6<br />

dB.<br />

In [MRA93] <strong>and</strong> [MEJ91] radio propagati<strong>on</strong> differences between 900 <strong>and</strong> 1800 MHz centre-frequencies<br />

have been compared in different envir<strong>on</strong>ments. In both the studies it was found that path-loss values at<br />

900 <strong>and</strong> 1800 MHz were highly correlated, <strong>and</strong> there was no significant difference in distance dependent<br />

behaviour.<br />

Path loss <strong>and</strong> delay spread results from Japanese urban metropolitan envir<strong>on</strong>ment at 3 GHz, 8 GHz, <strong>and</strong><br />

15 GHz frequencies are <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [OTTH01]. The average building heights in were 20…30 meters, <strong>and</strong><br />

BS height both clearly above (macrocell) <strong>and</strong> at rooftop <strong>level</strong> (microcell) were measured. In the<br />

measurements power delay profiles were recorded simultaneously for each of the three frequencies.<br />

Shadow fading st<strong>and</strong>ard deviati<strong>on</strong>s did not show c<strong>on</strong>siderable differences between frequencies, but values<br />

in the range 5…10 dB were obtained. Path loss frequency dependency was found to directly follow freespace<br />

characteristics, i.e. 20log(f). Path loss exp<strong>on</strong>ents were not <str<strong>on</strong>g>report</str<strong>on</strong>g>ed.<br />

Page 123 (167)


WINNER D5.4 v. 1.4<br />

A path-loss model for <strong>system</strong> simulati<strong>on</strong>s is needed for c<strong>on</strong>siderably greater distances than generally<br />

<str<strong>on</strong>g>report</str<strong>on</strong>g>ed in literature, <strong>and</strong> obtained from Helsinki PL measurements. Therefore, based <strong>on</strong> the widely<br />

varying informati<strong>on</strong> available from existing literature <strong>on</strong> propagati<strong>on</strong> at 5 GHz frequency range, we<br />

propose to use 2 GHz COST231-Hata model with free-space correcti<strong>on</strong> to model path loss around 5 GHz.<br />

A reas<strong>on</strong>able resemblance was achieved at least within 2 kilometer range with urban macrocellular<br />

measurements in Helsinki.<br />

5.5.5.4 RMS delay spread<br />

Ref. [WHL94] presents wideb<strong>and</strong> propagati<strong>on</strong> measurements taken in the 1850-1990 MHz b<strong>and</strong> with 10<br />

MHz chip rate in flat rural, hilly rural <strong>and</strong> urban high-rise envir<strong>on</strong>ments. In flat rural scenario typical<br />

delay spreads are of the order of ~100 ns, whereas in urban high-rise cells median delay spread was ~700<br />

ns. Typical delay spread values of 800-1200 ns in urban macrocellular envir<strong>on</strong>ments at 1.8 GHz are<br />

<str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [AlPM02].<br />

In comparis<strong>on</strong> of path losses <strong>and</strong> delay spreads between 430 <strong>and</strong> 5750 MHz frequencies [Pa05] it was<br />

found that the delay spread decreased at higher frequencies: in urban macrocellular envir<strong>on</strong>ment the<br />

median delay spread decreased from 700 to 300 ns. Smaller delay spreads at higher frequencies indicate<br />

reflected signals were attenuated <strong>and</strong> fell below the 20 dB cutoff used for delay spread calculati<strong>on</strong>s. Also<br />

results <strong>on</strong> microcellular envir<strong>on</strong>ment <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [Bul03] show that RMS delay spreads were c<strong>on</strong>sistently<br />

lower at 6 GHz compared to those at 2 GHz by factor of 0.86.<br />

Somewhat c<strong>on</strong>tradictory c<strong>on</strong>clusi<strong>on</strong>s have been <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [OTTH01], which presents path-loss <strong>and</strong><br />

delay spread results from Japanese urban metropolitan envir<strong>on</strong>ment at 3 GHz, 8 GHz <strong>and</strong> 15 GHz<br />

frequencies. In the measurements power delay profiles were recorded simultaneously for each of the three<br />

frequencies in micro- <strong>and</strong> macrocellular scenarios. For multipath characterizati<strong>on</strong> 50 MHz b<strong>and</strong>width was<br />

used, <strong>and</strong> 15 dB dynamic range was applied in delay spread calculati<strong>on</strong>s. Typical (median) RMS delay<br />

spread values were ~100 ns, <strong>and</strong> maximum excess delays (50%) ~300 ns. Differences between<br />

frequencies were found to be very small.<br />

The following table summarizes the RMS delay spread <strong>and</strong> maximum excess delay statistics extracted<br />

from urban macrocellular measurements in Helsinki city centre. Dynamic range of 20 dB was used.<br />

10% 50% 90%<br />

σ τ [ns] 85 265 532<br />

Max excess delay [ns] 575 2210 3490<br />

5.5.5.5 Angle-spread<br />

In [PLN+99] directi<strong>on</strong>al wideb<strong>and</strong> <strong>channel</strong> measurements at 2.1 GHz centre-frequency <strong>and</strong> 50 MHz<br />

b<strong>and</strong>width in urban <strong>and</strong> suburban areas have been performed. It was found that in urban areas a BS<br />

antenna installed at lamppost <strong>level</strong> lead to more severe azimuth spread than a BS at rooftop <strong>level</strong>.<br />

Correlati<strong>on</strong> between angle-spread <strong>and</strong> delay spread was low. In urban city envir<strong>on</strong>ment the macrocellular<br />

BS positi<strong>on</strong> was at 25 meters, which is slightly above surrounding rooftop <strong>level</strong>s. BS-MS distances ranges<br />

from 20 to 360 meters. In suburban envir<strong>on</strong>ment with low residential wooden houses the BS height was 7<br />

meters, which was around the rooftop <strong>level</strong> of most the surrounding buildings. In this scenario BS-MS<br />

distances were 50…510 meters. Typical azimuth spread values (50 precentile value in cdf) in urban<br />

macrocellular envir<strong>on</strong>ment were 7.6…11.8 degrees, with mean value of 9.9 degrees. For the same<br />

envir<strong>on</strong>ment typical delay delay spreads were 20…92 ns, with mean value of 56 ns. In suburban<br />

measurements azimuth spread values 12.9…18.4 degrees with mean value of 15 degrees were obtained.<br />

Corresp<strong>on</strong>ding delay spread values were 45…233 ns, with mean 119 ns.<br />

In [KRB00] angle power distributi<strong>on</strong>s at the MS were measured in urban macrocellular envir<strong>on</strong>ment in<br />

Paris at 890 MHz. It was found that street cany<strong>on</strong>s force the l<strong>on</strong>g-delayed waves to come from street<br />

directi<strong>on</strong>s, but street crossings can cause additi<strong>on</strong>al signal comp<strong>on</strong>ents. For smaller delays local scatterers<br />

c<strong>on</strong>tribute to power spectra. Propagati<strong>on</strong> over the roofs was significant: typically 65% of energy was<br />

incident with elevati<strong>on</strong> angles larger than 10 degrees. In 1.8 GHz measurements <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [AlPM02]<br />

median angle-spreads at BS are in the range 8-13 degrees.<br />

In [KSL+02] elevati<strong>on</strong> angle distributi<strong>on</strong>s at the mobile stati<strong>on</strong> in different radio propagati<strong>on</strong><br />

envir<strong>on</strong>ments have been <str<strong>on</strong>g>report</str<strong>on</strong>g>ed at 2.15 GHz centre-frequency. Results show that in n<strong>on</strong>-line-of-sight<br />

situati<strong>on</strong>s, the power distributi<strong>on</strong> in elevati<strong>on</strong> has a shape of a double-sided exp<strong>on</strong>ential functi<strong>on</strong>, with<br />

different slopes in the negative <strong>and</strong> positive sides of the peak. The slopes <strong>and</strong> the peak elevati<strong>on</strong> angle<br />

depend in the envir<strong>on</strong>ment <strong>and</strong> BS antenna height. In urban macrocells mean elevati<strong>on</strong> angles of arrival<br />

are ~7…14 degrees, with st<strong>and</strong>ard deviati<strong>on</strong>s of 12…18 degrees.<br />

Page 124 (167)


WINNER D5.4 v. 1.4<br />

5.5.6 Scenario D1<br />

5.5.6.1 Path-loss<br />

5.5.6.1.1 LOS path-loss<br />

The basic theoretical equati<strong>on</strong> for LOS path-loss is<br />

where d is the distance between BS <strong>and</strong> MS <strong>and</strong> A <strong>and</strong> B are c<strong>on</strong>stants.<br />

PL = A + B*log 10 (d) (5.45)<br />

Normally, A <strong>and</strong> B are near the free-space values. For example, in our measurements at 5.25 GHz the<br />

values were: A = 41.8 <strong>and</strong> B = 22.<br />

The model in (5.34) can be extended until so called break-point distance, which depends <strong>on</strong> the wavelength<br />

? <strong>and</strong> base stati<strong>on</strong> <strong>and</strong> mobile stati<strong>on</strong> antenna heights, h BS <strong>and</strong> h MS respectively [28].<br />

d BP = 4 · h BS · h MS / ? (5.46)<br />

where h BS is the height of the base stati<strong>on</strong>, h MS is the height of the mobile stati<strong>on</strong>, <strong>and</strong> ? is the wave length<br />

at f c .<br />

After this break-point, the loss is proporti<strong>on</strong>al to another, greater path-loss exp<strong>on</strong>ent. By flat earth theory,<br />

this exp<strong>on</strong>ent should be 4, but in practice it can be also greater. The model is based <strong>on</strong> the assumpti<strong>on</strong><br />

about two rays arriving at the receiver antenna, <strong>on</strong>e direct ray, the other <strong>on</strong>e reflected from the flat earth.<br />

This model is also called two-ray model.<br />

The model can be written in the form [28]:<br />

PL = A + B log 10 (d), d d BP (5.48)<br />

where C = 10 n, <strong>and</strong> n is the path-loss exp<strong>on</strong>ent for the distances greater than the break-point distance.<br />

Other c<strong>on</strong>stants are as given above.<br />

About LOS path-loss, there is a statement in [13] about trials in an rural envir<strong>on</strong>ment that show that the<br />

two-ray model woks well there. For the urban microcellular envir<strong>on</strong>ment it has been modified slightly to<br />

make it agree with the measurement results. Measurements for the two-ray modeling were <str<strong>on</strong>g>report</str<strong>on</strong>g>ed at 1.9<br />

GHz <strong>and</strong> cover the range of 0 to 1800 m with antenna heights of 6 m (BS) <strong>and</strong> 1.7 m (MS). With these<br />

values the distance of 1800 m is far bey<strong>on</strong>d the break-point distance.<br />

Also, [AlPM02] shows results for LOS c<strong>on</strong>diti<strong>on</strong>s, where the path-loss exp<strong>on</strong>ent is near 2 with st<strong>and</strong>ard<br />

deviati<strong>on</strong> of 6.9 dB. The behavior of the path loss is thus like in free-space. The envir<strong>on</strong>ment is called<br />

residential. It can be classified also suburban. Measurements were c<strong>on</strong>ducted using 100 MHz b<strong>and</strong>width.<br />

For NLOS c<strong>on</strong>diti<strong>on</strong>s, the path-loss exp<strong>on</strong>ent was 3.5 <strong>and</strong> the st<strong>and</strong>ard deviati<strong>on</strong> was 9.5 dB.<br />

In the reference [Zha02], based <strong>on</strong> measurements performed at 5.3 GHz with RF BW 30 MHz, <strong>and</strong> omnidirecti<strong>on</strong>al<br />

antennas, the path-loss <strong>models</strong>, excess delay <strong>and</strong> RMS delay-spread statistical values were<br />

obtained. In the rural envir<strong>on</strong>ments, the transmitter was placed at a hill with a mast, the total height was<br />

55 m from ground <strong>level</strong>, the height of the mobile stati<strong>on</strong> was 2.5 m <strong>on</strong> top of a car. The path-loss equati<strong>on</strong><br />

is expressed as follows:<br />

5.5.6.1.2 NLOS path-loss<br />

PL ( dB) = 21.8 + 33log<br />

10<br />

( d ) , σ = 3. 7 dB (5.49)<br />

The model has been based partly <strong>on</strong> measurements <strong>and</strong> partly <strong>on</strong> literature. There are numerous path-loss<br />

<strong>models</strong> for lower frequencies than 5 GHz, <strong>and</strong> especially for urban <strong>and</strong> suburban envir<strong>on</strong>ments. For the<br />

rural envir<strong>on</strong>ment at 5 GHz there are not many results available. One alternative is to use results of lower<br />

frequencies, e.g. 2 GHz <strong>and</strong> translate them to 5 GHz. This can be justified with results presented in the<br />

paragraph 5.5.6.1.4, which show that the path-loss properties at 2 <strong>and</strong> 6 GHz are closely related. Mean<br />

difference was found to be 8.1 dB, when the difference due to the free-space losses should be 9.7 dB.<br />

Although the results were measured in an urban envir<strong>on</strong>ment, they suggest that the rural 2 GHz path-loss<br />

model can be c<strong>on</strong>verted to 5 GHz by increasing the path loss with the difference in the free-space losses.<br />

One potential <strong>channel</strong> model for the D1 scenario is the COST-231-Hata model [26] that is c<strong>on</strong>verted for 5<br />

GHz. COST-231-Hata model for sub-urban envir<strong>on</strong>ment is<br />

d<br />

PL = ( 44.9 − 6.55log10 ( hBS<br />

))log10<br />

( ) + 45.5 + (35.46−1.1h<br />

MS<br />

)log10(<br />

f<br />

c<br />

) −13.82log10<br />

( hBS<br />

) + 0. 7hMS<br />

(5.50)<br />

1000<br />

where<br />

h BS = the height of the base stati<strong>on</strong><br />

h MS = the height of the mobile stati<strong>on</strong> (m)<br />

f c = the centre-frequency (MHz)<br />

Page 125 (167)


WINNER D5.4 v. 1.4<br />

d = distance between BS <strong>and</strong> MS (m).<br />

The original model is applicable up to 2 GHz, <strong>and</strong> in the distance range 1 – 20 km.<br />

It should be noted that COST-231-Hata model is not a NLOS model, but it does not make difference<br />

between the propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. However, at l<strong>on</strong>ger distances the propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s are mostly<br />

NLOS. So it can be applied for NLOS in spite of the afore-menti<strong>on</strong>ed fact.<br />

In reference [Zha02] the path-loss model for NLOS was<br />

PL ( dB) = −27.8<br />

+ 59log10 ( d)<br />

, σ =1. 9 dB (5.51)<br />

The parameters differ quite much from the values found out in this campaign. One reas<strong>on</strong> is the hilly<br />

terrain, the other could be the relatively small number of routes measured.<br />

One interesting empirical <strong>channel</strong> model for suburban envir<strong>on</strong>ment is [Erc99]. The suburban envir<strong>on</strong>ment<br />

is divided to three sub-envir<strong>on</strong>ments according to the tree density <strong>and</strong> the height variati<strong>on</strong> of the<br />

envir<strong>on</strong>ment. One of these envir<strong>on</strong>ments could well be applied to the rural envir<strong>on</strong>ment. This model is<br />

created for 1. 9 GHz. With the same reas<strong>on</strong>ing as afore it can be also extended to 5 GHz. The model is<br />

PL = 20 log10 (4p 100/?) + 10 (a – b hBS + c/hBS) log10 (d/100) (5.52)<br />

where the parameters a, b <strong>and</strong> c may get three sets of values depending <strong>on</strong> the envir<strong>on</strong>ment (see below)<br />

<strong>and</strong> the other parameters are the same as in the previous formula. The model is applicable for distances<br />

100 m – 20 km.<br />

The parameter set that is closest to the rural envir<strong>on</strong>ment in Tyrnävä is the <strong>on</strong>e for low tree density <strong>and</strong><br />

flat terrain. Then the c<strong>on</strong>stants are: a = 3.6, b = 0.005 <strong>and</strong> c = 20. The complete model defines also a<br />

distance dependent st<strong>and</strong>ard deviati<strong>on</strong> for the path loss, but it is not discussed further in this document.<br />

5.5.6.1.3 Probability of LOS<br />

There are few references about the LOS probability. Especially for the rural envir<strong>on</strong>ment where relatively<br />

high BS antenna heights are likely to be used. Reference [30] discusses LOS probability in a peer to peer<br />

<strong>and</strong> ad-hoc envir<strong>on</strong>ment, where the antenna heights are low. The result is that the LOS probability<br />

decreases from <strong>on</strong>e to zero approximately in the interval 30 m to 300 m. No formula for the decay is<br />

given.<br />

In [SCM] there is a model given for the LOS probability. The probability formula proposed is<br />

p LOS (d) = (d 0 - d) / d 0 , 0 < d


WINNER D5.4 v. 1.4<br />

5.6 Interpretati<strong>on</strong> of results<br />

5.6.1 Path-loss<br />

5.6.1.1 Scenario A1<br />

5.6.1.1.1 Proposed path-loss model<br />

The results for path loss <strong>and</strong> shadowing have been summarized in Table 5.45.<br />

Table 5.45: Path-loss <strong>and</strong> shadowing characteristics in the indoor envir<strong>on</strong>ment.<br />

Indoor LOS (c-c) NLOS (r–c)<br />

PL at 5.25 GHz<br />

SF st<strong>and</strong>ard deviati<strong>on</strong> at<br />

5.25 GHz<br />

46.8 +18.7 log10(d),<br />

d >1m<br />

s = 3.1 dB<br />

PL (d)= 38.8+36.8 log10(d)<br />

d >5m<br />

s = 3.5 dB<br />

5.6.1.1.2 Probability of LOS<br />

The probability of line-of-sight (LOS) propagati<strong>on</strong> vs. distance is a functi<strong>on</strong> we denote the pLOS<br />

functi<strong>on</strong>. For scenario A1, this characteristic can be derived analytically because the geometry of the<br />

scenario is known exactly.<br />

A simple ad-hoc fit of the derived pLOS functi<strong>on</strong> is given as:<br />

p LOS (d) = 1 – (1 – x 3 ) 1/3 * (1 – 5 / 50), (5.55)<br />

where x = 1 - log 10 (d / 2.5) / log 10 (100 / 2.5).<br />

5.6.1.2 Scenario B5a<br />

We use the path-loss model of [PT00] as given below. We assume that it is applicable from 30 meters to<br />

2km distance with a correcti<strong>on</strong> term for frequency, i.e.<br />

( fc<br />

/ 2.5GHz) + 23.5log10( d + δ<br />

slow<br />

Loss = 36 .5 + 20log10<br />

) , 300 m < d < 8 km (5.56)<br />

We note that for the 30m to 300m range (for which [PT00] presents no measurements), the path-loss<br />

model almost coincides with cases in [Dug99] with the smallest path-loss. These cases are probably the<br />

<strong>on</strong>es with the least obstructed LOS. The model of [PT00] is for 2.5GHz. For other centre-frequencies, fc,<br />

it seems reas<strong>on</strong>able to translate by using the free-space frequency dependence as the propagati<strong>on</strong> scenario<br />

(e.g. path-loss exp<strong>on</strong>ent) is close to free-space propagati<strong>on</strong>.<br />

The shadow fading is Gaussian with mean zero <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> σ SF = 3.4 dB based <strong>on</strong> [PT00].<br />

5.6.1.3 Scenario B5b<br />

Based <strong>on</strong> the observati<strong>on</strong> of numerous papers that the path loss follows approximately a free-space law<br />

before the breakpoint distance we will assume that loss is given by<br />

Loss<br />

( r) = −20log( /( 4πr)<br />

) + σ free + δ free,<br />

r ≤ rb<br />

λ , d < 1 km (5.57)<br />

where the first part is recognized as the free-space path-loss, d free is a Gaussian distributed r<strong>and</strong>om<br />

variable (shadow fading), with st<strong>and</strong>ard deviati<strong>on</strong> s free = 3 dB. This path-loss model (i.e., (4.10)) can be<br />

used for a maximum distance of 1 km. Many measurements seem to show path loss lower than the freespace<br />

before the breakpoint <strong>and</strong> indeed it can happen due to c<strong>on</strong>structive multi-path. However, to avoid<br />

producing overly optimistic results the extra loss s free has been introduced such that the probability of a<br />

lower than free-space loss is <strong>on</strong>ly some 14%. The breakpoint distance is calculated as<br />

( h − h )( h − h )<br />

b 0 b 0<br />

rb = 4<br />

(5.58)<br />

λ<br />

where we, somewhat pessimistically, set the effective ground <strong>level</strong> h0<br />

to 1.6 meter. For distances larger<br />

than r b we set the path loss to<br />

Loss<br />

( r) = free − 20log10( λ /( 4πrb<br />

)) + 40log( r / rb ) + δ bey<strong>on</strong>d,<br />

r > rb<br />

σ , (5.59)<br />

where the first two teRMS corresp<strong>on</strong>d to the (mean) path-loss at b<br />

r <strong>and</strong> d bey<strong>on</strong>d is a Gaussian shadowfading<br />

term with mean zero <strong>and</strong> st<strong>and</strong>ard deviati<strong>on</strong> 7 dB.<br />

Page 127 (167)


WINNER D5.4 v. 1.4<br />

5.6.1.4 Scenario C1<br />

5.6.1.4.1 LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s<br />

The measured path-loss formula is<br />

where d is the distance.<br />

PL(d) = 41.6 + 23.8 log 10 (d), s = 4 dB (5.60)<br />

The measurement range was limited to 600 m. However, like e.g. in the rural scenario, it is assumed that<br />

the range can be extended to the break-point distance, because the LOS propagati<strong>on</strong> is not very sensitive<br />

to the envir<strong>on</strong>ment. As well the frequency range can be extended to other frequencies. The path-loss<br />

formula can be expressed as<br />

where d = distance<br />

PL(d) = 41.6 + 23.8 log 10 (d), s = 4.0 dB, 20 m < d d BP<br />

The formula above can be adapted for the frequencies between 2000 <strong>and</strong> 6000 MHz by replacing the<br />

c<strong>on</strong>stant 41.6 by a factor<br />

5.6.1.5 Scenario D1<br />

5.6.1.5.1 LOS model<br />

C(f c ) = 33.2 + 20 log10(f c /(2·10 9 )) (5.62)<br />

The path loss is shown in the figure below for the two centre-frequencies 2.45 <strong>and</strong> 5.25 GHz in LOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. The measurements have been c<strong>on</strong>ducted in a 100 MHz b<strong>and</strong>width. The curve for<br />

2.45 GHz c<strong>on</strong>tains also a part that is measured in NLOS (or nearly NLOS) c<strong>on</strong>diti<strong>on</strong>s around 1000 m<br />

distance.<br />

130<br />

path loss (dB)<br />

120<br />

110<br />

100<br />

90<br />

80<br />

70<br />

PL(d) = -105 + 75.0*log10(d), σ = 2.2 dB<br />

PL(d) = 41.8 + 22.0*log10(d), σ = 2.6 dB<br />

5.25 GHz<br />

Free space<br />

2.45 GHz<br />

PL(d) = 38.3 + 21.1*log10(d), σ = 2.9 dB<br />

100 200 500 1000<br />

distance from MS to BS (m)<br />

Figure 5.88: Rural path-loss at 2.45 <strong>and</strong> 5.25 GHz.<br />

The measured model for LOS has been extended from the [D5.3] for l<strong>on</strong>ger ranges, because it has<br />

become evident that a path-loss model for LOS c<strong>on</strong>diti<strong>on</strong>s with l<strong>on</strong>ger BS – MS distances is needed. The<br />

maximum distance for the model should be 10 km. Theoretically the model shown in (5.63) can be<br />

extended until so called break-point distance, which depends <strong>on</strong> the wave-length ? <strong>and</strong> base stati<strong>on</strong> <strong>and</strong><br />

mobile stati<strong>on</strong> antenna heights h BS <strong>and</strong> h MS , respectively. After this break-point the loss is proporti<strong>on</strong>al to<br />

another, greater path-loss exp<strong>on</strong>ent. By flat earth theory this exp<strong>on</strong>ent should be 4, but in practice it can<br />

be also smaller or greater. In practice the break point distance varies around the theoretical value. The<br />

Page 128 (167)


WINNER D5.4 v. 1.4<br />

break-points or the dual-slope behaviour could not be found in our measurements. However, this depends<br />

most probably <strong>on</strong> the r<strong>and</strong>omness of the practical situati<strong>on</strong>: We have decided to take it as part of our rural<br />

LOS model.<br />

For the line of sight (LOS) c<strong>on</strong>diti<strong>on</strong>s the measurements suggest the path-loss equati<strong>on</strong>s:<br />

with the st<strong>and</strong>ard deviati<strong>on</strong> s = 3.5 dB.<br />

PL(d) = 44.6 + 21.5 log 10 (d) (5.63)<br />

The model is based partly <strong>on</strong> our measurements <strong>and</strong> partly <strong>on</strong> the literature research, where we have<br />

adopted the two ray model for distances higher than the break-point. For the LOS envir<strong>on</strong>ment we get<br />

then:<br />

where d = distance<br />

PL(d) = 44.6 + 21.5 log 10 (d), s = 3.5 dB, d d BP<br />

The path losses behave very similarly at the two frequencies. As a matter of fact the mean behaviour is<br />

very near free-space path-loss in LOS c<strong>on</strong>diti<strong>on</strong>s. The formula above can be adapted for the frequencies<br />

between 2000 <strong>and</strong> 6000 MHz by replacing the c<strong>on</strong>stant 44.6 by a factor<br />

C(f c ) = 36.2 + 20 log 10 (f c /(2·10 9 )) (5.65)<br />

Note that the LOS path-loss depends <strong>on</strong> antenna heights <strong>on</strong>ly through the break-point distance.<br />

5.6.1.5.2 NLOS model<br />

The NLOS model is based <strong>on</strong> our measurements which have been fitted to results found in the literature<br />

research. The measured path-loss curve for the NLOS c<strong>on</strong>diti<strong>on</strong>s has the equati<strong>on</strong><br />

with s = 6.7 dB.<br />

PL = 55.8 + 25.1 log 10 (d) (5.66)<br />

This path-loss equati<strong>on</strong> was measured for the BS antenna height 23 m <strong>and</strong> MS antenna height 1.7 m. This<br />

equati<strong>on</strong> will be compared to some theoretical path-loss curves. From the literature we found that mostly<br />

the st<strong>and</strong>ard deviati<strong>on</strong> was a little bit higher than 6.7. Because our measurement was limited, we decided<br />

to use the value 8 dB for the st<strong>and</strong>ard deviati<strong>on</strong>.<br />

The formula (5.66) above can be adapted for the frequencies between 2000 <strong>and</strong> 6000 MHz by replacing<br />

the c<strong>on</strong>stant 55.8 by a factor<br />

C(f c ) = 46.9 + 20 log 10 (f c / (2·10 9 )) · 1.063 (5.67)<br />

= 46.9 + 21.3 log 10 (f c / (2·10 9 ))<br />

The c<strong>on</strong>stant 1.063 is based <strong>on</strong> our finding that the path loss difference in between 5.25 <strong>and</strong> 2.45 GHz for<br />

NLOS c<strong>on</strong>diti<strong>on</strong>s was about 2.5 dB higher than for free space, see paragraph 5.6.9.<br />

After D5.3 the following equati<strong>on</strong> was proposed for the D1 rural NLOS scenario path-loss by the WP5 to<br />

other work-packages. This model is called. COST231-Hata model, originally for urban <strong>and</strong> suburban<br />

envir<strong>on</strong>ments. Slightly simplified it reads as<br />

f<br />

c<br />

d<br />

PL = 20log10 ( ) + [ 44.9 −6.55log10<br />

( hBS<br />

) ] log10<br />

( ) −13.82log10<br />

( hBS<br />

) + 153. 39 (5.68)<br />

2000<br />

1000<br />

where h BS is the height of the base stati<strong>on</strong>, f c is the centre-frequency (MHz), <strong>and</strong> d is the distance between<br />

BS <strong>and</strong> MS (m).<br />

This model <strong>and</strong> another well-known model, Erceg model 1, will be compared to the measured curve.<br />

The modificati<strong>on</strong> of the BS antenna height is probably needed, because the envir<strong>on</strong>ment of the<br />

measurements is extremely flat. This can be compensated by adding 25 m to get an effective BS antenna<br />

height that is greater in flat than in hilly envir<strong>on</strong>ments.<br />

Both <strong>models</strong> work equally well, if the BS antenna height is between, say, 10 <strong>and</strong> 100 m. For higher<br />

antenna heights the curves begin to differ. According to the comparis<strong>on</strong>, the modified COST231-Hata<br />

Page 129 (167)


WINNER D5.4 v. 1.4<br />

model can be used from 100 m to 10 km, although originally the model has been defined for distances<br />

greater than 1 km. The modified Erceg model 1 could be applied as well.<br />

5.6.1.5.3 Unified model<br />

For the overall path-loss in the D1 scenario we get from measurements the formula<br />

PL(d) = 50.4 + 25.8 log 10 (d) (5.69)<br />

The drawback of this model is that it is based <strong>on</strong> relatively few measurements. In additi<strong>on</strong> the LOS<br />

c<strong>on</strong>diti<strong>on</strong> disappeared after quite a small distance from the BS. This depends e.g. <strong>on</strong> the fact that the BSs<br />

were located slightly off the roads for practical reas<strong>on</strong>s. Real BS:s would be located probably in more<br />

beneficial way.<br />

The unified model can be formed also by combining the LOS model <strong>and</strong> the NLOS model using the LOS<br />

probability p LOS (d):<br />

PL(d) = p LOS (d) PL LOS (d) + [1- p LOS (d)] PL NLOS (d) (5.70)<br />

where d is the distance between MS <strong>and</strong> BS. p LOS will be defined in Secti<strong>on</strong> 5.6.1.5.4<br />

The drawback of this model is that the p LOS used here is <strong>on</strong>ly theoretical <strong>on</strong>e. However, it gives<br />

reas<strong>on</strong>able results, <strong>and</strong> is used therefore as basis of comparis<strong>on</strong> of our model <strong>and</strong> the <strong>models</strong> found in the<br />

literature.<br />

5.6.1.5.4 Probability of LOS model<br />

Probability of LOS in the D1 scenario is proposed to be modelled with an exp<strong>on</strong>ential functi<strong>on</strong><br />

P LOS<br />

d<br />

( d)<br />

= exp( − )<br />

(5.71)<br />

where d is the distance between the BS <strong>and</strong> the MS <strong>and</strong> d 0 is the c<strong>on</strong>stant defining the steepness of the<br />

exp<strong>on</strong>ential decay.<br />

Default value for d 0<br />

is proposed to be 1 km. The reas<strong>on</strong> for proposing this model is the following: It is<br />

very near the model for LOS probability defined in [SCM] at small distances. In additi<strong>on</strong> it does not go to<br />

zero at the cell boundary, so that it can be used in the <strong>system</strong>-<strong>level</strong> modelling of interference.<br />

5.6.1.5.5 Model comparis<strong>on</strong><br />

In the figure below there are the measurement based PL curves obtained in the WINNER project<br />

compared to some well-known PL <strong>models</strong> from the literature. In additi<strong>on</strong> there is the free-space loss<br />

curve. In the comparis<strong>on</strong> the base stati<strong>on</strong> antenna height was 24.5 m; mobile antenna height was the<br />

default value 1.7 m.<br />

d 0<br />

Figure 5.89: Comparis<strong>on</strong> of <strong>channel</strong> <strong>models</strong> for a rural envir<strong>on</strong>ment.<br />

Page 130 (167)


WINNER D5.4 v. 1.4<br />

When adjusting the PL curves of the <strong>models</strong> to closely follow the measured over-all curve, the following<br />

acti<strong>on</strong>s were needed:<br />

1. Cost231-Hata: Subtract 15 dB <strong>and</strong> apply h BS = 50 m (instead of the 25 m). I.e. use an effective<br />

BS antenna height h BS + 25 m.<br />

2. Erceg: Apply h BS = 60 m. I.e. use an effective BS antenna height h BS + 35 m.<br />

Both <strong>models</strong> could be used, after these modificati<strong>on</strong>s. The subtracti<strong>on</strong> of 15 dB needed with the<br />

COST231-Hata model is caused by the fact that the model is not originally planned for rural<br />

envir<strong>on</strong>ments, but for urban <strong>and</strong> suburban envir<strong>on</strong>ments.<br />

5.6.2 Power-delay profile<br />

5.6.2.1 Scenario A1<br />

The PDP at a corridor to corridor envir<strong>on</strong>ment is modelled as a decaying exp<strong>on</strong>ential. The measured PDP<br />

has a spike that can be identified coming due to a reflecti<strong>on</strong> from the end of the corridor, see Secti<strong>on</strong><br />

5.4.7.1. In our model we have neglected it, because the delay of the spike depends <strong>on</strong> the locati<strong>on</strong> of the<br />

BS in the corridor. The c<strong>on</strong>stants of the decay have been determined in the same paragraph. It is relatively<br />

easy to extend the model to include the spike, when the locati<strong>on</strong> of the BS is fixed. However, it is not<br />

included in the current model. This is justified by the model simplicity <strong>and</strong> also the relative low <strong>level</strong> of<br />

the measured spike. For the corridor to room envir<strong>on</strong>ment the model fits quite well in the exp<strong>on</strong>ential<br />

model.<br />

5.6.2.2 Scenario B5a<br />

The power-delay profile (of all paths except the direct) is set as exp<strong>on</strong>ential, based <strong>on</strong> the results in<br />

[OBL+02] <strong>and</strong> [SCK05].<br />

5.6.2.3 Scenario B5b<br />

The power-delay profile (of all paths except the direct) is set as exp<strong>on</strong>ential, based <strong>on</strong> the results in<br />

[SMI+00]. A per-path shadow fading of 3 dB is used to obtain some variati<strong>on</strong> in the impulse resp<strong>on</strong>ses.<br />

5.6.3 Delay spread<br />

5.6.3.1 Scenario B5a<br />

The RMS-delay-spread is set to 40 ns, based <strong>on</strong> [PT00]. In order to have a valid model, it requires beamwidths<br />

comparable to those employed in [PT00].<br />

5.6.3.2 Scenario B5b<br />

Based <strong>on</strong> the delay-spread formula in [MAS02] i.e.<br />

s<br />

[ ns] exp( β )<br />

= (5.72)<br />

we select the delay spread to be 30 ns when the path loss is less than 85 dB, 110 ns when the path loss is<br />

between 85 dB <strong>and</strong> 110 dB, <strong>and</strong> finally 380 ns when the path loss is greater than 110 dB. With these<br />

settings the delay-spread used here is a factor 40%-156% of the delay-spread formula of [MAS02] for<br />

path losses up to 137 dB. We call these path-loss intervals range1, range2 <strong>and</strong> range3 <strong>and</strong> different<br />

clustered-delay line <strong>models</strong> will be provided for the three cases.<br />

5.6.3.3 Scenario C1<br />

In the C1 LOS suburban scenario we model the PDP with a single exp<strong>on</strong>ential. It can be seen that another<br />

exp<strong>on</strong>ential cluster could be included in the PDP. For the same reas<strong>on</strong> as in the scenario A1 it was<br />

decided to be neglected.<br />

5.6.3.4 Scenario D1<br />

In the D1 scenario the PDP is best modelled with two decaying exp<strong>on</strong>entials in both LOS <strong>and</strong> NLOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s, i.e. with a dual-slope model. However, also now we model the PDP was decided<br />

to be modelled with a single exp<strong>on</strong>ential in both cases. In the LOS c<strong>on</strong>diti<strong>on</strong>s the first part is modelled<br />

with <strong>on</strong>e spike <strong>and</strong> the sec<strong>on</strong>d part with an exp<strong>on</strong>ential.<br />

In the NLOS c<strong>on</strong>diti<strong>on</strong>s a single slope model is fitted to the measured PDP for simplicity. The fitting is<br />

performed to preserve the modelled RMS-delay spread equal to the measured <strong>on</strong>e.<br />

PL dB<br />

Page 131 (167)


WINNER D5.4 v. 1.4<br />

5.6.4 K-factor<br />

5.6.4.1 Scenario B5a<br />

A static (n<strong>on</strong>-fading) <strong>channel</strong> comp<strong>on</strong>ent is added to the impulse resp<strong>on</strong>se. We select this parameter to be<br />

10 dB. This is based <strong>on</strong> the worst case (smallest value) in [SCK05]. In [OBL+02] a somewhat smaller<br />

average of 2.3 dB is seen but this is probably due to the LOS obstructi<strong>on</strong>s by trees.<br />

5.6.4.2 Scenario B5b<br />

A static (n<strong>on</strong>-fading) <strong>channel</strong> comp<strong>on</strong>ent is added to the impulse resp<strong>on</strong>se. Based <strong>on</strong> [FDS+94] we select<br />

this parameter to be 10 in range1, 2 in range2, <strong>and</strong> 1 in range3 (for a definiti<strong>on</strong> of the ranges see the<br />

secti<strong>on</strong> <strong>on</strong> delay-spread above.)<br />

5.6.5 Cross-polarizati<strong>on</strong> discriminati<strong>on</strong> (XPR)<br />

5.6.5.1 Scenario B5a<br />

The polarizati<strong>on</strong> scrambling (i.e. the power transfer between a transmitted vertically polarized to a<br />

received horiz<strong>on</strong>tally polarized antenna, <strong>and</strong> vice versa) is highly related to reflecti<strong>on</strong> <strong>on</strong> rough surfaces.<br />

This effect should be small in LOS scenarios. A high XPR means that there is little power transfer<br />

between the comp<strong>on</strong>ents. This means that we should be able to use the highest XPR values measured in<br />

[Dug99]. However, in order to avoid overly optimistic results we chose the mean value of [Dug99] i.e.<br />

30dB.<br />

5.6.5.2 Scenario B5b<br />

Based <strong>on</strong> the results in [KVV05] we set the XPR to 9 dB.<br />

5.6.6 Doppler<br />

5.6.6.1 Scenario B5a<br />

The Doppler is modelled by moving the scatterers appropriately. We chose the spectrum of [DGM+03]<br />

since it is assumed to be the most similar to the applicati<strong>on</strong> here.<br />

5.6.6.2 Scenario B5b<br />

We propose the introducti<strong>on</strong> of individual Doppler frequencies similar to the model in [TPE02]. We<br />

select the Doppler model [Erc01] which has somewhat larger Doppler spread than [DGM+03] probably<br />

due to the influence of traffic.<br />

5.6.7 Angle-spread<br />

5.6.7.1 Scenario B5a<br />

Based <strong>on</strong> our visual inspecti<strong>on</strong> of the plots in [SCK05] we set the AoD <strong>and</strong> AoA of the n<strong>on</strong>-direct paths<br />

to be Gaussian with composite power weighted angle-spread of 2 degrees. The ZDSC angle-spread is set<br />

to 0.5 degree.<br />

5.6.7.2 Scenario B5b<br />

Based <strong>on</strong> our visual inspecti<strong>on</strong> of the plots in [MIS01] we set the AoD of all the paths to be uniformly<br />

distributed between 0 <strong>and</strong> 360 degrees. The direct path is aligned with the geometrical angle between the<br />

transmitter <strong>and</strong> receiver. The intra-cluster angle-spread is set to 2 degrees.<br />

5.6.8 Antenna gain<br />

5.6.8.1 Scenario B5<br />

The antenna pattern that can be used in the simulati<strong>on</strong> is specified by<br />

A<br />

( γ )<br />

⎡ ⎛ γ ⎞<br />

=−min⎢12<br />

⎜ ⎟<br />

⎢ ⎝γ<br />

3dB<br />

⎣ ⎠<br />

2<br />

A<br />

,<br />

m<br />

⎤<br />

o<br />

o<br />

⎥, where 180 < φ


WINNER D5.4 v. 1.4<br />

5.6.9 Frequency dependence of the propagati<strong>on</strong> parameters<br />

5.6.9.1 Path-loss <strong>and</strong> shadowing properties<br />

In the Figure 5.90, the path losses at 5.25 GHz <strong>and</strong> 2.45 GHz are compared in a rural LOS envir<strong>on</strong>ment in<br />

the same route. It is obvious that the path-loss functi<strong>on</strong>s can be modeled so that the <strong>on</strong>ly the difference<br />

between them is the difference between the free-space path-losses, i.e. 6.62 dB.<br />

Figure 5.90: Propagati<strong>on</strong> at 2.45 GHz <strong>and</strong> 5.25 GHz in rural LOS c<strong>on</strong>diti<strong>on</strong>s.<br />

In the figure Figure 5.91, the path losses at 2.45 GHz (a) <strong>and</strong> 5.25 GHz (b) are compared in a rural<br />

LOS/NLOS envir<strong>on</strong>ment in the same route. The over-all behaviour of the path loss is almost identical.<br />

a<br />

Figure 5.91: Path-loss at 2.45 GHz (a) <strong>and</strong> 5.25 GHz (b) measured in the same route.<br />

b<br />

It can be seen that the path losses are almost identical except for the difference due to the free-space<br />

losses. However, we have calculated that there is a difference of 1.7 dB in the overall st<strong>and</strong>ard deviati<strong>on</strong><br />

<strong>and</strong> a difference of 1 dB in the NLOS st<strong>and</strong>ard deviati<strong>on</strong> of the path losses between the centre-frequencies<br />

5.25 GHz <strong>and</strong> 2.45 GHz. From this we c<strong>on</strong>clude that that the propagati<strong>on</strong> is attenuated more in 5 GHz<br />

than in 2 GHz due to shadowing. We assume that the higher st<strong>and</strong>ard deviati<strong>on</strong> is caused by the fact that<br />

the obstacles attenuate more at 5.25 Hz than in 2.45 GHz so that the fades caused by the shadowing are<br />

deeper for the 5.25 GHz. From the measurements we can calculate that the extra loss is about 2 - 3 dB.<br />

For the model we decided to select the value 2.5 dB. For LOS/OLOS c<strong>on</strong>diti<strong>on</strong>s we could find a similar<br />

result, but the difference was negligible.<br />

Page 133 (167)


WINNER D5.4 v. 1.4<br />

When comparing the shadow fading autocorrelati<strong>on</strong>, we got the following results for the autocorrelati<strong>on</strong>s<br />

of the over-all shadowing: For 2.45 GHz the correlati<strong>on</strong> distance was approximately 330 m <strong>and</strong> for the<br />

5.25 GHz it was approximately 320 m. Note that the route used for the comparis<strong>on</strong> was the <strong>on</strong>e with the<br />

greatest correlati<strong>on</strong> distance.<br />

5.6.9.2 Rms-delay spread<br />

The RMS delay spread is shown in the Table 5.46 for the centre-frequencies 2.45 <strong>and</strong> 5.25 GHz. The<br />

difference is c<strong>on</strong>siderable. For 2.45 GHz the mean delay spread for LOS is 35 % <strong>and</strong> for NLOS 80 %<br />

higher. Same trend can be found in the references cited in Secti<strong>on</strong> 5.5.<br />

Table 5.46: Rms-delay spread percentiles at 2.45 <strong>and</strong> 5.25 GHz.<br />

Rms delay spread<br />

(ns)<br />

LOS<br />

NLOS<br />

2 GHz 5 GHz 2 GHz 5 GHz<br />

10% 6.4 2.5 12.3 4.3<br />

50% 22.7 15.4 61.0 37.1<br />

90% 64.0 84.4 130.0 89.5<br />

mean 30.2 36.8 69.0 42.1<br />

Page 134 (167)


WINNER D5.4 v. 1.4<br />

6. Channel Model Implementati<strong>on</strong><br />

The purpose of this chapter is to discuss issues c<strong>on</strong>cerning implementati<strong>on</strong> of the WINNER <strong>channel</strong><br />

model.<br />

6.1 Overview for implementing the model<br />

WINNER <strong>channel</strong> model needs as an input the general informati<strong>on</strong> like <strong>channel</strong> scenario <strong>and</strong> MIMO<br />

setup, antenna c<strong>on</strong>figurati<strong>on</strong>s like radiati<strong>on</strong> patterns <strong>and</strong> array geometries <strong>and</strong> <strong>system</strong> layout informati<strong>on</strong><br />

like relative distances <strong>and</strong> orientati<strong>on</strong>s of the transceivers. Output of the model is a set of discrete <strong>channel</strong><br />

impulse resp<strong>on</strong>ses with matrix coefficients (see eq 3.26). Entries of the matrices are complex <strong>channel</strong><br />

coefficients for each transmitter receiver antenna element pairs. Channel impulse resp<strong>on</strong>ses are<br />

realisati<strong>on</strong>s of the radio <strong>channel</strong> for discrete time instants <strong>and</strong> for different radio <strong>link</strong>s.<br />

6.1.1 Time sampling <strong>and</strong> interpolati<strong>on</strong><br />

Channel sampling frequency has to be finally equal to the simulati<strong>on</strong> <strong>system</strong> sampling frequency. To have<br />

feasible computati<strong>on</strong>al complexity it is not possible to generate <strong>channel</strong> realisati<strong>on</strong>s <strong>on</strong> the sampling<br />

frequency of the <strong>system</strong> to be simulated. The <strong>channel</strong> realisati<strong>on</strong>s have to be generated <strong>on</strong> some lower<br />

sampling frequency <strong>and</strong> then interpolated to the desired frequency. A practical soluti<strong>on</strong> is e.g. to generate<br />

<strong>channel</strong> samples with sample density (over-sampling factor) two, interpolate them accurately to sample<br />

density 64 <strong>and</strong> to apply zero order hold interpolati<strong>on</strong> to the <strong>system</strong> sampling frequency. Channel impulse<br />

resp<strong>on</strong>ses can be generated during the simulati<strong>on</strong> or stored <strong>on</strong> a file before the simulati<strong>on</strong> <strong>on</strong> low sample<br />

density. Interpolati<strong>on</strong> can be d<strong>on</strong>e during the <strong>system</strong> simulati<strong>on</strong>.<br />

6.1.2 Coordinate <strong>system</strong><br />

System layout in the Cartesian coordinates is for example the following:<br />

Figure 6.1: System layout of multiple base stati<strong>on</strong>s <strong>and</strong> mobile stati<strong>on</strong>s.<br />

All the BS <strong>and</strong> MS have (x,y) coordinates. MS <strong>and</strong> cells (sectors) have also array broad side orientati<strong>on</strong>,<br />

where north (up) is the zero angle. Positive directi<strong>on</strong> of the angles is the clockwise directi<strong>on</strong>.<br />

Table 6.1: Transceiver coordinates <strong>and</strong> orientati<strong>on</strong>s.<br />

Tranceiver Co-ordinates Orientati<strong>on</strong> [°]<br />

Page 135 (167)


WINNER D5.4 v. 1.4<br />

BS1 cell1 (x bs1 ,y bs1 ) Ω c1<br />

cell2 (x bs1 ,y bs1 ) Ω c2<br />

cell3 (x bs1 ,y bs1 ) Ω c3<br />

BS2 cell4 (x bs2 ,y bs2 ) Ω c4<br />

cell5 (x bs2 ,y bs2 ) Ω c5<br />

cell6 (x bs2 ,y bs2 ) Ω c6<br />

MS1 (x ms1 ,y ms1 ) Ω ms1<br />

MS2 (x ms2 ,y ms2 ) Ω ms2<br />

MS3 (x ms3 ,y ms3 ) Ω ms3<br />

Both the distance <strong>and</strong> line of sight (LOS) directi<strong>on</strong> informati<strong>on</strong> of the radio <strong>link</strong>s are calculated for the<br />

input of the model. Distance between the BS i <strong>and</strong> MS k is<br />

d<br />

( 2<br />

2<br />

BS ,<br />

) ( )<br />

i MS<br />

x<br />

k BS<br />

− x<br />

i MS<br />

+ y<br />

k BS<br />

− y<br />

i MS k<br />

= . (6.1)<br />

The LOS directi<strong>on</strong> from BS i to MS k with respect to BS antenna array broad side is (see Figure 6.2)<br />

⎧ ⎛ y ⎞<br />

MS<br />

− y<br />

⎪ ⎜<br />

k BSi<br />

− arctan<br />

⎟ + 90° − Ω<br />

BS<br />

, when x ≥<br />

⎪<br />

i<br />

MS<br />

x<br />

k BSi<br />

⎝<br />

xMS<br />

− x<br />

k BSi<br />

⎠<br />

θ<br />

BS<br />

= ⎨<br />

i , MS<br />

(6.2)<br />

k<br />

⎪ ⎛ yMS<br />

− y ⎞<br />

⎜<br />

k BSi<br />

⎟<br />

⎪−<br />

arctan<br />

− 90° − Ω<br />

BS<br />

, when x <<br />

i<br />

MS<br />

x<br />

k BSi<br />

⎩ ⎝<br />

xMS<br />

− x<br />

k BSi<br />

⎠<br />

The angles <strong>and</strong> orientati<strong>on</strong>s are depicted in the figure below.<br />

Ω BSi<br />

θ BS i , MS k<br />

Ω MSk<br />

θ MS , k BS i<br />

Figure 6.2: BS <strong>and</strong> MS antenna array orientati<strong>on</strong>s.<br />

Pairing matrix A is in the example case of Figure 6.2 a 3x6 matrix with values {0,1}. Value 0 st<strong>and</strong>s for<br />

<strong>link</strong> MSm to celln is not modelled <strong>and</strong> value 1 for <strong>link</strong> is modelled.<br />

Page 136 (167)


WINNER D5.4 v. 1.4<br />

⎡χ<br />

ms1,<br />

c1<br />

χ<br />

ms1,<br />

c2<br />

L χ<br />

ms1,<br />

c6<br />

⎤<br />

⎢<br />

⎥<br />

= ⎢<br />

χ<br />

ms2,<br />

c1<br />

χ<br />

ms1,<br />

c2<br />

L χ<br />

ms1,<br />

c6<br />

A ⎥<br />

(6.3)<br />

⎢ M M O M ⎥<br />

⎢<br />

⎥<br />

⎣χ<br />

ms3,<br />

c1<br />

χ<br />

ms3,<br />

c2<br />

L χ<br />

ms3,<br />

c6<br />

⎦<br />

The pairing matrix can be applied to select which radio <strong>link</strong>s will be generated <strong>and</strong> which will not.<br />

6.1.3 Generati<strong>on</strong> of correlated large-scale parameters<br />

The <strong>system</strong> <strong>level</strong> modelling will introduce some locati<strong>on</strong> dependency between the radio <strong>link</strong>s. This is<br />

d<strong>on</strong>e by correlated large-scale <strong>channel</strong> parameters for the radio <strong>link</strong>s. There can be identified five<br />

different cases in the correlati<strong>on</strong> point of view:<br />

1. One MS is c<strong>on</strong>nected to two different BS<br />

2. One MS is c<strong>on</strong>nected to two different sectors of a single BS<br />

3. Two different MSs are c<strong>on</strong>nected to <strong>on</strong>e sector of a BS<br />

4. Two different MSs are c<strong>on</strong>nected to two different sectors of a single BS<br />

5. Two different MSs are c<strong>on</strong>nected to two different BSs<br />

The radio <strong>link</strong>s in the cases 1 <strong>and</strong> 5 are n<strong>on</strong> correlated, case 2 is fully correlated <strong>and</strong> in the cases 3 <strong>and</strong> 4<br />

the correlati<strong>on</strong> is a functi<strong>on</strong> of distance between MSs. Excepti<strong>on</strong> is the shadow fading, which is correlated<br />

also in case 1 with a fixed factor.<br />

Currently, the following large-scale parameters to be correlated are:<br />

1. Delay-spread (DES)<br />

2. AoD angle-spread (ASD)<br />

3. AoA angle-spread (ASA)<br />

4. Shadow fading (SHF)<br />

5. AoD elevati<strong>on</strong> spread (ESD)<br />

6. AoA elevati<strong>on</strong> spread (ESA)<br />

? (all of which have<br />

mean zero <strong>and</strong> variance <strong>on</strong>e) in the positi<strong>on</strong>s x<br />

i<br />

, yi<br />

where the mobiles are located. The elements of<br />

?( x, y)<br />

are uncorrelated, see Secti<strong>on</strong> 4.1.4.2. However the auto-correlati<strong>on</strong> of is n<strong>on</strong>-zero. More prisecely<br />

the correlati<strong>on</strong> between element c of the ?( x, y)<br />

vector, i.e. ?<br />

c<br />

( x,<br />

y)<br />

, in two points x , y 1 1 <strong>and</strong> x , y 2 2 is<br />

given by<br />

The first step is to generate the vector of four real-valued Gaussian variables ( x, y)<br />

E<br />

⎛<br />

2<br />

2 ⎞<br />

⎜ ( x1<br />

− x2)<br />

+ ( y1<br />

− y2)<br />

( = −<br />

⎟<br />

1 1 c 2 2<br />

exp<br />

(6.4)<br />

⎜<br />

λ ⎟<br />

⎝<br />

c<br />

⎠<br />

{ ξ x , y ) ξ ( x , y )}<br />

c<br />

To obtain these values for the K <strong>link</strong>s between a base-stati<strong>on</strong> <strong>and</strong> K users we may start by defining a<br />

correlati<strong>on</strong> matrix C of size KxK <strong>and</strong> then for the square root of this matrix as C = MM T <strong>and</strong> then obtain<br />

the samples as<br />

where = [ ξc<br />

( x<br />

1, y1) , K,<br />

ξc<br />

( xK<br />

, yK<br />

)]<br />

with mean zero <strong>and</strong> variance <strong>on</strong>e. Alternatively, ( x y)<br />

G = Mn , (6.5)<br />

G <strong>and</strong> n is Kx1 vector of independent real-valued Gaussian variables<br />

?<br />

c<br />

, can be generated for a grid of points by first<br />

generating a grid of independent samples <strong>and</strong> then apply an appropriate two-dimensi<strong>on</strong>al filter. <str<strong>on</strong>g>Final</str<strong>on</strong>g>ly,<br />

interpolati<strong>on</strong> is used to find the value for a specific x<br />

i<br />

, yi<br />

. In this approach the resoluti<strong>on</strong> of the grid<br />

should be much finer that the correlati<strong>on</strong> distance λ c .<br />

After having obtained ( x, y)<br />

? the actual large-scale parameters are obtained as<br />

( µ )<br />

( ) = − 1 0.5<br />

x , y g R ( 0) ?( x,<br />

y)<br />

s +<br />

, (6.6)<br />

0.5<br />

T 0.5<br />

5<br />

where R ( 0)<br />

is obtained from the eigendecompositi<strong>on</strong> R( 0) = EΛE<br />

as R ( 0) = EΛ<br />

0.<br />

required parameters are found in Sectri<strong>on</strong> 3.<br />

, <strong>and</strong> the<br />

Page 137 (167)


WINNER D5.4 v. 1.4<br />

6.2 Interfaces<br />

This secti<strong>on</strong> describes example input <strong>and</strong> output interfaces of the WIM <strong>channel</strong> model functi<strong>on</strong> in Matlab<br />

format.<br />

6.2.1 Example input parameters<br />

There are four input arguments, all of which are MATLAB structs. The first three arguments are<br />

m<strong>and</strong>atory. The following tables describe the fields of the input structs.<br />

Table 6.2: General <strong>channel</strong> model parameters. Comm<strong>on</strong> for all <strong>link</strong>s.<br />

Parameter name Definiti<strong>on</strong> Default value Unit<br />

NumBsElements<br />

NumMsElements<br />

Scenario<br />

PropagC<strong>on</strong>diti<strong>on</strong><br />

SampleDensity<br />

NumTimeSamples<br />

UniformTimeSampling<br />

NumSubPathsPerPath<br />

FixedPdpUsed<br />

FixedAnglesUsed<br />

CenterFrequency<br />

The number of elements in the BS array. This<br />

parameter is ignored if antenna patterns are defined<br />

in the input struct ANTPAR. In this case the number<br />

of BS elements is extracted from the antenna<br />

definiti<strong>on</strong>.<br />

The number of elements in the MS array. This<br />

parameter is ignored if antenna patterns are defined<br />

in the input struct ANTPAR. In this case the number<br />

of BS elements is extracted from the antenna<br />

definiti<strong>on</strong>.<br />

Selected WIM <strong>channel</strong> scenario (‘A1’, ‘B1’, ‘C2’ or<br />

‘D1’)<br />

Line of sight c<strong>on</strong>diti<strong>on</strong> (‘LOS’, ’NLOS’). Select<br />

either line of sight or n<strong>on</strong> line of sight model.<br />

Time sampling interval of the <strong>channel</strong>. A value<br />

greater than <strong>on</strong>e should be selected if Doppler<br />

analysis is to be d<strong>on</strong>e.<br />

Number of <strong>channel</strong> samples (impulse resp<strong>on</strong>se<br />

matrices) to generate per <strong>link</strong>.<br />

If this parameter has value ‘yes’ time sampling<br />

interval of the <strong>channel</strong> for each user will be equal.<br />

Sampling interval will be calculated from the<br />

SampleDensity <strong>and</strong> the highest velocity found in the<br />

input parameter vector MsVelocity. If this<br />

parameter has value ’no’, then the time sampling<br />

interval for each <strong>link</strong> will be different, if MSs have<br />

different speeds (see userpar.MsVelocity). Setting<br />

this parameters ‘yes’ may be useful in some <strong>system</strong><strong>level</strong><br />

simulati<strong>on</strong>s where all simulated <strong>link</strong>s need to<br />

be sampled at equal time intervals, regardless of MS<br />

speeds.<br />

Number of rays per path. The <strong>on</strong>ly value supported<br />

in the WIM implementati<strong>on</strong> is 10 rays.<br />

Use tabulated delays instead of drawing r<strong>and</strong>om<br />

values for each drop yes/no. If FixedPdpUsed='yes',<br />

the delays <strong>and</strong> powers of paths are taken from a<br />

table.<br />

Use tabulated angles instead of drawing r<strong>and</strong>om<br />

values for each drop yes/no. If<br />

FixedAnglesUsed='yes', the AoD/AoAs are taken<br />

from a table. R<strong>and</strong>om pairing of AoDs <strong>and</strong> AoAs is<br />

not used.<br />

Carrier centre-frequency. Affects path loss <strong>and</strong> time<br />

sampling interval.<br />

2<br />

2<br />

‘A1’<br />

‘NLOS’ -<br />

2<br />

Page 138 (167)<br />

-<br />

samples/half<br />

wavelength<br />

100 -<br />

‘no’ -<br />

10 -<br />

‘no’ -<br />

‘no’ -<br />

DelaySamplingInterval Delay sampling interval (delay resoluti<strong>on</strong>). 10e-8 sec<br />

PathLossModelUsed<br />

Path-loss included in the <strong>channel</strong> matrices yes/no (if<br />

‘no’, PL is calculated <strong>and</strong> returned in the sec<strong>on</strong>d<br />

output argument, but not multiplied with the<br />

<strong>channel</strong> matrices)<br />

2E9<br />

Hz<br />

‘no’ -<br />

ShadowingModelUsed Shadow fading included in the <strong>channel</strong> matrices ‘no’ -


WINNER D5.4 v. 1.4<br />

PathLossModel<br />

AnsiC_core<br />

LookUpTable<br />

R<strong>and</strong>omSeed<br />

yes/no (if ‘no’ shadow fading is still computed <strong>and</strong><br />

returned in the sec<strong>on</strong>d output argument, but not<br />

multiplied with the <strong>channel</strong> matrices). Note that if<br />

both path loss <strong>and</strong> shadowing are switched off the<br />

average power of the <strong>channel</strong> matrix elements will<br />

be <strong>on</strong>e (with azimuthally uniform unit gain<br />

antennas).<br />

The name of the path-loss functi<strong>on</strong>. Functi<strong>on</strong> ‘pathloss’<br />

implements the default WIM path-loss model.<br />

If the default is used, centre-frequency is taken from<br />

the parameter CenterFrequency. One can define<br />

his/her own path-loss functi<strong>on</strong>. For syntax, see<br />

PATHLOSS.<br />

Use optimized computati<strong>on</strong> yes/no. With ‘yes’<br />

faster C-functi<strong>on</strong> is used instead of m-functi<strong>on</strong>.<br />

Note the C-functi<strong>on</strong> SCM_MEX_CORE.C must be<br />

compiled before usage. For more informati<strong>on</strong>, see<br />

SCM_MEX_CORE.M.<br />

If optimized computati<strong>on</strong> is used, complex<br />

exp<strong>on</strong>ents can be either taken from a look-up table<br />

to speed up computati<strong>on</strong> or calculated explicitly.<br />

This parameter defines the table size, if 0 table is<br />

not used, if –1 default table size 2 14 =16384 is used.<br />

The size of the table must be a power-of-two. If<br />

AnsiC_core = ‘no’ this parameter is ignored.<br />

R<strong>and</strong>om seed for fully repeatable <strong>channel</strong><br />

generati<strong>on</strong> (if empty, seed is generated r<strong>and</strong>omly).<br />

Note that even if R<strong>and</strong>omSeed is the fixed, two<br />

<strong>channel</strong> realizati<strong>on</strong>s may still not be the same<br />

between different MATLAB versi<strong>on</strong>s.<br />

‘path-loss’ -<br />

‘no’ -<br />

0 integer<br />

integer<br />

Table 6.3: Link-dependent parameters. All the parameters are vectors of length K, where K is the<br />

number of <strong>link</strong>s. The values are r<strong>and</strong>omly generated; they are not based <strong>on</strong> any specific network<br />

geometry or user behaviour model.<br />

Parameter name<br />

6.2.1.1 Definiti<strong>on</strong><br />

Default value<br />

MsBsDistance Distance between BS <strong>and</strong> MS 1965*RAND(1,K) + 35 m<br />

ThetaBs θ BS (see Figure 6.2) 360* RAND(1,K) deg<br />

ThetaMs θ MS (see Figure 6.2) 360* RAND(1,K) deg<br />

MsVelocity MS velocity 10 m/s<br />

MsDirecti<strong>on</strong> θ v (see Figure 6.2) 360* RAND(1,K) deg<br />

BsNumber<br />

StreetWidth<br />

Dist2<br />

MsNumber is a positive integer defining the index<br />

number of MS for each <strong>link</strong>. This parameter is used in<br />

generati<strong>on</strong> of inter-site correlated shadow fading<br />

values; shadow fading is correlated for <strong>link</strong>s between a<br />

single MS <strong>and</strong> multiple BSs. There is no correlati<strong>on</strong> in<br />

shadow fading between different MSs. Examples: The<br />

default value is the case where all <strong>link</strong>s in a call to the<br />

SCM functi<strong>on</strong> corresp<strong>on</strong>d to different MSs. Setting<br />

MsNumber=<strong>on</strong>es(1,K) corresp<strong>on</strong>ds to the case where<br />

the <strong>link</strong>s from a single MS to K different BSs are<br />

simulated.<br />

Street width is utilized <strong>on</strong>ly with path-loss model in<br />

[D5.3, sec 2.3.1.13.2]<br />

This is utilized <strong>on</strong>ly with path-loss model in [D5.3, sec<br />

2.3.1.13.2]. Parameter is defined in Figure 2-37 in<br />

[D5.3] <strong>and</strong> generated r<strong>and</strong>omly if empty.<br />

Unit<br />

[1:K] -<br />

20 m<br />

[empty matrix] 1xK<br />

m<br />

Table 6.4: Antenna parameters. The following parameters characterize the antennas. Currently<br />

<strong>on</strong>ly linear uniform arrays with dual-polarized elements are supported. The antenna patterns do<br />

Page 139 (167)


WINNER D5.4 v. 1.4<br />

not have to be identical. The complex field pattern values for the r<strong>and</strong>omly generated AoDs <strong>and</strong><br />

AoAs are interpolated.<br />

Parameter name Definiti<strong>on</strong> Default value Unit<br />

BS antenna field pattern values in a 4D array. The<br />

dimensi<strong>on</strong>s are [ELNUM POL EL AZ] =<br />

SIZE(BsGainPattern), where<br />

BsGainPattern<br />

ELNUM is the number of physical antenna elements in<br />

the array. The elements may be dual-polarized.<br />

POL – polarizati<strong>on</strong>. The first dimensi<strong>on</strong> is vertical<br />

polarizati<strong>on</strong>, the sec<strong>on</strong>d is horiz<strong>on</strong>tal. If the polarizati<strong>on</strong><br />

opti<strong>on</strong> is not used, vertical polarizati<strong>on</strong> is assumed (if both<br />

are given).<br />

EL – elevati<strong>on</strong>. This value is ignored. Only the first<br />

element of this dimensi<strong>on</strong> is used.<br />

AZ – complex-valued field pattern in the azimuth<br />

dimensi<strong>on</strong> given at azimuth angles defined in<br />

BsGainAnglesAz.<br />

1<br />

BsGainAnglesAz<br />

BsGainAnglesEl<br />

BsElementPositi<strong>on</strong><br />

If NUMEL(BsGainPattern)=1, all elements are assumed to<br />

have uniform gain defined by the value of BsGainPattern<br />

over the full azimuth angle, <strong>and</strong> the number of BS antenna<br />

elements is defined by NumBsElements. This speeds up<br />

computati<strong>on</strong> since field pattern interpolati<strong>on</strong> is not<br />

required.<br />

Vector c<strong>on</strong>taining the azimuth angles for the BS antenna<br />

field pattern values. These values are assumed to be the<br />

same for both polarizati<strong>on</strong>s. This value is given in degrees<br />

over the range (-180,180) degrees. If<br />

NUMEL(BsGainPattern)=1, this variable is ignored.<br />

Vector of elevati<strong>on</strong> angles for definiti<strong>on</strong> of BS antenna<br />

gain values. This parameter is for future needs <strong>on</strong>ly; its<br />

value is ignored in this implementati<strong>on</strong> (WIM does not<br />

support elevati<strong>on</strong>).<br />

Element spacing for BS linear array in wavelengths. This<br />

parameter can be either scalar or vector. If scalar, uniform<br />

spacing is applied. If vector, values give distances between<br />

adjacent elements.<br />

MS antenna field pattern values in a 4D array. The<br />

dimensi<strong>on</strong>s are [ELNUM POL EL AZ] =<br />

SIZE(MsGainPattern), where<br />

linspace(-<br />

180,180,90)<br />

deg<br />

- -<br />

0.5 wavelength<br />

MsGainPattern<br />

ELNUM – the number of physical antenna elements in the<br />

array. The elements may be dual-polarized.<br />

POL – polarizati<strong>on</strong>. The first dimensi<strong>on</strong> is vertical<br />

polarizati<strong>on</strong>, the sec<strong>on</strong>d is horiz<strong>on</strong>tal. If the polarizati<strong>on</strong><br />

opti<strong>on</strong> is not used, vertical polarizati<strong>on</strong> is assumed (if both<br />

are given).<br />

EL – elevati<strong>on</strong>. This value is ignored. Only the first<br />

element of this dimensi<strong>on</strong> is used.<br />

AZ – complex-valued field pattern in the azimuth<br />

dimensi<strong>on</strong> given at azimuth angles defined in<br />

MsGainAnglesAz.<br />

1 complex<br />

If NUMEL(MsGainPattern)=1, all elements are assumed<br />

to have uniform gain defined by the value of<br />

MsGainPattern over the full azimuth angle, <strong>and</strong> the<br />

number of MS antenna elements is defined by<br />

wimpar.NumMsElements. This speeds up computati<strong>on</strong><br />

since field pattern interpolati<strong>on</strong> is not needed.<br />

Page 140 (167)


WINNER D5.4 v. 1.4<br />

MsGainAnglesAz<br />

MsGainAnglesEl<br />

MsElementPositi<strong>on</strong><br />

InterpFuncti<strong>on</strong><br />

InterpMethod<br />

Vector c<strong>on</strong>taining the azimuth angles for the MS antenna<br />

field pattern values. These values are assumed to be the<br />

same for both polarizati<strong>on</strong>s. This value is given in degrees<br />

over the range (-180,180) degrees. If<br />

NUMEL(BsGainPattern)=1, this variable is ignored.<br />

Vector of elevati<strong>on</strong> angles for definiti<strong>on</strong> of MS antenna<br />

gain values. This parameter is for future needs <strong>on</strong>ly; its<br />

value is ignored in this implementati<strong>on</strong> (WIM does not<br />

support elevati<strong>on</strong>).<br />

Element spacing for MS linear array in wavelengths. This<br />

parameter can be either scalar or vector. If scalar, uniform<br />

spacing is applied. If vector, values give distances between<br />

adjacent elements.<br />

The name of the interpolating functi<strong>on</strong>. One can replace<br />

this with his own functi<strong>on</strong>. For syntax, see interp_gain.m,<br />

which is the default functi<strong>on</strong>. For faster computati<strong>on</strong>, see<br />

interp_gain_c.m<br />

The interpolati<strong>on</strong> method used by the interpolating<br />

functi<strong>on</strong>. Available methods depend <strong>on</strong> the functi<strong>on</strong>. The<br />

default functi<strong>on</strong> is based <strong>on</strong> MATLAB’s interp1.m<br />

functi<strong>on</strong> <strong>and</strong> supports e.g. ‘linear’ <strong>and</strong> ‘cubic’ (default)<br />

methods. Note that some methods, such as ‘linear’, cannot<br />

extrapolate values falling outside the field pattern<br />

definiti<strong>on</strong>.<br />

linspace(-<br />

180,180,90)<br />

deg<br />

- -<br />

0.5 wavelength<br />

‘interp_gain’ -<br />

‘cubic’ -<br />

Parameter matrices BsGainPattern <strong>and</strong> MsGainPattern 2nd dimensi<strong>on</strong> is either 1 or 2. If polarizati<strong>on</strong><br />

opti<strong>on</strong> is in use, the field pattern values have to be given for vertical <strong>and</strong> horiz<strong>on</strong>tal polarizati<strong>on</strong>s (in this<br />

order). If polarizati<strong>on</strong> is not used <strong>on</strong>ly the first dimensi<strong>on</strong>, i.e. vertical, is used, if both are given.<br />

Note that the mean power of narrowb<strong>and</strong> <strong>channel</strong> matrix elements (i.e. summed over delay domain)<br />

depends <strong>on</strong> the antenna gains. The default antenna has unit gain for both polarizati<strong>on</strong>s. Hence, the mean<br />

narrowb<strong>and</strong> <strong>channel</strong> coefficient power is two for ‘polarized’ opti<strong>on</strong>, <strong>and</strong> <strong>on</strong>e for all other opti<strong>on</strong>s.<br />

The fourth input argument, is opti<strong>on</strong>al. It can be used to specify the initial AoDs, AoAs, cisoid phases,<br />

path losses <strong>and</strong> shadowing values when WIM is called recursively, or for testing purposes. If this<br />

argument is given, the r<strong>and</strong>om parameter generati<strong>on</strong> as defined in WIM is not needed. Only the antenna<br />

gain values will be interpolated for the supplied AoAs <strong>and</strong> AoDs.<br />

The fields of the MATLAB struct are given in the following table. Notati<strong>on</strong>: K denotes the number of<br />

<strong>link</strong>s, N denotes the number of paths, M denotes the number of subpaths within a path.<br />

Table 6.5: Initial values, fourth opti<strong>on</strong>al input argument.<br />

Parameter name Definiti<strong>on</strong> Unit<br />

InitDelays A K x N matrix of path delays. Sec<br />

InitSubPathPowers A K x N x M array of powers of the subpaths. -<br />

InitAods A K x N x M array Degrees<br />

InitAoas A K x N x M array Degrees<br />

InitSubPathPhases<br />

A complex-valued K x N x M array. When polarizati<strong>on</strong> opti<strong>on</strong><br />

is used, this is a K x P x N x M array, where P=4. In this case<br />

the sec<strong>on</strong>d dimensi<strong>on</strong> includes the phases for [VV VH HV<br />

HH] polarized comp<strong>on</strong>ents.<br />

degrees<br />

InitPathLosses A K x 1 vector Decibel<br />

InitShadowLosses A K x 1 vector Decibel<br />

6.2.2 Example output parameters<br />

There are three output arguments: CHAN, DELAYS, FULLOUTPUT. The last two are opti<strong>on</strong>al output<br />

parameters. Notati<strong>on</strong>: K denotes the number of <strong>link</strong>s, N is the number of paths, T the number of time<br />

samples, U the number of receiver elements, <strong>and</strong> S denotes the number of transmitter elements.<br />

Page 141 (167)


WINNER D5.4 v. 1.4<br />

Table 6.6: The three output arguments.<br />

Parameter name Definiti<strong>on</strong> Unit<br />

CHAN<br />

DELAYS<br />

FULLOUTPUT<br />

delays<br />

A 5D-array with dimensi<strong>on</strong>s U x S x N x T x K<br />

A K x N vector of path delay values. Note that delays<br />

are, for compatibility with the INITVALUES, also<br />

included in FULLOUTPUT.<br />

A MATLAB struct with the following elements:<br />

A K x N matrix of path delays. This is identical to the<br />

sec<strong>on</strong>d output argument.<br />

subPathPowers A K x N x M array of subpath powers. -<br />

Aods A K x N x M array of subpath angles of departure degrees<br />

Aoas A K x N x M array of subpath angles of arrival degrees<br />

subpath_phases<br />

A complex-valued K x N x M array giving the final<br />

phases of all subpaths. When polarizati<strong>on</strong> opti<strong>on</strong> is<br />

used, a K x P x N x M array, where P=4. In this case<br />

the sec<strong>on</strong>d dimensi<strong>on</strong> includes the phases for [VV VH<br />

HV HH] polarized comp<strong>on</strong>ents.<br />

sec<br />

sec<br />

degrees<br />

Path_losses A K x 1 vector linear scale<br />

shadow_fading A K x 1 vector linear scale<br />

Delta_t<br />

Xpr<br />

A K x 1 vector defining time sampling interval for all<br />

<strong>link</strong>s.<br />

A K x 2 x N array of cross-polarizati<strong>on</strong> coupling<br />

power ratios. The sec<strong>on</strong>d dimensi<strong>on</strong> is the [V-to-H H-<br />

to-V] coupling ratios.<br />

sec<br />

linear scale<br />

6.3 Guidelines <strong>and</strong> examples <strong>on</strong> performing <strong>system</strong>-<strong>level</strong> simulati<strong>on</strong>s<br />

Chapter 6.1 has provided an overview <strong>on</strong> the c<strong>on</strong>cept of our implementati<strong>on</strong>. In the following, we want to<br />

show how this generic interface can be used to simulate some special types of <strong>system</strong>-<strong>level</strong> situati<strong>on</strong>s.<br />

This can serve as a quick guide <strong>on</strong> how to implement these special cases <strong>and</strong> as proof of the versatility of<br />

our implementati<strong>on</strong>.<br />

6.3.1 H<strong>and</strong>over<br />

A h<strong>and</strong>over situati<strong>on</strong> is characterized by a MS moving from the coverage are of <strong>on</strong>e BS to the coverage<br />

area of another BS. Figure 6.3 illustrates this setup.<br />

Figure 6.3: H<strong>and</strong>over scenario.<br />

There are two base-stati<strong>on</strong>s or cells denoted c1 <strong>and</strong> c2, <strong>and</strong> <strong>on</strong>e mobile stati<strong>on</strong>. Note that WIM is a quasistati<strong>on</strong>ary<br />

<strong>channel</strong> model; it does not provide the means to generate smooth evoluti<strong>on</strong> of <strong>channel</strong>s for a<br />

l<strong>on</strong>g, c<strong>on</strong>tinuous period. What we generate instead is the <strong>channel</strong>s for a sequence of short, separated<br />

Page 142 (167)


WINNER D5.4 v. 1.4<br />

periods. Path-loss will be determined according to the geometry, large-scale parameters correlate<br />

properly, but first-order bulk parameters change abruptly from a segment to segment. Thus, while there is<br />

<strong>on</strong>ly <strong>on</strong>e mobile stati<strong>on</strong> in the scenario, each locati<strong>on</strong> of the mobile <strong>on</strong> its path is assigned a unique label<br />

ms1 to msM. This is equivalent to a scenario with multiple mobile stati<strong>on</strong>s at different positi<strong>on</strong>s ms1 to<br />

msM. The resulting procedure is as follows.<br />

1. Set base stati<strong>on</strong> c1 <strong>and</strong> c2 locati<strong>on</strong>s <strong>and</strong> array orientati<strong>on</strong>s according to geometry.<br />

2. Set MS locati<strong>on</strong>s ms1 to msM <strong>and</strong> array orientati<strong>on</strong>s al<strong>on</strong>g the route. Choose the distance<br />

between adjacent locati<strong>on</strong>s according to desired accuracy.<br />

3. Set all the entries of the pairing matrix to 1.<br />

4. Generate all the radio <strong>link</strong>s at <strong>on</strong>ce obtain correct correlati<strong>on</strong> properties. It is possible to generate<br />

more <strong>channel</strong> realizati<strong>on</strong>s, i.e. time samples, for each <strong>channel</strong> segment afterwards. This can be<br />

d<strong>on</strong>e by applying the initial values of small scale parameters in the Table 6.5.<br />

5. Simulate <strong>channel</strong> segments c<strong>on</strong>secutively to emulate moti<strong>on</strong> al<strong>on</strong>g the route.<br />

6.3.2 Interference<br />

Interference situati<strong>on</strong>s are quite similar to h<strong>and</strong>over situati<strong>on</strong>s, except that in this case the sec<strong>on</strong>d BS<br />

transmits a n<strong>on</strong>-desired signal which creates interference. That doesn’t change the parameters of the<br />

<strong>channel</strong>. What changes, however, is the degree of realism that is needed for the interference <strong>channel</strong>. This<br />

has been discussed in Chapter 4.1.3.<br />

6.3.3 Multi-cell <strong>and</strong> multi-user<br />

The h<strong>and</strong>over <strong>and</strong> interference situati<strong>on</strong>s from the previous secti<strong>on</strong>s were an example of single-user<br />

multi-cell setups. Other cases of such a setup are for example found in the c<strong>on</strong>text of multi-BS protocols,<br />

where a MS receives data from multiple BS simultaneously.<br />

The extensi<strong>on</strong> to multiple users (<strong>and</strong> <strong>on</strong>e or more base stati<strong>on</strong>s) is straightforward. Because locati<strong>on</strong> <strong>and</strong><br />

mobile stati<strong>on</strong> index are treated equivalently, it follows that all locati<strong>on</strong>s of all mobiles have to be<br />

defined. C<strong>on</strong>sider the drive-by situati<strong>on</strong> in Figure 6.4.<br />

Figure 6.4: Drive-by scenario (with multiple mobile stati<strong>on</strong>s).<br />

Here, M locati<strong>on</strong>s of mobile stati<strong>on</strong> 1, <strong>and</strong> N locati<strong>on</strong>s of mobile stati<strong>on</strong> 2 are defined yielding a total of<br />

M+N points or labels. The resulting procedure is as follows.<br />

1. Set BS c1 <strong>and</strong> c2 locati<strong>on</strong>s <strong>and</strong> array orientati<strong>on</strong>s according to layout.<br />

2. Set MS locati<strong>on</strong>s ms11 to ms2N <strong>and</strong> array orientati<strong>on</strong>s according to layout.<br />

3. Set the <strong>link</strong>s to be modelled to 1 in the pairing matrix.<br />

4. Generate all the radio <strong>link</strong>s at <strong>on</strong>ce obtain correct correlati<strong>on</strong> properties. It is possible to generate<br />

more <strong>channel</strong> realizati<strong>on</strong>s, i.e. time samples, for each <strong>channel</strong> segment afterwards. This can be<br />

d<strong>on</strong>e by applying the initial values of small scale parameters in the Table 6.5.<br />

5. Simulate <strong>channel</strong> segments in parallel or c<strong>on</strong>secutively according to the desired moti<strong>on</strong> of the<br />

mobiles.<br />

Page 143 (167)


WINNER D5.4 v. 1.4<br />

6.3.4 Multihop <strong>and</strong> relaying<br />

Typically, the <strong>link</strong>s between the MS <strong>and</strong> the <strong>link</strong>s between the BS are not of interest. Cellular <strong>system</strong>s are<br />

traditi<strong>on</strong>ally centric networks where all traffic goes through <strong>on</strong>e or more BS. The BS themselves again<br />

<strong>on</strong>ly talk to a BS hub <strong>and</strong> not between them.<br />

Multihop <strong>and</strong> relaying networks break with this limitati<strong>on</strong>. In multihop networks, the data can take a route<br />

over <strong>on</strong>e or more successive MS. Relaying networks, <strong>on</strong> the other h<strong>and</strong>, employ another <strong>level</strong> of network<br />

stati<strong>on</strong>s, the relays, which depending <strong>on</strong> the specific network, might offer more or less functi<strong>on</strong>ality to<br />

distribute traffic intelligently.<br />

<<br />

06<br />

%6<br />

%6<br />

06<br />

06<br />

%6<br />

%6<br />

06<br />

;<br />

Figure 6.5: Multihop <strong>and</strong> relaying scenarios.<br />

In the example figure above the signal from MS1 to BS3 is transmitted via MS3 <strong>and</strong> BS2 act as a repeater<br />

for BS1. These scenarios can be generated by introducing a BS-MS pair into positi<strong>on</strong> of a single BS<br />

serving as a relay or into positi<strong>on</strong> of a single MS serving as a multihop repeater. In these cases <strong>on</strong>e can<br />

apply path-loss <strong>models</strong> of feeder scenarios described in secti<strong>on</strong> 3.2.4. The resulting procedure is as<br />

follows.<br />

1. Set base stati<strong>on</strong> BS1 to BS3 locati<strong>on</strong>s <strong>and</strong> array orientati<strong>on</strong>s according to layout.<br />

2. Set mobile locati<strong>on</strong>s MS1 to MS3 <strong>and</strong> array orientati<strong>on</strong>s according to layout.<br />

3. Add extra base stati<strong>on</strong> BS4 to positi<strong>on</strong> of MS3 <strong>and</strong> extra mobile MS4 to positi<strong>on</strong> of BS2 with<br />

same array orientati<strong>on</strong>s <strong>and</strong> array characteristics as MS3 <strong>and</strong> BS2 respectively.<br />

4. Set the pairing matrix to<br />

⎡0<br />

1 0 0⎤<br />

⎢ ⎥<br />

⎢<br />

0 0 0 1<br />

A =<br />

⎥<br />

⎢0<br />

0 1 0⎥<br />

⎢ ⎥<br />

⎣1<br />

0 0 0⎦<br />

5. Generate all the radio <strong>link</strong>s at <strong>on</strong>ce.<br />

6. Simulate the <strong>channel</strong> segments in parallel.<br />

Page 144 (167)


WINNER D5.4 v. 1.4<br />

7. Test <strong>and</strong> Verificati<strong>on</strong> of the Channel Model <strong>and</strong> Its Implementati<strong>on</strong><br />

The goal of <strong>channel</strong> modelling is to imitate real radio <strong>channel</strong> with high accuracy but with low<br />

complexity. WINNER <strong>channel</strong> model parameters are based <strong>on</strong> measurements c<strong>on</strong>ducted during the<br />

project <strong>and</strong> prior measurement results available <strong>on</strong> public literature. The ultimate verificati<strong>on</strong> of the<br />

model would be to compare the <strong>channel</strong> model output parameters to measurement results. Practical<br />

comparis<strong>on</strong> should be d<strong>on</strong>e between the statistical distributi<strong>on</strong>s of the <strong>channel</strong> parameters. To perform<br />

any verificati<strong>on</strong>, we need implementati<strong>on</strong> of the model.<br />

Verificati<strong>on</strong> of the WINNER <strong>channel</strong> model is a twofold task. The first task is to test the implementati<strong>on</strong>,<br />

i.e. to verify that implementati<strong>on</strong> is in line with the model descripti<strong>on</strong>. WINNER <strong>channel</strong> model has three<br />

<strong>level</strong>s: generati<strong>on</strong> of large-scale parameters like sec<strong>on</strong>d moments of delay <strong>and</strong> directi<strong>on</strong> distributi<strong>on</strong>s,<br />

generati<strong>on</strong> of small scale parameters like delays <strong>and</strong> mean powers, <strong>and</strong> generati<strong>on</strong> of matrix coefficient<br />

<strong>channel</strong> impulse resp<strong>on</strong>ses for the radio <strong>link</strong>s. Testing of implementati<strong>on</strong> is mostly verifying generati<strong>on</strong> of<br />

large-scale parameters. This work is <str<strong>on</strong>g>report</str<strong>on</strong>g>ed in [WP5TS].<br />

The sec<strong>on</strong>d task is to compare implementati<strong>on</strong> output to measurement results. Statistical distributi<strong>on</strong>s of<br />

<strong>channel</strong> parameters can be extracted from the output of the implementati<strong>on</strong>. Extracti<strong>on</strong> can be d<strong>on</strong>e<br />

applying the same methods as with the measured radio <strong>channel</strong>. However comparis<strong>on</strong> of output statistics<br />

to measurement results is not feasible because of reduced complexity of the model. The number of<br />

<strong>channel</strong> multipath comp<strong>on</strong>ents is limited in the model compared to a number that <strong>on</strong>e can observe in the<br />

reality. Even though parameters of the multipath comp<strong>on</strong>ents are drawn from the specific distributi<strong>on</strong>, it<br />

is not possible to compute back the same distributi<strong>on</strong> from a very limited number of samples. Thus<br />

verificati<strong>on</strong> of the model against the measurements c<strong>on</strong>ducted in real envir<strong>on</strong>ments is not viable due to<br />

limited number of samples (multipath comp<strong>on</strong>ents per <strong>channel</strong> segment).<br />

7.1 Test cases<br />

7.1.1 General test cases<br />

All test cases assume reference MIMO antenna c<strong>on</strong>figurati<strong>on</strong> B <strong>and</strong> reference antenna field pattern I,<br />

unless otherwise menti<strong>on</strong>ed.<br />

7.1.1.1 General features<br />

Test id Test descripti<strong>on</strong> Expected outcome Notes<br />

1.1.A<br />

1.1.B<br />

1.1.C<br />

See that the installati<strong>on</strong> package (zip file)<br />

of the distributed versi<strong>on</strong> includes all the<br />

files <strong>and</strong> that the installati<strong>on</strong> instructi<strong>on</strong>s<br />

are intelligible <strong>and</strong> informative. No vital<br />

or important informati<strong>on</strong> is missing.<br />

Give suggesti<strong>on</strong>s for improvement.<br />

Installati<strong>on</strong> <strong>and</strong> compilati<strong>on</strong> of the<br />

optimized ANSI C core functi<strong>on</strong><br />

(scm_core.c) succeeds <strong>and</strong> is adequately<br />

documented.<br />

Check that MATLAB help texts of all<br />

functi<strong>on</strong>s are intelligible <strong>and</strong> informative.<br />

Functi<strong>on</strong>s are wim.m,<br />

generate_bulk_par.m, path-loss.m,<br />

wimparset.m, <strong>link</strong>parset.m, antparset.m,<br />

scenpartables.m<br />

All files included,<br />

informative<br />

installati<strong>on</strong><br />

instructi<strong>on</strong>s within the<br />

distributi<strong>on</strong> zip file.<br />

Nothing vital missing.<br />

Installati<strong>on</strong> <strong>and</strong>/or<br />

compilati<strong>on</strong> is<br />

successful <strong>on</strong> various<br />

platfoRMS: Linux,<br />

Unix, Windows.<br />

Documentati<strong>on</strong> is<br />

sufficient.<br />

No mistakes or<br />

ambiguities in help<br />

texts.<br />

7.1.2 Input/output parameters<br />

7.1.2.1 Validity <strong>and</strong> range of basic input <strong>and</strong> output arguments<br />

Test id Test descripti<strong>on</strong> Expected outcome Notes<br />

2.1.A<br />

Field names of all input <strong>and</strong> output<br />

parameters corresp<strong>on</strong>d to those in the<br />

Full match between<br />

documentati<strong>on</strong><br />

Page 145 (167)


WINNER D5.4 v. 1.4<br />

2.1.B<br />

2.1.C<br />

implementati<strong>on</strong> specificati<strong>on</strong>.<br />

Ranges, units <strong>and</strong> sizes of all input <strong>and</strong><br />

output parameters corresp<strong>on</strong>d to those in<br />

the implementati<strong>on</strong> specificati<strong>on</strong>.<br />

Angles units in the formulas <strong>and</strong> in the<br />

variables as input/ouput.<br />

functi<strong>on</strong> input <strong>and</strong><br />

output<br />

Full match between<br />

documentati<strong>on</strong><br />

functi<strong>on</strong> input <strong>and</strong><br />

output<br />

Angles in the formula<br />

should be in radians;<br />

angles in input/output<br />

should be in degrees;<br />

7.1.3 Validati<strong>on</strong> of computati<strong>on</strong><br />

7.1.3.1 Deterministic behaviour of MIMO <strong>channel</strong> matrices<br />

Test id Test descripti<strong>on</strong> Expected outcome Notes<br />

3.1.A<br />

3.1.B<br />

Test a SISO <strong>system</strong> with NumPaths=1<br />

<strong>and</strong> <strong>on</strong>ly <strong>on</strong>e subpath<br />

(NumSubPathsPerPath=1). These must<br />

be fed as initial values using the fourth<br />

input argument. Check that the amplitude<br />

of the <strong>channel</strong> coefficient over time is<br />

c<strong>on</strong>stant. Check that the phase of the<br />

<strong>channel</strong> coefficient changes as expected<br />

based <strong>on</strong> e.g. MSVelocity, array<br />

orientati<strong>on</strong> <strong>and</strong> AoA. Compute Doppler<br />

shift using FFT. Check that the shift is at<br />

the correct sideb<strong>and</strong> of the centrefrequency<br />

<strong>and</strong> of correct magnitude.<br />

Repeat 3.1.A for two subpaths coming<br />

from different AoAs.<br />

Amplitude is c<strong>on</strong>stant.<br />

Phase is changing<br />

accordingly. Doppler<br />

shift is correct.<br />

Amplitude is fading<br />

accordingly. Phase is<br />

changing accordingly.<br />

Doppler shifts of the<br />

subpaths are correct.<br />

In the test result<br />

descripti<strong>on</strong> of values<br />

used during testing <strong>and</strong><br />

expected output. For<br />

special settings, new<br />

scenario ‘Test’ must<br />

be added to the<br />

ScenParTables.m <strong>and</strong><br />

to the other functi<strong>on</strong>s.<br />

For special settings,<br />

new scenario ‘Test’<br />

must be added to the<br />

ScenParTables.m <strong>and</strong><br />

to the other functi<strong>on</strong>s.<br />

7.1.3.2 Stochastic behaviour of output MIMO <strong>channel</strong> matrices<br />

Test id Test descripti<strong>on</strong> Expected outcome Notes<br />

3.2.A<br />

3.2.A/L<br />

3.2.B<br />

3.2.B/L<br />

Mean power of each matrix element,<br />

summed over delay domain, should be<br />

<strong>on</strong>e.<br />

Narrowb<strong>and</strong> mean power for the LOS<br />

opti<strong>on</strong> should be <strong>on</strong>e for all matrix<br />

elements.<br />

Narrowb<strong>and</strong> amplitude distributi<strong>on</strong> of<br />

<strong>channel</strong> coefficients. Sum the <strong>channel</strong><br />

taps over delay domain for each time<br />

instant <strong>and</strong> each element of the MIMO<br />

matrix. The amplitude distributi<strong>on</strong> of<br />

each MIMO matrix element should be<br />

approximately Rayleigh.<br />

Narrowb<strong>and</strong> amplitude distributi<strong>on</strong> of<br />

<strong>channel</strong> coefficients for the LOS<br />

c<strong>on</strong>diti<strong>on</strong>. Sum the <strong>channel</strong> taps over<br />

delay domain for each time instant <strong>and</strong><br />

each element of the MIMO matrix. The<br />

amplitude distributi<strong>on</strong> of each MIMO<br />

matrix element should be approximately<br />

All matrix elements<br />

have unit narrow-b<strong>and</strong><br />

power.<br />

3.2.A<br />

Both cdf <strong>and</strong> pdf of<br />

narrowb<strong>and</strong> matrix<br />

elements are Rayleigh<br />

Both cdf <strong>and</strong> pdf of<br />

narrowb<strong>and</strong> matrix<br />

elements are Ricean<br />

Page 146 (167)


WINNER D5.4 v. 1.4<br />

3.2.C<br />

Ricean with the corresp<strong>on</strong>ding K factor.<br />

Narrowb<strong>and</strong> phase angle distributi<strong>on</strong> of<br />

<strong>channel</strong> coefficients. Sum the <strong>channel</strong><br />

taps over delay domain for each time<br />

instant <strong>and</strong> each element of the MIMO<br />

matrix. The distributi<strong>on</strong> of the phase of<br />

each MIMO matrix element should be<br />

approximately uniform over (0,2*pi].<br />

Uniform pdf over<br />

(0,2*pi]<br />

7.1.3.3 Stochastic behaviour of the large-scale parameters<br />

Test id Test descripti<strong>on</strong> Expected outcome Notes<br />

3.3.A<br />

3.3.B<br />

3.3.C<br />

Bulk parameter statistics. Repeat all<br />

the results in Appendix 4.<br />

Path-loss <strong>models</strong> for all the scenarios<br />

except B1 NLOS. Repeat results in<br />

[D5.3, Sec. 2.3.1.13]. Calculate pathloss<br />

exp<strong>on</strong>ent <strong>and</strong> intercept <strong>and</strong><br />

compare to given values.<br />

Path-loss model for B1 NLOS. Fit a<br />

plane to resulting<br />

triplets.<br />

( log10<br />

d<br />

1,log10<br />

d<br />

2,<br />

PL)<br />

Compare the coefficients of plane<br />

equati<strong>on</strong> to values based <strong>on</strong> [D5.3, eq.<br />

2.6].<br />

The obtained results<br />

are very close to the<br />

‘input’ values.<br />

The obtained results<br />

are very close to<br />

[D5.3, sec. 2.3.1.13].<br />

Coefficients should<br />

be close to:<br />

a = 20.1<br />

b = 35.97<br />

c = 9.55<br />

Testing of mu, epsil<strong>on</strong>, <strong>and</strong> r<br />

values may be easier to do<br />

“within” the code (after step<br />

3) than from the output<br />

AoDs/AoAs. Note that, for<br />

example,<br />

E[log10(sigma_AS)]=mu_AS<br />

<strong>and</strong> STD(log10(sigma_AS))<br />

= epsil<strong>on</strong>_AS.<br />

The general equati<strong>on</strong> of<br />

plane:<br />

z = a*x + b*y + c<br />

7.1.3.4 Stochastic behaviour of CDL <strong>models</strong> output<br />

Test id Test descripti<strong>on</strong> Expected outcome Notes<br />

3.4.A<br />

3.4.B<br />

Estimate power delay profile from output<br />

<strong>channel</strong> matrices. Compare it to the <strong>on</strong>es<br />

given in [D5.3, Tables 4.7-16].<br />

Estimate amplitude probability density<br />

functi<strong>on</strong>s for <strong>models</strong>/clusters with LOS<br />

comp<strong>on</strong>ent. Compare distributi<strong>on</strong>s to<br />

Ricean distributi<strong>on</strong>s with desired K-<br />

factor.<br />

Resulting PDPs should<br />

match to <strong>on</strong>es given in<br />

[D5.3, Tables 4.7-16].<br />

Estimated PDFs<br />

should match<br />

theoretical <strong>on</strong>es.<br />

CDL <strong>models</strong> are<br />

selected by setting<br />

fixed PDP <strong>and</strong> Angles<br />

<strong>on</strong>.<br />

Theoretical PDFs must<br />

be generated by the<br />

test pers<strong>on</strong>.<br />

Page 147 (167)


WINNER D5.4 v. 1.4<br />

8. References<br />

[3GPP SCM] 3GPP TR 25.996, “3rd Generati<strong>on</strong> Partnership Project; technical specificati<strong>on</strong> group radio<br />

access network; spatial <strong>channel</strong> model for MIMO simulati<strong>on</strong>s (release 6)”, V6.1.0.<br />

[802.11n] IEEE 802.11-03/940r2, “IEEE P802.11 Wireless LANs, TGn Channel Models, Jan. 9,<br />

2004.<br />

[AlPM02]<br />

[Bal02]<br />

[BBK+02]<br />

[BBK+04]<br />

A. Algans, K. I. Pedersen, <strong>and</strong> P. E. Mogensen, “Experimental analysis of the joint<br />

statistical properties of azimuth spread, delay spread, <strong>and</strong> shadow fading,” IEEE J. Select.<br />

Areas Commun., vol. 20, no. 3, pp. 523-531, Apr. 2002.<br />

Baltitude, R.J.C., “A Comparis<strong>on</strong> of Multipath-Dispersi<strong>on</strong>-Related Micro-Cellular Mobile<br />

Radio Channel Characteristics at 1.9 GHz <strong>and</strong> 5.8 GHz”, in Proc. ANTEM’02, M<strong>on</strong>treal,<br />

Jul. 31 – Aug. 2, 2002, pp. 623-626.<br />

M. D. Batariere, T. K. Blankenship, J. F. Kepler, T. P. Krauss, I. Lisica, S. Mukthvaram,<br />

J. W. Porter, T. A. Thomas, F. W. Vook, “Wideb<strong>and</strong> MIMO mobile impulse resp<strong>on</strong>se<br />

measurements at 3.7 GHz”, IEEE 55 th VTC, pp. 26-30, 2002.<br />

M. D. Batariere, T. K. Blankenshiop, J. F. Kepler, T. P. Krauss, ”Seas<strong>on</strong>al variati<strong>on</strong>s in<br />

path-loss in the 3.7 GHz b<strong>and</strong>”, IEEE RAWCON, pp. 399-402, 2004.<br />

[BGS+05] D. S. Baum, G. Del Galdo, J. Salo, P. Kyösti, T. Rautiainen, M. Milojevic, <strong>and</strong> J.<br />

Hansen, “An Interim Channel Model for Bey<strong>on</strong>d-3G Systems,” in Proc. IEEE VTC’S05,<br />

May 2005.<br />

[Bul03]<br />

[Cor01]<br />

R. J. C. Bultitude, ”Microcellular mobile radio <strong>channel</strong> transmissi<strong>on</strong> loss <strong>and</strong> temporal<br />

dispersi<strong>on</strong> characteristics at 1.9 GHz <strong>and</strong> 5.9 GHz”, COST273 TD(03)015, Barcel<strong>on</strong>a,<br />

Spain, 15-17 Jan, 2003.<br />

L. M. Correia, Wireless Flexible Pers<strong>on</strong>alised Communicati<strong>on</strong>s: COST 259, European<br />

Co-operati<strong>on</strong> in Mobile Radio Research, John Wiley & S<strong>on</strong>s Ltd, 2001.<br />

[Cost231] “Digital Mobile radio towards future generati<strong>on</strong>s”, Cost-231. final <str<strong>on</strong>g>report</str<strong>on</strong>g>.<br />

http://www.lx.it.pt/cost231/<br />

[D5.1]<br />

[D5.2]<br />

WINNER WP5, “A set of <strong>channel</strong> <strong>and</strong> propagati<strong>on</strong> <strong>models</strong> for early <strong>link</strong> <strong>and</strong> <strong>system</strong> <strong>level</strong><br />

simulati<strong>on</strong>s,” WP5 Deliverable D5.1, March. 2004.<br />

WINNER WP5, “Determinati<strong>on</strong> of Propagati<strong>on</strong> Scenarios,” WINNER WP5, Deliverable.<br />

30.6.2004.<br />

[D5.3] WINNER WP5, “Interim Channel Models”, WINNER project deliverable D5.3, ver 2.4,<br />

April 2005.<br />

[D7.2]<br />

[DC99]<br />

[DGM+03]<br />

[DRX98]<br />

[Dug99]<br />

J. Nyström et. al., “System Assessment Criteria Specificati<strong>on</strong>”. WINNER WP7, D7.2.<br />

June 2004. (Currently, 4.5.04, a draft versi<strong>on</strong>.)<br />

E. Damosso <strong>and</strong> L. Correia, eds., COST Acti<strong>on</strong> 231, Digital Mobile Radio Towards<br />

Future Generati<strong>on</strong> Systems, <str<strong>on</strong>g>Final</str<strong>on</strong>g> Report. No. EUR 18957, European Commissi<strong>on</strong>, 1999.<br />

A. Domazetovic, L. J. Greenstein, N. B. M<strong>and</strong>ayam, <strong>and</strong> I. Seskar. “Estimating the<br />

Doppler Spectrum of a Short-Range Fixed Wireless Channel”, IEEE Comm. Lett.,<br />

7(5):227-229, May 2003.<br />

G. Durkin, T. S. Rappaport, H. Xu, ”Measurements <strong>and</strong> <strong>models</strong> for radio path-loss <strong>and</strong><br />

penetrati<strong>on</strong> in <strong>and</strong> around homes <strong>and</strong> trees at 5.85 GHz, IEEE Trans. Comm., Vol. 46,<br />

pp.1484-1496, 1998.<br />

A. Dugas, “5 GHz MILTON Propagati<strong>on</strong> Study”, slide-set downloadable from<br />

http://www.crc.ca/en/html/milt<strong>on</strong>/home/documents.<br />

[EBITAR] EBIT “Analysis Report of EBIT”, WINNER Internal Report, 2005,<br />

https://bscw.eurescom.de/bscw/bscw.cgi/0/124779<br />

[EGT+99]<br />

V. Erceg, L. J. Greenstein, S. Y. Tj<strong>and</strong>ra, S. R. Parkoff, A. Gupta, B. Kulic, A. A. Julius,<br />

<strong>and</strong> R. Bianchi, “An empirically based path-loss model for wireless <strong>channel</strong>s in suburban<br />

envir<strong>on</strong>ments”, IEEE J. Selected Areas Comm., Vol. 17, No. 7, pp. 1205-1211, 1999.<br />

[Erc01] V. Erceg et al., “Channel Models for Fixed Applicati<strong>on</strong>s”, Technical <str<strong>on</strong>g>report</str<strong>on</strong>g>, IEEE 802.16<br />

Broadb<strong>and</strong> Wireless Working Group, Jul. 2001.<br />

Page 148 (167)


WINNER D5.4 v. 1.4<br />

[FBR+94]<br />

[FDS+94]<br />

[Fle00]<br />

[Gald04]<br />

[GEYC]<br />

[Gud91]<br />

[Hat80]<br />

[HOHB02]<br />

[JXP01]<br />

[KHK+01]<br />

[KKM02]<br />

[KRB00]<br />

[KSL+02]<br />

[KVV05]<br />

[LUI99]<br />

[Maw92]<br />

[MBX93]<br />

[MBX94]<br />

[Medav]<br />

[MEJ91]<br />

Feuerstein, M.J.; Blackard, K.L.; Rappaport, T.S.; Seidel, S.Y.; Xia, H.H., “Path-loss,<br />

delay spread, <strong>and</strong> outage <strong>models</strong> as functi<strong>on</strong>s of antenna height for microcellular <strong>system</strong><br />

design”, in Proc. IEEE VTC’94, Vol. 43, Iss. 3, Aug. 1994, pp. 487 – 498.<br />

Foster, H.M.; Dehghan, S.F.; Steele, R.; Stefanov, J.J.; Strelouhov, H.K., “Microcellular<br />

measurements <strong>and</strong> their predicti<strong>on</strong>”, IEE Colloquium <strong>on</strong> “Role of Site Shielding in<br />

Predicti<strong>on</strong> Models for Urban Radiowave Propagati<strong>on</strong>” (Digest No. 1994/231), Nov. 1994,<br />

pp. 2/1 - 2/6.<br />

B. H. Fleury, “First- <strong>and</strong> sec<strong>on</strong>d-order characterizati<strong>on</strong> of directi<strong>on</strong> dispersi<strong>on</strong> <strong>and</strong> space<br />

selectivity in the radio <strong>channel</strong>”, IEEE Trans. Inform. Theory, Sep. 2000, vol. IT-46, no.<br />

6, pp. 2027-2044.<br />

G. Del Galdo, “The SCM b<strong>and</strong>width extensi<strong>on</strong>,” WINNER WP5.5, Tech. Rep., Sep.<br />

2004.<br />

L. J. Greenstein, V. Erceg, Y. S. Yeh, <strong>and</strong> M. V. Clark, “A new path-gain/delay-spread<br />

propagati<strong>on</strong> model for digital cellular <strong>channel</strong>s,” IEEE Trans. Vehicular Technology, vol.<br />

46, p. 477, 1997.<br />

M. Gudmunds<strong>on</strong>, “Correlati<strong>on</strong> Model For Shadow Fading in Mobile Radio Systems,”<br />

Electr<strong>on</strong>. Lett., Vol. 27, No. 23, Nov. 1991.<br />

M. Hata, “Empirical Formula for Propagati<strong>on</strong> Loss in L<strong>and</strong> Mobile Radio Services,”<br />

IEEE Trans. Vehicular Technology, Vol. VT-29, Aug. 1980.<br />

M. Herdin, H. Ozcelik, H. Hofstetter, <strong>and</strong> E. B<strong>on</strong>ek, „Variati<strong>on</strong> of measured indoor<br />

MIMO capacity with receive directi<strong>on</strong> <strong>and</strong> positi<strong>on</strong> at 5.2 GHz,“ Electr<strong>on</strong>ic Letters, vol.<br />

38, no. 21m pp. 1283-1285, Oct. 2002.<br />

J. Kivinen, X. Zhao, <strong>and</strong> P. Vainikainen, “Empirical characterizati<strong>on</strong> of wideb<strong>and</strong> indoor<br />

radio <strong>channel</strong> at 5.3 GHz,” IEEE Trans. Antennas Propagt., vol. 49, no. 8, pp. 1192-1203,<br />

Aug. 2001.<br />

K. Kalliola, H. Laitinen, K. Sul<strong>on</strong>en, L. Vuokko, <strong>and</strong> P. Vainikainen, ‘’Directi<strong>on</strong>al Radio<br />

Channel Measurements at Mobile Stati<strong>on</strong> in Different Radio Envir<strong>on</strong>ments at 2.15 GHz’’,<br />

Feb. 2001.<br />

J. F. Kepler, T. P. Krauss, S. Mukthvaram, ”Delay spread measurements <strong>on</strong> a wideb<strong>and</strong><br />

MIMO <strong>channel</strong> at 3.7 GHz”, IEEE 56 th VTC, pp. 2498-2502, 2002.<br />

A. Kuchar, J-P. Rossi, <strong>and</strong> E. B<strong>on</strong>ek, ”Directi<strong>on</strong>al macro-cell <strong>channel</strong> characterizati<strong>on</strong><br />

from urban measurements”, IEEE Trans. Antennas <strong>and</strong> Propagati<strong>on</strong>, Vol. 48, pp.137-145,<br />

2000.<br />

K. Kalliola, K. Sul<strong>on</strong>en, H. Laitinen, O. Kivekäs, J. Krogerus, <strong>and</strong> P. Vainikainen, ”Angle<br />

power distributi<strong>on</strong> <strong>and</strong> mean effective gain of mobile antenna in different propagati<strong>on</strong><br />

envir<strong>on</strong>ments”, IEEE Trans. Veh. Techn., Vol. 51, pp. 823-838, 2002.<br />

A. Kainulainen, L. Vuokko, <strong>and</strong> P. Vainikainen, “Polarizati<strong>on</strong> Behavior in Different<br />

Urban Radio Envir<strong>on</strong>ments at 5.3 GHz,” COST273 TD(05)18, Bologna, Jan. 19-21.<br />

L. M. Correia, `COST 259: Wireless Flexible Pers<strong>on</strong>alized Communicati<strong>on</strong>s`<br />

Mawira, A., “Models for the spatial correlati<strong>on</strong> functi<strong>on</strong>s of the (log)-normal comp<strong>on</strong>ent<br />

of the variability of VHF/UHF field strength in urban envir<strong>on</strong>ment,” PIMRC 1992, pp.<br />

436-440.<br />

L. R. Maciel, H. L. Bert<strong>on</strong>i, <strong>and</strong> H. H. Xia, “Unified Approach to predicti<strong>on</strong> of<br />

Propagati<strong>on</strong> Over Building for all Ranges of Base Stati<strong>on</strong> Antenna Height,” IEEE Trans.<br />

Vehicular Technology, Vol. 42, No. 1, Feb. 1993.<br />

L. R. Maciel, H. L. Bert<strong>on</strong>i, <strong>and</strong> H. H. Xia, “Unified Approach to predicti<strong>on</strong> of<br />

Propagati<strong>on</strong> Over Building for all Ranges of Base Stati<strong>on</strong> Antenna Height,” IEEE Trans.<br />

Vehicular Technology, Vol. 42, No. 1, Feb. 1993.<br />

http://www.<strong>channel</strong>sounder.de<br />

P. E. Mogensen, P. Eggers, <strong>and</strong> C. Jensen, ”Urban area radio propagati<strong>on</strong> measurements<br />

for GSM/DCS 1800 macro <strong>and</strong> micro cells”, ICAP 91, pp. 500-503, 1991.<br />

[MET99] IST-1999-11729 METRA, D2, MIMO <strong>channel</strong> characterisati<strong>on</strong>, February 2001.<br />

Page 149 (167)


WINNER D5.4 v. 1.4<br />

[MIS01]<br />

[MIT+00]<br />

[MKA02]<br />

[MRA93]<br />

[OBL+02]<br />

[OHWW03]<br />

[OOKF68]<br />

[OTH00]<br />

[OTTH01]<br />

[Pa03]<br />

[Pa05]<br />

Masui, H.; Ishii, M.; Sakawa, K.; Shimizu, H.; Kobayashi, T.; Akaike, M., “Spatiotemporal<br />

<strong>channel</strong> characteristics at base stati<strong>on</strong> in microwave urban propagati<strong>on</strong>,” in Proc.<br />

IEEE NRSC’01, Vol. 2, Mar. 2001, Mansoura Univ. Egypt, pp. 609 – 615.<br />

Masui, H.; Ishii, M.; Takahashi, S.; Shimizu, H.; Kobayashi, T., ”Microwave propagati<strong>on</strong><br />

characteristics in an urban LOS envir<strong>on</strong>ment in different traffic c<strong>on</strong>diti<strong>on</strong>s”, IEEE<br />

Antennas <strong>and</strong> Propagati<strong>on</strong> Society Internati<strong>on</strong>al Symposium, Vol. 2, Jul. 2000, pp. 1150 -<br />

1153.<br />

Masui, H.; Kobayashi, T.; Akaike, M., “Microwave path-loss modeling in urban line-ofsight<br />

envir<strong>on</strong>ments,” IEEE Journal <strong>on</strong> Selected Areas in Communicati<strong>on</strong>s, Vol. 20, Iss. 6,<br />

Aug. 2002, pp. 1151-1155.<br />

L. Melin, M. Rönnlund, <strong>and</strong> R. Angbratt, “Radio wave propagati<strong>on</strong> – a comparis<strong>on</strong><br />

between 900 <strong>and</strong> 1800 MHz”, IEEE 43 rd VTC c<strong>on</strong>ference, pp. 250-252, 1993.<br />

J. Ojala, R. Böhme, A. Lappeteläinen <strong>and</strong> M. Uno, ”On the propagati<strong>on</strong> characteristics of<br />

the 5 GHz rooftop-to-rooftop meshed network,” IST Mobile & Wireless<br />

Telecommunicati<strong>on</strong>s Summit 2002, Jun. 2002, Thessal<strong>on</strong>iki, Greece.<br />

H. Ozcelik, M. Herdin, W. Weichselberger, J. Wallace, <strong>and</strong> E. B<strong>on</strong>ek, “Deficiencies of<br />

‘Kr<strong>on</strong>ecker’ MIMO radio <strong>channel</strong> model,” Electr<strong>on</strong>ic Letters, vol. 39, iss. 16, pp. 1209-<br />

1210, Aug. 2003.<br />

Y. Okumura, E. Ohmori, T. Kawano, <strong>and</strong> K. Fukuda, ”Field strength <strong>and</strong> its variability in<br />

VHF <strong>and</strong> UHF l<strong>and</strong>-mobile radio services,” Review of the Electrical Comm. Lab., Vol.<br />

16, No 9. 1968.<br />

Oda U, Tsunekawa, K. <strong>and</strong> Hata, M., “Advanced LOS path-loss mode in microcellular<br />

mobile communicati<strong>on</strong>s”, IEEE Trans. Vehicular Technology, vol. 49, (6), Nov. 2000, pp.<br />

2121-2125.<br />

Y. Oda, R. Tsuchihashi, K. Tsunekawa, M. Hata, ”Measured path loss <strong>and</strong> multipath<br />

propagati<strong>on</strong> characteristics in UHF <strong>and</strong> microwave frequency b<strong>and</strong>s for urban mobile<br />

communicati<strong>on</strong>s,” IEEE 53 rd VTC, pp. 337-341, 2001.<br />

P. Pajusco, ”Double characterisati<strong>on</strong>s of power angule spectrum in macrocell<br />

envir<strong>on</strong>ment,” Electr<strong>on</strong>ics Letters, Vol. 39, pp. 1565-1567, 2003.<br />

P. Papazian, ”Basic transmissi<strong>on</strong> loss <strong>and</strong> delay spread measurements for frequencies<br />

between 430 <strong>and</strong> 5750 MHz,” IEEE Trans. Antennas <strong>and</strong> Propagati<strong>on</strong>, Vol. 53, pp. 694-<br />

701, 2005.<br />

[Par04] J. D. Pars<strong>on</strong>s, The Mobile Radio Propagati<strong>on</strong> Channel, Pentech Press, L<strong>on</strong>d<strong>on</strong>, 1994.<br />

[PLN+99]<br />

[PMF00]<br />

[PT00]<br />

[RIMAX]<br />

[RKJ05]<br />

[SBA+02]<br />

[SCK05]<br />

M. Pettersen, P. H. Lehne, J. Noll, O. Rostbakken, E. Ant<strong>on</strong>sen, <strong>and</strong> R. Eckhoff,<br />

”Characterizati<strong>on</strong> of the directi<strong>on</strong>al wideb<strong>and</strong> radio <strong>channel</strong> in urban <strong>and</strong> suburban areas,”<br />

IEEE 50th VTC, pp. 1454-1459,1999.<br />

K. Pedersen, P. Mogensen, <strong>and</strong> B. H. Fleury, “A stochastic model of the temporal <strong>and</strong><br />

azimuthal dispersi<strong>on</strong> seen at the base stati<strong>on</strong> in outdoor propagati<strong>on</strong> envir<strong>on</strong>ments,” IEEE<br />

Trans. Veh. Technol., Mar. 2000, vol. VT-49, no. 2, pp. 437-447.<br />

J. W Porter <strong>and</strong> J. A Thweatt, “Microwave Propagati<strong>on</strong> Characteristics in the MMDS<br />

Frequency B<strong>and</strong>,” in Proc. IEEE ICC’00, Jun. 2000, Vol. 3, pp. 1578-1582.<br />

R. S. Thoma, M. L<strong>and</strong>mann, <strong>and</strong> A. Richter, “RIMAX – a Maximum Likelihood<br />

framework for parameter estimati<strong>on</strong> in multidimensi<strong>on</strong>al <strong>channel</strong> sounding,” 2004 Intl.<br />

Symp. <strong>on</strong> Antennas <strong>and</strong> Propagati<strong>on</strong>, Aug. 17-21, 2004, Sendai, JP.<br />

T. Rautiainen, K. Kalliola, <strong>and</strong> J. Juntunen, ”Wideb<strong>and</strong> radio propagati<strong>on</strong> characteristics<br />

at 5.3 GHz in suburban envir<strong>on</strong>ments”, 16 th Annual IEEE Int. Symp. <strong>on</strong> Pers<strong>on</strong>al Indoor<br />

<strong>and</strong> Mobile Radio Communicati<strong>on</strong>s, Berlin, Germany 11-14 Sep., 2005.<br />

Schenk, T.C.W., Bultitude, R.J.C., Augustin, L.M., van Poppel, R.H., <strong>and</strong> Brussaard, G.,<br />

“Analysis of Propagati<strong>on</strong> loss in Urban Microcells at 1.9 GHz <strong>and</strong> 5.8 GHz,” in Proc.<br />

URSI Commisi<strong>on</strong> F Open Symposium <strong>on</strong> Radiowave Propagati<strong>on</strong> <strong>and</strong> Remote Sensing,<br />

Garmisch-Patenkirchen, Germany, Feb. 2002.<br />

N. Skentos, C<strong>on</strong>stantinou <strong>and</strong> A. G Kanatas, “Results from Rooftop to Rooftop MIMO<br />

Channel Measurements at 5.2 GHz,” COST273 TD(05)59, Bologna, Jan. 19-21.<br />

Page 150 (167)


WINNER D5.4 v. 1.4<br />

[SCME]<br />

[SDD00]<br />

[SG00]<br />

[SMI+00]<br />

[SV87]<br />

[Sva02]<br />

[THL+01]<br />

[TLS+05]<br />

[TPE02]<br />

[TSS+03]<br />

[WET+04]<br />

[WHL+93]<br />

[WHL94]<br />

D. S. Baum, G. Del Galdo, J. Salo, P. Kyösti, T. Rautiainen, M. Milojevic, <strong>and</strong> J. Hansen,<br />

“SCM Extensi<strong>on</strong>s,” Technical Report, WINNER WP5.5 Internal, Oct. 2004.<br />

J. Sydor, A. Dugas, <strong>and</strong> J. Duggan, “A new broadb<strong>and</strong> wireless network for 5 GHz<br />

licence-exempt applicati<strong>on</strong>s,” in Proc. IEEE RAWCON’00, pp. 33-37.<br />

T. Schwengler <strong>and</strong> M. Gilbert, ”Propagati<strong>on</strong> <strong>models</strong> at 5.8 GHz - path loss <strong>and</strong> building<br />

penetrati<strong>on</strong>”, IEEE Radio <strong>and</strong> Wireless C<strong>on</strong>ference 10-13 Sep. 2000, pp. 119-124.<br />

H. Shimizu, H. Masui, M. Ishi, K. Sakawa, <strong>and</strong> T. Kobayashi, “LOS <strong>and</strong> NLOS Path-Loss<br />

<strong>and</strong> Delay Characteristics at 3.35 GHz in a Residential Envir<strong>on</strong>ment,” IEEE Antennas <strong>and</strong><br />

Propagati<strong>on</strong> Society Internati<strong>on</strong>al Symposium 2000, Vol. 2, Jul. 2000, pp. 1142 - 1145.<br />

A. Saleh, <strong>and</strong> R. A. Valenzuela, “A statistical model for indoor multipath propagati<strong>on</strong>”,<br />

IEEE J. Select. Areas Commun., vol. SAC-5, no. 2, Feb. 1987, pp. 128-137.<br />

Svantess<strong>on</strong>, T., “A double-bounce <strong>channel</strong> model for multi-polarized MIMO <strong>system</strong>s,” in<br />

Proc. IEEE VTC’02-Fall, Vol. 2, Sep. 2002, pp. 691 – 695.<br />

Thoma, R.S.; Hampicke, D.; L<strong>and</strong>mann, M.; Richter, A.; Sommerkorn, G., “MIMO<br />

Measurement for Double-Directi<strong>on</strong>al Channel Modelling,” IEE MIMO Seminar, L<strong>on</strong>d<strong>on</strong>,<br />

UK, Dec. 2001.<br />

Trautwein, U.; L<strong>and</strong>mann, M.; Sommerkorn, G.; Thomä, R., “System-Oriented<br />

Measurement <strong>and</strong> Analysis of MIMO Channels,” COST 273 TD(05) 063, Bologna, Italy,<br />

Jan. 19-21, 2005, Jan. 2005.<br />

S. Thoen, L. Van der Perre, <strong>and</strong> M. Engels, “Modeling the Channel Time-Variance for<br />

Fixed Wireless Communicati<strong>on</strong>”, IEEE Communicati<strong>on</strong> Letters, Vol. 6, No. 8, Aug. 2002.<br />

Trautwein, U.; Schneider, C.; Sommerkorn, G.; Hampicke, D.; Thoma, R., Wirnitzer, W.,<br />

“Measurement Data for Propagati<strong>on</strong> Modeling <strong>and</strong> Wireless System Evaluati<strong>on</strong>,”<br />

COST273, MCM, TD-03-021, Barcel<strong>on</strong>a, Jan. 2003<br />

Zhenyu Wang, Eustace K. Tameh <strong>and</strong> Andrew R. Nix, “Statistical Peer-to-Peer Channel<br />

Models for Outdoor Urban Envir<strong>on</strong>ments at 2 GHz <strong>and</strong> 5 GHz,” IEEE 2004.<br />

J. A. Wepman, J. R. Hoffman, L. H. Loew, W. J. Tanis, <strong>and</strong> M. E. Hughes, “Impulse<br />

resp<strong>on</strong>se measurements in the 902.928 <strong>and</strong> 1850.1990 MHz b<strong>and</strong>s in macrocellular<br />

envir<strong>on</strong>ments,” 2 nd internati<strong>on</strong>al c<strong>on</strong>ference <strong>on</strong> Universal Pers<strong>on</strong>al Communicati<strong>on</strong>s, Vol.<br />

2, pp. 590-594, 1993.<br />

J. A. Wepman, J. R. Hoffman, <strong>and</strong> L. H. Loew, ”Characterizati<strong>on</strong> of macrocellular PCS<br />

propagati<strong>on</strong> <strong>channel</strong>s in the 1850-1990 MHz b<strong>and</strong>, 3rd Annual Internati<strong>on</strong>al C<strong>on</strong>ference<br />

<strong>on</strong> Universal Pers<strong>on</strong>al Communicati<strong>on</strong>s, pp. 165-170, 1994.<br />

[WP5AI] WP5 “List of Measurement Analysis Items”, Internal Report WINNER, 2005,<br />

https://bscw.eurescom.de/bscw/bscw.cgi/0/124779.<br />

[WP5AR]<br />

[WP5TS]<br />

[Xia96]<br />

[YMI+04]<br />

[YTL02]<br />

[ZKVS02]<br />

[ZRKV04]<br />

WP5 “Analysis Reports EBIT, KTH, NOK, TKK <strong>and</strong> TUI”, WINNER Internal Reports,<br />

2005. https://bscw.eurescom.de/bscw/bscw.cgi/0/124779.<br />

WP5 “Test Specificati<strong>on</strong> of WP5 WIM Matlab implementati<strong>on</strong>”, WINNER Internal<br />

Reports, 2005. https://bscw.eurescom.de/bscw/bscw.cgi/0/XX.<br />

H. H. Xia, ” An analytical model for predicting path loss in urban <strong>and</strong> suburban<br />

envir<strong>on</strong>ments”, Seventh IEEE Int. Symposium PIMRC, Vol 1, pp. 19.23, 1996<br />

K. Y<strong>on</strong>ezawa, T. Maeyama, H. Iwai, <strong>and</strong> H. Harada, ”Path loss measurement in 5 GHz<br />

macro cellular <strong>system</strong>s <strong>and</strong> c<strong>on</strong>siderati<strong>on</strong>s of extending existing path-loss predicti<strong>on</strong><br />

<strong>models</strong>”, IEEE WCNC, Vol. 1, pp. 279-283, 2004.<br />

Yacoub, D.; Teich, W.; Lindner, J., „Capacity of Vehicle-Bridge MIMO Channels”,<br />

TD(02)118, COST 273, 5th Management Committee Meeting, Lisb<strong>on</strong> / Portugal, Sep. 19-<br />

20, 2002<br />

X. Zhao, J. Kivinen, P. Vainikainen, <strong>and</strong> K. Skog, ”Propagati<strong>on</strong> characteristics for<br />

wideb<strong>and</strong> outdoor mobile communicati<strong>on</strong>s at 5.3 GHz”, IEEE Sel. Areas Comm., Vol. 20,<br />

pp. 507-514, 2002<br />

X. Zhao, T. Rautiainen, K. Kalliola, P. Vainikainen, “Path-loss <strong>models</strong> for urban<br />

microcells at 5.3 GHz,” COST 273, TD(04)207, Duisburg, Germany, Sep. 2004.<br />

Page 151 (167)


WINNER D5.4 v. 1.4<br />

[ZJY+05]<br />

[ZJ05]<br />

P. Zetterberg, N. Jaldén, K. Yu, <strong>and</strong> M. Bengtss<strong>on</strong>, ”Analysis of MIMO Multi-Cell<br />

Correlati<strong>on</strong>s <strong>and</strong> Other Propagati<strong>on</strong> Issues Based <strong>on</strong> Urban Measurements”, IST Mobile<br />

<strong>and</strong> Wireless Communicati<strong>on</strong>s Summit, Dresden, Germany, Jun. 2005.<br />

P. Zetterberg <strong>and</strong> N. Jaldén, “Comparis<strong>on</strong> of Angle-Spread in Outdoor-to-Outdoor <strong>and</strong><br />

Outdoor-to-Indoor Cases in an Urban Macro-Cell”, Wireless Pers<strong>on</strong>al Multimedia<br />

Communicati<strong>on</strong>s WPMC, WINNER sessi<strong>on</strong>, Aalborg, Denmark, Sep. 2005.<br />

Page 152 (167)


WINNER D5.4 v. 1.4<br />

9. Appendix<br />

9.1 Other scenarios<br />

9.1.1 Scenario definiti<strong>on</strong>s<br />

Here we present WP5 view to the envir<strong>on</strong>ments bey<strong>on</strong>d the five prioritized scenarios.<br />

9.1.1.1 Scenario “high mobility short range hot spot”<br />

This scenario is new <strong>and</strong> therefore not described in [D7.2]. We call it “High Mobility Short Range Hot<br />

Spot”. It represents an outdoor hot spot applicati<strong>on</strong> for short range distances, where the distance can be<br />

range from few meters to 250m. The Rx can be fixed e.g. <strong>on</strong> the side or even under a bridge. The Tx is<br />

mounted <strong>on</strong> the roof of a car, van or lorry. Within this scenario high mobility <strong>and</strong> traffic throughput with<br />

high density can be expected. Such scenarios can be found in rural, urban <strong>and</strong> suburban envir<strong>on</strong>ments,<br />

e.g. for traffic informati<strong>on</strong> or other user applicati<strong>on</strong>s. Depending <strong>on</strong> the envir<strong>on</strong>ment different MIMO<br />

<strong>channel</strong> characteristics must be c<strong>on</strong>sidered.<br />

Technische Universität Ilmenau (TUI) measured this particular scenario “High Mobility Short Range Hot<br />

Spot”, where the measurement car (Tx) was driving <strong>on</strong> a highway <strong>and</strong> the Rx was fixed at a bridge in a<br />

rural envir<strong>on</strong>ment. No buildings were in this area <strong>on</strong>ly high trees <strong>and</strong> shrubs surrounding the highway <strong>and</strong><br />

bridge. Both LOS <strong>and</strong> NLOS propagati<strong>on</strong> situati<strong>on</strong>s can be expected. Mostly LOS is obvious when the<br />

car is approaching the bridge, under <strong>and</strong> after the bridge NLOS is dominating.<br />

9.1.1.2 Scenario “outdoor to indoor”<br />

See [ZJ05].<br />

9.2 Measurement campaigns for other scenarios<br />

9.2.1 Scenario “high mobility short range hot spot”<br />

9.2.1.1 TUI campaign<br />

TUI outdoor high mobility short range hot spot measurement scenario c<strong>on</strong>sists of a bridge-to-car scenario<br />

in a highway, shown in Figure 9.1. These measurements were d<strong>on</strong>e with RUSK ATM MIMO sounder by<br />

Medav [Medav]. The Carrier frequency in the measurements was 5.2 GHz <strong>and</strong> b<strong>and</strong>width of 120 MHz.<br />

During the measurement campaign the Tx was driving <strong>on</strong> the right <strong>and</strong> left side lane of a two lane<br />

highway road (2 lanes per directi<strong>on</strong>). The Tx antenna was mounted <strong>on</strong> the roof of a car <strong>and</strong> the Rx as<br />

mounted at the bridge, whereby it was down tilt by 45 degrees. Pictures of high-resoluti<strong>on</strong> antennas used<br />

in TUI measurement campaign are shown in Figure 5.7. The maximum distance between Tx <strong>and</strong> Rx<br />

antenna positi<strong>on</strong> was found to be 250m, which defines the short range area for the hot spot applicati<strong>on</strong>.<br />

Figure 9.1: Bridge-to-car hot spot scenario.<br />

Page 153 (167)


WINNER D5.4 v. 1.4<br />

9.2.2 Urban ad-hoc peer-to-peer<br />

9.2.2.1 NOK <strong>and</strong> HUT campaign<br />

In urban ad hoc peer-to-peer measurements both the receiver <strong>and</strong> transmitter ends of the radio <strong>link</strong> were<br />

equipped with spherical array antennas shown in Figure 5.8. Both Tx <strong>and</strong> Rx were installed to a trolley,<br />

<strong>and</strong> during the measurements the transmitter was kept fixed while the receiver unit was moving. The Tx<br />

<strong>and</strong> Rx antenna heights were ~1.5 m from the ground <strong>level</strong>. Measurements c<strong>on</strong>sist of 10 to 20 meter l<strong>on</strong>g<br />

routes in LOS <strong>and</strong> NLOS scenarios in urban outdoor <strong>and</strong> indoor envir<strong>on</strong>ments. Surrounding buildings<br />

were 4-6 floors high, <strong>and</strong> the Rx <strong>and</strong> Tx positi<strong>on</strong>s were located outdoors in open areas (market square),<br />

street cany<strong>on</strong>s <strong>and</strong> restaurant terrace. Outdoor-to-indoor peer-to-peer measurements were d<strong>on</strong>e in metro<br />

stati<strong>on</strong>, <strong>and</strong> indoor peer-to-peer cases were measured in metro stati<strong>on</strong> lobby <strong>and</strong> in a supermarket.<br />

9.3 Measurement results for other scenarios<br />

9.3.1 Scenario C2: typical urban macro-cell - KTH campaign<br />

These results are narrow b<strong>and</strong> measurements close to 2 GHz frequency range.<br />

9.3.1.1 Inter-sector <strong>and</strong> inter-site correlati<strong>on</strong>s<br />

The log-fading based <strong>on</strong> the four MS transmit antennas, as well as combing all MS transmit antennas<br />

together to form a basically omni-directi<strong>on</strong>al antenna, has been calculated. The correlati<strong>on</strong> of the logfading<br />

between sectors is shown in Table 9.1. The correlati<strong>on</strong> between Sector A <strong>and</strong> C has been excluded<br />

as it c<strong>on</strong>tains much less data than the correlati<strong>on</strong> between Sector A <strong>and</strong> B. The cross-site correlati<strong>on</strong> (A<br />

<strong>and</strong> B) is virtually very small while it is substantial for the cross-sector (B <strong>and</strong> C) measurement although<br />

not full.<br />

Table 9.1: Correlati<strong>on</strong> of log-normal fading.<br />

MS Antenna<br />

Sectors 1 2 3 4 All<br />

A & B 27% 7% 9% -9% 10%<br />

B & C 86% 77% 84% 77% 84%<br />

The correlati<strong>on</strong> of the angle-spread results is shown in Table 9.2. The correlati<strong>on</strong> of the angle-spread is<br />

virtually zero between the sites <strong>and</strong> very small between the sectors.<br />

Table 9.2: Correlati<strong>on</strong> of angle-spread.<br />

MS Antenna<br />

Sectors 1 2 3 4 All<br />

A & B 3% -2% -5% -19% -3%<br />

B & C 25% 17% 24% 26% 34%<br />

The reas<strong>on</strong> for the n<strong>on</strong>-full correlati<strong>on</strong> between the sectors of the same site (B & C) we believe is<br />

primarily due to the difference in base-stati<strong>on</strong> antenna pattern differences between the two sectors <strong>and</strong><br />

sec<strong>on</strong>dly because the two sectors are mounted 20-meters apart. Note that the sector cross correlati<strong>on</strong> is<br />

evaluated in an angle segment of width 20-degrees where the element patterns are oscillating. If the<br />

sectors had been more closely located <strong>and</strong> pointing in more similar directi<strong>on</strong>s we believe the correlati<strong>on</strong><br />

would be full. The correlati<strong>on</strong> between A <strong>and</strong> B sectors was evaluated mostly in the main beam of the two<br />

sectors.<br />

In [Maw92] the correlati<strong>on</strong> between sites of the log-normal fading is experimentally found to be given<br />

approximately 0.9-|θ |/200, where θ is the angle between the two base-stati<strong>on</strong>s as seen from the mobile<br />

stati<strong>on</strong>. For this measurement campaign this would corresp<strong>on</strong>d <strong>on</strong> average to a correlati<strong>on</strong> coefficient of<br />

40%. However, the correlati<strong>on</strong> herein is much lower.<br />

Even if the properties at the base-stati<strong>on</strong>s are different it could be theorized that the <strong>channel</strong>s at the<br />

mobile-stati<strong>on</strong>s were similar, for instance if the same scatterers are active in both c<strong>on</strong>necti<strong>on</strong>s. This<br />

property was investigated by indicating <strong>on</strong> map where the same mobile-stati<strong>on</strong> antenna is str<strong>on</strong>gest<br />

(summed over all transmit antennas) for the <strong>link</strong>s to sector A <strong>and</strong> B (i.e. different sites) in Figure 9.2. The<br />

Page 154 (167)


WINNER D5.4 v. 1.4<br />

results show no indicati<strong>on</strong> of any correlati<strong>on</strong>. The corresp<strong>on</strong>ding plot for the cross-sector correlati<strong>on</strong> is<br />

shown in Figure 9.3. The cross-site results in Figure 9.2 show no correlati<strong>on</strong>. The same antenna was<br />

selected in <strong>on</strong>ly 29% of the cases. In fact, since not every antenna is selected with equal probability<br />

(probably due to unintended differences in tilt angle <strong>and</strong> reflecti<strong>on</strong>s from the car), the 29% is c<strong>on</strong>sistent<br />

with completely uncorrelated antenna selecti<strong>on</strong>. In Figure 9.3 the cross-sector results are shown. Here, the<br />

correlati<strong>on</strong> for distances larger than 300meters or so is almost full. The differences close to the basestati<strong>on</strong><br />

may be due to 20meter distance between the two sector antennas.<br />

Figure 9.2: Illustrati<strong>on</strong> of where the same MS<br />

antenna is the str<strong>on</strong>gest in sector A <strong>and</strong> B (which<br />

are <strong>on</strong> different sites).<br />

Figure 9.3: Illustrati<strong>on</strong> of where the same MS<br />

antenna is the str<strong>on</strong>gest in sector B <strong>and</strong> C<br />

(different sectors <strong>on</strong> same site).<br />

9.3.1.2 (Joint) DoA/DoD distributi<strong>on</strong>s<br />

The main DoA directi<strong>on</strong> is estimated for each local area by means of beamforming (the pointing directi<strong>on</strong><br />

in which the most energy is received). Since there are four mobile antennas four estimates are available<br />

from each of the mobile stati<strong>on</strong> transmitting antennas. Here we investigate the dependence between the<br />

pointing angle of the four mobile stati<strong>on</strong> antennas relative to the directi<strong>on</strong> of the base-stati<strong>on</strong>, a, <strong>and</strong> the<br />

DoA of the incoming signal ß, relative to the geographical angle, a, of the mobile, see Figure 9.4. The<br />

angle a is obtained by combining the estimated main DoA <strong>and</strong> the GPS informati<strong>on</strong>, while ß is obtained<br />

from the GPS informati<strong>on</strong> together with knowledge of the directi<strong>on</strong> of travel <strong>and</strong> the mounting of<br />

antennas <strong>on</strong> the vehicle.<br />

Figure 9.4: Illustrati<strong>on</strong> of the dependence between the pointing angle of the MS antenna (relative<br />

directi<strong>on</strong> of BS) <strong>and</strong> the main DoA at the base-stati<strong>on</strong>.<br />

If a <strong>on</strong>ce-bounce model is valid, as indicated in the figure, then a positive a should imply a negative ß. To<br />

investigate this c<strong>on</strong>jecture Figure 9.5 <strong>and</strong> Figure 9.6 were generated where the x-value of each 'x' marks<br />

the pointing directi<strong>on</strong> of an MS antenna, a, <strong>and</strong> the y-axis the main DoA offset ß estimated at the basestati<strong>on</strong><br />

(not all points are included to increase clarity). Also included are the mean of the DoA offset<br />

Page 155 (167)


WINNER D5.4 v. 1.4<br />

values ß as a functi<strong>on</strong> of the MS-antenna pointing angle a, the mean of this curve, <strong>and</strong> curves indicating<br />

the range of plus minus <strong>on</strong>e st<strong>and</strong>ard deviati<strong>on</strong>. In additi<strong>on</strong>, a sinusoid curve has been fitted. The results<br />

show that there is a very small tendency of the effect indicated by Figure 9.4.<br />

Figure 9.5: Main DoA offset at base-stati<strong>on</strong> as a<br />

functi<strong>on</strong> of mobile-stati<strong>on</strong> pointing angle at<br />

Kårhuset.<br />

Figure 9.6: Main DoA offset at base-stati<strong>on</strong> as a<br />

functi<strong>on</strong> of mobile-stati<strong>on</strong> pointing angle<br />

Kårhuset.<br />

A related measure is the probability that the MS antenna pointing most directly towards the base-stati<strong>on</strong><br />

has the smallest DoA offset when received at the base-stati<strong>on</strong>. This probability is found to be around<br />

20%. These two statistics indicate that the very l<strong>on</strong>g-term DoD <strong>and</strong> DoA distributi<strong>on</strong>s may be modeled as<br />

independent. This should not be c<strong>on</strong>fused with the Kr<strong>on</strong>ecker model being valid as it operates <strong>on</strong> a<br />

shorter term. The correlati<strong>on</strong> of the DoA offset between antennas is found to be 0-50%.<br />

The angle-spread at the base-stati<strong>on</strong> was also investigated as a functi<strong>on</strong> of mobile stati<strong>on</strong> antenna pointing<br />

angle <strong>and</strong> found independent. The probability that the antenna pointing mostly towards the base stati<strong>on</strong><br />

should have the smallest angle-spread was found to be 28% at Kårhuset <strong>and</strong> 41% at Vanadis. In the<br />

Vanadis case this is a significant result. This is somewhat surprising in the light of the small (or<br />

practically n<strong>on</strong>-existing) average spread versus MS pointing angle dependence. Therefore this<br />

dependence is also not worthwhile modelling. The correlati<strong>on</strong> of angle-spreads am<strong>on</strong>g the MS antennas if<br />

found to be 26-55% at Kårhuset <strong>and</strong> 50-70% at Vanadis.<br />

9.3.2 Scenario “high mobility short range hot spot”<br />

9.3.2.1 Path-loss <strong>and</strong> shadow fading<br />

In Figure 9.7 path loss for this scenario is presented.<br />

-70<br />

-75<br />

-80<br />

PL [dB]<br />

-85<br />

-90<br />

-95<br />

-100<br />

-105<br />

-110<br />

10 1 10 2<br />

d [m]<br />

Figure 9.7: Path loss under LOS (as example <strong>on</strong>e curve out of 8 measurement runs).<br />

The table below highlights the PL exp<strong>on</strong>ents, offset K <strong>and</strong> the variance (shadow fading) within this<br />

scenario.<br />

Page 156 (167)


WINNER D5.4 v. 1.4<br />

Under LOS c<strong>on</strong>diti<strong>on</strong> the equati<strong>on</strong> for the path loss was to be found as:<br />

PL = 60.6 + 19.3 log 10 (d), with s = 3.1 dB, (9.1)<br />

where d is the distance <strong>and</strong> s is the st<strong>and</strong>ard deviati<strong>on</strong> of the shadow fading.<br />

In Figure 9.8 distributi<strong>on</strong> of the shadow fading <strong>and</strong> SF versus distance are shown.<br />

PDF<br />

0.2<br />

0.18<br />

0.16<br />

0.14<br />

0.12<br />

0.1<br />

0.08<br />

0.06<br />

0.04<br />

0.02<br />

0<br />

-10 -8 -6 -4 -2 0 2 4 6 8 10<br />

SF [dB]<br />

(a)<br />

SF [dB]<br />

10<br />

8<br />

6<br />

4<br />

2<br />

0<br />

-2<br />

-4<br />

-6<br />

-8<br />

-10<br />

0 50 100 150 200 250<br />

d [m]<br />

(b)<br />

Figure 9.8: Shadow Fading distributi<strong>on</strong> in LOS envir<strong>on</strong>ment (a) <strong>and</strong> SF with distance (b).<br />

9.3.2.2 Modelling of PDP<br />

Power delay profile (PDP) in an outdoor high mobility short range hot spot LOS envir<strong>on</strong>ment is presented<br />

in Figure 9.9 <strong>and</strong> time c<strong>on</strong>stant in table Table 9.3.<br />

Table 9.3: Time c<strong>on</strong>stants for PDPs (MHz).<br />

Time<br />

c<strong>on</strong>stant<br />

[MHz]<br />

LOS<br />

95.9<br />

0<br />

-5<br />

Power (dB)<br />

-10<br />

-15<br />

-20<br />

-25<br />

0 10 20 30 40<br />

Excess delay [ns]<br />

(a) LOS.<br />

Figure 9.9: Modeling of power delay profile (PDP) in an outdoor high mobility short range hot spot<br />

envir<strong>on</strong>ment LOS (a) propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s.<br />

Page 157 (167)


WINNER D5.4 v. 1.4<br />

9.3.2.3 Probability of LOS<br />

In measurement data of this scenario we have probability of LOS equal to100% since the measurements<br />

were d<strong>on</strong>e <strong>on</strong> the highway without the traffic.<br />

NLOS could happen if there is a big truck in fr<strong>on</strong>t of the car (Tx). Highways always have at least 2 lines -<br />

cars are faster than trucks so there is very low probability that car is in the shadow of the truck. Also by<br />

law the distance between cars <strong>on</strong> highways should be approximately 30 m <strong>and</strong> therefore a car shouldn’t<br />

make a shadow to a car behind it if the Rx antenna is put high enough.<br />

Having this in mind probability of LOS is very high <strong>and</strong> it is lower at the higher distances. Therefore for<br />

this scenario we propose:<br />

P LOS<br />

= 1 − d *0.0004,0 < d < 250m<br />

, (9.2)<br />

where d is in meters. This functi<strong>on</strong> is presented in the figure below.<br />

1<br />

Probability of LOS<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

50 100 150 200 250<br />

distance [m]<br />

Figure 9.10: Probability of LOS.<br />

9.3.2.4 DS <strong>and</strong> maximum excess-delay distributi<strong>on</strong><br />

The 10, 50 <strong>and</strong> 90 % values for the Cumulative Distributi<strong>on</strong> Functi<strong>on</strong>s of the distributi<strong>on</strong> of the RMSdelay<br />

spread <strong>and</strong> maximum excess delays are given below for the 5.2 GHz centre-frequency <strong>and</strong> LOS<br />

propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s for this scenario.<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 10 20 30 40 50<br />

RMS delay spread [ns]<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 50 100 150 200 250<br />

Max excess delay [ns]<br />

Figure 9.11: RMS delay spread, LOS<br />

Figure 9.12: Maximum excess delay, LOS<br />

Table 9.4: Percentiles RMS delay spread.<br />

Rms delay spread (ns)<br />

LOS<br />

Page 158 (167)


WINNER D5.4 v. 1.4<br />

10% 4.4<br />

50% 4.5<br />

90% 5.8<br />

mean 5.6<br />

Table 9.5: Percentiles of maximum excess delay [ns].<br />

Maximum excess delay (ns)<br />

LOS<br />

10% 16.7<br />

50% 24.7<br />

90% 41.5<br />

mean 31.9<br />

9.3.2.5 Ricean K factors per tap<br />

The Ricean K-factor as a functi<strong>on</strong> of the distance <strong>and</strong> the CDF of it are shown in Figure 9.13. The K<br />

factor decreases slowly with increase of the distance.<br />

K factor [dB]<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0<br />

-5<br />

-10<br />

measurement based result<br />

linear fitting<br />

K [dB] = 6.4-0.001*d[m]<br />

-15<br />

0 50 100 150 200 250 300<br />

distance [m]<br />

(a)<br />

CDF<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

measurement<br />

based result<br />

Percentiles:<br />

10%: 1.7 dB<br />

50%: 5.9 dB<br />

90%: 11.5 dB<br />

0<br />

-20 -10 0 10 20<br />

K factor [dB]<br />

(b)<br />

Figure 9.13: Scenario B3, LOS: (a) Ricean K factor as a functi<strong>on</strong> of distance, (b) CDF of the Ricean<br />

K factor.<br />

9.3.2.6 Cross-polarizati<strong>on</strong> ratio (XPR)<br />

Prob(XPD)<br />

0.08<br />

0.07<br />

0.06<br />

0.05<br />

0.04<br />

0.03<br />

0.02<br />

0.01<br />

0<br />

-5 0 5 10<br />

XPD [dB]<br />

(a)<br />

Prob(XPD < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-5 0 5 10<br />

XPD [dB]<br />

(b)<br />

Page 159 (167)


WINNER D5.4 v. 1.4<br />

Figure 9.14: XPR 1 under LOS with (a) as PDF <strong>and</strong> (b) as CDF.<br />

Table 9.6: Percentiles of the cross-polarizati<strong>on</strong> ratio XPR 1 .<br />

Propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong><br />

Percentile<br />

(degrees)<br />

LOS<br />

NLOS<br />

10 -0.9 n.a.<br />

50 2.4 n.a.<br />

90 5.3 n.a.<br />

mean 2.4 n.a.<br />

The st<strong>and</strong>ard deviati<strong>on</strong> for the XPR 1 under LOS was found to be 2.50 dB.<br />

9.3.2.7 Azimuth AS at BS <strong>and</strong> MS<br />

The cumulative distributi<strong>on</strong> functi<strong>on</strong>s of the RMS angle-spreads at 5.20 GHz (120 MHz b<strong>and</strong>width) are<br />

shown in Figure 9.15 for LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. The RMS angle-spread is calculated using the<br />

circular angle-spread formula [3GPP SCM]. No statistical fitting comparis<strong>on</strong> based <strong>on</strong> some well known<br />

techniques like KS test is applied. The percentiles for the CDF functi<strong>on</strong>s for the angle-spreads are shown<br />

in the Table 9.7 .<br />

Table 9.7: Percentiles of the RMS azimuth spread.<br />

Propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong> LOS<br />

Link end BS MS<br />

10% 1 5<br />

Percentile (degrees) 50% 5 20<br />

90% 50 88<br />

Prob(angular spread @BS < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 20 40 60 80 100<br />

angular spread @BS [deg]<br />

(a)<br />

Prob(angular spread @MS < Abscissa)<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 20 40 60 80 100<br />

angular spread @MS [deg]<br />

(b)<br />

Figure 9.15: RMS angle-spreads at (a) BS (AoA) <strong>and</strong> (b) MS (AoD) for the Bridge-2-Car scenario<br />

under LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>.<br />

9.3.2.8 Distribtui<strong>on</strong> of the azimuth angles of the multipath comp<strong>on</strong>ents (AI 7.3)<br />

The cumulative distributi<strong>on</strong> functi<strong>on</strong>s of the AoAs <strong>and</strong> AoDs for the multipath comp<strong>on</strong>ents at 5.20 GHz<br />

(120 MHz b<strong>and</strong>width) are shown in Figure 9.16 Figure 9.16Figure 9.16Figure 9.16for LOS propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong>s. The percentiles for the CDF functi<strong>on</strong>s for the AoAs <strong>and</strong> AoDs are shown in the table below.<br />

Table 9.8: Percentiles of the distributi<strong>on</strong> of azimuth.<br />

Page 160 (167)


WINNER D5.4 v. 1.4<br />

Link end BS MS<br />

Propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong><br />

Percentile<br />

(degrees)<br />

LOS NLOS LOS NLOS<br />

10 -23.6 n.a. -121.8 n.a.<br />

50 -1.8 n.a. -1.8 n.a.<br />

90 16.4 n.a. 107.3 n.a.<br />

mean -0.2 n.a. -1.8 n.a.<br />

1<br />

1<br />

Prob(angle @BS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Prob(angle @MS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

-150 -100 -50 0 50 100 150<br />

angle @BS [deg]<br />

0<br />

-150 -100 -50 0 50 100 150<br />

angle @MS [deg]<br />

(a)<br />

(b)<br />

Figure 9.16: CDFs of azimuth angles at (a) BS (AoA) <strong>and</strong> (b) MS (AoD) for the High Mobility<br />

Short Range Hot Spot – (Bridge-2-Car) scenario under LOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>.<br />

9.3.2.9 Angle proporti<strong>on</strong>ality factor<br />

The angle proporti<strong>on</strong>ality factor (r AS ) has been extracted for signals arrive at (or depart from) the MS both<br />

in LOS <strong>and</strong> NLOS propagati<strong>on</strong> c<strong>on</strong>diti<strong>on</strong>s. Figure 9.17 shows the results at the MS <strong>and</strong> BS for LOS. The<br />

percentiles for the CDF of the angle proporti<strong>on</strong>ality factor are shown in Table 9.9.<br />

1<br />

1<br />

Prob(r-factor @BS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Prob(r-factor @MS < Abscissa)<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

0 10 20 30 40<br />

r-factor @BS [deg]<br />

0<br />

0 10 20 30 40<br />

r-factor @MS [deg]<br />

(a)<br />

(b)<br />

Figure 9.17: Angle proporti<strong>on</strong>ality factor at the (a) BS <strong>and</strong> (b) MS under LOS.<br />

Table 9.9: The percentiles for the CDF of the angle proporti<strong>on</strong>ality factor at the BS <strong>and</strong> the MS,<br />

LOS.<br />

Link end BS MS<br />

Propagati<strong>on</strong><br />

c<strong>on</strong>diti<strong>on</strong><br />

LOS NLOS LOS NLOS<br />

Page 161 (167)


WINNER D5.4 v. 1.4<br />

Percentile<br />

(degrees)<br />

10 0 n.a. 1.6 n.a.<br />

50 1.6 n.a. 4.0 n.a.<br />

90 8.1 n.a. 9.3 n.a.<br />

mean 3.7 n.a. 5.3 n.a.<br />

9.3.2.10 Number of ZDSC<br />

The results for the number of the ZDSCs shown in Figure 9.18 are calculated by resampling the data with<br />

100 MHz sampling rate. This has to be c<strong>on</strong>sidered an upper limit of the number of n<strong>on</strong>zero power delay<br />

bins. Table 9.10 presents the 10, 50 <strong>and</strong> 90 percentiles of the cumulative distributi<strong>on</strong> of the number of<br />

ZDSCs.<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 2 4 6 8 10 12 14 16<br />

Number of ZDSC<br />

LOS.<br />

Figure 9.18: Number of ZDSCs.<br />

Table 9.10: CDF values for number of ZDSCs.<br />

Number of ZDSC<br />

LOS<br />

10% 3<br />

50% 4<br />

Percentile<br />

90% 5<br />

mean 4<br />

9.3.2.11 Distributi<strong>on</strong> of ZDSC delays<br />

In Figure 9.19 distributi<strong>on</strong> of the ZDSC delays for LOS envir<strong>on</strong>ment is shown<br />

Page 162 (167)


WINNER D5.4 v. 1.4<br />

0.05<br />

0.04<br />

0.03<br />

PDF<br />

0.02<br />

0.01<br />

0<br />

0 100 200 300 400 500<br />

ZDSC delays [ns]<br />

Figure 9.19: Distributi<strong>on</strong> of the ZDSC delays.<br />

9.3.2.12 Delay proporti<strong>on</strong>ality factor<br />

Figure 9.20 shows the empirical cumulative distributi<strong>on</strong> functi<strong>on</strong> of the delay proporti<strong>on</strong>ality factor, LOS<br />

case. Percentiles of delay proporti<strong>on</strong>ality factor are given in the Table 9.11.<br />

CDF<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

0 1 2 3 4 5<br />

r ds<br />

(a) LOS.<br />

Figure 9.20: Delay proporti<strong>on</strong>ality factor r DS .<br />

Table 9.11: Percentiles of delay proporti<strong>on</strong>ality factor.<br />

Delay proporti<strong>on</strong>ality factor<br />

LOS<br />

10% 1.86<br />

50% 2.20<br />

Percentile<br />

90% 2.59<br />

mean 2.22<br />

9.3.2.13 Channel model tables<br />

Table 9.12: Distributi<strong>on</strong> functi<strong>on</strong>s of large-scale parameters.<br />

Bridge2Car<br />

LOS<br />

Page 163 (167)


WINNER D5.4 v. 1.4<br />

Delayspreadσ<br />

AoD<br />

spreadσ<br />

AoA<br />

spreadσ<br />

τ<br />

φ<br />

ϕ<br />

Shadowing<br />

LN<br />

LN<br />

LN<br />

LN<br />

Table 9.13: Cross-correlati<strong>on</strong> between large-scale parameters.<br />

Bridge2Car<br />

Scenario LOS NLOS<br />

σ<br />

φ vs σ<br />

τ 0.23 n.a.<br />

Cross-Correlati<strong>on</strong>s<br />

σ<br />

ϕ vs σ<br />

τ -0.02 n.a.<br />

σ<br />

ϕ vs LNS 0.17 n.a.<br />

σ<br />

φ vs LNS 0.17 n.a.<br />

σ<br />

τ vs LNS -0.19 n.a.<br />

σ<br />

φ vs σ<br />

ϕ 0.13 n.a.<br />

Table 9.14: Additi<strong>on</strong>al parameters required for generati<strong>on</strong> of large-scale parameters.<br />

σ τ<br />

Scenarios<br />

ν<br />

(S-dB)<br />

ζ<br />

(S-dB)<br />

∆ τ<br />

Bridge2<br />

Car<br />

LOS<br />

-8.26<br />

0.2<br />

0.28<br />

σ φ<br />

(m)<br />

ν<br />

(deg-dB)<br />

ζ<br />

(deg-dB)<br />

∆ φ<br />

1.07<br />

0.31<br />

3<br />

σ ϕ<br />

(m)<br />

ν<br />

(deg-dB)<br />

ζ<br />

(deg-dB)<br />

1.24<br />

0.48<br />

∆ 4.6<br />

ϕ<br />

Page 164 (167)


WINNER D5.4 v. 1.4<br />

LNS<br />

(m)<br />

ζ<br />

(dB)<br />

∆ LNS<br />

(m)<br />

2.3<br />

3.2<br />

Table 9.15: Lambda parameters.<br />

Bridge2Car<br />

LOS<br />

λ<br />

1 (m) 0.28<br />

λ<br />

2 (m) 3.0<br />

λ<br />

3 (m) 4.6<br />

λ<br />

4 (m) 3.2<br />

Figure 9.21: The auto correlati<strong>on</strong> functi<strong>on</strong>s obtained from (*) using the λ parameters of the table<br />

above <strong>and</strong> the single exp<strong>on</strong>ential functi<strong>on</strong>s obtained from measurements from Scenario<br />

Bridge2Car.<br />

Table 9.16: K factor formulae for LOS scenarios.<br />

Scenarios Bridge2Car<br />

K [dB] 6.4 - 0.001*d [m]<br />

Table 9.17: Distributi<strong>on</strong>s of azimuth <strong>and</strong> departure angles.<br />

Scenarios<br />

AoD<br />

distributi<strong>on</strong><br />

LOS<br />

Bridge2Car<br />

NLOS<br />

Wrapped Gaussian<br />

Page 165 (167)


WINNER D5.4 v. 1.4<br />

AoD<br />

scaling<br />

parameter<br />

AoA<br />

distributi<strong>on</strong><br />

AoA<br />

scaling<br />

parameter<br />

5.3σ<br />

φ<br />

n.a.<br />

Wrapped Gaussian<br />

3.7σ<br />

ϕ<br />

n.a.<br />

Table 9.18: Number of ZDSCs <strong>and</strong> the number of rays in each cluster.<br />

Scenarios<br />

Bridge2Car<br />

LOS<br />

NLOS<br />

Number of ZDSC 4 n.a.<br />

rays per ZDSC 10<br />

AS<br />

φ (deg) 5.5 n.a.<br />

AS<br />

ϕ (deg) 17.8 n.a.<br />

Table 9.19: Path loss <strong>models</strong>.<br />

Scenario path loss [dB] shadow fading<br />

st<strong>and</strong>ard dev.<br />

applicability<br />

range<br />

Bridge2Car<br />

LOS 19.3 log 10 (d[m]) + 60.6 s = 3.1 dB 3m < d < 250m<br />

NLOS n.a n.a n.a<br />

Table 9.20: Scenario: LOS Clustered delay line model.<br />

ZDSC<br />

#<br />

delay<br />

[ns]<br />

Power<br />

[dB]<br />

AoD<br />

[º]<br />

AoA<br />

[º]<br />

K-<br />

factor<br />

[dB]<br />

MS speed = 1.5 km/h,<br />

directi<strong>on</strong> U(0 o ,360 o )<br />

1 0 0 4,6 -1,6 21 -0.03 * -31 **<br />

2 5 -3.1 -0,6 1,7 4 -4.56 -18.5<br />

3 10 -6.2 -4,2 2,6 -16.2<br />

4 15 -9.3 -13,0 3,6 -19.3<br />

5 20 -12.4 -17,7 6,4 -22.4<br />

6 25 -15.5 -23,7 6,5 -25.5<br />

7 30 -18.6 -35,6 3,7 -28.6<br />

8 40 -24.8 -37,1 2,4<br />

*<br />

**<br />

+<br />

- ∞<br />

Power of dominant ray,<br />

Power of each other ray<br />

Clusters with high K-factor will have 11 rays.<br />

Number of rays/ZDSC = 10<br />

Ray Power [dB]<br />

-24.8<br />

ZDSC AS at MS [º] =<br />

17.8<br />

ZDSC AS at BS [º] = 5.5<br />

Composite AS at MS [º] =<br />

5.6<br />

Composite AS at BS [º] =<br />

1.4<br />

Page 166 (167)


WINNER D5.4 v. 1.4<br />

9.4 Literature review for other scenarios<br />

9.4.1 Scenario “high mobility short range hot spot”<br />

9.4.1.1 Reference data<br />

The measurement data for the bridge to car setup were gathered with partly support by the WINNER<br />

project at a highway bridge close to Ulm (Germany). Measurement b<strong>and</strong>width <strong>and</strong> centre-frequency were<br />

selected to be 120 MHz <strong>and</strong> 5.2 GHz. The BS was mounted at the bridge (height of ~5.5m with a down<br />

tilt of 45°) whereby the transmit antennas were fixed <strong>on</strong> the roof of a car (~2m height). During the<br />

measurement LOS was dominating the propagati<strong>on</strong> characteristics, after the car went under the bridge to<br />

situati<strong>on</strong> changed to NLOS.<br />

9.4.1.2 Publicati<strong>on</strong>s<br />

Only few publicati<strong>on</strong>s to <strong>channel</strong> measurements <strong>and</strong> modelling for this scenario can be found. The spatial<br />

<strong>channel</strong> in [YTL02] for the bridge to car scenario is modelled as LOS propagati<strong>on</strong> <strong>and</strong> was used for<br />

MIMO capacity analysis.<br />

In [THL+01] first measurement results for the c<strong>on</strong>sidered scenario were published. Those data were used<br />

for the measurement based parametric <strong>channel</strong> modelling (MBPCM) approach for rec<strong>on</strong>structing <strong>channel</strong><br />

impulse resp<strong>on</strong>ses with different antenna c<strong>on</strong>figurati<strong>on</strong>s.<br />

The measurement data presented in [TSS+03] fit into the c<strong>on</strong>sidered scenario for the WINNER project.<br />

SIMO measurements with a uniform rectangle array (URA) at the receive side (bridge) were performed<br />

<strong>and</strong> are available for download. Results of super resoluti<strong>on</strong> estimati<strong>on</strong>s for angle of arrival in azimuth <strong>and</strong><br />

elevati<strong>on</strong> can be found when downloading the data. No further analysis results were published.<br />

In [TLS+05] the measured bridge to car scenario partly supported by the WINNER project was published.<br />

Analysis results in teRMS of delay window, RMS delay spread, RMS angle-spreads for transmit <strong>and</strong><br />

receive azimuth, number of paths, relative power of dense multipath comp<strong>on</strong>ents <strong>and</strong> joint probability<br />

densities between RMS delay <strong>and</strong> angle-spreads are shown. Those measurements <strong>and</strong> analysis results are<br />

part of the D5.4.<br />

Page 167 (167)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!