Heterrogen Network

Università degli Studi di Bologna
Dipartimento di Elettronica Informatica e Sistemistica, DEIS
Dottorato di Ricerca in Ingegneria Elettronica,

Informatica e delle Telecomunicazioni
XIX Ciclo
Wireless Heterogeneous Networks
Coordinatore:
Tesi di:
Chiar.mo Prof. Ing. Paolo Bassi
Ing. Claudio Gambetti
Relatore:
Chiar.mo Prof. Ing. Oreste Andrisano
Settore scientifico-disciplicare:
ING-INF/03 TELECOMUNICAZIONI
Contents
Introduction 1
1 TD-SCDMA System Simulator design 9

1.1 TD-SCDMA air interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2 TD-SCDMA System Simulator . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.1 Link Level simulator . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.2 Link-to-Network level Interface module . . . . . . . . . . . . . . . . 17
1.2.3 Network Level simulator . . . . . . . . . . . . . . . . . . . . . . . . 19
1.2.4 Upper Layer simulator . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3 How to convert the TD-SCDMA system simulator to the UTRA FDD option 24
2 Performance of TD-SCDMA in mixed CS/PS traffic scenarios 31

2.1 Packet scheduler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Power control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3 Scenario and propagation environment . . . . . . . . . . . . . . . . . . . . 34
2.4 Performance Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.5 Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5.1 Circuit switched services . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5.2 Packet and circuit switched services . . . . . . . . . . . . . . . . . . 40
2.5.3 CS/PS performance enhancements . . . . . . . . . . . . . . . . . . 41
3 Link Level aspects modelling in the simulation of packet switched wire-

less networks 47
3.1 Wireless system simulations . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Link and Network Level Analysis . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.1 Network level simulator with a large simulation step . . . . . . . . . 49
3.2.2 Network level simulator with small value of simulation step . . . . . 50
3.3 An Example on how to interface Link and Network Levels . . . . . . . . . 53
3.3.1 Link Level to Interface Module communication . . . . . . . . . . . . 54
i
ii Contents
3.3.2 Interface Module to Network Level communication . . . . . . . . . 54

3.4 An experiment for the validation of the proposed link-to-network interface
method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4 Architectures for Heterogeneous Wireless Networks 61

4.1 Service interworking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2 Proposed architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3 Architectural and implementation issues . . . . . . . . . . . . . . . . . . . 65
4.3.1 System Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3.2 Implementation issues . . . . . . . . . . . . . . . . . . . . . . . . . 67
5 Common Radio Resource Management UMTS & WLAN 71

5.1 The CRRM challange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.2 Interactions between CRRM and RRMs . . . . . . . . . . . . . . . . . . . 73
5.2.1 Network topology information . . . . . . . . . . . . . . . . . . . . . 74
5.2.2 Network load report . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2.3 RRM report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.2.4 CRRM decision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.3 Local RRM functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.4 CRRM Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.4.1 CRRM ”Service Based” . . . . . . . . . . . . . . . . . . . . . . . . 79
5.4.2 CRRM ”Coverage Based” . . . . . . . . . . . . . . . . . . . . . . . 81
5.4.3 CRRM ”QoS Based” . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.5 Software simulation platform settings . . . . . . . . . . . . . . . . . . . . . 84
5.5.1 Upper Layers Simulator-ULS . . . . . . . . . . . . . . . . . . . . . 84
5.5.2 UMTS LLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.5.3 WLAN LLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.6 Performance measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.7 Traffic scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.7.1 Network topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.7.2 Traffic distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.8 Numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Conclusions 107
A SHINE: Simulation platform for Heterogeneous Interworking Networks109

A.1 Simulation platform Structure . . . . . . . . . . . . . . . . . . . . . . . . . 110
A.1.1 Flexibility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Contents iii
A.1.2 Time efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

A.2 ULS and LLSs main tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
A.3 Time axis management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
A.4 ULS-LLSs communications . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Bibliography 119
Acknowledgments 125
iv Contents
List of Tables
1.1 3GPP modes: FDD, TDD 3.84 Mcps, TDD 1.28 Mcps . . . . . . . . . . . 25
2.1 System parameters fixed in the numerical results . . . . . . . . . . . . . . . 36

2.2 Packet switched session parameters for web browsing services . . . . . . . . 37
2.3 Default values for (Eb /I0 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1 Definition of quantities exchanged between the interface module and the
network simulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.1 Initial-RAT selection algorithm in the hotspot. . . . . . . . . . . . . . . . . 80

5.2 Set of parameters adopted for IEEE 802.11e . . . . . . . . . . . . . . . . . 89
5.3 Traffic distribution and arrival rates . . . . . . . . . . . . . . . . . . . . . . 95
5.4 Adopted traffic classes: parameters and requirements for satisfaction . . . . 96
v
vi Contents
List of Figures
1 Scenarios for wireless heterogeneous networks . . . . . . . . . . . . . . . . 4
1.1 TD-SCDMA physical channel signal format . . . . . . . . . . . . . . . . . 12

1.2 Structure of the TD-SCDMA sub-frame . . . . . . . . . . . . . . . . . . . . 13
1.3 TD-SCDMA System Simulator: block diagram . . . . . . . . . . . . . . . . 15
1.4 TD-SCDMA Link Level simulator: block diagram . . . . . . . . . . . . . . 16
1.5 Link-to-Network level Interface module: block diagram . . . . . . . . . . . 18
1.6 Main functional blocks of the TD-SCDMA simulator . . . . . . . . . . . . 20
1.7 event 1G for TD-SCDMA: a P-CCPCH RSCP becomes better than the
previous best P-CCPCH RSCP . . . . . . . . . . . . . . . . . . . . . . . . 28
1.8 W-CDMA basic handover algorithm . . . . . . . . . . . . . . . . . . . . . . 29
2.1 The considered network scenario . . . . . . . . . . . . . . . . . . . . . . . . 35

2.2 Voice satisfaction rate (Tsat ) vs. cell radius . . . . . . . . . . . . . . . . . . 39
2.3 Areas of user satisfaction in the scenario . . . . . . . . . . . . . . . . . . . 39
2.4 Voice satisfaction rate (Tsat ) vs. downlink data packet traffic (TP S ), for
different values of voice offered traffic (TCS ) . . . . . . . . . . . . . . . . . 40
2.5 Downlink active session throughput (AST ) vs. downlink data packet traffic
(TP S ), for different values of voice offered traffic (TCS ) . . . . . . . . . . . . 41
2.6 Network Performance as a function of (Eb /Io )U L−CS . . . . . . . . . . . . . 42
2.7 Network Performance as a function of (Eb /Io )DL−CS . . . . . . . . . . . . . 43
2.8 Network Performance as a function of (Eb /Io )U L−P S . . . . . . . . . . . . . 44
2.9 Network Performance as a function of (Eb /Io )DL−P S . . . . . . . . . . . . . 45
3.1 Link and Network levels in operation . . . . . . . . . . . . . . . . . . . . . 49

3.2 Example of data collection of Nbit−err of the transmitted blocks . . . . . . 52
3.3 Link tool-to-interface module communication: block diagram . . . . . . . . 54
3.4 Interface module-to-network simulator communication: block diagram . . . 55
3.5 Validation experiment: block diagram . . . . . . . . . . . . . . . . . . . . . 58
3.6 BER as a function of h(Eb /Io )i and h(Eb /Io )iT B . . . . . . . . . . . . . . . . 59
vii
viii Contents
4.1 Loose coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.2 Gateway approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3 Tight coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 UMTS - WLAN Inter-working architecture: proposed logical scheme . . . . 66
4.5 Node-B and WLAN integration: implementation . . . . . . . . . . . . . . . 68
4.6 Evolved multi-standard (UMTS & WLAN) NodeB communicating to the
Common Radio Resource Management (CRRM) . . . . . . . . . . . . . . . 68
5.1 RRM & CRRM relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5.2 Flow diagram of interactions between CRRM and RRMs . . . . . . . . . . 77
5.3 CRRM Service Based . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.4 Flow diagram of CRRM Service Based . . . . . . . . . . . . . . . . . . . . 81
5.5 CRRM Coverage Based . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.6 CRRM QoS Based . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.7 Flow diagram of CRRM QoS Based . . . . . . . . . . . . . . . . . . . . . . 83
5.8 Simulation platform architecture. . . . . . . . . . . . . . . . . . . . . . . . 84
5.9 Investigated scenario: WLAN APs in hotspot of high density traffic . . . . 93
5.10 Simulated scenario. The grey square indicates the area considered for nu-
merical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.11 fa (t): voice call arrival rate VoiceHS in the hotspot . . . . . . . . . . . . . 96
5.12 Voice QoS in the investigated 100x100 m2 area: users’ Satisfaction Rate
SatR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.13 Speech Service Access QoS in the investigated 100x100 m2 area: users’ Call
Setup Success Rate CSSR . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.14 Speech Service Retainability QoS in the investigated 100x100 m2 area:
users’ Drop Call Rate DCR . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.15 Speech Service Integrity QoS in the investigated 100x100 m2 area: users’
Outage Rate OutR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.16 Number of Intersystem Handover procedures per voice call . . . . . . . . . 103
5.17 Distribution of voice call users in the two networks covering the hotspot
(k = 0.5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.18 Distribution of voice call users in the two networks covering the hotspot
(k = 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.19 Ftp sessions in the hot spot: average perceived throughput . . . . . . . . . 105
A.1 Simulator platform global architecture. . . . . . . . . . . . . . . . . . . . . 111

A.2 Simulation platform architecture. Access networks side. . . . . . . . . . . . 112
Contents ix
A.3 ULS-LLS communications. . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

A.4 ULS-LLSs communications scheme (cellular network notation) implemented
in SHINE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Introduction
In the last years, mobile communications have become pervasive to all activities of soci-
ety; the number of mobile phones and wireless Internet users has increased significantly.
Changing private and professional lifestyles has created a surging demand for communi-
cations on the move, reachability and wireless broadband.
In fact, latest industrial surveys1 reveal 2.255 billion subscriptions in the family of
all fully open standard cellular technologies GSM, GPRS, EDGE and WCDMA-HSDPA
(31.12.2006); 511 million of these subscriptions were just added in 2006, corresponding to
a growth of 29% in 2006. The third generation networks (WCDMA-HSDPA) now count
almost 100 million subscriptions (31.12.06), with an average monthly growth in 2006 of
over 4 millions and an annual growth of more than 100% in 2006; the forecast is that by
end-2009, WCDMA-HSxPA subscriptions will be half billion.
Mobile networks evolution

Among all these different mobile technologies, traditionally, first generation (in Italy it
was TACS - Total Access Communication System) and second generation (GSM - Global
System for Mobile Communications) wireless networks were primarily targeted at voice
communications and finally at data communications occurring at low data rates. The first-
phase GSM specifications provided only basic transmission capabilities for the support of
1
GSA - Global mobile Suppliers Association - representing GSM/EDGE/WCDMA suppliers globally;
mobile subscribers data source: Informa Telecoms & Media.
1
2 Introduction
data services, with the maximum data rate in these early networks being limited to 9.6
Kbps on one timeslot in each radio frame.
About 5 years ago, the most advanced cellular technology for mobile internet access
became the GSM implementing the High-Speed Circuit-Switched Data (HSCSD) evolution,
specified in ETSI Rel’96; it was the first GSM Phase 2+ work item that clearly increased
the achievable data rates in the GSM system. The maximum radio interface bit rate
of an HSCSD configuration with 14.4 Kbps channel coding (corresponding to the best
radio conditions) multiplying 4 timeslots is 57.6 Kbps: this was broadly equivalent to
providing the same transmission rate as that available over one ISDN B-Channel. It
seems prehistory, but just 5 years ago this was a great achievement!
Quickly, the GSM networks were upgraded to 2.5G by introducing the General Packet
Radio System (GPRS) technology. GPRS provides GSM with a packet data air interface
and an IP-based core network, with bit rates varying from 9 Kbps to more than 150 Kbps
per user when all eight timeslots of a GSM carrier are assigned to a single GPRS Mobile
Station for exclusive use.
The Enhanced Data Rates for Global Evolution (EDGE) was a further innovation step
of GSM packet data and now EDGE is widely deployed on global GSM networks. Thanks
to the 8 phase shift keying (8PSK) modulation scheme, EDGE can handle about three
times more data subscribers than GPRS, or triple the data rate for one end-user. EDGE
is specified in a way to enhance the throughput per timeslot for both HSCSD and GPRS.
The enhancement of HSCSD is called ECSD (Enhanced Circuit Switched Data), whereas
the enhancement of GPRS is called EGPRS (Enhanced General Packet Radio System).
In ECSD, the maximum data rate does not exceed 64 Kbps because of the restrictions
in the A-interface, but the data rate per timeslot triples. Similarly, in EGPRS, the data
rate per timeslot triples and the peak throughput, with all eight timeslots in the radio
interface, can reach 473 Kbps.
On the other hand, in the last couple of years, we have also seen a strong development
of third-generation (3G) wireless systems that incorporate the features provided by broad-
band. In addition to support mobility, broadband aims at supporting multimedia traffic,
with quality of service (QoS) assurance; four class of services are considered: conversa-
tional (both speech and video calls), streaming, interactive and background. In Europe,
the 3G standard has been initially developed by ETSI (European Telecommunication
Standard Institute), then the work has been continued by Third Generation Partnership
Project (3GPP) under the designation of UMTS (Universal Mobile Telecommunications
System).
The radio access interface of UMTS comprises two standards for operation adopting
Introduction 3
the Frequency Division Duplex (FDD) and Time Division Duplex (TDD) modes. The
present UMTS FDD networks, based on wideband CDMA (WCDMA), adopt new spec-
trum and new radio network configurations while using the same GSM core infrastructure.
The maximum data rate in the first WCDMA release (Rel’99) is 2 Mbps, but in practice
the widely used maximum downlink rate (i.e. direction from NodeB to User Equipment)
is equal to 384 Kbps.
With the really recent addition of the High Speed Downlink Packet Access (HSDPA)
specified in 3GPP Rel’5, that is a sort of 3.5G technology, WCDMA network operators
aim at providing extremely high data rate multimedia services and to improve spectral
efficiency by higher order modulation using 16-QAM: in March 2007, some of the deployed
WCDMA-HSDPA networks are starting to support calls at 7.2 Mbps gross data rate
(corresponding to a 6.7 Mbps net data rate) with category 8 mobiles: about 5 years are
passed since the HSCSD introduction, and the speed of data transfer over cellular networks
has increased of more than 100 times! Moreover, these WCDMA-HSDPA networks will
soon achieve data rates in downlink up to 14 Mbps with category 10 mobiles.
Similarly, in the next months in 2007, the High Speed Uplink Packet Access (HSUPA),
standardized in 3GPP Rel’6, will complement HSDPA by significantly reducing latency on
the uplink and offering data speeds up to 5.8 Mbps (peak) on the uplink channel. Together
with HSDPA, it means a huge stride in WCDMA-HSxPA2 network performance.
In this extremely fast changing and widened context of mobile networks, my work
was initially addressed to evaluate the performance of the innovative 3G networks and to
study the impact of physical layer parameters on the network performance. Based on one
of the main requirements for 3G systems, that is the ability to support asymmetric up-
link/downlink traffic, the choice of the 3G radio interface to be studied has been directed
to one of the two TDD modes, the Time-Division Synchronous Code Division Multiple
Access (TD-SCDMA): thanks to its TDD/TDMA characteristics, the TD-SCDMA net-
work can adapt the uplink/downlink ratio according to the data load within a single
unpaired frequency thus utilizing the spectrum more efficiently. This is especially helpful
in an environment with increasing data traffic (mobile data), which tends to be asymmet-
ric, often requiring little uplink throughput, but significant bandwidth for downloading
information (mobile Internet).
Wireless heterogeneous networks

Coming back to the development of wireless networks, we can observe that some alter-
native operators are already offering wireless broadband Internet access with WiFi or
2
Here with the term HSxPA, we mean both the HSDPA and the HSUPA technologies.
4 Introduction
(a) WLAN AP in high traffic density area glob- (b) WLAN AP to improve indoor coverage
ally covered by 3G
Figure 1: Scenarios for wireless heterogeneous networks
(pre-)WiMAX 802.16d networks. For example Wireless Local Area Networks (WLAN)
are achieving a great penetration in the mass market as a really effective solution to
provide mobile access to the Internet; companies all over the world are already offering
WLAN connections in particular locations, such as airports, hotels or caffés (see figure
1). In these areas, the so-called ”hot spots”, anyone owning the appropriate technology
on his laptop can connect to the Internet at a reasonable price and with a satisfactory
connection speed.
Nonetheless, the request for higher bit rates is expected to further increase in the next
future and more capacity will be necessary; on these conditions, 3G/WLAN interworking
becomes a really significant issue to be investigated: provided that the WLAN hot spot is
within the 3G network coverage and that the final user is equipped with a dual mode ter-
minal, integrating the two technologies, thus increasing the ”pool” of available resources,
would considerably increase both users’ satisfaction and networks’ utilization efficiency.
In this phase, I therefore extended my analysis from the initial scenario of a 3G ”stand-
alone” network to a full 3G/WLAN heterogeneous network: firstly, the feasibility of the
integration of these two technologies in a single system has to be evaluated; afterwards,
the possible methods for a Common Radio Resource Management of the two radio access
networks have been studied in depth.
Introduction 5
Thesis Outline
This dissertation mainly deals with the study of third-generation and wireless heteroge-
neous networks.
Chapter 1 presents the design of the third-generation dynamic system simulator I
developed during this activity: the standard that has been chosen is the UTRA TDD
1.28 Mcps, also called TD-SCDMA. The main functional blocks composing this tool are
the Link Level simulator, the Network Level simulator, an Interface module between the
link and the network levels and finally the Upper Layer simulator, which receives the
input from a Mobility simulator and a User Activity simulator. An overview of the main
changes required to implement UTRA FDD are also shown.
Chapter 2 deals with the analysis of the performance of TD-SCDMA network through
system simulation. The influence of packet switched applications on the overall network
performance is investigated; in order to balance the combined quality of voice and data
services, the tuning of physical layer parameters for the power control algorithm is eval-
uated.
Chapter 3 describes the method we propose on how to interface link and network level
tools through an ”instant value” interface. Although this approach has been adopted in
the TD-SCDMA system simulator, it’s quite general and can be implemented in arbitrary
wireless networks with a time slotted physical frame structure. The proposed approach
allows a thorough analysis of performance in networks with mixed circuit and switched
services.
Chapter 4 introduces a possible architectural solution for a heterogeneous wireless
system exploiting the complementary characteristics of the radio interfaces of a third-
generation network and a wireless LAN. Both a logical scheme and a feasible realization
are provided; a link between each local UMTS and WLAN Radio Resource Management
(RRM) entity and a Common RRM (CRRM) entity is proposed.
Chapter 5 describes the possible advantages introduced by the CRRM for a heteroge-
neous integrated and interworking UMTS-WLAN network; the performance evaluation is
carried out through full simulations from the physical to the application layer. Firstly, the
required interactions and information exchanged among the CRRM entity and the local
RRM entities are presented; afterwards a fully configurable CRRM algorithm is proposed:
the project is composed of a Coverage Based, a Service Based and a Quality of Service
(QoS) Based CRRM algorithm. Finally, the trend in the system capacity provided with
the various CRRM options is discussed in a realistic scenario.
Introduction 7
This work includes parts from papers under IEEE copyrights. In particular, text and
figures are here reprinted with permission from:
c
°[2004] IEEE. C.Gambetti, A.Zanella, R.Verdone, O.Andrisano, “Performance of a
TD-SCDMA cellular system in the presence of circuit and packet switched services”, IEEE
VTC 2004 Spring, Milan (Italy), 17-19 May, 2004.
c
°[2004] IEEE. C.Gambetti, A.Zanella, “Trade-off between Data Throughput and
Voice User Satisfaction in TD-SCDMA Networks: the Impact of Power Control”, IEEE
WPMC 2004, Abano Terme (Italy), 12-15 Sept., 2004.
c
°[2005] IEEE. O. Andrisano, A. Bazzi, M. Diolaiti, C. Gambetti, G. Pasolini, “UMTS
and WLAN Integration: Architectural Solution and Performance”, IEEE PIMRC 2005,
Berlin (Germany), 11-14 Sept., 2005.
c
°[2006] IEEE. A.Bazzi, M.Diolaiti, G.Pasolini, C.Gambetti, “WLAN Call Admission
Control Strategies for Voice Traffic over Integrated 3G/WLAN Networks”, IEEE CCNC
2006, Las Vegas (USA), January, 2006.
c
°[2006] IEEE. L.Zuliani, C.Gambetti, A.Zanella, O.Andrisano, “Link Level Aspects
Modelling in the Simulation of Packet Switched Wireless Networks”, IEEE WCNC 2006,
Las Vegas (USA), 3-6 April, 2006.
c
°[2006] IEEE. A. Bazzi, C. Gambetti, G. Pasolini, “SHINE: Simulation platform for
Heterogeneous Interworking Networks”, IEEE ICC 2006, Istanbul (Turkey), 11-15 June
2006.
Chapter 1
TD-SCDMA System Simulator

design
Research activities on current and future communication systems are more and more
carried out by means of simulation tools, either developed ad hoc for specific purposes or
already existing, such as Opnet [1], ns-2 [2], Glomosim [3].
This trend is mainly due to the increasing complexity of current and forthcoming
technologies as well as to the frequent need of complete investigations, embracing the
whole protocol stack (from physical to application levels), for which simulation is the
only feasible way to get some insights of the performance provided to the final user.
However, as emphasized also in [4], the realization of a reliable simulator of a commu-
nication system is quite a tricky task, especially when wireless technologies are concerned
which require an accurate modelling of the physical level as well as an adequate charac-
terization of the radio channel behavior.
Off-the-shelf network simulation tools, such as the aforementioned Opnet, ns-2 and
Glomosim, generally adopt simplified approaches to model the physical level behavior of
the supported wireless technologies. Usually they only account for the path-loss effect or
at most a simplified channel model; in all cases, however, accurate bit-level simulations
9
10 TD-SCDMA System Simulator design
are neglected [4, 5, 6, 7].
At first sight, this choice is acceptable since bit level simulations, which require an
accurate implementation of all physical level aspects of the investigated technology (prop-
agation, channel coding and decoding, interleaving, modulation and demodulation,etc.),
are very consuming in terms of simulation time. Nonetheless the possible lack of accuracy
of physical level characteristics of such tools is felt as problem to be overcome by many
researchers [6, 7, 8] and some effort has been made in this direction [9].
Let us emphasize that in this work Physical Level simulators are conceived as a part
of complete system simulators aimed at reproducing the entire protocol stack (from ap-
plication to propagation). Within this context, the task of a physical simulator is no
more to simply provide curves of bit error rate or packet error rate characterizing the
performance of the investigated technology at physical level, but, on the contrary, its task
is to interact with the simulation tool which reproduces the upper layers behaviors (from
MAC to Application), hereafter denoted as Network Level simulator. Let us observe that
this task has to be performed for each user within the investigated scenario, that is, for a
number of links which could be very relevant.
Moreover, without such an integrated approach from physical to application level, the
simulation of advanced wireless heterogeneous networks, which is the main objective of
this thesis, would be quite rough: actually, the final direction of our study is to investigate
whether a Common Radio Resource Management entity (see chapter 5) could better the
system capacity, by exploiting in real time the complementary characteristics offered by
the different radio access technologies. In this context, it’s therefore strongly required to
build for each radio access stratum a System Simulator reproducing with accuracy the
main characteristics of the physical layer, as well the main aspects of the related datalink
layer, the local Radio Resource Management entity and the upper layer properties.
In this first chapter, the design of the 3G (third generation) system simulator I devel-
oped during the thesis is presented.
In section 1.1, an overview of the technology is given, with reference to the selected UMTS
standard, the UTRA TDD 1.28 Mcps option, ordinary called TD-SCDMA. Afterwards,
in section 1.2 the design of the TD-SCDMA system simulator is described, whose main
functional block are the Link Level simulator, the Network Level simulator and the Upper
Layer simulator. Finally in section 1.3, the main changes required to implement UTRA
FDD will be shown.
TD-SCDMA System Simulator design 11
1.1 TD-SCDMA air interface

TD-SCDMA, which stands for Time Division Synchronous Code Division Multiple Ac-
cess, is an innovative mobile radio standard for the physical layer of a 3G air interface.
It has been adopted by ITU and by 3GPP as part of UMTS release 4, becoming in this
way a global standard, which covers all radio deployment scenarios: from rural to dense
urban areas, from pico to micro and macrocells, from pedestrian to high mobility.
TD-SCDMA combines an advanced TDMA/TDD system with an adaptive CDMA com-
ponent operating in a synchronous mode.
TD-SCDMA offers several unique characteristics for 3G services. In particular its TDD
nature allows TD-SCDMA to master asymmetric services more efficiently than other 3G
standards (for example, the ordinary UTRA FDD, known as W-CDMA from the ITU
terminology). Up and downlink resources are flexibly assigned according to traffic needs,
and flexible data rate ranging from 1.2 Kbit/s to 2Mbit/s are provided. This is especially
helpful in an environment with increasing data traffic (mobile data), which tends to
be asymmetric, often requiring little uplink throughput, but significant bandwidth for
downloading information (mobile Internet).
Many radio technology, such as GSM, EDGE, W-CDMA or cdma2000, require separate
bands for uplink and downlink (paired FDD spectrum). In this case with asymmetric
loads, such as Internet access, portions of the spectrum are occupied but not used for
data transfer. These idle resources cannot be utilized for any other service, leading to an
inefficient use of the spectrum. On the contrary, TD-SCDMA adapts the uplink/downlink
ratio according to the data load within a single unpaired frequency band, thus utilizing
the spectrum more efficiently.
Highly effective technologies like smart antennas, joint detection and dynamic channel
allocation are integral features of the TD-SCDMA radio standard. They contribute to
minimize intra-cell interference (typical of every CDMA technology) and inter-cell inter-
ference leading to a considerable improvement of the spectrum efficiency. This is especially
helpful in high-populated areas, which are capacity driven and require an efficient use of
the available spectrum. TD-SCDMA can also cover large areas (up to 40 Km) and sup-
ports high mobility. It is therefore well suited to provide mobile services to subscribers
driving on motorways or travelling on high-speed trains.
In order to mitigate the effect of interference and improve the coverage at the cells
edge, conventional CDMA 3G systems have to use the so-called soft handover when an
ongoing call needs to be transferred from one cell to another as a user moves through
the coverage area. During soft handover, however, the users terminal has concurrent
traffic connections with more than one base station. To handle this increased traffic more
Figure 1.1: TD-SCDMA physical channel signal format
channel units and leased lines are required, resulting in higher operating costs. Thanks
to joint detection, smart antennas and an accurate terminal synchronization TD-SCDMA
does not need to rely on soft handover.
Here the basic technological principles on which the TD-SCDMA technology is based
are summarized:
• TDD (Time Division Duplex) allows uplink and downlink on the same fre-
quency band and does not require paired bands. In TDD, uplink and downlink are
transmitted in the same frequency channel but at different times. It is possible to
change the duplex switching point and move capacity from uplink to downlink or
vice versa, thus utilizing spectrum optimally. It allows for symmetric and asymmet-
ric data services.
For symmetric services used during telephone and video calls, where the same
amount of data is transmitted in both directions, the time slots are split equally
between the downlink and uplink.
For asymmetric services used with Internet access (download), where high data vol-
umes are transmitted from the base station to the terminal, more time slots are
used for the downlink than the uplink.
• TDMA (Time Division Multiple Access) is a digital technique that divides

each frequency channel into multiple time-slots and thus allows transmission chan-
nels to be used by several subscribers at the same time (see figure 1.1).
TD-SCDMA [56] uses a 5 ms sub-frame subdivided into 7 time slots of 675 µs du-
ration each, which can be flexibly assigned to either several users or to a single user
Figure 1.2: Structure of the TD-SCDMA sub-frame
who may require multiple time slots.

Each TD-SCDMA sub-frame contains 3 special time slots DwPTS (Downlink Pilot),
GP (Guard Period) and UpPTS (Uplink Pilot) lasting for 75 µs (96 chips), 75 µs
(96 chips) and 125 µs (160 chips) respectively (see figure 1.2). The total duration
of these 3 slots is 275 µs.
• CDMA (Code Division Multiple Access) increases the traffic density in each
cell by enabling simultaneous multiple-user access on the same radio channel. Yet
each user can interfere with another, which leads to multiple access interference
(MAI).
In TD-SCDMA, within each time slot a number of up to 16 CDMA codes may
be transmitted (maximum CDMA loading factor). Using a chip rate of 1.28 Mcps
allows a carrier bandwidth of 1.6 MHz. According to its operating license, the
network operator can deploy multiple TD-SCDMA 1.6 MHz carriers. Each radio
resource unit is thus identified by a particular time slot and a particular code on a
particular carrier frequency.
• Joint Detection (JD) allows the receiver to estimate the radio channel and works
for all signals simultaneously. Through parallel processing of individual traffic
streams, JD eliminates the multiple access interference (MAI) and minimizes intra-
cell interference, thus increasing the transmission capacity.
The efficiency of the Joint Detection receiver in TD-SCDMA technology is based on
the TDMA/TDD operation and on the limited number of codes employed per time
slot: the total number of users per radio carrier is distributed over the different time
slots of the basic TDMA frame, so that a maximal number of 16 codes per time slot
per radio carrier can be easily processed in parallel and detected.
To compare the two systems, in UTRA FDD 256 CDMA codes might be transmit-
ted: due to the high number of codes, the implementation of an optimal multi-user
receiver in FDD is difficult, since the implementation complexity is an exponential
function of the numbers of codes. In order to combat MAI, UTRA FDD emploies
suboptimal detection schemes, such as the Rake receiver, which do not extract all
CDMA codes in parallel.
• Dynamic Channel Allocation (DCA): the advanced TD-SCDMA air interface

takes advantage of all available Multiple Access techniques (TDMA (Time Division
Multiple Access), FDMA (Frequency Division Multiple Access), CDMA (Code Di-
vision Multiple Access) and SDMA (Space Division Multiple Access)).
Making an optimal use of these degrees of freedom, TD-SCDMA provides an adap-
tive allocation of the radio resources according to the interference scenario, mini-
mizing intercell interference.
• Mutual Terminal Synchronization: like all TDMA systems, TD-SCDMA needs

an accurate synchronization between mobile terminal and base station. This syn-
chronization becomes more complex through the mobility of the subscribers, because
they can stay at varying distances from the base station and their signal presents
varying propagation delays.
Thanks to synchronization, TD-SCDMA does not need soft handover, which leads
to a better cell coverage, reduced inter-cell interference and low infrastructure and
operating costs.
• Smart Antennas are beam steering antennas which track mobile usage through
the cell and distribute the power only to cell areas with mobile subscribers. Without
them, power would be distributed over the whole cell. Smart antennas reduce multi-
user interference; increase system capacity by minimizing intra-cell interference,
increase reception sensitivity and lower transmission power while increasing cell
range.
1.2 TD-SCDMA System Simulator

The TD-SCDMA system simulator is a complete home-made test-bed reproducing the
main characteristics of a UMTS network, compliant to the standard UTRA TDD 1.28
Mcps, from the physical to the application layer.
Due to the complexity and the vastness of the behaviors to be reproduced for such
wireless system, the overall problem has been subdivided in smaller logical blocks (see
Figure 1.3: TD-SCDMA System Simulator: block diagram
figure 1.3) which can be developed in different software programs, thanks to appropriate
communication interfaces.
This planning choice has many advantages: first of all, during the initial phase of
software implementation, the effort of the developer is focused on smaller procedures, thus
allowing a tidier work. Secondly, some parts can be easily re-used for system simulators
of other wireless technologies (for example, the mobility simulator doesn’t strictly depend
on the radio access technology). Finally, this choice permits to easily project each logical
block with a different time scale (for example, the link level simulator works with the
bit, whereas the user activity simulator setups voice calls with a much larger time step),
provided that there is a clear interface definition between the various blocks.
In the next subsections, a description of each logical block in figure 1.3 is given.
1.2.1 Link Level simulator

The block diagram of the implemented TD-SCDMA Link Level simulator is presented in
figure 1.4.
Figure 1.4: TD-SCDMA Link Level simulator: block diagram
• The block ”S” is the source of the information to be transmitted.

Every timeslot, the information bits are randomly generated. Each active user may
transmit with different bit rates, depending on variable Spreading Factor codes (1,
2, 4, 8, 16) in uplink and depending on multi-coding till a maximum of 16 parallel
codes with fixed spreading factor equal 16 in downlink.
• The channel coding block performs the coding operation through convolutional codes
with variable code rate and constraint length [49], depending on the settings of the
configuration file (in the next future, turbo-codes will be implemented).
• The interleaving block [49] operates at different periods: 10, 20, 40 or 80 ms.
Let us recall that the TD-SCDMA radio frame duration is 10 ms, that is the same
duration of UTRA TDD 3.84 Mcps and UTRA FDD; nevertheless, the TD-SCDMA
peculiarity is that the radio frame is split in two sub-frames of duration 5 ms. In this
way the interleaving procedure allocates the bits by spanning the various timeslots
over at least two sub-frames.
• The QPSK modulator generates the QPSK symbols.

The number of symbols per burst depends on the type of used burst [48] and on the
selected spreading factor.
• The spreading procedure performs the ”Direct Sequence Spread Spectrum” tech-
nique by multiplying the input signal with an Orthogonal Variable Spreading Factor
(OVSF) code [50].
• The propagation channel block generates the channel impulse response for AWGN
channels or non-AWGN channels with frequency-selective fading and intersymbol
interference (ISI). The latter uses Jake’s doppler channel model for pedestrian and
vehicular channels, or a FIR filter for indoor channels.
• Samples of AWGN noise are then summed to the signal previously generated with
the channel impulse response.
• Several receivers have been implemented: the Matched Filter (MF), the Zero Forcing
(ZF) and the Minimum Mean Square Error (MMSE) one. These receivers perform
the operation of despreading.
• The QPSK demodulator carries out the de-mapping of the QPSK symbols, that is
the association between each symbol of the constellation with the related couple of
encoded bits.
• The de-interleaving block executes the dual operation of the interleaver, returning
the ordered sequence of encoded bits that were previously mixed in various timeslots
and subframes.
• The channel decoding block performs the decoding through Viterbi algorithm. The
decoding can be ”hard” or ”soft” (that is, by addition of reliability information of
the decoding).
• The last block records the result of each transmission and produce the desired
output: as explained in chapter 3, different methods shall be used to generate the
appropriate information for the rest of the system simulator.
1.2.2 Link-to-Network level Interface module

Unlike systems such as pure GSM (i.e. with no GPRS services) in which link and system
level issues can be investigated separately, in case of third generation cellular systems,
the interaction between link and system level is much more involved and a large number
of link level parameters have to be considered. Such parameters are: the class of service
(speech or data CS, data PS), type of ITU propagation channel, direction of the link
(DL/UL), number of timeslots and codes occupied in every subframe, spreading factor
(SF), code rate, interleaving period, type of receiver.
Figure 1.5: Link-to-Network level Interface module: block diagram
As already explained at the beginning of this chapter, a thorough performance inves-

tigation of the TD-SCDMA system requires the analysis of both link and network level
aspects. Owing to the complexity of this approach, a unique simulation program address-
ing both these aspects at the same time is expected to requires an inconceivable amount
of time to obtain some results.
This suggests to develop two different simulators: one for link level and another one
for network level issues (see figure 1.3). While the minimum time step of the TD-SCDMA
link level tool described in subsection 1.2.1 is equal to the chip duration, at network level
the selected time resolution has a much longer duration, equal to the physical time slot.
Therefore, in the design of the TD-SCDMA system simulator, the project of a suitable
interface integrating link and network level parts and a careful definition of the parameters
exchanged through the interface is required.
As explained in more detail in chapter 3, in this thesis an advanced link-to-network
interface module working with ”instantaneous values” has been developed; in figure 1.5,
a simple diagram of the main interactions between the link and the network level through
the interface module.
• First of all, the interface module receives from the link level simulator the BER/BLER
look-up tables for each allowed parameter combination and the vector of fading sam-
ples for each propagation channel.
• Afterwards, during the dynamic simulation, the interface module communicates

real-time with the Network Level Simulator. The exchange time is TS (Time Slot)
based for fast fading enforcement or TB (Transport Block) based for BER/BLER
evaluation after decoding.
The transport block represents the elementary data unit managed by encoder/decoder
blocks and we assume the TB duration equal to the interleaving period.
Since a large part of the information needed for certain calculations is already included
in the network level simulator tool, the interface module could be directly implemented
in the same software program, although logically separated (see the dotted rectangular
containing both the network level simulator and the interface module in figure 1.3): this
choice has been adopted in our project and allows to speed up the overall simulation time.
1.2.3 Network Level simulator

Different simulation approaches for UMTS network performance evaluation can be char-
acterized.
The first one is the static approach: in this case, the capacity of a given UMTS
network layout can be estimated based on the propagation conditions in a very fast way.
This is very useful for a first rough network planning with a few iterations necessary to
find a proper network layout that fulfills the needs of the operator.
Starting with a fixed downlink load the coverage area is evaluated as well as the serving
area for each cell. For the obtained coverage areas the cell capacity is evaluated using
a distributed UE method, which means that for each prediction pixel a fractional UE
is considered. The fractional coefficient is determined by assigning the total available
downlink power resources to all pixels according to their radio channel conditions. The
resulting cell capacities are valid for the given downlink load and the procedure can be
repeated for different loads to evaluate the dependencies between coverage and capacity.
The traffic density can be defined either homogeneous or location dependent. During
the static capacity prediction the distributed UE per pixel (fractional UE) are weighted
according to the traffic distribution. This enables for example that areas with poor radio
channel conditions having a low expected traffic density do not decrease the cell capacity
very much, despite these areas require considerable resources to be connected.
A second common possibility for TD-SCDMA network simulation is based on the well
known Monte Carlo approach. Compared to the static approach, the Monte Carlo
simulation tool provides more detailed outputs and gives a better understanding of the
network behavior. Compared to the full dynamic simulations, the Monte Carlo approach
is much faster, thus enabling the simulation of much larger areas.
The Monte Carlo (MC) method consists in repeating an experience many times with
different randomly determined data in order to draw statistical conclusions; in mobile
network case, the users are deployed in the network with random positions. The results
given by a high number of tests are considered to be representative for all the possible
states of the network.
The third main method for UMTS network study is the dynamic simulation: this
is the approach that has been chosen in this work. With this methodology, time variant
effects like the fast power control, the Radio Resource Management algorithms and the
UE mobility are taken into account and their impact on the overall system performance
Figure 1.6: Main functional blocks of the TD-SCDMA simulator
can be studied in a very detailed manner.

For example, in this approach, a new radio connection is reproduced according to
arrival and ending statistics. If there are enough radio resources available to serve the new
mobile station, the initial transmission power is determined and the closed loop fast power
control process is started. Considering all transmission links from all mobiles in uplink and
downlink directions with their corresponding transmission powers, the interference levels
can be determined for each mobile in uplink and downlink, respectively. The resulting
signal-to-noise and interference ratio levels can be compared with the target levels specified
within the service definition, and the required power control command is sent back to
update the transmission power in the next active slot.
The main characteristics of the TD-SCDMA dynamic network level simulator devel-
oped in this study are summarized in the following (see also figure 1.6):
• The simulation step is equal to the time slot duration (675 µs): all the power
measurements refers to this time resolution. Other system operations occur with
larger time intervals: frame (10 ms), transport block duration (10, 20, 40 or 80 ms)
or longer periods for the averaging of the measurements (0.5-1.0 s).
• Channel propagation model: Pathloss, Walfish Ikegami model: K1 + K2 · log10 (d)

with arbitrary values of K1 and K2 ; shadowing, modelled by means of log normal
random variables with zero mean and exponential correlation function. Fast Fading:
different ITU channels are simulated (pedestrian, vehicular, indoor, both A and B)
[54].
• Power Control: both fast closed loop power control (rate 200 Hz) and slow outer
loop power control are implemented. Each service is characterized by a proper
value of the (Eb /I0 )target . Different TPC(Transmit Power Control) step sizes can be
selected.
• Antennas: either omnidirectional or sectorized antennas with arbitrary (2D) an-

tenna pattern.
• Links: both uplink and downlink BER/BLER are considered to estimate the quality
of radio links. This information is obtained through communication with the link-
to-network level interface module (see subsection 1.2.2).
• At datalink layer (level 2), Automatic Repeat reQuest (ARQ) is implemented for
packet switched services [19].
• Radio Resource Management (RRM) - Call Admission Control: a new user is ac-
cepted provided that codes are available, the estimated interference is less than a
given threshold and the power required in downlink is available. Other specifications
might be the maximum number of RUs occupied per timeslot by all CS users and
allocation strategies that subdivide various class of service users (i.e. speech and
data users) on different timeslots.
• RRM - Dynamic channel allocation (DCA): based on code-timeslot-frequency alloca-

tion. Three carriers with a separation of ∆f =1.6 MHz are considered. A scheduler,
which combines throughput maximization and fairness guaranteeing, is proposed
for multi-service TD-SCDMA system.
• RRM - Handover Control: in TD-SCDMA, only hard-handover is defined. In the

simulator, the event 1G for Change of Best Cell in intra-frequency scenario has
been implemented [51]; the condition is that ”a P-CCPCH RSCP becomes better
than the previous best P-CCPCH RSCP”. In similar way, the event 2A applies for
inter-frequency scenario.
• RRM - intersystem Handover: due to the Common Radio Resource Management

(see chapter 5) for heterogeneous networks, calls can be moved from/to TD-SCDMA
network from/to an alternative radio access technology.
• Both conversational (speech and video) and interactive/background services are

simulated. Every class of traffic has the following characteristics: number of RUs
(number of timeslots, number of codes used in downlink and uplink), number of

information bits carried in a transport block.
The mobility characteristics and the type of user activity (i.e. voice call arrival
rate, packet distribution statistics, etc.) are defined offline by other simulators,
respectively the Mobility simulator and the User Activity simulator; these data are
communicated to the network simulator by the Upper Layer simulator.
The TD-SCDMA dynamic network simulator provides an extensive set of radio perfor-
mance measures (on the contrary, the overall system performance in terms of call block,
call drop, system throughput, etc. is evaluated by the Upper Layer simulator):
• System efficiency: in TD-SCDMA network, it’s essential to exploit the oscillating

switching point between downlink and uplink timeslots in order to fulfill asymmetric
traffic request.
• Coverage, capacity and handover maps.
• Radio statistics: the TD-SCDMA network simulator provides statistics of transmit

power (both in the NodeB and in the UE), received power, interference, signal-to-
interference ratio, etc.
The union of the TD-SCDMA network level simulator with the link-to-network level
interface module constitutes a Lower Layer Simulator (see figure 1.3), according to the
definitions and the terminology introduced in our home-made platform, SHINE, Simula-
tion platform for Heterogeneous Interworking Networks, see appendix A for more details.
In practice, all aspects related to the access technology adopted, hence related to the phys-
ical and data-link layers as well as the local Radio Resource Management, are managed
by the lower layer simulator.
1.2.4 Upper Layer simulator

The Upper Layer simulator manages the end-to-end aspects of each connection, no matter
the supported access technology at the physical and data-link levels. This means the upper
layer simulator design is independent from the specific wireless network implemented
in the lower layer simulator, therefore the same upper layer simulator program can be
connected to different lower layer simulators (i.e. TD-SCDMA, W-CDMA, WLAN, etc.).
In SHINE (see appendix A), a further step has been done: a unique upper layer
simulator may be connected to several lower layer simulators at the same time, in order
to reproduce the conditions of a wireless heterogeneous network. Actually, the end-to-end
connection control is terminated at the upper layer simulator side, whereas peculiar radio
access aspects are reproduced by each different lower layer simulator.
The tasks of the upper layer simulator are mainly concerned with user activity man-
agement and its corresponding mobility in the simulated scenario. For these purposes, the
upper layer simulator receives as input the results of a proper User Activity simulator
and a Mobility simulator (see figure 1.3); aiming at the reproducibility of the system
simulation results, the mobility and the user activity simulators provide their results to
the upper layer simulator in offline mode.
The user activity simulator, according to customized call arrival statistics and IP
traffic models, generates the profile of the service usage of each user: the instant of call
initiation and termination for both voice and data users are defined as well as the instant
of generation and the dimension of each IP packet based on FTP download and web
browsing application characteristics.
On the other hand, the mobility simulator reproduces, based on the specific character-
istics of each traffic class (for example, vehicular voice users as well as static email service
users), the movements of each mobile station within a realistic scenario. The movements
can be completely randomly generated or forced to follow certain rules (for example, users
moving along the streets).
The main tasks of the upper layer simulator are therefore defined:
• communication to the TD-SCDMA lower layer simulator of the starting instant

of each new traffic session according to the user activity simulator as well as the
time-varying user position within the investigated scenario according to the mobility
simulator;
• generation of the bit-flows up(down)loaded by users in each session according to

the user activity simulator: each packet transmission shall be simulated by the
lower layer simulator, but the transport protocol is transparent to the lower layer
simulator.
The upper layer simulator implements the most important transport level protocols
(TCP, UDP, etc.);
• execution of all Common Radio Resource Management (CRRM) functions (see chap-
ter 5): the upper layer simulator selects through which technology should each user
be connected on the basis of customized rules and the available networks’ infor-
mation; it can also decide to move a connection from a lower layer simulator to
another (that is, from a given technology to another) at any time, thus simulating
the interworking;
• collection of all simulation results in order to provide application level performance,

that is from an end-to-end point of view (i.e. call setup success rate, call drop rate,
user throughput, etc.).
1.3 How to convert the TD-SCDMA system simula-

tor to the UTRA FDD option
In the previous section 1.2, the characteristics of the TD-SCDMA system simulator de-
veloped during this work are illustrated. In this section, an overview of the required
changes to convert this environment to the UTRA FDD system (called W-CDMA in the
terminology of ITU) is given.
Let us recall that in Europe, the 3G standard has been initially developed by ETSI
(European Telecommunication Standard Institute) under the designation of UMTS (Uni-
versal Mobile Telecommunications System). The radio access interface of the UMTS
(UTRA) comprises two standards for operation in the FDD and TDD modes. Both in-
terfaces have been accepted by ITU and are designated IMT-DS (Direct Spread) and
IMT-TD (Time Division) respectively.
The UMTS standard is being currently defined by Third Generation Partnership
Project (3GPP): a joint venture of industry organizations and of several Standards De-
veloping Organizations from Europe (ETSI), US (T1), Japan (ARIB), Korea (TTA), and
China (CWTS).
According to the second 3GPP release (called Release 4) the UMTS terrestrial radio
access standard includes the following modes:
• UTRA F DD (W-CDMA)
• UTRA T DDHCR (3.84 Mcps, 5 MHz bandwidth, TD-CDMA air interface)
• UTRA T DDLCR (1.28 Mcps, 1.6 MHz bandwidth, TD-SCDMA air interface)
where HCR stands for High Chip Rate and LCR stands for Low Chip Rate (in the first
release, called Release’99, TD-SCDMA wasn’t yet defined).
These systems share the same Core Network and a common set of features within
the UTRAN (UMTS Terrestrial Radio Access Network). On the other side, the Radio
Access Technology (RAT) (collecting with this term the radio frame format, channel
coding procedures, definition of transport channels and so on) is the distinguishing aspect
among them.
FDD TDD
Air Interface W-CDMA TD-CDMA TD-SCDMA
Bandwidth 2 * 5 MHz paired 1 * 5 MHz unpaired 1 * 1.6 MHz unpaired
Frequency re-use 1 1 1 or 3
Handover soft, softer (interfreq:hard) hard hard
Receiver Rake Joint Detection/Rake Joint Detection/Rake
Chip rate 3.84 Mcps 3.84 Mcps 1.28 Mcps
Spreading factor 4-256 1,2,4,8,16 1,2,4,8,16
fast: every 667 µs slow: 100 cycles/s slow: 200 cycles/s
Power control
closed loop UL:open, DL:closed loop closed loop
Frame duration 10 ms 10 ms 5ms (2 subframes in 10ms)
Timeslot duration 0.667 µs 0.667 µs 0.675 µs
Time slot/frame 15 15 7
Table 1.1: 3GPP modes: FDD, TDD 3.84 Mcps, TDD 1.28 Mcps
In this thesis, the main studies have been focused on TD-SCDMA because of its
compelling characteristics, one above all the possibility to manage asymmetric traffic
thanks to its TDD nature. Nevertheless, here a comparison with W-CDMA is given and
the main changes are described.
In table 1.1, a summary of the main differences between TD-SCDMA and W-CDMA
(for the sake of completeness, also the characteristics of TD-CDMA, that is the TDD high
chip rate option, are included).
It’s evident that the main differences between TD-SCDMA and W-CDMA are in the
physical signal (i.e. bandwidth, frame structure, etc.) and in the physical procedures
(power control, receiver characteristics, handover type, etc), but also Radio Resource
Management procedures like Dynamic Channel Allocation have a different (reduced) im-
pact in the W-CDMA simulator.
The simulators (see figure 1.3) that are impacted by the implementation of W-CDMA
are the TD-SCDMA link level simulator and the TD-SCDMA network level simulator.
On the contrary, the link-to-network level interface module is not impacted, since the
”instant values” approach (see chapter 3) has a quite general validity. Also the current
upper layer simulator doesn’t need any change to communicate with a W-CDMA network
simulator.
The main modifications in the existing TD-SCDMA link and network simulators are
here described.
W-CDMA is not TDMA

First of all, the Time Division Multiple Access characteristic of the TD-SCDMA system
implies that in the time domain different connections can be multiplied (for example, two
users can use the same spreading factor 16 code, but in different timeslots). This access
technique is not present in the W-CDMA mode, therefore a W-CDMA simulator can
simplify the structure of the so called Resource Units (RU), by eliminating one dimension:
RUT D−SCDM A [code, timeslot] −→ RUW −CDM A [code]

Both the W-CDMA link and the W-CDMA network level simulators, shall assign a
code for each connection, and this code is used continuously by the same user, until the
call release (or a RAB modification, still not implemented in the current TD-SCDMA
network simulator).
Note that in a further step of the development of the W-CDMA system simulator to
comply to 3GPP Rel5 specifications, the introduction of HSDPA (High Speed Downlink
Packet Access) channels will require a structure in [code, subf rame], since in every HSDPA
sub-frame (2 ms), the HS-DSCH channels (1 to 15, with fixed spreading factor 16) can
be flexibly assigned by the MAC-hs scheduler to different Rel5 users.
W-CDMA is FDD
Since W-CDMA uses paired bands, the resources are doubled for downlink and uplink
usage. This mainly impacts the current TD-SCDMA network simulator, because the TD-
SCDMA link level simulator is already split in two part, one for downlink simulation, the
other one for uplink simulation.
Frame organization
Given the differences of timeslot and frame duration specified in table 1.1, the major
impacts in the W-CDMA simulators are that it is not anymore requires to cycle each
frame two times to reproduce the behavior of the two sub-frames of 5 ms, typical of
TD-SCDMA.
On the other hand, the FDD characteristic involves a doubled effort for the calculation
of the transmitted powers and the received signal-to-interference plus noise ratio (SINR)
of each connection, because these calculations shall be repeated each timeslot both in
downlink and in uplink (on the contrary, in the TD-SCDMA system, each timeslot is
exclusively used in downlink or in uplink).
Power Control
Both the TD-SCDMA and the W-CDMA modes define a closed loop power control.
However, the mechanism of W-CDMA is slightly simpler to be implemented, because
in each timeslot, both in downlink and in uplink, the measured SINR is immediately
compared to the SIN Rtarget and a consequent Transmit Power Control (TPC) command
is sent to the transmitting entity (UE or NodeB) in order to control its transmitted power
level and therefore satisfy the quality criterion.
On the contrary, in TD-SCDMA, the receiving entity measures the SINR in each
active timeslot (i.e. the timeslots in which there are data for that user), and waits the
next switching point (i.e. the point in which the direction of transmission change from/to
downlink/uplink) to evaluate the average value of the accumulated SINR measurements
and finally send the related TPC command.
Receiver
Because of the higher number of active codes at the same time in W-CDMA compared
to TD-SCDMA, the joint detection is not feasible in W-CDMA. The reason of this large
difference in the number of codes is that W-CDMA user codes at higher spreading factor:
for example, for a speech AMR 12.2 Kbps connection, in W-CDMA a code with spreading
factor equal 128 in downlink is used; on the contrary, TD-SCDMA for the AMR call uses
2 parallel codes at spreading factor 16 in one timeslot every 5 ms sub-frame.
Handover
Many differences distinguish W-CDMA and TD-SCDMA regarding the intrafrequency
handover procedure (here in this work, the interfrequency handover is neglected): since
TD-SCDMA implements the hard handover only, whereas W-CDMA uses soft and softer
handover, different measurements shall be defined in the W-CDMA network simulator.
Currently, in the TD-SCDMA network simulator, the event 1G for Change of Best
Cell (TDD) has been implemented [51]; the condition is that a P-CCPCH RSCP becomes
better than the previous best P-CCPCH RSCP (see figure 1.7); this measurement shall
trigger the replacement of the previously best cell with the current evaluated cell.
On the contrary the W-CDMA simulator shall implement at least the following three
different events (see figure 1.8):
• event 1A: a Primary CPICH enters the reporting range; this measurement shall
trigger a cell addition to the current cell active set.
• event 1B: a primary CPICH leaves the reporting range; this measurement shall
trigger a cell deletion form the current cell active set.
Figure 1.7: event 1G for TD-SCDMA: a P-CCPCH RSCP becomes better than the pre-
vious best P-CCPCH RSCP
• event 1C: a primary CPICH that is not included in the active set becomes better
than a primary CPICH that is in the active set; this measurement shall trigger a
cell replacement within the current active set.
Apart from the different measurement triggers, two new issues shall be solved by the
W-CDMA network simulator in order to correctly implement the FDD soft/softer han-
dover: first of all, the data structure of each radio connection in the W-CDMA simulator
shall implement the possibility to manage more than one radio link at the same time
(typically a maximum of 3 radio links in active set). Whereas the uplink transmit power
in the mobile is obviously the same for all radio links, in downlink each cell shall transmit
with a proper value defined by the closed loop power control. In order to avoid drift effects
among the different active NodeBs, the implementation of a Downlink Power Balancing
algorithm should be considered.
Secondly, in the TD-SCDMA network simulator, since the P-CCPCH is transmitted
only on the timeslot 0 of each subframe without other dedicated physical channels, the
only appropriate quantity used for handover measurement is the Received Signal Code
Power (RSCP) from the P-CCPCH. On the other hand, in the W-CDMA mode, the
reference for handover measurements is the Common Pilot Channel (CPICH), which is
transmitted in all timeslots, and multiplied with all the other common and dedicated
Figure 1.8: W-CDMA basic handover algorithm
physical channels, therefore for handover control it’s appropriate to measure also the
downlink Ec /N0 , that is the received energy per chip from CPICH divided by the power
density in the band.
Chapter 2
Performance of TD-SCDMA in
mixed CS/PS traffic scenarios
In this chapter, the performance of a time-division synchronous code-division multiple ac-

cess (TD-SCDMA) system is analyzed. In particular, we investigate the impact of packet
switched applications (for instance web browsing sessions) on the overall performance of
the network. Here, we quantify the degradation of the voice users quality in the presence
of packet data services and viceversa. Finally, we investigate the impact of some physical
layer parameters for the power control algorithm (PCA) on the overall quality of service
and we show that these parameters should be carefully chosen in order to balance the
quality of voice and data users. The analysis of the influence of physical layer on network
performance is carried out through the method based on an ”instant value” interface
depicted in chapter 3.
The numerical results have been obtained through the system simulator developed
during this research which takes both link and network level issues into account; to the
author’s knowledge only few papers in literature provide a thorough investigation of TD-
SCDMA networks since the majority of them deals with either link or network level
analysis, separately (see for instance [15, 16]).
31
32 Performance of TD-SCDMA in mixed CS/PS traffic scenarios
The main characteristics of the used TD-SCDMA system simulator are described in
chapter 1.
In the next sections 2.1 and 2.2, we present in more detail two of the main algorithms
directly impacting the performance of TD-SCDMA in mixed CS/PS traffic scenarios, re-
spectively the packet scheduler and the power control. Afterwards in section 2.3, the
simulated scenario is described and in section 2.4 the merit figures for performance evalu-
ation are defined; finally, in section 2.5 the results obtained with the TD-SCDMA system
simulator are discussed.
2.1 Packet scheduler

The main goal of a scheduling algorithm should be the achievement of as many satisfied
data packet users as possible at a given system load. It turns out therefore necessary to
define a criterion based on which a user is considered satisfied; in fact, various criteria of
evaluation carry to different choices for the allocation algorithms [19].
For a minimum delay criterium a first-in-first-out (FIFO) scheduling strategy would
be a good choice: the data of all users are appended to a single queue in the order they
are requested and they are read out on a FIFO basis. An other choice is a throughput
based criterium: a user is defined to be satisfied if the average data rate during the
entire connection, the active session throughput (AST), does not remain under a certain
threshold.
In this work, we adopted the second criterium: in particular, ETSI [18] specifies that
the AST threshold value should be 10% of the nominal bit rate, BitRX . AST represents
a performance figure reported for every single user, and is defined as the ratio between
the total number of bits at level of application received during the whole session (block
retransmissions and overhead due to lower protocol layers are not considered) and the
duration of the session, Tsession , excluded the time intervals in which no information from
upper layers request to be transmitted, Tinact (that is the transmission queues do not
contain packets from that user).
BitRX
AST = ≥ 10% · Brnom (2.1)
Tsession − Tinact
In this work we decided that, for Non-Real Time (NRT) packet data services, uplink
and downlink shared channels (respectively, USCH and DSCH [48]) can be used to allow
efficient allocations for a short period of time: these transport channels exist only in
TDD mode. The MAC-c/sh packet scheduler, localized in the Radio Network Controller
Performance of TD-SCDMA in mixed CS/PS traffic scenarios 33
(RNC)1 , redistributes dynamically the RUs between all active users, that is the users with
data in the buffers, waiting to be transmitted.
We considered three Quality of Service packet bearers, with different requests of max-
imum rate, at 64 Kbps, 144 Kbps and 384 Kbps. In downlink, for every cell, three
buffers are used (one for every class of service): they hold the packets addressed to mobile
terminals connected to that cell.
The packet scheduler algorithm proposed here considers two level of priority to estab-
lish a data packet transmission. First of all, transmission requests are ordered based on
the service class to which they belong; level of priority, from highest to lowest, is following:
384 kbps, 144 kbps, 64 kbps. Second level of priority orders the requests of users who
belong to the same class of service: a fair round robin (RR) algorithm is used.
When a data packet transmission request is finally admitted to air interface by the
scheduler, the packet is fragmented into transport blocks: the number of TBs necessary
for the complete transmission of the considered packet is obtained by dividing the number
of bits at level of application contained in the packet by the number of information bits
carried by a TB (every bearer service has its own characteristics, in terms of number of
bits, timeslots and codes per TB [52, 53]). During a packet transmission, if the link-to-
system level interface (see chapter 3) evaluates that an error has been occurred in the
reception of a transport block, only that TB is retransmitted.
2.2 Power control

Power control is the mechanism which takes care of maintaining the received signal-to-
interference ratio (or the received power level) at a constant value. In modern cellular
network this is commonly carried out in both uplink and downlink. In our TD-SCDMA
system simulator three different kinds of the power control are implemented. [21]:
• Open-loop power control: owing to the correlation among the average path loss
of downlink and uplink, the user equipment (UE) can estimate the initial power
needed in uplink and downlink based on the path loss calculations in the downlink
direction. This mechanism is commonly used to set the initial value of received
value of signal-to-interference ratio.
• Closed-loop power control: this PCA uses the feedback information from the opposite
end of the radio link. This allows the considered terminal (UE in uplink and base
1
On the other hand, the MAC-hs scheduler using the latest Rel5 HSDPA transport channels is located
in the NodeB. This function is not yet implemented in our system simulator.
station in downlink) to adjust the value of the transmitted power. Let us consider
for instance the uplink power control, user equipment transmits using a given value
of power, the base station measures the value of received power and compared it
with its target value of signal-to-interference ratio, if this value is smaller than the
threshold, the base feeds back, through the TPC control bits, this information and
the UE increases its value of transmitted power. As few bits are carried by TPC,
the step between the old and the new value of transmitted power is quantized (PC
step).
If the speed of the power update is sufficiently high, terminal (or base) can compen-
sate for the fast fading contributions. In W-CDMA network, as terminal transmits
in each timeslot, the value of transmitted power is changed with a frequency of 1500
Hz (15 slot/10 ms). In case of TD-SCDMA, owing to time division multiple access
nature, transmitted power is updated one time each sub-frame, so the frequency is
200 Hz.
• Outer loop power control: The aim of outer-loop power control is that of maintaining
the quality of the communication at the level defined by the quality requirements of
the bearer service. This is carried out by changing the target value of the received
Eb /I0 ; this is necessary when propagation conditions change, i.e. a change in the
mobile speed, and therefore the previous value of received Eb /I0 does not guarantee
an acceptable value of bit error rate.
2.3 Scenario and propagation environment

As far as channel model is considered, in this chapter we assume that the received power
(in uplink and downlink) is calculated according to the following expression:
Pr [dBW ] = Pt [dBW ] − K1 − K2 · log10 (d) + S[dB] + F [dB] (2.2)
where Pt represents the power transmitted by the mobile user (in uplink) or the base
station (downlink), pathloss is modelled by means of the coefficients K1 and K2 (expressed
in dB), and d is the distance between transmitter and receiver. Several other pathloss
models can be considered: Walfish Ikegami, taking both line-of-sight (LOS) and non-LOS
conditions into account; K1 + K2 · log10 (d) with arbitrary values of K1 and arbitrary
coverage maps.
S stands for log-normal Shadowing, modelled by means of Gaussian random variable
(expressed in dB) with zero mean and exponential correlation function. Finally fast
Figure 2.1: The considered network scenario
fading is modelled using the time-correlated random variable F : different ITU channels
are simulated (pedestrian,vehicular,indoor, both A and B).
The simulated scenario considered in this chapter is depicted in figure 2.1. Here a
regular lay-out with 18 sites is designed. However, the simulation tool is quite general
and allows the evaluation of a larger number of sites located in arbitrary positions.
As far as antenna pattern is considered, a two-dimensional antenna pattern is consid-

ered. Both omnidirectional and sectorized antennas can be considered by the simulation
tool. Here, we assume that each site uses three sectorized antennas. The total number of
sectors considered in this chapter is 54 (see figure 2.1).
The total bandwidth assigned to each site is 5 MHz; we have assumed that every
sector uses a different 1.6 MHz carrier (1/3 frequency reuse).
Main simulation parameters are summarized in table 2.1.

Parameter Default value Description

K1 15.3 dB Pathloss model
K2 37.6 dB Pathloss model
σ 5 dB dB spread of the log normal shadowing process
do 20 m Correlation distance of the log normal shadowing
process
No DL -169 dBm/Hz downlink noise power spectral density
No U L -174 dBm/Hz uplink noise power spectral density
α 0.2 downlink orthogonality factor
β 50% MUD efficiency
BERout 10−2 Threshold on averaged BER for outage
BERdrop 2 · 10−2 Threshold on averaged BER for the evaluation of the
dropout
Table 2.1: System parameters fixed in the numerical results
Traffic characteristics
Both voice and data bearer services are considered in this chapter in order to achieve the
main target of this study that is to quantify the degradation of the voice users quality in
the presence of packet data services and viceversa.
Every class of traffic is characterized by the following parameters: the number of Resource
units, classified in number of time slots required, number of codes used in both downlink
and uplink, the number of information bits carried by the transport block [52][53].
The main characteristics of traffic generated in the simulations of this chapter are
listed as follows for every class of service:
• Voice services (CS): call arrival and departure processes are Poisson distributed;
average call duration is 120 s.
TCS denotes the voice offered traffic per site (in Erlang).
• Data users (PS): the data packet application we have simulated is Web browsing.
Three layered stochastic processes are considered: the session arrival process, the
packet call arrival process and the packet arrival process. Session arrival process is
Poisson distributed. Packet size is Pareto distributed; all the parameters have been
fixed according to [18].
Mean traffic parameters are specified in table 2.2, where Npc is the average number
of packet calls per session, RTpc is the average reading time between consecutive
packet calls, Np is the average number of packets per packet call, IAT is the av-
erage packet inter-arrival time. Number of packet calls, reading time, number of
QoS service Npc RTpc Np IAT

64 Kbps DL 5 120 s 25 60 ms
144 Kbps DL 5 120 s 25 20 ms
384 Kbps DL 5 120 s 25 10 ms
64 Kbps UL 5 200 s 12 10 ms
144 Kbps UL 5 180 s 12 10 ms
384 Kbps UL 5 180 s 12 10 ms
Table 2.2: Packet switched session parameters for web browsing services
packets per packet call and inter-arrival time are all modelled by means geometric
distributions.
Because of bursty and asymmetric nature of data packet sources, we do not charac-
terize the data traffic by means of the average number of data user in the system.
To measure the data packet traffic per site (using separate metrics for downlink and
uplink) we have defined TP S , which represents the ratio between the total number
of bit at application level received (if we consider the downlink) or transmitted (up-
link) by all users connected to the site and the total time of simulation (i.e. 100
minutes). Block retransmissions and overhead due to lower protocol layer are not
considered.
2.4 Performance Metrics

Several metrics can be evaluated by the simulation tool for circuit-switched applications:
• Blocking rate (Tb ): defined as the ratio between the number of voice connections
blocked by the call admission control algorithm and the total number of voice con-
nections generated.
• Dropping rate (Td ): defined as the ratio between the number of voice connections
dropped owing to quality problems and the total number of terminated connections.
The simulator decides to drop a call using a ”leaky bucket” algorithm, which com-
pares the averaged measured BER with a given threshold BERdrop . If the BER is
larger than the threshold the call gains two points, otherwise it looses one point.
The call is dropped when the counter reaches a given value.
• Outage rate (Tout ): an event occurs when the averaged value of the BER exceed a
threshold - BERout .
• Satisfaction rate (Tsat ): a user is said to be satisfied if his call is neither blocked
nor dropped, and during the call the outage events have a duration which is smaller
than 10% of the total duration of the call.
In case of packet switched applications, the simulation tool evaluates the following
performance metric:
• Transport block error rate (BLER): defined as the rate of transport blocks erro-
neously detected by the receiver owing to interference.
• Active session throughput (AST ): defined by (2.1).
• Distribution of packet arrival delays: it gives the distribution in terms of probability

density function (pdf) and cumulative density function (cdf) of the delays in the
delivery of the packets.
2.5 Simulation results

The results of three different simulation sessions are here presented. Firstly, in subsec-
tion 2.5.1 we show the results in a basic scenario with only speech users; afterwards, in
subsection 2.5.2, we analyze the performance in a mixed traffic scenario with both voice
and data user, when varying the network load; finally, in subsection 2.5.3, we discuss
how to balance the performance of voice and data users through tuning of physical layer
parameters.
2.5.1 Circuit switched services

In the first simulation session, we have considered three values of offered traffic for voice
users (using a AMR codec at bit rate of 12.2 Kbps): TCS = 21.6, 36 and 43,2 Erlang per
site; we repeated each simulation for different cell radius values ranging from 150 m to
3600 m.
Figure 2.2 shows that for a fixed value of offered traffic, when the cell radius is smaller
than 450 meters, the performance in terms of voice satisfaction rate Tsat remains prac-
tically constant; for radius up to 900 meters it starts to decreases, and it is remarkably
reduced for larger values. The behavior can be explained by remembering that, for large
radius, the system is limited by coverage and the limiting factor is the maximum power
transmitted by the user’s terminals.
On the other hand, fixed the cell radius, due to capacity limitations, the less loaded
configuration (TCS = 21.6 Erlang/site) provides the best performance.
Figure 2.2: Voice satisfaction rate (Tsat ) vs. cell radius
These results are in agreement with those obtained in literature for W-CDMA [20,
21]. The clear dependence of Tsat from the network load could be minimized with the
introduction in the simulation tool of smart antennas which would reduce the levels of
interference (for future implementation in the TD-SCDMA system simulator).
The geographical distribution of unsatisfied users for the three values of network load,
at R = 450m is shown in figure 2.3: the pixels in green represent the areas in which speech
users are satisfied, the pixels in yellow and red represent different ascending levels of users
unsatisfaction, whereas the black squares are located in the positions of the NodeBs. It’s
clear from the figure that, for higher network loads, the users at border cell suffer with
Figure 2.3: Areas of user satisfaction in the scenario

Figure 2.4: Voice satisfaction rate (Tsat ) vs. downlink data packet traffic (TP S ), for
different values of voice offered traffic (TCS )
larger probability outage situations or call drops.
2.5.2 Packet and circuit switched services

In this second simulation session, mixed scenarios with both circuit and packet switched
services are considered: voice at 12.2 Kbps, and data packet services at 64 Kbps, 144
Kbps and 384 Kbps; in each simulation the offered traffic of data users TP S is varied.
For what concerns the performance in terms of active session throughput for packet
switched classes, we focus on downlink; however, for web browsing sessions, our program
simulates also uplink traffic: in this scenario the asymmetry between uplink/downlink
traffic is about 1:3. On the other hand, the metric for voice users (Tsat ) considers both
uplink and downlink.
Figure 2.4 shows the voice satisfaction rate versus the offered downlink data packet
traffic: Tsat decrease with the increasing amount of downlink data traffic due to the
increasing level of interference. In this simulation session, to augment the downlink data
packet traffic, the average arrival frequency of data sessions has been increased; this have
an impact also on the data traffic generated in uplink: the degradation of the quality
of voice users is due to the combined effect of data packet traffic in both downlink and
uplink.
Figure 2.5: Downlink active session throughput (AST ) vs. downlink data packet traffic
(TP S ), for different values of voice offered traffic (TCS )
Figure 2.5 shows the downlink active session throughput for data packet users with 64
Kbps QoS versus the downlink data packet traffic; AST2 , that is the average throughput
per session, decreases firstly with the increasing amount of downlink data traffic because
of higher latency time in buffers and secondly it decreases when a higher number of voice
users is served in the network.
2.5.3 CS/PS performance enhancements

In this final simulation session, we analyze a more specific optimization problem. In fact,
here we investigate the parameterization of the fast power control algorithm (PCA) of
TD-SCDMA (see section 2.2): we evaluate the impact of the target signal-to-interference
and noise ratio (Eb /Io ) of the PCA on the combined performance of voice and data users.
In all the next simulations, we have considered a value of TCS of 18 Erlang/site and a
value of TP S of 161 Kbs/site in the downlink and 71 kbs/site in the uplink.
The default values of the target (Eb /I0 ) for voice and data users, in uplink and down-
link, are shown in table 2.3. Only one target (Eb /I0 ) has been changed in each of the
2
The maximum value of AST for 64 Kbps bearer service should be 64 Kbps, but because of fragmen-
tation of packets from the application level in the transport blocks of fixed dimension, there is a reduction
on maximum AST.
Service (Eb /I0 )DL [dB] (Eb /I0 )U L [dB]

Voce (CS) 5 7
Data (PS) -0.5 -1.5
Table 2.3: Default values for (Eb /I0 )
Figure 2.6: Network Performance as a function of (Eb /Io )U L−CS
following figures, the remaining parameters were fixed to the default values indicated in
table 2.3.
In this session, we’ve disabled the outer-loop power control feature, so the target value
of (Eb /I0 ) is not changed dynamically during the simulation of each connection.
Impact of (Eb /I0 )U L−CS and (Eb /I0 )DL−CS

We start the discussion by considering the impact of the target values of (Eb /I0 ) of voice
users on the overall performance of the network.
Figure 2.6 shows all the main performance metrics considered in this work (see section
2.4). In this first scenario, we have considered AST in uplink and downlink, BLER of
data connection in uplink, Tsat , Td and Tout in the uplink as a function of (Eb /I0 )U L−CS .
Let us consider the curve corresponding to the satisfaction of voice users Tsat : as
expected for small values of (Eb /I0 )U L−CS the quality of service for voice users is very
low, since the associated average BER values are rather high. Let us highlight that this
kind of analysis is allowed thanks to the tight integration between link and network level
simulators (see chapter 3).
If we increase (Eb /I0 )U L−CS , Tsat increases and it reaches a maximum around 5-9 dB;
the use of larger values of (Eb /I0 )U L−CS forces the PCA to increase unnecessarily the UE
transmitted power with a consequent degradation in terms of Tsat . In fact, the distribution
Figure 2.7: Network Performance as a function of (Eb /Io )DL−CS
of the UE transmitted power (which has not been shown here for the sake of conciseness)
shows that many terminals are transmitting the maximum allowed power in order to
achieve the target (Eb /I0 )U L−CS .
Note also that both Tout and Td increases for large values of (Eb /I0 )U L−CS .
Now, if we consider the quality of packet data users in this simulation, we observe
that AST in uplink is almost constant for small values of (Eb /I0 )U L−CS , with a value
around 48 kb/s, and decreases for large values of (Eb /I0 )U L−CS , as BLER in uplink of
data connections increases because of the high level of interference introduced by the voice
connections. As expected AST in downlink is not influenced by (Eb /I0 )U L−CS .
Figure 2.7 shows the same performance metrics of figure 2.6 as a function of (Eb /I0 )DL−CS .
Here we can observe that Tsat does not have the same trend: this depends on the fact that
in the downlink the intercell interference does not play the same role owing to the multi-
user detection, here considered through the parameter β. Large values of (Eb /I0 )DL−CS
do not increase significantly the level of interference in the network. However, it can be
expected that for a higher network load, a too high target for (Eb /I0 )DL−CS will impact
significantly the system capacity. As far as the performance of packet-switched services
is concerned, here the curves corresponding to AST in uplink and downlink are almost
constant.
Comparing the results of figures 2.6 and 2.7, we can observe that the impact of the
target value of (Eb /I0 ) is much more emphasized in the uplink owing to the strongest
contribution of the uplink intercell interference.
Figure 2.8: Network Performance as a function of (Eb /Io )U L−P S
Impact of (Eb /I0 )U L−P S and (Eb /I0 )DL−P S

In this part, we evaluate the performance of the network by varying the target values of
(Eb /I0 )U L−P S and (Eb /I0 )DL−P S .
Figure 2.8 shows the main performance metrics as a function of (Eb /I0 )U L−P S . Small
values of (Eb /I0 )U L−P S provide large values of Tsat ; however, Tsat remains almost con-
stant until a value of (Eb /I0 )U L−P S of 8 dB. This behavior can be explained by looking
at the curve corresponding to Tout in uplink: this value is constant for small values of
(Eb /I0 )U L−P S but becomes significant for (Eb /I0 )U L−P S > 4dB because of the increase of
interference introduced by data users during uplink transmissions.
Concerning the performance of data services, small values of (Eb /I0 )U L−P S cause a
large number of re-transmissions (see the curve corresponding to the uplink BLER) with
a consequent reduction in terms of AST in uplink. Larger values of (Eb /I0 )U L−P S increase
the uplink throughput until a maximum of about 56 kb/s. Again AST in downlink is not
influenced by the variation of (Eb /I0 )U L−P S .
Finally, figure 2.9 shows the main performance metrics as a function of (Eb /I0 )DL−P S .
Here, AST in downlink reaches a value of about 58 kb/s for values of (Eb /I0 )DL−P S larger
than 2 dB; this values corresponds to a BLER of about 0.01. As expected Tsat decreases
when (Eb /I0 )DL−P S is configured to an unnecessary high value; however, it maintains an
acceptable value for (Eb /I0 )DL−P S smaller than 4 dB.
To summarize: in this chapter we have evaluated the performance of a TD-SCDMA

cellular network in the presence of both circuit and packet switched services. The analysis
has been carried out using our home-made simulation environment in which both link and
network level issues are taken into account. Results show that the presence of data services
Figure 2.9: Network Performance as a function of (Eb /Io )DL−P S
can reduce the quality perceived by voice users and viceversa. Moreover, the target values
of (Eb /Io ) in the Power Control Alogorithm play a crucial role in the balancing of the
performance of voice and data services. Owing to the latter considerations, the different
values of target (Eb /Io ) should be carefully chosen: the optimum value can be found as a
trade-off between the obtainable quality of packet and circuit-switched applications.
Chapter 3
Link Level aspects modelling in the

simulation of packet switched
wireless networks
The investigation of both link and network level aspects is fundamental when a thorough
analysis of mobile radio systems is required. The conventional approach, based on a
complete separation between the two levels, adopted in the past for the performance
evaluation of second generation systems, is not enough accurate for third generation ones,
characterized by the presence of both voice and data services as well as circuit and packet
switched applications.
Here we describe the methodology we followed on how to interface link and network
level tools; this approach has been adopted in the TD-SCDMA system simulator developed
during this thesis and whose results are described in the numerical sections of chapters
2 and 5. Nevertheless, the proposed method is quite general and can be adopted for
arbitrary wireless networks with a time slotted physical frame structure and packed-
switched applications.
3.1 Wireless system simulations

Third generation systems such as UMTS [21], that will make available high data rate
transmission are expected to facilitate the diffusion of a large class of new services that
should increase the number of subscribers of wireless applications. Moreover, the presence
of services characterized by different quality requirements, are making the design and the
management of such networks very challenging.
In the past, system performance evaluation was carried out by separating the aspects
related to the radio transmission chain, i.e. coding, modulation, fading mitigation, etc.,
47
48 Link Level modelling in the simulation of packet switched wireless networks
from those related to the multiple access, radio resource management (RRM) and so
on. This division was possible because second generation systems (i.e. GSM [22]) were
originally designed to provide mainly voice services, i.e. at constant bit rate.
With the advent of third generation systems, the focus moves from voice to data
services and from circuit to packed switched networks. This latter aspect requires the use
of performance evaluation tools characterized by the ability to follow all the rapid changes
(power control, packet scheduling,..) of the network. The design of such a platform is
extremely challenging as link level aspects cannot be separated by those of network level
and viceversa [17, 23, 24]. As a matter of fact, since the Call Processing part of UMTS
takes decisions every 10 ms (duration of a radio frame) or 2 ms (duration of a sub-frame
when supporting HSDPA physical channels in FDD mode), and since fast power control
at rate of 1500 Hz shall be implemented and bit rate shall be adapted to actual radio
frequency condition, a different design of network level simulators is required.
In the next sections, we address the problem of how to combine link and network level
analysis in order to realize a thorough simulation tool able to investigate a modern mobile
radio network from a system (”system” is defined in this chapter as the union of link and
network level issues) point of view. Here, we show that the investigation of both link and
network level issues requires the design of a suitable interface integrating link and network
level parts and a careful definition of the parameters exchanged through the interface.
Since our work is focused on the analysis of this interface and that of the quantities to
be taken in account, a performance comparison of different simulation results is out of the
scope of this chapter. However, we demonstrate that, in a typical 2G scenario with speech
service, the proposed method provides the same results of the conventional approach. On
the other hand, our approach becomes the unique adoptable when simulating complex
scenarios with both circuit (CS) and packet switched (PS) bearers and mixed (voice and
data) applications (i.e. refer to to the simulated traffic scenarios in chapter 2).
A similar approach has been proposed in [57] however this proposal is simpler to
implement as it takes into account only the most relevant aspects of the link-level.
3.2 Link and Network Level Analysis

A thorough performance investigation of mobile radio systems requires the analysis of both
link and network level aspects. Owing to the complexity of this approach, an integrated
simulation tool addressing both these aspects is expected to requires hundred of CPU
time to obtain results. This suggests to separate the two aspects, and exchanging data
through a suitable interface [23] -[24].
Link Level modelling in the simulation of packet switched wireless networks 49
Figure 3.1: Link and Network levels in operation
Referring to figure 3.1, at link level, all aspects related to the transmission chain
(modulation, coding, bit interleaving, etc) have to be taken into account in order to
evaluate the performance metrics, typically expressed in terms of bit error rate (BER),
frame error rate (FER) and block error rate (BLER) as a function of a suitable definition
of the carrier to interference ratio.
At network level, the macroscopic aspects of the scenario (user mobility, cellular layout,
shadowing, handover, power control and so on) have to be considered to evaluate the
system performance (i.e. the merit figure Rb of rate of blocked users as a function of
the offered traffic To ). In order to measure the system performance, the network level
simulator exploits the information provided by the link level (BER/FER/BLER) in many
RRM procedures.
Obviously, in such a integrated approach, the interface between the two simulation
tools has to be carefully designed since the time simulation step at link and at network
level are usually different. While the simulation step TL of link level tools is typically
equal to the bit or chip duration, at network level the time resolution TN has usually
a much longer duration, equal to the physical time slot (see section 3.2.2) or to a time
interval in the order of the second (see section 3.2.1).
3.2.1 Network level simulator with a large simulation step

Network level simulators with large value of simulation step (that is a step with a duration
of hundreds of transmission frames) are usually used to investigate scenarios with CS
services only. In such a network, the conventional approach [23] consists in considering
that all the RRM procedures rely on averaged channel measurement reports provided
by the mobile stations and the base stations, over the fast fluctuations due to fading to
provide stable information about the current electromagnetic status of the network.
Assume this averaging is performed over an interval of Ta seconds. For instance in
GSM, Ta could be the duration of a slow associated common channel multiframe (SACCH)
equal to about 0.48 s [22]. So, the network level evaluation could be performed by choosing
this interval as the time resolution (network level simulation step). Therefore, suitable
definition of the measurements performed by mobiles and bases stations at physical level
has to be given, according to the network level simulation step.
All changes in the scenario occurring with a rate larger than 1/Ta (the effect of fast
fading, fast power control and so on) cannot be explicitly modelled, they have to be taken
into account at link level and reported in averaged terms at network level. Then, by
means of this approach the network level simulation tool is a computer program which is
driven by an internal clock equivalent to Ta seconds. All the measurements at this level
have to be considered averaged over a period of time equal to Ta .
At each simulation step, the network level simulator evaluates the carrier to interfer-
ence ratio (i.e. C/I, Eb /I0 or similar) averaged over a Ta period, and takes information
about the quality of radio links by means of look-up tables containing the values of
BER/FER/BLER computed at link level. It should be noticed that owing to the size of
Ta , the fast fading has a mean value equal to zero within this period, and therefore the
BER as a function of Eb /I0 averaged over a period Ta can be considered equivalent to
that of the average over infinitive period of time.
The corresponding link level simulator generates the look-up tables of BER/FER/BLER
following these steps: a given reference value of Eb /Io is selected and the whole trans-
mission chain (modulation, coding, interleaving, etc.) is simulated by considering a huge
number of transmitted bit. At the end of simulation, it evaluates the number of wrong
detected bits and the corresponding BER. Look-up tables will contain on the y-axes the
values of BER and on the x-axes the corresponding reference Eb /Io value obtained by av-
eraging the Eb /Io over the whole simulation time. Note that, by means of this approach,
the presence of fast fading generates, in each (link level) simulation step, instantaneous
values of Eb /Io different from the reference one.
3.2.2 Network level simulator with small value of simulation

step
In the presence of systems where RRM procedures have a decision rate in the order of
several hundreds of Hz (this is the case of WCDMA for 3G networks [21]), the size of
the simulation step of the network level should be considerably reduced, by considering
observation intervals corresponding to one time slot or frame, according to the system
considered [25, 24].
In this case fast fading effects and the sudden changes in the level of radio interference
have to be considered both at network and link level, whereas averaging radio measure-
ments over seconds like in section 3.2.1 would lead to a rough approach. This approach,
which implies a network level simulation step Ta equal to the time slot duration (typically
less than 1 msec), makes network level simulations much more complex, but allows us to
extend the investigation to PS services. With this time resolution, it is also possible to
evaluate the correct/wrong transmission of each transport block and analyze the impacts
of MAC-hs or RLC protocols on the TCP-IP level. Obviously, the whole system analysis
is not allowed with the approach depicted in section 3.2.1.
Since we have to consider the effects of fast fading both at link and network level, in
order to maintain the synchronization between the two simulators, it is necessary that the
link level tool, which generally considers several fast fading contributions according to the
selected channel (pedestrian, vehicular, etc.) generates also a suitable look-up table of
correlated fast fading samples at a rate equal to the network simulator time step. These
values will be used by the network level simulator.
Besides fading samples, the link simulator, that has typically a simulation step equal
to bit or chip durations has to simulate the whole transmission chain to obtain the
BER/BLER curves. Nevertheless, the number of erroneous bits have to be evaluated
at the end of channel decoding phase. This means that, if we consider a mobile radio
system with a time structure organized in frame and slots (i.e. GSM and UMTS), the
evaluations on the link quality have to be computed with a rate equal to 1/TB , where
TB is the duration of a transport block, the elementary data unit managed by the en-
coder/decoder blocks [21], which corresponds to an interleaving length of K frames of
duration Tf rame (TB = K ∗ Tf rame ).
Using this approach, at network level, for each simulation step Ta , the software obtains
the value of ”instantaneous” carrier to interference ratio (Eb /Io ) affected by fast fading
samples generated at link level and subsequently the network simulator calculates the
averaged value over the duration of a transport block (h(Eb /Io )iTB ). This value will be
provided as input to link level which will return the corresponding BER/BLER related
to that particular value of h(Eb /Io )iTB . Note that look-up tables generated following this
approach are completely different from those created using the conventional approach
(see subsection 3.2.1). Furthermore, coding gain effect is still present, even if (Eb /Io ) is
averaged over all the duration of the transport block: the prerequisite is that for every
single transport block, both link and network simulator use the same fading channel,
Figure 3.2: Example of data collection of Nbit−err of the transmitted blocks
power control settings, code rate, etc.

Now, link level tool has to provide values of BER/BLER as a function of h(Eb /Io )iTB .
In fact, in this case the value of h(Eb /Io )i is averaged over a very short (few frames) period
of time. To summarize, the link level tool analyzes the transmission of a huge number
of bits, and considering the interleaving length, for each received transport block Bi , the
algorithm calculates the averaged value (over the duration of the transport block) of the
signal to interference ratio, h(Eb /Io )iTB (Bi ) and the number of wrong detected bits after
the decoding of the transport block (Ne (Bi )).
The overall range of interest of Eb /Io ((Eb /I0 )min , (Eb /I0 )max ) is subdivided in N in-
tervals {(Γ1 , Γ2 ), (Γ2 , Γ3 ), . . . , (ΓN −1 , ΓN )}, where Γ1 = (Eb /I0 )min and ΓN = (Eb /I0 )max .
Each interval is identified by its mean value (Eb /I0 )j = (Γj + Γj+1 )/2. Each value of
Ne (Bi ) is assigned to the interval corresponding to the value of h(Eb /Io )iTB (Bi ). The
values of Ne (Bi ) assigned to the same interval (Γj , Γj+1 ) are then averaged to provide
(Ne )j .
The look-up table contains on the x-axes the values of (Eb /I0 )j (in the following j will
be omitted and the notation < (Eb /Io ) >TB will be used to emphasize that the average
is performed over the transport block) and in the y-axes (Ne )j , that is the expected BER
value when < (Eb /Io ) >TB is measured in the interval (Γj , Γj+1 ) . We behave in similar
way when obtaining the BLER look-up table: in this case the number of wrongly decoded
transport blocks is counted, where a TB is considered to be wrong when at least a wrong
decoded bit is present.
The example in figure 3.2 summarizes the main algorithm steps. In this case, two
different values of reference h(Eb /Io )i (that is averaged over an infinite period of time)
have been used to transmit a large amount of Transport Blocks; let us recall that TB =
K ∗ Tf rame and in each frame a certain number of timeslots will be required for the
transmission of each transport block.
The effect of fast fading becomes evident by observing that the values of Ne (Bi ) of
each transmitted block are associated to different intervals although related to the same
reference value h(Eb /Io )i. At the end of the transmission of all the different reference
h(Eb /Io )i, in each interval (Eb /I0 )j , the average value of the contained Ne (Bi ) is calculated
to derive the value of BERj . Similarly, to obtain BLERj , the average value of wrongly
decoded transport blocks in (Eb /I0 )j is calculated.
3.3 An Example on how to interface Link and Net-

work Levels
In this section, we present an example of interface between link and network level with
reference to the UTRA TDD 1.28 Mcps system, also well-known as TD-SCDMA, which
is characterized by high spectrum efficiency and by the capability of managing asymmet-
ric traffic conditions [14, 15, 16]. This design has been implemented in the TD-SCDMA
System Simulator developed during this thesis. However, the methodology previously de-
scribed can be used to investigate mobile radio systems based also on different technologies
(i.e. UTRA FDD).
The system is reproduced by means of a time discrete model with a simulation step
equal to the time slot (675 µs) of the TD-SCDMA frame. The approach described in
subsection 3.2.2 has been used.
The simulation environment we have created is composed of three main blocks: a
Link Level tool, a Network Level simulator and an Interface module logically positioned
between the two tools.
3.3.1 Link Level to Interface Module communication
Figure 3.3: Link tool-to-interface module communication: block diagram
As previously described, in case of third generation cellular systems, the interaction

between link and network level is much more involved and a large number of link level
parameters have to be considered [16, 17]. Such parameters are: type of ITU propagation
channel, direction of the link (DL/UL), number of timeslots and codes allocated in every
radio frame, Spreading Factor (SF), code rate, interleaving period, receiver characteristics,
etc.
The first step for the integration of the link and the network level aspects in the same
system simulator is represented in figure 3.3: for each allowed configuration of physical
parameters in the scenario that will be later simulated at system level, the link level
simulator generates the related BER/BLER look-up tables (as described in section 3.2.2),
which are then recorded in a new tool, called Link-to-Network level Interface module or
more in short interface module
In this phase, a lot of curves have to be generated by the link level simulator; however,
this effort is required only at the beginning of the study and doesn’t depend on the network
level configurations, i.e. these tables shall not be created again for different traffic statistics
or different RRM algorithms, but only when the configuration of physical parameters
changes which isn’t expected to happen frequently during the system simulation study.
Moreover, for each propagation channel, the link level simulator provides to the in-
terface module the vector of the time-correlated fading samples of resolution equal to the
duration of the timeslot.
3.3.2 Interface Module to Network Level communication

In the system simulation, where ”system” is defined as the union of link and network level
issues, there is a continuous communication between the network level simulator and the
interface module which contains all the link level characteristics in terms of BER/BLER
curves and fast fading profiles (see previous subsection). The scheme describing the
interactions between these two blocks is depicted in figure 3.4.
Figure 3.4: Interface module-to-network simulator communication: block diagram
The interface module communicates real-time with the network level simulator; the
quantities exchanged by the two blocks can be classified in time slot (T S)-based (here
denoted as Mode A) for fast fading enforcement and transport block (T B)-based (Mode
B) for BER/BLER evaluation. We assume that the interleaving period is coincident with
the duration of a TB.
Mode A - Time Slot Based
The system model considers a time discrete version, with a resolution equal to the timeslot
duration, of the correlated fast fading. The link-to-network level interface computes the
received carrier, CP LSH , and the contribution of the interference coming from each single
source (terminal or base station), IK P LSH ; both terms are affected by pathloss (PL) and
log normal shadowing (SH): the scope of this phase is to determine the final signal-to-
interference ratio affected by fast fading.
CP LSH and the various terms IK P LSH represent an input for the interface block which
adds a sample of fast fading (ITU channels are considered) to both CP LSH and each
IK P LSH giving:
C = CP LSH · Af ad (3.1)
where Af ad is a factor that represents the additional attenuation of the user signal caused
by multipath propagation.
An expression similar to (3.1) can be written for each interfering signal IK P LSH :
IK = IK P LSH · AK f ad (3.2)
Indeed no assumptions on gaussian distribution of inter-cell interference is made. This

more realistic way of proceeding is necessary in our simulations, since high data rate packet
transmissions are characterized by sudden variations of the interfering power that cannot
be modelled using the gaussian hypothesis.
Once the fast fading samples have been added, the quantity (Ebc /Io )f ading is evaluated
as:
µ ¶ µ ¶
Ebc C Q C Q
= · = · (3.3)
Io f ading I log2 L Iinter + α · Iintra + N log2 L
where C is the received power, Iinter is the inter-cell interference, Iintra is the intra-cell
interference, N is the thermal noise power, Q is the Spreading Factor and L denotes
the levels of modulation (L=4 in QPSK), and α represents the orthogonality factor. In
uplink we substitute α with (1 − β), where β represents the Multi User Detection (MUD)
efficiency. All the terms of power and interference are affected by fast fading as explained
in equations 3.1 and 3.2.
Ebc denotes the encoded bit energy in the section before the decoder. Since the inter-
leaving duration is equal to 10, 20, 40 or 80ms, corresponding to 2, 4, 8 or 16 subframes,
the interface module through the link simulator results couldn’t evaluate the signal-to-
interference ratio after decoding with a resolution equal to the time-slot.
Mode B - Transport Block Based
At the end of the reception of a transport block, the network level simulator evaluates
(Eb /Io ) averaged over the duration of the transport block (where Eb is the information
bit energy after decoding):
PM ³ Ebc ´
¿µ ¶À
Eb i=1 Io
f ading 1
= · (3.4)
Io TB
M r
where r denotes the code rate and M is the number of timeslot required for the trans-
mission of a single transport block (in mode A, link level has previously provided these
M values of Ebc /Io to the system simulator).
Eb /Io represents an input for the link-to-network level interface module which returns,
through look up tables, the BER and the BLER referred to the considered block.
In our network simulator, during a CS connection (RLC protocol in transparent mode),
long term averaged BER values determine the quality of the connection. Viceversa, in case
of PS sessions, through the value of the measured BLER, the system evaluates whether
a transport block belonging to a data packet is correctly received. For PS services we
use a pair of RLC protocol istances in acknowledge mode, providing a reliable radio
bearer service, including error correction by automatic retransmission, thus when the
transmission fails, the transport block is re-transmitted.
Thanks to the proposed approach, described in section 3.2.2, it is actually possible
to calculate the Block Error Rate related to each transport block transmission; on the
contrary, the more conventional approach in section 3.2.1 doesn’t allow such a possibility:
a BLER averaged over a long period wouldn’t be useful when evaluating the data link
performance of a PS call.
Direction Time Quantity Definition
interval
TS CP LSH Received user code power affected by path loss
and shadowing
From NETWORK Level TS Ik P LSH Received power coming from the generic k−nth
interference source (UE or BS), affected by path
To INTERFACE module loss and shadowing
D³ ´E
Eb
TB Io Signal-to-Interference ratio, averaged over the
TB
duration of the Transport Block, after de-
spreading, demodulation and decoding
³ ´
Ebc
TS Io Signal-to-Interference ratio affected by fast fad-
f ading
From INTERFACE module ing, before decoding
TB BER Information bit error probability in a Transport
To NETWORK Level Block (CS)
TB BLER Transport Block error probability (PS)
Table 3.1: Definition of quantities exchanged between the interface module and the net-
work simulator
In table 3.1, the definition of the relevant quantities exchanged between the interface
module and the network simulator is summarized; the time interval in which they are
calculated and the direction of the information are reported.
3.4 An experiment for the validation of the proposed

link-to-network interface method
In this final section we will show the results of a simulative session drawn to compare in a
basic situation the BER calculated in the two approaches explained in sections 3.2.1 and
3.2.2. For more complex scenarios with both CS and PS users, refer to the simulation
results in chapter 2 which were obtained with the described link-to-network level interface
module implementation.
As the aim of this investigation is the comparison between the two approaches, a simple
case is here considered: the simulation was performed in the absence of convolutional
coding and fast power control; all the results are related to only circuit switching sessions,
because the approach in section 3.2.1 isn’t suitable for packet data analysis.
In figure 3.5, the logical steps adopted during this test are described.
Firstly, for each (Eb /Io ) reference, in the range of interest, through the link level
simulator implemented with the conventional approach described in section 3.2.1, it has
been measured the BER as a function of h(Eb /Io )i, that is as a function of the the Signal-
to-Interference ratio averaged over a long time period - see the red curve in figure 3.6.
Figure 3.5: Validation experiment: block diagram
Secondly, by using the link-to-network level interface approach proposed in this work
and described in section 3.2.2, it has been generated the curve of BER as a function
of h(Eb /Io )iTB , that is as a function of the Signal-to-Interference ratio averaged over the
duration of the Transport Block - see the green curve in figure 3.6 - that is over a duration
much shorter compared to the first method.
As it can be observed in figure 3.6, the BER curves generated by the link level simulator
based on the two approaches are significantly different (see the red and the green curve).
In particular the curve based on the averaged values of (Eb /Io ) over a period TB shows
smaller values of BER for the same nominal value of (Eb /Io ). The behavior can be
explained if we consider for instance a value of (Eb /Io ) of 5 dB. In case of conventional
approach, the (Eb /Io ) on the x-axes is averaged over a large time-interval, this means that,
owing to fading, in some cases the ”instantaneous” value of (Eb /Io ) maybe significantly
smaller than 5 dB, with a consequent increase of the ”instantaneous” BER. As the BER
plotted in figure 3.6 is averaged over a large time interval, the values of BER corresponding
to small (Eb /Io ) have a weight within the average larger than those corresponding to large
values of ”instantaneous” (Eb /Io ).
Afterwards, for this specific validation activity, it has been added to the network
simulator a function for the calculation of the average of all h(Eb /Io )iTB samples received
during the system simulation and for the calculation of the related averaged BER. For
Figure 3.6: BER as a function of h(Eb /Io )i and h(Eb /Io )iT B .
each (Eb /Io ) reference, a new system simulation has been performed, and the blue curve
in figure 3.6 of BER as a function of h(Eb /Io )i as been obtained by the network simulator.
In this case h(Eb /Io )i has been calculated as an average over a huge number of transport
blocks; in figure 3.6, it can be observed that this method provides therefore similar results
compared to the conventional approach when focusing the attention on merit figures
averaged over long term period (see the red and the blue curve).
Although in this validation experiment the agreements between the conventional ap-
proach performance and the ”instantaneous” one with post-averaging at network level is
very good, we have to highlight that the proposed interface is the only suitable approach
when a more accurate analysis is needed and PS traffic is simulated.
To conclude, in this chapter we have described a methodology to interface link and

network level tools. Performance evaluation of modern mobile radio systems requires the
joint consideration of link and network level aspects; this requires the use of an approach
which is different from that used in the past for 2G systems when the two aspects were
completely separated and few look-up tables, with values of BER averaged over fast fading,
were sufficient to take link level into account.

We stress that although our approach becomes equivalent to the conventional one when
circuit switched services are considered, the conventional approach cannot be used in case
of packet switched applications owing to the necessity to evaluate sudden variations of the
link quality. In this latter case, only approaches based on the proposed implementation
can be applied. The methodology presented here is quite general and is valid for different
radio access technologies.
Chapter 4
Architectures for Heterogeneous

Wireless Networks
Third generation for cellular telephony is already a reality. A new way to communicate
and a growing number of different services will be the challenge for UMTS. High data
speed will allow video-communication and mobile Internet on cellular terminals. But the
request for bit rate will never stop, and greater bandwidth will be necessary.
At the same time WLAN technology is starting to be an ordinary way to realize mobile
connection to the Internet, while companies all over the World realize wireless connections
in particular locations (hot spots), like airports or hotels. Anyone, owning the appropriate
technology on his laptop, can connect to the Internet in these places, at a reasonable price
with satisfactory connection speed.
On these conditions, UMTS and WLAN interworking becomes a really significant
issue: combining the two technologies would double the available resources. How to gain
this interwork is a large field for researches.
The heterogeneous technologies employed in 3G cellular networks and WLANs bring
many challenges to the interworking. Based on different radio access techniques, cellular
networks and WLANs present distinct characteristics in terms of mobility management,
61
62 Architectures for Heterogeneous Wireless Networks
security support, and quality of service (QoS) provisioning. In order to achieve seamless
integration, these issues should be carefully addressed while developing the interworking
schemes [28]).
Different aspects have to be considered while discussing about UMTS and WLAN
interworking: in this chapter, the state of the art in services, the most important network
architecture solutions, comprising an original one, and some functional issues will be
discussed; in the next chapter 5, an advanced Common Radio Resource Management
algorithm will be presented and the related performance will be simulated through the
simulation platform SHINE, developed in our laboratories (see appendix A).
4.1 Service interworking
In [61] 3GPP defines six levels of interworking, focusing on the service provided:
Level 1: ”Common billing and customer care”. Users pay a common bill to connect
both to UMTS and to WLAN. This level needs only a commercial agreement among
the different operators, sharing their costumer information; no technologic advance is
required. Each system offers the same service, as there was no interworking.
Level 2: ”Common access control and charging”. WLAN reaches the same level of
security of UMTS. Users notice the same interface connecting to both the systems.
Level 3: ”Access of all UMTS Packet Switching based services”. No handover is
provided, but both technologies offer the same services: QoS is managed even by WLAN.
Level 4: ”Service continuity”. Handover is managed, but the user can experience
service interruption or noticeable degradation.
Level 5:” Seamless mobility”. Users do not notice any difference using any of the
two systems neither during the handover. Circuit Switching services of UMTS are not
provided by WLANs.
Level 6: ”Access to Circuit Switching services”. Even Circuit Switching services are
supported by WLANs.
The basic target of our investigations was the interworking level 5. Nevertheless,
although the road map of the development of the Core Network foresees in the next future
an all-IP Core Network, we propose an integrated UMTS-WLAN architecture which allows
the WLAN users to exploit also the same Circuit Switching services provided by the Core
Network to the UMTS users, thus achieving the interworking level 6.
Architectures for Heterogeneous Wireless Networks 63
4.2 Proposed architectures

Different hypothesis are considered about how should the networks be connected. Three in
particular are proposed (for example in [28]): loose coupling, gateway and tight coupling.
Loose coupling (fig.4.1) considers the two networks as separated. To permit mobility,
maintaining the session open, Mobile-IP is implemented. A user is referred to a server
called home agent, owning the IP address of its connection. When it moves to a different
server (called foreign agent), the two servers communicate to each other. Every message
to the user is now sent to the home agent (maintaining the same IP address), whose duty
is to redirect it to the foreign agent. This solution makes new networks easy to be installed
and it does not require a tight cooperation among different operators. Real-Time services
hardly will survive during a handover, because of the time latency introduced by mobile
IP.
Figure 4.1: Loose coupling
Another solution is to introduce a new node, a Gateway (fig.4.2), connecting the

WLAN network to SGSN and GGSN in the UMTS core network. The Gateway to SGSN
connection is used while a WLAN customer is roaming on UMTS, while the Gateway to
GGSN connection is used when a UMTS customer is roaming on WLAN. Handover can
so be realized in a quicker way, while maintaining the two networks enough separated.
Operators of the two networks must cooperate, while hardware of the two parts needs to
be connected together. Introduction of new networks is not that easy.
The third solution is to use the WLAN technology as an access stratum to the UMTS
network (Tight coupling, fig.4.3). An interworking node is required to connect the WLAN
AP to the UMTS network. This connection can be done in the RNC or in the SGSN. In
the former case the AP acts as a Node-B. In the latter it acts as an RNC; a connection
even to the neighboring RNCs can be thought in this case. This solution reduces the
handover latency, making the UMTS Core Network to be the bottle neck for the traffic
Figure 4.2: Gateway approach
speed; the Interworking Unit must be designed appropriately.

Based on the Tight Coupling concept, a number of leading international companies
in the field of wireless communications have jointly developed a set of open specifications
(that can be found in [35]) in order to encourage the deployment of interworking networks
involving GSM and unlicensed spectrum networks (i.e. WLAN, bluetooth, etc.): the result
was the design of a new technology called UMA (Unlicensed Mobile Access).
Although the Loose Coupling approach seems to be the easier and thus the more
practicable way for 3G and WLAN interworking, the UMA specifications draft was a
clear demonstration that the Tight Coupling approach was much more than a ”proposed
solution”: actually, the 3GPP continued the work based on UMA, and under the designa-
tion of Generic Access Network (GAN) in 2006 the first specifications for such a system
architecture were officially released [60].
Figure 4.3: Tight coupling

4.3 Architectural and implementation issues

Assuming that UMTS/WLAN dual mode terminals are available, two aspects have to be
detailed in order to make UMTS and WLAN interwork: an architectural proposal for the
integrated network and a feasible practical realization.
4.3.1 System Architecture

Among the different solutions proposed from an architectural point of view (as detailed in
subsection 4.2), in this thesis, the performance of the UMTS-WLAN integrated network is
investigated with reference to the tight coupling approach, which is at the same time the
most challenging and the most promising architectural solution, as already mentioned.
The basic prerequisite to deploy a tight interworking is that the technology of dual-
mode UMTS&WLAN mobiles is available: this type of equipment should be able
to switch between the two radio access network, i.e. when in UMTS mode, the mobile
should periodically sense 802.11 coverage, for example every 2-3 sec.
Different solutions can be envisioned to realize the tight coupling between the WLAN
Access Point (AP) and the UMTS Terrestrial Radio Access Network (UTRAN): the AP
could be seen, for instance, as an additional cell or as an additional Node-B connected to
the same Radio Network Controller (RNC) which controls the adjacent UMTS Nodes B.
Let us observe, however, that the Iub control plane (see left side of figure 4.4) should be
adapted to support an ”enhanced” cell or Node-B: consider, for instance, that an NBAP
(Node-B Application Part) Cell Setup procedure for WLAN AP should be introduced in
3GPP specifications. The user plane implementation is also a limitation, since the frame
protocol should be adapted to the higher WLAN rates.
Thus, in our opinion, the best architectural solution is moving to a Radio Network
Subsystem (RNS) perspective (see right side of figure 4.4): an RNC emulator is inserted
between the AP and the Core Network acting as an interface between the WLAN world
and the UTRAN, without any impact on both the two technologies and taking advantage
of the same MSC/SGSN services already provided to UTRAN by the Core Network.
This way, we can exploit the essential fact that neither the adjacent RNCs nor the Core
Network need any knowledge about the internal resource management of RNSs. Moreover,
we think that an Iur logical interface between the standard RNC and the RNC emulator
is not required (see figure 4.4): since macro-diversity between cells under Serving-RNC
(UMTS cells under standard RNC) and cells under Drift-RNC (WLAN AP under RNC
emulator) can not be exploited, Iur user plane would not be used.
Regarding signalling for inter-system handover procedure, we could then use the Serv-
Figure 4.4: UMTS - WLAN Inter-working architecture: proposed logical scheme
ing RNS Relocation function already provided by 3GPP specifications, that manages the
mobility of the Iu connection from a RNS to another (e.g., from Iu-1 to Iu-2 in figure 4.4)
without loss of Packet Data Protocol -PDP- context or any other session management
information (see [28] for a possible message flow allowing the execution of the relocation
procedure): in this way, a seamless handover UMTS-WLAN without perceiving any
service interruption can be performed.
Concerning data flows, two cases have to be considered: interactive/background ser-
vices and conversational services (voice or video) [62]. In the first case, when the mobile
terminal is camped within the WLAN hot-spot the Iu-PS data packets coming from SGSN
and carried over Iu-2 (see figure 4.4) should be forwarded by the RNC Emulator to the
AP and vice versa; no procedure change is needed when, after a Serving RNS Relocation,
the new Iu-PS is Iu-1.
Regarding conversational services, our proposal is to exploit real-time transport char-
acteristics of Iu-CS also when the mobile terminal is connected to the WLAN AP and data
are packetized, as in the case of VoIP calls: the main issue is how to adapt RTP/UDP/IP
packets coming from the mobile terminal to a circuit switched based context.
To this aim, without any impact on the mobile terminal functionality, an H.323 [45]
gateway between the AP and the RNC Emulator could be inserted [46], providing protocol
translation and media transcoding between the endpoint of the PS domain (the WLAN
AP) and endpoint of the CS domain (RNC Emulator).
As far as the protocol translation is concerned, the easiest solution could be that Non-
Access Stratum messages from MSC (call setup messages, for instance) are transparently
forwarded by the RNC Emulator towards the H.323 Gateway; vice versa for messages
coming from the AP WLAN. Note that Non Access Stratum messages in UMTS only sce-
narios are already exchanged between mobile terminals and MSC without RNC knowledge
(Radio Access Network Application Part -RANAP- messages on Iu and Radio Resource
Control -RRC- Direct Transfer messages on Uu, for instance).
An H.323 Gatekeeper should also be introduced to translate E.164 addresses (i.e.,
phone numbers) into Transport Addresses (e.g., IP address and port address).
4.3.2 Implementation issues

Having defined the conceptual architectural solution for UMTS/WLAN integration, a
feasible practical solution has to be investigated in order to conveniently provide a logical
and a physical link between the RNC emulator and the Core Network.
Here we imagine that the same network provider will manage both the WLAN and the
UMTS networks, hence it is possible that the WLAN AP, as well as the H.323 Gateway
and the RNC emulator will be co-located with an UMTS Node-B, to reduce costs and to
ease maintenance.
In this case, the physical contiguity between the RNC emulator and the Node-B sug-
gests to adopt already existing UTRAN interfaces for the physical and logical WLAN-
UMTS integration. Due to the fact that Layer 1 Iub and Iu comply with the same
requirements [47], and that the physical layer provides the same ATM services according
to ITU-T I.361, we could multiplex different logical interfaces over the same physical link.
In particular, an ATM switch could be inserted in the physical connection already exist-
ing between the Node-B and the RNC, in order to switch the flows to/from the Node-B
(logical Iub interface) and the flows to/from the RNC Emulator (logical Iu interface) (see
figure 4.5).
In this scenario the ATM switch should be located close to the Node-B and con-
sequently also close to the AP, which is also convenient for network management and
maintenance.
Thus, the physical link between the ATM switch and the standard RNC carries two
different logical interfaces: the Iub from Node-B and the Iu from WLAN RNS. The latter
should then transparently cross the standard RNC exploiting the RNC switching capa-
bility, thus making the RNC emulator directly connected to the Core Network, according
Figure 4.5: Node-B and WLAN integration: implementation
to the conceptual scheme of figure 4.4.

It follows that the link between the standard RNC and the CN conveys Iu for both
the standard RNS and the WLAN RNS.
It is worth noting that, apart from the short physical links connecting the AP to
the the ATM switch through the H.323 Gateway and the RNC emulator, no significant
physical interconnections have been introduced by this solution; obviously, the capacity
of the link between the ATM switch and RNC should be adapted in order to convey also
the WLAN traffic.
Figure 4.6: Evolved multi-standard (UMTS & WLAN) NodeB communicating to the
Common Radio Resource Management (CRRM)
Finally, another logical link could be added between the standard RNC and the RNC
emulator, (the path would be the same used for Iu by WLAN RNS - see figure 4.6): this
link could be dedicated for Operations and Maintenance (O&M) functions: we expect
that a Common Radio Resource Management functionality between different RNSs and
particularly between a standard RNS and a WLAN RNS could be carried out through
this O&M communication link, leading to an increase of network performance.
In the figure 4.6, we considered the CRRM element located in the RNC, since it could
communicate at the same time to several WLAN RNSs concentrated to the same RNC.
As will be discussed in detail in the next chapter, the CRRM communicates with both
the UMTS Radio Resource Management (RRM) block located in the standard RNC and
the WLAN RRM block.
In the next chapter we’ll show the benefit introduced with our architectural solution
by the proposed inplementation of this new network element, the CRRM entity.
Chapter 5
Common Radio Resource

Management UMTS & WLAN
In this chapter, we will investigate the possible advantages introduced by the Common
Radio Resource Management (CRRM) for a heterogeneous integrated and interworking
UMTS-WLAN network; the performance evaluation will be carried out through simula-
tions, performed adopting our advanced SW platform called SHINE illustrated in Appen-
dix A.
The work has been organized in this way: in section 5.2, the required interactions
between the CRRM entity and the local RRM entities will be described; in section 5.3, the
main functions of the RRM UMTS and the RRM WLAN will be presented; afterwards, in
section 5.4, the CRRM algorithm we projected will be explained in detail. After a proper
definition of performance measurement in section 5.6 for such a heterogeneous network,
we will present the studied scenario in section 5.7 based on a hotspot of high density
traffic covered by both UMTS and WLAN radio access networks. In the numerical results
section 5.8, the benefit in the system capacity provided with the implementation of a
CRRM QoS based will be shown.
71
72 Common Radio Resource Management UMTS & WLAN
5.1 The CRRM challange

In the last few years several projects were dedicated to the issue of 3G/WLAN interwork-
ing networks (e.g. [36], [37]) and many papers appeared, investigating specific aspects.
Most of them are related to architectural and/or signalling issues (see [27], [28], [38]) while
some papers were focused on TCP performance in the presence of vertical handover (i.e.
handover among different technologies) [39]. Handover is indeed a very relevant aspect
in such networks and has been investigated also in [29], [30], [40] and [41]. The manage-
ment of the integrated network, with particular attention to profiles and parameters to
be monitored, has also been investigated in [42] and [43].
However, although many operating schemes have been proposed, no deep investigation
of the impact of networks integration on the overall performance experienced at applica-
tion level by the final user has been carried out so far. In this study it is envisioned the
possibility to serve by means of the WLAN technology, which can operate at bit rates
considerably higher than 3G systems, those voice calls that within the hot spot cannot
be served by the UMTS network (and would therefore be blocked) in case of saturation
of its radio resources.
A format conversion from circuit-switched (CS) voice flows into packet-switched (PS)
Voice over IP (VoIP) flows (and vice versa) is required in order to allow the transcoding
between the speech bearers over the Iu-CS interface and RTP/UDP (see chapter 4).
In previous chapter, we already described in detail the required aspects for a tight
integration between UMTS and WLAN. As a reminder, here we summarize the main
points:
• Availability of a dual-mode mobile UMTS - 802.11, capable of switching between

the two technologies.
• All type of services (i.e. speech and data transmission) shall be offered by the two
networks.
• Handover ”seamless” between UMTS and WLAN.
• Feasible interworking architecture: re-use of already existing logical interface Iu-

CS/Iu-PS and creation of a new Network Element providing policies for the Common
Radio Resource Management (CRRM).
In this chapter, we focus on the functionalities and benefits provided by the CRRM
block. The Common Radio Resource Management refers to the set of functions that
are devoted to ensure an efficient and coordinated use of the available radio resources in
Common Radio Resource Management UMTS & WLAN 73
heterogeneous networks scenarios. We’ll show algorithms we implemented in the CRRM

in a position to improve the overall system performance. The performance evaluation of
a heterogeneous integrated and interworking UMTS-WLAN network will be carried out
through simulations, performed adopting the SHINE platform.
Actually, although some tools developed for the simulation of heterogeneous network
have been already realized and described (see [37] and [44], for instance), we differentiate
from previously proposed solutions since we investigate the performance from the users’
point of view, reproducing in details all aspects related to all levels of the protocol stack
affecting the perceived Quality of Service (QoS), from the application to the physical
layers, including channel characteristics of each different radio technology (that is, for
example, unlike the just quoted tools which work on pathloss information only and a
rough model of the physical channel, we consider also the fast fading effect in our home-
made simulation tool). Moreover, none of the published papers considers the voice flows
somehow served by WLANs, thus limiting the capacity enhancement to data packets only.
The feasibility and the benefits of 3G/WLAN interworking are here investigated with
reference to the Chinese TD-SCDMA UMTS technology [47] and to the IEEE802.11e [31]
WLAN technology (adopting the IEEE802.11a physical layer [13, 10]).
5.2 Interactions between CRRM and RRMs

Before discussing the algorithms to be implemented in the CRRM block, the type of
relation between the CRRM block and the already existing Radio Resource Management
(RRM) blocks has to be defined.
Let us recall that one RRM function is already located in each Radio Access Technol-
ogy (RAT) to optimally manage the resources available in the single radio technology. In
our vision of different interworking networks, a centralized CRRM is needed to coordinate
and define policies to be followed at local level by each RRM. For the sake of simplicity
we’ll consider a system constituted of a single CRRM entity and many (2 or more) RRM
entities, that is we’re not interested in this study to the communication among different
CRRM entities. One feasible realization of this system is to locate the CRRM in the
UMTS RNC (see chapter 4); the local RRMs shall be located in the RNC (UMTS) and
in the RNC Emulator (WLAN).
We estimate that the interactions between CRRM and RRM entities (that is the arrows
between each RRM block and the CRRM in figure 5.1) involve mainly these functions
we’ll present in the next subsections:
• Network topology information.

Figure 5.1: RRM & CRRM relations
• Network load report.
• RRM report.
• CRRM decision.
5.2.1 Network topology information

The CRRM block needs to know the geographic position of the UMTS cells and of the
WLAN APs. This allows the CRRM to automatically create the external adjacent cell
relations for the Inter-system Handover between WLAN and UMTS. As an alternative,
CRRM may use the Location-area identification comprising the mobile country code,
mobile network code, and location area code (LAC) corresponding to the cell/AP [58]
(i.e. by using a different LAC for each AP WLAN).
This type of information may be exchanged between the RRM and the CRRM only
once, at the setup of the cell/AP.
In SHINE, our simulation platform for heterogeneous networks, the geographic infor-
mation is communicated at the start up of the simulation.
5.2.2 Network load report

Each already existing RRM shall be enhanced to report to the CRRM some merit figures
directly in relation with the load of the network. The purpose of these reports is that
the CRRM algorithm should use this information to increase the overall system capacity,
therefore a thoughtful choice of the merit figures shall be done. Each wireless technology
has different characteristics in terms of air interface technology, cell-size, coverage, etc.,
therefore the merit figure reported may differ from a RRM to another: for example the
UMTS cell could report the level of uplink interference, the value of transmitted carrier
power, the OVSF code tree usage, the latency time of HS-DSCH queues in the UMTS
NodeB, whereas the WLAN AP could report the channel occupation rate or the collision
rate.
Because of the huge set of merit figures that each local RRM can choose in order to
report the load of its cells, the tasks of the CRRM could become extremely complicated.
As an example, the CRRM may have difficulties to interpret the value of the transmitted
carrier power (TCP) of a UMTS cell in order to take the right action (i.e. to move users
from UMTS to WLAN because of high TCP in the UMTS cell): if the UMTS cell reports
TCP equal to 10W, the CRRM doesn’t understand if it is a high or low value because
the UMTS cell could be a micro cell (for example, maximum TCP equal to 12.5W) or a
macro cell (for example, maximum TCP equal to 40W). One solution in order to make
easier the task of the CRRM is to ask the local RRMs to report percentage values and not
absolute values (in our example, compared to the maximum TCP, the micro cell would
report 80%, the macro cell would report 25%).
Another point to be discussed is the amount of network load information to be ex-
changed between the RRM and the CRRM: main concern is on the RRM located in the
UMTS RNC which already handles several hundreds of UMTS cells. In order to reduce
the complexity of the UMTS RRM and the quantity of signalling between the RRM and
the CRRM, we decided the two following points:
1. the local RRMs send triggered measurements to the CRRM, that is reports from the
RRM to the CRRM aren’t periodic, but only when a threshold is exceeded (above
or below) a report is generated;
2. only the cells which are deployed in areas covered by different radio technologies
shall send their network load information through RRM (i.e. areas covered by
UMTS cells only don’t require the UMTS RRM sends any report to the CRRM).
Final remark about the network load information exchange is on the time scale: in
order to better the system capacity, the CRRM needs to exploit in real time the comple-
mentary characteristics offered by the different radio access technologies. Therefore each
local RRM has to provide to the CRRM up-to-date information, averaged over short time
intervals (i.e. 100ms, 1s): for example, in order to solve a radio congestion situation in
one system, it’s required the CRRM quickly receives this information and immediately
takes the decision to command an intersystem handover to an alternative radio technology
before any call is dropped.
In SHINE, as basic report, we decided the WLAN RRM entity reports the channel
occupation rate of the AP, whereas the UMTS RRM entity reports the downlink (NodeB
to UE) and uplink (UE to NodeB) OVSF code tree usage rate. However, the platform
may be extended to work with other merit figures.
5.2.3 RRM report

The local RRM entities shall send to the CRRM some of the outcome of their RRM
algorithms.
First of all, in order to implement a CRRM ”Service Based” (see subsection 5.4.1),
it’s necessary each local RRM informs the CRRM when a call setup is attempted in its
own RAT: the RRM shall send the QoS profile received from the Core Network in the
RANAP RAB Assignment message [59]. With this information, the CRRM ”Service
Based” communicates to the local RRM whether the Radio Bearer shall actually be setup
on its own RAT, or whether a Directed retry procedure shall be started: directed retry
is the process of assigning a user to a radio resource that does not belong to the serving
RRM by employing relocation procedures. Let us highlight that in the latter case, the
relocation procedure between the 2 RRMs uses standard 3GPP signalling [59] between the
RNC/RNC Emulator and the Core Network, that is the implementation effort is reduced.
Secondly, some of the local RRM functions (see section 5.3), like Admission Control,
may decide to restrict the service availability in one system because of a radio resource
shortage condition. If interworking UMTS-WLAN is available, the RRM shall also be
enhanced to inform the connected CRRM that such a type of event has happened and
therefore ask ”help” to the CRRM. In this case, there is a possibility that a blocked call
on one system, may be re-directed by the CRRM to an alternative system.
For this issue, the main RRM blocks reporting their decisions to the CRRM should
be the Resource Allocation (both admission control and code allocation), the Congestion
Control, the Pre-Emption and the Restriction Control.
In SHINE, the RRM decision report we use is the Resource Allocation function. How-
ever, the platform may support other RRM reports.
5.2.4 CRRM decision

The CRRM has the fundamental function of coordinating the use of the available radio
resources in heterogeneous networks scenarios in order to improve the overall system
performance. The CRRM decisions, like moving one user from one congested system to
another, are communicated by the CRRM to the local RRM; the RRM shall be enhanced
in order to start the standard Intersystem Handover procedure once it receives this order
from the CRRM.
In SHINE, the CRRM decisions are the initial RAT selection and the Intersystem
Handover command.
In figure 5.2, a diagram resuming the main interactions between the CRRM and each
RRM. At the system setup, every RRM sends the NETWORK TOPOLOGY INFORMATION;
at each call establishment the local RRM ask through RRM DECISION whether the call
shall be established on the own RAT or an alternative RAT: this decision is communicated
by the CRRM through CRRM DECISION. Finally, each RRM may send the triggered
NETWORK LOAD REPORT to drive the CRRM DECISION of an Intersystem Han-
dover based on the network load. It is evident, moreover, that no direct communication
between the two RRMs is required; actually this approach allows to use this CRRM block
connected to many different RAT (UMTS, WLAN, WIMAX, etc.).
Figure 5.2: Flow diagram of interactions between CRRM and RRMs
In the next two sections we present the functions normally implemented in the local
RRMs and the functions we propose to implement in the CRRM.
5.3 Local RRM functions

In order to understand the impact introduced by the CRRM, it’s appropriate to give a
brief description of the principal local RRM functions implemented in the UMTS and the
WLAN networks.
Main RRM functions of a UMTS network

• Radio Bearer Translation: to establish a radio access bearer between UE and Core
Network (CN), this function maps the Radio Access Bearer (RAB) QoS parameters
from the Core Network to the Radio Bearer (RB) parameters. This mapping has
to be performed for each new incoming bearer request.
• Radio Bearer Control: This set of functions aims to optimize the usage of radio
resources by adapting the amount of resources assigned to an UE depending on its
traffic load. This is achieved by managing the bit rate adaptation of the radio bearers
to the source bit rate and quality of service (QoS) requirements. The mapping has
to take into account the actual system load as well as the actual bit rate and quality
of service requirements of the considered radio bearer.
• Power Control: Since in UMTS all subscribers share the same frequency band,
lowering the interference within the cell while maintaining transmission quality is
of paramount importance. For this purpose power control adjusts the transmission
power in uplink and downlink.
• Handover Control : For UEs moving from a cell to another, this function transfers
connected calls to the new cell. The functionality is different for different scenarios:
Intrafrequency Handover Control (hard or soft handover within the same frequency),
Interfrequency Handover Control (handover between different radio frequency lay-
ers), Intersystem Handover Control (3G-2G).
• Resource Allocation (Admission Control): The admission control function decides

whether a new radio link can be admitted in a particular cell based on the cell load
and the available channelization codes.
• Pre-Emption: Pre-emption function allows the establishment of new higher-priority

calls in the face of high load by degrading lower-priority bearers. It interacts with
admission control.
• Congestion Control: Congestion control is a vital function, which detects an over-

load situation in a particular cell based on the load information from the Node Bs.
It resolves congestion by invoking functions for reducing cell load, such as channel
type switching, bit rate adaptation or bearer dropping.
• Restriction Control: This functions permits to restrict the available bit rates within
a cell by restricting the allowed minimum spreading factor.
Main RRM functions of a WLAN network

• Enhanced Distributed Channel Access (EDCA): IEEE 802.11e allows service differ-
entiation delivery.
• Call Admission Control (CAC): in order to ensure the QoS, a CAC algorithm may
decide whether an incoming user canor cannot access the network.
5.4 CRRM Algorithm

In our platform, we defined three main algorithms for the Common Radio Resource Man-
agement UMTS-WLAN (see the interaction from CRRM to RRM ”CRRM DECISION”
in section 5.2):
• CRRM ”Service Based”.
• CRRM ”Coverage Based”.
• CRRM ”QoS based”
The three CRRM options aren’t exclusive, in the section 5.8 we’ll show results with
different combination of CRRM algorithms. The final solution will be to use them all
together.
5.4.1 CRRM ”Service Based”

The algorithm of the CRRM, first of all, defines the RAT to be selected at the call setup
for each type of service. Given that UMTS can serve data traffic at excellent rates with
HSDPA technology, however the choice we made in this project is to preferably serve the
data traffic requests coming from the high traffic hotspot area with the WLAN network
because it can offer a higher bandwidth than UMTS. On the contrary, our choice for
speech users is to preferably serve them with the UMTS network, because of the better
speech quality UMTS can provide over its dedicated channels.
Bearer Preferred RAT Retry

Conversational UMTS WLAN (VoIP)
Best Effort WLAN UMTS
Table 5.1: Initial-RAT selection algorithm in the hotspot.
In SHINE, the CRRM Service Based policy can be easily changed, i.e. we could define
that for all calls in the hotspot the preferred RAT is the WLAN.
Figure 5.3: CRRM Service Based
Moreover, the CRRM allows to retry the call setup on an alternative RAT, in case of
radio resource shortage, i.e. if a voice call is rejected in the UMTS network, the CRRM
may order a ”Directed Retry” in the WLAN network.
In figure 5.4, we show three possible scenarios for CRRM Service Based.
In case (A), the RRM UMTS informs the CRRM that a user requiring QoS 1 is
attempting to establish the call on cell k; if QoS corresponds to a speech call, the CRRM
confirms that the call shall be established in UMTS.
In case (B), the RRM WLAN communicates that a user is attempting to setup a call,
for example a speech call, when connected to the AP z of WLAN; the CRRM Service
Based, because of the choices in table 5.1, shall inform the RRM WLAN that a Directed
Retry procedure to the UMTS cell k has to be performed: automatically the RRM WLAN
entity shall start a relocation to UMTS.
Finally, in case (C), the RRM UMTS attempts to establish a speech call, the CRRM
confirms the call shall be established in UMTS; after that, the RRM UMTS realize there
aren’t enough radio resource on cell k, therefore it communicates this Admission Control
rejection to the CRRM, which eventually decides to attempts the voice call establishment
in WLAN by sending to the RRM UMTS a CRRM DECISION of Directed Retry to
WLAN AP z.
Figure 5.4: Flow diagram of CRRM Service Based
5.4.2 CRRM ”Coverage Based”

Secondly, the UMTS-WLAN interworking is ”coverage based”: when a user connected
to the WLAN AP exits from the WLAN coverage, without service interruption, through
Intersystem Handover (see chapter 4), it is moved to the UMTS system.
Figure 5.5: CRRM Coverage Based
If UMTS rejects the relocation (for example, because of Admission Control), the call
is dropped and the relative counter is incremented: we decided in this case to avoid to
maintain the call in the WLAN, because of the well-known negative effect introduced by
WLAN stations far away from the AP.
5.4.3 CRRM ”QoS Based”

Finally, the UMTS-WLAN interworking is ”QoS based”: together with the huge effort
to develop the SHINE simulation environment, this part is one of the most important
contribution of this work. The purpose of the CRRM QoS based is to serve the voice
users in the RAT that currently, based on its load, best fits the QoS requirements and to
avoid to congest a single RAT when there are many RAT available in the same area.
In section 5.2.2, we described how each local RRM shall send NETWORK LOAD REPORT
to the CRRM. Here we explain how actually the CRRM uses these radio measurements
coming from each RAT. First of all, based on the threshold mechanism, the CRRM locates
each UMTS cell and each WLAN AP in one of the following three states:
• Unloaded state.
• Highly loaded state.
• Congested state.
The interaction with the basic CRRM Service Based is the following: the CRRM QoS
based decides that each conversational call (that is the most demanding service regarding
the packet delay and error rate) is setup on the less loaded RAT (in ascending order,
unloaded - highly loaded - congested); in case the UMTS cell and the WLAN AP are in
the same state, it is the CRRM Service Based that decides the preferred RAT as defined
in table 5.1, that is the speech call is setup in UMTS.
The CRRM QoS based we defined specifies also that the best effort calls in the hotspot
shall be always preferably served by the WLAN, that is for the best effort service, the
CRRM continues to follow the policy in table 5.1. Nevertheless, the CRRM QoS based
doesn’t setup any call when the WLAN state is ”congested”: in this case, the call is
established in UMTS.
Besides initial RAT reselection, the CRRM QoS based may command Intersystem
Handover procedures: when the CRRM QoS based detects that a UMTS cell or a WLAN
AP change the state from ”highly loaded” to ”congested”, the CRRM checks if there is
an alternative RAT in not congested state and in case this is true it starts to move users
through standard 3GPP RANAP Relocation procedure [59] from the congested RAT to
the alternative RAT (one handover every ∆T ) till the critical condition is solved.
In figure 5.6 the scenario of interest for the study of the behavior of the CRRM QoS
based: in a congested UMTS cell, the CRRM moves with Intersystem Handover some
users in the hotspot from UMTS to WLAN. In figure 5.7, it’s schematized the interaction
between the RRMs and the CRRM: after a NETWORK LOAD REPORT with ”con-
gestion” measurement in UMTS cell k, since the related WLAN AP z is unloaded, the
CRRM orders through CRRM DECISION to handover a user to AP z every ∆T . Once
the UMTS RRM informs in NETWORK LOAD REPORT that the cell k is not anymore
congested, the critical situation is solved and the QoS of all the calls in the system (both
UMTS and WLAN) has been maintained.
Figure 5.6: CRRM QoS Based
Figure 5.7: Flow diagram of CRRM QoS Based

Figure 5.8: Simulation platform architecture.
5.5 Software simulation platform settings

The simulation platform that we developed and adopted for the study of Heterogeneous
Networks is explained in Appendix A. In this subsections, a brief recall and some further
details on the parameters adopted for the next numerical results are given. As a reminder,
in figure 5.8 the block scheme of the platform is reported: the main concept is to divide
the overall problem in two areas, simulated by one Upper Layer Simulator (red area) and
two or more Lower Layer Simulators (blu areas). The Lower Layer Simulators used in
this study are the radio access network simulators for UMTS and WLAN.
5.5.1 Upper Layers Simulator-ULS

The main tasks of ULS are hereafter recalled:
• it sets the starting instant of each new traffic session originated by users, according
to the statistics of the traffic class it belongs to, as well as the users’ positions within
the investigated scenario;
• it generates the bit-flows up(down)loaded by each user in each traffic session ac-
cording to the statistics of its class of traffic;
• it reproduces the transport protocol behavior: both UDP and New Reno TCP are
implemented;
• it performs packet segmentation and reassembly;
• it executes all CRRM functions: it selects through which technology should each
user be connected on the basis of user-defined rules and the available networks’
information; it can also decide to reject a connection or to move it from a LLS to
another (that is, from a given technology to another) at any time, thus simulating
the network interworking;
• it finally collects all simulation outcomes and generates the outputs (throughput,
packet delivery delays,...) from an end-to-end point of view.
In our simulation platform each LLS manages its own time axis; the ULS, for its part,
communicates to the LLSs the next instant in which some event concerning the upper
protocol levels happens (sessions begin, start of bit transfers, TCP timeouts, etc.). In this
way, when the time counter of an LLS reaches that instant, the related LLS simulation
stops and a ”call” to the ULS is performed asking for the event-related information and
providing to the ULS, at the same time, information on the lower protocol levels events.
After the ULS replies, the simulation of the calling LLS resumes, the consequent actions
(new session start, MAC level frame queuing,etc.) are executed and the instant of the
next ULS event (TCP time-outs,etc.) is updated.
An LLS can perform a call to the ULS not only when the instant of the next ULS
event has come but also whenever an LLS event which is of interest for the ULS takes
place, such as, for instance, the correct transmission of a MAC level frame.
The above described stop-and-wait procedure is the basis for the coordination among
LLSs, which is obviously needed when simulating interworking networks: in case an ULS
event is of interest for more than one LLS, no ULS reply is granted to the calling LLSs
until all the interested LLSs have stopped waiting for it, then the reply is issued on the
basis of the LLSs reports provided to the ULS. It follows that although LLS executions
can take place at different speeds (LLSs complexity could be different and they could be
running on different PCs), the faster LLSs periodically stop and wait for the LLSs they
are interworking with, thus re-synchronizing simulations.
5.5.2 UMTS LLS

The UMTS simulation tool (that is, the UMTS LLS) I developed reproduces the main
characteristics of the TD-SCDMA radio interface reported in [47], Release 4.
UMTS Physical level Protocol

TD-SDCMA, which stands for Time Division Synchronous Code Division Multiple
Access, combines an advanced TDMA/TDD system with a synchronous CDMA. This
technology is characterized by a chip rate of 1.28 Mcps and a bandwidth of 1.6 MHz,
allowing 3 carriers in a 5 MHz band. Channelization code is OVSF (Orthogonal Variable
Spreading Factor) and a QPSK modulation scheme is adopted.
TD-SCDMA frame is constituted by 2 subframes of duration of 5ms: each subframe
is subdivided in 7 time slots, which can be flexibly assigned either to several users or to
a single user who may require multiple time slots.
Main features of our TD-SDCMA LLS are:
• power attenuation due to propagation was taken into account according to the
Walfish-Ikegami model, hence the pathloss (PL) dependence on the propagation
distance d is given by P L(d) = K1 + K2 log10(d). Here we assumed K1 = 15.3 and
K2 = 37.6;
• the shadowing is modelled by means of lognormal random variables with zero mean
and an exponential time correlation function; the spread here assumed for the log-
normal shadowing process is σ = 5dB;
• fast fading: different ITU channels are simulated (pedestrian, vehicular, indoor,
both A and B) [54]; here the pedestrian channel model has been assumed with an
user speed of 3.5Km/h;
• both fast closed loop power control (rate 200 Hz) and slow outer loop power control
are implemented; each bearer service is characterized by a proper value of initial
target signal-to interference ratio and different Transmitted Power Command step
sizes (∆) can be selected;
• the maximum transmitted power for each cell is 10W;
• both multi-slot and multi-code combinations are supported.
UMTS Layer 2 Protocols

Basic procedures are defined for the MAC simulator, providing a logical channel ser-
vice, and for the RLC simulator, providing a Radio Bearer service.
MAC main task is the priority-based management of the shared resource among users;
RLC provides, on the other hand, segmentation functionality and different reliability
modes: for CS speech connections RLC transparent mode is used and the quality of
service is estimated by averaging bit error rate measurements over long periods (here we
assumed 1.0 s); in case of PS sessions the simulator evaluates for each transport block
belonging to a data packet whether it is correctly received or not through the value of the
related measured block error rate (see chapter 3).
For PS calls a pair of RLC instances in acknowledge mode is used, providing a reliable
radio bearer service which includes error correction by means of automatic retransmission
of the transport block in case of reception failure.
At top of Layer 2, Radio Resource Control (RRC) block implements Call Admission
Control: a new radio link is successfully setup provided that the necessary OVSF codes
are available, the estimated interference is less than a given threshold and the initial power
required in Node-B is available.
The simulator decides to drop a call using a ”leaky bucket” algorithm, which com-
pares the average measured bit error rate, BER, with the threshold BERdrop = 2 · 10−2 :
if BER > BERdrop → increase counter (+1), otherwise → decrease counter (-2) ; if
counter ≥ 4, the call is dropped
Different classes of traffic are supported:
• CS Adaptive Multi Rate -AMR- speech at 12.2 Kbps;
• PS Best Effort: three Unconstrained Delay Data bearer services at 64/64 Kbps,
64/144 Kbps and 64/384 Kbps are considered, where x/y Kbps stands for x Kbps
for the uplink and y Kbps for the downlink.
In the following, when dealing with UMTS data services, we always considered the 64/384
Kbps bearer.
5.5.3 WLAN LLS

As hereafter detailed, the WLAN network simulator (that is, the WLAN LLS) carefully
reproduces the IEEE802.11a [13, 10] PHY level behavior as well as the MAC protocol of
the IEEE 802.11e [31] technology. Moreover, a Call Admission Control (CAC) strategy
has been implemented on top of IEEE802.11e MAC layer in order to prevent WLAN
saturation.
WLAN Physical level Protocol
As for the PHY issues, let us recall that IEEE802.11a is based, at the Physical level,
on eight operating modes adopting in the 5 GHz band the Orthogonal Frequency Division
Multiplex (OFDM) technique to counteract the effects of frequency selective fading [10].
Each mode is characterized by a different combination of modulation scheme and code rate
(R) of the punctured convolutional code adopted to correct transmission errors. Mode 1
(the slowest one), for instance, adopts a BPSK-based OFDM scheme with R = 12 , while
Mode 8 (the fastest one) adopts a 64QAM-based OFDM scheme with R = 43 .
All IEEE 802.11a PHY level aspects (propagation, modulation, channel coding, ...)
have been carefully taken into account by the WLAN LLS, in particular:
• according to ETSI recommendation, the multipath channel is generated in a time

and frequency correlated way following [11]. In this work we considered the channel
model A (see [11]) that corresponds to an indoor environment with 18 Rayleigh
distributed paths. The time-correlated channel variations are taken into account
considering a relative user speed of 3 km/h for the Clarke [12] Doppler spectrum of
each path;
• the Auto Rate Fallback (ARF) [32] link adaptation algorithm is assumed to select
the proper transmission mode (i.e., the combination of modulation scheme and
coding rate) following channel variations; MAC level frames are discarded after 7
consecutive failed transmissions; hard decision convolutional decoding is assumed;
• according to the technical specifications of most of commercially available WLAN

devices, here we assumed 19dBm for EIRP, 2dB for the antennas’ gains and 12dB
for the receiver noise figure. The signal power has been assumed decaying with the
fourth power of the distance.
WLAN Medium Access Control Protocols

As for the MAC protocol of IEEE 802.11e, here we considered the contention-based
access mode (Distributed Coordinated Function DCF for version ’a’, Enhanced DCF
EDCF for version ’e’, see [13, 31] for details) which is based on the Carrier Sensing Medium
Access with Collision Avoidance (CSMA-CA) strategy; the hidden terminal problem is
supposed to be negligible, hence the two-ways handshake procedure [13] is assumed in the
following.
Let us recall that IEEE802.11e differs from IEEE802.11a, as well as from IEEE802.11b/g,
since it introduces the concept of Quality of Service in the frame delivery procedure at
the MAC level.
The idea at the basis of the IEEE802.11e MAC is to manage traffic flows with different
throughput/delay requirements by means of different queues, which are given different
probabilities to gain the access to the channel: the more the requirements of the traffic
assigned to a queue are stringent, the higher is the probability its queue is given to gain
the access to the channel. IEEE802.11a MAC layer, on the contrary, does not differentiate
among different traffic flows.
Let us recall, now, that IEEE802.11e specifications [31] define only the MAC level
strategies, which can be combined with anyone of the IEEE802.11a, IEEE802.11b or
IEEE802.11g PHY layers; here, when dealing with IEEE 802.11e we mean the combination
of IEEE802.11e MAC level and IEEE802.11a PHY level.
When considering IEEE802.11e, the different traffic classes are assigned different pri-
orities: in table 5.2 we reported the values of IEEE802.11e MAC protocol parameters
adopted in our simulation to differentiate the access probability of each queue (see [31]
for details on their meaning). According to the values of table 5.2 the VoIP traffic is given
the highest priority (that is, the highest probability to gain the access to the channel)
while FTP traffic is given the lowest priority.
Voice Web browsing Ftp

Parameter AIFS CWmin m AIFS CWmin m AIFS CWmin m
value 1 8 2 4 16 6 4 32 6
Table 5.2: Set of parameters adopted for IEEE 802.11e
Call Admission Control Protocol

In order to prevent the saturation of WLAN resources, that would lead to a poor QoS
level perceived by final users, here we implemented on the IEEE802.11a/e simulator a
simple but efficient CAC solution, based on a centralized evaluation of the current network
congestion level. Let recall that this is done by monitoring the channel occupation through
the assessment of the Channel Occupation Rate parameter, CO [33], defined as the ratio
between the amount of time the medium is busy, TB , and the related observation time
∆T :
TB
CO = .
∆T
The evaluation of the channel occupation rate is particularly simple to be implemented
in existing Access Points, owing to their carrier sensing capability. The AP simply adds
the busy medium sensed time TSB to its transmission time TAP , as follows:
TB = TAP + TSB .
In order to improve the accuracy of the estimation of TB , also the mandatory idle
times DIFS, SIFS [13] and AIFS [31] are considered in the assessment of TAP and TSB :
in particular, a DIFS period is considered for each uncorrect transmission (i.e. each
packet not acknowledged) and both DIFS (or the appropriate AIFS, for IEEE802.11e)
and SIFS periods are considered for each correct transmission; please note that as long as
the transmission is successfully performed, the AP automatically knows the correspondent

class of traffic and thus its AIFS interval.
For high values of CO the frame collision probability increases and consequently the end
to end delay and the throughput experienced by users get worse. The CAC algorithm
has to keep the Channel Occupation under the saturation point, hereafter denoted as
Congestion Threshold (CT ). When CT is exceeded any admission is denied, till CO goes
below a Decongestion Threshold (DT ).
5.6 Performance measurements

An integrated UMTS-WLAN system needs also an integrated definition of Quality of Ser-
vice requirements and merit figures. The main idea is that the end-to-end user experience
should be independent from the radio technology - UMTS or WLAN - chosen by the
CRRM, i.e. the integrated system wouldn’t be attractive if in UMTS the speech quality
were satisfactory whereas in the hotspot the VoIP over WLAN were highly affected by
jitter and packet loss.
The merit figures shall provide the performance of the overall UMTS&WLAN system,
for example a call established in UMTS, moved to WLAN because of congestion, and
moved again to UMTS because of coverage shall be considered as a single call, and not
as three different calls. The only entity that can evaluate the global merit figures is the
CRRM, because it’s the only network element which can keep track of each call in the
heterogeneous network.
Quality of Service requirements
The quality requirements of the services provided in the scenario we will simulate are
the followings:
• Speech traffic. When served by UMTS, the speech traffic is a circuit switched bidi-
rectional flow. A speech user is assumed to be satisfied if its call is neither blocked
nor dropped, and the overall outage time is lower than 5% of the call duration in
each direction (an outage event for CS voice calls occurs when the average value of
the bit error rate exceeds the threshold BERout = 10−2 ).
When served by WLAN AP, the speech traffic is a VoIP flow; the user in this case
is satisfied if 97% of packets are received in less than 0.15 sec in each direction [46];
following [46], since the voice/VoIP conversion adopting a G.729 codec in the H.323
gateway takes around 120 ms (including coding, decoding, bufferization delay, etc.)
and assuming that the circuit switched Core Network introduces a negligible delay,
the maximum tolerable delay introduced by the WLAN is thus 30 msec.
It is clear the parameters used to measure the quality of speech calls are different
in UMTS and WLAN; in order to measure the outage of calls that were served by
UMTS and WLAN in different periods, we decided to count the number of time
intervals (for example, slot duration of 1 second) in which the user experienced
outage in UMTS or in WLAN, then to divide this number by the total duration of
the call and finally to compare this ratio with a 5% threshold.
• Web browsing traffic. Characterized by bursts (packet calls) of small application-

level packets followed by inactivity periods (reading time). A web browsing user is
satisfied if 90% of packets are received in less than 5 sec;
• FTP traffic. Bursty traffic characterized by requests of huge application-level pack-

ets followed by inactivity periods (reading time). An FTP user is satisfied if the
average experienced throughput is higher than 800 kbit/s.
Merit figures
The merit figures we will consider in the numerical results section 5.8 cover the different
phases of Quality of Service [26] aspects during service use from the customer’s point of
view. They are here defined:
• Call Setup success rate (CSSR): it regards the Service Access, that is if the cus-
tomer wants to use a service, the network operator of an integrated UMTS-WLAN
network should provide him as much as possible access to the service. CSSR de-
scribes the ratio between the calls successfully connected to the system (either
UMTS or WLAN) and the overall number of call attempts:
Nattempt − Nblock
CSSR = (5.1)
Nattempt
where Nattempt is the number of call attempts and Nblock is the number of blocked
calls (for example, for Admission Control). Note that if a call attempt in the
hotspot fails in both UMTS and WLAN, this is counted only once by the CRRM:
the perspective is to provide a performance indication from the user point of view;
each single system may internally count this failure for its own statistics, but at
the end the CRRM will count only one blocked call as only one user ”pushed the
button” to start a call.
• Drop Call rate (DCR): it regards the Service Retainability, that is it describes
the termination of services in accordance with or against the will of the user. DCR is
the ratio between the abnormal releases (i.e. dropped calls for poor radio conditions
or for congestion) and the overall number of call releases (that is both the normal
and the abnormal releases):
Ndrop
DCR = (5.2)
Nrelease
where Ndrop is the number of dropped calls and Nrelease in the number of call releases.
• Outage rate (OutR): only for speech calls, it regards the Service Integrity, that
is the Quality of Service during service use. OutR describes the ratio between
the number of calls which, although normally released, perceived an unacceptable
outage time lasting more than 5% of the duration of the call, and the overall number
of call releases (that is both the normal and the abnormal releases):
Noutage
OutR = (5.3)
Nrelease
where Noutage is the number of calls with unacceptable outage time.
• Satisfaction rate (SatR): only for speech calls, it summarize the degree of voice
user satisfaction, by considering altogether Service Access, Service Retainability
and Service Integrity. A call is considered satisfied if it isn’t blocked, nor dropped,
nor it didn’t feel outage:
SatR = CSSR · (1 − (DCR + OutR)) = (5.4)

Nattempt − Nblock Nrelease − (Ndrop + Noutage )
= · = (5.5)
Nattempt Nrelease
Nattempt − Nblock Nsat
= · (5.6)
Nattempt Nrelease
where Nsat is the number of satisfied users, that is the calls which were successfully
connected to the network and were normally released without perceiving significant
outage during the call.
In case of a stationary system, in which Nattempt = Nblock + Nrelease , from equation
5.6 it is easy to obtain an easier formula for the satisfaction rate SatR:
Nsat
SatR = (5.7)
Nattempt
Figure 5.9: Investigated scenario: WLAN APs in hotspot of high density traffic
• Average perceived throughput: only for best effort calls, it is the average value of the
application level throughput perceived by users; the application level throughput is
defined as the ratio between the amount of bits of each application-level ’packet’
and the time elapsed from the instant in which the TCP packet containing the first
fragment of the application-level ’packet’ reached the head of the TCP transmission
queue and the instant in which the TCP packet containing the last fragment of the
same application-level ’packet’ was completely successfully received;
• Average delivery delay: it is the average time interval occurring from the generation
of an application-level ’packet’ in the transmitter to its successful reception, thus
including in the delay calculation the time spent in the TCP and MAC levels queues.
5.7 Traffic scenario

The architecture for an integrated UMTS-WLAN system we’ve proposed in chapter 4
and the CRRM project we’ve explained in the previous sections 5.2-5.4, cover the general
scenario of a system in which both UMTS and WLAN radio access network are deployed
and there is an interest to exploit the complementary characteristics provided by the two
Figure 5.10: Simulated scenario. The grey square indicates the area considered for nu-
merical results
technologies.
We haven’t yet mentioned the location or the number of NodeBs and the number of
WLAN APs: in figure 5.9 an example, in which 2 WLAN APs cover high traffic areas (red
circles), inside a global UMTS coverage. In this section we define the network topology
and the traffic scenario we used for the simulation with our platform SHINE. Actually, in
order to carry out significant investigations on the benefits of networks integration and to
derive meaningful conclusions, particular attention has to be paid not only to the accuracy
of the simulation platform but also to the realistic modelling of all aspects characterizing
the operating conditions of real networks, such as the scenario, the variety of services
requested by users and the statistics and geographical distribution of traffic flows.
5.7.1 Network topology

The scenario investigated is depicted in figure 5.10 and consists of 18 UMTS Nodes-B
with tri-sectorial antennas over an area of 3720x2775 m2 which also includes a WLAN
hot spot. The WLAN AP is assumed co-located with a Node-B (one of the central ones in
figure 5.10) and is equipped with a dipole antenna offering an omnidirectional coverage in
the horizontal plane; to avoid border effects all merit figures investigated in the numerical
results section refer to an area of 100x100 m2 centered in the AP (see figure 5.10).
5.7.2 Traffic distribution

In order to reproduce the variety of services provided in a real network, here we assumed
that users could perform voice calls as well as web browsing and file transfer sessions, thus
generating three different kinds of traffic whose statistical description have been carefully
reproduced.
Regarding the geographical distribution of traffic, here we considered the superim-
position of a background traffic generated by users uniformly distributed in the whole
3720x2775 m2 region, and of a hot spot traffic generated by users uniformly distributed
in a circular area (hereafter denoted as the hot spot region) centered in the AP, with a
radius of 35m, which could represent a highly crowded area where the WLAN hot spot
is deployed (see figure 5.10):
• Background traffic: it is generated by ”background users” that are assumed to

perform voice calls or web-browsing sessions in the whole area covered by UMTS.
• Hot spot traffic: it is the further traffic contribution which is added to the ”back-
ground traffic” only in the hot spot region and it is constituted by a huge additional
amount of voice, web-browsing and file transfer (hereafter, FTP) traffics.
The straightforward consequence of the above reported assumptions is that the hot spot
region is characterized by a higher user density than the surrounding region: this is typical
of crowded areas such as, for instance, an airport gate or a shopping mall where, actually,
the realization of a hot spot is envisioned.
The above reported characteristics are summarized in the first two columns of table
5.3, in the third one the mobility type is reported (we assumed only the voice users
enters/exits from the Hot spot traffic), whereas in the fourth column the average call
arrival rate is reported.
Traffic Area Mobility Arrival rate
Voicebackgr. whole scenario Pedestrian/Vehicular [54] 2.7 calls/s
Web br.backgr. whole scenario Static 1.08 sessions/s
VoiceHS hot spot Static varying: k · fa (t)
Web br.HS hot spot Static 0.02 sessions/s
FTPHS hot spot Static 0.012 sessions/s
Table 5.3: Traffic distribution and arrival rates
We assumed the additional average voice call arrival rate in the hot spot (VoiceHS )
is a variable quantity depending on the time, fa (t), in order to simulate the behavior of
a hot spot of traffic in which the density of users change dynamically: in figure 5.11, it
is shown that in 4800 seconds of simulation there are two peaks of traffic of the same
Figure 5.11: fa (t): voice call arrival rate VoiceHS in the hotspot
maximum value (0.64 calls/s, corresponding to 76.8 Erlang, based on traffic parameters
of table 5.4), but with different variation step. In the numerical results section, the merit
figures are evaluated as a function of k, that is the multiplier factor of fa (t) in table 5.3:
for each run of simulation, the parameter k will assume the values 0.125, 0.25, 0.5, 0.75
and 1.
For the sake of clarity we finally summarize in table 5.4 the traffic categories we
considered (first column), their characteristics (second column), the references where the
models of each traffic class were taken (third column) and the satisfaction thresholds
(fourth column).
Class Characteristics Refer to Satisfaction thresholds

CS Voice Poissonian duration, 120 sec in aver- Natural conclusion with
age outage lower than 5% in
each direction
VoIP CBR traffic, 20 bytes packets with [46] (codec G.729) 97% of packets received in
rate 8 kbit/s per direction less than 0.03 sec (wireless
link) in each direction
Web B. 1 to 8 packet calls of 70 kbytes in [55] with α = 0.6, 90% of packets received in
average, divided into 1 to 30 packets average reading time less than 5 sec
each; 60 sec average reading time halved
FTP 1 to 6 packet calls of 500 kbytes in Each packet call fol- Average throughput of
average; 180 sec average reading time lows [34], with µ = 800 kbit/s
13.06 and σ = 0.12
Table 5.4: Adopted traffic classes: parameters and requirements for satisfaction
5.8 Numerical results

Given that we defined the CRRM strategy in sections 5.2-5.4 and a realistic traffic scenario
in the previous section, here the system performance of such an integrated UMTS-WLAN
network is finally investigated through our simulation platform SHINE; the main simu-
lation setting were described in section 5.5. Actually, the final scope of this work is to
evaluate the possible benefit of a CRRM approach when in the hotspot covered by WLAN
AP and UMTS there is a high distribution of voice call users that cannot be completely
served by UMTS only.
In the figures 5.12-5.16, the impact of different levels of integration between UMTS
and WLAN on Quality of Service is investigated in the 100x100 squared meters area
around the location of the WLAN AP. The results are provided for four different CRRM
functions, of progressive complexity:
• No CRRM: the voice calls in the hotspot are served only by UMTS: during the
peaks of traffic, the calls that are not admitted to the system won’t be redirected
to WLAN. This is the basic situation when no interworking between UMTS and
WLAN is realized.
• CRRM Service Based (see subsection 5.4.1): the voice calls in the hotspot are
preferably served by UMTS, but in case of resource shortage, call setup is redirected
to WLAN. Moreover CRRM Coverage Based (see subsection 5.4.2) is executed: users
exiting from WLAN hotspot do Intersystem handover to UMTS.
• CRRM QoS Based - 50% (see subsection 5.4.3): same algorithm used for previous
point (CRRM Service Based); moreover, the CRRM QoS based algorithm is enabled
too. Once the UMTS system reports a load of 50% (see subsection 5.2.2 for Network
load report) the congestion of UMTS is declared. This is the most advanced proposed
CRRM option, allowing users to be served by the best system in terms of congestion
of the radio interface.
• CRRM QoS Based - 85%: same as previous point, but with higher threshold -
85% - for UMTS congestion detection.
In figure 5.12, it’s shown the Satisfaction Rate (SatR) of voice users in the grey
square of figure 5.10, including the area covered by the WLAN AP; it’s evaluated the per-
formance when varying the multiplier factor k of the traffic function fa (t) (see VoiceHS
traffic definition in table 5.3). The worst result for SatR is obtained when no UMTS-
WLAN integration is established (No CRRM). The first substantial improvement is
Figure 5.12: Voice QoS in the investigated 100x100 m2 area: users’ Satisfaction Rate
SatR
achieved with the CRRM Service Based : during the peaks of traffic, the voice calls re-
jected by UMTS maybe redirected to WLAN, ensuring an increase of the number of voice
calls that can be managed by the system, as could be expected since without UMTS-
WLAN integration the available bandwidth is clearly lower. For k = 0.5 (corresponding
to traffic peaks of 38.4 Erlang), SatR increases from 70% to around 85%, whereas for the
higher value k = 1 (corresponding to traffic peaks of 76.8 Erlang) the Satisfaction Rate
increases from 40% to 60%.
The next step in the performance improvement is obtained thanks to the full CRRM
algorithm implementation, both Service and QoS based (for convenience, here we denote
this option only with the label CRRM QoS Based ): many parameterizations have been
tried, in this analysis two of the most interesting ones from the results point of view
have been reported. By using two different thresholds at 50% and 85% for Network Load
report of congestion in UMTS, the performance of the CRRM QoS Based in terms of
SatR improves the results of the CRRM Service Based.
For different ranges of traffic in the hotspot (k), a different parameterization of CRRM
Service Based improves the Satisfaction Rate: with k < 0.5, the CRRM QoS Based - 85%
option provides the best performance, whereas for k > 0.5, the CRRM QoS Based - 50%
option outperforms both the CRRM QoS Based and the CRRM QoS Based - 85%. In
Figure 5.13: Speech Service Access QoS in the investigated 100x100 m2 area: users’ Call
Setup Success Rate CSSR
particular, when k = 1, the Satisfaction Rate is about 80% with the CRRM QoS Based
- 50% : this result is the double compared to the performance of the No CRRM option,
as could be expected in general terms, but it’s also remarkable this CRRM QoS Based
option satisfies about more 20% users than the simpler CRRM Service Based option.
A single CRRM QoS Based algorithm changing dynamically its congestion thresholds
based on the current traffic (k) will be evaluated in future studies, in order to deter-
mine if it is possible to achieve or even improve with a single algorithm parameterization
the results provided by the envelope of the curves obtained adopting the two proposed
parameterizations (CRRM QoS based 50% and 85%).
Let us recall that the merit figure SatR is a Performance Measurement providing the
result of the combination of several phase of the service: access, retainability and integrity.
Now we separately study each of these aspects to investigate the root causes of the trends
of the Satisfaction Rate.
In figure 5.13, the results of Call Setup Success Rate (CSSR) of voice users in the
hotspot are presented. It can be observed that the system without interworking UMTS-
WLAN (No CRRM), starts to block voice users for k = 0.25: from this point on, the
service access can be improved only with the CRRM. At k = 0.5, both the CRRM Service
Based and the CRRM QoS Based provide CSSR > 99%, that is this integrated network
would satisfy an ordinary Service Level Agreement (i.e. CSSR > 96%); this is due to the
fact that the calls blocked in UMTS can be served by WLAN.
In the highest simulated traffic scenario, k = 1, apart from giving service access to
almost 50% additional users compared to the No CRRM scenario (93% vs 45%), it is
notable that the CRRM QoS Based - 50% provides service to more 10% users compared
to the CRRM Service Based algorithm. At first, this improvement of the CRRM QoS
Based compared to the CRRM Service Based might surprise since both these two CRRM
options have at their disposal the same amount of radio resources of UMTS and WLAN,
therefore it could be generally expected that the CRRM QoS Based should impact only
on service retainability and integrity as will be explained later, and not on service access.
From these simulation results, on the contrary, we have realized that the strategy of CRRM
QoS Based to establish and move calls on the best system in terms of radio interface load
leads to a better radio resource management and at last to provide access in the system to
a larger number of users: instead of exhausting the UMTS resources and then forwarding
the next call setups to WLAN (CRRM Service Based algorithm), it’s preferable to serve
calls in the less congested system.
Moreover, in the CRRM QoS Based configuration, once the UMTS cell in the hotspot
reports a High Load state, the CRRM starts to move calls in the hotspot from UMTS
to WLAN: this procedure allows the UMTS RRM to free resources for next incoming
calls that cannot be served by WLAN, for example, calls established just outside the area
covered by the AP WLAN.
In figure 5.14, the performance in terms of Call Drop Rate (DCR) of voice users in
the hotspot is shown. The basic configuration with No CRRM presents the worst results:
due to the high traffic load concentrated in the hotspot area, the service retainability
is reduced; calls can be accepted by the UMTS RRM, but because of peaks of uplink
interference, calls maybe abnormally terminated. When k = 0.5, the performance of No
CRRM is similar to the result of CRRM Service Based (DCR about 4.5%), even if the
latter can use the radio resources of both UMTS and WLAN; nevertheless, even if the
level of service retainability is similar, figure 5.13 at k = 0.5 shows that with No CRRM
25% less users accessed the system, therefore the system load is reduced, thus affecting
the comparison in terms of service retainability.
At k = 0.5, the CRRM QoS Based (both 50% and 85% options) halves the Drop Call
Rate to 2.4%, a value that can be acceptable for a Service Level Agreement; in this case
the comparison with the DCR of the CRRM Service Based is relevant since both of them
offer the same voice traffic quantity (CSSR > 99%). This clear improvement is provided
by the combined management of the load levels of the two radio interfaces UMTS and
Figure 5.14: Speech Service Retainability QoS in the investigated 100x100 m2 area: users’
Drop Call Rate DCR
WLAN: when one system achieves the congestion level, before any call is abnormally
terminated, the CRRM QoS Based attempts to move voice calls to the alternative radio
access technology, if that one isn’t congested.
Moreover, in the CRRM QoS Based configuration, since its general strategy is to avoid
to use all UMTS resources until calls start to be blocked, the voice call users exiting from
the area covered by the WLAN AP will likely perform a successful Intersystem Handover
to UMTS; on the contrary, with the CRRM Service Based only, the possibility to exit
from the WLAN area and then to not find enough radio resources in UMTS and therefore
being dropped is higher.
In figure 5.15, the results of Outage Rate (OutR) of voice users in the hotspot
are presented. The service integrity provided with No CRRM shows the best results,
but as explained above, this basic option offers a lower voice traffic, therefore a straight
comparison is tricky. Only in the most unloaded scenario (k = 0.125) CSSR = 100% for
all four curves (see figure 5.13), therefore a reasonable analysis can be performed. In this
specific case we can observe that OutR obtained with CRRM QoS Based - 50% is much
higher than with No CRRM : the reason is that, since UMTS radio channels are optimized
to transport the voice flows, for lower traffic values the best choice is always to serve the
voice calls in UMTS (No CRRM ). On the contrary, when the CRRM QoS Based - 50%
Figure 5.15: Speech Service Integrity QoS in the investigated 100x100 m2 area: users’
Outage Rate OutR
is used, some short spikes in the UMTS cell load may trigger the CRRM to move voice
calls to WLAN, in which voice users will likely perceive a worst VoIP experience. This is
confirmed by the trends in figure 5.16 - Number of Intersystem Handover procedures per
voice call - in which it is reported that with the option CRRM QoS Based - 50%, the
users execute about 0.5 Intersystem Handovers per call when k = 0.125, much more than
with the other configurations.
The same phenomenon is less relevant with CRRM QoS Based - 85% : in this case,
the higher threshold for UMTS congestion detection is less sensitive to spikes of cell load,
therefore for low values of k (k < 0.5) less voice calls are moved to WLAN, that is only
during real high load situations the CRRM will react and order Intersystem Handover.
The drawback of the option of CRRM QoS Based at 85% is that by reacting more slowly
to radio congestion events or by recognizing a lower number of critical network load
situations, when the offered traffic increases (k > 0.5), this CRRM option may solve too
late some congestion situation, therefore in this case OutR is much higher compared to
the CRRM QoS Based at 50%.
This distinction finally explains the trends detected in the first figure 5.12 of numerical
results, in which the highest Satisfaction Rate SatR is obtained with the CRRM QoS
Based algorithm, by choosing the two different thresholds for congestion at 50% and at
Figure 5.16: Number of Intersystem Handover procedures per voice call
85% in different ranges of traffic (k).

Apart from the number of Intersystem Handover procedures between the two radio
access networks, another interesting merit figure of the interworking between UMTS and
WLAN is the distribution of voice calls between the different networks. In the figures
5.17 and 5.18, the average subdivision of voice call users between UMTS and WLAN is
reported, respectively for k = 0.5 and for k = 1.
From figure 5.17, it’s immediate to verify how the global performance in terms of
Satisfaction Rate (see figure 5.12) has been improved from the condition of No CRRM
to the implementation of the CRRM QoS Based - 85% : the latter serves 30% of the
users in WLAN, most of them wouldn’t otherwise obtain service access. On the contrary
the option of CRRM QoS Based with threshold at 50% triggers a higher number of
Intersystem Handover from UMTS to WLAN: even 55% of the voice users are served
by WLAN, therefore, because of the lower performance of VoIP in WLAN, the service
integrity of the system is reduced compared to the case of CRRM QoS Based - 85%, .
In the case of high traffic load, k = 1, in figure 5.18, almost 65% of the voice calls
established in the hotspot are served by WLAN with the CRRM QoS Based - 50% : this is
firstly due to the high number of call attempts that cannot be handled by UMTS because
of Admission Control; secondly, the high UMTS network load reports trigger several
intersystem handover of voice call connections from UMTS to WLAN. The CRRM QoS
Figure 5.17: Distribution of voice call users in the two networks covering the hotspot
(k = 0.5)
Figure 5.18: Distribution of voice call users in the two networks covering the hotspot
(k = 1)
Figure 5.19: Ftp sessions in the hot spot: average perceived throughput
Based option at 85%, on the contrary, triggers less handovers, thus reducing the balancing
of radio resources between UMTS and WLAN and finally leading to a higher call block
and a worst service integrity and retainability.
In the last figure of this work - figure 5.19 - it is reported the main drawback introduced
by adopting a common radio resource management approach to improve the quality of
service for speech users in the hotspot, which was the main purpose of this study. As
could be expected, the average perceived throughput during FTP sessions in the hotspot
is reduced from the case in which the WLAN is dedicated to serve only data traffic (No
CRRM ) to the case of any CRRM configuration in which the WLAN channel is much
more busy to ”help” the UMTS to serve a larger amount of voice users.
However, we consider acceptable in the worst case scenario (k = 1) a reduction of
less than 25% for the average FTP throughput, that is a best effort service. As a matter
of fact, for k = 1, the benefit obtained thanks to the CRRM QoS Based is that the
Satisfaction Rate of voice users (see figure 5.12) is doubled.
Just for information, when No CRRM is used, the performance of WLAN is inde-
pendent from the voice traffic k in the hotspot; the small oscillations of average FTP
throughput in this configuration are due to the fact that each simulation uses seeds gen-
erated by different random sequences.
In this chapter we’ve presented the interactions between the Common Radio Resource
Management (CRRM) entity and the local RRM entities in UMTS and WLAN; differ-
ent CRRM algorithms have been proposed, and simulated with our advanced platform
(SHINE) in a realistic scenario of a hotspot of high density traffic covered by both UMTS
and WLAN radio access networks. In this final section of the numerical results we have
shown that a CRRM QoS Based algorithm can largely improve the performance of such
an integrated UMTS-WLAN network in terms of served voice users.
Actually, the aim of this study wasn’t to find the exact optimum, but it was to prove
that it is suitable to realize an integrated heterogeneous wireless network and that with
an appropriate CRRM algorithm combining radio interface measurements of both UMTS
and WLAN, the quality of service provided by the system can be clearly increased.
Conclusions
In this thesis, the problem of performance evaluation of third-generation and heteroge-

neous wireless networks is faced, focusing the attention on all main aspects of the radio
interface protocol layers, that is, from the physical to the application layer.
To this aim, a TD-SCDMA (Time-Division Synchronous Code Division Multiple Ac-
cess) system simulator has been realized (TD-SCDMA is one if the three UMTS standard
modes): the main functional blocks of this tool are the Link Level simulator, the Network
Level simulator, an Interface module between the link and the network levels and finally
the Upper Layer simulator, which receives the input from a Mobility simulator and an
User Activity simulator.
Thanks to an ”instant value” interface approach, the impact of the physical layer
parameters for the power control algorithm on the overall end-to-end user performance
has been analyzed: we showed that these parameters should be carefully chosen in order
to balance the quality of voice and data users in a third-generation mobile network.
Afterwards, the study of system performance has been extended to a wider and really
topical scenario, the wireless heterogeneous networks, in which many wireless radio access
technologies with complementary characteristics cover the same high traffic areas: the TD-
SCDMA and the WLAN IEEE802.11x simulators were used to reproduce this situation.
Let me highlight that at the beginning of my activity, three years ago, only few works
on the topic of UMTS-WLAN integration were available in the technical literature and
most of them were on signalling and architectural issues only. Now, a growing number of
107
108 Conclusions
research groups is focusing on this topic and international projects are active on it.
The main contribution of my work in this area has been the design of a Common Radio
Resource Management algorithm, which can be Coverage Based, Service Based and Qual-
ity of Service (QoS) Based. Thanks to an innovative simulation platform called SHINE,
the full dynamic simulation of a UMTS-WLAN wireless heterogeneous network has been
carried out: the proposed CRRM QoS Based algorithm appears to largely improve the
performance of such an integrated UMTS-WLAN network in terms of served users.
This important result shall not be considered only as the final arrival point of this
activity, but also as the demonstration that an intelligent tight interworking in hetero-
geneous networks allows to widely improve the system capacity and that further studies
focused on CRRM algorithms based on radio interface measurements can be a prolific
field of research.
Appendix A
SHINE: Simulation platform for

Heterogeneous Interworking
Networks
The timeliness of the issue of access-networks interworking is witnessed by the large

number of papers appeared in the open Literature in the recent past on this topic (see
for instance [27, 28, 29, 30]), especially with reference to the integration of UMTS and
wireless local area networks (WLANs).
Indeed, this issue is very challenging for researchers since investigations on heteroge-
neous interworking networks can hardly be carried out analytically, owing to the high
complexity of technologies involved as well as the large number of aspects to be simul-
taneously considered, besides those specifically related to the access technologies (traffic
characteristics, TCP protocol behavior, realistic channel modelling, ...).
It follows that the only feasible way to accurately investigate the performance of inter-
working access-networks, in order for instance to design effective common radio resource
management strategies, is to realize a software simulator able to capture all aspects of
access-networks behaviors, including, as is obvious, the networks interworking strategies.
The realization of such a simulator is still, however, a hard task and should be carried
109
110 SHINE: Simulation platform for Heterogeneous Interworking Networks
out having two important issues in mind: flexibility and time efficiency.
The former is mainly related to the reusability of most of the software independently
on the particular interworking technologies to be investigated, in such a way to make
possible to move, for instance, from an UMTS-WLAN simulator to a GPRS-WiMAX one
without rewriting almost the entire code. In few words, flexibility is related to a change of
perspective in the design of the simulation tool which should be conceived as a simulation
platform rather than a dedicated simulator.
As for the time efficiency, it is straightforward to understand that simulating the
behavior of two or more interworking access-networks, which obviously operate simulta-
neously, could be really prohibitive in terms of simulation time, unless parallel execution
is performed. The point is: can parallel execution be performed with commonly available
computers and without any experience on parallel programming?
In order to achieve both the above mentioned objectives, we developed the simula-
tion platform SHINE (Simulation platform for Heterogeneous Interworking NEtworks),
hereafter described.
A.1 Simulation platform Structure

SHINE was developed with the objective to reproduce the behavior of interworking access-
networks, taking care of all aspects related to every single protocol level of each access
technology affecting the achieved performance.
Its realization was planned in order to overcome the limitations of off-the-shelf network
simulation tools, such as, for instance, Opnet [1] or ns-2 [2], in terms of capability to
simulate the behavior of heterogeneous networks operating simultaneously and exchanging
users.
Hereafter the main criteria we adopted in order to design SHINE in a flexible and time
efficient manner are reported.
A.1.1 Flexibility.
To achieve the objective of flexibility we adopted a client-server structure for the simula-
tion platform, which is constituted, in particular, by one server-core simulator (hereafter
called Upper Layers Simulator, ULS) and a client simulator for each access technology
considered (Lower Layers Simulators, LLSs) (see figure A.1 where, for the sake of clarity,
only one LLS is depicted).
The ULS simulator is, in its turn, constituted by an access network(s) side and a Core
Network side: at the access network(s) side the ULS takes care of all information related
SHINE: Simulation platform for Heterogeneous Interworking Networks 111
Figure A.1: Simulator platform global architecture.
to those users operating within the region covered by the simulated access-networks, such
as their mobility, class of service, etc. and of the end-to-end aspects of each connection,
such as the generation of the application-level traffic and the users’ TCP or UDP dy-
namics; at the Core-Network side, instead, the ULS takes care of all aspects concerning
communications.
Focusing the attention on the access network(s) side, it is worth noting that the ULS
structure, being related to the end-to-end aspects of communications, is independent
on the particular access technology (UMTS, WLAN, ...) adopted to establish the user
connection.
All aspects related to the access technologies adopted, hence related to the data-link
and physical layers and the local Radio Resource Management, are managed by the LLSs,
which are the client simulators and are specific for each access technology, so that our
simulation platform provides the presence of a LLS for each radio technology adopted in
the investigated scenario (see figure A.2).
What is really remarkable is that ULS and LLSs are distinct executables; nonetheless
the ULS communicates run time with LLSs through the TCP sockets of the computer
operating-system, thus simulating both vertical communications among the protocol lay-
ers of the single access technology and, since ULS is on top of all LLSs and manages all
end-to-end communications aspects, the access-networks interworking.
Figure A.2: Simulation platform architecture. Access networks side.
Given this structure, any change in the access technologies investigated requires only
to write the related LLS simulators (that is, the data-link and physical level simulators
of each new technology including the corresponding RRM entity), which obviously have
to be provided with the standardized communication interface to interact with the ULS.
A.1.2 Time efficiency.

As previously stated, the ULS and LLSs are distinct executables which communicates one
with the others through the TCP sockets of the PC operating-system; this means that
ULS and LLSs can run independently on different personal computers, provided that they
are connected to the same computer network. In particular the simultaneous activities
of the interworking access-networks is reproduced by the simultaneous executions of the
related LLSs.
It is straightforward to understand that with this solution the objective of parallel
computing, and therefore of simulation time efficiency, has been easily achieved without
resorting to multiprocessor computers and to the paradigms of parallel programming.
A.2 ULS and LLSs main tasks

As previously stated, the ULS manages the end-to-end aspects of each connection (no
matter the access technology supporting them at the physical and data-link levels), hence
its tasks are mainly concerned with communications management (connections setup and
closure, management of application level traffic flows, ...), the simulation of transport
level protocols (TCP, UDP, ...) and the processing of simulation outcomes to provide
application level performance. In particular, the main tasks of ULS are:
• to set the starting instant of each new traffic session originated by users according
to the arrival statistics of the traffic class it belongs to (http, e-mail, voice calls, ...),
as well as users positions within the investigated scenario;
• to manage connection setup and closure procedures;
• to generate the bit-flows up(down)loaded by users in each session according to the

statistics of their class of traffic;
• to reproduce the transport protocol behavior;
• to perform packet segmentation and reassembly;
• to select through which technology (that is, through which LLS) should each user
connect to the network on the basis of user-defined rules and available information
on the current status of each network; it can also decide to reject a connection or
to move it from an LLS to another (that is, from a given technology to another) at
any time, thus simulating vertical (i.e. inter-system) handovers;
• to collect, finally, all simulation outcomes and to generate the outputs (user sat-
isfaction rate, throughput, packet delivery delays,...) from an end-to-end point of
view.
As for the LLSs, since they are specific for the particular access technologies inves-
tigated, their tasks are mainly concerned with data-link and physical level aspects of
communications and, in particular, are:
• to perform, if required, the call admission control specific of the technology it sim-
ulates and all technology specific radio resource management;
• to manage, if required, the transmission scheduling at the data-link layer level;
• to perform MAC or RLC fragmentation and reassembly of TCP-IP level packets;
• to simulate MAC/RLC behavior of the given technology;

• to reproduce all physical layer procedures related to each transmission and reception:
power control, handover, radio frequency measurements, channel coding, modula-
tion, information detection, decoding, etc.
• to simulate all main Radio Resource Management (RRM) functions (see section 5.3
for more details).
• to collect, finally, all simulation outcomes and to generate the outputs (user satis-
faction rate, throughput, packet delivery delays,...) from the wireless links point of
view (that is, at data-link and physical level).
It is important to underline that the above reported structure and tasks division
allows an easy management of vertical (inter-system) handovers, which have obviously
to be simulated when investigating interworking access-networks. Vertical handovers, in
fact, essentially determine a change of the access technology adopted by users; from the
simulation point of view they mainly implies to switch the management of users physical
and data-link levels from one LLS to another, and this can be easily managed by the ULS
which is on top of all LLSs.
A.3 Time axis management

In SHINE each LLS manages its own time axis; the ULS, for its part, communicates
to the LLSs the next instant in which some event concerning the upper protocol levels
happens (connection request, start of bit transfers, TCP timeouts, etc.) properly setting a
parameter called ULProxEvent, which represents a ”time-flag”. This way, when the time
counter of an LLS reaches that instant, the related LLS simulation stops and a ”call” to
the ULS is performed asking for the event-related information. After the ULS reply, the
simulation of the calling LLS is resumed, the consequent LLS actions (MAC level frame
queuing, ...) are taken, the time-flag ULProxEvent is updated and the LLS time counter
is restarted.
An LLS can perform a call to the ULS not only when the instant of the next ULS
event has come but also whenever an LLS event which is of interest for the ULS takes
place, such as, for instance, the correct transmission of a MAC frame.
This mechanism is illustrated in figure A.3 where both lower layers (A.3-b) and upper
layers events (A.3-c) are depicted.
The above described stop-and-wait procedure is the basis for the coordination among
LLSs, which is obviously needed when simulating interworking networks: in case an ULS
event is of interest for more than one LLS (for instance a new traffic session searching for
Figure A.3: ULS-LLS communications.
the best connection), no ULS reply is granted to the calling LLSs until all the interested
LLSs have stopped waiting for it.
It follows that although LLS executions take place at different speeds (LLSs complexity
could be different and they could be running on different PCs), the faster LLSs periodically
stop and wait for the LLSs they are interworking with, thus re-synchronizing simulations.
Please note that the depicted architecture leaves the LLS’ implementer free to decide
whereas an event driven or a time slotted simulator is the best choice for the specific
task. For instance, while a WLAN simulator is generally implemented as an event driven
simulator, a cellular system, characterized by a fixed radio frame structure, may be more
easily reproduced adopting a slotted time axis.
A.4 ULS-LLSs communications

As previously stated, ULS is the server simulator which always waits for LLSs calls (that
is, the clients calls), either because something happened at the lower layers (case b in figure
A.3), or because the LLSs current times reached ULProxEvent (the flag was reached in
case c of figure A.3).
All communications from LLS to ULSs and vice-versa can be classified adopting a
Figure A.4: ULS-LLSs communications scheme (cellular network notation) implemented

in SHINE
cellular network formalism, as shown in figure A.4. On the left side of the figure the
generic LLS is represented, whereas the ULS is depicted on the right side. From top to
bottom of the central part of the figure, we find the logical planes of the overall system:
from the call control/session management down to the physical layer, that is, from all
the Non Access Stratum functions to the Access Stratum ones. Thus, the arrows and
their directions clearly show all possible communications occurring between the ULS and
LLSs. Obviously, the arrows from left to right are related to lower layers events to be
communicated to the ULS, while the others are due to the flag reaching, hence due to
ULS events to be communicated to the LLSs.
Bibliography
[1] OPNET Technologies, Inc. http://www.opnet.com
[2] NS-2 Network Simulator. http://www.mash.cs.berkley.edu/ns/
[3] Glomosim Simulator. http://pcl.cs.ucla.edu/projects/glomosim/
[4] M.Wellens, M.Petrova, J.Riihijrvi, P.Mhnen, “Building a Better Wireless Mousetrap:

Need for More Realism in Simulations”, Proc. of WONS 2005, St. Moritz, Switzer-
land, January 2005.
[5] R.J.Punnoose, P.V.Nikitin, D.D.Stancil, “Efficient Simulation of Ricean Fading

within a Packet Simulator”, Proc. IEEE 2000 Fall Vehic. Tech. Conf., pp. 764-767.
[6] J.Heidemann et al., “Effects of Detail in Wireless Network Simulation” SCS Multi-
conference on Distributed Simulation, Phoenix, Arizona, USA January, 2001.
[7] D.Cavin, Y.Sasson, A.Schiper, “On the accuracy of MANET Simulators”, POMC’02,
October 30-31, 2002, Toulouse, France.
[8] M.Takai, J.Martin, R.Bagrodia, “Effects of Wireless Physical Layer Modeling in Mo-
bile Ad Hoc Networks”, Proc. MobiHoc 2001.
[9] G.Yeung et al, “Detailed OFDM modeling in network simulation of mobile ad hoc
networks ”, 18th Workshop on Parallel and Distributed Simulation, 2004. PADS
2004. May 2004.
[10] IEEE Std. 802.11a, “High-Speed Physical Layer in the 5 GHz Band”, 2000.
[11] J. Medbo, H. Andersson, P. Schramm, H. Asplund, J-E. Berg, “Channel Models for
HIPERLAN/2 in different indoor scenarios,1998-04”.
[12] W.C. Jakes, “Microwave Mobile Communications”. Piscataway, NJ. IEEE Press,
1994.
[13] IEEE Std. 802.11, “Wireless LAN Medium Access Control (MAC) And Physical
Layer (PHY) Specifications”, 1999.
119
120 Bibliography
[14] K. Kammerlander, ”Benefits and implementation of TD-SCDMA,” International

Conference on Communication Technology, WCC-ICCT 2000., pp. 1013-1016, 21-
25 Aug. 2000.
[15] J. Zhang, W. Wang, ”Dynamic system level simulation for radio resource manage-
ment of TD-SCDMA high speed downlink packet access,” International Conference
on Communications, Circuits and Systems and West Sino Expositions, pp. 187-191,
29 June-1 July 2002.
[16] Y. Zhang, Y. Chang, D Yang, ”Link performance of TD-SCDMA systems,” Spread

Spectrum Techniques and Applications, 2002 IEEE Seventh International Sympo-
sium on , pp.14-18, 2002.
[17] S. Hmlinen, P. Slanina, M. Hartman, A. Lappetelinen and H. Holma, ”A Novel

Interface Between Link and System Level Simulations”, ACTS Mobile Telecommu-
nications Summit’97, Aalborg, Denmark, in proceedings pp. 599-604, October 7-10,
1997.
[18] ETSI TS 123 107 v 5.10.0, ”Universal Mobile Telecommunication System (UMTS);
Quality of Service (QoS) concept and architecture”, septembre 2003
[19] P. Slanina, F. Raji, C. Mecklenbräuker, C. Hölzl, G. Ostermayer, T. Stadler, T.

Gruhn, ”Performance of hybrid ARQ types for TDD mode”, Proc. ACTS 4th Mobile
Communications Summit, 1999
[20] V.V. Veeravilli, A. Sendonaris, ”The coverage-capacity tradeoff in cellular CDMA

systems”, IEEE Transactions on Vehicular Technology, Vol. 48, No.5, September
1999, pp. 1443-1450
[21] H. Holma, A. Toskala, ”WCDMA for UMTS Radio Access for Third Generation
Mobile Communications”, WILEY, September 2000.
[22] A. Mehrotra, ”GSM System Engineering”, Artech House, 1997.
[23] E. Malkamaki, F. de Ryck, C. Mourot, ”A Method for Combining Radio Link Sim-
ulations and System Simulations for a Slow Frequency Hopped Cellular System”,
IEEE Veh. Technol. Conf. 1994 (VTC’94), pp. 1145-9, Phoenix, AZ, May 1994.
[24] H. Olofsson, M. Almgren, C. Johansson, M. Hook, F. Kronestedt, ”Improved Inter-

face Between Link Level and System Level Simulations Applied to GSM”, IEEE Int.
Conf. On Univ. Pers. Comm. (ICUPC’97), Page(s): 79 -83, vol.1, S. Diego, CA, Oct.
1997.
[25] Z. Kostic, X. Qiu, J. Chuang, L.F. Chang, K. Chawla, N. Sollenberger, “Dynamic

Frequency Hopping in Cellular Systems with Network Assisted Resource Allocation”,
IEEE VTC’2000spring, pp. 2459-2463, Tokyo, May, 2000.
Bibliography 121
[26] ITU-T Recommendation E.800, “Terms and definitions related to quality of service
and network performance including dependability”. Augusto 1994.
[27] Kaveh Pahlavan et al., “Handover in Hybrid Mobile Data Networks”, IEEE Personal
Communications, April 2000.
[28] Shiao-Li Tsao, Chia-Ching Lin, “Design and Evaluation of UMTS-WLAN Interwork-
ing Strategies”, 2002 Computer & Communications Research Laboratories, Industrial
Technology Research Institute, Hsinchu, Taiwan, R.O.C.
[29] M. Lott et al., “Interworking of WLAN and 3G systems”, IEE Proceedings-

Communications Volume 151, Issue 5, 24 Oct. 2004 Page(s):507 - 513.
[30] Hongyang Bing,Chen He,Lingge Jiang, “Performance analysis of vertical handover in

a UMTS-WLAN integrated network”, IEEE PIMRC 2003.
[31] IEEE 802.11 WG, IEEE 802.11e/D5.0, “Draft Supplement to Standard for Telecom-
munications and Information Exchange Between Systems LAN/MAN Specific Re-
quirements Part 11: Wireless Medium Access Control (MAC) and Physical Layer
(PHY) Specifications: Medium Access Control (MAC) Enhancements for Quality of
Service (QoS), Aug. 2003”.
[32] Ad Kamerman and Leo Monteban, “WaveLANr - II: A High-Performance Wireless

LAN for the Unlicensed Band”, Bell Labs Technical Journal, Summer 1997.
[33] D. Gu, J. Zhang, “A new measurement-based admission control method for

IEEE802.11 wireless local area networks” Personal, Indoor and Mobile Radio Com-
munications, 2003. PIMRC 2003. 14th IEEE Proceedings on , Volume: 3 , 7-10 Sept.
2003 Pages:2009 - 2013 vol.3
[34] IEEE 802.20 Working group on Mobile Broadand Wireless Access, “Traf-
fic Model for IEEE 802.20 MBWA System Simulations (Baseline Draft)”,
N.K.Shankaranarayanan, July 2003.
[35] http://www.umatechnology.org
[36] Http://www.ist-mobydick.org
[37] Http://www.Monasidre.com
[38] N.Vulic, I.Niemegeers, S.H.de Groot, “Architectural options for the WLAN integra-
tion at the UMTS radio access level”, IEEE VTC 2004-Spring.
[39] Kim Sung-Eun, J.A.Copeland, “Interworking between WLANs and 3G networks :

TCP challenges” IEEE WCNC 2004.
[40] M.Bernaschi, F.Cacace, A.Pescape, S.Za, “Analysis and experimentation over het-
erogeneous wireless networks”, Tridentcom 2005.
122 Bibliography
[41] J.Indulska, S.Balasubramaniam, “Context-aware vertical handovers between WLAN

and 3G networks”, IEEE VTC 2004-Spring.
[42] http://www.ist-scout.org
[43] http://www.freeband.nl/kennisimpuls/projecten/4gplus
[44] G.F.Riley et al., “Distributed network simulations using the dynamic simulation
backplane” 21st International Conference on Distributed Computing Systems, 16-19
April 2001.
[45] “ITU-Reccomendation H.323: Packet-based multimedia communications systems”,

International Telecommunication Union, 1997.
[46] Bur Goode, “Voice Over Internet Protocol (VoIP)”, Proceedings of the IEEE, vol.90,
No.9, sept. 2002.
[47] 3GPP, TS 25 series.
[48] 3GPP TS 25.221 V5.2.0, ”Physical channels and mapping of transport channels onto
physical channels (TDD)”
[49] 3GPP TS 25.222 V5.7.0, ”Multiplexing and channel coding (TDD)”
[50] 3GPP TS 25.223 V5.3.0, ”Spreading and modulation (TDD)”
[51] 3GPP TS 25.331 V5.8.0, ”Radio Resource Control (RRC); Protocol specification”
[52] 3GPP TS 25.102 V5.5.0, appendix A, ”UE Radio Transmission and Reception
(TDD)”, june 2003
[53] 3GPP TS 25.105 V5.4.0, appendix A, ”Base Station (BS) radio transmission and
reception (TDD)”, april 2003
[54] ITU-R Recommendation M.1225, “Guidelines for Evaluation of Radio Transmission

Technologies for IMT-2000”. 1997.
[55] 3GPP TR 101 112 V3.2.0, “Technical Report Universal Mobile Telecommunications
System (UMTS); Selection procedures for the choice of radio transmission technolo-
gies of the UMTS”. 1998-04.
[56] CWTS 3G digital cellular telecommunications system TSM , ”Multiplexing and mul-
tiple access on the radio path”, 05-02, Dec. 2001., Dec. 2001.
[57] ”cdma2000 Evaluation Methodology”, 3GPP2 C.R1002-0 v1.0, December 10 2004.
[58] 3GPP TS 24.008 V5.16.0, ”Mobile radio interface Layer 3 specification”, June 2006
[59] 3GPP TS 25.413 V5.12.0, ”UTRAN Iu interface RANAP signalling”, June 2005
Bibliography 123
[60] 3GPP TS 43.318 V7.0.0, ”Generic access to the A/Gb interface”
[61] 3GPP TR 22.934 V6.0.0, “Feasibility study on 3GPP system to Wireless Local Area
Network (WLAN) interworking (Release 6), 2002”
[62] 3GPP TS 23.107. Technical Specification Group Services and System Aspects; QoS
Concept and Architecture
Acknowledgments
It is impossible to acknowledge every one who had contributed in some way to the results
of my doctoral course. Nonetheless, to the best of my ability I shall attempt to do so.
I would thus thank Professor Oreste Andrisano, who always supported and trusted
my job, also in the periods out of Bologna. Besides these three years of studies, I’ve
undertaken a long journey with him, which started on the summer of 2002 when I joined
for the first time the IEIIT-CNR institute for my degree thesis. Working in his research
team has been a fantastic opportunity.
Special thanks are dedicated to Gianni Pasolini and Alessandro Bazzi. I’m grateful to
have worked with them, they have been really my guidance in this period, and I’m only
sorry not being able to spend much more time with them in Bologna; thanks for your
help and encouragements to finalize this thesis!
I’m happy to convey my appreciation to the other researchers and professors of the
University of Bologna, Alberto Zanella (thanks to let me go to Las Vegas!), Barbara
Masini, Davide Dardari, Andrea Conti and Roberto Verdone.
Great thanks to Lorenzo Faggioni from Siemens Mobile. We have enjoyed together
many epic battles in the UMTS test lines; he has given me a lot of freedom to deepen
my research interests, he has transmitted to me the passion for teamwork, but most of all
he’s a great friend.
I wish to thanks also my unforgettable colleagues (friends) in Siemens Milano, Giovanni
De Pascalis, Marco Guidone, Manuel Piras, Simone Buzzi, Alberto Fontana, Andrea Sala
and Arnó Valerio: how much fun we had in these years! And how many discussions about
football, potatoes and UTRAN faults.
A special thank to Andrea Brambilla from Siemens. We left Milan together and now
we share an amazing apartment in Dusseldorf and an exceptional period of our life; I’m
very happy of that!
I would like also to thanks Augusto Tomas and Doctor René Jahnkow from Siemens
Dusseldorf. We experienced how to setup a full UMTS network alone in the desert of
Saudi Arabia, unrepeatable adventure!
125
126 Acknowledgments
I wish to thanks Gianluigi Liva for his constant support, also when things weren’t
going in the best way; he’s another Italian Ph.D. who immigrated in Germany and I
share with him the idea to go back to Romagna, asymptotically.
Thanks to Mauro Gibertoni, for his friendship for a life.
I want finally to express unique thanks to my family. My parents and my two sisters,
Laura and Gabriella, supported and always encouraged me during the long period of my
studies. By dedicating them this thesis, I know that I’m just giving back a little bit of
what they did (and they are still doing) for me. Nevertheless, they know how much I love
them.
13th March 2007

Heterrogen Network

Uploaded by

Copyright:

Available Formats

Heterrogen Network

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Heterrogen Network

Uploaded by

Copyright:

Available Formats

Università degli Studi di Bologna

Dipartimento di Elettronica Informatica e Sistemistica, DEIS

Dottorato di Ricerca in Ingegneria Elettronica,

Wireless Heterogeneous Networks

1 TD-SCDMA System Simulator design 9

2 Performance of TD-SCDMA in mixed CS/PS traffic scenarios 31

3 Link Level aspects modelling in the simulation of packet switched wire-

3.3.2 Interface Module to Network Level communication . . . . . . . . . 54

4 Architectures for Heterogeneous Wireless Networks 61

5 Common Radio Resource Management UMTS & WLAN 71

A SHINE: Simulation platform for Heterogeneous Interworking Networks109

A.1.2 Time efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

2.1 System parameters fixed in the numerical results . . . . . . . . . . . . . . . 36

5.1 Initial-RAT selection algorithm in the hotspot. . . . . . . . . . . . . . . . . 80

1 Scenarios for wireless heterogeneous networks . . . . . . . . . . . . . . . . 4

1.1 TD-SCDMA physical channel signal format . . . . . . . . . . . . . . . . . 12

2.1 The considered network scenario . . . . . . . . . . . . . . . . . . . . . . . . 35

3.1 Link and Network levels in operation . . . . . . . . . . . . . . . . . . . . . 49

4.1 Loose coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.1 RRM & CRRM relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

A.1 Simulator platform global architecture. . . . . . . . . . . . . . . . . . . . . 111

A.3 ULS-LLS communications. . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

Mobile networks evolution

Wireless heterogeneous networks

Figure 1: Scenarios for wireless heterogeneous networks

TD-SCDMA System Simulator

are neglected [4, 5, 6, 7].

1.1 TD-SCDMA air interface

Figure 1.1: TD-SCDMA physical channel signal format

• TDMA (Time Division Multiple Access) is a digital technique that divides

Figure 1.2: Structure of the TD-SCDMA sub-frame

who may require multiple time slots.

• Dynamic Channel Allocation (DCA): the advanced TD-SCDMA air interface

• Mutual Terminal Synchronization: like all TDMA systems, TD-SCDMA needs

1.2 TD-SCDMA System Simulator

Figure 1.3: TD-SCDMA System Simulator: block diagram

1.2.1 Link Level simulator

Figure 1.4: TD-SCDMA Link Level simulator: block diagram

• The block ”S” is the source of the information to be transmitted.

• The QPSK modulator generates the QPSK symbols.

(OVSF) code [50].

1.2.2 Link-to-Network level Interface module

Figure 1.5: Link-to-Network level Interface module: block diagram

As already explained at the beginning of this chapter, a thorough performance inves-

• Afterwards, during the dynamic simulation, the interface module communicates

1.2.3 Network Level simulator

Figure 1.6: Main functional blocks of the TD-SCDMA simulator

can be studied in a very detailed manner.

• Channel propagation model: Pathloss, Walfish Ikegami model: K1 + K2 · log10 (d)

• Antennas: either omnidirectional or sectorized antennas with arbitrary (2D) an-

• RRM - Dynamic channel allocation (DCA): based on code-timeslot-frequency alloca-

• RRM - Handover Control: in TD-SCDMA, only hard-handover is defined. In the

• RRM - intersystem Handover: due to the Common Radio Resource Management

• Both conversational (speech and video) and interactive/background services are

(number of timeslots, number of codes used in downlink and uplink), number of

• System efficiency: in TD-SCDMA network, it’s essential to exploit the oscillating

• Coverage, capacity and handover maps.

• Radio statistics: the TD-SCDMA network simulator provides statistics of transmit

1.2.4 Upper Layer simulator