Femtocell UBS-21!92!028 Femto Cell Macro Layer Study

© 2007 UbiquiSys Limited –Company Confidential

UBS-21-92-028 V1.0.0 (2007-1-13)

White paper

ZoneGate

Impact of 3G Femto Cells on Existing 3G

Networks

© 2007 UbiquiSys Limited. All rights are reserved; reproduction in whole or in part is prohibited without

written consent of the copyright owners.

ZoneGate Impact of 3G Femto Cells on Existing 3G Networks /40 UBS-21-92-028 V1.0.0 (2007-1-13)


Contents

1 EXECUTIVE SUMMARY........................................................................................................................... 3

2 REFERENCES.............................................................................................................................................. 6

3 DEFINITIONS AND ABBREVIATIONS .................................................................................................. 7

4 INTRODUCTION......................................................................................................................................... 8

5 A BRIEF SUMMARY OF 3GPP HOME NODEB INTERFERENCE STU DIES................................ 10

5.1 HOME NODEB WORKING GROUP PRELIMINARY CONCLUSIONS............................................................ 10 5.2 CLOSED SUBSCRIBER GROUP AND CO-CHANNEL/ADJACENT CHANNEL DEPLOYMENT......................... 11 5.3 OPEN SUBSCRIBER GROUP AND CO-CHANNEL DEPLOYMENT................................................................ 13 5.4 HNB TO HNB DEPLOYMENTS............................................................................................................... 14

6 DESCRIPTION OF THE NETWORK SIMULATION TOOL ......... .................................................... 16

7 FEMTO CELL AND MACRO LAYER NETWORK LAYER SIMULATION S................................. 20

7.1 FEMTO CELLS IN AN URBAN ENVIRONMENT.......................................................................................... 21 7.2 FEMTO CELLS IN A SUBURBAN ENVIRONMENT ...................................................................................... 25

8 DOWN LINK AND UP LINK POWER ADAPTATION TEST RESULTS .......................................... 33

9 CONCLUSIONS ......................................................................................................................................... 39

ANNEX A CHANGE HISTORY...................................................................................................................... 40



1 Executive Summary

Prior to any large scale deployment of 3G femto cells mobile network operators (MNOs) will want to understand the possible effect that the co-channel and adjacent channel radio frequency (RF) interference caused by this new technology may have on their existing Macro Layer networks. Regulatory bodies will also need to understand the effect one operator‘s femto cell deployment may have on another operator’s existing Macro Layer to ensure harmonious coexistence. Furthermore with limited availability of 3G spectrum worldwide it is important to understand the effect interference will have as the femto cell market grows and matures. A range of internal and external activities have been underway to understand the effect of interference. The objective of this paper is to provide a brief summary of these studies and then to extend the current understanding of this topic by focussing on a number of key issues.

Over the last two years UbiquiSys in collaboration with various network operators and an independent consulting company Red-M, have been investigating the effect of interference and the management thereof. This work has resulted in various RF trials [2], [3] and theoretical studies [1] which have focussed on the following key areas:

• Coverage of the femto cell when operating on a co-channel and adjacent channel as occupied by the Macro Layer.

• The impact on both the Macro Layers Down Link (DL) and Up Link (UL) coverage caused by the femto cell and its mobiles (UEs) when operating on the same or adjacent carrier.

• The performance evaluation of the femto cell radio resource management algorithms (RRM) known as network monitor mode and the associated DL and UL power adaptation and data rate adaptation algorithms. Where applicable the evaluation of the effect of these algorithms on the Macro Layer.

• The evaluation that one femto cell may have on others when they are collocated (i.e. office or apartment scenario)

In general the trials and studies have shown that it is very possible for a small population of femto cells to coexist harmoniously with each other, or the Macro Layer, both on the same or adjacent carrier. Due to limitations in the analytical tools (link budgets) and availability of large quantities of femto cells these studies and trials have not been able to evaluate the scaling effect of large deployments of femto cells on the Macro Layer. To this effect, UbiquiSys engaged with Red-M in the development of a static radio network simulator which incorporates key aspects of the Radio Resource Management (RRM) algorithms that UbiquiSys has developed in order to manage interference. The purpose of the network simulator is to model the effect of large deployments of femto cells and to understand the positive or negative impacts they may have on the Macro Layer.

In the meantime, the 3GPP, through one of its study groups, initiated a femto cell technical study to evaluate the effect of large deployments of femto cells (which they refer to as Home NodeBs) to determine the feasibility of co-channel and adjacent channel deployments. The objective of this study group is also to decide which radio compliance requirements [4] need to be changed to reflect the femto cell RF requirements. Within the technical study group, leading 3G infrastructure/mobile vendors and operators including, but not limited to, Ericsson, Vodafone, T-Mobile, Nortel, Qualcomm, Nokia-Siemens, Orange, Motorola and Alcatel-Lucent have created a framework within which to analyse the interference issue. In particular the 3GPP technical study group identified three key deployment scenarios that will affect interference in a large deployment of femto cells, namely:

• Mass deployment of femto cells on the same or adjacent carrier as the Macro Layer. Mass deployment implies an increasing density of femto cells.

• The maximum allowed Down Link and Up Link transmission power levels of the femto cell community. It was recognized early that power management is crucial in an ‘unplanned’ femto cell network.



• Supporting a Closed or Open subscriber group for femto cells. A Closed Subscriber Group implies that only those users registered to use a particular femto cell may establish calls on that femto cell. Open Subscriber Group implies that any subscriber registered with that particular operator can establish calls on that operator’s femto cells. Partially open implies that part of the capacity of the femto cell can be made available to support a limited number of members from the open subscriber group.

Following a significant number of contributions from various interested parties [5] it has in general been concluded1 that large femto cells deployments on a dedicated carrier can harmoniously co-exist with Macro Layer networks with minimal impact assuming that the Femto Cell transmit power levels are either adapted (based on the coverage requirements) or capped at approximately +5dBm. This conclusion holds for both Open and Closed Subscriber Groups. They have also in general concluded that operation on a co-channel is possible if the Femto Cell is made open. Concerns over ownership of the backhaul i.e. the DSL line, or similar broadband connection in the home, could yet make this option more challenging than it first appears. However, the members of the working group have all acknowledged that operation on the same carrier, with a closed subscriber group, still presents a significant interference challenge. The working group have also acknowledged the positive benefit that RRM algorithms (e.g. adaptive power management) will bring to this scenario. Finally the 3GPP WG has also concluded that power management alone may not be sufficient to ensure that the impact on the Macro Layer is always below an acceptable limit.

At this stage it is important to note that the 3GPP submissions assume that the femto cells were added in addition to the Macro Layer without accounting for any traffic off load from the Macro Layer to the femto cells. The reason for this approach was to understand the worst case interference effects and to simplify an already complex problem. Furthermore it would appear that the study group did not relate the expected femto cell densities (i.e. number of femto cells per cell site) to the expected densities that would occur at different stages of market maturity. For example, it is expected that the density of femto cells would be dependant on the existing 3G operator subscriber base, expected femto cell market penetration and population demographics (i.e. population densities in urban, suburban, dense urban) and other factors all of which will affect levels of interference.

In this paper UbiquiSys attempts to extend the understanding of the impact on the Macro Layer caused by the deployment of a large population of closed, co-channel femto cells focusing on the following aspects:

• The paper investigates the effect of moving all indoor users onto the femto layer. It is shown that there is an overall increase in network capacity mainly due to the Macro Layer using less DL transmit power (which is freed up by moving the indoor users onto the femto layer) even though the femto layer would create interference to the Macro Layer through the transmission of its control channels. Furthermore on the Up Link, since the indoor users would generally operate at a lower transmit power when served by the femto layer, there is a smaller up link noise rise at the Macro Layer nodeBs2. Clearly, taking this somewhat obvious result in the context of the negative capacity effect of simply placing Femto Cells in the Macro network (as shown in [15] without traffic migration) implies that there must be a ‘cross over’ point, where, assuming a certain percentage of the Macro network subscribers have migrated to the femto layer (i.e. some Femto Cells are in use while others are not) results in a zero impact to the Macro Layer. Beyond the ‘cross over’ point (i.e. larger percentage migration) there should always be a positive capacity benefit to the Macro network. A positive capacity benefit implies that the sum of the users (assuming their service remains unchanged or is improved) served by both Macro Layer and femto layer exceeds that of just the Macro Layer.

• The paper then investigates the ‘cross over’ point for a range of different scenarios, traffic mixes and network loadings of a suburban network in Swindon assuming that the demographics remain constant (i.e. there is no mass migration into or out of an area, nor mass migration of the subscriber base from one operator to another) for both fixed and adaptive power management of the femto layer. Based on a range of simulations it is concluded that for a typical suburban network the cross over point occurs when the femto layer is power managed, and approximately 10 to 15% of the Macro Layer traffic is carried by the femto layer. This assumes a worse case density of femto cells similar to a mature market.

1 At this stage further work is ongoing within the 3GPP Home NodeB study group and these general conclusions are subject to change.

2 An increase in uplink noise rise would potentially lead to a decrease in Macro Layer coverage through cell shrinkage.



For a non-power adapted femto layer this increases to approximately 40% hence demonstrating the significant advantage of Ubiquiys’s RRM power management algorithms. In essence these simulations extend the work carried out by Nortel Vodafone [14] but incorporate femto cell power management and investigate the effect of different levels of network traffic migration.

• The paper then provides a number of on-air test results of the DL and UL power adaptation algorithms that were carried out by UbiquiSys with their ZoneGate Access Point (ZAP) femto cell in two residences. The test results demonstrate the significant range of maximum DL and UL transmit power levels that would occur in different building types and for different levels of Macro Layer interference.

The UbiquiSys study concludes that as long as a certain proportion of the user population is served by the Femto Cells, the capacity benefits of Femto Cells will outweigh any impact on 3G Macro networks.



2 References

The following references are used in this document.

[1] UBS-21-92-005 ZoneGate Network Study Ver 3

[2] UBS-21-63-001 Vodafone Phase 1 RF Test Results

[3] UBS-21-63-006 RF Coverage Test Results

[4] 3GPP TS25.104 Base Station Radio Transmission and Reception (FDD).

[5] 3GPP TR25.820 V1.0.0 (2007-11) 3G Home NodeB Study Item Technical Report (Release 8)

[6] UBS-21-92-004, “ZoneGate Radio Resource Management”.

[7] ‘Radio Network Planning and Optimization for UMTS’, Edited by Jaana Laiho, Achim Wacker and Tomas Novosad. Wiley.

[8] ‘Matlab Impelementation of a Static Radio Network Planning Tool for Wideband CDMA”, Jaana Laiho, Achim Wacker, Kari Heiska, Kai HeikKinen.

[9] ZNPSW User Manual, Ver 1.0, Red-M, Dom Jenkins

[10] Definition of the path loss models as implemented in ZNPSW Simulation Tool, Ver 1.3, Red-M, Dr. Bachir Belloul

[11] Validation Report for ZNPSW Network Simulation Tool. Ver 1.0. Red-M, Dr. Bachir Belloul

[12] Technical note on the practical implementation of the RRM variables in ZNPSW, July 2007, Dr. Bachir Belloul.

[13] A simulator to assess the impact of 3G ZAPs on ML Performance: Requirements Document, Version 1.0, 30 March, 2007, Red-M, Dr. S.Mitchell

[14] R4-071231 3GPP TSG-RAN WG4 Meeting #44 Athens, Greece, 20-24th August, 2007, Nortel, Vodafone

[15] 3GPP R4-070902 Initial home nodeB coexistence simulation results, Nokia Siemans.

[16] 3GPP R4-071578 Simulation results of macro-cell and co-channel Home NodeB with power configuration and open access, Alcatel – Lucent.

[17] 3GPP R4-071231 Open and Closed Access for Home NodeBs, Nortel – Vodafone.

[18] 3GPP R4-0701494 Spectrum Arrangement to enable Co-channel deployment of Home NodeBs, Nortel

[19] 3GPP R4-071661 Impact of HNB with controlled output power on macro HSDPA capacity, Ericsson

[20] 3GPP R4-071660 Impact of HNB with fixed output power on macro HSDPA capacity, Ericsson

[21] 3GPP R4-071940 Simulation results for Home NodeB to macro UE downlink co-existence considering the impact of HNB HS utilization, Ericsson.

[22] 3GPP R4-071941 Simulation results for Home NodeB to Home NodeB downlink co-existence considering the impact of HNB HS utilization. Ericsson.

[23] Ofcom Site Finder. http://www.sitefinder.ofcom.org.uk/. Accessed September 2006



3 Definitions and Abbreviations

For the purposes of the present document, the following abbreviations apply:

DL Down Link

HNB Home NodeB (i.e. Femto Cell)

MS Mobile Station

RRM Radio Resource Management

UL Up Link

WG Working Group

ZG ZoneGate

ZAP ZoneGate Access Point



4 Introduction

To date, through measurements, analysis and trials Ubiquisys has shown that small communities of femto cells can operate on a carrier already in use by an existing Macro Layer network with limited impact on the local 3G network users. Similarly UbiquiSys has also shown that small communities of femto cells operating on carriers adjacent to a carrier in use by an existing 3G network will have little or no impact on the 3G network users [1],[2] and [3].

Key to this harmonious coexistence is the application of proprietary DL and UL power management techniques within the UbiquiSys femto cell which have been designed to minimise interference to existing 3G networks and other femto cells whilst providing coverage in the femto cell’s expected coverage area.

Having shown the power management techniques achieve the desired effects on a small scale, the question still remains as to whether such techniques will scale to large communities of femto cells similar to what would be expected in a commercial deployment at its peak. In this paper UbiquiSys aims to provide insight into some key questions surrounding large femto cell deployments, including:

• How does a large femto cell deployment impact a typical 3G network capacity when operating on the same or adjacent carrier?

• At what deployment density do the benefits of deploying large numbers of femto cells outweigh any negative impacts?

• What capacity improvement can be expected from a deployment of large number of femto cells on the same carrier as an existing 3G network assuming different percentages of traffic migration from the Macro Layer to the femto layer?

• What are the impacts of femto cells if deployed on a dedicated carrier adjacent to a second operator?

In an attempt to answer these questions UbiquiSys, in collaboration with Red-M, have developed a static network simulation platform that provides a snapshot of a 3G macro network with femto cells deployed within its coverage area.

With the 3GPP Home NodeB working group encouraging analysis from a range of different view points so as to ensure a robust and diverse analysis of the issue, Ubiquisys has taken the approach of analysing the roll out of femto cells on existing 3G networks with femto cell densities based on demographics related to the 3G coverage area being simulated as well as accounting for factors such as MNO market share. This paper contains simulation results based on this approach of typical 3G networks within the UK representing suburban and dense urban 3G networks.

The rest of the paper is structured as follows:

• Section 5 provides a summary of the contributions and conclusions made by various major OEMs and operators as part of the 3GPP Home NodeB working group four (WG4) activity. The papers submitted to WG4 have shown a range of results and opinions due to the variability of user deployments (different buildings, locations) and simulation assumptions (femto cell deployments, interference mitigation etc). The section captures key results representing the ‘middle ground’ and includes the conclusions from the latest status report from this working group. Finally, although this paper focuses on the interaction between the Macro Layer and femto cell layers on the same carrier it also includes some of the published femto cell on femto cell interference study results as they extend the information provided in [1].

• Section 6 provides a brief overview of the UbiquiSys/Red-M Macro Layer static traffic simulation platform. The section highlights the key functions within the platform and the validation that was carried out on the platform prior to running the simulations reported in this study.

• In section 7 the paper first investigates through simulations a heavily loaded 3G network in a dense urban environment (London) and the effect of moving all indoor users onto the femto layer. It shows that there is an overall increase in network capacity as the Macro Layer requires less DL transmit power even though the femto layer is creating additional interference in the Macro Layer through the



transmission of its control channels. Clearly taking this somewhat obvious result in the context of the negative capacity effect of simply placing Femto Cells in the Macro network without traffic migration (as reported in [14]) implies that there must be a ‘cross over’ point where assuming a percentage of the users have migrated to the femto layer (i.e. some femto cells are in use while others are not) where there is a zero impact to the Macro Layer. Beyond the ‘cross over’ point it can then be concluded that there would always be a positive capacity benefit to the Macro Layer.

• Section 7 then investigates the ‘cross over’ point for a range of different scenarios, traffic mixes and network loadings of a suburban network in Swindon assuming that the demographics remain constant (i.e. there is no mass migration into or out of an area or there is no mass migration of the subscriber base from one operator to another) for fixed power and adaptive power management of the femto layer. In essence these simulations extend the work carried out by Nortel Vodafone [14] by incorporating Femto Cell DL/UL power adaptation and investigating the effect of different levels of user migration.

• Section 8 provides a range of on-air test results of the DL and UL power adaptation algorithms that were carried out by UbiquiSys with their ZAP Femto Cell in two residences. The test results demonstrate the significant range of maximum DL and UL transmit power levels that would occur in different building types and different deployment positions hence supporting the view that adaptive power management is key to minimising interference.

• Finally, section 9 concludes the paper, recalling the objectives of the study side by side with the results of the UbiquiSys study and important conclusions considering the impacts of expected changes in user behaviour as femto cell usage approaches that of WLAN.

Although this report was written by UbiquiSys it is important to acknowledge the significant contribution of Red-M3 in the development and validation of the network simulator including the RRM power management algorithms. UbiquiSys also gratefully acknowledges the Red-M technical review of this report.

3 Note: UbiquiSys engaged with Red-M (http://www.red-m.com/about/) due to their well established reputation in the area of RF network planning, and in particular their portfolio of wireless services which include Consulting, Audit, Network Design, Planning, System Integration and Systems Management. The company’s capabilities span all wireless technologies and its customers include Ofcom, Department of Health, Network Rail, network operators: T-Mobile, O2, 3, Airwave, major corporations: British Land, BAA and Bullring.



5 A Brief Summary of 3GPP Home NodeB Interference Studies

This section provides a summary of contributions that have been made the OEM and Operator community into RAN Working Group 4 (WG4) as part of the Home NodeB technical study activity. As part of this activity there have been a significant number of excellent contributions made by Nokia-Siemens, Alcatel-Lucent, Qualcomm, Vodafone, Ericsson, Nortel, T-Mobile, Motorola and Orange. Since femto cells (or Home NodeBs) are not network planned in the traditional sense there has been a range of opinions that have been expressed within this forum. This summary is not an evaluation of the technical/business merit of submissions but rather an attempt to summarize the collective thinking and then isolate those areas which require further analysis/simulations within the context of the Radio Resource Management (RRM) principles and algorithms adopted by UbiquiSys in their femto cell product.

This subsection is structured in the following manner:

• It begins by explaining the purpose and focus of the 3GPP Home NodeB activity with extracts from [5] to reflect the preliminary conclusions of the wider community. The paper emphasises the areas still open to further investigation.

• Following the WG4 overview the next subsections provide a selection of the results presented by different companies to support the conclusions made within WG4. Where applicable these results are cross compared with the results presented in the simulation results in section 7.

5.1 Home NodeB Working Group Preliminary Conclusions

The following text has been taken largely from [5] to reflect the current view of the Home NodeB (HNB) working group. The diverse input to this study item on Home Node B / eNode B has revealed that a wide range of possible deployment configurations are envisioned for the HNB. This study has used interference scenarios to investigate the impact on Home Node B deployment on the existing basestation requirements. However, the interference scenarios are dependent on the deployment configurations. Specifically, the most important deployment characteristics considered are as follows:

• Open access or CSG (Closed Subscriber Group)

o Open access HNBs can serve any UE in the same way as a normal NodeB o CSG HNBs only serve UEs which are a member of a particular Closed Subscriber Group

• Dedicated carrier or co-channel o Whether HNBs operate in their own separate channel, or whether they share a carrier with an

existing (e)UTRAN network

Furthermore, how an operator chooses to manage Home Node B power has a strong impact on the interference analysis. Therefore these studies distinguished between the following methods of managing the HNB transmit power:

• Fixed: HNBs have a set fixed maximum transmit power.

• Adaptive: HNBs sense interference to existing networks, and adjust maximum transmit power accordingly4.

Home Node B’s extend and enhance the coverage of a UMTS Radio Access Network in the home environment. However, it is not feasible to completely control the deployment of the HNB layer within the UMTS Radio Access Network (RAN). Therefore, interference due to the HNB is a concern and [5] concludes that interference mitigation techniques are required in the case of closed access. No single method has been identified that completely eliminates interference while maintaining HNB performance for closed access. It is

4 This is the approach taken by UbiquiSys in their Femto Cell product. At power up and regular intervals the Femto Cell uses a mobile measurement mode to sense the Macro Layer interference and set its DL and UL Tx powers. Ongoing UE measurements are used to adapt the DL and UL Femto transmit powers. Femto Cell power adaptation and interference management is provided for both dedicated and co-channel scenarios. Interference management between a group of collocated Femto Cells is also provided by the UbiquiSys solution.



not the intention of [5] to recommend a set of specification or an algorithm that ensures feasibility of the Home Node B. Rather, [5] evaluates the effectiveness of interference control with an acceptable trade-off between macro layer and HNB performance over a set of deployment configurations.

The analysis of the various configurations resulted in the following observations:

• Open access configuration will result in lower Macro Layer interference levels than Closed Subscriber Group Operation.

• Dedicated carrier deployment results in much lower interference levels than co-channel deployment.

• A CSG HNB deployment (whether dedicated or co-channel) requires interference mitigation techniques in order to control the inter-HNB interference for both the downlink and uplink.

• It is not possible to control the downlink co-channel interference through fixed maximum HNB transmit power setting in case of co-channel CSG HNB deployment.

• A "partial co-channel" approach for UTRAN operating on two channels can provide higher spectral efficiency than obtained with a dedicated carrier approach while maintaining the same cell edge performance.

• In case of CSG co-channel HNB deployment it is possible to control the uplink and downlink interference levels to the macro layer through appropriate selection of interference mitigation parameters and thus maintain a suitable performance trade-off between the HNB and Macro layers.

The following statement taken from [5] represents the preliminary conclusions drawn by WG4 HNB study group:

Dedicated Carrier Deployment

• To the extent investigated so far, dedicated carrier deployment is feasible for both open and closed subscriber group systems. Further work is required to investigate uplink co-existence between HNB and Macro layers and to investigate interference mitigation for very high CSG HNB deployment densities.

Co-channel Deployment

• To the extent investigated so far co-channel deployment is feasible for open access.

• For closed access, analysis conducted so far indicates that co-channel deployment is feasible if adaptive interference mitigation techniques are used. Further work is required to summarise the trade-off between HNB performance and the impact on the macro layer and to determine whether an acceptable tradeoff can be identified.5

Further work is still required by the 3GPP HNB study group for high density scenarios for the dedicated carrier scenario.

5.2 Closed Subscriber Group and Co-Channel/Adjacent Channel Deployment

Early simulations [15] presented initial Home NodeB coexistence simulation results in downlink (macro cell victim, home NodeB interferer). Home NodeBs were placed randomly in each macro cell. It was assumed the transmit power of each home NodeB was equal either to 20dBm or 24dBm. Femto cell penetration loss of 10dB and macro cell penetration loss of 0dB was considered (macro cell terminals are outside the buildings). The following two figures provide the capacity degradation caused by various densities of Femto Cells (measured as Femto Cells per cell site) for a transmit power of +20dBm or +24dBm and Closed Subscriber Group. It was clear from these early results that some form of interference mitigation for a co-channel deployment would be required although the impact on capacity due to adjacent channel deployments would appear to be manageable.

5 Hence the purpose of the simulation results presented in this paper are to investigate the effect of a power adaptation within the maximum DL transmit power range of +10dBm to -10dBm while considering the capacity enhancement of offloading the traffic from the Macro Layer to the Femto Layer.



Figure 1: Capacity Degradation due to Femto Cells placed on the same carrier as the Macro Layer assuming fixed transmit power levels [15]. Macro Cell Terminals are outside the building.

Figure 2: Macro cell capacity degradation vs number of home NodeBs per macro cell, ACIR=30dB, macro cell terminals are outside the buildings



Later Ericsson produced an excellent set of technical papers [19], [20], [21] where they investigated the effect of a CSG on the Macro and Femto Layer HSPDA performance where the Home NodeBs are deployed on a co-channel and adjacent channel for fixed transmit and adaptive transmit power levels. Their results quantify the reduction in relative average bit rate. In general it could be concluded from their results that for a CSG on the adjacent carrier that for maximum HNB transmit power levels of less than +10dBm and Femto Cell densities of up to 500 per cell site that the impact to the Macro Layer HSDPA average bit rate was approximately 2%. For the co-channel CSG scenario for the same scenario the impact could be as large as a 50% reduction in relative average bit rate. Their results clearly showed the benefit of adaptive power management based on Macro Layer interference. However, their study didn’t include the effect of migration of users from the Macro Layer to the Femto Layer.

5.3 Open Subscriber Group and Co-Channel Deployment

The results in this subsection were taken from [16] where Alcatel-Lucent investigated the feasibility of user deployed Home NodeB on the same carrier as an existing macro-cell network for open access and power management of Home nodeBs. Key requirements for co-channel operation of Home NodeB such as adaptive power management and open access were analyzed. The resulting impact on the existing macro-cell network was also investigated.

A key requirement for co-channel Home NodeB deployment is to keep the interference caused by Home NodeB low enough to ensure a low impact on the performance of the existing macro-cell network, while still ensuring enough transmit power for Home NodeB to achieve the target coverage and quality of services. In the downlink, both the pilot power that defines the cell range, and the maximum transmit power (to limit interference) must be configured. In the Alcatel-Lucent simulation, the transmit power for each Home NodeB is set to a value that is on average equal to the power received from the closest macro-cell at the target cell radius, subject to a maximum power. The pilot power is also set with respect to the total power (around 1/10th of the total power) to achieve the target range. In the uplink of the Alcatel-Lucent simulation, the UE power was limited to a value that limits the aggregate interference of all Home NodeB UEs to the closest macro-cell to a pre-defined value (subject to a maximum power). This way it was guaranteed that the uplink of the macro-cell is not degraded significantly.

Another key requirement for co-channel operation, when only one macro-cell frequency is available was to allow open access on all Home NodeBs in order to prevent excessive interference for UEs of the same operator located close to the Home NodeB.

0 5 10 15 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

← 64-QAM rate 1/2

← 16-QAM rate 1/2

← 4-QAM rate 1/2

adding femtocells has onlylittle impact on themacrocell performance

Downlink throughput 3dB from the capacity limit [Mbit/s]

macrocell, N = 0macrocell, N = 10macrocell, N = 100femtocell, N = 10femtocell, N = 100

Figure 3: CDF of downlink throughputs for co-channel macro-cell and Home NodeBs

The Down Link and Up Link results are summarised in Figure 3 and Figure 4 respectively and it is apparent that there is little impact on the Macro Layer throughput as a result of reasonable densities (100 Femto Cells per cell site) and Lucent-Alcatel concluded that co-channel deployment of Home NodeB can be achieved with only minor impact on the macro-cell throughput if power configuration and open access are used in the system.



0 5 10 15 20 25 30 35 400

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

← 64-QAM rate 1/2

← 16-QAM rate 1/2

← 4-QAM rate 1/2

adding femtocells has onlylittle impact on themacrocell performance

Uplink throughput 3dB from the capacity limit [Mbit/s]

macrocell, N = 0macrocell, N = 10

macrocell, N = 100femtocell, N = 10femtocell, N = 100

Figure 4: CDF of uplink throughputs for macro-cell for both macro-cell and Home NodeBs.

5.4 HNB to HNB Deployments

The following subsection provides a subset of the results taken from [22] where the HSDPA performance of collocated Femto Cells is evaluated. Although these results are not strictly relevant to this white paper (as they do not also consider the effect to the Macro Layer) they are included in order to complete the 3GPP review of interference studies.

The paper [22] discusses the downlink co-existence scenario between co-channel HNBs HSDPA performance and considers the impact of the HNB traffic and the end-user behaviour. Even though the system may have a large density of HNBs, it is not that likely that all of them are scheduling a user, or users, with full downlink power at the same time. In the paper, a random process was assumed, defining which of the modelled HNBs are active, i.e. scheduling a user with PHNBmax, during the simulated time instant (“snapshot”), while the other HNBs are assumed to transmit only the common control channels.

In the model assumed in the paper, the system area is divided into 10x10 m bins, “apartments”. At the beginning of each simulation snapshot, a predefined number of HNBs, either 200 or 500, are placed in random locations within each of the macro cells. The process to define the location of each HNB consists of the following steps: a) the apartments at which the HNBs were located are randomly selected from the group of apartments belonging to a certain cell b) each HNB was assigned a floor level between 1 and 6, c) a check was made that none of the apartments contain more than one HNB, d) the HNBs are placed in random coordinates within the selected apartments.

In order to study the impact of the level of HNB HS activity (traffic models, end-user behaviour), each HNB had a certain probability, “HNB HS utilization”, of being active, i.e. scheduling a HS-DSCH. The decision of whether a certain HNB was active or not affects the output power of that HNB: if the HNB was not active, it was assumed to be transmitting only the common control channels, 20% of PHNBmax, while an active HNB is assumed to be transmitting with the maximum power, PHNBmax.

Next, a home UE (HUE) was placed in a random location within each of the apartments with an active HNB. A check was made that the P-CPICH Ec/I0 towards the serving HNB is at least equal to -18 dB. If that is not the case, the HUE is re-dropped into a new location within the apartment until an acceptable P-CPICH quality has been reached.

The simulation results presented in [22] show that the interference between HNBs increases as the HNB density and/or the HNB HS utilization increase. However, the results demonstrate that even with a maximum HNB output power equal to 5 dBm, the assumed 10x10m apartment would be fully covered.



0 20 40 60 80 1000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

HNB HS Utilization [%]

CP

ICH

Out

age

Pro

babi

lity

[%]

Home Areas

200 HNB/cell. 5 dBm200 HNB/cell. 10 dBm200 HNB/cell. 15 dBm500 HNB/cell. 5 dBm500 HNB/cell. 10 dBm500 HNB/cell. 15 dBm

Figure 5: P-CPICH outage probability for the simulated HUEs with different levels of HNB HS utilization and with different HNB densities and PHNBmax.

0 20 40 60 80 1004

4.1

4.2

4.3

4.4

4.5

4.6

4.7

4.8

4.9

5

HNB HS Utilization [%]

HU

E A

vera

ge B

it R

ate

[bps

/Hz]

Home Areas. Max 19.3 Mbps

200 HNB/cell. 5 dBm200 HNB/cell. 10 dBm200 HNB/cell. 15 dBm500 HNB/cell. 5 dBm500 HNB/cell. 10 dBm500 HNB/cell. 15 dBm

Figure 6: The reduction on HNB Average Bit Rate as a function of HNB transmit power and densities.

With reference to Figure 5 and Figure 6 the authors of [22] state that according to the results shown here, the inter-HNB interference has a clear impact on the HNB downlink performance, but it is by no means dramatic, and in any case, the achievable HNB downlink performance is splendid, exceeding in many cases the capabilities of HSDPA with MIMO and 64QAM. Measurements of collocated Femto Cells in terms of CPICH quality [2] (which can be related to HSDPA throughput) support the conclusions drawn by Ericsson.



6 Description of the Network Simulation Tool

To investigate the impact of large femto cells deployments, UbiquiSys decided to use a network simulation approach whereby a layer of femto cells in various deployments scenarios could be analysed alongside a macro 3G network. A review of UMTS macro network simulation platforms available at the time revealed the lack of a suitable platform that could model a layer of femto cells with adaptive power management [13]. As an alternative, UbiquiSys teamed up with leading wireless propagation experts Red-M to modify an existing open source network simulation platform [8] to support a layer of femto cells. This particular platform has been widely used in academic papers. In addition, Red-M performed an independent validation of the platform prior to work commencing [13].

The Red-M network simulator is a link-level simulator that models network coverage and throughput in the “steady state”. It does this by balancing powers in the uplink and downlink so they remain within design parameters. To ensure convergence the simulator can put UEs to “outage” (i.e. drop the call). Inputs to the simulator include the network elements i.e. the macro layer and femto layer base station positions and parameters and UE positions and parameters. The network simulator can use pathloss data from a planning tool or it can calculate pathloss using various models and antenna patterns. The network simulator contains an iteration loop which includes the outcome of the RRM power management algorithms. A block diagram of the network simulator is provided in Figure 7

Scenario Generator

Planning ToolPathlossFiles

PreprocessPathloss

Model Pathloss

Iteration Loop

Store

Previous:ZAP

positionsUE positions

Pathloss

ZAP RRM

Building Types

Macro Layer UEDistribution

ZAP Layer Distribution

KPIs

If available

Simulator Structure

Set Initial ZAP PowersZAP Layer on/off Monte-Carlo Analysis

KPI Stats

Statisticallikelihood of key

KPI values

Figure 7: Block Diagram of Static Network Simulator

The following details are applicable to the network simulator:



• The blocks within the “Scenario Generator” represent the functionality required to allow editing, saving and loading of the Macro Layer and ZAP Layer entities required for the simulation.

• The “Iteration Loop” block in the diagram above represents the simulator iterative process, which

attempts to find a stable set of uplink and downlink powers that satisfy the power and service requirements of the network(s). Note the “ZAP Layer on/off” block that switches the ZAP Layer on once convergence has been reached with the Macro Layer only.

• The “Store” block represents the saved scenarios, which may be reused or used as the starting point for

new scenarios if required.

• The “KPIs” blocks are the Key Performance Indicators that are available once the simulation has finished running.

A key requirement for this simulator was to have the ability to capture the randomness and variability likely to be associated with the deployment of femto cells, thus providing a realistic dimension to the analysis. Subscribers with femto cells will install them in locations within their homes outside the control of the network operator. To capture this variability the simulator has the capability to model various building types ranging from terraced houses, bungalows, and apartment blocks in a 3-dimensional space each with its own unique radio propagation properties within which the femto cells and mobiles can be randomly located. Fifteen building types were selected to provide the range of buildings most commonly encountered in the UK. These buildings were then given typical dimensions and radio characteristics (external wall penetration loss, floor penetration and variability) derived from the literature. These can be seen in Table 1.

BuildingT

ype Description

Width

[m]

Depth

[m]

Height

[m]

Nbr of

Floors

Internal

Loss

Internal

Loss Dev

Floor

Loss

Floor Loss

Dev

Ext wall

Loss

Ext wall

Loss Dev

1 Modern_Apartment 5 6 5.6 2 4 4 4 4 8 6



4 Modern_Apartment 5 6 14 5 4 4 4 4 8 6

5 Old_Apartment 6 8 6.4 2 7 6 6 4 10 6

6 Old_Apartment 6 8 9.6 3 7 6 6 4 10 6

7 Old_Apartment 6 8 12.8 4 7 6 6 4 10 6

8 Old_Apartment 6 8 16 5 7 6 6 4 10 6

9 Large_Detached 18 12 8 2 6 5 6 4 12 8

10 Small_Detached 10 6 6 2 6 5 6 4 12 8

11 Terraced 6 8.5 6 2 4 4 4 4 10 6

12 Large_Semi 20 10 8 2 6 5 6 4 12 8

13 Small_Semi 16 8 6 2 6 5 4 4 10 6

14 Town_House 10 9 11 3 7 6 10 6 12 8

15 Bungalow 10 10 5 1 4 4 4 4 9 7

Table 1- List of buildings types used in the simulations with their dimensions. Losses due to external wall penetration [dB], internal loss and loss through floors [dB] are also provided, together with the standard

deviation of the lognormal variability associated with each of these losses.

Key to any network type simulator is the link performance prediction which requires accurate path loss prediction as this directly impacts the interference power levels within the network and hence the final results. The path loss modelling in this network simulator places particular emphasis on predicting path loss for the network elements located indoors i.e. the femto cells and mobiles located indoors.

The network simulator uses established empirical models to predict path losses between network elements which include the COST231–Hata, ITU P.1411 and the ITU P.1238, with additional losses added to cater for the signal propagation through external walls. The following is a summary of the path loss modelling used in the simulator:

• For the low height femto cells to mobile the ITU P1411 is used, as this model satisfies most of the requirements in terms of frequency range, height of base station and distance validity.



• For the indoor links (femto cell to indoor UE or internal portions of the femto basestation to macro basestation and femto basestation to outdoor UE), the ITU P.1238 is used.

• An extra variability is added to the median loss predicted by the above models as a result of shadowing. The shadowing contributions from the various mechanisms (outdoor propagation, through wall/floor) are treated as independent and uncorrelated quantities and are therefore added separately to the median path loss to provide realistic values encountered during measurements. Due to the randomness assumed in the deployment Monte Carlo simulation runs were carried out to determine a mean of the results.

• Finally, due to the random nature of the added variability, a check is performed on all resulting losses against the output from the free space model to ensure no loss can fall below free space loss.

During the validation of the pathloss module, the outputs of the link loss models were compared against real measurements. The measurements shown in Figure 8 were obtained from a live network in suburban Swindon using an outdoor receiver. The graph shows that there is a very good agreement between the two sets of data in the distance range where measurements were made. Further measurements were obtained in the same area using an indoor receiver. The measurements were made in a semi-detached house located about 200m from a macro basestation. The receiver was moved around the house while measuring RSCP levels both from a Femto Cell positioned in the house and from the node B. Link loss levels were then derived from the RSCP measurements and the probability density function of the link loss from the femto cell and from the node B are shown in Figure 9.

Figure 8 Comparison between linkloss model predictions (blue) and measurements (red)



Figure 9 Linkloss distribution measured in a house 200m away from a macro basestation (red). Measured linklosses from a ZAP located in the house are also shown in blue.

The measured link loss from the femto cell ranged between ~50dB and 90dB and were found to be comparable to those computed by the model.

At two hundred metres, COST231 predicts the linkloss to vary between ~120dB and 155dB and is in good agreement with the range of measured levels shown in Figure 9, bearing in mind that the external wall loss used in the model is an estimate and that there is a 7dB standard deviation for the lognormal fading accounted for in the model predictions.

The end result was a network simulation platform capable of modelling a macro network with up to 2 carriers and the option of deploying femto cells on either of the two carriers. The network simulator was thoroughly tested and validated by UbiquiSys and Red-M [11]. Some of the additional features of the modified network simulator are summarised as follows:

• Deployment of femto cell layer on the same carrier as the Macro Layer.

• A range of Rel 99 3GPP speech and data services.

• Up to two carriers adjacent to each other for macro and/or femto deployment. The simulator could model any combination of macro layer and femto cell deployment across 2 carriers. Inter-carrier handover is supported on the Macro Layer but not supported on the femto layer.

• Closed subscriber group for Femto Layer and migration of users from Macro Layer to Femto Layer.

• Femto cell carrier selection. The simulator models the carrier selection algorithms used in the UbiquiSys femto cell. When the femto cell has the choice of operating on more than one carrier, the carrier selection algorithm will select the appropriate carrier where it will create the least amount of interference.

• Fixed or adaptive power management within the femto cells. The adaptive power management techniques used in the UbiquiSys femto cell is modelled within the simulator. As the simulator provides a snapshot at a given instant in time the power settings from the adaptive power management in the simulator represent the expected converged power levels of the femto cell layer over time.

• Modelling of network elements inside a range of buildings. The simulator is capable of modelling different buildings as 3 dimensional spaces within which the femto cells and mobiles can be located.

A detailed description of the simulation platform design and validation is provided in [10], [11], [12] and [13].



7 Femto Cell and Macro Layer Network Layer Simulations

In this section we present the analysis of simulated deployments of femto cells in typical dense urban and suburban 3G Macro Layer coverage areas. In this section we refer to spectral efficiency which we define as a measure of the combined macro and femto layer throughput on a given carrier as bits/sec/Hertz/unit area.

The analysis presented aims to extend previous work by the 3GPP, and in particular the analysis carried out by Nortel Vodafone [14] . We build up our arguments as follows:

� For a given network subscriber base it is a reasonable assumption in today’s market that a sudden and large change in the subscriber base is unlikely to occur. Bearing this in mind we present the case for femto cells as a means of relieving the macro layer base stations of the more resource demanding indoor subscribers.

� Femto cells are more suited to the indoor environment as this is the environment they have been designed and optimised to operate within. The femto cell should achieve a higher throughput per subscriber than an outdoor macro cell within an indoors environment since effectively a carrier of capacity is available for each femto cell.

� We acknowledge that every femto cell introduced into the network is likely to be a constant additional source of interference as it is likely that the femto cells will be left on 24 hours per day. However, we also believe, that in between the extremes where the femto cells are not serving any of the subscriber base, to where the femto cells are serving all the subscribers, there must be a percentage traffic split between the macro and femto layers whereby the overall capacity is the same as the macro capacity without the deployment of femto cells. Beyond such points of breakeven, any traffic offload onto the femto cells is a benefit to the overall network.

� Adaptive power setting is crucial to femto cell operation not only from an interference point of view but also from a coverage area point of view. It is desirable that the adaptive power management takes into account the femto cell coverage area so as it ensures that femto cells can adapt in order to cover a range of different sizes and composition. For example within the same interference environment a femto cell in a studio flat should ideally set its power much less than in a three storey mansion as the coverage area is likely to be a fraction of that of a three storey mansion.

� Finally, with the promise of new waves of exciting mobile applications being unleashed partly due to the femto cell capabilities, it is likely that indoor usage patterns will grow at a much faster rate than usage pattern outdoors. This is where the femto cells will provide the cushion required as the macro network continues to grow on its planned evolution path.

The results provided in this section include:

� In subsection 7.1 we provide results of a simulated femto cell deployment in a dense urban 3G coverage area. An area in central London (UK) was selected which has all the properties of a dense residential area in the heart of an urban centre. The results show the potential improvements in spectrum efficiency that can be brought about by a layer of femto cells by comparing against the optimal spectrum efficiency that can be achieved with such a 3G macro network.

� In subsection 7.2 we analyse the deployment of femto cells in a suburban 3G coverage areas. An area in Swindon (UK) was selected which has all the properties of a typical suburban residential area. The analysis shows the proportion of network traffic that needs to be offloaded to reach a throughput-interference break-even point for both adaptive power setting and fixed power levels .

As the simulation is of the static type which gives a snapshot of the network averaged over multiple simulation runs (i.e. Monte Carlo analysis), the femto cell adaptive power management simulated is an approximation to the steady state value of the UbiquiSys femto cell power management techniques.



7.1 Femto Cells in an Urban Environment

7.1.1 Scenario

In this section we analyse the deployment of femto cells in a 2.5km2 dense urban the London (UK) area known as Bayswater. Bayswater is a built up residential area typical of dense urban environments in the UK. Most of the buildings in this area are apartment blocks and Georgian terraces.

Existing 3G sites from a UK operator were positioned in and around the area based on information gathered from Ofcom’s site finder data [23]. An aerial view of the area, with potential building containing femto cell and the location of the macro layer node Bs is shown in Figure 10. This particular 3G coverage area is representative of a 3G network deployment designed to cope with a high traffic levels. The area consists of 4 tri-sectored sites with antenna heights of over 20m and two low height micro cells interweaved within the macro coverage areas. The simulated scenario consists of these base stations as well as randomly positions buildings within which the femto cells are deployed.

Figure 10 Femto deployment scenario in London. Node B sectors are in red and buildings with femto are in blue. The simulated area is highlighted by the white overlay.

To capture a more realistic femto cell deployment densities, likely densities were approximated as follows:

� From the aerial shot shown in shown in Figure 10 of the 2.5Km area highlighted by the white overlay, we can assume that roughly 25% of the simulation area is covered by roads, hence leaving approximately 0.7km2 of buildings

� Assuming an average house horizontal footprint of 10m by 10m, and on average the building are apartment blocks with 4 storeys then we can fit approximately 30, 000 residences in this area.

� We can further assume that 25% of the area is occupied by buildings that cannot have femto cells e.g. school classrooms, hospitals, shopping centres etc, which leaves us in the region of 20,000 residences

� If across the four UK operators we assume equal market share and further assume that only 25% of these residences are likely to have broadband, 3G phones and femto cells then we have about 1,250 potential residences per operator in this 2.5km2 area.



Based on the above, we assume for this particular area an operator can likely have up to 1000 femto cells (160 femto cells per site) deployed which forms the basis of our simulated femto cell density. In addition the following points apply to the simulated scenario:

� Macro cell powers in region of 40dBm with antenna heights above 20 meters (Ofcom’s site finder data [23]). 18dBi antenna with 65 degree down tilt used.

� Micro cell powers in region of 35dBm with antenna heights below 5 meters (Ofcom’s site finder data [23]). Omni directional antenna with 11 dBi gain.

� Maximum macro UE allowed transmission power of 24dBm

� Femto cell powers total downlink power fixed at 10dBm across network. The femto cell pilot signal power (CPICH) is set to 10% of the total femto cell downlink power

� Maximum femto cell UE transmit power set to 10dBm

� Buildings

o 400 Randomly distributed apartment blocks

o Apartments blocks located at least 10m from the nearest base station

o Apartment blocks located at least 5m to the nearest building. This also implies no 2 buildings are located in the same position

o Apartment blocks containing 5 apartment units stacked on top of each other with dimension 6 meters wide, 8meters long and 2.8 meters high (see Table 1 for more details)

o Up to 3 femto cells allowed in a given apartment block (i.e a maximum of 3 out of 5 floors can have a femto cell) , no more than 1 femto cell per apartment, each serving a single UE

� Mix of speech & data traffic at 3kmph & 50kmph (indoor users all simulated at 3kmph). 15% of traffic on data services.

7.1.2 Results

In this section we summarise some of the observations from a snapshot of the dense urban network described in the previous section.

Figure 11 shows the status of users indoors and outdoors attempting to access the macro network without femto cells deployed when the macro network is experiencing high level of loading. In this case there is 23% outage level across the network, mainly due to the lack of lack of downlink power. In this state the network is able to achieve a spectrum efficiency of 0.45 bits/sec/Hertz/km2.

The case of the same dense urban network is shown with a deployment of 1000 femto cells in Figure 12, out of which 250 are serving the indoor users previously served by the macro network i.e. 25% of the femto cells are actively carrying traffic and a 100% of the networks indoor traffic has been offloaded onto the macro layer. It is clear that in this case the femto cells provide huge benefits as the overall network outage is down to 1.4%. Also the spectrum efficiency has increased by 28% to 0.68 bits/sec/Hertz/km2



0 20 40 60 80 100 120 140 160 180 200 220 240-6

-4

-2

0

2

Macro-UE index

Car

rier

num

ber

Outdoor Macro UE

0 20 40 60 80 100 120 140 160 180 200 220 240-6

-4

-2

0

2

Macro UE Index

Car

rier

num

ber

Indoor Macro UEs

Figure 11 Outage levels in dense urban 3G coverage area without femto cells at high network loading6

0 20 40 60 80 100 120 140 160 180 200 220 240-6

-4

-2

0

2

Macro UE index

Car

rier

num

ber

Outdoor Macro UE

0 20 40 60 80 100 120 140 160 180 200 220 240-6

-4

-2

0

2

Femto UE index

Car

rier

num

ber

Indoor Femto UEs

Figure 12 Outage level in dense urban 3G coverage area with 1000 femto cells. All of the networks indoor traffic has been offloaded onto 25% of the femto cells. More outdoor users are now served.

6 A carrier number of 1 implies that the UE is being served. Any other number (-1, -2, -3, -4, -5 and -6) implies that the UE is in outage.



Figure 14 gives an idea of the coverage quality indoor versus outdoor with and without femto cells in the network respectively. Figure 14 shows that with the indoor traffic offloaded on the femto cells, the indoor UEs served by the femto cells enjoy a much better signal quality. In addition the outdoor UEs appear to be enjoying the same signal quality prior to when the femto cells were deployed. This could be explained by the fact that macro intra-cell interference caused by the high powered indoor users has been replaced by interference from the femto cells. However, the interference from the femto cells appears not to be significant enough to cause a noticeable impact on coverage as indicated by the comparison of Figure 13 and Figure 14.

-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 0

10

20

30

40

CPICH Ec/Io, dB

% o

f m

obile

s

CPICH Ec/Io distribution of outdoor macro mobiles

-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 0

10

20

30

CPICH Ec/Io, dB

% o

f m

obile

s

CPICH Ec/Io distribution of indoor macro mobiles

Figure 13 CPICH EcNo distributions for network mobiles without femto cells in dense urban network

-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 0

5

10

15

20

CPICH Ec/Io, dB

% o

f m

obile

s


-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 0

20

40

60

CPICH Ec/Io, dB

% o

f m

obile

s

CPICH Ec/Io distribution of indoor femto mobiles

Figure 14 CPICH EcNo distributions of network mobiles with 1000 femto cells deployed in dense urban network. All of the network’s indoor traffic has been offloaded onto 25% of the femto cells (i.e. only 250

out of 1000 femto cells are carrying traffic).



The results in this section show that a closed access fixed power co-channel femto cell deployment can coexist with the macro layer. However these results show a favourable scenario, in that all the indoor network traffic is assumed to be offloaded onto the femto cells. At this stage the dense urban deployment case is still under study with a detailed analysis of femto cell deployment in this type of 3G coverage area using adaptive power management at various deployment densities and migration ratios to follow. It is expected that the application of adaptive power management will reduce the power levels of the femto cells hence reducing the interference impact on the macro network.

7.2 Femto Cells in a Suburban Environment

7.2.1 Scenario

In the 2nd study, a 4km2 suburban area in Swindon (UK) known as Abbey Meads is analysed. The Abbey Meads area in Swindon can be considered as a typical suburban area that provides a good mix of residential buildings and pockets of open parks typical of suburban areas encountered in the UK.

Existing 3G sites from a UK operator were positioned in and around the area based on information gathered from Ofcom’s site finder data [23]. An aerial view of the area, with potential buildings containing femto cell buildings and the location of the ML node Bs is shown in Figure 15. This particular 3G coverage area is representative of a 3G network deployment designed such that cells have a relatively larger converge area and cope with lower traffic levels than the case presented for the dense urban environment. In the Abbey Meads scenario, the macro layer consists of 12 node Bs distributed over 4 sites (the sectors from the 3 sites are marked in green on Figure 15; the 4th site is located outside the image to the southeast of the area). The simulated scenario consists of these base stations as well as randomly deployed buildings within which the femto cells are deployed.

Figure 15: Femto deployment scenario in Abbey Meads. Node B sectors are in green and buildings with ZAPs are in red. The area is about 2km x 2km in size

To capture more realistic femto cell deployment densities, likely densities were approximated as follows:

� From the aerial shot we can assume that 50% of the simulation area covered by green area and roads, hence leaving approximately 2km2 of buildings

� Assuming an average house horizontal footprint of 10m by 10m, then we can fit 20, 000 buildings. (Swindon planning authority quote 40 to 75 units per hectare in suburban areas, this would place Abbey Meads within the 50 units per hectare range)



� We can further assume that 25% of the area is occupied by buildings that cannot have femto cells e.g school classrooms, hospitals, shopping centres etc, which leaves 15,000 residences

� If across the four UK operators we assume equal market share and further assume that only 25% of these residences are likely to have broadband, 3G phones and femto cells then we have about 937 potential residences per operator in this 4km2 area.

Based on the above, we assume for this particular area an operator can likely have up to 1000 femto cells (250 femto cells per site) deployed which forms the basis of our simulated femto cell density. In addition, the following points apply to the simulated scenario.

� Macro cell powers in region of 40dBm with antenna heights above 15 meters (Ofcom’s site finder data [23]). 18dBi antenna with 65 degree down tilt used.

� Maximum macro UE allowed transmission power of 24dBm

� Femto cell powers set via simulated UbiquiSys femto cell power management used for femto cell Dl/UL power settings. Powers allowed to vary over the following range

o Femto total downlink power -10dBm to 10dBm

o Maximum allowed femto cell UE transmission power-10dBm to 10dBm

� The femto cell pilot signal power (CPICH) is set to 10% of the total femto cell downlink power

� Buildings

o Randomly distributed semi-detached buildings with dimension 8 meters wide, 16meters long and 8 meters high (see Table 1 for more details)

o Buildings located at least 50m from the nearest base station

o Buildings located at least 10m to the nearest building. This also implies no 2 buildings are located in the same position

o One femto cell per building, each serving no more than one UE at a given time

� Mix of speech & data traffic at 3kmph & 50kmph (indoor users all simulated at 3kmph). 15% of traffic on data services.

7.2.2 Results

In this section we present results of various analyses of femto cell deployments in a suburban environment. Results, analysing a scenario where the network traffic is entirely voice, are presented as well as a scenario where 15% of the network traffic is assumed to be for data services.

Figure 11 shows the status of 150 users indoors and outdoors attempting to access the macro network in the Abbey Meads area without femto cells deployed when the network is at an optimal loading level. In this case there is a 15% outage level across the network, mainly due to a combination of both uplink loading and lack of downlink power.



0 20 40 60-6

-4

-2

0

2

UE index

Car

rier

nu

mb

er

Outdoor UEs

0 20 40 60-6

-4

-2

0

2

UE index

Car

rier

nu

mb

er

Indoor UEs

Figure 16 Outage level in suburban 3G coverage area

Figure 17, Figure 18 and Figure 19 shows the impact on spectrum efficiency with various proportions of total network traffic offloaded onto the femto cells for the case where adaptive power management, fixed power of +10dBm and fixed power of +20dBm are used respectively. For the expected deployment densities in this area in the region of 700 and 1200 femto cells an impact of -5% can be maintained when only 16% of the total network traffic (i.e 25 out of the 150 users) is offloaded onto the femto cells and the femto cells are using adaptive power management as shown in Figure 17. However, this is not the case for the fixed power cases. As shown in Figure 18 it requires a 50% traffic offload on to the femto cells to achieve any benefit, whereas it appears almost impossible to achieve any benefit when using a fixed power of +20dBm. Figure 20 shows a comparison of the adaptive power management and the fixed power cases which further highlights the importance power management as implemented by UbiquiSys in its ZAPs. Figure 21 shows a typical power distribution of the UbiquiSys RRM approach in such a 3G coverage area. An idea of the coverage quality with this power levels is shown in the CPICH EcNo distribution of the network mobiles in Figure 22.

From the results a couple of observations are made:

• For the following example where 16% of the macro layer traffic (150 calls) are running on the femto cell layer it implies that only 25 calls are running on approximately 700 to 1200 femto cells (at the break-even point) which is clearly a low utilization of the femto cell population (<5%).

• Offloading data calls off the macro layer onto the femto layer improves the spectral efficiency more than offloading voice traffic as a greater proportion of macro layer transmit power is required to service a data call7. Hence by analysing the effect of voice traffic provides a lower bound in terms of the percentage of traffic of load that is required to meet the break even point. The break even point for data only users was found to occur at a relatively low proportion of offload (<10%) and hence a value of between 10 to 15% is currently expected to be the norm.

• It can be seen that the femto cell layer interference effects are at their worst when there is little traffic on the femto layer. It is likely however that in this event there will be little traffic on the macro layer (i.e. a consumer purchases a femto cell to use it) and hence it could be argued that the femto cell takes

7 This was confirmed through simulations although the results are not published as yet in this report.



away capacity when there is spare capacity in the network but improves spectral efficiency when most needed (i.e. when the macro layer is heavily loaded).

0 200 400 600 800 1000 1200 1400-20

-15

-10

-5

0

5

10

15

20

25

Number of Femto Cells in Network

Spe

ctru

m E

ffic

einc

y Im

pact

, %

Femto Cells with Adaptive Power Management. 150 Users, 75 Indoor, 75 outdoor. Mixed Traffic;

0% migrated

16% Migrated

33% migrated

50% migrated

Figure 17 Spectrum efficiency impact for femto cells with adaptive power management in a suburban 3G coverage area

0 200 400 600 800 1000 1200 1400-60

-50

-40

-30

-20

-10

0

10

20

30


Spe

ctru

m E

ffic

einc

y Im

pact

, %

Femto Cells with Fixed Power +10dBm. 150 Users, 75 Indoor, 75 outdoor. Mixed Traffic;

0% migrated

16% Migrated

33% migrated

50% migrated

Figure 18 Spectrum efficiency impact for femto cells with fixed power of 10dBm in a suburban 3G coverage area



0 200 400 600 800 1000 1200 1400-80

-70

-60

-50

-40

-30

-20

-10

0

10

20


Spe

ctru

m E

ffic

einc

y Im

pact

, %

Femto Cells with Fixed Power +20dBm. 150 Users, 75 Indoor, 75 outdoor. Mixed Traffic;

0% migrated

16% Migrated

33% migrated

50% migrated

Figure 19: Spectrum efficiency impact for femto cells with fixed power of +20dBm in a suburban 3G coverage area

0 200 400 600 800 1000 1200 1400-60

-50

-40

-30

-20

-10

0

10


Spe

ctru

m E

ffic

einc

y Im

pact

, %

Impact with 16% Network Traffic Migrated150 Users, 75 Indoor, 75 outdoor. Mixed Traffic;

Adaptive Power

Fixed +10dBm

Fixed +20dBm

Figure 20 Spectrum efficiency impact for femto cells in a suburban 3G coverage area with 16% of macro network traffic offloaded onto femto cells



.

-10 -8 -6 -4 -2 0 2 4 6 8 100

20

40

60

Power, dB

% f

emto

s

femto total DL power

-10 -8 -6 -4 -2 0 2 4 6 8 100

20

40

60

80

Power, dB

% o

f fe

mto

s

femto maximum UE power

Figure 21 UbiquiSys Adaptive power management distribution in typical suburban 3G coverage area

-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 0

10

20

CPICH Ec/Io, dB

% o

f m

obile

s


-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 0

10

20

CPICH Ec/Io, dB

% o

f m

obile

s

CPICH Ec/Io distribution of indoor macro mobiles

-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 0

50

CPICH Ec/Io, dB

% o

f m

obile

s

CPICH Ec/Io distribution of femto mobiles

Figure 22 EcNo distribution of network mobiles with femto cells deployed in typical suburban 3G coverage area

The next set of results shows the case of a larger number of subscribers accessing the network for voice only services. Figure 25, Figure 24 and Figure 23 show consistency with the previous results. Again it appears that with adaptive power management this scenario will breakeven when 16% of network traffic is offloaded onto the femto cells and as expected it will be difficult to breakeven with the +20dBm fixed power case even when 40% of the network traffic is migrated. The comparison of the three power setting approaches in Figure 26 again



highlights the importance of the UbiquiSys adaptive power management. It is also clear that, as the number of users increases in the macro layer, the break even point occurs at a lower percentage of traffic migration. In the results in Figure 23 the break even point occurs at approximately 10% migration of traffic as apposed to 16%.

0 200 400 600 800 1000 1200 1400-30

-20

-10

0

10

20

30

40

50

60


Spe

ctru

m E

ffic

einc

y Im

pact

, %

Femto Cells with Adaptive Power Management. 300 Users, 150 Indoor, 150 outdoor. All user on Voice;

8% migrated

16% Migrated

25% migrated

33% migrated

Figure 23 Spectrum efficiency impact for femto cells with adaptive power management in a suburban 3G coverage area

0 200 400 600 800 1000 1200 1400-80

-60

-40

-20

0

20

40

60


Spe

ctru

m E

ffic

einc

y Im

pact

, %

Femto Cells with Fixed Power +10dBm. 300 Users, 150 Indoor, 150 outdoor. All user on Voice;

8% migrated

16% Migrated

25% migrated

33% migrated

Figure 24 Spectrum efficiency impact for femto cells with fixed power of 10dBm in a suburban 3G coverage area



0 200 400 600 800 1000 1200 1400-100

-80

-60

-40

-20

0

20


Spe

ctru

m E

ffic

einc

y Im

pact

, %

Femto Cells with Fixed Power +20dBm. 300 Users, 150 Indoor, 150 outdoor. All user on Voice;

8% migrated

16% Migrated25% migrated

33% migrated

Figure 25 Spectrum efficiency impact for femto cells with fixed power of +20dBm in a suburban 3G coverage area

0 200 400 600 800 1000 1200 1400-70

-60

-50

-40

-30

-20

-10

0

10

20

30


Spe

ctru

m E

ffic

einc

y Im

pact

, %

Network with 16% of traffic migrated onto femto cells

Adaptive Power -10dBm to +10dBm

Fixed Power +10dBm

Fixed Power +20dBm

Figure 26 Spectrum efficiency impact for femto cells in a suburban 3G coverage area with 16% of macro network traffic offloaded onto femto cells



8 Down Link and Up Link Power Adaptation Test Results

In this section we provide results of the adaptive power setting process used in the UbiquiSys femto cell in a range of residences of different sizes and different radio propagation properties. The Ubiquisys femto cell adaptive power setting aims to minimise the impact on the existing macro layer as well as ensuring that the femto cell provides coverage within the residence it has been deployed in. The adaptive power setting is done in two phases: in the first phase which occurs at power up, an initial estimate of the power levels are made based on measurements made on the surrounding macro layer and information decoded from their Broadcast Channels (BCH). The second phase is a continuous process whereby the power levels are optimised based on statistics calculated over time as the femto cell mobiles move around the residence.

In all the cases presented after the power setting process had completed, coverage measurements were performed as well as test calls to evaluate user perception of quality.

Figure 27 shows the total downlink and maximum femto cell uplink powers converging over time for a femto cell deployed in a 100m2 1960 semi detached home with internal brick and concrete walls in the Swindon Green Meadows area whereas Figure 33 shows the powers in a 50m2 modern 2 bed apartment with plaster board internal walls near the Swindon city centre. The initial power estimates were biased towards the high end to demonstrate the optimisation process over time. The time axis in Figure 27 and Figure 33 represents the total call time for all the mobiles on the femto cell.

In all cases the call remained up and the voice quality was good. Macro layer walk tests were conducted outside of the residence. The macro layer mobile was forced into 3G mode (i.e. 3G to 2G handover was disabled). In the case that the femto cell was located in the semi detached home a Macro Layer voice call was sustained all the way past the house in the street. The Macro Layer call also remained up until approximately 2 metres from the front door of the residence and then it dropped.

Conditions Initial Down Link Tx power

(dBm)

Initial Uplink Tx power

(dBm)

Down Link power after convergence

(dBm)

Up Link power after

convergence (dBm)

1960s Home on strong carrier with 90% CPICH quality of -10dB. Unit at door

8 0 10 -14

1960’s Home on average carrier with 90% CPICH quality of -6dB. Unit at door

8 10 10 0

1960’s home on average carrier with 90% CPICH quality of -10dB. Unit at the door

8 10 6 -3

1960s home on average carrier with 90% CPICH quality of -10dB. Unit in kitchen

4 2 3 3

Small apartment on a relatively weak carrier with 90% CPICH quality of -10dB

4 10 -26 7

Table 2: Summary of the test results in both residential home and modern apartment.



Figure 27: Down Link and Up Link Power convergence when 90% CPICH quality factor is set at -10dB and the Femto Cell was placed in the kitchen

Figure 28: The femto cell indoor path loss profile and the Macro Layer (nodeB) to femto cell UE path loss profile with the femto cell in the Kitchen.



Figure 29: Indoor femto cell path loss and femto cell to Macro Layer nodeB path loss measurements with the femto cell at front door.

Figure 30: 1960’s Semi Detached Home



Figure 31: Ground floor CPICH Ec/Io femto cell and Macro Layer Coverage after power adaptation with femto cell placed in the kitchen

Figure 32: Ground floor CPICH Ec/Io femto cell and Macro Layer Coverage after power adaptation with femto cell placed in Kitchen



Femto Cell CPICH

Ec/Io (dB)

Femto Cell CPICH RSCP

(dBm)

Macro Layer CPICH Ec/Io

(dB)

Macro Layer RSCP (dBm)

Dining Room -3.0 -68.9 n/a n/a

Sitting Room -2.8 -70 -21 -87

Kitchen -2.7 -56.5 n/a n/a

Scullery -3.0 -80.8 n/a n/a

Study/Backroom -3.0 -86.2 n/a n/a

Top of stairs -2.7 -78.8 -23.2 -98.8

Box Room -5.5 -84.4 -11.7 -90.3

Master Bedroom -9.5 -87.3 -7.1 -84.3

Second Bedroom -3.3 -82.9 -18.8 -98.3

Bathroom -2.5 -78.7 n/a n/a

Toilet -3.2 -89.2 n/a n/a

On road in front of adjacent house

-14.3 -101 -10.5 -96.5

On road in front of house

-19.2 -103 -6.8 -9.2

On road in front of adjacent house

-15.5 -98.7 -8.8 -92.5

In the garden at back of house

-5.4 -88.8 -12.4 -95

Table 3: The Femto Cell CPICH quality when the 90% CPICH Ec/Io target was set at -10dB and the Femto Cell was located in the kitchen

Figure 33: The Down Link and Up Link power adaptation in a small apartment where the 90% CPICH quality factor was set at -10dB.



Figure 34: The femto cell apartment indoor path loss and the Macro Layer to femto cell path loss measurements as collected from the femto cell UE.

Figure 35: Apartment block containing 2 bed modern apartment



9 Conclusions

This white paper is a detailed technical summary as a sequel to the earlier technical report written by UbiquiSys and Red-M [1] and addresses the following aspects:

• It provides a summary of the current 3GPP Home NodeB working group early conclusions in terms of the feasibility of coexistence of femto cells on co-channel and adjacent carriers with the Macro Layer network for both open and closed subscriber groups. In general the 3GPP community currently concludes that operation on an adjacent carrier with both open and closed subscriber groups and on the co-channel with an open subscriber group should be feasible.

• The paper then presents simulation results which analyse the network capacity impacts of DL and UL power management RRM algorithms for a closed subscriber group (CSG) operating on the same carrier as the Macro Layer and the effect of migrating traffic from the Macro Layer to the femto layer. The results suggest that typically when approximately 10 to 15% of the traffic has been migrated from the Macro Layer to the femto layer that there is a break even point where the capacity gain provided by the femto layer equals the capacity lost on the Macro Layer due to femto layer interference. If higher levels of traffic migration occur then there is a net positive gain in capacity.

• The paper also demonstrates the significant impact that power management can have in the position of the break even point. It is expected that future revisions of this paper will explore the effect of different operator network lay outs, different call densities supported by the femto cells and partially open access points.

• The simulations have typically assumed a 50:50 split between indoor and outdoor users. The view from a large UK MNO is that 70% of all voice calls take place within a building. Consequently it could be expected that even higher proportion of all data calls (>90%?) would take place indoors. Furthermore the average time duration of each call is 3 minutes and the average number of cell sites used per call is 1.06 (i.e. in general users are stationary). These factors imply that the split between indoor/outdoor users is somewhat conservative and hence in later simulation different splits of indoor/outdoor users will be investigated.

• From on-air testing it has been found that femto cells with transmit power levels of +10dBm and lower are capable of providing coverage that exceeds that of both Wireless LAN and DECT. Assuming that MNOs when rolling out femto cells suitably price this new technology it is possible, that over time there could be migration of some Wireless LAN and DECT traffic to the femto layer. This expected migration could be attributed to the simple installation process, advantages of managed licensed spectrum and single handset. From the simulations, it is clear that, as the expected utilization of femto cells increases, the Macro Layer alone would not be sufficient to carry the additional traffic if the femto layer was later removed.

Finally the paper has demonstrated that co-channel closed access operation with the Macro Layer is feasible and has explored in detail the cross over point caused by traffic migration when the femto cell is power adapted. Further simulations will be carried out to investigate different network topologies, traffic profiles and user locations to understand the effect of these parameters on the cross over point.



Annex A Change History

Version Authors Date Status Comment

0.0.1 Aminu Maida 20-Dec-2008 In work Initial Draft

1.0.0 Aminu Maida, Alan Carter 13-Jan-2008 In work Base lined after Red-M Review

Femtocell UBS-21!92!028 Femto Cell Macro Layer Study

Documents

Transcript of Femtocell UBS-21!92!028 Femto Cell Macro Layer Study