Aspects of WINNER+ Spectrum Preferences - Celtic-Plus

Project Number: 

Project Title: 

Document Type: 

CELTIC / CP5-026 

Wireless World Initiative New Radio – WINNER+ 

I (Internal) 

Document Identifier: D3.2 

Document Title: 

Source Activity: 

Editor: 

Aspects of WINNER+ Spectrum Preferences 

WP3 

Matthias Siebert 

Authors: 

Anne-Lyse Bouaziz, Jean-Philippe Desbat, Jean-Marc Kelif, Horst 

Mennenga, Albena Mihovska, Werner Mohr, Miia Mustonen, 

Matthias Siebert 

Status / Version: 1.0 

Date Last changes: 26.05.09 

File Name: 

D3.2_WINNER+SpectrumPreferences_v1.0.doc 

Abstract: 

This deliverable covers selected items on WINNER+ spectrum 

preferences. A large part of this work is dedicated to spectrum 

aggregation issues covering scheduling aspects, transceiver 

concepts and an analysis of deployment scenarios as discussed 

within ITU and 3GPP. Further, Femto-Macro coexistence is 

investigated as an exemplary Intra-Operator sharing scenario, 

complemented by some multi-operator resource sharing 

considerations. 

Keywords: Spectrum Aggregation, Spectrum Sharing, Femto-Macro 

Coexistence, Multi-Operator Spectrum Usage, Scheduling, 

Spectrum Usage Scenarios, Transceiver Concepts 

Page 1 (70)

WINNER+ D3.2 

Version: 1.0 Page 2 (70)


Table of Contents 

1. Introduction............................................................................................ 5 

2. Spectrum Regulation and Demand ...................................................... 5 

2.1 Worldwide Standardisation & Regulation .............................................................................. 5 

2.2 European Standardisation & Regulation................................................................................. 7 

2.3 Spectrum Estimation Methodologies ...................................................................................... 8 

2.4 Spectrum Requirement Estimation ......................................................................................... 8 

3. Spectrum Sharing and Co-Jointly Usage ............................................ 9 

3.1 Intra-Operator Femto-Macro coexistence ............................................................................... 9 

3.1.1 Scenario 1: Interference between femtocells ................................................................. 11 

3.1.1.1 Assumptions.......................................................................................................... 11 

3.1.1.2 Interference between femtocells: SINR results ..................................................... 12 

3.1.1.3 Interference between femtocells: Throughput results............................................ 16 

3.1.1.4 Scenario 1 conclusion............................................................................................ 18 

3.1.2 Scenario 2: Interference between femto and macrocells................................................ 20 

3.1.2.1 Assumptions.......................................................................................................... 20 

3.1.2.2 Interference between femto and macro cells: SINR results................................... 22 

3.1.2.3 Interference between femto cells: Throughput results........................................... 24 


3.1.3 Scenario 3: Interference between femto and macrocells................................................ 28 

3.1.3.1 Assumptions.......................................................................................................... 28 

3.1.3.2 Interference between femto and macro cells: SINR results................................... 30 

3.1.3.3 Interference between femtocells: Throughput results............................................ 31 


3.2 Multi-Operator Resource Sharing Scenario in the Context of IMT-A Systems.................... 32 

3.2.1 Spectrum Aggregation Requirements ............................................................................ 33 

3.2.2 Hypotheses for System and Spectrum Resource Usage................................................. 34 

3.2.3 Intra-and Inter-Operator Sharing Aspects...................................................................... 34 

3.2.3.1 Intra-Operator Sharing Considerations.................................................................. 34 

3.2.3.2 Inter-Operator Considerations............................................................................... 36 

4. Spectrum Aggregation ........................................................................ 37 

4.1 Spectrum Aggregation with Multi-Band User Allocation over Two Frequency Bands........ 38 

4.1.1 Problem Statement......................................................................................................... 38 

4.1.2 General Multi-band Scheduling Spectrum Sharing as a General Assignment Problem 39 

4.1.2.1 Profit Function....................................................................................................... 39 

4.1.2.2 Resource Allocation .............................................................................................. 40 

4.1.2.3 Differences between the 2-GHz and 5-GHz frequency bands............................... 41 

4.1.3 Suboptimal Multiband Allocation Algorithm ................................................................ 41 

4.1.4 Results ........................................................................................................................... 41 

4.2 Basic Transceiver Concepts for aggregated Spectrum.......................................................... 41 

4.2.1 Approach........................................................................................................................41 

4.2.2 Brief Investigation of Several Topics Affecting the System Design ............................. 44 

4.2.2.1 Frequency Dependent Path Loss ........................................................................... 44 

4.2.2.2 Doppler Frequency and Doppler Spectrum........................................................... 45 



4.2.2.3 Effective Noise Power........................................................................................... 46 

4.2.2.4 Receiver Input Signal ............................................................................................ 46 

4.2.2.5 Nonlinearities in the Analogue Receiver Components.......................................... 47 

4.2.2.6 Image Rejection..................................................................................................... 47 

4.2.2.7 Receiver Performance (Desensitization, Blocking, Intermodulation) ................... 49 

4.2.2.8 Reciprocal Mixing and Receiver Noise Figure ..................................................... 49 

4.2.2.9 Band Pass Filters and Filter Slope for Stop-band Attenuation .............................. 51 

4.2.2.10 Maximum Input Signal........................................................................................ 53 

4.2.2.11 Sampling Theorem for Both Receiver Versions.................................................. 54 

4.2.2.12 A/D-Converter Dynamic Range and Output Data Rate ...................................... 54 

4.2.2.13 A/D-Converter Performance Development......................................................... 55 

4.2.3 Conclusions.................................................................................................................... 56 

4.3 Actual discussions in 3GPP and ITU-R on the implementation of identified frequency 

spectrum in WRC 2007 ....................................................................................................................... 57 

4.3.1 Frequency arrangements in the ITU-R WP5D meeting in Geneva, Switzerland, from 

February 10 – 17, 2009 ................................................................................................................ 57 

4.3.2 Status of discussion on the implementation of identified additional IMT frequency 

bands in the 3GPP TSG-RAN meting in Athens, Greece, in February 3 – 13, 2009 ................... 59 

4.3.3 Conclusions.................................................................................................................... 68 

5. References ........................................................................................... 69 



1. Introduction 

This deliverable covers spectrum aspects as considered within WINNER+. 

Starting with a summary on worldwide, European and national regulation activities and fora in 

Section 2, sharing options for single- and multi-operator environments are discussed in 

Section 3. A large part of this deliverable finally is dedicated to spectrum aggregation as 

addressed within Section 4. 

2. Spectrum Regulation and Demand 

This section gives an overview of national as well as international groups and authorities being concerned 

with spectrum. The presentation is chosen to be top down as illustrated in Figure 2-1, starting with 

worldwide fora followed by associations on a European level. 

Level of 

participation 

Standardization organizations / results 

Regulatory 

Organizations 

Results 

World 

-wide 

ISO/IEC 

ITU 

Recommendation 

International Standard 

reference 

Conferences 

ITU-Plenipotentiary 

ITU-Administrative 

Voluntary 

Standard 

International 

convention 

Regulations 

ETSI 

EN 

TCAM 

Europe 

CEN/CENELEC 

EN/ENV 

reference 

reference 

European 

Council 

European 

Commission 

Harmonised 

Norm 

Directive 

Decisions 

National 

(Germany) 

DIN/DKE 

TBETSI 

DIN Standard 

reference 

Federal 

Government 

(BMWi / BNetzA) 

Acts 

Ordinances 

Ordinances 

Regulations 

Figure 2-1: Relation between the fields of voluntary standardization and regulatory activities 

2.1 Worldwide Standardisation & Regulation 

There are three organisations which take care about worldwide standardisation, see also Figure 2-1: 

- ISO: International Organization for Standardization (General) 

- IEC: International Electrotechnical Commission 

- ITU: International Telecommunication Union 

ITU is the leading United Nations agency for information and communication technology issues, and the 

global focal point for governments and the private sector in developing networks and services. For nearly 

145 years, ITU has coordinated the shared global use of the radio spectrum, promoted international 

cooperation in assigning satellite orbits, worked to improve telecommunication infrastructure in the 

developing world, established the worldwide standards that foster seamless interconnection of a vast 



range of communications systems and addressed the global challenges of our times, such as mitigating 

climate change and strengthening cyber security. 

ITU is committed to connecting the world. 

– Membership: 191 Member countries and over 700 companies and organisations; 

– Provides the international regulatory framework within which the various 

communication services can be operated 

The ITU has three Sectors: 

– Telecommunication Development (ITU-D) 

– Telecommunication Standardization (ITU-T) 

– Radiocommunication (ITU-R), 

The main task of International Telecommunication Union Radiocommunication sector (ITU-R) is to 

manage international radio-frequency spectrum and satellite orbit resources. ITU-R develops standards 

for radiocommunication systems to ensure the effective use of the radio-frequency spectrum and carries 

out studies concerning the development of radiocommunication systems. Furthermore, ITU-R carries out 

studies for the development of radiocommunication systems used in disaster mitigation and relief 

operations. ITU-R consists of five Study Groups (SG) two of which are of main interest to the Winner+ 

project and are briefly described below. 

Study Group 1 of ITU-R deals with issues related to spectrum management. Study group 1 is responsible 

for the overall development of principles and techniques for spectrum management and long-term 

strategies for effective spectrum use. SG 1 is divided into three Working Parties (WPs). Working Party 

1A (WP 1A) is responsible of spectrum engineering techniques, WP 1B of spectrum management 

methodologies and economic strategies, and WP 1C of spectrum monitoring. WP 1B is responsible of the 

World Radio Conference 2011 (WRC-11) agenda item 1.19 concerning software-defined radio and 

cognitive radio systems (an overview about the WRC agenda items is provided later in this section). 

Study Group 5 is called terrestrial services and it deals with systems and networks for fixed, mobile, radio 

determination, amateur and amateur-satellite services. The responsibility of WP 5A are land mobile 

services excluding International Mobile Telecommunications (IMT); amateur and amateur-satellite 

service WP 5A is responsible of providing technical input to WP 1B concerning the WRC-11 agenda item 

on software-defined radio and cognitive radio systems. The main focus of the Winner+ project and its 

predecessor EU-projects WINNER I and II is WP 5D (before WRC-07 WP 8F), which is concentrated on 

IMT systems. WP 5D is at the moment in the process of evaluating IMT-Advanced proposals. WP 5D is 

also responsible for developing band plans for frequency bands identified for IMT in WRC-07. 

The ITU WRC, which convenes every three to four years, is at the core of the international spectrum 

management process and constitutes the starting point for national practices. WRC reviews and revises 

the Radio Regulations, an international treaty establishing the framework for the utilization of radio 

frequencies and satellite orbits among ITU member countries, and considers any question of a worldwide 

character within its competence and related to its agenda. 

WRC-07 

The WRC-07 has identified the following new spectrum bands for IMT systems, some of which Region 1 

specific: 

• 450-470 MHz globally, 

• 698-806 MHz in Region 2 and nine countries in Region 3, 

• 790-862 MHz in Region 3 and part of Region 1 countries, 

• 2.3-2.4 GHz globally, 

• 3.4-3.6 GHz in a large number of countries in Regions 1 and 3. 

1 Region 1: Europe, the Middle East and Africa (EMEA) 

Region 2: North- and South America (Americas) 

Region 3: Asia Pacific (APAC) 



The intention is that the bands previously identified in the Radio Regulations for IMT-2000 are now 

identified for IMT. The WRC’07 allocations are used as the starting point in the considerations of 

preferred spectrum use in WINNER+. 

A considerable amount of work on preferred spectrum use has been done in WINNER II. In [WIN2D591] 

TDD / FDD inter-mode operation is investigated, whereas in [WIN2D592] and [WIN2D593] scenarios 

for wide area, metropolitan, and local area are considered. 

WRC-11 

There are three agenda items (AI) in WRC-11 that could be considered the most interesting ones from 

Winner+ perspective: 

• AI 1.17 to consider results of sharing studies between the mobile service and other services in 

the band 790-862 MHz in Regions 1 and 3, in accordance with Resolution 749 [COM4/13] 

(WRC-07), to ensure the adequate protection of services to which this frequency band is 

allocated, and take appropriate action; 

• AI 1.19 to consider regulatory measures and their relevance, in order to enable the introduction 

of software-defined radio and cognitive radio systems, based on the results of ITU-R studies, in 

accordance with Resolution 956 [COM6/18] (WRC-07); 

• AI 1.25 to consider possible additional allocations to the mobile-satellite service, in accordance 

with Resolution 231 [COM6/21] (WRC-07); 

The proposals for action items to the WRC-15 can be made via the following action item: 

• AI 6 to identify those items requiring urgent action by the Radiocommunication Study 

Groups in preparation for the next world radiocommunication conference 

2.2 European Standardisation & Regulation 

There are three organisations which take care about European standardisation: 

- CEN: European Committee for Standardization (General) 

- CENLEC: European Committee for Electrotechnical Standardization 

- ETSI: European Telecommunications Standards Institute 

The European Telecommunications Standards Institute (ETSI) is an independent, non-profit, 

standardization organization in the telecommunications industry (equipment makers and network 

operators) in Europe, with worldwide projection. ETSI has been successful in standardizing the GSM cell 

phone system and the TETRA professional mobile radio system. 

ETSI is developing telecommunication standards (EN) for Europe. The part of the EN which contains the 

regulatory requirements may be published in the official journal of the European Commission, after that 

the EN will be known as Harmonized Standard (HN). 

For the frequency related requirements ETSI works together with the European Conference of Postal and 

Telecommunications Administrations (CEPT). CEPT and ETSI have created a Memorandum of 

Understanding (MoU) to solve this task. 

CEPT: Conference of 48 European Post and Telecommunication Administrations 

CEPT is a body of policy-makers and regulators for: 

• establishing a European forum for discussions on sovereign and regulatory issues in the 

field of post and telecommunications issues 

• influencing, through common positions, developments within ITU and UPU (Universal 

Postal Union) in accordance with European goals 



Table 2-1: Some national regulators of European countries 

National Spectrum Authority Body (Regulator) 

Denmark 

National IT and Telecom Agency 

France ANFR Agence Nationale des Fréquences 

Germany BNetzA Bundesnetzagentur für Elektrizität, Gas, Telekommunikation, 

Post und Eisenbahnen (The Federal Network Agency for 

Electricity, Gas, Telecommunications, Post and Railway) 

Netherlands 

Agentschap Telecom 

United Kingdom Ofcom Office of Communications 

ECC: Electronic Communications Committee 

The ECC is the main group of the CEPT. Its tasks are related to: 

• Development of European radiocommunications policies 

• Harmonisation of the frequency usages within the CEPT area 

• Coordination of frequency management aspects and regulation matters 

• Providing of guidelines regarding cooperation with the ITU 

• ECC PT1 is making the European preparation for ITU-R WP 5D 

ECC PT1 is responsible for the overall system aspects of IMT. Currently the work in ECC PT1 includes 

developing appropriate ECC Deliverables on the designation and frequency arrangements for the 

spectrum bands 3.4-3.6 GHz and 790-862 MHz. The sharing and compatibility issues on all these bands 

will be considered and reported in ECC Recommendations and Reports. ECC PT1 is also going to address 

the band 3.6-3.8 GHz for IMT. Additionally, ECC PT1 will co-ordinate positions on the ITU process for 

the development of IMT-Advanced as well as the compatibility studies and frequency arrangements for 

other bands identified for IMT in WRC-07. 

EU Regulations 

The European Commission had set up a Group (Telecommunications Conformity Assessment and Market 

Surveillance Committee, TCAM) which handles the requirements concerning the R&TTE Directive. 

Spectrum related activities were handed in the Radio Spectrum Committee (RSC) and Radio Spectrum 

Policy Group (RSPG) 

2.3 Spectrum Estimation Methodologies 

The methodology for calculating the spectrum requirements for IMT has been developed in ITU-R WP 

8F, and most parts of the methodology originate from the EU-project WINNER. The methodology 

accommodates a complex mixture of services from market studies with service categories having 

different traffic volumes and QoS constraints. The methodology is to apply a technology neutral approach 

to handle emerging as well as established systems. A detailed description of the methodology is provided 

in [WIND6.2] and [WIND6.5]. A summary of the methodology with the main steps and parameters can 

be found in [WIN2D5.10.2]. This methodology is implemented into a tool (SPECULATOR) calculating 

the predicted spectrum needs for 2010, 2015 and 2020. The spectrum calculation tool has been developed 

in the WINNER project and it has been also adapted as an official ITU-R tool. A detailed description of 

the tool can be found in [WIND6.5] and a brief guide to the usage of the tool can be found in 

[WIN2D5.10.2]. 

2.4 Spectrum Requirement Estimation 

Based on the methodology, an estimation of the additional spectrum needed for IMT by the year 2020 has 

been calculated to be 1280 - 1720 MHz for lower and higher market setting, respectively [WIN2D5.10.2]. 

Including the spectrum identified in WRC-07 the total amount of spectrum for IMT is still left far below 

the spectrum prediction for lower market setting in all of the ITU regions as illustrated in Table 2.1. This 

indicates that additional spectrum would be needed in the future to provide the predicted services. 



However, there is not an agenda item for WRC-11 identification of additional spectrum for IMT and 

therefore at earliest additional spectrum can be allocated for IMT in WRC-15. 

Identified 

before WRC- 

07 [MHz] 

693 whole region 

120, partially 

392 

Table 2-2: Amount of spectrum in different ITU Regions 

Region 1 Region 2 Region 3 

Identified in 

WRC-07 

[MHz] 

Total 

[MHz] 

Identified 

before 

WRC-07 

Identified in 

WRC-07 

[MHz] 

Total 

[MHz] 

Identified 

before 

WRC-07 

Identified in 

WRC-07 

[MHz] 

whole 

region 813, 

partially 

1085 

[MHz] 

723 whole region 

228 

whole 

region 951 

Total 

[MHz] 

[MHz] 

749 whole region whole 

192, partially region 941, 

484 partially 

1233 

3. Spectrum Sharing and Co-Jointly Usage 

Mobile network operators are now preparing for the next generation of mobile communication systems. 

This is a transformation of cellular telephone networks into ubiquitous wireless broadband, which will be 

based on a new technology, called International Mobile Telecommunications Advanced (IMT-A) [ITU- 

CL 2008]. Affordable, high-bandwidth mobile access improves the quality of experience for users, 

enabling them to get more out of existing services, and opens up opportunities for new mobile broadband 

services. By delivering more capacity through faster radio access technologies (RATs), a universally 

accessible broadband infrastructure will improve the value of these services. The standardization of 

IMT-A technologies for preferred spectrum use in frequency bands announced by WRC-07 is, however, 

still a challenge to overcome [WRC’07]. The existing highly fragmented radio frequency spectrum does 

not match the actual demand for transmission and network resources, and there is a need for a revolution 

on how regulators manage spectrum. Mobile devices that implement various RAT standards (i.e., 

multiband devices) increase the complexity of coordination, reuse of resources and interference 

management. 

Mobile operators might be forced to aggregate spectrum of two or more separated sub-bands for downlink 

(DL) and uplink (UL) bands. 

Spectrum aggregation can be performed in the same or in different bands. It can appear when the 

operator’s dedicated DL or UL band is not contiguous but split into two or more parts. In addition, 

spectrum aggregation can happen in scenarios, in which an operator accesses both dedicated band and a 

spectrum sharing band, which is separated in frequency from the dedicated operator’s band. Spectrum 

aggregation has been proposed for Long Term Evolution-Advanced (LTE-A) [3GPP]. This process is 

investigated here also in relation to other IMT-A candidate systems, such as the WINNER system. The 

provided results are based on the preliminary investigations reported in [WIN+D1.2], [WIN+D3.1]. 

A mobile transmission system’s ability to support a wide range of services lies across all elements of the 

network (i.e., core, distribution and access), and across all layers of the OSI model [WIN+D3.1]. This 

becomes challenging in a scenario of dynamic spectrum use (e.g., spectrum aggregation) where 

information is required about how to aggregate contiguous parts of the highly fragmented spectrum for 

the use of a single operator and how to achieve this when access is available to a dedicated and a shared 

band separated by frequencies. Currently, there is no information available yet about how to cope with 

multiple bands and what are the added complexities both at the network managing and at the user 

equipment (UE) level. 

Section 3 investigates the aspects related to intra-and inter-operator sharing in the above context. 

3.1 Intra-Operator Femto-Macro coexistence 

In this section, the study focuses on the three following general scenarios: 

• Interference between femtocells, 

• Interference between femtocells and femto plus macrocells, 

• Interference between macrocells and femto plus macrocells, 

This aim of this study is to known if a femtocell network can be deployed on the same frequency band as 

a macrocell network and to deduce dimensioning aspects like cell size in femto LTE. For additional 



coexistence scenarios e.g. Femto-Macro interference in a close-access scenario, please refer to 

WINNER+ D1.2 Chapter 4. 

Figure 3-1: Representation of the femtocell concept 

In the given three scenarios, we assess the impact of various parameters on interference factor, SINR and 

throughput. These parameters are: 

• the femtocell transmitting power in mW, 

• the macro cell transmitting power in mW, 

• the cell load, 

• the pathloss, 

• the distance between the serving cell and the MS (=UE), 

• the number of walls between the serving cell and the MS, 

• the mean number of walls between the interfering cells and the MS, 

The location configurations are defined by: The number of walls between the serving cell and the MS and 

the mean number of walls between the interfering cells and the MS, as represented in Figure 3-2. As 

illustrated, the number of walls between the serving femtocell and the MS varies from 0 to 3 and from 1 

to 4 between the others femtocells and the MS. 

Figure 3-2: Different location configurations 



3.1.1 Scenario 1: Interference between femtocells 

The scenario 1 focuses on the assessment of the interference caused in between femtocells for particular 

location configurations. Indeed in this scenario, the impact of the number of walls between the serving 

HBS (Home Base Station) and the MS and the impact of the number of walls between the interfering 

HBS and the MS are assessed. 

3.1.1.1 Assumptions 

In this scenario, we assume that all the femtocells transmit at the same power and we use the ITU indoor 

pathloss model from Recommendation ITU-R P.1238-2 [INDOOR]: 

L 

dB 

= 20log10 ( f 

MHz 

) + 28log10 

( rm 

) − 28 + N 

walls 

× 12 . 

This leads in linear to: 

2 2.8 1.2 N walls − 8 η 

Llinear = f × r × 10 2. = Kr , 

where N walls is the number of walls. 

The SINRs are calculated as follows: 

SINR = 

F 

F 

OMNI 

( η − 2) 

Nother 

η 

OMNI 

+ 

1 

, 

Noise 

2 × π × ρ BS × K Nown × r 

2−η 

η 

Thermal Noise K Nown r 

Noise 

⎡ 

2− 

× × 

+ = 

( 2Rc 

− r) − ( R − r) 

⎤ + 

× K ⎢⎣ 

⎥⎦ 

P 

femto 

η 

, 

where r represents the varying distance between the serving femtocell and MS, η the exponential pathloss 

factor, ρ BS the BS's density, 2Rc the distance between two interfering femtocells (Figure 3), R the 

network's size in meters and P femto represents the femtocells transmitting power in mW. We define K Nown 

and K Nother as a function of N walls for the serving femtocell and the others femtocells, respectively. The 

general formula is: 

2 1.2 N walls −2.8 

K 

N 

= f × 10 . 

Validation of the use of these formulas has been assessed and counterchecked with comparison with 

system simulations developed within Orange. The carrier frequency is set to 2.6 GHz. The distance 

between two interfering HBS is about 56.4 meters. 



Figure 3-3: Input parameters 

Figure 3-4: Symbolisation of the 1 st ring of femtocells 

3.1.1.2 Interference between femtocells: SINR results 

We start the study by the assessment of the SINR, we focus on DL transmission. In this study simulation 

parameters are shown in Figure 3-3. Figure 3-5 shows the SINR in different configurations, considering 

interference from only the first ring of femtocells as symbolically represented in Figure 3-4. Figure 6 

shows the SINR in different configurations, considering interference from the entire network (N other is the 

number of walls between the interfering cells and the MS and N own is the number of walls between the 

serving femtocell and the MS). 

From Figure 3-5 we can say that, if there is one wall between the interfering cells and the MS and no wall 

between the serving femtocell and the MS, the user experiences an SINR of about 34 dB when he is at 6 

meters from his serving HBS. 



100 

Plot of SINR 

80 


80 

60 

60 

40 

40 

dB 

20 

dB 

20 

0 

0 

-20 

-20 

-40 

0 5 10 15 20 25 30 

Distance between the serving femto cell and the MS in meters 

Figure 3-5: 1 st ring of cells 

-40 

0 5 10 15 20 25 30 


Figure 3-6: The entire network 

The most important result to notice is that interference mainly comes from the 1 st ring of cells. From 

Figure 3-5 and Figure 3-6, it clearly appears that the best SINRs are obtained for the best location 

configurations. Indeed, when there is no wall between the serving HBS and the MS and 3 walls between 

the interfering femtocells and the MS (which corresponds to N other =3 N own =0), the SINR reaches 55.9 dB 

when the MS is at 5 meters from its serving HBS (Figure 3-5). 

Whereas, when there are 3 walls between the serving HBS and the MS and one wall between the 

interfering femtocells and the MS (which corresponds to N other =1 N own =3), the SINR reaches 0.44 dB 

when the MS is at 5 meters from its serving HBS (Figure 3-5). 

Another important result to notice is that, as expected, some location configurations are equivalent. 

Indeed, it is due to the ratio calculation: 

K 

K 

Nown 

Nother 

= 

2600 

2600 

MHz 

MHz 

× 10 

× 10 

1.2Nother 

−2.8 

1.2Nown 

−2.8 

= 10 

1.2 

( Nother 

−Nown 

) 

. 

This leads to four equivalences as symbolized in Figure 3-7: 

• N other =1 N own =2 ≡ N other =2 N own =3, 

• N other =1 N own =1 ≡ N other =2 N own =2 ≡ N other =3 N own =3, 

• N other =1 N own =0 ≡ N other =2 N own =1 ≡ N other =3 N own =2, 

• N other =2 N own =0 ≡ N other =3 N own =1. 



Figure 3-7: Equivalent configuration for SINR calculation 

In the following analysis we set an SINR threshold which depends on the number of users in the cell and 

on the requested service as can be seen in Table 3-1. As the targeted services in LTE need high 

throughput performance, this study focuses on a 5 Mbps service. 

Table 3-1: SINR threshold for a requested service of 5 Mbps 

SINR threshold 

1 user using the whole bandwidth (50RB/TTI) 1 dB 

5 users using 10RB/TTI 13.9 dB 

These SINR threshold values are obtained from some throughput performance curves generated by a 

system simulator for an Indoor A 3km/h channel (6 paths) (see Table 3-2) in a 2x2 MIMO with spatial 

multiplexing configuration which is a classical LTE configuration. 

Table 3-2: ITU Indoor A 3 km/h channel 

Tap Channel A Doppler 

Rel. Delay (nsec) Avg. Power (dB) Spectrum 

1 0 0.0 CLASSIC 

2 50 -3.0 CLASSIC 

3 110 -10.0 CLASSIC 

4 170 -18.0 CLASSIC 

5 290 -26.0 CLASSIC 

6 310 -32.0 CLASSIC 

This result has been validated with a 3GPP document [3GPP1]. Under these thresholds the link quality is 

not good enough to satisfy the users. 

These thresholds are represented in Figure 3-8 by the red and black dashed lines. From these figures, we 

notice that, as above: the smaller the number of walls between the serving HBS and the MS is and the 

higher the number of walls between the other HBSs and the MS is, the greater is the SINR. This is 

because, as the desired signal, the interference suffers from fading. 




Cdf plot of SINR 

80 

60 

1 

0.9 

0.8 

1user 

40 

0.7 

0.6 

5 users 

dB 

20 

0 

-20 

0.5 

0.4 

0.3 

0.2 

0.1 

0 5 10 15 20 25 30 


0 

-20 0 20 40 60 80 

dB 

Figure 3-8: SINR and CDF of SINR with thresholds 

In Table 3-3, we can see in which location configurations a femtocell network can achieve at least the 

SINR threshold corresponding to a 5 Mbps service (HBS density =0.0004m -2 ). 

Table 3-3: Summary of the possible configurations 

SINR Thresholds 

1 dB 13.9 dB 

N other =3 N own =0 User is always satisfied. Users are always satisfied. 

N other =2 N own =0 

≡ 



≡ 


≡ 



≡ 


≡ 

N other =3 N own =3, 


≡ 



User is always satisfied. 

User is always satisfied. 

User is satisfied if he is NOT further 

than 26.2 meters from the serving 

HBS. 



HBS. 



HBS. 

Users are always satisfied. 

Users are satisfied if they are NOT further 

than 24.9 meters from the serving HBS. 







From Table 3-3, we can conclude that a femtocell network of 400 HBS in 1km 2 , on a dedicated frequency 

band can be deployed. Indeed, when there is only one user requesting a 5 Mbps service all spatial 

configurations are satisfying, except the more pessimistic N other =1, N own =3, which is an improbable 

configuration. For five users the same results are observed. As expected, when users are closer to the 

interfering HBSs than to their HBS, the interference is too strong. 



Considering the entire network, it is 

reasonable to think that the average 

number of walls between the interfering 

HBS and the MS is about 2. For this 

reason the following parts of the study will 

focus on this spatial configuration. 

Figure 3-9 is an example of a dense 

internet box deployment in which each 

logo of France Telecom represents an 

internet box equipped with an HBS. 

Figure 3-9: Example of internet box deployment 

3.1.1.3 Interference between femtocells: Throughput results 

In this section we focus on the throughput performance for a 2x2 MIMO configuration with spatial 

multiplexing. We choose what is seen as the most realistic deployment scenario for LTE. The simulation 

has been done with the following parameters: 

• Downlink OFDM LTE, Carrier frequency 2.6 GHz, 

• Indoor A 3km/h channel (6 paths), 

• Channel bandwidth 10MHz 

• DTxAA MIMO scheme (spatial multiplexing, the highest MCS used is 64 QAM, code rate 5/6), 

• Maximum transmit power 15 dBm, 

• Without shadowing, 

• BER=0.01, 

• 1 user in the cell (using 10RB). 

The link curve has been generated by a system simulator calibrated for 3GPP studies in Orange labs for 

an Indoor A 3km/h channel (6 paths) and it has been validated by comparing the obtained throughput 

performance curves to simulations done by a NEC 3GPP contribution [3GPP2]. As the curve does not go 

under 0.5 dB, smaller values have been extrapolated. 

Figure 3-10 shows the maximum throughput that 

can be achieved when the total system bandwidth 

is used. 

60 

50 

Throughput per TTI 

40 

Mbps 

30 

20 

As expected, the greatest total throughput is 

achieved by the best location configuration (N other 

=2 N own =0), as the throughput is calculated from 

the SINR. Another important result to notice is 

that the slope of the curve increases as the 

number of walls increases. Indeed, when there is 

no wall between the user and his serving HBS, 

the throughput per TTI decreases from about 8.5 

Mbps between 16 meters and 26 meters. 

10 

0 

0 5 10 15 20 25 30 


Figure 3-10: Throughput performance for the whole 

bandwidth (50 RB/TTI) 



Number of satisfied users N other 

= 2 Service=5 Mbps 

10 

Whereas, when there are 3 walls between the user 

and his serving HBS, the throughput per TTI 9 

decreases from about 42.3 Mbps between 2 

8 

meters and 12 meters. Furthermore, in this 

configuration the user must not be further than 13 

N 

7 

ow n 

=0 

meters. 

N ow n 

=1 

The same observation is done for the number of 

satisfied users shown in Figure 3-11. Indeed, 

when there are 3 walls between the user and his 

serving HBS, the number of satisfied users 

decreases dramatically. When there are 3 walls 

between the user and his serving HBS, the 

number of satisfied users decreases from 10 users 

at 1 meter from the serving HBS to 2 users at 5 

Number of users satisfied 

6 

5 

4 

3 

2 

1 

N ow n 

=2 

N ow n 

=3 

0 

meters. 0 5 10 15 20 25 30 


Figure 3-11: Number of satisfied users 

The impact of the number of walls is easily seen on average total throughput of the cell given in Table 

3-4. 

Space configurations 

• Table 3-4: Average total throughput and average number of satisfied users 

Average total throughput 

Average satisfied users 

requesting 5 Mbps 

Average satisfied users 

requesting 10 Mbps 

N other =2 N own =0 53 Mbps 9.4 4.7 

N other =2 N own =1 28.5 Mbps 4.9 2.3 

N other =2 N own =2 9 Mbps 1.2 0 

N other =2 N own =3 1.4 Mbps 0 0 

Table 3-4 also shows the average number of users satisfied depending on the requesting service and on 

the location configurations. From all the results shown above, it appears that to be at 3 walls from his 

serving HBS (which will not be a common situation) does not satisfy users requesting high data rate 

services. However, this configuration allows basic low data rate services (VoIP…). 

We also imagined a pretty dense scenario in which the internet box would have an impressive success. 

The following figures (12, 13 and 14) show that almost the same performances are achieved when there 

are three times more HBS (1200 HBS in 1km 2 ) which means that the distance between two interfering 

HBS is about 32 meters. The throughput performances are very similar. However the cell size is reduced, 

but still acceptable. 


60 



60 


50 

40 

30 

N other 

=2 N ow n 

=0 

N other 

=2 N ow n 

=1 

N other 

=2 N ow n 

=2 

N other 

=2 N ow n 

=3 

50 

40 

N other 

=2 N ow n 

=0 

N other 

=2 N ow n 

=1 

N other 

=2 N ow n 

=2 

dB 

20 

Mbps 

30 

N other 

=2 N ow n 

=3 

10 

20 

0 

-10 

10 

-20 

0 2 4 6 8 10 12 14 16 18 


Figure 3-12: SINR as a function of the distance. 

Table 3-5: Average total throughput and average number of 

satisfied users. 

Space 

configurations 

N other =2 

N own =0 

N other =2 

N own =1 

N other =2 

N own =2 

N other =2 

N own =3 

Average 

total 

throughput 

Average 

satisfied 

users 

requesting 5 

Mbps 

Average 

satisfied 

users 

requesting 

10 Mbps 

52.6 Mbps 9.4 4.7 

28 Mbps 4.8 2.3 

8.7 Mbps 1.2 0 

1.3 Mbps 0 0 


0 

0 2 4 6 8 10 12 14 16 18 



bandwidth (50 RB/TTI). 

10 

9 

8 

7 

6 

5 

4 

3 

2 

1 



N ow n 

=0 

N ow n 

=1 

N ow n 

=2 

N ow n 

=3 

0 

0 2 4 6 8 10 12 14 16 18 



3.1.1.4 Scenario 1 conclusion 

The most important result to notice is that interference mainly comes from the 1 st ring of cells. All results 

shown above lead to the conclusion that a femtocell network on a dedicated frequency band (here 

10MHz) can be deployed. 

As seen on Figure 3-15, for a femtocell 

network of 400 HBS in 1km 2 (which means 

that the distance between two interfering HBS 

is about 56.4 meters), if a user wants to have a 

guaranteed throughput of 10 Mbps (HDTV + 

other services): 

• he can be further than 28.2 meters from 

his serving HBS when the MS is in 

the same room, 

• he must not be further than 28.2 meters 

from his serving HBS when there is 1 

wall between the serving HBS and 

the MS, 




from his serving HBS when there are 

2 walls between the serving HBS and 

the MS, 

• he must not be further than 5.7meters 



the MS. 

• 

Figure 3-15: Size of a femtocell depending on the 

spatial configuration. 

As can be seen on Figure 3-16, for a femtocell 

network of 1200 HBS in 1km 2 (which means 

that the distance between two interfering HBS 

is about 32 meters), if a user wants to have a 

guaranteed throughput of 10 Mbps (TVHD + 

other services): 

• he can be further than 16.2 meters from 

his serving HBS when the MS is in 

the same room, 


from his serving HBS when there is 1 

wall between the serving HBS and 

the MS, 




the MS, 




the MS. 

Figure 3-16 : Size of a femtocell depending on the 

spatial configuration. 

Considering only downlink transmission, we observed that interference mainly comes from the 1st 

ring of cells and we recommend to have no more than 2 walls between the serving HBS and the MS. 

To guarantee a 10 Mbps rate to deliver high QoS for multiplay offer (TVHD plus other 

services) to one user using the whole bandwidth, the cell size radius is about (when there are 2 walls 

between the serving HBS and the MS) : 

• 14,3 meters (400 HBS in 1km 2 , distance between two interfering HBS is about 56.4 meters), 

• 8.1 meters (1200 HBS in 1km 2 , distance between two interfering HBS is about 32 meters). 



3.1.2 Scenario 2: Interference between femto and macrocells 

The scenario 2 focuses on the assessment of the interference induced on a femtocell by both femtocell and 

macrocell networks, for particular location configurations and different pathloss. In this scenario, we use 

almost the same assumptions as in scenarios 1. 

The way we model interference from femto and macro is based on superposition of macrocells on 

femtocells as shown in Figure 3-17. 

Figure 3-17: scenario 2 modelling 


In this scenario, we used the same pathloss formula as in the scenario 1 for all the HBSs: 

L log ( f ) + 28log ( r ) − 28 + N 12 . 

dB 

= 20 

10 MHz 

10 m 

walls 

× 

And the Cost231-Walfish-Ikegami Dense Urban pathloss formula for the macrocells at which we added a 

12dB loss for the penetration: 

L 38× 

log10 ( r ) + 150.9 . 

dB 

= 

Km 

This means for instance that a first ring macro cell interference on femtocell victim will be composed of: 

o dense urban path loss on a distance corresponding to the Macro radius 

o plus 12 dB due to outdoor to indoor penetration 

o plus femtocell radius with indoor path loss 



In this study the HBS have all 

omnidirectional and macrocells are 

three-sectored. The antenna pattern is 

shown in Figure 18 and is calculated 

as follows: 

1 

0.9 

0.8 

0.7 

0.6 

Antenna pattern 

G 1 

(θ) 

G 2 

(θ) 

G 3 

(θ) 

2 

( θ ) = − min(12× 

( θ / 70) ,20) 

G , 

( θ ) G( ), 

G = 

G 

1 

θ 

2 

( θ ) = G( θ + 2π 

), 

3 

( θ ) = G( θ + 4π 

) 

G 

3 

. 

3 

0.5 

0.4 

0.3 

0.2 

0.1 

-250 -200 -150 -100 -50 0 50 100 150 200 250 

Angle Theta 

Figure 3-18: Antenna Pattern. 

The SINR is calculated as shown hereafter. In this scenario we assess the interference induced on a 

femtocell by both femtocell and macrocell networks. The analytical model which takes the assumption is 

used to calculate the interference factor coming from macrocell network. 

F 

OMNI Macro 

2 × π × ρ 

= 

BS macro 

( η − 2) 

m 

× K 

× K 

Nown 

macro 

× P 

× P 

macro 

femto 

× r 

η 

f 

2−η 

m 

2−η 

m 

[( Rc + Rcm − r) − ( R − r) 

] 

Where r represents the varying distance between the HBS and MS, η f the exponential pathloss factor of 

HBS, η m the exponential pathloss factor of macrocells, ρ BS macro the BS's density, Rc+Rcm the distance 

between two interfering cells, R the network's size in meters, P femto represents the HBS power in mW and 

P macro represents the macrocell power in mW. We define K Nown as a function of N walls for the serving HBS 

and MS: 

K 

Nown 

2 1.2 N walls −2.8 

= f × 10 

And K MACRO is calculated as follows: 

From the FLUID model: 

Where: 

a 

K macro 

= f 

× 10 

0 15.09−(38× 

log10 

( 1000 )/10) 

( θ ) × F b( θ ) 

FSect Macro 

= a 

OMNI Macro 

+ 

2π 

1 S 

omni 

G( ) 

0 

( ) ∫ θ 

θ = × 

And b ( θ ) = 0 

2π 

S 

sec t 

The Analytical model developed in Orange Labs is also used to calculate the interference factor coming 

from other HBS [Kel07]. 

F 

OMNI Femto 

2× 

= 

× ρ 

( η − 2) 

× K 

× K 

× r 

f 

π 2−η 

f 

BS femto Nown 

f 

Nother 

η 

⎡ 

⎢⎣ 

2−η 

f 

( 2Rc 

− r) − ( R − r) 

⎤ 

⎥⎦ 



Where r represents the varying distance between HBS and MS, η f the exponential pathloss factor, ρ BS femto 

the HBS's density, 2Rc the distance between two interfering HBS, R the network's size in meters and 

P femto represents the femtocell power in mW. We define K Nown and K Nother as a function of N walls for the 

serving femtocell and the others femtocells respectively: 

2 1.2 N walls −2.8 

K = f × 10 

N 

The total interference factor is obtained by summing the two interference factors above, as shown below: 

F = F + F 

TOTAL 

Sect Macro 

OMNI Femto 

The SINR is obtained as follows: 

SINR = 

F 

TOTAL + 

1 

Noise 

in which 

⎛ 

⎜ 

Thermal Noise × K 

Nown 

× r 

Noise = 

⎝ 

Pfemto 

η f 

⎞ 

⎟ 

⎠ 

The input parameters of the simulation studied in the next section are shown in Figure 3-19. 


3.1.2.2 Interference between femto and macro cells: SINR results 

Let's start by the assessment of the SINR. Figure 26 shows the SINR considering the interference coming 

from the entire network (N other is the number of walls between the interfering cells and the MS and N own is 

the number of walls between the serving femtocell and the MS). In this scenario the HBS maximum 

transmitting power is set to 15 dBm, the macrocell maximum transmitting power is set to 47.8 dBm and 

cell load is uniformly set to 70%. 




From Figure 3-20 we can say that if there is one wall 

between the interfering cells and the MS and no wall 

between the serving HBS and the MS (which corresponds to 

N other =1 N own =0), the user experienced a SINR of about 

29.5 dB when he is at 6 meters from his serving HBS. 

As expected, it clearly appears that the best SINRs are 

obtained for the best location configurations. Indeed, when 

there is no wall between the serving HBS and the MS and 3 

walls between the interfering femtocells and the MS (which 

corresponds to N other =3 N own =0), the SINR reaches 55.2 dB 

when the MS is at 5 meters from its serving HBS (Figure 

19). Whereas, when there are 3 walls between the serving 

HBS and the MS and one wall between the interfering 

femtocells and the MS (which corresponds to N other =1 N own 

=3), the SINR reaches -4.1 dB when the MS is at 5 meters 

from its serving HBS. 

dB 

80 

60 

40 

20 

0 

-20 

-40 

0 5 10 15 20 25 30 



Another important result to notice is that some location configurations are almost equivalents. Indeed as 

in scenario 1, it is due to the analytical model and pathloss (underlined by the ratio calculation hereafter), 

which means that in this scenario the macrocells have a small impact on the femtocell network. 


K 

K 

Nown 

Nother 

10 

= 

10 

1.2Nother 

−2.8 

1.2Nown 

−2.8 

= 10 

1.2 

( N −N 

) 

other 

own 

55 

50 

45 

Three equivalences (Figure 28): 



• N other =1 N own =2 ≡ N other =2 N own =3. 

It is also interesting to see that in this scenario the 

importance of N own grows as the number of walls 

between the interfering cells and the MS grows. 

dB 

40 

35 

30 

25 

20 

15 

2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3 


Figure 3-21: Equivalences 

In this report, we set an SINR threshold which depends on the number of users in the cell and on 

the requested service as can be seen in Table 3-6. As the targeted services in LTE need high throughput 

performances, this study focuses on a 5 Mbps service. 



Table 3-6: SINR threshold for a requested service of 5 Mbps. 

SINR threshold 

1 user using the whole bandwidth (50RB/TTI) 1 dB 

5 users using 10RB/TTI 13.9 dB 

The SINR threshold values are obtained from the throughput performance curves generated by 

the simulator, called SOFFA, for an Indoor A 3km/h channel (6 paths) in a 2x2 MIMO with spatial 

multiplexing configuration which is a classical LTE configuration. Under the given thresholds the link 

quality is not good enough to satisfy the users. 

These thresholds are represented in Figure 3-22 by the red and black dashed lines. Considering the entire 

network, it is reasonable to think that the average number of walls between the interfering HBS and the 

MS is about 2. For this reason the following parts of the study will focus on this spatial configuration. 

N other =2 

N own =0 

N other =2 

N own =1 

N other =2 

N own =2 

N other =2 

N own =3 

Table 3-7: Configuration summary 


1 dB 13.9 dB 

Ok 

Ok 

Ok 

Ok If d


In this section we focus on the throughput performances for a 2x2 MIMO configuration with spatial 

multiplexing. Indeed, with regard to the time required to assess radio link curves, we took what is seen as 

the most realistic deployment scenario for LTE. The simulation has been done with the following 

parameters: 


• Indoor A 3km/h channel (6 paths), 

• Channel bandwidth 10Mhz 

• DTxAA MIMO scheme (spatial multiplexing the highest MCS used is 64 QAM, code rate 5/6), 

• Maximum power 15 dBm, 


• BER=0.01, 

• 1 user in the cell (using 10RB). 

The link curve has been generated by the simulator called SOFFA for an Indoor A 3km/h channel (6 

paths) and has been validated by comparing the obtained throughput performance curves to simulations 

done by a NEC 3GPP contribution. As the curve does not go under 0.5 dB, the values below have been 

extrapolated. 

Figure 3-23 shows the maximum throughput that 

can be achieved when the total system's 

bandwidth is used. 

As expected, the greatest total throughput is 

achieved by the best location configuration (N other 

=2 N own =0), as the throughput is calculated from 

the SINR. 

Another important result to notice is that the slope 

of the curve increases as the number of walls 

increases. Indeed when there is no wall between 

the user and his serving HBS, the throughput per 

TTI decreases from about 8,5 Mbps between 16 

meters and 26 meters. 

Whereas, when there are 3 walls between the user 

and his serving HBS, the throughput per TTI 

decreases from about 42.3 Mbps between 2 meters 

and 12 meters. Furthermore, in this configuration 

the user must not be further than 13 meters. 

The same observation is done for the number of 

satisfied users shown in Figure 24. Indeed when 

there are 3 walls between the user and his serving 

HBS, the number of satisfied users decreases 

dramatically. When there are 3 walls between the 

user and his serving HBS, the number of satisfied 

users decreases from 10 users at 1 meter from the 

serving HBS to 2 users at 5 meters. 

From these results it can easily be seen that the 

performances are very similar to the ones obtained 

in scenario 1. This means that a femtocell network 

can be deployed on the same frequency band (here 

10MHz) as the outdoor macrocell network. 

Mbps 


60 

50 

40 

30 

20 

10 


N other 

=2 N ow n 

=0 

N other 

=2 N ow n 

=1 

N other 

=2 N ow n 

=2 

N other 

=2 N ow n 

=3 

0 

0 5 10 15 20 25 30 



bandwidth (50 RB/TTI) 

10 

9 

8 

7 

6 

5 

4 

3 

2 

1 



N ow n 

=0 

N ow n 

=1 

N ow n 

=2 

N ow n 

=3 

0 

0 5 10 15 20 25 30 






In Table 3-9, we can see in which location configurations the system can achieve at least the SINR 

threshold corresponding to a 5 Mbps service (HBS density =0.0004m -2 , Node B density =2*10 -6 m -2 ). In 

Table 3-10, we can see in which location configurations the system can achieve at least the SINR 

threshold corresponding to a 5 Mbps service (HBS density =0.0004m -2 ). 

Table 3-9: Configuration summary (femto + macro) 


1 dB 13.9 dB 

N other =2 N own =0 Ok Ok 

N other =2 N own =1 Ok Ok If d


femtocell network can be deployed on the carrier bandwidth frequency band (here 10MHz) as the 

outdoor macrocell network. 



3.1.3 Scenario 3: Interference between femto and macrocells 

The scenario 3 focuses on the assessment of the interference induced on a macrocell by both femtocell 

and macrocell networks, for particular location configurations and different pathloss. In this scenario, we 

use almost the same assumptions as in scenario 2. 


In this scenario, we use the same pathloss formula as in the scenario 1 for all the HBSs: 

L 

dB 

= 20 

10 

walls 

log10 ( f 

MHz 

) + 28log ( rm 

) − 28 + N × 12 

And the Cost231-Walfish-Ikegami Dense Urban pathloss formula for the macrocells: 

L 

dB 

= 

Km 

38× 

log10 ( r ) + 150.9 

In this study the HBS have all 

omnidirectional and macrocells are 

three-sectored. The antenna pattern 

is shown in Figure 26 and is 

calculated as follows: 

1 

0.9 

0.8 

0.7 

Antenna pattern 

G 1 

(θ) 

G 2 

(θ) 

G 3 

(θ) 

G 

2 

( θ ) = − min(12× 

( θ / 70) ,20) 

0.6 

0.5 

( θ ) G( ), 

G = 

1 

θ 

0.4 

G 

2 

( θ ) = G( θ + 2π 

), 

3 

( θ ) = G( θ + 4π 

) 

G 

3 

. 

3 

0.3 

0.2 

0.1 

-250 -200 -150 -100 -50 0 50 100 150 200 250 

Angle Theta 

Figure 3-26: Antenna Pattern 

The SINR is calculated as shown hereafter. In this scenario we assess the interference induced on a 

femtocell by both femtocell and macrocell networks. The analytical model is used to calculate the 

interference factor coming from macrocell network. 

F 

OMNI Macro 

2 × π × ρ 

= 

BS macro 

( η − 2) 

m 

× K 

× K 

macro 

macro 

× P 

× P 

macro 

macro 

× r 

η 

m 

2−η 

m 

2−η 

m 

[( 2Rcm 

− r) − ( R − r) 

] 

Where r represents the varying distance between the serving macrocell and MS, η m the exponential 

pathloss factor of macrocells, ρ BS macro the BS's density, 2Rcm the distance between two interfering cells, 

R the network's size in meters and P macro represents the macrocell power in mW. We define K MACRO as 

follows: 

K macro 

= f 

× 10 

0 15.09−(38× 

log10 

( 1000 )/10) 

From the analytical model: 



( θ ) × F b( θ ) 

FSect Macro 

= a 

OMNI Macro 

+ 

Where: 

a 

2π 

1 S 

omni 

( θ ) 

( θ ) ∫ G 

0 

= × 

And b( θ ) 

2π 

S × G ( θ ) 

sec t 

1 

= 

3 

∑ 

S = 2 

G 

1 

( θ ) 

( θ ) 

G 

The FLUID model is also used to calculate the interference factor coming from other HBS. 

F 

OMNI Femto 

2 × 

= 

ηm 

π × ρ × K × P × r 

2−η 

f 

BS femto macro femto 

( η − 2) × K × P × G ( θ ) 

f 

Nother 

macro 

1 

⎡ 

⎢⎣ 

2−η 

f 

( Rcm + Rc − r) − ( R − r) 

Where r represents the varying distance between the HBS and MS, η f the exponential pathloss factor of 

HBS, η m the exponential pathloss factor of macrocells, ρ BS macro the BS's density, Rc+Rcm the distance 

between two interfering cells, R the network's size in meters, P femto represents the HBS power in mW and 

P macro represents the macrocell power in mW. We define K Nown and K Nother as a function of N walls : 

⎤ 

⎥⎦ 

K 

N 

2 1.2 N walls −2.8 

= f × 10 

And K MACRO is calculated as follows: 

K macro 

= f 

× 10 

2 15.09−(38× 

log10 

( 1000 )/10) 

The total interference factor is obtained by summing the two interference factors above, as shown below: 

F = F + F 

TOTAL 

Sect Macro 

OMNI Femto 

The SINR is obtained as follows: 

SINR = 

F 

TOTAL + 

1 

Noise 

in which 

⎛ Thermal Noise × K 

macro 

× r 

⎜ 

Noise = 

⎝ 

Pmacro 

η m 

⎞ 

⎟ 

⎠ 

The input parameters of the simulation studied in the next section are shown in Figure 3-27. 




3.1.3.2 Interference between femto and macro cells: SINR results 

Let's start by the assessment of the SINR. Figure 3-28 shows the SINR considering the interference 

coming from the entire network (N other is the number of walls between the interfering cells and the MS 

and N own is the number of walls between the serving macrocell and the MS). In this scenario the HBS 

maximum transmitting power is set to 15 dBm, the macrocell maximum transmitting power is set to 47.8 

dBm and cell load is uniformly set to 70%. 

The most important phenomenon to notice is that the femtocells have an important impact on the 

macrocell SINR. 

20 


15 

10 

5 

dB 

0 

-5 

-10 

-15 

0 50 100 150 200 250 300 350 400 


Indeed, from Figure 3-28 focusing on the results for two walls 


between the interfering cells and the MS and no wall between 

the serving macrocell and the MS (which corresponds to N other 

=2 N own =0), the user experiences an 

SINR of about: 

• 8.3 dB when he is at 300 meters from his serving e-Node B (considering 400 HBS in 1km 2 ), 

• 10.6 dB when he is at 300 meters from his serving e-Node B (considering 200 HBS in 1km 2 ), 

• 14.4 dB when he is at 300 meters from his serving e-Node B (considering no HBS). 



As expected, it clearly appears that the best SINRs are obtained for the best location configurations. 

Indeed, when there is no wall between the serving macrocell and the MS and 4 walls between the 

interfering cells and the MS (which corresponds to N other =4 N own =0), the smallest achieved SINR is 

about 9.7 dB when the MS is at 400 meters from its serving e-Node B. 

Another important result to notice is that there is a maximum value of SINR (about 17 dB) which is 

underlined by the "No HBS" curve (which means that there is 0 femtocell). This limitation is explained 

by the fact that as the e-Node B are three-sectored, the studied sector suffers from interference coming 

from the two other sectors. 

3.1.3.3 Interference between femtocells: Throughput results 

As the extension to LTE and associated study were planed within 3 months long, simulation with impact 

from femto on macro cell throughput could not be performed. Hence, following scenario for a 2x2 MIMO 

configuration with spatial multiplexing for following parameters remains to be done: 


• Outdoor pedestrian B 3km/h channel (6 paths), 

• Channel bandwidth 10MHz 

• DTxAA MIMO scheme (spatial multiplexing the highest MCS used is 64 QAM 5/6), 

• Maximum power 47.8 dBm (ENode B), 


• BER=0.01, 

• 1 user in the cell (using 10RB) 

A study on the throughput performance cannot be done. 


The most important phenomenon to notice is that the femtocells have an important impact on the 

macrocell SINR. Another important result to notice is that there is a maximum value of SINR 

(about 17 dB) which is underlined by the "No HBS" curve (which means that there is 0 femtocell). 

This limitation is explained by the fact that as the e-Node B are three-sectored, the studied sector 

suffers from interference coming from the two other sectors. 



3.2 Multi-Operator Resource Sharing Scenario in the Context of IMT-A Systems 

This subsection addresses spectrum and network sharing for IMT-A systems, and proposes an innovative 

way to coordinate and manage spectrum and network resources in the context of user centric mobile 

broadband services. The proposed concept is for the integration of the spectrum and resource coordination 

into a single framework with the objective to enable improved performance and capacity gains. It has not 

been researched before in the context of IMT-A systems. 

Decentralized spectrum and joint radio resource management are investigated in [SALL08], [PER04], 

[GRA05]. It is shown in [PER04] that competing UMTS operators can cooperate without operational 

information exchange and share a block of UMTS carriers simultaneously, to deploy multimedia services. 

An algorithm based on the total transmitted energy thresholds is proposed that limits the loading by 

individual operators in the shared band and allows for coexistence. It is shown that dynamic spectrum 

sharing offers capacity gain due to the increase of the trunking efficiency in a spectrum-shared system. 

This improvement is achievable by the pooling of the radio channels belonging to different operators 

together. In [GRA05], it is shown that proper underlying spectrum sharing coordination processes are 

required to manage the interference and to further increase the achievable capacity gains. The 

investigations are performed for UMTS and with the use of joint radio resource management (JRRM). 

JRRM is a distributed approach to resource management [PER04]. Besides, it is shown that the different 

proposed schemes have to be dynamically managed as a function of the traffic loads through the 

activation of the right strategies. This appropriate capacity management facilitates to get the best 

spectrum reuse opportunities. Finally, [SALL08] extends the concept to a scenario of cognitive networks 

using a decentralized approach to both spectrum and radio resource management (RRM), i.e., JRRM. The 

proposed solution is based on a layered approach where spectrum sharing and JRRM are identified and 

integrated at both inter-and intra-operator levels. 

Work reported in [MIH09] goes beyond the above-mentioned concepts by exploring the possibilities of 

spectrum sharing integration (e.g., spectrum aggregation) and RRM functionalities in the context of IMT- 

Advanced systems. The concept supports cooperation between operators also when different RATs must 

share resources. The required functionalities and interactions are defined and an integrated framework is 

proposed. A combined centralized and distributed approach to RRM (i.e, cooperative RRM) that is 

proposed for the WINNER system and analyzed in [MIH08] for inter and intra-system interworking is 

assumed. The spectrum management functionalities can be integrated with this framework to use further 

the advantages of higher performance and capacity gains. 

The WINNER project considers spectrum management and sharing but does not investigate it together 

with use over fragmented frequency bands and network resource coordination. Spectrum aggregation is 

proposed for LTE-A, an IMT-Advanced candidate concept evolving from LTE. Although LTE-A 

considers use of fragmented bands, it does not consider it together with network management strategies. It 

is proposed here to use the state of the art of the above system concepts and elevate them further by 

providing research results that are valuable to the definition of the deployment and user requirements of 

IMT-A systems and mobile broadband services. 

The ITU-R M.1645 Recommendation provides the initial requirements for the IMT-Advanced system 

concept. Legacy systems, such as GSM, UMTS, and WLAN, would continue to provide services to users; 

therefore, a generic framework for the support of the interworking between these, essentially, different 

systems is required. Interworking implies coexistence and sharing of resources. Coexistence means the 

concurrent operation of different radio systems in the same or in adjacent frequency bands without 

causing degradation to any other system, with emphasis on the indicated limitations in terms of, e.g., 

frequency separation, physical separation, and transmission resources. Sharing in the context of spectrum 

considers the use of the same frequency band by different radio systems, either with coordination or 

possibly without any coordination between the systems, with emphasis on the spectrum access schemes 

and methods. ‘Cooperation’ is the term used here to jointly refer to inter-working, coexistence and 

sharing. 

The system concept developed in [WINDIID6.13.4] proposed that cooperation (i.e., JRRM) mechanisms 

for intra-system interworking are developed at the radio segment level of new RANs, following a 

combined centralized and distributed approach, while inter-system interworking is handled by a central 

entity located outside of the RAN architecture following a centralized approach. A gateway (GW) 

functionality is an IP anchor to external data networks (e.g., Internet, corporate networks, operator 

controlled core networks) and operator services. It also terminates flows on the network side and serves as 

the anchor point for external routing. Thus, all functions that operate on user data traffic are located here. 

Figure 3-29 shows the interworking of cooperative RRM and spectrum management functions at different 

levels of the radio access network architecture. 



Measurements 

(neighbouring cell lists; 

KPI calculation; 

Triggers) 

Multi‐RAN/multioperator 

Scenario? 

YES 

Cooperative 

RRM/Spectrum 

Manager 

(centralized 

approach) 

Inter‐system 

interworking and 

spectrum 

management 

Handover 

Admission control, 

Congestion control 

Load control 

Coexistence and 

sharing 

NO 

Generic RRM/ 

Spectrum Server 

(centralized and 

distributed 

approach) 

Intra‐system 

Interworking 

and 

spectrum 

control 

Intra‐system 

handover, 

load balancing, 

vertical 

specturm 

sharing 

Policy‐Based 

network and 

spectrum 

management 

Multi‐stage 

admission 

control 

Specific 

RRM/ 

Spectrum assignment 

(distributed 

approach) 

Intra‐mode 

Interworking and 

spectrum control 

Intra‐mode 

handover, 

load balancing, 

horizontal 

specturm 

sharing 

Multi‐band 

Scheduling, 

admission 

Control, 

power control 

Figure 3-29: Cooperation framework for an IMT-Advanced candidate system 

The cooperative RRM and spectrum manager contain centralized RRM and spectrum sharing functions, 

and enable interworking, sharing and co-existence with other systems (RATs). These interface with the 

GW, which contains functionalities related to generic RRM and spectrum management on intra-system 

level in order to obtain information about the associated UEs and their characteristics. The generic RRM 

is in essence a centralized type of control but it allows a distributed approach, while the loads in the 

network are low-to-medium (i.e., specific RRM located at the level of BS or relay node (RN). The 

spectrum server communicates spectrum sharing and spectrum assignment decisions to the BS. The 

spectrum server is accessible via the GW for spectrum negotiations. The BS containing the specific RRM 

functions is a point of radio access. The BS performs all radio related functions for active UEs (i.e., 

sending data) and is responsible for governing radio transmission to and reception from the UE in one 

cell. The interconnectivity of the functions in the support of the above framework is shown in the control 

and user plane from Figure 3-30. In practice, the amount of required spectrum bandwidth for a network 

operator will depend on the traffic / capacity requirements, the modulation scheme used, the cell sizes, 

and the frequency reuse factor. The optimum balance is a commercial decision, based on these technical 

considerations. 

3.2.1 Spectrum Aggregation Requirements 

The concept of spectrum aggregation consists of exploiting multiple, small spectrum fragments 

simultaneously to deliver a wider band service (i.e., not otherwise achievable when using a single 

spectrum fragment. The best choice of a frequency band for a mobile communications system depends on 

many different factors [DIX08]. In addition, once the spectrum has been obtained the problem of 

managing the shared band implies proper user allocation mechanisms. Spectrum aggregation allows that 

new high data rate wireless communication systems can coexist while reusing the spectrum of legacy 

systems. In general, the lower frequency bands are better suited to longer range, higher mobility, lower 

capacity systems, while higher frequency bands are better suited to shorter range, lower mobility, and 

higher capacity systems. Therefore, for any given network the optimum frequency would vary depending 

on the required range, mobility and capacity [DIX08]. For a commercial mobile network, the required 

range and capacity varies by changing the number of BSs (i.e., changing the size of their coverage area) 

so that the level of investment in the network infrastructure can affect the optimum balance. The White 

Paper of WWRF WG 8 [DIX08] reports that spectrum for higher mobility applications (which are usually 

operating in interference limited scenarios) should be separated in the frequency domain from short-range 

applications (often noise limited scenarios) in order to avoid unnecessary complexity in sharing scenarios. 

This can be combined with a traffic split at higher RAN levels (i.e., GW level or generic RRM level) to 

explore the benefits of interworking at lower layers, while perceiving the network conditions and acting 

upon them. 



Access to 

shared 

band 

IP Handover 

Admission 

Control and Flow 

Establishment 

Radio Handover 

Adaptive 

modulation and 

coding 

Resource 

access 

Flow Control 

Load Control 

Control Plane 

Congestion Control 

Scheduling 

Multi‐band 

Scheduling (MAC 

level) 

User Plane 

Figure 3-30: Interconnectivity of cooperative functions in the control and user planes 

Some factors, which should be taken into account to determine the best frequency band, depend on the 

individual service requirements and the network requirements. These are explained here together with 

some considerations on how radio frequency bands should be managed once access to frequency 

spectrum has been obtained [MIH09]. 

3.2.2 Hypotheses for System and Spectrum Resource Usage 

The following research hypotheses for the system and the spectrum usage are made here: 

• It is possible to reuse the existing infrastructure in terms of signalling, measurement, and 

protocol procedures with novel interference and resource management schemes; 

• An integration of spectrum and network resource management functionalities can be exploited 

within an IMT-A scenario, while benefiting from higher performance and system capacity gains. The key 

to such integration is the pooling of the resources together; the integration allows for mapping of the 

service requirements onto an available spectrum amount and translates the latter into network loads; 

• It is possible to use the widely separated frequency bands for achieving lower delays and jitter 

and higher user throughput; this can be achieved by use of multi-band scheduling algorithms and 

exploiting the channel diversity; 

• User allocation on various frequency bands can be performed based on the UE capabilities and 

performance gains in terms of increased peak data rate that can be obtained by use of specially designed 

algorithms for the management of the shared band; 

• Information from the network about the system state (e.g., received signal strength, transmitted 

power, UE velocity, etc.) and used in RRM procedures based on cooperative RRM [MIH08], such as 

load, admission and congestion control can successfully be combined with dynamic spectrum use and 

reduce the need of spectrum aggregation in some cases; this will be achieved by the integration of 

functionalities for spectrum and RRM use and further optimized by the implementation of joint policies; 

3.2.3 Intra-and Inter-Operator Sharing Aspects 

3.2.3.1 Intra-Operator Sharing Considerations 

For simplicity, in the following discussions it is assumed that the UE is a single-band device. Spectrum 

aggregation can appear when an operator’s dedicated band is not continuous but is split in two or more 

parts. In addition, spectrum aggregation can happen in scenarios, in which an operator accesses both a 

dedicated band, and a spectrum sharing band which is separated in frequencies from the dedicated 

operator’s band. This is also valid for the inter-operator scenario. Once a portion of the spectrum has been 

obtained, it is important to identify the load on the network that the subsequent allocation in the bands 

will create. This requires that service requirements are translated into load values and in addition mapped 

to the available spectrum amount. A possible traffic split at generic RRM level (i.e., centralized control) 

can act upon the current load network conditions and assist the distributed approach to allocating users in 

the frequency bands without disturbing the balance of the overall network. The envisioned relationships 



between RRM and spectrum entities are shown in Figure 3-31. During spectrum aggregation, the amount 

of available spectrum resources in both shared band and dedicated bands will change. 

Cooperative 

RRM 

1. Estimation of available 

radio resources in controlling 

sub network (BS) and traffic split 

according to service 

requirements 

V1 V2 Vn 

A1 A2 An 

GW 

BS 

3. GW maps split traffic streams to allocation in bands 

inform Cooperative RRM server about timing requirements 

between sub-streams 

BS 

2. Users are allocated 

according to specific 

(CQIs) by multi-band 

scheduler and generic 

(network load) RRM 

requirements 

Multi‐band 

scheduler 

V D 

A C 

A D 

A D 

V C 

V C 

Band D 

Band C 

Figure 3-31: Translation of service requirements into network resources and mapping to spectrum 

availability 

A mechanism is required to match the available data in the queues to the available bandwidth. The 

scheduling of users over multiple frequency bands can be modelled in its most general form as a General 

Multi-Band Scheduling (GMBS) problem. A multi-band scheduler to manage the balance between the 

data pipe and the obtained extra source of spectrum is proposed in [WINIID5.9.1]. When the bandwidth 

from the shared band becomes available, the scheduler must be capable of realizing such a change in the 

spectrum pipe and shift some of the traffic load from the dedicated band to the shared band or vice versa. 

The scheduler must also be capable of further load balancing by actively monitoring the forthcoming 

changes in the spectrum and traffic data in order to shift the load from the shared band to dedicated one 

and vice- versa, if so required. 

Interworking between RRM and spectrum functions then can be translated into the objective of 

maximizing the total throughput of the operator, while considering the QoS requirements of the service 

classes. The more resources are available for a system, the higher the multiuser diversity gain that can be 

obtained [LUO03], [E2RD5.3]. The performance is heavily dependent on the radio channel qualities for 

each user in the considered bands. The channel quality indicators (CQIs) depend on the path loss and on 

the distance from the BS. The operator will have good improvements when the UEs have heterogeneous 

spatial distribution in the cell (variable distances from the BS), i.e., different channel qualities in the 

considered bands. Here we can view the joint allocation on the dedicated and shared bands as cooperation 

at the scheduling level. 

Figure 3-30 shows that at intra-system cooperation level a multi-stage admission control mechanism is 

active that uses the benefits of a joint distributed and centralized control for the admission of users to the 

network, while balancing the load over a number of BSs and thus increasing the network throughput 

[MIH208]. It is shown that through load balancing both gains from interworking at MAC level (i.e., 

multiplexing gain) as well as above the MAC (at generic RRM level) are realizable. It is proposed here to 

exploit this concept in relation to spectrum aggregation. The multi-band scheduling mechanism will be 

busy allocating users over the two frequency bands until the key performance indicators (KPIs) inform 

about a degradation of the QoS, which would mean that the allocation over the frequency bands does not 

match the current network (i.e., traffic) conditions. The operator then would rely on a generic or 

cooperative (i.e., centralized) RRM mechanism for restoring the network balance. Cooperative RRM 

would mean that the intra-system RRM has been exhausted and inter-system cooperation (still between 

systems belonging to the same operator) will be activated (i.e., cooperative RRM level). Such actions 

need to work concurrently with the scheduling over the available frequency bands. In this case, the 

achievable gain will be a combination of the gain from scheduling over separated frequency bands (i.e., 

diversity gain) and the trunking gain from resource sharing (i.e., multiplexing gain achievable through 



traffic splitting and BSs interworking [MIH208]). Future work will compare the achievable performance 

for each of the two gains. 

3.2.3.2 Inter-Operator Considerations 

The inter-operator sharing deployment is described in Section 3.2. In this scenario, each operator has 

access to a shared band, in addition to owning a dedicated band [WIN+D1.2]. Two general spectrumsharing 

deployments can be distinguished: 

• Common frequency pool, which is shared by all operators (considered here); 

• Separated frequency pools, which are shared by two neighbouring operators. 

In the inter-operator case, the interworking can be seen as interworking at cooperative RRM level and 

allocation in the shared band. The two mechanisms for coping with the traffic demands can be seen as 

exchange of resources between operators at a system (i.e., cooperative RRM) and at a user level (i.e., 

sharing of the band). The multi-band scheduling mechanism can use the priority levels of the split traffic 

for assigning the traffic over the frequency bands and in case of changes in availability of the shared 

band, it can relocate traffic or decide to drop or reduce the bandwidth of less important flows. For this, an 

input from the sharing negotiation functionalities in the network is required. The exchange of resources at 

a system level is realized by using the logical functionality of the GW [WINIID6.13.4]. There are two 

types of GW logical nodes, the IP Anchor GW, and the control GW. The control GW provides control 

functions for UEs that are not active (i.e., not sending data) and functions that control and configure the 

IP Anchor GW. In practice, there would be several IP Anchor GWs as well as an independent number of 

control GWs. One BS can be connected to multiple GW logical pairs and conversely, one GW logical pair 

can be connected to multiple BS. Thus, the logical pairs of the GW form a pool of equipment that may 

cover large areas, e.g., cities. In [TRA07] a hybrid handover is proposed based on the concept of using 

pools of GWs as a means to reduce the number of IP handovers (i.e., support of macro mobility). By 

using the logical association between the UE and GW independently of the BSs, the set of GWs can be 

seen as a pool of resources. This in essence is a trunking gain realized by the sharing of resources 

between GWs and can be extended to the concept of GWs belonging to different operators. When 

degradation in the QoS is observed based on monitoring information from the KPIs, adding or removing 

of GWs can help to bring the network in a balanced state (i.e., load balancing). This would reduce the 

observed delays due to signalling required for the user context transfer or the degradation of the QoS for 

some users, and even dropping of users due to the need to reallocate users in the shared band. 

Without the use of the pool of GWs, when the shared band is unavailable a fast transfer to the dedicated 

band might be required to maintain the connectivity, and provide basic and priority services. In addition, 

there can be various reasons to initiate a frequency band transfer, for example the load in the other system 

with whom the resources are shared may change, a user terminal might move into an area where it would 

interfere with the other system’s transmissions, and so forth. Realizing the pool of GWs reduces the need 

for such actions because it allows for a centralized balancing of the load and resources, while preserving 

the logical associations of the UEs. Anticipation of a band reallocation improves the system performance 

because the contexts can be exchanged before the connection on the band is actually lost [WINIID5.9.1]. 

The information about this can be obtained by scanning other networks, or from the Cooperative RRM 

functionality that also implements a spectrum manager. The pooling concept is shown in Figure 3-32. 



Figure 3-32: Pooling of resources of different operators 

Spectrum assignment is dictated on a short-term basis by the services used at the UE. As a result, it is a 

rather distributed method even though the information exchange between the different systems (belonging 

or not to different operators) can be distributed or centralized. The knowledge of the existing spectrum 

resources within a RAN may be centralized since this information belongs to each RAN, can be accessed 

through the spectrum server, and the available spectrum resources belong to a common pool between 

RANs. The spectrum information can be distributed to the BSs, which, in scenarios of low to medium 

loads, can take fast decisions and provide for multiplexing gain. 

The system performance is affected by the subsystem capacities, the selection of an RRM algorithm, and 

the use of available radio resources. Both, single operator scenarios as well as multi-operator scenarios 

can benefit from a framework that supports a joint centralized and distributed approach to resource 

sharing. If the decision comes from the spectrum manager or spectrum server entities located higher in the 

network, the delays associated with the signalling will interfere with the timely assignment of resources 

during radio access. Therefore, the spectrum assignment should be performed as a distributed approach. 

Interworking between entities for network resource management and entities for spectrum management 

can provide a two-fold gain (i.e., trunking and diversity gain). The performance gain in terms of diversity 

would depend on the maximum acceptable distance between fragmented bands that are aggregated; in this 

context, it would be useful to investigate the upper limits on the separation and the achievable gain. 

4. Spectrum Aggregation 

As illustrated in Figure 4-1 and Figure 4-2 spectrum aggregation can appear when operator’s dedicated 

DL or UL band is not continuous but is split in two or more parts. In addition, spectrum aggregation can 

happen in scenarios in which an operator accesses both dedicated band and a spectrum sharing band 

which is separated in frequencies from the dedicated operator’s band. The separated bands can be situated 

in the same frequency band (e.g., in C-band) or in separated bands (e.g., part in 2GHz band and part in C- 

band). Both options are illustrated in the case of DL in Figure 4-1 and Figure 4-2. More detailed options 

on spectrum sharing can be found in [R4051146]. 

DL 

DL 

frequency 

Figure 4-1: Spectrum aggregation of operator’s dedicated bands in the same band, e.g., C-Band 



DL 

DL 

frequency 

Figure 4-2: Spectrum aggregation of operator’s dedicated bands in two bands, e.g., 2 GHz band 

and C-band 

Note that the scenario depicted in Figure 4-2, i.e., spectrum aggregation from different frequency bands, 

represents a more complex solution and should be avoided if possible. 

The main open questions considering spectrum aggregation are: 

• What is the maximum acceptable distance between fragmented bands that are aggregated at the 

receiver? Does this constraint come due to hardware or some other restrictions? 

• How does the complexity grow with the number of fragmented bands aggregated at the receiver? 

What is the upper limit on the number of aggregated fragmented bands? 

4.1 Spectrum Aggregation with Multi-Band User Allocation over Two Frequency 

Bands 

Widely separated bands can offer a significant level of diversity due to very different propagation 

characteristics of the radio waves [LEE93]. These show independent channel quality indicators (CQIs) 

over time and space. This is a source of diversity at the physical (PHY) layer, which offers an important 

chance to achieve better quality of service (QoS) and spectrum efficiency [MEU09]. 

The focus of the following investigation is on the aspects of how one operator can manage the user 

allocation over the dedicated and shared bands with the objective to achieve higher throughput. The 

evaluations are performed for a single RAT, which is assumed to be High Speed Downlink Packet Access 

(HSDPA), which was proposed as an enhancement to 3G by 3GPP. Because the goal is to prove that 

integration of dynamic spectrum use and RRM techniques leads to an overall increase in performance 

gains, the choice of RAT is not of consequence for the final results. Spectrum sharing mechanisms are 

beyond the scope of the investigation here, and therefore, it is assumed that the operator considered here 

has gained access to the frequency pool with a certain portion of the available spectrum. Once a part of 

the spectrum has been obtained, the operator still faces the problem of allocating users on both bands. 

Depending on the capabilities at the UEs, each user could be allocated to a single frequency band or to 

both frequency bands. In the latter case, the UEs have multi-radio transceivers and can transmit and 

receive data on both bands. Here, focus is only on single-band UEs that need to be allocated over two 

possible bands. 

4.1.1 Problem Statement 

The objective is to determine the best user allocation for a single operator over two (or more) frequency 

bands in order to maximize the total network throughput. Two bands are analyzed. The operator has 

exclusive usage of the 2-GHz band and can access to the shared frequency pool at 5 GHz. The quantity of 

radio resources available at 5 GHz is determined by spectrum trading (or bargaining) among all the 

operators that have been granted access to the frequency pool. The performance gains are analyzed in 

terms of data throughput. 

After the operator has gained access to a certain portion of the frequency pool there is still the problem of 

allocating users in both bands. The performance is heavily dependent on the radio channel qualities for 

each user in the considered bands. The CQI depends on the path loss and on the distance from the BS. 

The operator will have good improvements when the UEs have heterogeneous spatial distribution in the 

cell (variable distances from the BS), i.e., different channel qualities in the considered spectrum bands. 

The problem of scheduling the users into two bands (2 GHz and part or all of the frequency pool at 5 

GHz) can be formulated as an Integer Programming (IP) optimization problem [KEL05], [KAR05]. The 

total throughput of the operator is the profit function to be maximized. QoS requirements are not 

considered at this stage. The profit function depends on the channel qualities and on the number of users 

in each band. Because the goal is to determine the user allocation, the optimization problem cannot be 

solved directly. Therefore, the problem is formulated in two steps: first, the number of users that can be 

allocated in the primary band (i.e., 2 GHz) are determined based on the load thresholds; then the multiband 

scheduling is determined as a General Assignment Problem (GAP) where the number of users in 

each band is upper bounded. 

The scheduling algorithm is implemented as a reduced-complexity sub-optimal allocation. 



4.1.2 General Multi-band Scheduling Spectrum Sharing as a General Assignment 

Problem 

4.1.2.1 Profit Function 

The scheduling of users over multiple frequency bands can be modelled in its most general form as a 

General Multi-Band Scheduling (GMBS) problem, i.e., GAP problem. 

Optimization problems can have single or multiple objectives. Here, the objective is to maximize the total 

throughput of the operator without considering fairness and QoS requirements of the service classes. 

Other objectives can be easily implemented in the problem, such as the cost or profit functions, e.g., 

maximizing the total rate while minimizing the QoS dissatisfaction indexes [MEU08]. Solving Multiple- 

Objectives GAP (MO-GAP) can be very difficult and usually the objectives are combined together via a 

linear combination, called “scalarization”. 

The profit function (PF) depends on the ratio between the service request and the goodput available for 

each user (on each band). Therefore, the problem is formulated with a load constraint for each 

band: 

The PF is thus defined considering the ratio between the requested rate by the service flow and the rate 

available on a single downlink channel. This weight accounts for a real usage of the capacity considering 

the source traffic generator. The PF to be maximized is the following: 

with 

(4-1) 

, (4-2) 

where is the service goodput request. is the throughput for user u on band b, and is an 

0-1 integer variable defining the user allocation over the bands that is mathematically 

represented by: 

For a single transceiver UEs, the GMBS has two constraints: 

1) Each user can be allocated only to a single frequency band with a single allocation channel. This 

results in an Allocation Constraint (AC); 

2) The total number of users on each band is upper bounded by the maximum load that can be handled 

in the band, i.e., the Bandwidth Constraint (BC) 

. 

, (4-3) 

. (4-4) 

In general, the throughput is a function of the channel quality of user u on band b: , 

where is dependent on the Signal to Interference Ratio (SIR) that depends on the propagation 

conditions, on the interference from other cells and on power per code available: 

. 

The higher the number of users allocated in the band, the lower the CQI. This is mainly due to the power 

division by the used codes, which are using completely orthogonal codes. The decrease in the CQI that 

occurs when more users arrive to the system does not reflect the throughput decrease per user. Rbu can be 

changed ahead to depend on other variables, however, for simplicity, here it is dependent only on the 

CQI. 



The GMBS problem is formulated into two steps. First, the maximum number of users that can be 

allocated on each frequency band is determined, then (i.e., throughput for user u on band b) is 

defined as the available throughput considering that users are allocated in the band. This is clearly a 

lower bound because 

and there is a lower level of interference. This allows for defining 

only as a function of the link budget, 

, and to proceed with the problem solving. 

The number of maximum users in each band is calculated depending on the load, as follows: 

, 

where L normalized is estimated based on the resources available for the cell. The normalized load is 

estimated as follows: 

L 

normalized 

( i) 

Nb 

∑ 

u= 

= 1 

Load 

where N b is the number of HSDPA users in band b, R HSDPA is the number of High Speed Downlink Shared 

Channels (HS-DSCH) allocated in the cell, and Load(u) is the average number of HS-DSCH required by 

user u to support its service rate, R(u). This number is given by the following 

Load 

( u) 

= 

R( 

CQI 

bu 

R 

R 

) • N 

HSDPA 

( u) 

( u) 

HS−PDSCH 

( CQI ) 

bu 

where, the average propagation condition determines the channel quality indicator ID of user u on band b, 

, R( ) is the achieved bit rate when one block is allocated in every frame and N HS- 

PDSCH( ) is the number of HS-DSCH associated to . 

Once has been determined, R bu , or the expected data rate for each user should be dependent on 

CQI bu , the effective packet error rate (PER) experienced (or predicted based on the position in case of a 

first transmission), and the number of users: 

The direct mapping between CQI bu and the throughput can be expressed as: 

4.1.2.2 Resource Allocation 

The resource allocation (RA) allocates the user packets to the available radio resources in order to satisfy 

the user requirements, and to ensure efficient packet transport to maximise spectral efficiency. The RA, 

an entity within the set of RRM algorithms, should have inherent tuning flexibility to maximise the 

spectral efficiency of the system for any type of traffic QoS requirements. The RA adopted here maps 

packets of variable size into variable length radio blocks for transmission over the PHY layer, and the 

length is dependent on the channel quality. The following events are performed: 

1. User packets awaiting transmission are prioritized according to the scheduling algorithm criteria. 

2. A CQI identifier is selected according to the link adaptation algorithm, using the available CQI 

options from the PHY layer. 

3. The scheduler calculates the number of MAC transport blocks required to transmit the scheduled 

packet. The number of HS-PDSCH channels is calculated according to 

where is the lowest integer higher or equal to x; 

4. An idle ARQ channel j is selected to hold and manage the ARQ transmission. 

5. The packet is transmitted and received at the UE. Soft retransmissions are combined with previous 

packet transmissions (chase combining) and the ARQ messages are generated accordingly. These 

are then signalled to the BS, and the ARQ processes are released if the messages are positive 

acknowledgments (ACKs). 



4.1.2.3 Differences between the 2-GHz and 5-GHz frequency bands 

The 2-GHz and 5-GHz frequency bands are characterized over the same HSDPA architecture by 

assuming the models summarized in Table 4-1: 

Table 4-1: Models used for 2- and 5 GHz 

Carrier frequency 2 GHz 5GHz 

Bandwidth 5MHz 5MHz 

Path loss model: 128.1 + 37.6 Log 10 (R) 141.52+28log 10 (dkm) 

Shadowing decorrelation 

5 m 20 m 

length 

4.1.3 Suboptimal Multiband Allocation Algorithm 

UEs are able to use both of the frequencies. The CRRM entity keeps track of the CQI in both frequencies 

by making use of the pilot channel. By keeping the load at the same level in both frequencies the users 

may be scheduled on one frequency or another, depending also on the CQI available for the UE. When a 

user arrives to the system, the UE is allocated to the 2-GHz band. The load is checked in both bands. If 

the load is higher in the 2-GHz than in the 5-GHz one, the user with the highest CQI in the 2 GHz will be 

moved to the 5-GHz band. In a situation of higher network loads, the users with the lowest CQIs will be 

allocated to the non-shared 2-GHz band while the users with highest CQIs will be allocated on the shared 

band. The load is estimated based on the resources available for the cell, and actually consumed by user 

connections. The procedure is the same as the one described in [MON08]. 

4.1.4 Results 

The performance of the algorithm is assessed by using the service throughput that is the total number of 

bits that have been transmitted and correctly received by the all users in the cell: 

where b serv [p] is the number of bits received in given period p, T is the transmit time interval, k is the 

number of steps, and k⋅T is the simulation duration. Users are displaced in the cell within a distance from 

the BS from 300 to 3000 km with a uniform distribution. The NRTV calls are modelled by a Poisson 

distribution, the call duration is exponentially distributed with an average of 180s. 

4.2 Basic Transceiver Concepts for aggregated Spectrum 

Future mobile and wireless communications will provide aggregated total throughput data rates up to 100 

Mbit/s for new mobile access and up to 1 Gbit/s for new nomadic local area access [M.1645]. The 

expected carrier bandwidth from technical reasons for these data rates is in the order of 100 MHz 

[Moh03]. However, the result WRC 2007 may not allow such wide carrier bandwidth with contiguous 

frequency bands for new radio systems. In this section a generic theoretical investigation is presented, 

which shows the basic impact of different parameters and topics affecting the system design and 

performance and the associated problems. The technical feasibility from the perspective of the 

infrastructure (base stations) and terminal devices is being assessed in the conclusions. 

4.2.1 Approach 

Today, several frequency bands are allocated to mobile and wireless communications. WRC 2007 

discussed the potential identification of new frequency bands to mobile and wireless communications. In 

the following the basic idea of use of fragmented spectrum bands is presented without detailed figures on 

the today’s allocated frequency bands. 

The approach is based on a joint simultaneous use of different frequency bands in the frequency range of 

about 400 MHz up to about 6 GHz including the unlicensed frequency bands for WLAN applications. 



Figure 4-3 shows the basic map of frequency bands for mobile and wireless applications excluding 

broadcast spectrum, because this is currently not available for such applications. 

Potentially 

available 

frequency 

bands 

1 2 3 4 5 6 

f [GHz] 

Figure 4-3: Basic allocated frequency band for mobile & wireless applications (qual. presentation) 

In the following it is assumed that the transceiver is processing the entire available frequency spectrum 

comprising the different part bands. There are now two basic concepts for a wideband communication 

system, which will be described hereafter in more detail: 

• Multi-band transmission system with several parallel transceivers for the different frequency 

bands, which are used simultaneously. 

• Wideband receiver, which is processing the entire frequency band from about 400 MHz to about 

6 GHz. The filtering should be done in the digital domain. 

The two basic transceiver concepts are shown in Figure 4-4 and Figure 4-5. The block diagrams are rather 

high-level in order to discuss the major impacts. Both concepts are explained only for the receiver 

section. 

RF 

bandpass 

per band 

RF 

frontend 

A 

D 

RF 

frontend 

A 

D 

Digital 

signal 

Detected signal 

processing 

RF 

frontend 

A 

D 

Figure 4-4: Multi-band received with parallel receivers for different bands 



Broadband 

Antenna 

Broadband 

RF 

frontend 

A 

D 

Digital 

signal 

processing 

Detected signal 

RF 

bandpass 

0.4 to 6 GHz 

Figure 4-5: Wideband receiver for the entire band 

A first basic assessment of the different concept is as follows: 

Multi-band receiver: 

• The RF band pass filter per branch has a bandwidth corresponding to the allocated band. This 

filter is also needed as anti-aliasing filter. 

• The channel filtering per link will be done in the digital domain. 

• This concept will be feasible. It is similar to today’s multi-band terminals. 

• Several RF branches are needed, which need volume (RF filters). 

• The dynamic range should be feasible due to the limitation to useful bands and the avoidance of 

adjacent disturbing signals as much as possible due to RF filtering. 

• However, this concept is rather complex. The different RF chains require filters, amplifiers, 

mixers and local oscillators. 

• Potentially, amplifiers, mixers and local oscillators can be re-used for different part-bands by 

changing RF-filters and local oscillator frequencies. 

• The number of parallel available RF chains corresponds to the number of part bands, which are 

simultaneously used. 

Wideband receiver: 

• The RF band pass is needed as anti-aliasing filter for wideband sampling. Its bandwidth covers 

the entire receiver bandwidth. 

• The channel filtering per link will be done in the digital domain. 

• This concept requires very wideband RF components. 

• A very high sampling rate is needed. 

• A much higher dynamic range is necessary in order to mitigate the impact of strong adjacent 

disturbing signals (e.g. broadcast and radar signals), which are within the wide input RF 

bandwidth but outside of one of the useful frequency bands. 

• This requires A/D-converters with very high resolution and linearity. 

• The RF components have to be very linear and to be able to handle high-amplitude signals. 

• With respect to the high required sampling rate and dynamic range (A/D-converter resolution) 

this concept will not be feasible with today's available technology. 

There are two basic concepts for the implementation of the wideband receiver according to Figure 3: 

• Either the A/D-conversion is performed directly in the RF domain with down-conversion (mixing) 

and channel filtering in the digital domain (version A) or 

• the wideband RF signal is down-converted in an IQ-mixer in the analogue domain to the base band 

and the I- and Q-channel are separately A/D-converted according to the band pass sampling theorem 

(version B) [Lue75]. 

The two basic receiver implementations in the case of a general asymmetric input signal and band pass 

transfer function with respect to the center frequency show Figure 4-6 and Figure 4-7 [Lue75]. 



Broadband 

Antenna 

RF 

bandpass 

0.4 to 6 GHz 

Broadband 

RF 

frontend 

A 

D 

cos(2πf LO t) 

LO 

-sin(2πf LO t) 

h Tr (t) 

h Ti (t) 

-h Ti (t) 

h Tr (t) 

+ 

+ 

I 

Q 

Complex detected signal 

Digital signal processing 

Figure 4-6: Potential receiver implementation: Version A 

Broadband 

Antenna 

RF 

bandpass 

0.4 to 6 GHz 

Broadband 

RF 

frontend 

cos(2πf LO t) 

LO 

-sin(2πf LO t) 

A 

A 

D 

D 

h Tr (t) 

h Ti (t) 

-h Ti (t) 

h Tr (t) 

+ 

+ 

I 

Q 

Complex detected signal 

Figure 4-7: Potential receiver implementation: Version B 

Digital signal processing 

4.2.2 Brief Investigation of Several Topics Affecting the System Design 

The following investigation is mainly made for the wideband concept. The figures can easily be 

transformed to the multi-band concept. However, due to the different band pass filters the impact of 

interference and other issues is basically smaller than in the wideband concept. 

4.2.2.1 Frequency Dependent Path Loss 

The frequency dependence of the path loss L(f c ) [dB] on the carrier frequency f c depends on the actual 

propagation conditions. For free-space propagation the frequency dependence corresponds to [MG86] pp. 

H12 and H30. 

⎛ fc,2 

L( 

fc, 2) 

L( 

fc,1) 

20 log [ dB] 

f ⎟ ⎞ 

− = ⋅ ⎜ 

(4-5) 

⎝ c,1 

⎠ 



With increasing carrier frequency the path loss is increasing quadratic. For shadowing propagation the 

increase of path loss is bigger with about 25 – 30 dB/decade for suburban areas and about 30 dB/decade 

for urban areas [Pa92][Ra96][Ha80][WB88]. 

⎛ fc,2 

L( 

fc, 2) 

L( 

fc,1) 

30 log [ dB] 

f ⎟ ⎞ 

= ⋅ ⎜ 

⎝ c,1 

⎠ 

− (4-6) 

For some propagation conditions this increase may be even bigger. Figure 4-8 shows this dependence in 

the frequency range 0.4 to 6 GHz. The relative path loss is normalized to 400 MHz. 

L(f c,2 ) – L(f c,1 ) [dB] 

40 

35 

30 

25 

20 

15 

10 

5 

0 

≈ 

0.4 1 2 3 4 5 6 

Free space 

Shadowing 

f c [GHz] 

Figure 4-8: Relative path loss versus carrier frequency 

In the upper frequency bands the radio channel shows a significantly higher path loss. This results directly 

in a significant reduction of range for the upper bands. Within the bandwidth under discussion a huge 

difference in path loss has to be considered. 

4.2.2.2 Doppler Frequency and Doppler Spectrum 

Due to moving receivers, transmitters and/or movements in the environment the radio channel is timevariant. 

This is described by the Doppler shift f d or in the case of multi-path propagation by a Doppler 

spectrum S E (f), which represents the different Doppler shifts from different received waves from different 

directions [MG86][Ja74]. Eq. (4-7) shows the Doppler shift versus mobile speed ν, the carrier wave 

length λ, the speed of light c 0 , the angle of arrival θ between the mobile direction and the incoming wave: 

v v 

f 

d 

= ⋅cos θ = ⋅ fc 

cosθ 

= f 

d , m 

⋅cosθ 

. (4-7) 

λ c 

The maximum Doppler shift is 

f 

v 

= ± ⋅ f 

0 

d , m 

c 

(4-8) 

c0 

with the Doppler spectrum under the assumption of a uniform distribution of incoming waves from all 

directions 

SE 

z 

( f ) = 

π ⋅ f 

d , m 

1 

1.5 

⎛ 

− ⎜ 

⎝ 

f − f 

f 

d , m 

c 

⎞ 

⎟ 

⎠ 

2 

for 

− . (4-9) 

f 

d , m 

≤ f − fc 

≤ f 

d , m 

Outside of this range the Doppler spectrum is zero. The maximum relative Doppler shift in the frequency 

range 0.4 to 6 GHz is directly proportional to the ratio of carrier frequencies (Figure 4-9): 

f 

f 

f 

d , m2 

c2 

= . (4-10) 

d , m1 

fc 

1 



f d,m2 /f d,m1 14 

12 

10 

8 

6 

4 

2 

0 

0.4 1 2 3 4 5 6 f c [GHz] 

Figure 4-9: Relative maximum Doppler shift 

In the upper frequency bands the radio channel shows a significantly higher time variance than in the 

lower bands, which requires a higher update rate for adaptation algorithms such as equalization and 

power control. This may result in higher overhead in the upper bands for signalling. 

4.2.2.3 Effective Noise Power 

The thermal noise density power at an absolute temperature of 300 K corresponds to N 0 = -174 dBm/Hz 

[MG86]. Including the receiver noise figure F the effective noise power within the receive bandwidth B 

follows as: 

N eff 

= N 0 

⋅ B ⋅ F . (4-11) 

With a receiver noise figure F = 5 dB and the entire receive bandwidth B = 6 GHz – 0.4 GHz = 5.6 GHz 

the total noise power is given by 

N 

= N dBm/ 

Hz] 

+ 10⋅ 

Log( 

B )[ dBHz] 

+ F[ 

dB] 

= 71. dBm . (4-12) 

[ 

eff , total 0 total 

− 5 

From the perspective of terminal devices a noise figure of F = 5 dB is not feasible with reasonable cost. 

The receiver noise figure is frequency dependent due to the frequency dependent matching, which results 

in increasing effective noise figure with increasing relative receiver bandwidth. Base stations can provide 

lower noise figures than terminal devices. The noise power can directly be scaled by other noise figures. 

This noise power is significantly below the power levels of the receiver, where nonlinear effects might be 

expected. The noise power per band and the noise power density needs to be investigated. 

Under the assumption that the smallest signal bandwidth after base band filtering equals B min.ch = 1.25 

MHz the effective noise power follows as 

N N [ 

eff , 1.25MHz 

0 

dBm/ 

Hz] 

+ 10⋅ 

Log(1.25MHz)[ 

dBHz] 

+ F[ 

dB] 

= −108. 

dBm 

= . (4-13) 

This might be the smallest noise power level, which has to be considered for the dynamic range of the 

receiver. 

4.2.2.4 Receiver Input Signal 

The receiver input signal is rather complex. After the RF band pass, which is also needed as anti-aliasing 

filter, useful signals in different part-bands and disturbing signals from other services between the 

different part-bands are present. Such disturbing signals may have very high transmit power such as 

broadcast and radar signals. 

The RF pre-selection band pass filter transfers all receive signals u r (t) according to (4-15) within the RF 

reception band according to its transfer function. 

All signals out of the RF reception band u r,out (t) are attenuated with respect to the band pass characteristic. 

– These signals are not considered in the following. – 

Useful receive signals within part-band j: 



u 

( ω ⋅t 

+ ϕ ( t ) 

I ( j) 

I ( j) 

r, band , j 

( t) 

= ∑ur, 

j, 

i 

( t) 

= ∑ar, 

j, 

i 

( t) 

⋅sin 

r, 

j, 

i r, 

j, 

i 

) 

i= 1 

i= 

1 

Useful receive signals in all part-bands: 

u 

( ⋅t 

+ ( t ) 

J 

J I ( j) 

J I ( j) 

r, useful 

( t) 

= ∑ur, 

band , j 

( t) 

= ∑∑ur, 

j, 

i 

( t) 

= ∑∑ar, 

j, 

i 

( t) 

⋅sin 

r, 

j, 

i 

ϕr, 

j, 

i 

) 

j= 1 j= 1 i= 1 j= 1 i= 

1 

ω (4-14) 

Disturbing signals between part-bands: 

u 

( ω ⋅t 

+ ( t ) 

K 

K 

r, disturbing 

( t) 

= ∑ur, 

k 

( t) 

= ∑ar, 

k 

( t) 

⋅sin 

r, 

k 

ϕr, 

k 

) 

k = 1 k = 1 

The total receive signal after the receive RF band pass including noise is as follows: 

u ( t) 

= u 

r 

= 

r, 

useful 

+ 

J 

( t) 

+ u 

I ( j) 

∑∑ 

( t) 

= 

r, 

disturbing 

( t) 

= 

r, 

j, 

i 

j= 1 i= 1 j= 1 i= 

1 

K 

∑ 

u 

u 

K 

∑ 

r, 

k 

k = 1 k = 1 

a 

J 

r, 

k 

( t) 

+ n( 

t) 

= 

I ( j) 

∑∑ 

a 

r, 

j, 

i 

( t) 

⋅sin 

( t) 

⋅sin 

( ω ⋅t 

+ ϕ ( t) 

) 

( ω ⋅t 

+ ϕ ( t) 

) 

r, 

k 

r, 

j, 

i 

r, 

k 

+ 

r, 

j, 

i 

+ 

(4-15) 

+ n( 

t) 

Even if the sampling of the input signal is performed directly in the band pass domain without a mixing 

process to an intermediate frequency (IF), the signal to be detected has to be transformed to the base band 

domain in the digital signal processing. This does only correspond to one possibility of implementation. 

Therefore, issues like image rejection and reciprocal mixing etc. have to be considered in general. 

4.2.2.5 Nonlinearities in the Analogue Receiver Components 

In a simplified description the analogue active part of the receiver may be represented by a nonlinear 

characteristic (Taylor series): 

2 

3 

um ( t) 

= c0 + c1 

⋅ur 

( t) 

+ c2 

⋅ur 

( t) 

+ c3 

⋅ur 

( t) 

+ ... 

(4-16) 

Under overload conditions additional signals are generated due to distortion and intermodulation, which 

may disturb the demodulation process. 

• For one input signal a reduction of amplification due to nonlinear distortion for a cubic characteristic 

with increasing input signal amplitude (compression) occurs [Car90]. 

• For – in minimum – two input signals (useful plus disturbing signal) intermodulation products result 

in reduction of the output signal amplitude of the useful signal for increasing disturbing signal input 

amplitude (desensitization or blocking) [Car90]. 

• For multiple input signals intermodulation products are generated. 

The receiver nonlinearity can be characterized by the 2nd and 3rd order intercept points. 

4.2.2.6 Image Rejection 

The RF band pass is needed for some pre-selection with respect to the entire input signal and as image 

reject filter. However, due to the very complex input signal within the bandwidth of the RF band pass 

according to (4-15) disturbing signals with potentially high transmit power cannot be avoided directly by 

a band pass filter in front of the active receiver components (low noise RF amplifier, potential mixers). 

Such disturbing signals may be reduced by dedicated notch filters as part of the RF pre-selection filter. 

In the case of a mixing process with an IF f IF , different input signals at different receive carrier 

frequencies f r may lead to nonlinear distortion and harmonics of the local oscillator frequency f LO to this 

same IF-range. 

f 

IF 

= m⋅ 

f ± n⋅ 

f with m, n = 1, 2, 3 ... (4-17) 

r 

LO 

Therefore, the complex input signal may result in the active and usually nonlinear receiver components in 



• linear spurious reception due to the mixing process (m = 1) 

• image reception 

• reciprocal mixing 

• nonlinear spurious reception due to receiver nonlinearities (m > 1) 

• overload, cross modulation, limiting and blocking 

• possible spurious reception frequencies (one signal at receiver input) 

In the case of f LO > f IF and no overloading of the RF pre-amplifier and mixer (m = 1) two possible receive 

frequencies per local oscillator harmonic (image reception) may be observed: 

f 

= n ⋅ f 

− f 

r, n /1 LO IF 

and 

r n / 2 LO IF 

f 

, 

= n ⋅ f + f . (4-18) 

Usually, contributions resulting from local oscillator harmonics (n > 1) are negligible. Therefore, the RF 

pre-selection band pass filter has to avoid signals at frequencies 2 f IF apart from useful signals with (c.f. 

Figure 4-10): 

2 ⋅ f ≥ . (4-19) 

IF 

B RF 

f r,n/1 f LO f r,n/2 

f IF 

f IF 

receive band 

B RF 

Figure 4-10: Relation between B RF and f IF 

In the case of a very big RF bandwidth in the order of B RF = 5.6 GHz the IF has to be f IF > 2.8. The 

requirements on the pre-selection filter slope are relaxed for increasing f IF . For very wideband RF band 

pass filter no high stop-band attenuation can be expected. Therefore, the main channel selection with high 

stop-band attenuation is performed in the IF- or base band stage to reduce the effective receiver noise 

bandwidth and to avoid overload in the main IF amplification due to high power out of IF band signals. 

However, this may not be possible in the concept according to Figure 3, where the main filtering is done 

in the digital domain. Therefore, the analogue receiver has to show a high resistance against overload 

conditions. 



4.2.2.7 Receiver Performance (Desensitization, Blocking, Intermodulation) 

The basic performance of the receiver is summarized in Figure 4-11 [MG86]. 

level of input signal P E [dBm] 

D 

F' 

IPIP3 

E' 

IPIP2 

C' 

C 

A 

F E B 

minimum required 

signal/noise-ratio 

level of input disturbing signals in [dBm] 

Figure 4-11: Receiver performance 

A: Receiver sensitivity due to noise, no interference. 

B – C: Impact of reciprocal mixing, required useful signal has to be increased proportionally to 

additional noise. 

C – C': Limit due to cross modulation for modulated interfering signal. 

D: Maximum level of useful input signal for one disturbing signal with same amplitude, in-band 

intermodulation products and/or harmonics determine output signal/intermodulation-ratio. 

E – E': For two input signals 3rd order intermodulation products proportional to third power of 

disturbing signal level. 

F – F': For two input signals 2nd order intermodulation products proportional to second power of 

disturbing signal level (less important due to RF pre-selection). 

IPIP3: Construction of 3rd order intercept point. 

IPIP2: Construction of 2nd order intercept point. 

4.2.2.8 Reciprocal Mixing and Receiver Noise Figure 

The receiver is not only affected by thermal noise and receiver internal additional noise sources. In a 

system with several carrier input signals in front of the mixer reciprocal mixing especially in the RF stage 

of the receiver may play a significant role. Figure 4-12 shows the effect of reciprocal mixing [Car90]. 



Amplitude 

Desired 

IF signal 

(noiseless) 

• Differencemixingwith high-side 

injection, down conversion 

Amplitude 

DesiredRF signal 

Local oscillator 

IF passband 

( unmodulated and 

(noiseless) 

f 

noiseless) 

Noise in IF 

Desired 

passband caused 

IF signal 

by oscillator noise 

Noise 

( noisy) 

sidebands 

• Extension of abovewhen oscillator 

noise is significant 

Amplitude 

Noise 

sidebands 

DesiredRF signal 


IF passband 

(unmodulated and 

(noisy) 

noiseless) 

Additional noisein 

Noise 

IF passband caused 

sidebands 

by oscillator noise 

f 

• Reciprocalmixing occurswhen an 

undesired signal mixes with 

oscillator noise to produceadditional 

noise in IF passband 

Undesired 

IF signal 

( noisy) 

UndesiredRF signal 


(unmodulated and 

(noisy) 

noiseless) 

f 

Figure 4-12: Reciprocal mixing 

The three figures are explained in the following: 

• Top: Difference mixing with high-side injection and down conversion based on the difference 

frequency between the local oscillator and the received signal. 

• Middle: Extension of the top figure when the oscillator noise is significant. The phase noise is down 

converted towards the IF. 

• Bottom: Reciprocal mixing occurs when an undesired signal (adjacent channel) mixes with the 

oscillator noise to produce additional noise in IF pass band. 

Reciprocal mixing results in an additional noise figure F RM proportional to the amplitude of the disturbing 

signal (adjacent channel) within the RF pass band and the single side band oscillator phase noise [MG86]. 

However, F RM is independent of the receiver noise figure. 

F 

[ dB] P [ dBm] + ( f ) [ dBc / Hz] 174 [ dBm / Hz] 

= Φ (4-20) 

RM s 

e 

+ 

The additional noise components due to the oscillator phase noise increase the effective receiver noise 

figure with increasing number of adjacent disturbing carriers. The resulting receiver noise figure for N 

input signals follows as: 

⎪⎧ 

N 

F 

/ 10 

[ dB ] = 10log⎨10 

[ dB ] + ∑ 

⎪⎩ 

i= 

1 

F [ dB ] / 10 

F 

receiver RMi 

receiver,res 

10 

. (4-21) 

In the case of the wideband received according to Figure 4-5 there is a high likelihood for many high 

power disturbing signals. These relations have to be considered in the design of the RF filters and the 

local oscillator phase noise requirements. The relation in (4-20) is shown in Figure 4-13 [MG86]. 

⎪⎫ 

⎬ 

⎪⎭ 



adjacent carrier P s 

[dBm] 

40 

20 

0 

-20 

-40 

-60 

-80 

F RM 

= 140 dB 

120 

100 

80 

60 

40 

20 

0 

example point 

-40 -60 -80 -100 -120 -140 -160 ⎝ e 

(f) [dBc/Hz] 

Figure 4-13: Additional noise figure F RM due to reciprocal mixing 

When the receive level of an adjacent carrier is below -44 dBm and the oscillator phase noise reaches - 

130 dBc/Hz, the additional noise figure due to reciprocal mixing would be F RM = 0 dB. The effect of 

reciprocal mixing can be reduced further by oscillators with low phase noise and for low effective 

adjacent channel amplitude. 

The receiver chain noise figure (Friis formula) according to [MG86] follows as 

F 

receiver 

N 

Fadd 

,2 

Fadd 

,3 

, res 

= F1 

+ + + ∑10 

L L ⋅ L 

v1 

v1 

v2 

FRMi 

[ dB] 

/10 

... + 

(4-22) 

i= 

1 

with, e.g., 

• index 1 – antenna cable plus image reject filter, 

• index 2 – pre-amplifier and 

• index 3 – mixer (without reciprocal mixing). 

A high receiver sensitive requires 

• low noise first receiver stages (F 1 and F add,2 ) plus high amplification (L v1 ⋅ L v2 ) to reduce the impact 

of mixer conversion loss and 

• oscillators with low phase noise to reduce the impact of reciprocal mixing. 

The additional noise figure F RM due to reciprocal mixing increases with an increasing number N and 

amplitude of input signals independent of the receiver noise figure F receiver without reciprocal mixing. On 

the other hand a reduced impact of intermodulation is achieved through a low gain RF amplifier. 

Therefore, a trade-off between the receiver sensitivity and the impact of intermodulation has to be 

implemented for overall optimal performance. 

4.2.2.9 Band Pass Filters and Filter Slope for Stop-band Attenuation 

The following general description of the RF band pass filter is done for lossless filters. The feasibility 

concerning 

• insertion loss, 

• achievable stop band attenuation, 

• group delay distortion, 

• possible filter order, 

• filter size and 

• cost 

is not considered. 

The overall filtering is usually distributed between the base band, the IF and the RF domain. In the RF 

band the maximum receive power at the input of the RF low noise amplifier has to be sufficiently low to 

avoid blocking of the receiver. As shown in Section 4.2.2.10 very high input signal amplitudes from 

disturbing signals within the wideband band pass bandwidth have to be expected. 

The low pass (LP) and the band pass (BP) domain are related by the LP-BP-transformation with the band 

pass cut-off frequencies f c1 and f c2 [Zv67][Sa79], the normalizing frequency 



f 

n = fc1 

⋅ fc2 

, (4-23) 

the normalized band pass frequency 

f 

F BP = , (4-24) 

f 

n 

and the normalized low pass frequency for filter design 

F 

LP 

= 

F 

BP, 

c2 

1 

− F 

BP, 

c1 

⎛ 

⋅ 

⎜ F 

⎝ 

BP 

1 

− 

F 

BP 

⎞ 

⎟ 

⎠ 

. (4-25) 

The RF frequency range is allocated between f c1 = 0.4 to f c2 = 6 GHz. With an assumed reference RF 

bandwidth of 

f 

− f 5. GHz 

(4-26) 

c2 c1 

= 6 

the relative RF bandwidth with an average carrier frequency of 

fc 1 

+ fc2 

= 3. 2GHz 

(4-27) 

2 

is in the order of 175 %, which results in a very low Q-factor. The RF filter should be designed according 

to Figure 4-5 as band pass filter for the entire receive and transmit band. The channel filtering should be 

performed in the digital base band or possibly IF domain. Due to the very high relative bandwidth the RF 

band pass shows a very asymmetric attenuation slope for the upper and the lower bound with a different 

but low increase in stop band attenuation. 

Under the assumption that the maximum attenuation in the pass band at the band edge corresponds to 

3 dB, the RF band edge frequencies 0.4 and 6 GHz correspond in the equivalent LP domain to the band 

edge frequency f LP,c = 2.8 GHz corresponding to the normalized frequency F LP,c = 1. The following 

figures apply: 

f n = 1.549 GHz 

f c1 = 400 MHz F BP,c1 = 0.258 (4-28) 

f c2 = 6 GHz F BP,c2 = 3.873 . 

In Figure 4-14 the equivalent normalized LP frequency in the stop band is presented for the lower and the 

upper slope of the wideband RF band pass. 

F LP 

0 

F LP 

45 

40 

35 

30 

25 

20 

15 

Lower bound slope 

Upper bound slope 

10 

5 

0 

0 100 200 300 390 

Offset frequency [MHz] 

Figure 4-14: Normalized LP frequency for the lower and upper bound slope of the RF band pass 

At the upper bound a reasonable attenuation can only be achieved for very high frequency separation. 

Therefore, disturbing signals beyond 6 GHz have to be taken into account. In addition, the band pass does 



not serve as anti-aliasing filter. At the lower bound, a reasonable stop band attenuation can be achieved 

for a separation of several 100 MHz. However, also here disturbing signals below 400 MHz have to be 

considered. 

In the case of a Butterworth filter of the order n the attenuation follows the function: 

2⋅n 

( + ) 

a( F )[ dB] 

= 10 ⋅ log 1 . (4-29) 

LP 

F LP 

Figure 4-15 shows the achievable attenuation versus offset frequency from the band edge for a 

Butterworth filter of the order of n. Here a filter order n = 6 is assumed. 

Attenuation [dB] 

250 

200 

150 

100 

Lower bound 

Upper bound 

50 

0 

0 100 200 300 390 

Offset frequency [MHz] 

Figure 4-15: Attenuation at the lower and upper bound slope of the RF band pass 

The low attenuation at the upper bound will result is serious problem with disturbing signal above 6 GHz. 

4.2.2.10 Maximum Input Signal 

Within the RF bandwidth very strong disturbing signals such as broadcast of radar signals may occur, 

which potentially impact the receiver performance. Therefore, the receiver has to have a sufficient 

dynamic range in order to avoid overload conditions. 

In a simplified worst-case estimation a broadcast transmitter with the following parameters is assumed: 

• Transmit power: 20 kW 

• Broadcast transmitter antenna gain: 20 dBi 

• Terminal antenna gain: 0 dBi 

• Distance between transmitter and receiver: 300 m 

• Transmitter carrier frequency: 800 MHz 

In the case of broadcast transmitters the main lobe of the transmitter antenna is horizontally oriented. 

Therefore, close to the broadcast transmitter location a terminal device does see the broadcast antenna for 

negative elevation angles θ. This results in a lower effective antenna gain than in the main lobe. The 

effective transmitter antenna gain G t,eff (θ)


Elevation diagram 

θ 

Broadcast 

transmitter 

Distance d 

Transmitter antenna 

Terminal 

Figure 4-16: Interference scenario between a broadcast transmitter and a mobile terminal 

This scenario with the assumptions above results in a maximum receive power in the order of 

G t,i (θ=0) = 20 dBi 

G t,eff (θ) = 0 dBi 

P r 

2. 8mW 

P r 

, max 

= corresponding to 4.5 dBm 

, max 

= 28μW 

corresponding to -15.5 dBm 

(4-31) 

at the terminal. However, blocking levels of terminal device receivers are usually in the order of -40 to - 

20 dBm. Under these conditions the receiver should not show desensitization or blocking as well as 

nonlinear distortion. However, without RF-filtering to reduce such high power interfering signals the 

blocking levels would significantly be exceeded even with lower effective transmitter antenna gain 

G t,eff (θ). 

4.2.2.11 Sampling Theorem for Both Receiver Versions 

Figure 4 presents the two versions of the receiver implementation: 

• Version A with direct sampling in the RF domain, 

• Version B based on the band pass sampling theorem. 

Both versions have a different impact on the sampling rate. 

Under the assumption of an ideal rectangular RF band pass as anti-aliasing filter the following sampling 

rates are necessary [Lue75]. – However, such filters are not feasible and with respect to Section 4.8 the 

attenuation at the upper bound is very small. – 

• Version A: Maximum signal frequency, which needs to be sampled in the RF domain 

corresponds to f max = 6 GHz. This results in a minimum sampling rate of 12 G-samples/s for real 

signal. In this case a A/D-converter with a sampling rate of 12 G-samples/s is required. 

• Version B: According to the band pass sampling theorem for a band pass bandwidth B = 5.6 

GHz the sampling rate per I- and Q-channel corresponds to two times the RF bandwidth B. 

Therefore, the sampling rate for the complex base band signals corresponds to 5.6 G-samples/s 

or to 5.6 + 5.6 G-samples/s for the I- and Q-channel. In this case two A/D-converters with a 

sampling rate of 5.6 G-samples/s each are needed. 

These high necessary sampling frequencies are theoretical values under the assumptions above for very 

wideband transceivers. It is shown in the following that such requirements are technically not feasible. 

4.2.2.12 A/D-Converter Dynamic Range and Output Data Rate 

The A/D-converter has a limited resolution. It generates quantization noise. According to [TS80] the 

signal/noise-ratio between the effective values of a sinusoidal input signal U s,eff and the triangular 

quantization error U q,eff for a resolution of N bits is given by: 

S 

N 

quantization 

U 

s, 

eff 

= 20⋅log 

[ dB] 

= N ⋅6dB 

+ 1.8dB 

. (4-32) 

U 

q, 

eff 



The quantization noise should be below the effective receiver noise level in order not to increase the 

effective noise figure. With respect to [MG86] the A/D-converter dynamic range D ADC follows under this 

condition as 

D ADC 

= N ⋅6dB 

− (5 to 10) 

dB . (4-33) 

With respect to the maximum input power according to (4-31) and the minimum noise power for narrow 

band signals in (4-13) a minimum dynamic range of 112.5 dB is needed. Under the assumption that for 

the narrow band signal with 1.25 MHz carrier bandwidth (like TD-SCDMA, cdma2000) spread spectrum 

modulation may be applied, receive signals below the noise level may be detected using processing gain. 

Then the required dynamic range is in the order of 

D ADC 

= 120 to 130dB 

. (4-34) 

This corresponds to a required A/D-converter resolution of 21 to 24 bit. 

This results with Section 4.10 in the following data rates for the output signal of the A/D-conversion 

before digital signal processing: 

• Version A: Minimum sampling rate of 12 G-samples/s for real signal with an A/D-converter 

resolution of 21 to 24 bit provides a data rate for digital signal processing of 

252 to 288 Gbit/s (31.5 to 36 Gbyte/s) . (4-35) 

• Version B: Minimum sampling rate for the complex base band signals corresponding to 5.6 + 5.6 G- 

samples/s for the I- and Q-channel provides a data rate for digital signal processing of 

117 + 117 to 134 + 134 Gbit/s (2 ⋅ 14.6 to 2 ⋅ 16.75 Gbyte/s) (4-36) 

Such high data rate signals would need to be processed in the receiver. Such processing rates are not 

feasible with today's technology. In addition, the A/D-converter power consumption would be huge and 

not be available in terminal devices. 

4.2.2.13 A/D-Converter Performance Development 

There is a relation between the sampling rate and the resolution of the A/D-converters. Figure 4-17 shows 

the currently (status 2006) valid limiting curve between the numbers of bits versus sampling rate. With 

respect to the requirements in Sections 4.10 and 4.11 there will be no A/D-converters in the foreseeable 

future available, which will support 

• a sampling rate in the order of several GHz and 

• a dynamic range in the order of 120 to 130 dB. 

Basic limitations for the progress of the development of A/D-converters are also confirmed by Analog 

Devices, Inc. in [BH06]. According to [BH06] no significant progress is expected in the near future. New 

processes for semiconductor technology will be needed. 



Figure 4-17: Resolution-bandwidth trade-off 

For a potentially expected channel bandwidth of about 100 MHz with an ideal rectangular channel filter 

for anti-aliasing purposes in minimum 100 M-samples for the complex base band signal or 100 + 100 M- 

samples for the real and imaginary component are required. Such sampling rates correspond with Figure 

4-17 to in maximum 11 bits and thereby in a A/D-converter dynamic range in the order of 71 to 76 dB. 

Therefore, adjacent carrier interference has to be avoided in front of the A/D-converter by using a channel 

bandpass filter in the IF-domain in order to reduce the necessary dynamic range. In addition, an AGC 

(Automatic Gain Control) in the RF receiver in the IF-domain after channel filtering is required. 

In conclusion, if fragmented frequency spectrum should be applied in order to support wideband systems 

with a total channel bandwidth in the order of 100 MHz, a receiver concept according to Figure 4-4 is a 

feasible solution. A wideband receiver according to Figure 4-5 will not be feasible due to the huge 

necessary dynamic range. In the ideal case the necessary channel bandwidth should be available in a 

single consecutive frequency block. 

4.2.3 Conclusions 

In this document the basic concept of a very wideband receiver for the use of fragmented frequency 

spectrum for broadband application for IMT-Advanced systems has been theoretically investigated under 

the assumption of a very wideband and flexible receiver to the entire frequency band from about 400 

MHz to about 6 GHz. It is not the intention of this Chapter to provide any final decisions. This Chapter is 

mainly summarising such major issues, which need further evaluation such as: 

• Frequency Dependent Path Loss 

• Doppler Frequency and Doppler Spectrum 

• Effective Noise Power 

• Impact of receiver nonlinearities due to very high disturbing input signals within the receive 

band 

• RF band pass filtering for image rejection, anti-aliasing and rejection of strong disturbing signal 

within the RF receiver bandwidth 

• Estimation of required receiver dynamic range 

• Necessary dynamic range of the A/D-converter 

• Data rate for digital signal processing. 

These issues have to be evaluated for potential future implementation with respect to technology 

roadmaps for 

• RF receiver components, 



• A/D-converters and 

• available signal processing power 

with respect to implementation cost, form factor and power consumption. 

First investigations have already been made in the WINNER project [WIND22] under less challenging 

requirements. A more in-depth investigation will require significant effort to look at the different aspects. 

However, the available information on the expected performance of A/D-converters in terms of number 

of bits versus sampling rate provide a clear indication that due to the strict limitations for the achievable 

dynamic range and sampling rate only a receiver concept according to Figure 4-4 will be feasible. The 

impact of adjacent channels has to be avoided by channel filtering to reduce the necessary dynamic range. 

In the ideal case the number of receiver chains per transmission channel should be as small as possible; 

therefore, a single block of spectrum is desired. 

The concept of a very wideband receiver according to Figure 4-5 with very high sampling rates of several 

GHz and huge dynamic range or A/D-converter resolution, which may be affected by strong adjacent 

carrier signals, is not feasible with today's technology. 

Therefore, only a transceiver concept with several parallel RF chains for the different part bands 

according to Figure 4–4 is a feasible technical solution, if aggregated or fragmented spectrum use is 

required. 

4.3 Actual discussions in 3GPP and ITU-R on the implementation of identified 

frequency spectrum in WRC 2007 

WRC 2007 identified additional frequency spectrum for mobile and wireless applications. However, these 

identifications are not accepted on global level. There are regional differences. Regional regulators bodies 

such as CEPT ECC PT1 in Europe and system specification bodies like 3GPP are currently investigating 

options how to use these bands. WRC 2007 identified the following frequency bands for further 

consideration: 

• 450 – 470 MHz 

• 698 – 806 MHz 

1) 

• 2300 – 2400 MHz 

2) 

• 3400 – 3600 MHz 

3) 

The following remarks have to be taken into account for these bands: 

1) 

The whole band 698 – 960 MHz is not identified on a global basis for IMT due to variation in the 

primary mobile service allocations and uses across the three ITU Regions. In particular 698 – 806 

MHz in Region 2 and nine countries in Region 3, 790 – 862 MHz in Region 1 and other countries in 

Region 3 

2) 

3) 

This band is not available in Europe. 

In some countries identified via footnotes in the ITU-R Radio Regulations. CEPT ECC PT1 will also 

address the band 3600-3800 MHz for IMT. 

In the following the status of the discussion in ITU-R and 3GPP are summarised as basis for an 

evaluation of potential receiver implementations according to Section 4.2.2. 

4.3.1 Frequency arrangements in the ITU-R WP5D meeting in Geneva, Switzerland, from 

February 10 – 17, 2009 

The different identified frequency bands have been discussed: 

Figure 4-18 shows the progress of discussion on the frequency band 450 – 470 MHz. 



Figure 4-18: Figure Potential frequency arrangements in the band 450 – 470 MHz 

The arrangements D1 – D6 have been implemented by some countries taking into account their national 

conditions. The arrangement D7 can be implemented by countries that have the whole 450 – 470 MHz 

band available for IMT. This band is requiring flexibility in the implementation of frequency allocation of 

transmitters and receivers. 

In the frequency band 698 – 806 MHz different options are being discussed. There are two new 

frequency plans added to the discussion: 

• A3 is a placeholder for the band 790 – 862 MHz for IMT systems operating in FDD mode, which are 

using a reversed duplex direction in order to avoid interference at the upper band edge, e.g. with 

GSM. This frequency arrangement will be developed in ECC PT1 and it is a harmonized solution for 

European countries. 

• In the A4 band plan administrations can use the band for TDD, FDD or some combination of TDD 

and FDD. In this case administrations can use any FDD duplex direction (Figure 4-19). 

M Hz 690 700 710 720 730 740 750 760 770 780 790 800 810 

A4 

MS Tx 

or TDD 

Un-paired 

BS Tx 

or TDD 

BS Tx 

or TDD 

MS Tx 

or TDD 

698 716 728 746 

763 776 793 

Figure 4-19: Potential frequency arrangement A4 in the band 698 – 806 MHz 

The frequency band 2300 – 2400 MHz is getting particular interest in China. A TDD and a flexible 

FDD/TDD option are under discussion (Figure 4-20). It has been decided that the APT Wireless Forum 

(AWF) will study and develop frequency arrangements for the 2 300-2 400 MHz band in the Asia-Pacific 

Region. This arrangement will be available for the 6 th meeting of ITU-R WP5D in October 2009. 



Figure 4-20: Potential frequency arrangements in the band 2300 – 2400 MHz 

Similar to the discussion on the band 2300 – 2400 MHz in the frequency band 3400 – 3600 MHz an FDD 

and a TDD solution are under discussion (Figure 4-21). Details are not yet fixed such as 

• the size of the segments for the FDD uplink (MS Tx) and downlink (BS Tx). Basically, also one of 

the links (uplink or downlink allocation could disappear (i.e., zero width); 

• the size of the centre gap for duplex separation; 

• the arrangement of the segments (i.e., FDD UL and DL direction); 

• the use of the external bands (i.e., combination of any FDD pairing with the bands other than 3400 – 

3600 MHz); 

• the possible frequency arrangement for the simultaneous accommodation of both FDD and TDD 

arrangements. 

MHz 

3400 3600 

F1 

(1 ), ( 2) 

MS Tx 

BS Tx 

3400 3600 

F2 

TDD 

3400 3600 

Figure 4-21: Potential frequency arrangements in the band 3400 – 3600 MHz 

The following notes have to be considered for the time being before final decisions are made: 

• NOTE 1 – MS Tx and/or BS Tx segments could be paired with the external bands. 

• NOTE 2 – The horizontal arrow pointing both directions between the “MS Tx” and “BS Tx” 

segments indicates that the size of segments for DL and UL and subsequently that of the centre gap 

and duplex separation, would be determined in the later stage. 

4.3.2 Status of discussion on the implementation of identified additional IMT frequency 

bands in the 3GPP TSG-RAN meting in Athens, Greece, in February 3 – 13, 2009 

According to TS 36.104 in chapter 5.3, E-UTRA is designed to operate in the following bands as shown 

in Table 4-2. These bands are partly the basis for considerations in 3GPP on the implementation of 

identified frequency bands. In particular 3GPP is investigating already allocated frequency bands for 

mobile and wireless communications and newly identified frequency bands. 

Table 4-2: E-UTRA frequency bands 

E_UTRA 

Band 

Uplink (UL) 

BS receive 

Downlink (DL) 

BS transmit 

Duplex 

Mode 

UE transmit 

UE receive 

F UL_Low – F UL_high 

F DL_Low – F DL_high 

1 1920 MHz – 1980 MHz 2110 MHz – 2170 MHz FDD 










9 1749.9 MHz – 1784.9 MHz 1844.9 MHz – 1879.9 MHz FDD 


11 1427.9 MHz – 1452.9 MHz 1475.9 MHz – 1500.9 MHz FDD 




… 


… 

33 1900 MHz – 1920 MHz 1900 MHz – 1920 MHz TDD 








3GPP RAN WG4 is currently discussing several implementation concepts for IMT-Advanced according 

to Table 4-3. These scenarios are partly using already allocated frequency bands to mobile wireless 

services and newly identified frequency bands in WRC 2007. Some of the scenarios are not using all the 

identified and/or allocated frequency bands based on the available information from 3GPP. 3GPP is 

proposing to limit pairing to a maximum of three bands per active connection/terminal in order to limit 

the overall implementation complexity (e.g., scenario 10). 

In the following sections the different scenarios are briefly analysed to understand their feasibility and 

potential constraints. 

Scenario 

No. 

Deployment 

Scenario 

1 Single-band 

contiguous spectrum 

allocation @ 3.5GHz 

band for FDD 

Scenario 1 

Table 4-3: 3GPP spectrum implementation scenarios 

Transmission 

BWs of LTE- 

A carriers 

UL: 40 MHz 

DL: 80 MHz 

No of LTE-A component 

carriers 

UL: Contiguous 2x20 MHz CCs 

DL: Contiguous 4x20 MHz CCs 

Bands for LTE-A 

carriers 

Duplex 

modes 

3.5 GHz band FDD 

Figure 4-22 illustrates this scenario. The following conditions and comments are relevant: 

• There is some relation to the F1 scenario in ITU-R (Figure 4-21). 

• FDD duplex scheme. 

• In the given frequency band only one UL and one DL carrier is possible. 80 MHz are not used. 

• Basically, there is flexibility, where to allocate the UL band within 3400 – 3500 MHz and the DL band 

within 3500 – 3600 MHz. 

• In order to support competition infrastructure sharing will be needed. 

• The implementation is feasible with a single transceiver, because a contiguous frequency band is used. 

• The maximum sampling bandwidth is 80 MHz. Technically, this will be feasible to implement. 



UL 

DL 

3300 3400 3500 3600 3700 

MHz 

Figure 4-22: Potential band and carrier allocation in scenario 1 

CEPT is considering to extend the frequency band up to 3800 MHz in CEPT countries. Under these 

conditions additional uplink and downlink FDD carriers could be added as long as the bandwidth of the 

uplink and downlink carrier bandwidth remains 40 MHz and 80 MHz. The carriers in this case could be 

allocated as follows: 

• Uplink: 

o Carrier 1: 3400 – 3440 MHz 

o Carrier 2: 3440 – 3480 MHz 

o Carrier 3: 3480 – 3520 MHz 

• Downlink: 

o Carrier 1: 3560 – 3640 MHz 

o Carrier 2: 3640 – 3720 MHz 

o Carrier 3: 3720 – 3800 MHz 

This would result in a much better spectrum usage than only providing the frequency band from 3400 – 

3600 MHz. 

2 Single-band 


allocation @ Band 40 

for TDD 

Scenario 2 

100 MHz Contiguous 5x20 MHz CCs Band 40 (2.3 

GHz) 

The frequency allocation in this scenario (Figure 4-23) corresponds to Band 40. 

• This corresponds to the E1 scenario in ITU-R (Figure 4-20). 

• TDD duplex scheme. 

• In the given frequency band only one UL and one DL connection is possible. The entire band is used. 




TDD 



UL/DL 

Band 40 

2200 2300 2400 2500 

MHz 


3 Single-band 100 MHz Contiguous 5x20 MHz CCs 3.5 GHz band TDD 



band for TDD 

Scenario 3 

Figure 4-24 shows scenario 3. 

• This corresponds to the F2 scenario in ITU-R (Figure 4-21). 


• In the given frequency band two UL and one DL carriers/connections are possible. The entire band is 

used. 

• In order to support competition some lighter approach for infrastructure sharing may be needed. 



UL/DL 

UL/DL 

3300 3400 3500 3600 3700 

MHz 


CEPT is considering extending the frequency band up to 3800 MHz in CEPT countries. Under these 

conditions two additional uplink and downlink TDD carriers could be added as long as the bandwidth of the 

TDD carrier bandwidth remains 100 MHz. The TDD carriers in this case could be allocated as follows: 

• TDD carriers up- and downlink: 

o Carrier 1: 3400 – 3500 MHz 

o Carrier 2: 3500 – 3600 MHz 

o Carrier 3: 3600 – 3700 MHz 



o Carrier 4: 3700 – 3800 MHz. 

This would result in a complete usage of the available frequency spectrum. 

4 Single-band, noncontiguous 

spectrum 


band for FDD 

Scenario 4 

UL: 40 MHz 

DL: 80 MHz 

UL: Non-contiguous 20 + 20 

MHz CCs 

DL: Non-contiguous 2x20 + 

2x20 MHz CCs 

3.5 GHz band FDD 

This scenario is shown in Figure 4-25. 

• There is some relation to the F1 scenario in ITU-R (Figure 4-21). 


• In the given frequency band only one UL (theoretically two) and one DL carrier is possible. 80 MHz 

are not used. 


within 3500 – 3600 MHz. 


• The implementation is feasible either with a single transceiver or two parallel transceivers, because a 

nearly contiguous frequency band is used, where some gaps may not be used. 


UL 

DL 

3300 3400 3500 3600 3700 

MHz 


CEPT is considering extending the frequency band up to 3800 MHz in CEPT countries. Under these 

conditions additional uplink and downlink FDD carriers could be added as long as the bandwidth of the 

uplink and downlink carrier bandwidth remains 40 MHz and 80 MHz. If a non-contiguous spectrum 

allocation is assumed, then only one additional uplink and downlink carrier could be added with the 

following carrier allocation: 

• Uplink: 

o Carrier 1: 3420 – 3440 MHz plus 3460 – 3480 MHz 


• Downlink: 


o Carrier 1: 3700 – 3740 MHz plus 3760 – 3800 MHz. 

If the condition of non-contiguous spectrum allocation is not considered anymore, this would result in the 

same case as in scenario 1 with the following carrier allocation: 

• Uplink: 

o Carrier 1: 3400 – 3440 MHz 

o Carrier 2: 3440 – 3480 MHz 

o Carrier 3: 3480 – 3520 MHz 

• Downlink: 

o Carrier 1: 3560 – 3640 MHz 

o Carrier 2: 3640 – 3720 MHz 

o Carrier 3: 3720 – 3800 MHz 



The first version would result in the same non-ideal spectrum usage as the original 3GPP scenario 4. The 

second version would result in a much better spectrum usage. 

5 Single-band noncontiguous 

spectrum 


for FDD 

Scenario 5 

UL: 10 MHz 

DL: 10 MHz 

UL/DL: Non-contiguous 5 MHz 

+ 5 MHz CCs 

Band 8 (900 

MHz) 

This scenario is using today's 2G frequency bands (Figure 4-26). Reforming of such bands would be 

needed. 


• In the given frequency band three UL and three DL carrier are possible. 2 times 5 MHz are not used. 


within 925 – 960 MHz. 

• In order to support more competition infrastructure sharing may be needed. However, three 

independent operators may be active. 



FDD 

UL UL UL DL DL DL 

Band 8 

950 880 915 925 960 1000 

MHz 

6 Single-band noncontiguous 

spectrum 


for TDD 

Scenario 6 


80 MHz Non-contiguous 2x20 + 2x20 

MHz CCs 

Band 38 (2.6 

GHz) 

Figure 4-27 shows the understanding of this scenario. Band 38 has a bandwidth of 50 MHz. On the other 

hand the scenario description is requesting a non-contiguous TDD carrier of effective bandwidth of 80 

MHz. This is not consistent (see also Table 4-2) and will be resolved be 3GPP in upcoming meetings. 


• In the given frequency band in maximum only one UL and one DL carrier/connection is possible. 


• The implementation may be feasible with a single transceiver, is the entire 80 MHz effective carrier 

bandwidth may be allocated within band 7. 

• Possible, band 7 is meant. 

• The maximum sampling bandwidth is not yet known. 

TDD 



UL/DL 

Band 38 

2500 2570 2620 2670 

MHz 

7 Multi-band noncontiguous 

spectrum 

allocation @ Band 1, 

3 and 7 for FDD 

Scenario 7 


UL: 40 MHz 

DL: 40 MHz 

UL/DL: Non-contiguous 10 

MHz CC@Band 1 + 10 MHz 

CC@Band 3 + 20 MHz 

CC@Band 7 

Band 3 (1.8 GHz) 



Scenario 7 is the most difficult one by combining several bands (Figure 4-28). 


• In the given frequency band four UL and four DL carrier are possible. 80 MHz are not used in total. 

This could be used for guard bands. 

• Basically, there is not much flexibility, where to allocate the different UL and DL carriers. 

• In order to support competition different operators may receive different carrier frequencies. However, 

in some cases more complex base stations for aggregated frequency bands would be needed, which is 

not very attractive. 

• The implementation is feasible in general with two parallel transceivers, because in some cases a 

contiguous frequency band is used and in some cases two bands are combined to achieve 40 MHz 

carrier bandwidth. 

• However, a single more wideband transceiver would also be feasible, because the maximum sampling 

bandwidth would be 100 MHz and it can be expected that no major interferers are active in the band 

gaps. 

• Therefore, both concepts seem to be technically feasible to implement. 

FDD 

UL UL 

UL UL 

DL 

DL 

DL 

DL 

Band 3 Band 1 

Band 7 

1710 1785 

19201980 2110 

1805 1880 

2170 

2500 2570 

2620 2690 

MHz 


8 Multi-band non- 30 MHz Non-contiguous 1x15 + 1x15 Band 1 (2.1 GHz) FDD> 




MHz CCs 

Band 3 (1.8GHz) 


and Band 3 for FDD 

Scenario 8 

This scenario shows a symmetric spectrum allocation for uplink and downlink (Figure 4-29). 


• In the given frequency nine UL and nine DL carriers are possible. The entire bands are used. 

• This spectrum allocation allows for high competition. 

• The implementation is feasible with a single transceiver, because a contiguous frequency band is used 

per carrier. 

• The maximum sampling bandwidth is 15 MHz. Technically, this is feasible to implement. 

UL 

DL 

UL 

DL 

Band 3 Band 1 

1710 1785 

1920 1980 2110 2170 

1805 1880 


MHz 


spectrum 

allocation @ 800 

MHz band and Band 

8 for FDD 

Scenario 9 

UL: 20 MHz 

DL: 20 MHz 

UL/DL: Non-contiguous 10 

MHz CC@UHF + 10 MHz 

CC@Band 8 

800 MHz band 

Band 8 (900 

MHz) 

Figure 4-30 illustrates this scenario. 


• In the given frequency band three UL and three DL carriers are possible. 22 MHz are not used. 

• Basically, there is some flexibility, where to allocate the UL band within the UHF band and the DL 

band within band 8. 

• This approach does allow for some competition. 

• The implementation is feasible with a single transceiver, because mostly a contiguous frequency band 

is used. or two aggregated bands within 35 MHz bandwidth. 


FDD 



UL DL DL 

UHF Band 8 


spectrum 

allocation @ Band 

39, 34, and 40 for 

TDD 

Scenario 10 

790 862 

925 960 MHz 

880 915 


90 MHz Non-contiguous 2x20 + 10 + 

2x20 MHz CCs 




This scenario is using three different bands (Figure 4-31). 

• There is some relation to the E1/E2 scenario in ITU-R (Figure 4-20). 

• Possibly, not the entire band 40 is used, because this is not available in several regions. 


• In the given frequency bands only one UL/DL carrier/connection is possible. 75 MHz are not used. 

• Basically, there is flexibility, where to allocate the UL/DL part bands within band 34 from 2010 – 2025 

and band 40 from 2300 – 2400 MHz. 


• The implementation is feasible with three parallel transceivers. The total bandwidth from 1880 – 2400 

MHz will be too big for a wideband transceiver with respect to the sampling bandwidth and potentially 

other interfering signals in the band gaps. In this case the maximum sampling bandwidth would be 520 

MHz which can not be supported with sufficient dynamic range with today's technology and 

reasonable power consumption in particular in the terminal. 

• The maximum sampling bandwidth per part band would be 20 MHz. Technically, this is feasible to 

implement. 

TDD 

UL/DL UL/DL UL/DL 

Band 39 Band 34 

Band 40 

1880 

1920 

2010 

2025 

2300 2400 

MHz 



11 Single-band 

Contiguous spectrum 


for FDD 

Scenario 11 


UL: 20 MHz 

DL: 40 MHz 

UL: 1x20 MHz CCs 

DL: 2x20 MHz CCs 


Figure 4-32 is an implementation in an already allocated frequency band. 


• In the given frequency band only one UL and one DL carrier is possible. 80 MHz are not used in this 

asymmetrical frequency allocation. 


within 2620 – 2690 MHz. 



• The maximum sampling bandwidth is 20 MHz. Technically, this is feasible to implement. 

FDD 

UL 

DL 

Band 7 

2500 2570 2620 2690 

MHz 


All scenarios, which are currently being discussed in 3GPP, can be implemented either with a single band 

transceiver or in maximum with up to three parallel transceivers in order to avoid too high requirement on 

sampling bandwidth and dynamic range of the receivers and A/D-converters. Therefore, such concepts 

are feasible to implement. 

However, the identified frequency bands are not fully used. In some cases significant parts of potentially 

available bands remain unused in the proposed scenarios. 

4.3.3 Conclusions 

Section 4.3 is looking at the currently being discussed spectrum implementation issues in ITU-R and 

3GPP. The 3GPP scenarios can be implemented mostly with single band transceivers. In some cases two 

or in maximum three parallel transceivers are needed in order to implement the system with sufficient 

dynamic range and low sampling bandwidth and to avoid the impact of interfering signals in band gaps 

between part bands. Three parallel transceivers seem to be feasible with respect to today's terminal 

implementation, which is supporting up to four frequency bands for 2G systems and in parallel 3G, 

WLAN GPS and DVB in a single high-end device. From a complexity perspective such implementations 

are feasible. 

However, the use of single band transceivers provides lower complexity and cost and will most probably 

be preferred. Therefore, the implementation of wideband carriers in contiguous bands would be a solution 

with lower complexity. With respect to the available and identified frequency bands this may require 

infrastructure sharing in order to support competition between operators. Such concepts may change the 

role models of operators. 



Several of the 3GPP scenarios do not fully exploit potentially available frequency bands. 

International specification, standardisation and regulatory bodies will discuss spectrum scenarios further 

by taking into account national, regional and international constraints on the availability and different 

frequency bands. 

5. References 

[DIX08] J. Dixton, C. Politis, and C. Wijting, “Considerations in the choice of suitable spectrum 

for mobile communications,” White Paper, WWRF WG 8, Oct. 2008. 

[E2RD5.3] IST 2003-507995 Project End-to-End Reconfigurability (E2R), Deliverable D5.3, 

“Algorithms and performance, including FSM& RRM/network planning,” June 2005. 

[ITU-CL 2008] ITU-R Circular Letter 5/LCCE/2, “Invitation for submission of proposals for candidate 

radio interface technologies for the terrestrial components of the radio interface(s) for 

IMT-Advanced and invitation to participate in their subsequent evaluation,” Mar. 2008. 

[M.1645] ITU-R: Framework and overall objectives of the future development. of IMT 2000 and 

systems beyond IMT 2000. Recommendation ITU-R M.1645, 2003. 

[Moh03] Mohr, W.: Spectrum Demand for Systems Beyond IMT-2000 Based on Data Rate 

Estimates. Wiley Journal Wireless Communication and Mobile Computing, 2003; No. 

3:pp. 1–19. 

[MG86] Meinke/Gundlach: Taschenbuch der Hochfrequenztechnik, Springer Verlag, Berlin, 

1986. 

[Pa92] Parsons, J.D.: The Mobile Radio Propagation Channel, Pentech Press, London, 1992. 

[Ra96] Rappaport, T.S.: Wireless Communications, Prentice Hall, 1996. 

[WB88] Walfisch, J. and H.L. Bertoni: Theoretical model of UHF propagation in urban 

environments. IEEE Trans. Antennas and Propagation, vol. 36, pp.1788-1796, 1988. 

[Ha80] Hata, M.: Empirical formula for propagation loss in land mobile radio services. IEEE 

Trans. Vehicular Technology, vol.29, pp.317-325, 1980. 

[Ja74] Jakes, W.C.: Microwave Mobile Communications. John Wiley & Sons, New York, 

1974. 

[Lue75] Lüke, H.-D.: Signalübertragung. Springer-Verlag, Berlin, 1975. 

[Car90] Carson, R.S.: Radio Concepts Analog, John Wiley & Sons, New York, 1990. 

[Zv67] Zverev, A.I.: Handbook of Filter Synthesis. John Wiles & Sons, New York, 1967. 

[Sa79] Saal, R.: Handbuch zum Filterentwurf. AEG-TELEFUNKEN, Berlin, 1979. 

[GRA05] D. Grandblaise, et al., “Techno-Economic of Collaborative Based Secondary Spectrum 

Usage - E2R Research project outcomes overview,” in Proc. of DySPAN 2005 - First 

IEEE International Symposium on New Frontiers in Dynamic Spectrum Access 

Networks, Baltimore, Maryland, USA, Nov. 2005. 

[KEL05] H. Kellerer and U. Pferschy D. Pisinger Knapsack Problems. 2005, Springer 

VerlagISBN 3-540-40286-1. 

[KAR05] J., K. Karlof, “Integer Programming: Theory and Practice”, 1st ed., CRC, 2005 

[LEE93] W. C. Y. Lee, Mobile Comm. Design Fundamentals, 2nd. ed., New York: Wiley, 1993 

[LUO03] J., Luo, et al., “Performance investigations of JRRM in a reconfigurable environment,” 

in Proc. of SCOUT Workshop, Paris, France, Sep. 2003. 

[MEU08] F. Meucci, A. Mihovska, B. Anggorojati, N. R. Prasad. “Joint Resource Allocation and 

Admission Control Mechanism for an OFDMA-Based System”. The 11th International 

Symposium on Wireless Personal Multimedia Communications (WPMC 2008), 

Finland, 2008. 

[MEU09] F. Meucci, O. Cabral, A. Mihovska, “Spectrum Aggregation with Multi-Band User 

Allocation over Two Frequency Bands,” to appear in Proc. Of IEEE MWS’09, July 

2009, Napa Valley, CA, USA. 

[MIH08] A. Mihovska, et al., “Requirements and algortihms for cooperation of heterogeneous 

networks,” in Springer International Journal On Wireless Personal Communications, 

DOI: 10.1007/s11277-008-9586-y, September 2008. 



[MIH208] A. Mihovska, et al., “A mulit-stage admission control for load balancing,” in Proc. of 

WPMC 2008 - The 11th International Symposium on Wireless Personal Multimedia 

Communications, Lapland, Finland, Sep. 2008. 

[MIH09] A. Mihovska, et al., “Multi-Operator Resource Sharing Scenario in the Context of IMT- 

Advanced Systems,” to be published in IEEE Wireless VITAE 09, Aalborg, Denmark, 

May 2009. 

[MON08] V. Monteiro, J. Rodriguez, A. Gameiro, O. Cabral, F. Velez, “RAT and Cell Selection 

Based on Load Suitability”, invited paper in Proc. of VTC2008 The 68th IEEE 

Vehicular Technology Conference, Calgary, Canada,Sep. 2008. 

[PER04] M. K. Pereirasamy, et al., “An approach for inter-operator spectrum sharing for 3G 

systems and beyond,” in Proc. of PIMRC 2004 - 15th IEEE International Symposium 

on IEEE Personal, Indoor and Mobile Radio Communications, Barcelona, Spain, Sep. 

2004. 

[SAL08] O. Sallent, et al., “Decentralized spectrum and radio resource management enabled by 

an on-demand cognitive pilot channel,” Springer Journal on Annals of 

Telecommunications, Vol. 63, No. 5-6, June 2008, pp. 281-294. 

[TRA07] E. Tragos, A. Mihovska, E. Mino, “Hybrid RRM architecture for future wireless 

networks,” in Proc. of PIMRC 2007 – 18th IEEE International Symposium on Personal, 

Indoor and Mobile Radio Communications, Athens, Greece, Sep. 2007. 

[TS80] Tietze, U. and Ch. Schenk: Halbleiter-Schaltungstechnik, 5. überarbeitete Auflage, 

Springer Verlag, Berlin 1980. 

[WIND22] WINNER project: Feasibility of multi-bandwidth transmissions. Deliverable D2.2, IST- 

2003-507581 WINNER, October 2004. 

[WIND6.2] IST-2003-50781 WINNER D 6.2 “Methodology for estimating the spectrum 

requirements for “further developments of IMT-2000 and systems beyond IMT-2000”; 

v.2.0, December 2005. 

[WIND6.5] IST-2003-50781 WINNER D 6.5 “Spectrum requirements for “further developments of 

IMT-2000 and systems beyond IMT-2000”, v.1.0, December 2005. 

[WIN2D5.10.2] IST-4-027756 WINNERII D 5.10.2 “Spectrum Requirements for System beyond IMT- 

2000“, v.1.0, March 2007. 

[WIN+D1.2] WINNER+, Deliverable D1.2, “Initial report on system aspects of flexible spectrum 

use”, January 2009. 

[WIN+D3.1] WINNER+, Deliverable D3.1, “IMT-Advanced: requirements and evaluation criteria ” 

March 2009. 

[WRC’07] “Key results of World Radiocommunication Conference (WRC-07)”, available online 

at: http://www.itu.int/dms_pub/itu-t/oth/21/04/ T21040000030014PPTE.ppt. 

[BH06] Brannon, B. and J. Hall: Understanding State of the Art in ADCs. Analog Devices, ADI 

– Preliminary Technical Data, not for external distribution, August 23, 2006 (received 

on private basis). 

[3GPP] Third Generation Partnership Project 3GPP, www.3gpp.org. 

[INDOOR] ITU-R P 1238-2: Propagation data and prediction methods for the planning of indoor 

radio communication systems and the radio local area networks in the frequency range 

900 MHz 

[3GPP1] 3GPP TSG-RAN1#48, 12th-16th February, 2007, St. Louis, MI, USA. R1-070674 

[3GPP2] 3GPP TSG-RAN1 WG1, LTE Downlink Performance, Conference Call, Apr. 24 th , 

2007. R1-071978. 

[Kel07] Jean-Marc Kelif, “Multiservice Admission on Sectored CDMA networks An Analytical 

Model,” IEEE WCNC Hong Kong, March 2007.

Aspects of WINNER+ Spectrum Preferences - Celtic-Plus

Create successful ePaper yourself

Delete template?

Save as template?