012 OFDMA Interference Management

www.femtoforum.
org
Interference
Management
in OFDMA
Femtocells
Published by the Femto Forum
March 2010
Interference Management in OFDMA Femtocells is
published by the Femto Forum
March 2010. All rights reserved.
www.femtoforum.org
telephone +44 (0)845 644 5823 fax +44 (0)845 644 5824 email info@femtoforum.org PO Box 23 GL11 5WA UK
What is the Femto Forum?
The Femto Forum is the only organisation devoted to promoting femtocell technology
worldwide. It is a not-for-prot membership organisation, with membership open to providers
of femtocell technology and to operators with spectrum licences for providing mobile services.
The Forum is international, representing more than 120 members from three continents and
all parts of the femtocell industry, including:
l Major operators
l Major infrastructure vendors
l Specialist femtocell vendors
l Vendors of components, subsystems, silicon and software necessary to create femtocells
The Femto Forum has three main aims:
l To promote adoption of femtocells by making available information to the industry and the
general public;
l To promote the rapid creation of appropriate open standards and interoperability for
femtocells;
l To encourage the development of an active ecosystem of femtocell providers to deliver
ongoing innovation of commercially and technically efcient solutions.
The Femto Forum is technology agnostic and independent. It is not a standards-setting body,
but works with standards organisations and regulators worldwide to provide an aggregated
view of the femtocell market.
A full current list of Femto Forum members and further information is available at
www.femtoforum.org
The Femto Forum: Interference Management in OFDMA Femtocells
.
page 1 www.femtoforum.org
Contents
1 Executive Summary ......................................................................................................... 7
2 Femtocells, Femtocell Access Points and the Femto Forum .............................................. 8
2.1 What are Femtocell Access Points? ............................................................................ 8
2.2 What Do Femtocells Offer? ....................................................................................... 9
2.3 What is the Femto Forum? ...................................................................................... 10
3 Introduction .................................................................................................................. 11
3.1 Objectives and Methods of this Paper ..................................................................... 11
4 Simulation Scenarios Description .................................................................................. 13
5 Abbreviations and Defined Terms ................................................................................. 16
6 Scenario A1: Macrocell Downlink Interference to the Femtocell
UE Receiver (Traffic Channel) ......................................................................................... 17
6.1 Description ............................................................................................................. 17
6.2 Results & Analysis ................................................................................................... 17
6.3 Conclusions ............................................................................................................. 18
7 Scenario B1: macrocell Uplink Interference to the
Femtocell Receiver (traffic channel) ............................................................................... 19
7.1 Description of Interference Scenario(s) .................................................................... 19
7.2 Methodology and Assumptions ............................................................................... 19
7.3 Results and Analysis ................................................................................................ 20
7.3.1 Suburban Deployment Model, No Power Cap, Full Buffer Traffic ............................ 20
7.4 Conclusions ............................................................................................................. 20
8 Scenario C1: Femtocell Downlink Interference to the Macrocell UE Receiver.................. 21
8.1 Distance Based Power Control ................................................................................. 21
.
8.1.1 Introduction ............................................................................................................... 21
8.1.2 Simulation assumptions ............................................................................................ 21
8.1.3 Simulation results ...................................................................................................... 23
8.1.4 Fixed Femtocell Power, and Distance Based Power Control Conclusions ................ 24
8.2 Pathloss Based Power Control and Detection of Victim UEs ..................................... 25
8.2.1 Methodology and Assumptions ................................................................................ 25
8.2.2 Interference Mitigation Approaches ......................................................................... 26
8.2.3 Results and Analysis .................................................................................................. 26
8.2.4 Mitigation Techniques and Recommendations ........................................................ 29
8.2.5 Pathloss Based Power Control and Detection of Victim UEs Conclusions ................ 29
8.3 Pathloss Based Power Control Aimed at Improving Macro UE Performance ............. 30
8.3.1 Discussion .................................................................................................................. 30
8.3.2 Simulation setup ........................................................................................................ 30
8.3.3 HeNB Tx power control ............................................................................................. 31
8.3.4 Simulation results ...................................................................................................... 32
8.3.5 Pathloss Based Power Control Aimed at Improving
Macro UE Performance Conclusions ......................................................................... 35
8.4 Scenario C.1 Overall Conclusion ............................................................................... 35
9 Scenario C2: Femto BS Tx to Macro UE Rx (control channel) ........................................... 36
9.1 Description of Interference Scenario........................................................................ 36
9.2 Methodology and Assumptions ............................................................................... 37
9.3 Results and Analysis ................................................................................................ 39
9.3.1 Main Findings ............................................................................................................ 39
9.3.2 Detailed Results ......................................................................................................... 40
9.4 Mitigation Technique Recommendation .................................................................. 44
.
9.5 Conclusions ............................................................................................................. 44
10 Scenario D2: Femtocell Uplink Interference to the Macrocell NodeB Receiver ................ 45
10.1 Methodology and Assumptions ............................................................................. 45
10.2 Results and Analysis .............................................................................................. 47
10.2.1 Suburban Deployment Model, Fixed Power Cap, Full Buffer Traffic ........................ 47
10.2.2 Urban Deployment Model, Fixed Power Cap, Full Buffer Traffic .............................. 51
10.2.3 Urban Deployment Model, Full Buffer or Bursty Traffic,
Fixed or Adaptive Power Cap ................................................................................... 55
10.3 Mitigation Techniques and Recommendations ....................................................... 60
10.3.1 Description of Mitigation Techniques ....................................................................... 61
10.3.2 Recommendations .................................................................................................... 61
10.4 Conclusions ........................................................................................................... 62
11 Scenario E: Femtocell Downlink Interference to nearby Femtocell UE Receiver. ............. 63
11.1 Full Buffer Traffic and Distributed FFR .................................................................... 63
11.1.1 Distributed FFR .......................................................................................................... 63
11.1.2 Simulation Assumptions ............................................................................................ 64
11.1.3 Results and Discussion .............................................................................................. 65
11.1.4 Conclusions ................................................................................................................ 66
11.2 Mixed Traffic and Dynamic Interference Avoidance ............................................... 67
11.2.1 System Model ............................................................................................................ 67
11.2.2 Resource Allocation Algorithms ................................................................................ 68
11.2.3 Numerical Results: Mix of QoS and Full-Buffer Traffic .............................................. 70
11.2.4 Conclusions ................................................................................................................ 73
11.2.5 Additional Results ...................................................................................................... 74
12 Scenario F1: Femtocell Uplink Interference to Nearby Femtocell Receivers .................... 76
.
12.1 Two Apartment Model .......................................................................................... 76
12.1.1 Description ................................................................................................................ 76
12.1.2 Main Findings ............................................................................................................ 79
12.1.3 Conclusions ................................................................................................................ 81
12.2 System Simulation Approach ................................................................................. 81
12.2.1 Results ....................................................................................................................... 82
12.2.2 Conclusion ................................................................................................................. 84
12.3 Scenario F.1 Overall Conclusion ............................................................................. 85
13 Scenario G3: Macrocell Downlink Interference to an adjacent
channel Femtocell Receiver ........................................................................................... 86
13.1 Description of the interference scenario ................................................................ 86
13.2 Analysis ................................................................................................................. 86
13.2.1 Parameter settings .................................................................................................... 87
13.3 Main Findings ........................................................................................................ 87
13.4 Conclusions ........................................................................................................... 89
13.4.1 Mitigation techniques ............................................................................................... 89
14 Scenario G4: Macrocell Downlink Interference to the adjacent
channel Femtocell UE Receiver ...................................................................................... 90
14.2 Analysis ................................................................................................................. 91
14.3 Main Findings ........................................................................................................ 91
14.4 Conclusions ........................................................................................................... 93
15 Scenario H1: Macrocell Uplink Interference to the adjacent
channel Femtocell Receiver ......................................................................................... 94
.
15.2 Analysis ................................................................................................................. 95
15.3 Main Findings ........................................................................................................ 97
15.4 Conclusions ........................................................................................................... 98
16 Detailed Summary of Findings ....................................................................................... 99
17 Overall Conclusions and Future Work .......................................................................... 103
17.1 Overall Conclusions ............................................................................................. 103
17.2 Future Work ........................................................................................................ 104
18 References .................................................................................................................. 105
19 Contact Information .................................................................................................... 106
20 Appendix: OFDMA Interference Scenario Evaluation Methodology .............................. 107
20.1 Simulation Parameters ........................................................................................ 107
20.1.1 Macrocell Parameters ............................................................................................. 107
20.1.2 Femtocell Parameters ............................................................................................. 109
20.2 Femtocell Deployment modelling ........................................................................ 111
20.2.1 Suburban modelling ................................................................................................ 111
20.3 Dense Femto Cell Deployment Modelling ............................................................ 113
20.3.1 Dual Stripe Model .................................................................................................... 113
20.4 Channel Models ................................................................................................... 115
20.4.1 Antenna Patterns .................................................................................................... 115
20.4.2 Pathloss Models ...................................................................................................... 115
20.4.3 Shadowing Models .................................................................................................. 118
20.4.4 Fast Fading Models .................................................................................................. 118
.
20.5 Link-to-System Mapping ...................................................................................... 119
20.5.1 Link to System Mapping (RAN1 approach) ............................................................. 119
20.5.2 Link to System Mapping (per sub-carrier capacity approach) ................................ 120
20.5.3 Link to System Mapping (RAN4 approach) ............................................................. 121
20.6 Scheduler ............................................................................................................ 123
20.7 Power Control ..................................................................................................... 123
20.8 Downlink ............................................................................................................. 123
20.8.1 Femto Downlink Power Control .............................................................................. 123
20.8.2 Interference Modelling ........................................................................................... 125
20.8.3 Traffic Models .......................................................................................................... 126
20.9 Simulation Procedure and Flow ........................................................................... 127
20.9.1 Cell layout ................................................................................................................ 127
20.9.2 3GPP RAN4 based Monte-Carlo Static Simulation Methodology ........................... 129
20.10 Performance Metrics ......................................................................................... 132
20.11 References ......................................................................................................... 133

.
1 Executive Summary
Interference management is an important consideration for femtocell deployment. This white
paper presents work done to address interference management in the context of OFDMA
(particularly LTE-FDD) femtocell systems.
Link budget analyses and system level simulations were done to study various interference
scenarios.
Of the scenarios studied, certain scenarios were found to be benign and do not seem to require
new interference mitigation technique. The benign scenarios are:
Macro UL to femto UE on the same carrier (B1)
Macro DL to femto UE on an adjacent carrier (G3, G4)
Macro UL to femto AP on an adjacent carrier (H1)
Initial results based on averages suggest that one particular scenario, Macro DL to femto UE on the
same carrier (A1), is also benign, though more study is needed to confirm this initial finding.
The following scenarios may benefit from interference mitigation techniques:
Femto DL to macro UE on the same carrier (C1, C2)
Femto UL to macro BS on the same carrier (D2)
Femto DL to other nearby Femtos UE on the same carrier (E)
Femto UL to other nearby Femto AP on the same carrier (F1)
Interference mitigation techniques already applied in WCDMA femtos may be applied in addition
to techniques made possible by new tools provided by the LTE standards and architecture (e.g.,
the flexibility of frequency-domain resource management, the X2 interface). These form the bases
for new interference management techniques that may be used to build operational OFDMA
femtocell systems.
.
2 Femtocells, Femtocell Access Points and the Femto Forum
2.1 What are Femtocell Access Points?
Femtocell access points (FAPs) are low-power radio access points, providing wireless voice and
broadband services to customers primarily in the home environment. The FAP provides cellular
access in the home and connects this to the operators network through the customers own
broadband connection to the Internet.
FAPs have low power output, similar to other wireless home network equipment, and will allow a
small number (typically on the order of 10) of simultaneous calls and data sessions at any time. By
making the access points small and low-power, they can be deployed far more densely than
macrocells (for instance, one per household). The high density of deployment means that the
femtocell spectrum is re-used over and over again, far more often than the re-use that the macro
network (with its comparatively large cells) can achieve. Trying to reach the same levels of re-use
with macrocellular technology would be prohibitively expensive in equipment and site acquisition
costs. By using femtocells, the re-use, spectrum efficiency, and therefore total capacity of the
network can be greatly increased at a fraction of the macrocellular cost.
A typical deployment scenario is shown in Figure 2-1.

Figure 2-1: Typical femtocell deployment scenario

.
2.2 What Do Femtocells Offer?
Plugnplay installation by end user: Femtocells are installed by the end user without intervention
from the operator. The devices will automatically configure themselves to the network, typically
using Network Listen capabilities to select settings that minimise interference with the macro
network.
Moveability: The end user may move their femtocells, for example to another room, or, subject to
operator consent, to another location entirely.
Backhaul via the end users fixed broadband connection: Femtocells will use the subscribers
broadband connection for backhaul, which will be typically shared with other devices in the home.
Access Control the Closed User Group: The operator and/or end user will be able to control
which mobile devices can access the femtocell. For example, subscribers may be able to add guest
phone numbers via a web page.
Supports a restricted number of simultaneous users: Femtocells will support a limited number
(less than ten) of simultaneous calls and data sessions.
Femtozone (homezone) tariffs: Mobile services accessed through the femtocell may be offered at
a cheaper rate than the same services on the macro network. End users are advised when services
are accessed via the femtocell, either by an advisory tone, or a display icon or some other means,
so they know when the femto-tariffs apply.
Ownership: Various ownership models are possible, for example: end users may own their
femtocells, just as they own their mobile phones, or the operator may retain ownership with end
users renting the equipment (like a cable modem).
Small cell size/millions of cells in the network: The femtocell network can easily extend to
millions of devices.
Femto as a service platform: Novel mobile services can be made available on the femtocell. For
example, a femtocell-aware application on the mobile handset could automatically upload photos
to a website when the user enters the home, and download podcasts.

.
2.3 What is the Femto Forum?
The Femto Forum is the only organisation devoted to promoting femtocell technology worldwide.
It is a not-for-profit membership organisation, with membership open to providers of femtocell
technology and to operators with spectrum licences for providing mobile services. The Forum is
international, representing over 120 members from three continents and all parts of the femtocell
industry, including:
major operators,
major infrastructure vendors,
specialist femtocell vendors, and
vendors of components, subsystems, silicon and software necessary to create
femtocells.
The Femto Forum has three main aims:
to promote adoption of femtocells by making available information to the industry
and the general public,
to promote the rapid creation of appropriate open standards and interoperability for
femtocells, and
to encourage the development of an active ecosystem of femtocell providers to
deliver ongoing innovation of commercially and technically efficient solutions.
The Femto Forum is technology agnostic and independent. It is not a standards-setting body, but
works with standards organisations and regulators worldwide to provide an aggregated view of
the femtocell market. A full current list of Femto Forum members and further information is
available at www.femtoforum.org.

.
3 Introduction
3.1 Objectives and Methods of this Paper
The benefits of femtocells are not straightforward to realise. While network operators will see
significant capacity gains and end-users can expect higher performance, to achieve this, the radio
layer must be carefully managed. The management of the radio interference between the macro
and femto layers is a key industry concern addressed by this paper.
Interference adversely affects the capacity of a radio system and the quality of the individual
communication links on that system. Adding capacity is always based on a trade-off between
interference, quality and capacity hence there is a need for interference management techniques
to minimise interference that might otherwise counteract the capacity gains and degrade the
quality of the network.
1. The principal objectives of this study are:
a) to develop an industry position on the interference risks from OFDMA femtocell
deployments;
b) to recommend mitigation techniques and any necessary associated RF parameters
and performance requirements to ensure minimal disruption to the macro
network or other femtocells.
2. To achieve these objectives, this paper develops detailed interference scenarios for
evaluation and inclusion in the interference management assessment. The scenarios will
cover worst-case deployment conditions and assess the respective system impact.
3. The focus is to develop the assessment for OFDMA technologies (primarily LTE), and in
doing so devise a process that should be consistent with alternative radio technologies.
4. Two main steps were identified in order to accomplish the above goal:
a. Firstly, a baseline set of interference analysis conclusions for LTE-FDD femtocells
was required. Interference mitigation techniques should also be considered on the
understanding that vendor independence be preserved wherever possible.
b. Secondly, a recommendation for a common set of behaviours (RF parameters
and/or test cases) that can be derived by any LTE femtocell was required. This is so
that the femtocell can configure itself for minimal disruption to either the
macrocell layer or other deployed femtocells.
5. We focus exclusively on the Closed User Group model. This is the most likely residential
deployment model, and restricts the pool of allowed users to small group authorised by
the operator or the owner of the femtocell. Non-authorised subscribers may suffer
coverage and service impairment in the vicinity of a closed access femtocell which is
important to assess.
.
6. The study will also investigate methods of controlling the impact of deploying large number
of femto on the macro network. For example, different frequency resource utilization and
adaptive power control may be used to manage the interference in the network.
7. This paper has limited itself in scope according to perceived priorities, as follows.
a. It is exclusively concerned with LTE as an air interface technology.
b. It is concerned exclusively with operation at 2GHz this being seen as the most
important band for early, in-building deployment.
c. It is exclusively a theoretical treatment, using link level and system level simulations
to draw its conclusions.
d. In view of the residential application that femtocells are addressing, this paper is
also exclusively concerned with femtocells operating with closed user groups.

Femtocell interference management is an area of active research and study within the Femto
Forum as well as other Standards and industry organizations. While some interference mitigation
techniques are suggested in this document, they are not exhaustive. Other techniques and
refinements may continue to be developed and adopted elsewhere.

.
4 Simulation Scenarios Description
The Femto Forum has identified various scenarios that explore the limits of operation of
femtocells and femtocell subscriber equipment.
The scenarios are summarised in the following tables and figure. Detailed description of the
scenarios and methodology used to study them are provided in the Appendix.
Table 4-1: Summary of Femtocell Interference Scenarios
Scenario Description
Macrocell Downlink Interference to
the Femtocell UE Receiver (A1)
System simulation of a dense urban apartment block
scenario which models femtocells within the macro coverage
area (see Evaluation Methodology in section 20).
Interference between the downlink traffic channels is
considered. Co-channel operation is assumed.
Macrocell Uplink Interference to
the Femtocell Receiver (B1)
System simulation of suburban housing scenario, which
model femtocells within the macro coverage area (see
Evaluation Methodology in section 20). Interference
between the uplink traffic channels is considered. Co-
channel operation is assumed.
Femtocell Traffic Channel Downlink
Interference to the Macrocell UE
Receiver (C1)
System simulation of both dense urban apartment block and
suburban housing scenarios, which model femtocells within
the macro coverage area. Interference between the
downlink traffic channels is considered. Co-channel
operation is assumed.
Femtocell Control Channel
Downlink Interference to the
Macrocell UE Receiver (C2)
A link budget analysis is performed and two types of MUE
are considered. Interference between the downlink control
channels is considered. Co-channel operation is assumed.
Femtocell Uplink Interference to
the Macrocell NodeB Receiver (D2)
System simulation of both dense urban apartment block and
suburban housing scenarios, which model femtocells within
the macro coverage area. Interference between the uplink
traffic channels is considered. Co-channel operation is
assumed.
Femtocell Downlink Interference to
Nearby Femtocell UE Receivers (E)
System simulation of a dense urban apartment block
scenario. Interference between the downlink traffic channels
is considered.
.
Scenario Description
Femtocell Uplink Interference to
Nearby Femtocell Receivers (F1)
Two nearby houses are considered. A femtocell is located
within each house (AP1 and AP2). One of the apartment
owners (AP2) is outside his house. Analysis is performed as a
function of distances of victim and aggressor UEs to the AP1.
Interference between the uplink traffic channels is
considered. Co-channel operation is assumed.
an adjacent channel Femtocell
Receiver (G3)
A link budget/statistical analysis is performed, where the
suburban homes are located uniformly in a sector, and FUEs
are located uniformly within a house. Interference between
the downlink traffic channels is considered. Adjacent channel
operation is assumed.
the adjacent channel UE Femtocell
Receiver (control channel) (G4)
suburban homes are located uniformly in a sector, and FUEs
are located uniformly within a house. Interference between
the downlink control channels is considered. Adjacent
channel operation is assumed.
Macrocell Uplink Interference to
the adjacent channel Femtocell
Receiver (H1)
Macrocell connected UE (MUE) is located within the same
home as the Femtocell, and has a call established at full
power. The Femtocell UE (FUE) has already established a
call. Interference between the uplink traffic channels is
considered. Adjacent channel operation is assumed.

The relationship between these scenarios is summarised in the following table and figure:
Table 4-2: Interference Scenario Relationships

Victim
Femto UE
DL Rx
Femto AP
UL Rx
Macro UE
DL Rx
Macro
NodeB
UL Rx
Neighbour
Femto UE
DL Rx
A
g
g
r
e
s
s
o
r

Macro NodeB
DL Tx
A, G
Macro UE
UL Tx
B, H
Femto AP
DL Tx
C E
.
Femto UE
UL Tx
D
Neighbour Femto
UE
UL Tx
F

AF are the interference scenarios for co-channel deployments
GH are the interference scenarios for adjacent channel deployments

The following diagram illustrates and summarises the femtoforum scenarios A-H:

FUE
F
FUE
MUE
A,G
D
B,H
C
E
F
F
FUE
MUE
Femto AP
Femto UE
Macro UE
Apartments
Macro
NodeB
Interference
path
UE Association
F
.
5 Abbreviations and Defined Terms
Throughout this paper a number of abbreviations are used to identify various system elements
and parameters. The most frequently used are presented here for quick reference however a
more extensive list has been produced and is available under separate cover [FFG08]

AP Access Point
BS Base Station (assumed to be a wide-area BS, as defined in [TS25.104])
FAP Femto AP
FUE Femto UE
HeNB Home eNode B
MUE Macro UE
QoS Quality of Service
UE User Equipment
RAN Radio Access Network
LOS Line-Of-Sight
P-CPICH Primary Common Pilot Channel
Victim Device negatively impacted by interference
Aggressor Device negatively impacting other through interference
Deadzone Area where the quality of service is so poor as a result of interference that it is not
possible to provide the demanded service. Deadzones are also characterised by the
fact that in the absence of any interference, a normal service would be possible.

.
6 Scenario A1: Macrocell Downlink Interference to the Femtocell UE Receiver
(Traffic Channel)
This section discusses the impact to users and possible mitigation solutions with respect to
scenario A.1: Macrocell Downlink Interference to the Femtocell UE Receiver (Traffic Channel).
6.1 Description
The monte carlo simulation method is used with the assumptions align with [OFDMA-EMD]. Duel
strip model is used for the femto cells. The parameters are listed in Table 6-1: Urban-dense
Femtocell Modelling ParametersTable 6-1.
Table 6-1: Urban-dense Femtocell Modelling Parameters
N (number of cells per row ) 10
M (number of blocks per sector) 1
L (number of floors per block) 6
R (deployment ratio ) 0.2
P (activation ratio) 50%
6.2 Results & Analysis
As the femto cells have deployed the interference from the macro cell also needs to be considered
and throughput degradation in DL is presented in Table 6-2. Compared with the inter femto
interference, the interference from macro cell is not severe and the corresponding degradation is
negligible. We found that the maximum loss is about 1% even when the femto BS Tx power is
0dBm.
Table 6-2: Relative throughput degradation in DL with different sector radius
activation ratio 0.5 1
Femto cell BS power (dBm) 0 5 10 15 20 0 5 10 15 20
Relative
degradation %
R=250 m 1.07 0.49 0.2 0.08 0.02 0.86 0.416 0.205 0.11 0.074
R=577 m 0.31 0.19 0.14 0.11 0.10 0.17 0.082 0.039 0.019 0.011
.
6.3 Conclusions
For the deployment model and parameters considered in this study, the impact of the interference
from macro cell on the average femto cell throughput is not severe in this scenario, and the
corresponding performance degradation is around 1% or less for a wide range of femto cell base
station power, and for various macro cell radius and femto cell activation ratio values. Note that
the impact for other deployment parameters and on other performance metrics such as median or
tail of user throughput CDF may be different and quantization of such impacts may require further
investigation.
.
7 Scenario B1: macrocell Uplink Interference to the Femtocell Receiver (traffic
channel)
This section discusses the impact of interference arising from scenario B.1: Macrocell Uplink
Interference to the Femtocell Receiver
7.1 Description of Interference Scenario(s)
The interference scenario is as follows:
B.1 - Macro UE Tx to Femto BS Rx (traffic channel)
The femtocell is assumed to have closed access, and shared spectrum is assumed between the
macro and femto layers.
7.2 Methodology and Assumptions
The simulation methodology defined in [OFDMA-EMD] is assumed with the following specific
parameters:
Table 7-1: Simulation Parameters
Parameter Assumption
Deployment Suburban
Macro layer has 7 sites (21 sectors) with wrap-
around, 500m ISD.
10% of femto UEs are outdoors and 20% of
macro UEs are outdoors.
Number of macro UEs per sector 10
Exterior Wall Loss 20dB
Shadowing correlation (one BS to multiple
UEs)
Correlated Shadowing
Macrocell Power uplink control Proprietary method based on limiting noise
rise to macro neighbours
Femtocell uplink power control Based on femtocell coverage (no protection of
macro neighbours)
.
Link to System Mapping Per sub-carrier capacity approach
Scheduler Both baseline round robin
Traffic model Full Buffer

7.3 Results and Analysis
7.3.1 Suburban Deployment Model, No Power Cap, Full Buffer Traffic
This sub-section looks at the impact of interference from macro UEs to the femtocell uplink. The
following simulation results were obtained in the case of 5 femtos per macro sector for the
baseline configuration, in the special case where no power cap is placed on the femtocell UEs:
Table 7-2: Impact of Interference from Macro UEs to Femto Uplink
Number macro UEs
per macro sector
Femto average
throughput (Mb/s)
Femto 5 percentile
throughput (Mb/s)
Femto mean IoT (dB)
0 18.1 18.1 -22.0
10 18.1 18.1 -4.2

Although the presence of the macro UEs causes a rise in the interference over thermal as seen at
the Femto, the interference level remains low relative to the thermal noise and the throughput is
not impacted.
7.4 Conclusions
Although the presence of the macro UEs was shown to cause a rise in the interference over
thermal as seen at the femto, the average interference level remains low relative to the thermal
noise and the average throughput is not impacted.
.
8 Scenario C1: Femtocell Downlink Interference to the Macrocell UE Receiver
This section describes the modelling of scenario C.1 presented in reference [OFDMA-Scen]. Impact
to users and possible mitigation solution are covered. As well as a baseline scheme with fixed
femtocell power (no interference mitigation), three approaches for interference mitigation have
been considered which are a) distance based power control; b) power control based on pathloss
and detection of the presence of victim UEs (aimed at improving HUE performance while
maintaining similar protection for MUEs); and c) power control based on pathloss (aiming at
improving MUE protection while maintaining similar HUE performance). These three schemes are
covered in the following sub-sections.
8.1 Distance Based Power Control
8.1.1 Introduction
In this section, we present results illustrating the impact of HeNBs on macro cell downlink
throughput. The results reveal the extent of throughput degradation suffered by macro UEs under
different HeNB densities and transmission power levels. We also study the throughput
degradation when a distance based power control mechanism is applied.
8.1.2 Simulation assumptions
Table 8-1 shows the simulation assumptions used for generating results that follow. Other
simulation assumptions are mostly inline with the assumptions in [R4-091731].
Table 8-1: Simulation Assumptions
Parameter Value
Inter-site distance 500 m
Carrier frequency 2 GHz
Carrier bandwidth 5 MHz
Total Macro BS transmission
power
20 W (43dBm)
External wall penetration loss 10 dB
Number of BS antennas 2 Rx, 1 Tx
.
Minimum separation HeNB to
Macro BS
46 m
HeNB total transmission power 0dBm, 10dBm, 20dBm, distance based
power control
Traffic model Full buffer with 200 UEs per sector
UE distribution UEs are dropped with uniform density.
All UEs are assumed to be outdoors.
HeNB operation mode Closed access
Scheduler Proportional fair (alpha=1 and beta=0.375)

It is known that the coverage of a HeNB is strongly correlated with its distance to the macro cell if
constant power is used by all HeNBs. In order to ensure consistent HeNB coverage, we consider a
distance based power control algorithm. The following formula is used to compute HeNB
transmission power:
<
<
<
=
m d m dBm
m d m dBm
m d m dBm
d P
HeNB
HeNB
HeNB
HeNB HeNB
289 156 0
156 85 10
85 46 20
) ( ,
where
HeNB
d is the distance between the HeNB and the closest macro cell. The macro cell is divided
into three strips such that all HeNBs located in a strip will use the same power level as given in the
formula. The strip boundaries are selected so that the variation of the received macro cell signal
strength is approximately equal to 10dB for outdoor UEs in each strip. Since the locations of
HeNBs and macro cells are known to operators, the distance between a HeNB and its closest
macro cell can be calculated after the HeNB has been deployed, and transmission power of this
HeNB will be configured accordingly. Different from measurement based power control algorithm
where a HeNB chooses its transmission power based on measured signal strength of the nearby
macro cells, our proposal does not rely on measurements made by a HeNB DL receiver, which
could incur additional cost and security/reliability issues.

.
8.1.3 Simulation results
When no HeNB exists in the network, the simulated macro cell downlink throughput is equal to
6.4Mbps, which is used as a benchmark reference for macro cell throughputs with HeNB
deployment. Table 2 shows the normalized DL cell throughputs (actual throughputs divided by
6.4Mbps) for different HeNB densities and transmission power levels.
Table 8-2: Normalized aggregate macrocell downlink throughput with HeNB deployment
Normalized macro
cell DL throughput
HeNB transmission power (dBm)
20 10 0 Distance based
PC
Number
of
HeNBs
per
macro
cell
50 0.60 0.66 0.84 0.70
40 0.61 0.70 0.86 0.75
30 0.69 0.92 0.98 0.95
20 0.76 0.95 1.00 0.98
10 0.80 0.97 1.00 0.97

We have following observations:
As highlighted in the shaded region in Table 2, the normalized cell throughput is very
close to 1 with a small number of HeNBs (<=30) and low HeNB transmission power
(<=10dBm). The cell throughput variation within this region is quite small with varying
number of HeNBs and their power levels. Beyond this region, however, cell
throughput drops very fast when adding more HeNBs and/or increasing HeNB power
levels. In order to limit HeNB impact on aggregate macro cell DL throughput, operators
may want to set joint upper bounds for CSG HeNB density and transmission power and
perform admission control when needed. For example, if the number of deployed
HeNBs under a certain macro cell is larger than the upper bound, some of them may
need to reduce its power or change to open/hybrid access mode.
Distance based power control algorithm provides close to benchmark results when the
number of HeNBs is 10, 20 or 30. With larger number (40 or 50) of HeNBs, the
algorithm consistently leads to larger throughput compared to the throughput when
constant HeNB power 10dBm/20dBm is applied.

.
The results in Table 2 show that the aggregate cell throughput will not have a large reduction
whenever HeNB density and transmission power level are within a certain range. However, for a
given UE, its throughput can decrease significantly if it is close to an interfering HeNB. For
example, a HeNB deployed in an apartment could generate large interference to UEs in
neighbouring apartments. Such a HeNB may even create big coverage holes in which UEs in
neighbouring apartment cannot connect to the macro cell. Although only a small number of UEs
are likely to be impacted, we cannot ignore this problem.
One solution to this problem is to limit the resource blocks being used by HeNBs, so that a UE
nearby can be scheduled by the macro cell using the remaining resource blocks. The part of
resource blocks not used by HeNBs needs to be preconfigured so that the macro eNB will know
how to perform resource allocation once a UE reports a nearby interfering HeNB. This issue has
also been discussed in [R4-091908].
8.1.4 Fixed Femtocell Power, and Distance Based Power Control Conclusions
We have studied the impact of HeNB interference on macro cell throughput performance. Based
on the results, we propose the following:
Distance based power control algorithm should be considered as an alternative to the
measurement based power control algorithm.
To limit HeNB impact on aggregate macro cell DL throughput, it is desirable to set joint
upper bounds for HeNB density and transmission power. If the number of deployed
HeNBs under a certain macro cell is larger than the upper bound, some of them may
need to reduce its power or change to open/hybrid access mode.
To limit HeNB impact on a nearby UE DL throughput, HeNBs shall be blocked from
using certain resource blocks. The index of such resource blocks needs to be
preconfigured so that the macro eNB will know how to perform resource allocation
once a UE reporting a nearby interfering HeNB.

.
8.2 Pathloss Based Power Control and Detection of Victim UEs
8.2.1 Methodology and Assumptions
parameters:
Deployment Suburban, 7 sites (21 sectors) with wrap-around,
500m ISD.
10% of femto UEs are outdoors and 0% of macro
UEs are outdoors, macro UEs can be dropped
within femto house.
(Note that having all macro UEs indoor is worst
case for SINR since for those UEs not dropped in
a femto house, both the wanted signal and the
interference are attenuated by same exterior
wall loss whereas thermal noise is unchanged).
UEs)
Femtocell downlink power control Max power based on limiting interference to
macro UEs (a similar approach to that described
in [3GPP 25.967] section 7.2.1 for WCDMA).
Scheduler Proprietary Frequency Selective/Proportional
Fair
Traffic model Full Buffer

.
8.2.2 Interference Mitigation Approaches
A baseline interference mitigation approach is implemented where a power cap is placed on
the femto HeNBs in order to protect the macro downlink for macro UEs in the vicinity of the femto
(using a similar approach to that described in [3GPP 25.967] section 7.2.1 for WCDMA). The
power cap is typically static or semi-static and is a function of the pathloss to the neighbouring
macro eNBs.
A refined scheme is also investigated. For this refined scheme, the femto HeNB detects if macro
UEs (or any non-served UEs) are in the vicinity of the femto by detecting their uplink
transmissions.
If they are detected, full protection is provided (per the baseline scheme)
If they are not detected, limited protection is provided e.g. provide full protection only
when certain control channels are transmitted from the macro eNB (to protect idle
UEs), otherwise use a more relaxed power cap.
One possible way to detect if there are nearby non-served UEs transmitting is to look for the
properties of the Zadoff-Chu reference signals in the noise+interference signal received at the
femto. Note that macro UEs will require most protection in the downlink when they are at the
edge of macrocell coverage. In this case the macro UEs will be transmitting at high power, and the
SNR of the reference signals as seen by a nearby femto will be high i.e. the reference signals will
easiest to detect.
Note that in the simulation results that follow, protection of idle mode UEs is not considered for
the enhanced approach.
8.2.3 Results and Analysis
Figure 8-1 below shows the average macrocell sector throughput as a function of the density of
femtocells in the sector. Results are shown for the baseline approach (labelled Fixed Protection)
and enhanced approach (labelled Proposed). It can be seen that both interference mitigation
approaches protect the macrocell UEs equally.

.
0 10 20 30 40 50 60 70 80
1.18
1.2
1.22
1.24
1.26
1.28
1.3
x 10
4
Macro Layer Downlink, Av. Sector Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)
Femtos per Macro Sector

Fixed Protection
Proposed

Figure 8-1: Macrocell downlink average sector throughput
Figure 8-2 below shows the cell edge (5 percentile) macro user throughput as a function of the
density of femtocells in the sector. It can be seen that the two approaches protect the macrocell
UEs almost equally.
0 10 20 30 40 50 60 70 80
0
50
100
150
200
250
300
350
400
Macro Layer Downlink, 5 percentile User Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

Fixed Protection
Proposed

Figure 8-2: Macrocell downlink 5 percentile user throughput

.
Figure 8-3 below shows the average femtocell throughput as a function of the density of
femtocells in the sector. Results are shown for the baseline approach (labelled Fixed Protection)
and enhanced approach (labelled Proposed). It can be seen that the femtocell performance is
improved with the enhanced interference mitigation approach.
0 10 20 30 40 50 60 70 80
0
0.5
1
1.5
2
2.5
3
x 10
4
Femto Layer Downlink, Av. Cell Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

Fixed Protection
Proposed

Figure 8-3: Femtocell downlink average sector throughput
Figure 8-4 below shows the cell edge (5 percentile) femtocell throughput as a function of the
density of femtocells in the sector. It can be seen that the femtocell performance is significantly
improved with the enhanced interference mitigation approach.

.
0 10 20 30 40 50 60 70 80
0
50
100
150
200
250
300
350
400
Femto Layer Downlink, 5 percentile User Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

Fixed Protection
Proposed

Figure 8-4: Femtocell downlink 5 percentile user throughput
8.2.4 Mitigation Techniques and Recommendations
As suggested above the required cap on the femto HeNB power could be made a function of
whether UL transmissions are detected at the femto from nearby non-served UEs. The properties
of the Zadoff-Chu reference sequences could be used to assist this detection. Furthermore non-
served UEs in idle mode (which are not transmitting on the UL but which might be listening to
control channels on the macrocell downlink) could be protected by providing full protection
when certain control channels are being transmitted by the macro eNB.
It is recommended that such techniques be considered further.
8.2.5 Pathloss Based Power Control and Detection of Victim UEs Conclusions
The scenario C.1 (Femto BS Tx to Macro UE Rx (traffic channel and co-channel operation)) was
considered. Simulations with a suburban deployment model suggest that an approach of setting a
cap on the femto HeNB power as a function of whether there are nearby UEs whose downlinks
need protecting is an effective way of improving the throughput of the femto HeNBs, without
sacrificing protection of the macrocell downlink. The properties of the Zadoff-Chu reference
sequences sent in the uplink could be used to assist the detection of nearby but non-served UEs.
.
8.3 Pathloss Based Power Control Aimed at Improving Macro UE Performance
8.3.1 Discussion
The downlink throughput has been studied for densely populated areas with multi-floor
apartment buildings [R4-071617]. Both fixed and controlled transmit power for HeNB are
considered. For the fixed power case, the HeNB transmit power is set to 8 dBm, which is the
minimum Tx power upper limit in all cases for adjacent operator channel protection according to
TS 25.104. For the controlled power case, the power control algorithm, which tries to determine
the proper transmit power in order to reduce MUE outage on the same carrier frequency, is
applied. This algorithm has been shown to reduce the MUE coverage outage [R4-091895].
8.3.2 Simulation setup
In this section we consider a dense-urban model that was used for HNB studies in [R4-081344].
The dense-urban model corresponds to densely-populated areas where there are multi-floor
apartment buildings with smaller size apartment units. In the dense-urban model, blocks of
apartments are dropped into the three center cells of a macro cell layout with ISD of 1 km. Each
block is 50m x 50m and consists of two buildings (north and south) and a horizontal street
between them as shown in Figure 8-5. The width of the street is 10 meters. Each building has K
floors. K is chosen randomly between 2 and 6. In each floor, there are 10 apartment units in two
rows of five. Each apartment is 10m x 10m (i.e., approximately 1076 square feet) and has a one-
meter-wide balcony. The minimum separation between two adjacent blocks is 10m. The
probability that a HUE is in the balcony is assumed to be 10%. We drop 2000 apartment units in
each cell which corresponds to a 6928 households per square kilometer. This represents a dense-
urban area. Taking into account various factors such as wireless penetration (80%), operator
penetration (30%) and HeNB penetration (20%), we assume a 4.8% HeNB penetration which
means 96 of the 2000 apartments in each cell have a HNB installed from the same operator. We
assume 12 simultaneously active HUEs per cell to calculate the throughput.
MUEs are also dropped randomly into the three center cells of the 57-cell macro layout such that
30% of the MUEs are indoor. In addition, we enforce a minimum path loss of 38dB between UEs
and HeNBs (i.e., one meter separation). In the dense-urban model, we use the 3GPP micro-urban
model for the outdoor path loss computation. The free-space component for the micro-urban
model is given by
d dB PL
micro fs 10 ,
log 40 28 ) ( + =

The other propagation models are similar to the ones in [R4-071617].
.
The set of simulation parameters are shown below:
System bandwidth 5 MHz
Macro Power = 43dBm
HeNB power between [-10 dBm, 10 dBm].
o Case 1, the HeNB power is fixed to 8 dBm
o Case 2, the adaptive HeNB power setting is used to reduce the MUE outage
ISD of 1km
Thermal noise power = -99dBm
10 drops of 30 simultaneously active MUEs in each drop.
10m 10m
11m 11m
10m 10m
10m 10m
11m 11m
10m 10m

Figure 8-5 Top view of the apartment block in dense-urban model
8.3.3 HeNB Tx power control
The specific HeNB power control algorithm studied in the section is as following:
HeNB measures RSRP and SINR from the macro base station and neighboring HeNB.
Assuming the HeNB has a small coverage area, a close by MUE or non-serving HUE is
assumed to have similar RSRP and SINR as the measurements.
HeNB power is determined by three conditions:
1. To maintain an SINR of -6 dB for a MUE located 80dB away from HeNB, taking into
account the additional interference from the HeNB.
2. To maintain an SINR of -5 dB for a non-serving HUE located 80dB away from HeNB,
taking into account the additional interference from the HeNB.
3. If the conditions above could not be met, transmit at the minimum Tx power (-10
dBm).
This is an iterative process where HeNBs updates the transmit power periodically based
on previous measurements.
.
The algorithm shown above has been shown to reduce MUE outage by reducing potential
interfering HeNB transmit power. One obvious trade-off is the shrinking of HeNB coverage, i.e.,
some HUE will be off-loaded to the eNB due to smaller HeNB transmit power.
The HeNB transmit power statistics are shown in Figure 8-6, where 82% of HeNBs are shown to be
transmitting at the minimum Tx power of -10 dBm.
-10 -8 -6 -4 -2 0 2 4 6 8 10
0.8
0.82
0.84
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
dBm
C
D
F

Figure 8-6: HeNB transmit power statistics
8.3.4 Simulation results
We assume 41% downlink overhead and 2x2 MIMO with wideband precoding to obtain the
results. First, it is assumed that a fixed Tx power of 8 dBm is used for all HeNBs. The C/I
distributions with and without HeNBs are shown in Figure 8-7. With HeNBs, 25% of the UEs reach
the 30dB RF noise floor, whereas without HeNBs less than 12% UEs can reach 10dB. The average
throughput results are given in Figure 8-8. Almost 33% of UEs average 24.4Mbps with HeNBs.
However, the percentage of UEs in outage is increased to 19% with HeNBs from 13% without
HeNBs.
.
-60 -50 -40 -30 -20 -10 0 10 20 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C/I
C
D
F

no HeNBs
with HeNBs

Figure 8-7: C/I for mixed macro and femto deployment with 8 dBm HeNB
10
5
10
6
10
7
10
8
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
UE Avg Throughput [bps]
C
D
F

no HeNBs
with HeNBs

Figure 8-8: UE average throughput for mixed macro and femto deployment with 8 dBm HeNB
Figure 4 presents the results for the power-controlled case. The C/I distribution is shown in Figure
8-9. The UE average throughput is shown in Figure 8-10. Controlling the power transmitted by the
HeNBs, the percentage of UEs in outage is decreased to 10%. On the other hand, the percentage
of UEs achieving the maximum average throughput of 24.4Mbps is 17%.
.
-40 -30 -20 -10 0 10 20 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
C/I
C
D
F

no HeNBs
with HeNBs

Figure 8-9: C/I for adaptive HeNB power setting
10
5
10
6
10
7
10
8
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
UE Avg Throughput [bps]
C
D
F

no HeNBs
with HeNBs

Figure 8-10: Average throughput for adaptive HeNB power setting

Table 8-4: Summary of results
Outage Probability
Average
throughput (kbps)
Median throughput
(kbps)
No HeNB 12.7% 202 151
HeNB with fixed Tx
power of 8 dBm
18.9% 10988 5623
HeNB with adaptive
Tx power
9.8 % 8987 3311

.
8.3.5 Pathloss Based Power Control Aimed at Improving Macro UE Performance
Conclusions
In this section, we studied the downlink throughput with HeNBs deployed at the same frequency
as the macro eNBs. We observed that the average UE throughput increase with HeNBs with
respect to no HeNBs case by more than 40 times. Two cases are simulated, where the first case
has fixed 8 dBm transmit power for HeNBs and the second case adaptive transmit power for
HeNBs. With controlled transmit power, the percentage of UEs in outage decreases below the
case with no HeNB is deployed. Therefore, with controlled transmit power for HeNBs, not only
better C/I and higher UE average throughput are achieved, but also fewer UEs are left in outage.
8.4 Scenario C.1 Overall Conclusion
This section considers operation of the femtocell and macrocell downlinks in the case of shared
spectrum and closed access femtocells. It is shown that suitable power control of the femtocells is
important in managing interference to the macrocells and that this becomes more important as
the density of femtocells increases. In general the femtocell power should be decreased as the
femto becomes closer to the edge of the macrocell, and power control schemes based on both
pathloss measurements and distance were shown to be effective in controlling interference. While
this power control may in some cases restrict the performance on individual femtocells due to
restrictions in power, it is shown that the overall system performance (throughout and outage) for
a given population of UEs is significantly higher with both femtocells and macrocells than
with macrocells alone. Enhancements were also considered whereby a femtocell operates with
restricted power, and/or on a restricted set of frequency resources, only if victim UEs are detected
to be in the vicinity of the femtocell.
.
9 Scenario C2: Femto BS Tx to Macro UE Rx (control channel)
This section presents analysis results of the interference evaluation scenario C.2. The title of this
scenario is Femto BS Tx to Macro UE Rx (control channel) for LTE-TDD system.
The purpose of the analysis of this scenario is to evaluate the impact of control channel
interference from a femtocell experienced by a MUE in LTE-TDD system. The MUE is connected to
a macrocell at the edge of coverage, while the MUE and femtocell are located in close proximity.
The control channel signal from macro cell can be degraded by femto cell interference when
femto cells are fully loaded in the downlink and sharing the same frequency bandwidth with
macro cells. Since the MUE is not allowed to handover to CSG femtocell, a weak signal is received
from the macrocell. The impact of interference is evaluated using two control channels, PDCCH
(Physical Downlink Control Channel) and SCH (Synchronization Channel), where SCH is composed
of primary/secondary synchronization signals.
9.1 Description of Interference Scenario
In this scenario two MUEs (one is visiting, the other is passing) are connected to the macrocell at
the edge of coverage. A visiting MUE is located in the same room as the femtocell. A passing MUE
is outside the apartment and very close to the outdoor wall. Two MUEs are not allowed to access
the femtocell (i.e., closed subscriber group). A FUE is connected to a femtocell at the edge of
coverage and the femtocell is fully loaded in the downlink, i.e., it transmits full power over the
whole band. The victim receivers in this case are the MUEs and the aggressor is the femtocell
downlink transmitter. Figure 9-1 illustrates the interference scenario C2.
.
Macrocell
Apartment Block
Weak Signal (S
Macro
)
Interference (I
Femto
)
FUE
Femtocell
Visiting
MUE
Passing
MUE

Figure 9-1: Illustration of Interference Scenario C2
According to 3GPP specification [TS 36.211] the physical control format indicator channel (PCFICH)
carries information about the number of OFDM symbols used for transmission of PDCCHs in a
subframe. If there is not PCFICH in a subframe it means that radio resources are not assigned to
PDCCH. To consider interfering situation of PDCCH between two cells(macro and femto cell) it is
assumed that control messages to UEs are always assigned in the same radio resources, i.e., both
macrocell and femto cell utilize the same resources to send control messages which carry uplink
grant, downlink resource assignment, TPC command, etc. However, the primary synchronization
signal and the secondary synchronization signal are mapped to predefined position. The primary
synchronization signal is mapped to the third OFDM symbol in subframes 1 and 6. The secondary
synchronization signal is mapped to the last OFDM symbol in subframe 0 and 5. In frequency
domain the signals are allocated in 62 subcarriers around dc-carrier. Accordingly synchronization
signals are simultaneously received both from macrocell and femto cell.
In this section analytical evaluation is carried out for the interference scenario C.2 based on link-
budget calculations and transceiver performance requirements as specified by 3GPP. The downlink
frequency is assumed to be 2GHz and the antenna gains of the femtocell and UEs are equal to
unity. The detailed parameters are given below:

.
Control channels
o PDCCH (Physical Downlink Control Channel)
o SCH (Synchronization Channel)
Femtocell parameters
o Static maximum total transmission power (P
max
): -10dBm to 20dBm
o Channel bandwidth: 10MHz
Used subcarrier: 600
Subcarrier spacing (f): 15kHz
o Downlink frequency: 2GHz
o Antenna gain including a cable loss: 0dBi
Minimum separation UE to femtocell: 0.2m
Macrocell parameters
o Total transmission power: 46dBm
o Channel bandwidth: 10MHz
Used subcarrier: 600
Subcarrier spacing (f): 15kHz
o Downlink frequency: 2GHz
MUE receiver parameters
o Required SNR for PDCCH: -1.6 dB from [TS 36.101] (miss-detection probability = 1%
or lower at single antenna port)
o SCH s/Iot: -6.0 dB from [TS 36.133]
o Noise figure: 9dB
o Thermal noise per subcarrier: P
N, sc
=-123.2dBm
o Received signal power from a macrocell per subscriber: S
Macro, sc

o Received interference from a femtocell per subscriber: I
Femto, sc

.
Path loss model from [OFDMA-EMD]
UE to Macro BS
1) UE is outside:
PL = 128.1 + 37.6log10(R) dB, R in km
= 15.3+37.6log10(R), R in m
2) UE is inside:
PL = 15.3+37.6log10(R) + L
ow
, R in m
L
ow
is the penetration loss of an outdoor wall, which is 20dB
UE to Femto BS
3) UE is inside the same room as femtocell
PL = 38.46 + 20 log10R + 0.7R , R in m
4) UE is outside the apartment
PL = MAX(15.3 + 37.6log10R, 38.46 + 20log10R) +
0.7d
2D,indoor
+ L
ow
, R and d
2D,indoor
in m
d
2D,indoor
is the distance inside the house
L
ow
is the penetration loss of an outdoor wall, which is 20dB
In this analysis, d
2D,indoor
is assumed as R (i.e. the passing MUE is very close to the outdoor wall)
9.3.1 Main Findings
The deadzones of control channels of a MUE are varied according to a femto transmission power
and a radius of macrocell. The deadzones of PDCCH and SCH are increased as a femtocell
transmission power is increased and/or a radius of macrocell is grown up in this interference
scenario. The deadzone of a visiting MUE is large but the deadzone of a passing MUE is very small
due to a heavy loss of a wall. The detailed results are given below.

.
9.3.2 Detailed Results
In order to demodulate a PDCCH at MUE, the value of S
Macro,sc
/(I
Femto,sc
+P
N,sc
) must be greater than
a required SNR of PDCCH. The region of operation of PDCCH shown in Figure 9-2 illustrates the
maximum strength of the downlink interference plus noise per subcarrier versus the minimum
strength of wanted signal per subcarrier. The dynamic range of UE receiver is -25dBm to -97dBm
from. Therefore, the received power per subcarrier (S
Macro,sc
) should be in the range of -52.7dBm
and -124.7dBm. Each point in the region of operation translates into distance of separation
between femtocell to MUE versus distance between macrocell and MUE.
-130 -120 -110 -100 -90 -80 -70 -60 -50
-130
-120
-110
-100
-90
-80
-70
-60
-50
Min. Macro cell rx power(S
Macro,sc
) [dBm]
M
a
x
.

F
e
m
t
o

c
e
l
l

i
n
t
e
r
f
e
r
n
c
e

p
l
u
s

N
o
i
s
e

p
o
w
e
r

[
d
B
m
]
(
I
F
e
m
t
o
,
s
c
+
P
N
,
s
c
)

Normal operation region
PDCCH
SCH

Figure 9-2: The required Rx power per subcarrier to receive PDCCH and SCH
As shown in Figure 9-3 the impact of downlink interference for PDCCH can be calculated as
functions of femtocell transmission power. The curves are obtained by converting maximum
allowed path loss into distance according to specified path loss models [TS 36.211]. It is assumed
that femtocell is transmitting at full power over entire frequency band. The Figure 9-3 gives a
general trend that as the distance between macrocell and MUE is increased the distance of
separation between femtocell and MUE needs to be increased as well, in order to serve the
PDCCH at the MUE. Consequently, without adaptive femto transmission power management, the
deadzone for a visiting MUE (i.e. indoor macro user) around the femtocell will be significant when
the indoor macro user is at macrocell-edge. When the indoor macro user is located at 290m from
macrocell (ISD=500m), the deadzone is 11m radius if a femto transmit its power at 0dBm whereas
the deadzone is 5m if a femto decreases its power to -10dBm. In order to make the deadzone
down to an acceptable level, the femto transmission power must be configured to an appropriate
level based on femto cell environment. For a passing MUE (i.e. outdoor macro user), the deadzone
created by a femtocell is not significant. When the outdoor macro user is located at 290m from
.
macrocell, the deadzone is about 2m even if a femto transmit its power at 20dBm whereas the
deadzone of a visiting MUE enlarges to 27m.
0 5 10 15 20 25 30
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Minimum MUE-Femtocell separation [m]
M
a
x
i
m
u
m

M
a
c
r
o
c
e
l
l

r
a
d
i
u
s

[
k
m
]
Interference Scenario C.2(PDCCH)
V-MUE, Pmax=20 dBm
V-MUE, Pmax=10 dBm
V-MUE, Pmax=0 dBm
V-MUE, Pmax=-10 dBm
P-MUE, Pmax=20 dBm
P-MUE, Pmax=10 dBm
P-MUE, Pmax=0 dBm
P-MUE, Pmax=-10 dBm
Passing MUE
Visiting MUE

Figure 9-3: Maximum macrocell radius as a function of femtocell MUE
separation, assuming PDCCH
Table 9-1 shows the deadzone of an indoor PDCCH that can be supported for different femtocell
transmission power levels, when the ISD (inter-site distance) of the macrocell is 1732m and 500m,
i.e. the radius of the macrocell is 1000m and 290m, respectively [TS 36.211]. Results are obtained
by converting maximum allowed path loss into distance using appropriate path loss model. The
deadzone of a visiting MUE varies from 19m (Femto transmission power = -10dBm) to 50m (Femto
transmission power = 20dBm) for 1732m ISD and from 5.1m to 27m for 500m ISD whereas the
deadzone of a passing MUE is small compared to that of a visiting MUE.
Table 9-1: Macrocell deadzone of PDCCH by a femtocell
Femtocell
transmission
power (dBm)
Deadzone (m)
Macrocell ISD=1732m
Deadzone (m)
Macrocell ISD=500m
Visiting MUE Passing MUE Visiting MUE Passing MUE
20 50 11 27 2.1
10 39 5.4 18 0.8
0 29 2.2 11 0.3
-10 19 0.8 5.1 less than 0.2

.
In order to camp on the macrocell, the value of S
Macro,sc
/(I
Femto,sc
+P
N,sc
) must be greater than a
required SCH s/Iot. The relation between the maximum strength of the downlink interference
plus noise per subcarrier and the minimum strength of wanted signal per subcarrier is also shown
in Figure 9-2. In the same manner for PDCCH analysis, the impact of downlink interference for SCH
is converted to functions of femtocell transmission power as shown in Figure 9-4. It gives a trend
that without femto transmission power management the visiting MUE to femto separation is too
large to camp on the macrocell whereas the passing MUE can get closer to femtocell. From [TS
36.101] the received power per subcarrier (S
Macro,sc
) should be higher than minimum sensitivity. In
other words, a cell is considered detectable if received power in SCH (SCH_RP) is greater than or
equal to -127 dBm. Generally handover is not triggered by the minimum sensitivity but
interference from other cells. Figure 9-5 shows the received power in SCH for a visiting MUE,
which is converted to received power from macro cell radius. In this analysis receiver sensitivity
always meets the requirement, i.e., SCH_RP-127dBm
0 5 10 15 20 25 30
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
M
a
x
i
m
u
m

M
a
c
r
o
c
e
l
l

r
a
d
i
u
s

[
k
m
]
Interference Scenario C.2(SCH)
V-MUE, Pmax=20 dBm
V-MUE, Pmax=10 dBm
V-MUE, Pmax=0 dBm
V-MUE, Pmax=-10 dBm
P-MUE, Pmax=20 dBm
P-MUE, Pmax=10 dBm
P-MUE, Pmax=0 dBm
P-MUE, Pmax=-10 dBm
Passing MUE
Visiting MUE

Figure 9-4: Maximum macrocell radius as a function of femtocell-MUE separation, assuming
SCH

.
0 5 10 15 20 25 30
-130
-120
-110
-100
-90
-80
-70
-60
-50
-40
-30
S
C
H
_
R
P

f
r
o
m

m
a
c
r
o

c
e
l
l

[
d
B
m
]
Receiver sensitivity for a visiting MUE(SCH)
V-MUE, Pmax=20 dBm
V-MUE, Pmax=10 dBm
V-MUE, Pmax=0 dBm
V-MUE, Pmax=-10 dBm
Minimum receiver sensitivity

Figure 9-5: Minimum SCH Rx power from macrocell
Table 9-2 gives the deadzone of a visiting MUE that can be supported for different femtocell
transmission power levels, when the ISD of the macrocell is 1732m and 500m, respectively. The
deadzone of a visiting MUE varies from 15m (Femto transmission power = -10dBm) to 45m (Femto
transmission power = 20dBm) for 1732m ISD and from 3.5m to 23m for 500m ISD whereas the
deadzone of a passing MUE is very small compared to that of a visiting MUE.
Table 9-2: Macrocell deadzone of SCH by a femtocell
Femtocell
transmission
power (dBm)
Deadzone (m)
Macrocell ISD=1732m
Deadzone (m)
Macrocell ISD=500m
Visiting MUE Passing MUE Visiting MUE Passing MUE
20 45 8.0 23 1.3
10 34 3.6 15 0.5
0 24 1.4 7.8 less than 0.2
-10 15 0.5 3.5 less than 0.2

.
9.4 Mitigation Technique Recommendation
The femtocell should manage its transmission power under this interference scenario and the
hand over of MUE to another FA or RAT is another method to mitigate femtocell interference.
9.5 Conclusions
This section presents the analytical results for the interference scenario C2 using the link-budget
type calculation and 3 GPP requirements for two control channels, i.e., PDCCCH and SCH.
From the results, the femtocell transmission power can affect the coverage of the macrocell. The
macrocell deadzone for a visiting MUE highly depends on the femtocell transmission power and
distance between macrocell and MUE. For example, the deadzone of a visiting MUE may vary
from 19m to 50m for PDCCH and from 15m to 45m for SCH at a large radius (ISD=1732m). For a
passing MUE, if femto transmission power is managed around -10dBm, the deadzone is very small
(less than 0.8m) when it is adjacent to a heavy wall. However if a femto cell is located next to a
light wall or window, the interference impact on a passing MUE may not be negligible.
One potential mitigation techniques for this scenario is adaptive management of femtocell
transmission power to provide adequate femtocell coverage and maintain macro cell control
channel deadzone within a small range. Another possibility is to allow for registration of MUEs as
femtocell guests since the visiting MUEs suffer severely from femtocell interference.
Subsequently, the MUEs can be handed over to a preferred cell. A third approach for interference
mitigation in this scenario is to have the macrocell handover MUE to another FA or RAT before
MUE connection is deteriorated and dropped.
.
10 Scenario D2: Femtocell Uplink Interference to the Macrocell NodeB Receiver
The interference scenario is as follows:
D.2 - Femto UE Tx to Macro BS Rx (traffic channel)
The femtocell is assumed to have closed access, and shared spectrum is assumed between the
macro and femto layers.
parameters:
Deployment Both suburban and urban are modelled
Macro layer has 7 sites (21 sectors) with wrap-
around, 500m ISD.
10% (suburban) or 0% (urban) of femto UEs are
outdoors and 20% of macro UEs are outdoors.
Number of macro UEs per sector 10 or 20
UEs)
Macrocell Power uplink control Both baseline Fractional Power control and a
Proprietary method are modelled
Femtocell uplink power control Max power based on limiting noise rise to
macro neighbours (a similar approach to that
described in [3GPP 25.967] section 7.5.1 for
WCDMA).
Scheduler Both baseline round robin and Proprietary
Frequency Selective/Proportional Fair
.
Traffic model Full Buffer or Bursty
In the case of bursty traffic being modelled, 70%
of UEs use the bursty traffic model with the
parameters are described in [OFDMA Interf], the
remaining UEs are full-buffer.
Apartment block model Dual stripe, 6 floors (=240 apartments), one
dual stripe randomly dropped per macro
sector. A variable probability of having active
femto in each apartment.
Pathloss model Full (rather than simplified) model

Results for the suburban scenario are presented for 2 cases (Baseline and Enhanced) which are
defined as follows:
Table 10-2: Simulation Cases
Case Name Definition
Baseline Fractional power control on macro
Power cap on the femto UEs to protect the macro
layer (similar to 3GPP 25.967 for WCDMA)
Round Robin Scheduling
Enhanced Enhanced power control on macro (similar to
3GPP 25.967 for WCDMA)
Power cap on the femto UEs to protect the macro
layer (similar to 3GPP 25.967 for WCDMA)
Proportional Fair Frequency Selective Scheduling

In both the baseline and enhanced cases a power cap is placed on the femto UEs in order to
protect the macro uplink. The power cap is calculated as a function of the smallest measured
pathloss to a macro, and also as a function of a target maximum noise rise that the femto UE
should generate at the macro.
For the urban scenario only the Enhanced Scheduling and Power Control approach is assumed.
Finally note that the system simulations model interference from both the femtos to the macros
and vice-versa and therefore scenarios B.1 and D.2 (see section 9) are addressed simultaneously.
.
10.2.1 Suburban Deployment Model, Fixed Power Cap, Full Buffer Traffic
The simulation results presented in this section assume full buffer traffic model, with a fixed
power cap (selected to give either 0.2 or 1.8 dB noise rise at the macro eNB per femto UE), and 10
macro UEs per macrocell sector.
Simulation results for the macrocell uplink average sector throughput as a function of femto
density are shown in Figure 10-1 below:
0 10 20 30 40 50 60 70 80
2000
3000
4000
5000
6000
7000
8000
9000
10000
11000
12000
Macro Layer Uplink, Av. Sector Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

Baseline 1.8 dB noise rise thresh
Enhanced 1.8 dB noise rise thresh

Figure 10-1: Macrocell uplink average sector throughput
Note that the macrocell average user throughput (not shown) is 10% of the average sector
throughput since there are 10 macro UEs per sector.
Simulation results for the macrocell 5-percentile uplink user throughput as a function of femto

.
0 10 20 30 40 50 60 70 80
0
50
100
150
200
250
Macro Layer Uplink, 5 percentile User Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)


Figure 10-2: Macrocell uplink 5 percentile user throughput
Simulation results for the macrocell interference over thermal are shown in Figure 10-3 below:
0 10 20 30 40 50 60 70 80
6
7
8
9
10
11
12
13
14
15
Macro Layer Uplink, Av. IoT (dB)
A
v
.

I
o
T

(
d
B
)


Figure 10-3: Macrocell Interference over Thermal
Simulation results for the femtocell uplink average sector throughput as a function of femto

.
0 10 20 30 40 50 60 70 80
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
x 10
4
Femto Layer Uplink, Av. Cell Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)


Figure 10-4: Femtocell uplink average throughput
Note that the femtocell average user throughput (not shown) is equal to average sector
throughput since there is 1 femto UE per femtocell.
Simulation results for the femtocell 5-percentile uplink user throughput as a function of femtocell
0 10 20 30 40 50 60 70 80
0
1000
2000
3000
4000
5000
6000
7000
8000
Femto Layer Uplink, 5 percentile User Throughput
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)


Figure 10-5: Femtocell uplink 5 percentile user throughput
.
Simulation results for the femtocell interference over thermal are shown in Figure 10-6 below:
0 10 20 30 40 50 60 70 80
-8
-7
-6
-5
-4
-3
-2
-1
0
1
Femto Layer Uplink, Av. IoT (dB)
A
v
.

I
o
T

(
d
B
)


Figure 10-6: Femtocell Interference over Thermal
10.2.1.1 Discussion
Points to note regarding these results:
The LTE macrocell uplink performance is significantly improved with the enhanced power
control and scheduling algorithms compared to the baseline.
The LTE macrocell uplink performance degrades with increasing femto density, however
with simple interference mitigation this can be managed such that the resulting
performance loss is reasonable while at the same time the total system throughout
(macro + femto) increases significantly.
Results are shown for two values of the target maximum noise rise that the femto UE
should generate at the macro. It can be seen that as the density of active femtos
increases, the target maximum noise rise should be reduced in order to maintain the
macrocell performance.
There is a small degradation to the femtos when the macro layer uses the enhanced
power control and scheduling scheme (due to a more aggressive use of uplink power in
the macro layer).
The femtocell performance is a function of the target maximum noise rise at the
macrocell, with best performance with higher (=more relaxed) target maximum noise
rise.
.
10.2.2 Urban Deployment Model, Fixed Power Cap, Full Buffer Traffic
The simulation results presented in this section assume full buffer traffic model, with a fixed
power cap (selected to give either 0.2 or 1.8 dB noise rise at the macro eNB per femto UE), and 10
macro UEs per macrocell sector. The enhanced scheduling and power control case is assumed.
Figure 10-7 below shows the average macrocell sector throughput as a function of the probability
that there is an active femtocell in an apartment. Results are shown for two values of the target
maximum noise rise that the femto UE should generate at the macro. It can be seen that as the
density of active femtos increases, the target maximum noise rise should be reduced in order to
maintain the macrocell performance.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
x 10
4
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)
Prob. active femto in apt.

0.2 dB noise rise thresh

probability that there is an active femtocell in an apartment. It can again be seen that as the
density of active femtos increases, the target maximum noise rise should be reduced in order to
maintain the macrocell performance.

.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
0
50
100
150
200
250
300
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)


Figure 10-9 below shows the mean Interference over Thermal (IoT) at the macrocell. It can be seen
that IoT can be controlled by adjusting the target maximum noise rise.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
7
8
9
10
11
12
13
A
v
.

I
o
T

(
d
B
)



.
Figure 10-10 below shows the average femtocell sector throughput as a function of the probability
maximum noise rise that the femto UE should generate at the macro. It can be seen that the
femtocell performance is a function of the target maximum noise rise at the macrocell, with best
performance with higher (=more relaxed) target maximum noise rise. There is also some
evidence of interference between femtocells since the performance degrades slightly with active
femtocell density.
0.1 0.15 0.2 0.25 0.3 0.35 0.4
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
x 10
4
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)


Figure 10-10: Femtocell uplink average sector throughput

Figure 10-11 below shows the cell edge (5 percentile) femto user throughput as a function of the
probability that there is an active femtocell in an apartment. Again it can be seen that the
femtocell performance is a function of the target maximum noise rise at the macrocell, with best
performance with higher (=more relaxed) target maximum noise rise.
.
0.1 0.15 0.2 0.25 0.3 0.35 0.4
0
1000
2000
3000
4000
5000
6000
7000
8000
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)


Figure 10-12 shows the mean Interference over Thermal (IoT) at the femtocell.
0.1 0.15 0.2 0.25 0.3 0.35 0.4
-10
-8
-6
-4
-2
0
2
4
A
v
.

I
o
T

(
d
B
)



.
10.2.2.1 Discussion
Similar to the suburban deployment model results:
The LTE macrocell uplink performance degrades with increasing femto density, however
with simple interference mitigation this can be managed such that the resulting
performance loss is reasonable while at the same time the total system throughout
(macro + femto) increases significantly.
Results are shown for two values of the target maximum noise rise that the femto UE
should generate at the macro. It can be seen that as the density of active femtos
increases, the target maximum noise rise should be reduced in order to maintain the
macrocell performance.
The femtocell performance is a function of the target maximum noise rise at the
macrocell, with best performance with higher (=more relaxed) target maximum noise
rise.
10.2.2.2 Urban Deployment Model, Full Buffer or Bursty Traffic, Fixed or Adaptive Power Cap
The results for suburban and urban deployment models with fixed power suggest that the
required cap on the femto UE power should be made a function of femto density within the
macro cell. This would provide adequate performance at the macrocell for higher femto densities,
and improved femtocell performance at lower femtocell densities. This suggests that the femto UE
power cap should be adaptive.
An adaptive approach must be robust such that an appropriate power cap is determined for
different densities of femtos, different traffic models, and different deployment scenarios. In this
section one such scheme is evaluated for varying femto density and for either full buffer or bursty
traffic. The results are compared to a fixed power cap of either 0.2 dB (labelled tight) or 7 dB
(labelled loose, and it should be noted that this is a very loose cap, for which in practice the
femto UE power will likely be set considering coverage requirements of the femto alone rather
than also considering interference to the macro layer).
For these simulations there are 20 macro UEs per sector. The enhanced scheduling and power
control case is assumed.
The adaptive scheme evaluated here makes use of X2 signalling from the macro. As mentioned
above the power cap at the femto UE would typically be derived from an estimate of the pathloss
to neighbouring macro eNBs together with a target noise rise at the macro eNB. Therefore the
power cap at the femto UE is adjusted via the adaptation of the target noise rise based on X2
signalling received from the macro eNB (possibly via an X2 proxy or gateway see below). For
example, if the X2 load indication indicates high interference at the macro eNB, then the target
.
noise rise is set to a smaller value than if the X2 load indication indicates medium or low
interference, etc.
The X2 signalling is assumed to be relatively slow i.e. no attempt is made the track varying
interference due to scheduler decisions or bursty traffic on a TTI by TTI basis, rather the X2
signalling is used to indicate more of the average interference conditions. In the simulation
results below a delay of 50ms is assumed in the X2 signalling.
Furthermore, in order to distinguish between inter-cell interference within the macro layer and
interference generated by the femtos, the macro (or an X2 proxy or gateway see below)
periodically tries signalling different X2 load indication values towards the femtos and measures
the resulting impact on IoT in order to refine the load indications sent to the femtos.
Figure 10-13 below shows the average macrocell sector throughput as a function of the probability
maximum noise rise that the femto UE should generate at the macro (tight and loose) and
for the adaptive approach. It can be seen that with a low density of active femtos the adaptive
approach provides protection similar to the loose fixed approach but as the density increases
the protection approaches that of the tight fixed approach, and overall the macro is suitable
protected for all densities. This goes for both the full buffer and the bursty traffic models.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
5000
6000
7000
8000
9000
10000
11000
12000
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

full buffer loose
full buffer tight
full buffer adaptive
bursty loose
bursty tight
bursty adaptive


.
probability that there is an active femtocell in an apartment. Again it can be seen that with a low
density of active femtos the adaptive approach provides protection similar to the loose fixed
approach but as the density increases the protection approaches that of the tight fixed
approach, and overall the macro is suitable protected for all densities. This goes for both the full
buffer and the bursty traffic models.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
0
50
100
150
200
250
300
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

full buffer loose
full buffer tight
bursty loose
bursty tight
bursty adaptive

Figure 10-15 below shows the mean Interference over Thermal (IoT) at the macrocell. It can be
seen that the IoT is suitable controlled by the adaptive approach.
.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
7
8
9
10
11
12
13
A
v
.

I
o
T

(
d
B
)

full buffer loose
full buffer tight
bursty loose
bursty tight
bursty adaptive

Figure 10-16 below shows the average femtocell sector throughput as a function of the probability
that there is an active femtocell in an apartment. It can be seen that with a low density of active
femtos the adaptive approach provides performance more similar to the loose fixed approach
but as the density increases the performance approaches that of the tight fixed approach, and
overall the throughput is optimised for all densities. This goes for both the full buffer and the
bursty traffic models.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
x 10
4
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

full buffer loose
full buffer tight
bursty loose
bursty tight
bursty adaptive

Figure 10-16: Femtocell uplink average sector throughput
.
Figure 10-17 below shows the cell edge (5 percentile) femto user throughput as a function of
the probability that there is an active femtocell in an apartment. Again it can be seen that with
a low density of active femtos the adaptive approach provides performance more similar to
the loose fixed approach but as the density increases the performance approaches that of
the tight fixed approach, and overall the throughput is optimised for all densities. This goes
for both the full buffer and the bursty traffic models, but is particularly pronounced for bursty
traffic on the femtos where the interference caused to macros tends to be less which the
adaptive scheme can exploit.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
0
2000
4000
6000
8000
10000
12000
14000
16000
T
h
r
o
u
g
h
p
u
t

(
k
b
/
s
)

full buffer loose
full buffer tight
bursty loose
bursty tight
bursty adaptive


.
Figure 10-18 shows the mean Interference over Thermal (IoT) at the femtocell.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-10
-8
-6
-4
-2
0
2
4
A
v
.

I
o
T

(
d
B
)

full buffer loose
full buffer tight
bursty loose
bursty tight
bursty adaptive

10.2.2.3 Discussion
The X2 based adaptive scheme for setting the power cap of the femto UEs provides protection
to the macrocell uplink in all cases (see section 10.2.2.2) while allowing the throughput, and
particularly the cell edge throughput, on the femto to be improved for low densities of
femtocells. This is particularly pronounced for bursty traffic on the femtos where the
interference caused to macros tends to be less which the adaptive scheme can exploit.
10.3 Mitigation Techniques and Recommendations
As suggested above the required cap on the femto UE power could be made a function of femto
density within the macro cell. This could provide adequate performance at the macrocell for
higher femto densities, and improved femtocell performance at lower femtocell densities. Such an
adaptive approach must be robust such that an appropriate power cap is determined for different
densities of femtos, different traffic models, and different deployment scenarios.

.
10.3.1 Description of Mitigation Techniques
One approach evaluated here is to set the power cap of the femto UE as a function of uplink
interference being experienced at the macro eNB by making use of X2 signalling.
The power cap at the femto UE would typically be derived from an estimate of the pathloss to
neighbouring macro eNBs together with a target noise rise at the macro eNB. Therefore the power
cap at the femto UE is adjusted via the adaptation of the target noise rise based on X2 signalling
received from the macro eNB (possibly via an X2 proxy or gateway see below). For example, if
the X2 load indication indicates high interference at the macro eNB, then the target noise rise is
set to a smaller value than if the X2 load indication indicates medium or low interference, etc.
The X2 signalling is assumed to be relatively slow i.e. no attempt is made the track varying
interference due to scheduler decisions or bursty traffic on a TTI by TTI basis, rather the X2
signalling is used to indicate more of the average interference conditions. Furthermore, in order
to distinguish between inter-cell interference within the macro layer and interference generated
by the femtos, the macro (or an X2 proxy or gateway see below) periodically tries signalling
different X2 load indication values towards the femtos and measures the resulting impact on IoT in
order to determine what load indications to send to the femtos.
The X2 load indications are given per resource block (RB) and therefore if the femto UE is only
operating in a portion of the macro eNB bandwidth then only the X2 load indications for the
corresponding RBs need to be considered.
X2 for HeNB is not part of 3GPP release 8. However it could be provided for future releases. One
potential issue is that there could be a large number of femtos within the macro coverage area
which could lead to complexity issues at the macro eNB if each femto had an X2 interface to the
macro eNB. Examples of possible approaches to solving this include [R3-082442]:
Have an X2 concentration function in an X2 proxy or gateway
Only send load indications in the direction from the macro to the femtos, rather than
being bidirectional.
Running X2 over UDP transport
10.3.2 Recommendations
Allow provision of X2 for femtocells/HeNBs to allow the protection offered to the macro UL from a
femto UE to be adjusted according to the density and traffic profile of the active femtocells and
the UL interference conditions at the macro.

.
10.4 Conclusions
Simulations suggest that a simple approach of placing a cap on a femto UEs power can be effective
in controlling UL interference to a macro eNB. However the need to protect the macro uplink does
impact the femtocell uplink performance. The required cap on the femto UE power could be
made a function of femto density and interference conditions within the macro cell. This could
provide adequate performance at the macrocell for higher femto densities, and improved
femtocell performance at lower femtocell densities. An adaptive scheme based on X2 signalling
has been evaluated for full buffer and bursty traffic models. Some simplifications are possible to
reduce the complexity at the macro eNB due to X2.

.
11 Scenario E: Femtocell Downlink Interference to nearby Femtocell UE Receiver.
This section describes the modelling of scenario E, suggests enhancements to improve system
performance in this scenario, and presents the results of numerical analysis and the corresponding
performance comparisons. It consists of two subsections. The first subsection considers the case
of full buffer traffic for all users and proposes a resource coordination algorithm through
distributed FFR. The second subsection considers a mixture of full buffer and bursty traffic for
different users in the system and proposes a dynamic interference avoidance mechanism.
11.1 Full Buffer Traffic and Distributed FFR
System level simulations have been performed to investigate the Femto to Femto DL jamming
scenario. The baseline Rel 8 performance is evaluated assuming frequency reuse 1 and no
coordination between the femto cells. A distributed FFR is shown to demonstrate OFDMA
interference management techniques. In the rest of the paper, we first describe the distributed
FFR algorithm, and then we describe the simulation assumptions. Finally we compare the
performance at different femto penetration rate with and without OFDMA interference
management techniques.
11.1.1 Distributed FFR
In an embedded network, a communication link between a cell and a UE is often interference
limited due to neighbouring cell transmissions. One solution is to power control the transmission
of the interfering and the victim link. This approach works efficiently without losing system
dimension. But power control algorithm may have limited performance gain in some scenarios
when the victim is closer to the interfering transmitter than the interfering receiver. An alternative
solution is planned frequency reuse, such as 1/3 reuse, which is most effective for regular layouts
that can be easily coloured with different carrier frequencies.
Fractional frequency reuse algorithm is designed to orthogonalize dominant interferers in an
OFDMA system. Instead of having carrier-reuse, frequency reuse could be tailored to each user in
an OFDMA system. LTE supports FFR via subband CQI reporting, which allows a scheduler to
schedule users based on the subband CQI reporting that reflects different interference levels on
different subbands. In a 10 MHz system with 6 RBs/subband, there are 8 regular subbands and
one short subband that could be used to implement fractional frequency reuse.
Distributed FFR algorithm is designed to cope with non-operator deployed networks, such as
HeNBs. Each HeNB could construct an RF neighbour list through network listening and user
reporting. In this section, the local RF neighbour information is called a Jamming Graph, where
.
each node denotes an active HeNB and an edge denotes jamming condition between two HeNBs.
A jamming condition is declared when the channel gain difference between the interfering and
serving links exceeds certain threshold. The distributed FFR planning problem is now converted
into a graph colouring problem, which could be solved in a distributed manner at low complexity.
11.1.2 Simulation Assumptions
Downlink full buffer capacity has been investigated in this section. The link level performance is
based on single user 2x2 MIMO with channel and interference estimation loss. Link adaptation
and HARQ are modelled in the link level simulation. The baseline performance has been calibrated
with NGMN Rel 8 performance for D1 scenario. The LTE system parameters are shown in Table
11-1, where the total system overhead is close to 41%.
The femto cell is assumed to transmit at a fixed power of 10 dBm. This is inline with the maximum
transmit power that satisfy the ACLR requirements for HNB. DL control channel outage is modelled
with outage threshold at -10 dB geometry. The femto layout is based on the 5x5 Grid Model in
the draft evaluation methodology [OFDMA-EMD]. The path loss is given by PL(dB) =
127+30log10(R/1000), where R is in meter. Note that the shadowing standard deviation with this
simplified path loss model is 10 dB since no walls have been modelled. Penetration rates of 5 to
20% have been studied to test the robustness of the algorithm at different density.
Table 11-1 LTE system parameters and overhead
Parameters Values
Cyclic prefix 4.69us (6.57%)
Guard band 600 tones used with 15kHz subcarrier spacing (10%
in 10MHz)
DL Control 3 out of 14 OFDM symbols
Reference Signal 1 out of every 3 tones over 4 OFDM symbols in each
subframe
Sync channel
broadcast channel
6 RBs at subframes 0 and 5
Total Overhead 40.45%
.
11.1.3 Results and Discussion
Three interference management schemes have been compared in this section: Rel 8 with
frequency reuse 1; distributed FFR with medium level of interference orthogonalization;
distributed FFR with high level of interference orthogonalization. The level of orthogonalization is
controlled by tuning the jamming graph threshold, i.e., the channel gain difference between
interfering and serving link. In this simulation, 0 dB and -6 dB thresholds have been studied.
The decoding C/I and mobile throughput statistics are shown in Figures 1, 2 and 3 for 5%, 10%, and
20% penetration rate, respectively. Note that the C/I of UEs that are not scheduled have not been
logged in the decoding C/I CDF; hence, the decoding C/I is never lower than -10 dB in all the
figures.

(a) Decoding C/I (b) Mobile throughput
Figure 11-1 Mobile throughput and decoding C/I for 5% penetration rate
As shown in Figure 11-1, the C/I and mobile throughput improve with distributed FFR schemes for
the 5% penetration rate. More specifically, the 4% UE outage is eliminated through FFR schemes.
Note that the outage is mainly due to control channel outage at -10 dB. Other than outage users,
the overall system fairness is also shown to improve significantly.
As shown in Figures 2 and 3, the mobile outage for systems with no reuse increases to 8 and 12%
as the penetrate increases to 10 and 20%, respectively. In both cases, distributed FFR schemes
eliminate the system outage and significantly improve the system fairness. For example, when FFR
is enabled more than 20% and 30% of users doubled their throughput at 10 and 20% penetration
rate, respectively. It is noted that there is a negligible loss in throughput for high throughput users.

.


11.1.4 Conclusions
This section investigated the Femto to Femto DL interference scenario. It was shown that 4 to 12%
of HUEs will be in outage without interference management for 5 to 20% penetration rate.
OFDMA interference management schemes (e.g., FFR) is shown to eliminate the DL outage and to
further improve the system fairness. It was shown that when FFR is enabled more than 20% and
30% of users doubled their throughput at 10 and 20% femto penetration rate, respectively.
.
11.2 Mixed Traffic and Dynamic Interference Avoidance
Coordinated Multi-Point Transmission (CoMP) is considered to be an important technique for
achieving the performance requirements of LTE-Advanced [TR 36.814 v0.3.1]. In this context, a
procedure for transmit power and beam coordination between interfering eNBs is described in
[R1-090867]. This procedure can be used over the backhaul if a reliable and fast X2 interface is
available. Alternatively, for cases where an X2 interface may not be available or for cases where
the X2 interface latency may be unpredictable or excessive, the coordination can be carried out
over-the-air as described in [R1-090867]. Coordinated transmission between interfering HeNBs is
particularly important in heterogeneous deployments, where different HeNBs have very different
transmit powers or support Closed Subscriber Groups [R1-083190, R1-083195]. Moreover, only a
few UEs will typically be associated with a single Home HeNB (HeNB) this allows for only a
limited amount of statistical multiplexing to ensure that resources negotiated on a slow time scale
between neighbouring cells are fully utilized.
In this section, we present simulation results characterizing user performance for a HeNB
deployment when some cells have full-buffer traffic while others carry delay-sensitive traffic with
random arrivals. The results demonstrate the advantage of fast interference coordination, and
supplement the results for non-full buffer traffic presented in [R1-090869]. We restrict ourselves
to Cooperative Silencing (CS) or transmit power coordination in these simulations.
We focus only on the downlink in this section. Also, note that we do not consider coordination of
spatial beams. In this case, the Channel Direction Information (CDI) contained in the Spatial
Feedback Information (SFI) report, described in [R1-090867], is not utilized by the receiving eNB;
transmit power coordination will be based on utility metrics contained in the SFI report. Further
performance improvements could be obtained by making use of the CDI information in the SFI
report; these will be explored in future contributions.
11.2.1 System Model
11.2.1.1 Deployment and Channel Model
We consider a 5x5 cluster of apartments. Each apartment is populated with a HeNB with
probability p. If an apartment is populated with a HeNB, a single UE is associated with it. Both the
HeNB and the UE are dropped uniformly at random in the apartment. Path loss and lognormal
shadowing are modelled, but not fast fading. More details about the deployment model can be
found in [R1-083195]. For a given signal to interference and noise ratio, the resulting spectral
efficiency is given by capacity curves obtained using [R1-090355]. Perfect rate prediction and rate
granularity are assumed.
.
11.2.1.2 Traffic Model
For QoS users, we use the traffic model in [R1-090368]. The traffic model is such that there can be
at most one packet in the buffer at a HeNB that has to be transmitted to the associated UE. Once a
packet completes transmission, the next packet arrives after a time interval equal to an
exponentially distributed random variable, with a mean specified in Table 11-2. Note that this
model ensures stability of the queues at each HeNB.
11.2.1.3 SFI Transmission
Spatial Feedback Information (SFI) reports are sent by each HeNB to all other HeNBs in the cluster,
possibly through UEs which it serves. Such reports may contain information about the buffer state
(e.g., a function of queue length, packet delay, priority) and the SINR (with some nominal
interference) on the link from the HeNB to its associated UE. As mentioned earlier, an SFI report
would typically also contain CDI information; this information is ignored by the receiving eNB for
the purposes of this contribution. On receiving such reports, a HeNB determines transmission
attributes (e.g. power, sub-bands to use etc.) to its associated UE based on both the channel and
buffer state corresponding to its own transmissions relative to the channel and buffer state
contained in the received SFIs. In order to compute the transmission attributes, a HeNB may use
information it can infer about the channel gain from it to a non-associated UE; such information
can be obtained using reference signals and associated measurement reports, through the
contents of the SFI, or an appropriate scaling of SFI power when sent over-the-air. The exchange
of this information helps prioritize the different transmissions across the HeNB cluster, mitigate
interference and achieve the desired tradeoff between QoS and efficiency. Resource allocation
algorithms based on such information are outlined in Section 11.2.2.
In this section, we model the delay from the transmission of a SFI from a HeNB to the time when
such information is available at the other HeNBs which are the intended recipients of the SFI. Such
a delay can be incurred when the SFIs are transmitted over the backhaul X2 interface, or to convey
the desired information using over-the-air (OTA) signalling as described in [R1-090867].
11.2.2 Resource Allocation Algorithms
The following resource allocation schemes are studied in terms of their ability to efficiently
mitigate interference while meeting QoS requirements:
1. Reuse One: Each HeNB transmits to its associated UE over the entire bandwidth if it has
enough data in the buffer for that UE. In case the buffer has less data than that can be
transmitted over the entire bandwidth, the HeNB randomly selects sub-bands to transmit
on.
.
2. Spatial Feedback Information (SFI) with K subframe delay: During each sub-frame (1 ms), a
HeNB sends an SFI (possibly through its associated UE in the OTA scheme) to interfering
HeNBs in the cluster. The SFI indicates the priority of the HeNB sending the SFI (based on
delays of packets in the buffer and past achieved data rate) and the spectral efficiency
impact of the interference caused by the receiving HeNB. If the SFIs are received by the
HeNBs (SFI recipients) after a delay of K subframes, then at time t+K all SFIs transmitted at
time t are resolved. Note that even though a HeNB has access to its own buffer state at
current time t+K, it uses only its buffer state at time t to prioritize this helps achieve
consensus by ensuring that all contending HeNBs use the same information to determine
their actions
1
a. At time (t+K) a HeNB only transmits over enough resources which it can utilize
even though it may have contended for more resources in the SFI it sent at time t.
. However, the following two simple optimizations are made:
b. At time (t+K) a HeNB can use its preferred sub-band (based on the colouring
algorithm in [R1-090869]) if its buffer is non-empty and it hears no SFIs at time t+K
(sent at time t) contending for that sub-band.
At time (t+K) each HeNB uses the incoming SFIs to determine an increase in the utility
metric for its own transmission (at full power) versus an increase in the utility metric for
transmissions with which it is interfering (if it remains silent). The HeNB then makes a
decision to transmit at full power (can be generalized to other power levels) only if the
increase in its own utility metric is higher than the loss caused to the utility metric for
other HeNBs that sent a SFI at time t.
2
3. Genie SFI with K subframe delay: For the case of large subframe delays, the above
algorithm results in many SFIs being sent while the first one still has not been received. If

The choice of the metric is a tradeoff between fairness and high spectral efficiency. We
use a metric of packet delay*expected rate for bursty traffic and expected rate/(average
rate) for full-buffer traffic. Using packet delays or average rate helps achieve fairness,
while the expected rate component tries to ensure high spectral efficiency. We would like
to emphasize that many such metrics can be used in the SFI framework, thus resulting in
different flavours of SFI algorithms. For example, the fairness component can
alternatively be based on other functions of packet delays, queue length, or average rate
in the past.

1
Lack of consensus can mean that different HeNBs react to different (partial) system states this
can lead to uncoordinated transmissions and hence, either low utilization of the spectral resources
or high interference.
2
This computation is motivated by iterative algorithms which can be shown to stabilize queues
and optimize delay performance; similar iterations can be designed to maximize utilities of
average user rates as well.
.
the buffer at the sending HeNB is relatively small, the amount of bandwidth reserved by
that HeNB will be substantially larger than the amount of bandwidth needed to transmit its
buffer. This can result in some amount of unused bandwidth in the network. In order to
remove the effect of this inefficiency, for the case of large SFI delay, we show results
assuming that the HeNB receiving the SFI is aware of whether or not the HeNB sending the
SFI has a non-empty buffer. If the buffer is empty, the receiving HeNB ignores the SFI that
was received corresponding to that sub-frame. These results should be treated as an upper
bound on the performance of the large delay SFI case. Realistic techniques to limit the
amount of unused bandwidth will be explored in future contributions.
11.2.3 Numerical Results: Mix of QoS and Full-Buffer Traffic
11.2.3.1 Modelling Assumptions
The numerical results were generated using the System Model described above. The system
parameters are summarized in the Table 11-2 below.
Parameter Value
Bandwidth 10MHz
Apt Size 10 m by 10 m
Noise PSD -174 dBm/Hz
Noise Figure 5dB
Max. HeNB power 100 mW
Num Tx Antennas 2
Num Rx Antennas 2
Shadowing Std. Deviation 10 dB
Packet Size for QoS Traffic 16.25 KB
Min. Per Antenna C/I for data transmission -10 dB

We consider a mix of delay sensitive (QoS) flows and full-buffer flows in our simulations. Again,
one UE is associated with each cell each UE has only one flow which is either delay-sensitive or
full-buffer. For full-buffer flows the user performance is dependent on the average rate. In
.
particular, we assume a log utility function (corresponding to proportional fair) for full-buffer
flows.
We plot the delay CDF for QoS flows and the rate CDF for full-buffer flows. The results are plotted
for reuse one, SFI:8ms, SFI:100ms, and SFI:100 ms genie schemes. For the SFI schemes, the QoS
flows have strict priority over full-buffer flows this is optimal in the absence of fading, since
channel sensitive scheduling in the time dimension is not needed. The buffer empty/non-empty
information for the genie scheme is made available with a delay of 8ms to make the comparison
with SFI:8ms scheme fair.
11.2.3.2 Results
We show a small set of representative results to illustrate the underlying dynamics of all the
algorithms. Similar trends hold for a wide range of HeNB penetration, packet sizes, and traffic
mixes. The results for 50% penetration are shown in Figs. 1 and 2 below. Results for 20% and 70%
penetration are in Section 11.2.5.

(a) Delay CDF for QoS

(b) Rate CDF for full-buffer

Figure 11-4: 50% penetration, 25% full buffer
.



From the above figures, reuse-one is clearly not a feasible scheme due to high interference seen
by the receivers of certain links. Reuse-one results in large outages as noted in the table below.

Table 11-3: Total outage (full buffer and QoS) for reuse one for six different scenarios
Scenario Total outage for reuse one
20% penetration, 25% full buffer 5%

We make specific observations for QoS users and full-buffer users for different schemes below.
11.2.3.2.1 Delay Performance for QoS Users
The following observations can be made:
1. The delay performance for QoS users for all the three interference coordination schemes is
dominated by the time to access the medium. For example, the SFI:8ms scheme performs
.
very well because packets need to wait for only 8 ms before being allocated resources. For
the SFI:100ms and SFI:100ms genie schemes, some packets can be transmitted with a delay
less than 100 ms; this is because a cell can use its preferred band (allocated by colouring as
described in [R1-090869]) if no other QoS users contend for it. However, for many users,
the performance is still dominated by the time to access.
2. The SFI:100ms genie scheme performs better than SFI:100ms scheme for a fraction of
users. The genie scheme prevents users from backing off to SFI messages when the SFI
transmitter has an empty buffer, and hence results in more efficient use of spectral
resources. (Note that this unnecessary backing off is more of a concern for SFI:100ms than
for SFI:8ms because the buffer state available at the SFI receivers is much more stale.)
3. Reuse-one can lead to a significant percentage of QoS users to be in outage because (a)
there is no interference management, and (b) QoS users cannot be given higher priority
than full-buffer users. The percentage of users in outage increases when either the
percentage of full-buffer users is increased or the penetration rate is increased.
11.2.3.2.2 Rate CDF for Full-Buffer Users
We can make the following observations:
1. All SFI schemes result in a more fair resource allocation for full-buffer users compared to
reuse one. For reuse one, some full-buffer users can be in outage.
2. The performance of SFI:100ms can be significantly worse than that for SFI:8ms and
SFI:100ms genie; this is especially true when the QoS load is high. This is because QoS users
have strict priority over full-buffer users, and for SFI:100ms, full buffer users may back off
even when the (QoS) SFI transmitter has no packets to send.
11.2.4 Conclusions
The SFI8ms scheme outperforms the SFI:100ms and reuse one scheme in most regimes; thus it is
robust across different penetrations, loads, and traffic mixes. We can make the following two
broad conclusions from the study:
1. Enabling dynamic interference management and scheduling across HeNBs can help
significantly improve system efficiency and ensure fairness the fairness criterion can be
changed to easily tune system behavior. In particular, lack of interference management can
lead to outage of latency-sensitive flows in the presence of strong interference from
loaded cells. Coordination is needed even for fairness between different full-buffer users
and to prevent outage.
2. Reducing the latency incurred in the exchange of SFI reports for dynamic coordination can
lead to significant performance benefits. This is especially true for flows with relatively
small packets which need to be delivered quickly in the presence of strong interference.
.
Similar trends were observed in [R1-090869], where significant increases in latency were observed
when a bursty traffic profile was used in all cells. Based on these observations, we propose that
femto forum further investigate dynamic coordination techniques across interfering eNBs.
Moreover, since low-latency coordination between interfering eNBs is seen to provide significant
performance benefits, investigation of fast coordination schemes (over X2 or over-the-air) is worth
pursuing.
11.2.5 Additional Results






.






.
12 Scenario F1: Femtocell Uplink Interference to Nearby Femtocell Receivers
This section presents analysis results of the interference evaluation scenario F.1 as defined by
[OFDMA-Scen]. The title of this scenario is Femto UE Tx to Other Femto BS Rx (traffic channel).
The study focuses on evaluating the uplink interference experienced by a Femtocell BS from
another Femto UE. It is assumed that the aggressor UE and the served UE are deployed on shared
spectrum.
Two types of evaluation are performed. One approach is a link budget based approach which
considers two nearby apartments with statistical selection of separation between the victim
femtocell and the UEs. The other approach is a static Monte-carlo system simulation. These two
evaluations are described in the following subsections.
12.1 Two Apartment Model
12.1.1 Description
The same setup as defined in [UMTS-WP] is used. A femtocell is located within each apartment
(AP1 and AP2). User equipment 2 (UE2) is outside and at the edge of his femtocell coverage. For
this reason, UE2 is transmitting at its maximum power of 20dBm. The distance between AP1 and
UE2 is assumed to be less than or equal 3m (d
2
3m).

Figure 12-1: Illustration of interference scenrio F1

A link budget/statistical analysis is performed, where UE is inside and UE2 is outside the
apartment 1.
.
12.1.1.1 Analysis and Parameter settings
Methodology recommended in [OFDMA-EMD] was followed, using log-normal shadowing. For the
link between UE1 and AP1, the shadowing standard deviation of 4dB and between UE2 and AP1 of
8dB is used. In the first approach, ITU-R P.1238 path loss model with N = 28 is assumed (see [UMTS-
WP] and earlier version of [OFDMA-EMD] for details). Therefore
( ) 28 log log 20
10 10
+ + = n L d N f PL
f

Eq. 1
where:
N: distance power loss coefficient
f: frequency in MHz (2000 is assumed)
d: distance in m
L
f
: floor penetration loss factor in dB (20dB between AP1 and UE2, 0dB between AP1 and UE2)
n: number of floors between base station and user equipment
is valid. The required signal to interferences plus noise ratios (SINRs) for targeted block error rates
(BLER) can be evaluated through link level simulations. In the following example a SINR of 1dB is
assumed. In this scenario, the noise power of P
noise
< 100dBm (sensitivity requirements in [TS
36.104]) can be neglected. Assuming
P
tx2
= 20dBm
d
1
= 10m
d
2
= 3m
f = 2000 MHz
Eq. 1 yields PL
1
= 66dB and PL
2
= 71.4dB for path loss between AP1 and UE1 and between AP1 and
UE2 respectively. The interference power I at AP1 can be calculated as
dBm 4 . 51
2 2
= = PL P I
tx

.
Eq. 2
For the signal power S at AP1
{ } { } ( ) dBm 4 . 50 log 10
10
+ + =
noise
P lin I lin SINR S
Eq. 3
is valid. Finally, for the transmitted power P
tx1
from the UE1 follows:
dBm 6 . 15
1 1
+ = PL S P
tx

Eq. 4
As shadowing was not considered in the above example, Eq. 4 provides an average value. In Figure
12-2 the average values of transmitted power P
tx1
for SINR = 1dB and different distances are shown.
-30 -20 -10 0 10 20 30
1
2
3
4
5
6
7
8
9
10
P
tx1
[dBm]
d
1

[
m
]

d
2
=1m
d
2
=2m
d
2
=3m
d
2
=5m
d
2
=10m

Figure 12-2: Average values of P
tx1
for SINR of 1 dB and different distances

.
Simulation parameters for statistical analysis
For the statistical analysis the simulation parameters are set as follows:
House size 12x12m
min(d
1
) 1m
max(d
1
) 17m ( sqrt(12
2
+12
2
) )
min(d
2
) 1m
max(d
2
) 3m
Shadowing 4dB between AP1 and UE1, 8dB between AP1
Penetration loss 20dB for external walls
P
tx2
20dBm
UE antenna gain 0dBi
BS antenna gain 0dBi
f

2000MHz
12.1.2 Main Findings
The resulting CDFs of co-channel SINR are plotted in Figure 12-3 for 0dBm, 10dBm and 20dBm UE1
EIRP. The distances d
1
and d
2
are randomly chosen between min(d
1,2
) and max(d
1,2
) with an uniform
distribution. Fast fading is included in the SINR target. Referring to the above example of the 1dB
required SINR, from Figure 12-3 and Figure 12-4 it can be observed that for P
tx1
= 20dBm the
probability of not fulfilling this requirement is about 50%, which is too high. In order to keep the
1dB SINR, the distance d
2
should be increased and/or aggressors power P
tx2
decreased.
.
-80 -60 -40 -20 0 20 40 60
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SINR [dB]
P
r
o
b
(
S
I
N
R

<

a
b
s
c
i
s
s
a
)

P
tx1
=0 dBm
P
tx1
=10 dBm
P
tx1
=20 dBm

Figure 12-3: Uplink co-channel SINR at femtocell BS due to a femtocell aggressor
-40 -20 0 20 40 60 80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
P
tx1
[dBm]
P
r
o
b
(
P
t
x
1

<

a
b
s
c
i
s
s
a
)

Figure 12-4: Required power P
tx1
for 1 dB SINR

.
12.1.3 Conclusions
We have carried out a simple link-budget analysis of the interference scenario F.1 based on Monte
Carlo simulations. In this scenario, it is shown that uplink co-channel interference from an aggressor
UE located near to the victim femtocell adversely affects the transmission quality. Power
management will not sufficiently mitigate the interferences in this worst-case scenario without
decreasing the aggressors power. Therefore, additional methods (e.g. scheduling algorithms) are
recommended.
12.2 System Simulation Approach
The simulation parameters mostly follow the evaluation methodology document in [OFDMA-EMD]
with the following specific parameters listed in Table 12-1 and Table 12-2.
Table 12-1: System simulation parameters
Parameter Femto cell
Cellular Layout Urban dual-strip
Carrier Frequency 2000 MHz
Shadowing standard deviation 4 dB
Exterior wall penetration loss 20 dB
Antenna pattern omnidirectional
BS antenna gain after cable loss 0 dBi
UE Antenna gain 0 dBi
UE noise figure 9 dB
Maximum/Minimum UE TX power 24 dBm / -30dBm [4]
Minimum distance between UE and
HeNB
1 m

.
Table 12-2: Urban dual-strip parameters
Parameter Urban
R (deployment ratio) 0.2, 1
Number of active HUEs per femto-cell 1

The following scenarios with/without interference management are considered as described in
[OFDMA-EMD]:
Full power.
Adaptive power control 1 (APC1): fractional power control as defined for the macro uplink,
both power control parameter set 1 and 2 are simulated.
Adaptive power control 2 (APC2): the pathloss from the vicinity of the HeNB to the
neighbouring HeNBs is estimated from measurements, and based on this and other related
parameters, the HeNB can then determine a maximum allowed UE transmit power, such
that the noise rise experienced at the neighbour cells is constrained to be within an
acceptable limit. The noise rise thresholds (NRT) in our simulation are set to 0.2, 1.8 and
18dB, respectively.
12.2.1 Results
Table 12-3: Average cell and cell edge throughput under different power control methods, R=0.2
Power control method
Full
power
APC1 APC2
Set1 Set2 NRT=0.2dB NRT=1.8dB NRT=18dB
Average cell
throughput (Mbps)
15.813 12.763 11.730 12.380 15.016 15.676
Cell edge throughput
(Mbps)
7.375 6.843 4.841 3.294 4.963 6.584

.
Table 12-4: Average cell and cell edge throughput under different power control methods, R=1
Power control method
Full
power
APC1 APC2
Set1 Set2
NRT
=0.2dB
NRT
=1.8dB
NRT =18dB
Average cell
throughput (Mbps)
10.628 8.715 8.799 9.061 9.762 10.841
Cell edge throughput
(Mbps)
0.883 1.188 1.162 0.658 0.789 0.771

-30-20-10 0 10 20 30 40 50 60 70 80 90100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
HeNB SINR(dB)
C
D
F
R=0.2

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

-30 -20 -10 0 10 20 30 40 50 60
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
HeNB SINR(dB)
C
D
F
R=1

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

Figure 12-5: HUE SINR CDF

0 2 4 6 8 10 12 14 16 18 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
HeNB Throughput(Mbps)
C
D
F
R=0.2

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

0 2 4 6 8 10 12 14 16 18 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
HeNB Throughput(Mbps)
C
D
F
R=1

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

Figure 12-6: HUE throughput CDF
.
-80 -60 -40 -20 0 20 40 60 80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
IoT(dB)
C
D
F
R=0.2

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

-40 -20 0 20 40 60 80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
IoT(dB)
C
D
F
R=1

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

Figure 12-7: HUE IoT CDF
-30-25-20-15-10 -5 0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
HUE TxPower(dBm)
C
D
F
R=0.2

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

-30-25-20-15-10 -5 0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
HUE TxPower(dBm)
C
D
F
R=1

Full power
APC1 set1
APC2 0.2dB
APC2 18dB

Figure 12-8: HUE transmit power CDF
12.2.2 Conclusion
In this section, we study the uplink performance of femto cells considering the interference from
other femto cells through system simulation. Specifically, two power control methods are assumed
and compared with full power. It is shown that
Under small deployment ratio (R=0.2), full power achieves better performance compared
to power control methods in terms of average cell and cell edge throughput.
Under large deployment ratio (R=1), fractional power control (APC1) can achieve better cell
edge throughput compared to full power, at the cost of lower average cell throughput.
Power control based on noise rise (APC2) can improve the average cell throughput slightly
compared to full power when the noise rise threshold is large (18dB), at the cost of lower
cell edge throughput.

.
12.3 Scenario F.1 Overall Conclusion
The section considers the uplink interference experienced by a Femtocell BS from another Femto
UE. It is assumed that the aggressor UE and the served UE are deployed on shared spectrum.
It is shown that in some corner cases the uplink co-channel interference from an aggressor UE
located near to the victim femtocell can significantly affect the transmission quality. However
system simulation results suggest that the probability of such occurrences is low such that overall
system performance is acceptable. Furthermore at low deployment densities allowing UEs to
transmit at full power gives the best results, whereas at high deployment densities it is
advantageous to use uplink power control such as fractional power control to improve
performance in those cases where interference is most significant.

.
13 Scenario G3: Macrocell Downlink Interference to an adjacent channel Femtocell
Receiver
This section presents analysis of the interference evaluation scenario G.3 as defined by [OFDMA-
Scen]. The title of this scenario is Macrocell Downlink Interference to the adjacent channel
Femtocell Receiver.
The goal of this study is to evaluate the downlink interference experienced by a Femtocell UE (FUE)
from a Macro Node B. It is assumed that the Macro and Femto cellular layers are deployed on
adjacent frequencies, so a relevant specification will be adjacent channel selectivity (ACS). The
impact of interference is evaluated in terms of SIR, and 3GPP transceiver specifications will be used
in the analysis.
13.1 Description of the interference scenario
The Femtocell UE (FUE) has already established a call, but the house location may be such that the
Macro signal is very strong. Figure 13-1 illustrates the interference scenario G.3 of [OFDMA-Scen].

Figure 13-1: Illustration of the Interference Scenario G.3.
13.2 Analysis
A link budget/statistical analysis is performed, where the suburban homes are located uniformly in
a sector, and FUEs are located uniformly within a house.

.
13.2.1 Parameter settings
A frequency separation of 5 MHz between Femtocell UE (FUE) and Macrocell UE (MUE) is
considered. The parameter settings that are used in the analysis are given below. Antenna gains
are not part of the analysis, as they affect all signals and thermal noise is negligible compared to the
interference.
Femtocell parameters -
Adjacent Channel Selectivity (ACS) of the Femtocell UE receiver is 33 dB in this analysis, Table 7.5
[TS 36.101].
13.3 Main Findings
Methodology recommended in [OFDMA-EMD] was followed, using macrocell separation of 1732 m,
and outdoor wall penetration loss L
ow
=10 dB. Other parameters included 60 dBm macro EIRP.
Small scale fading was not included in the calculations of received powers, under the assumption
that receive antenna diversity and/or frequency diversity are sufficient. The ensemble of received
signal powers was generated through statistical variation of distance from the macro base as well
as shadow fading.
The resulting CDF of co-channel SIR is plotted in Figure 13-2 for 20 dBm femto EIRP and in Figure
13-3 for 0 dBm femto EIRP. It may be observed that in this suburban environment adjacent-channel
macro-femto interference is not a significant problem, with 99% of locations observing SINR in
excess of 8 dB even when with femto power of 0 dBm. Increasing the Femto AP transmit power or
ACS shifts the CDF correspondingly to the right, i.e. improves the SINR further. On the other hand,
co-channel operations (ACS=0 dB) and low power femto AP (0 dBm) will result in SINR<5 dB in 40%
of cases.

.
-40 -30 -20 -10 0 10 20 30 40
10
-4
10
-3
10
-2
10
-1
10
0
SINR (dB)
P
r
o
b
(
S
I
N
R

<

a
b
s
c
i
s
s
a
)

ACS=33 dB
ACS=0 dB

Figure 13-2: Downlink co-channel SINR at FUE due to a macrocell for 20 dBm femto EIRP
and ACS of 0 and 33 dB
-40 -30 -20 -10 0 10 20 30 40
10
-4
10
-3
10
-2
10
-1
10
0
SINR (dB)
P
r
o
b
(
S
I
N
R

<

a
b
s
c
i
s
s
a
)
ACS=33 dB
ACS=0 dB

and ACS or 0 and 33 dB
.
13.4 Conclusions
We have carried out a simple analysis of the interference scenario G.3 based on link-budget
calculations and 3GPP specifications. 99% of the locations will see SINR of 8 dB or greater, with 0
dBm femto AP pwer and ACS of 33 dB.
13.4.1 Mitigation techniques
Maximum transmit power should not be made too small.
.
14 Scenario G4: Macrocell Downlink Interference to the adjacent channel
Femtocell UE Receiver
This section presents analysis of the interference evaluation scenario G.4. The title of this scenario
is Macro BS Tx to Adjacent Channel Femto UE Rx (control channel) for LTE. Operation of the
physical downlink control channel (PDCCH) will be evaluated. This control channel defines location
of downlink assignment in the time-frequency grid and modulation & coding.
Downlink interference is experienced by a Femtocell UE (FUE) from a Macro Node B. It is assumed
that the macro and femto cellular layers are deployed on adjacent frequencies, so the relevant
specification will be adjacent channel selectivity (ACS) and out-of-band transmit specification
(ACLR), which together make up the adjacent channel interference ratio (ACIR). Considering all
placements of the FUE within the home, and all placements of the home in the macro-cell area
yields a cumulative density function (CDF) of SIR. The CDF may be used to determine the
percentage of locations that the SIR is below a certain threshold, in this case the minimum required
for proper control channel operation.
The Femtocell UE (FUE) has already established a call, but the house location may be such that the
macro signal is very strong. Figure 14-1 illustrates the interference scenario G.4, which is identical
to scenario G.3.

Figure 14-1: Illustration of the Interference Scenario G.4.
.
14.2 Analysis
A link budget/statistical analysis is performed, where the suburban homes are located uniformly in
a sector, and FUEs are located uniformly within a house.
A frequency separation of 5 MHz between Femtocell UE (FUE) and Macrocell is considered. The
parameter settings that are used in the analysis are given below. Antenna gains are not part of the
analysis, as they affect all signals and thermal noise is negligible compared to the interference.
Adjacent Channel Selectivity (ACS) of the Femtocell UE receiver is 33 dB in this analysis, Table 7.5
[TS 36.101]. The Adjacent Channel Leakage Ratio (ACLR) of the eNodeB is 45 dB. Thus the ACIR is
approximately 33 dB and is limited by the UE.
The minimum SINR required to properly decode the control channel PDCCH is approximately -2 dB,
with the maximum amount of coding, R=1/12 , i.e. 8 control channel elements (CCE) in ETU70
channel [TS 36.101, pg. 53]. For the least amount of coding, R=1/3 1 CCE, approximately 7 dB is
required. These requirements could be higher, as indoor channels do not provide significant
frequency diversity in 5 MHz.
14.3 Main Findings
Methodology recommended in [OFDMA-EMD] was followed, using macrocell separation of 1732 m,
and outdoor wall penetration loss L
ow
=10 dB. Other parameters included 60 dBm macro EIRP. The
ensemble of received signal powers was generated through statistical variation of distance from
the macro base as well as shadow fading.
The resulting CDF of co-channel SIR is plotted in Figure 14-2 for 20 dBm femto EIRP and in Figure
14-2 for 0 dBm femto EIRP. It may be observed that in this suburban environment adjacent-channel
macro-femto interference is not a significant problem for the control channel, with 99% of locations
observing SINR in excess of 8 dB even when with femto power of 0 dBm. Increasing the Femto AP
transmit power or ACIR shifts the CDF correspondingly to the right, i.e. improves the SINR further.
On the other hand, co-channel operations (ACIR=0 dB) and low power femto AP (0 dBm) will result
in SINR<-2 dB in 2% of cases.

.

-40 -30 -20 -10
0
10 20 30 40
10
-4
10
-3
10
-2
10
-1
10
0
SINR (dB)
Pr ob(SINR < abscissa)

ACIR=33 dB
ACIR=0 dB

and ACIR of 0 and 30 dB

-40 -30 -20 -10
0
10 20 30 40
10
-4
10
-3
10
-2
10
-1
10
0
SINR (dB)
Pr ob(SINR < abscissa)
ACIR=33 dB
ACIR=0 dB

Figure 14-3: Downlink co-channel SINR at FUE due to macrocell for 0 dBm femto EIRP
and ACIR of 0 and 33 dB
.
14.4 Conclusions
We have carried out a simple analysis of the interference scenario G.4 based on link-budget
calculations and 3GPP specifications. 99% of the locations will see SINR of 8 dB or greater, with 0
dBm femto AP power and ACIR of 33 dB. This is sufficient to decode the PDCCH with coded at
R=1/3 or 1 CCE. 99.8% of the location will see SINR of -2 dB or greater, which is enough to decode
the PDCCH when maximum coding is used.
Maximum transmit power should not be made too small.

.
15 Scenario H1: Macrocell Uplink Interference to the adjacent channel Femtocell
Receiver
This section presents analysis of the interference evaluation scenario H.1 of [OFDMA-Scen]. The
title of this scenario is Macrocell Uplink Interference to the adjacent channel Femtocell Receiver.
The goal of this study is to evaluate the uplink interference experienced by a Femtocell from a UE
that is connected to a Macro Node B, when the UE and Femtocell are located in close proximity (i.e.
both located in a home). It is assumed that the Macro and Femto cellular layers are deployed on
adjacent frequencies, so the relevant specification is adjacent channel selectivity (ACS). The impact
of interference is evaluated in terms of SIR, and 3GPP transceiver specifications will be used in the
analysis.
The Macrocell connected UE (MUE) is located within the same home as the Femtocell, and has a
call established at full power. The Femtocell UE (FUE) has already established a call. Figure 15-1
illustrates the interference scenario H.1 of [OFDMA-Scen].

Figure 15-1: Illustration of the Interference Scenario H.1.
.
15.2 Analysis
A frequency separation of 5 MHz between Femtocell UE (FUE) and Macrocell UE (MUE) is
considered. The parameter settings that are used in the analysis are given below. Antenna gains
are not part of the analysis, as they affect all signals and thermal noise is negligible compared to the
interference.
Adjacent Channel Interference Rejection (ACIR) of the Femtocell receiver for is around 33 dB.
The following parameters were taken from the Evaluation Methodology Document [OFDMA-EMD]:
Femto AP TX power: 20 dBm EIRP
Macro UE TX power: 24 dBm EIRP
Building Penetration Loss: 20 dB
Pathloss Model UE to AP: PL (dB) = 38.46 + 20 log
10
R + 0.7R

Table 15-1: Suburban femtocell modelling parameters
Femtocell parameters -
Parameter Value
House size 12x12m
House+Lot size (12f) x (12f) m with f chosen to
give desired probability of femto
UE being outdoors when
randomly dropped in total area
of house plus lot.
Probability femto UE outdoors 10%
Macro UEs allowed in femto
house
Yes, macro UEs are randomly
dropped within macro coverage
area, and a macro UE may
happen to be dropped within the
.
12x12m femto house
Macro UEs outdoors No all macro UEs are indoors
(either in femto house or
another house)
Allow Femto houses+lots to
overlap
No
Minimum separation UE to femto
BS
1m
Minimum separation femto BS to
macro BS
35m
Number of active femto UEs per
femto-cell
1
Distribution of femto houses Random uniform within macro
coverage area, subject to
minimum separation to macro BS
and non-overlapping constraint.
Distribution of femto UE within
femto house
Random uniform, subject to
minimum separation to femto BS
Distribution of femto BS within
femto house
Random uniform

.
15.3 Main Findings
If the receiver operates in the linear region, then the interfering signal is attenuated by the ACIR,
and an outage is said to occur if the SIR is below the minimum requirement. It is assumed that as
the noise rises, due the approach of a MUE, the FUE will power up, up to its own maximum power.
To determine the probability of outage, we allow both the FUE and the MUE be randomly and
uniformly located in the home. Both are set to maximum power. Figure 2 gives the distribution of
SIR for ACIR equal to 33 dB. For 95% of the locations the SIR is greater than 14 dB. If the required
SIR is higher, the outage would be correspondingly higher. Since the observed outage is low, it is
concluded that allowing the FUE to adjust its transmit power to the noise rise is an effective
strategy.
-20 -10 0 10 20 30 40 50 60 70 80
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SIR (dB)
P
r
o
b
(
S
I
R

<

a
b
s
c
i
s
s
a
)

Figure 15-2: SIR for ACIR = 33 dB, for suburban house
Note, if the FUE increases power, this will increase co-channel interference at other Femtocells.
Since these other Femtocells are far way, i.e. high pathloss, this will in turn cause the neighbouring
FUEs to increase their own power by a small amount, and thus the effect of the MUE is spatially
limited.
.
15.4 Conclusions
We have carried out an analysis of the interference scenario H.1 based on link-budget, Monto Carlo
simulation, using 3GPP specifications, and following the Femto Forum Evaluation Methodology.
For 95% of the locations the SIR is greater than 14 dB. The dynamic range of the femto receiver will
need to be large to handle the possible large signal from the UEs due to the small path loss.
The Femtocell should command the UE to increase its transmit power under this interference
scenario.
.
16 Detailed Summary of Findings
Scenario Key Conclusions Level of Impact
A1 Macrocell
Downlink
Interference to
the Femtocell
UE Receiver
(Traffic
Channel)
The impact of the interference from macro
cell on the average femto cell throughput is
not severe and the corresponding
degradation is negligible. Note that
evaluation of the impact on other
performance metrics such as tial of the
user throughput CDF may require further
investigation.
It was found that the loss in the average
femto cell throughput is 1% or less, even
when the femto BS transmit power is as
little as 0dBm. Further study may be
required to quantify the level of impact
on the tail of user throughput CDF.
B1 Macrocell
Uplink
Interference to
the Femtocell
Receiver
(Traffic
Channel)
Although the presence of the macro UEs
was shown to cause a rise in the
interference over thermal as seen at the
femto, the average interference level
remains low relative to the thermal noise
and the average throughput is not
impacted.
In a macro-femto setup with 20 dB
exterior wall penetration, it was
observed that average IoT at the femtos
can rise from -22dB when no macro UEs
are present to -4.2dB when there are 10
macro UEs per cell. Being still well under
thermal, this IoT rise has no noticeable
impact on the average throughput seen
at the femtocells.
.
C1 Femtocell
Downlink
Interference to
the Macrocell
UE Receiver
(Traffic
Channel)
It is shown that suitable power control of
the femtocells is important in managing
interference to the macrocells and that this
becomes more important as the density of
femtocells increases. In general the
femtocell power should be decreased as
the femto becomes closer to the edge of
the macrocell, and power control schemes
based on both pathloss measurements and
distance were shown to be effective in
controlling interference. While this power
control may in some cases restrict the
performance on individual femtocells due
to restrictions in power, it is shown that the
overall system performance
(throughout and outage) for a given
population of UEs is significantly higher
with both femtocells and macrocells than
with macrocells alone.
Enhancements were also considered
whereby a femtocell operates with
restricted power, and/or on a restricted set
of frequency resources, only if victim UEs
are detected to be in the vicinity of the
femtocell.
The impact depends on the femto cell
density as well as the power control
algorithm used for the femtos. At 50
femtos/macro and femto cell transmit
power of 20dBm, the loss to macro cell
DL throughput can be as high as 40%
compared to the case of 10
femtos/macro and 0dBm transmit power
per femto cell. Distance based PC was
shown to reduce this loss to 30%.
At 4.8% femto penetration (12 active
femtos) with fixed tx power of 8dBm in a
macro layout with 1Km ISD, was shown
to increase the outage probability from
12.7% to 18.9%, which can be reduced to
9.8% if an adaptive tx power setting
algorithm was used.
A macro-UE aware power control
scheme was shown to provide similar
protection to macro layer as a path loss
based power controlled, but higher
throughput (around 25% for average
throughput and almost an order or
magnitude for cell edge throughput) to
the femto layer.
C2 Femtocell
Downlink
Interference to
the Macrocell
UE Receiver
(Control
Channel)
It is shown that the femtocell transmission
power can affect the coverage of the
macrocell. The macrocell deadzone for a
visiting MUE can be seriously varied with
respect to femtocell transmission power
and distance between macrocell and MUE.
Adaptive management of femtocell
transmission power is required to get
adequate femtocell coverage and maintain
control channel deadzone within a feasible
range.
The impact depends on the femtocell
transmission power and ISD of a
macrocell as well as type of MUEs. The
deadzone of a visiting MUE is varied from
19m to 50m for PDCCH and from 15m to
45m for SCH at a large radius
(ISD=1732m). For a passing MUE, if a
femto transmission power is managed
around -10dBm, the deadzone is very
small (1m or less than 1m) due to a
heavy wall.
.
D2 Femtocell
Uplink
Interference to
the Macrocell
NodeB
Receiver
Simulations suggest that a simple approach
of placing a cap on a femto UEs power can
be effective in controlling UL interference
to a macro eNB. The required cap on the
femto UE power could be made a function
of femto density and interference
conditions within the macro cell. This could
provide adequate performance at the
macrocell for higher femto densities, and
improved femtocell performance at lower
femtocell densities. An adaptive scheme
based on X2 signaling has been evaluated
for full buffer and bursty traffic models.
Some simplifications are possible to reduce
the complexity at the macro eNB due to X2.
The impact depends on the power
control algorithm used for femto and
macro UEs, as well as femto cell density.
At 50 femtos/macro and with fractional
power control and round robin
scheduling of macro UEs, losses of 20-
50% can be seen in macro layer average
throughput (depending on the power
control algorithm for femto UEs), with
even higher losses at the edge. The IoT
rises of 2-6 dB were observed at the
macro cell. An enhanced power control
and scheduling of macro UEs can
improve macro layer throughput. An
adaptive power control for femto UEs
can provide protection to the macro cell
uplink in all of the considered cases.
E Femtocell
Downlink
Interference to
Nearby
Femtocell UE
Receivers
For full buffer traffic, was shown that 4 to
12% of HUEs will be in outage without
interference management for 5 to 20%
penetration rate. OFDMA interference
management schemes (e.g., FFR) is shown
to eliminate the DL outage and to further
improve the system fairness. It was shown
that when FFR is enabled more than 20%
and 30% of users doubled their throughput
at 10 and 20% femto penetration rate,
respectively. For bursty traffic, a dynamic
resource coordination technique was
shown to remove outages of over 35% and
maintain a tail (90%-90%) latency of
around 10ms while significantly improving
the 5% throughput among users with full
buffer traffic.
Outages of 4% and 12% were observed
with 5% and 20% penetration rate with
all full buffer traffic. Higher outages (up
to 37%) were observed with mixed traffic
and higher penetration. FFR and
dynamic resource coordination was
shown to remove the outage and
significantly improve the performance
for both full buffer and bursty traffic
users.
.
F1 Femtocell
Uplink
Interference to
Nearby
Femtocell
Receivers
It is shown that in some corner cases the
uplink co-channel interference from an
aggressor UE located near to the victim
femtocell can significantly affect the
transmission quality. However system
simulation results suggest that the
probability of such occurrences is low such
that overall system performance is
acceptable. Furthermore at low
deployment densities allowing UEs to
transmit at full power gives the best
results, whereas at high deployment
densities it is advantageous to use uplink
power control such as fractional power
control to improve performance in those
cases where interference is most
significant.
For high deployment density, cell edge
throughput can be increased from 0.883
Mb/s to 1.188 Mb/s by employing
fractional power control.
G.3 Macrocell
Downlink
Interference to
the adjacent
channel
Femtocell UE
Receiver
(Traffic
Channel)
Link budget analysis shows that 99% of the
locations will see SINR of 8 dB or greater,
with 0 dBm femto AP power and ACS of 33
dB.
The impact is very small: at 0dBm femto
tx power, less than 1% of femto UEs will
see SINR of less than 8dB, assuming ACS
of 33dB and macrocell tx power of
60dBm and separation of 1732m.
G.4 Macrocell
Downlink
Interference to
the adjacent
channel
Femtocell UE
Receiver
(Control
Channel)
Link budget analysis shows that 99% of the
locations will see SINR of 8 dB or greater,
with 0 dBm femto AP power and ACIR of 33
dB. This is sufficient to decode the PDCCH
with coded at R=1/3 or 1 CCE. 99.8% of the
location will see SINR of -2 dB or greater,
which is enough to decode the PDCCH
when maximum coding is used.
The impact is very small: at 0dBm femto
tx power, less than 1% of femto UEs will
see SINR of less than 8dB, assuming ACS
of 33dB and macrocell tx power of
60dBm and separation of 1732m.
H1 Macrocell
Uplink
Interference to
the adjacent
channel
Femtocell
Receiver
Link-budget Monte Carlo simulations show
that for 95% of the locations the SIR is
greater than 14 dB. The dynamic range of
the femto receiver will need to be large to
handle the possible large signal from the
UEs due to the small path loss.
At ACIR of 33dB, 5% of the locations (for
macro and femto UEs) will see SIR of less
than 14dB. This is assuming that both
macro and femto UEs are transmitting at
full power.
.
17 Overall Conclusions and Future Work
17.1 Overall Conclusions
Link budget analysis or system level simulations were carried out for various interference
scenarios. Among the considered interference scenarios, Scenarios A1 (macro DL to femto),
B1 (macro UL to femto), G3 (macro DL to adjacent carrier femto, traffic), G4 (macro DL to adjacent
carrier femto, control), and H1 (macro UL to adjacent channel femto) were identified as benign
scenarios that do not seem to require new interference mitigation techniques. Note that the
conclusion for scenario A1 is based on the average throughput, and further study might be needed
to confirm the conclusion for other performance metrics such as tail of the user throughput CDF.
On the other hand, it was observed that interference in Scenarios C1 & C2 (femto DL to macro), D2
(femto UL to macro), E (femto DL to nearby femto), and F1 (femto UL to nearby femto) potentially
have significant impact on the performance of the victims, however it was found that interference
mitigation techniques can provide acceptable and robust performance.
Several interference mitigation techniques were proposed for the Scenarios in which significant
impact was observed, some of which are inline with the techniques discussed for WCDMA femto
deployments, such as adaptive femto power control based on distance or macro signal
measurements at the femto on the downlink. Enhancements such as macro UE-aware power
control of femto nodes have also been discussed. Having the flexibility of frequency-domain
resource management and coordination, LTE offers a new dimension for additional interference
mitigation techniques. This option has been explored in the interference mitigation techniques
proposed for Scenario D2 in the form of enhanced frequency selective scheduling, and for Scenario
E in the form of semi-static FFR and dynamic resource coordination. Another area which
distinguishes LTE from WCDMA is the potential use of the X2 interface in future 3GPP releases
between macros and femtos, or between femtos, which allows the interference mitigation
techniques to be adapted according to the needs of the victim. This has been explored for scenario
D2 where an adaptive power-capping algorithm is shown to provide adequate performance at the
macrocell for higher femto densities, and improved femtocell performance at lower femtocell
densities. Additionally, a number of methods to limit the complexity of such a scheme are also
discussed. The proposed interference mitigation techniques provide promising approaches for
enabling coexistence between femtocells within a femtocell deployment, and between femto cell
deployments and macro cellular networks for LTE.
.
17.2 Future Work
While analyses and work done to date suggests that there are hopeful, encouraging paths for
OFDMA femtocell development and deployment in the future as far as interference mitigation is
concerned, the Working Group also feels that such development and deployment will likely further
benefit from additional works in this are, which may include (but not limited to):
Further analyses on LTE-TDD (e.g. control channel aspects)
Study to understand the impact on individual users who may experience uncommon yet
extreme interference conditions and explore related mitigation techniques
Taking hybrid/open access models into account
Additional deployment scenarios (e.g. enterprise femto, metro hotspot femto)
Incorporating realistic service modeling (e.g., different traffic patterns for different
services)
Further analysis on low-band spectra.
Additional interference mitigation techniques and, where multiple techniques may be
applied to the same scenario, comparison of the effectiveness of those different
techniques for that scenario

.
18 References
[FFG08] Femto Forum Glossary, Femto Forum Working Group Document
[Law07] A. Law, Interference Management Evaluation Scenarios, v1.1, 07-04-2008
[OFDMA-EMD] OFDMA Interference Study: Evaluation Methodology Document, Femto Forum WG2
Document. Included in the Appendix.
[OFDMA-Scen] OFDMA Interference Study: Candidate Scenarios and Priorities, Femto Forum WG2
Document.
[picoChip 01] OFDMA Interference Management Study: Preliminary Results (scenarios B.1, D.2),
Femto Forum WG2 Document.
[R1-083190] New interference scenarios in LTE-A, Qualcomm Europe
[R1-083195] Range expansion for efficient support of heterogeneous networks, Qualcomm Europe
[R1-090355] Refinement of hot spot and femto deployment parameters, Qualcomm Europe
[R1-090368] Cooperative Silencing Hotzone DL Performance, Qualcomm Europe
[R1-090867] Signaling for spatial coordination in DL CoMP, Qualcomm Europe
[R1-090869] CoMP analysis in presence if bursty traffic, Qualcomm Europe
[R3-082442] Use of X2 to mitigate interference between HeNBs and macro eNBs, picoChip Designs.
[R4-071617] HNB and HNB-Macro Propagation Models, Qualcomm
[R4-081344] HNB and Macro Downlink Performance with Calibrated HNB Transmit Tower,
Qualcomm
[R4-091731] Simulation assumptions and parameters for FDD HeNB RF requirements, Alcatel-
Lucent, picoChip Designs, Vodafone.
[R4-091895] RSRQ measurement requirement in idle mode, Qualcomm
[R4-091908] Partial Bandwidth Control Channel Performance, Qualcomm Europe.
[UMTS-WP] Interference Management in UMTS Femtocells, Femto Forum White Paper
.
19 Contact Information
The Radio and Physical Layer working group (WG2) of the Femto Forum would be pleased to
respond to further queries on the aspects examined in this paper.

Email: info@femtoforum.org
Web: www.femtoforum.org
Postal: The Femto Forum
PO Box 23
GL11 5WA
UK
.
20 Appendix: OFDMA Interference Scenario Evaluation Methodology
A set of interference scenarios has been proposed for the Femto Forum OFDMA Interference study
[1]. System simulations are proposed in order to evaluate some of these scenarios. This section
proposes a methodology for these system simulations, and also proposes some parameters specific
to LTE.
20.1 Simulation Parameters
20.1.1 Macrocell Parameters
During 2007 operators within 3GPP requested that system simulations be performed to compare
LTE and HSPA performance. A set of macrocell assumptions for these simulations was specified by
the operators (see [2]). Note that based on discussions in femto forum, the following options are
provided:
inter-site distance is 500m or 1732m
exterior wall penetration loss is 10 or 20dB
multi-stream MIMO is optional (baseline is no MIMO).
The assumed macrocell parameters are as follows:
Cellular Layout Hexagonal grid, 3 sectors per site, reuse 1 (see
section 12).
Inter-site distance 500m or 1732m
Number sites 19 (=57 cells) or 7 (=21 cells) with optional
wrap-around (see section 12).
Carrier Frequency 2000 MHz
Distance-dependent path loss See section 6.2

Shadowing standard deviation 8 dB (see section 6.3)
Auto-correlation distance of Shadowing
(optional)
50 m (see section 6.3)
.
Shadowing
correlation
Between cells 0.5 (fixed, see section 6.3)
Between sectors 1.0 (see section 6.3)
Penetration Loss (assumes UEs are indoors) 10dB or 20dB (see section 6.2)
Antenna pattern [4] (horizontal)
(For 3-sector cell sites with fixed antenna
patterns)
see section 6.1
BS antenna gain after cable loss 14 dBi
BS noise figure 5 dB
UE Antenna gain 0 dBi
UE Noise Figure 9dB
Number of UE antennas 2 Rx, 1 Tx
Total BS TX power (Ptotal) 46dBm
UE power class 23 dBm (200 mW)
In order to keep the simulations simple it is not
necessary to model Maximum Power Reduction
(MPR) versus modulation scheme.
Inter-cell Interference Modelling Explicit modelling (all cells occupied by UEs), see
section 10 for more details)
Antenna Bore-sight points toward flat side
of cell (for 3-sector sites with fixed antenna
patterns)

Traffic model Full buffer initially with 10 UEs per sector.
Other traffic models (e.g. bursty traffic models)
should be looked at later, and that the
associated model(s) to use need to be selected
and documented.
UE distribution UEs dropped with uniform density within macro
coverage area, subject to a minimum separation
to macro and HeNBs.
.
The probability of a macro UE being indoors
should be a parameter rather than being fixed
at one, depending on the scenario being
investigated.
Minimum distance between UE and cell >= 35 meters
UE speeds of interest 3km/h
Fading model Ray based or correlation matrix based (see
section 6.3)
HARQ/Link to system mapping see section 7
DL Receiver Type MRC (single stream) or MMSE (multiple stream).
Multiple stream MIMO is optional.
Other MIMO modes should be looked at after
the initial studies.
UL Receiver Type MRC.
Table 20-1: Macrocell system assumptions
20.1.2 Femtocell Parameters
Femto Frequency Channel Either same frequency and same bandwidth as
macro layer, or adjacent channel and same
bandwidth as macro layer
Min separation UE to femto BS 20 cm [23]
Number Tx antennas femto BS 1 (baseline)
2x2 MIMO is optional
Other MIMO modes should be looked at after the initial
studies
Number Rx antennas femto BS 2
femto BS antenna gain 0 dBi or 5 dBi
Exterior wall penetration loss 10 or 20 dB (See section 6.2)
.
Interior path loss model See section 6.2
Interior to Exterior path loss model See section 6.2
Exterior path loss model femto BS to UE See section 6.2
Log-normal shadowing standard
deviation
4dB
10 dB in the simplified model in the Urban case
when no internal walls are modelled (See
section 6.2)
Shadowing auto-correlation distance for
femto (optional)
3m
Noise figure femto BS 8 dB (per [3])
Min/Max Tx power femto BS 0/20 dBm
Table 20-2: Femtocell system assumptions
20.1.2.1 LTE Specific Parameters
LTE-specific parameters are given in the table below:
Number RBs for PUCCH 4
Number of symbols for PDCCH 3
Table 20-3: LTE specific assumptions

.
20.2 Femtocell Deployment modelling
20.2.1 Suburban modelling
Femto cells are dropped within the macro coverage area with a random uniform distribution,
subject to minimum separation to the macro sites. The density of femtocells per macro cell will be a
variable in the simulations. Each of the dropped femtocells is assumed to be active i.e. there is at
least one active call. A sub-urban type deployment is considered where each femto cell is modelled
as a (2 dimensional) rectangular house. Within each house the femto base station and femto UEs
are randomly dropped within a specified distance of the centre point of the house.
All macro UEs are assumed to be indoors similar to the RAN1 assumptions [2]. A macro UE may be
within a femto house.
The details of the sub-urban model are as follows.
Parameter Value
House size 12x12m
House+Lot size (12f) x (12f) m with f chosen to give desired probability
of femto UE being outdoors when randomly dropped in
total area of house plus lot.
Probability femto UE outdoors 10%
Macro UEs allowed in femto house Yes, macro UEs are randomly dropped within macro
indoors coverage area, and a macro UE may happen to
be dropped within the 12x12m femto house
Allow Femto houses+lots to overlap No
Minimum separation UE to femto BS 20 cm [23]
Minimum separation femto BS to macro
BS
35m
Number of active femto UEs per femto-
cell
1
Distribution of femto houses Random uniform within macro coverage area, subject
to minimum separation to macro BS and non-
overlapping constraint.
.
Distribution of femto UE within femto
house
Random uniform, subject to minimum separation to
femto BS
Distribution of femto BS within femto
house
Random uniform
Table 20-4: Suburban femtocell modelling parameters

house
Max UE area
Max femto BS
area

Figure 20-1: Macro and Femto Geometry

.
20.3 Dense Femto Cell Deployment Modelling
20.3.1 Dual Stripe Model
In a dense-urban femtocell modelling, each block represents two stripes of apartments, each stripe
has 2 by N apartments (N is 10 in the example illustrated in Figure 20-2). Each apartment is of size
10m X 10m. There is a street between the two stripes of apartments, with width of 10m. Each
femtocell block is of size 10(N+2)m X 70m. This is to make sure that the femtocells from different
femtocell blocks are not too close to each other. In each macro cell sector, one or several femtocell
blocks are randomly dropped. It is assumed that each femtocell blocks are not overlapping with
each other.
Each femtocell block has L floors, L is chosen randomly (L could be a number between 1 and 10). If
more than one femtocell blocks are dropped, each femtocell block can have different number of
floors.
To simulate the realistic case that an apartment may not have a femtocell, we use a parameter
named deployment ratio to determine whether an apartment is deployed with a femtocell or
not. If the deployment ratio is 0.1, in our example, it means that on average, each floor has 8
(=0.2*40) femtocells and each block has 8L femtocells. The occupation ratio can vary from 0.0 to
1.0. Another parameter called activation ratio is defined as the percentage of active femtocells. If
a femtocell is active, it will transmit with suitable power at the traffic channel. Otherwise, it will
only transmit the control channels. Activation ratio can be from 0 to 100%. A femto cell block is
illustrated in Figure 20-2. An example parameter table is listed in
Table 20-5.

N (number of cells per row ) 10
M (number of blocks per sector) 1
R (deployment ratio ) 0.2

Table 20-5 Urban-dense femtocell modelling parameters
.
10 m
10 m
10 m
10 m
10 m

Figure 20-2. A femtocell block

Macro UEs are dropped uniformly and randomly throughout the sector. Its possible that some
Macro UEs will be dropped into the femtocell area. It is assumed that there is one femto UE per
femto cell, which is dropped randomly in the active femtocell. The femtocell BS is also randomly
placed in each femtocell.
20.3.1.1 5x5 Grid Model
An alternative simple femto cluster model has been defined as follows. We consider a single floor
building with 25 apartments. The apartments are 10mx10m and are placed next to each other on a
5x5 grid on each floor. In addition, we assume that with probability p, there is a femto in each
apartment. This probability represents the density of femto deployment. For the apartments that
have a femto, the femto and UE are dropped randomly and uniformly in the apartment with a
minimum separation of 20cm [23]

.
20.4 Channel Models
20.4.1 Antenna Patterns
The azimuth antenna pattern of the macro is modelled as:

( )
(
(
|
|
.
|
\
|
=
m
dB
A A , 12 min
2
3

where
dB 3
= 70 degrees, A
m
= 20 dB.
The azimuth antenna patterns for UEs and femto BSs are assumed to be omnidirectional.
20.4.2 Pathloss Models
Pathloss models for the suburban and urban deployments are described in the following. The path
loss models shall apply when the Tx-Rx separation is larger than or equal to 1m, otherwise the
following formula shall be used with no shadowing, which gives similar results to the 2 GHz
measurements in [23]:
PL (dB) = 38.46 + 20 log10R + 0.7d2D,indoor

.
20.4.2.1 suburban deployment
Cases Path Loss (dB)
UE to
macro BS
(1) UE is outside
PL (dB) =128.1 + 37.6log
10
R, R in km [18]
= 15.3 + 37.6log
10
R, R in m
(2) UE is inside a
house
PL (dB) =15.3 + 37.6log
10
R + L
ow
, R in m [19]
UE to
femto BS
(3) UE is inside
the same house
as femto BS
PL (dB) = 38.46 + 20 log
10
R + 0.7d
2D,indoor
+ 18.3 n
((n+2)/(n+1)-0.46)
[19]
R and d
2D,indoor
are in m
n is the number of penetrated floors
In case of a single-floor house, the last term is not needed
(4) UE is outside
PL (dB) = max(15.3 + 37.6log
10
R, 38.46 + 20log
10
R) + 0.7d
2D,indoor

+ 18.3 n
((n+2)/(n+1)-0.46)
+ L
ow

R and d
2D,indoor
are in m
(5) UE is inside a
different house
PL(dB) = max(15.3 + 37.6log
10
R, 38.46 + 20log
10
R) + 0.7d
2D,indoor

+ 18.3 n
((n+2)/(n+1)-0.46)
+

L
ow
,
1
+ L
ow,2

R and d
2D,indoor
are in m

Table 20-6 Path loss models for suburban deployment
R is the Tx-Rx separation
L
ow
is the penetration loss of an outdoor wall, which is 10dB or 20dB.
In Case (3), the path loss is modeled by free space loss, penetration loss due to
internal walls and floors. The loss due to internal walls is modeled as a log-linear
value, equal to 0.7dB/m.
In Case (4), the path loss modeling takes account of [19] and [20]. d
2D,indoor
is the
distance inside the house.
In Case (5), d
2D,indoor
is the total distance inside the two houses. L
ow
,
1
and L
ow,2
are the
penetration losses of outdoor walls for the two houses.

.
20.4.2.2 Urban deployment
Cases Path Loss (dB)
UE to
macro BS
(1) UE is outside PL (dB) =15.3 + 37.6log
10
R, R in m
(2) UE is inside an apt PL (dB) =15.3 + 37.6log
10
R + L
ow
, R in m
UE to
femto BS
(3) Dual-stripe model:
UE is inside the same
apt stripe as femto BS

PL (dB) = 38.46 + 20 log
10
R + 0.7d
2D,indoor
+ 18.3 n
((n+2)/(n+1)-0.46)
+
q*L
iw

R and d
2D,indoor
are in m
n is the number of penetrated floors
q is the number of walls separating apartments between UE and
femto BS
In case of a single-floor apt, the last term is not needed
UE is outside the apt
stripe
PL (dB) = max(15.3 + 37.6log
10
R, 38.46 + 20log
10
R) + 0.7d
2D,indoor

+ 18.3 n
((n+2)/(n+1)-0.46)
+ q*L
iw
+ L
ow

R and d
2D,indoor
are in m
femto BS
UE is inside a different
apt stripe
PL(dB) = max(15.3 + 37.6log
10
R, 38.46 + 20log
10
R) + 0.7d
2D,indoor

+ 18.3 n
((n+2)/(n+1)-0.46)
+

q*L
iw
+ L
ow
,
1
+ L
ow,2

R and d
2D,indoor
are in m
femto BS
(6) Dual-stripe model
or 5x5 Grid Model: UE
is within or outside the
apartment block
PL(dB) = 127+30log10(R/1000)
R in m
This is an alternative simplified model based on the LTE-A
evaluation methodology which avoids modelling any walls.

Table 20-7 Path loss models for urban (dense apartment) deployment
L
iw
is the penetration loss of the wall separating apartments, which is 5dB.
The term 0.7d
2D,indoor
takes account of penetration loss due to walls inside an apartment.

.
20.4.3 Shadowing Models
Log-normal shadowing applies to all links. For links between a femto BS and a UE served by this BS,
the standard deviation is assumed to be 4dB (10dB in the simplified model in the Urban case when
no internal walls are modelled). Otherwise for all other links (including interference links) the
standard deviation is 8dB.
Correlated shadowing is applied. The baseline models the shadowing correlation from one UE to
multiple BS, and assumes no shadowing correlation from one BS to multiple UEs no matter how
close the UEs are located.
Optionally shadowing correlation from one BS to multiple UEs can be modelled as a function of
separation between UEs. In this case the auto-correlation distance for macro BS is assumed to be
higher (50m) than for femto BS (3m).
Note that different methods for generating correlated shadowing exist. See [21] for example.
20.4.4 Fast Fading Models
Depending on the interference analysis methodology, fast fading may or may not be modelled. For
modelling in line with 3GPP RAN4 coexistence study assumptions, fast fading channel is not
modelled between the BS and UE. However more detailed models may include fast fading.
Both time and angular dispersion properties can be modelled for the fast fading. Two approaches
to modelling the fast fading are defined: one is a ray-based model, and the other is an approach
based on the use of pre-computed correlation matrices. The models are as defined for 802.16m
evaluations in [5].
In [5], parameters are defined for various scenarios (urban macrocell, urban microcell, etc.). The
mapping of scenarios in [5] to links in the present OFDMA interference study is as follows:
macro BS to outdoor UE Urban Macrocell ([5] section 3.2.5.1)
macro BS to indoor UE - Outdoor to indoor ([5] section 3.2.5.6)
femto BS to indoor UE Indoor hotspot ([5] section 3.2.5.5)
femto BS to outdoor UE - Outdoor to indoor ([5] section 3.2.5.6)

.
20.5 Link-to-System Mapping
Three alternative mapping approaches are described.
20.5.1 Link to System Mapping (RAN1 approach)
Link-to-system mapping (a.k.a PHY abstraction) is an important building block in a system-level
simulator. The objective of link-to-system mapping is to predict the packet error rate for a given
channel realization across the tones which are used to transmit the coded packet. The general
approach can be described as follows:
Link level simulator provides a set of AWGN curves for various MCS (modulation and code
scheme) as well as different HARQ terminations if necessary.
System level simulator calculates post-processing SINR per data symbol and generates a
packet effective SINR.
System level simulator maps the packet effective SINR to the corresponding AWGN curve
and finds the packet error rate (PER).
System level simulator flips a coin and decides whether the packet is in error or not
according to the PER.
The procedure of compressing SINR per data symbol of multiple data symbols to a single packet
effective SINR value is called effective SINR mapping (ESM). The general method of ESM can be
described as ( )
1
1
1
N
eff n
n
SINR SINR
N
=

=
`
)
, where N is the number of data symbols in the coded

packet. The widely used ESM methods include EESM (exponential effective SINR mapping) and Q-
factor. EESM has been used in 3GPP while Q-factor in 3GPP2.
In EESM, ) ( is derived based on the Chernoff bound for the error probability and
1
1
ln exp
N
n
eff
n
SINR
SINR
N
=
| | | |
=
| |
\ . \ .
. Appropriate values for various MCS levels, channel

models, coded packet sizes, MIMO schemes, receiver structures, channel estimations, HARQ
terminations etc. can be found via link level simulations. A set of reference values for WiMAX
system is provided in [7]. More details of EESM can be found in [5].
In Q-factor, ) ( is based on constrained capacity and
|
|
.
|
\
|
|
|
.
|
\
|
=

=
N
n
n
eff
Q
SINR
C
N
C Q SINR
1
mod
1
mod
1

.
where ( )
mod
C is the Gaussian channel capacity with input restricted to the specified modulation.
Similar to EESM, Q value could depend on MCS levels etc.. Details of Q-factor can be found in [8] [9
Appendix T] [10].
To match link level behaviour, backoff factors may be applied in addition to adjustments of or Q
value.
LTE UL employs DFT-SOFDMA in which each data symbol is effectively transmitted over multiple
subcarriers. Note that the SINRs for data symbols within one DFT-SOFDMA symbol are the same.
The packet effective SINR can be calculated as
|
|
.
|
\
|
)
`
=

=
symb
N
n
n
symb
eff
SINR
N
SINR
1
exp
1
ln
, where
symb
N is the number of DFT-SOFDMA symbols for the packet. The SINR of data symbols within nth
DFT-SOFDMA symbol can be calculated as
1
1
0 ,
,
1
1
1
1
=
(
(
(
(
+
=
M
k k n
k n
n
SINR
SINR
M
SINR , where
k n
SINR
,
is the
SINR of the kth subcarrier in the nth DFT-SOFDMA symbol. More details of link-to-system mapping
for DFT-SOFDMA can be found in [11].
20.5.2 Link to System Mapping (per sub-carrier capacity approach)
An alternative to the ESM approach has been used in some 3GPP evaluations which is based on
estimating throughput directly from channel capacity, for example [14] and [15]. For OFDMA
systems a measure of channel capacity is calculated for each assigned (data bearing) subcarrier.
This is then summed over all assigned subcarriers. The ideal Shannon channel capacity is backed-
off in order to take account of non-ideal receivers structures and also to take HARQ overhead into
account. In this way the system throughput can be estimated without the need to simulate MCS
selection and HARQ.
The per-subcarrier channel capacity with non-ideal receiver and including HARQ overhead is
determined as:
) 1 ( log
2
+ = C
where and are parameters (between 0 and 1) selected to match the link level performance (for
example for LTE these could be selected to match the UE and BS (post-HARQ) performance
requirements defined in 36.101 and 36.104 for the uplink in initial study suggests that values of
=0.9 and =0.5 would be appropriat e). is the (linear) per sub-carrier SNIR which is determined as
.
a function of assumed receiver type (e.g. MRC for single stream or MMSE for MIMO). Note that
with multiple streams is t he SNIRper stream at the output of the assumed receiver type.
The capacity is calculated for each data bearing sub-carrier and symbol, i.e. for each data-bearing
modulation symbol. The scheduler in the simulation may assign a resource to a UE taking into
account channel conditions, available resources and available power, and the resulting capacity
determines the achieved throughput of this assignment. Note that with a limitation in Tx power,
the capacity per sub-carrier will be a function of number of sub-carriers used for a transmission, so
the scheduler needs to select the number of assigned sub-carriers appropriately.
20.5.3 Link to System Mapping (RAN4 approach)
It is known that LTE was aimed at supporting Packet Switched services where throughput is an
important metric. Since the static simulation methodology does not capture the timing information,
it was agreed in 3GPP RAN4 to map the obtained SNR at each snapshot to the throughput. The
details on the mapping are given in this subsection.
A mapping method called attenuated and truncated Shannon bound has been used to obtain the
throughput of a radio channel with link adaptation for a given SNR. Specifically, given SNR, the
throughput is calculated according to the following equation:
MAX MAX
MAX min
MIN
SNIR SNIR for Thr Thr
SNIR SNIR SNIR for .S(SNIR) Thr
SNIR SNIR for 0 Thr
/ , ,
> =
< < =
< =
= Hz bps Thr Throughput (6)
where S(SNIR) is the Shannon bound, i.e., S(SNIR) = log
2
(1+SNIR) bps/Hz, is the attenuation
factor representing implementation losses, SNR
MIN
is the minimum SNIR of the codeset in dB and
Thr
MAX
is the maximum throughput of the codeset in bps/Hz, and SNIR
MAX
is the SNIR in dB at which
the maximum throughput is reached. Note that for different modem implementations and link
conditions, the parameters , SNR
MIN
and THR
MAX
can be chosen accordingly. The parameters
proposed in Table 20-8 represent a baseline case, which assumes:
1:2 antenna configurations
Typical Urban fast fading channel model (10kmph DL, 3kmph UL)
Link Adaptation (see Table 1 for details of highest and lowest rate codes)
Channel prediction
HARQ

.
Table 20-8. Parameters describing baseline Link Level performance for LTE static
system simulations
Parameter DL UL Notes
, attenuation 0.6 0.4 Represents implementation losses
SNIRMIN, dB -10 -10 Based on QPSK, 1/8 rate (DL) & 1/5 rate (UL)
ThruMAX, bps/Hz 4.4 2.0 Based on 64QAM 4/5 (DL) & 16QAM 3/4 (UL)

Figure 20-3 plotted the mapping curve from SNR to throughput.
0
1
2
3
4
5
-15 -10 -5 0 5 10 15 20 25
SNIR, dB
T
h
r
o
u
g
h
p
u
t
,

b
p
s
/
H
z
Shannon
DL
UL

Figure 20-3. Throughput vs. SNR for Baseline LTE System simulations

.
20.6 Scheduler
The baseline scheduler for initial simulations is the round robin scheduler.
For final evaluations, more complex proportional fair + frequency domain type scheduling schemes
may be assumed.
20.7 Power Control
20.8 Downlink
20.8.1 Femto Downlink Power Control
The baseline approach does not take account of interference to the macro layer or within the
femto layer, and can be either:
a fixed power (e.g. 20 dBm)
a variable power based on the coverage requirements of the UEs within the femto cell
(assuming no interference), up to a maximum power (e.g. 20dBm).
More sophisticated approaches in order to protect the macro may also be modelled taking account
of interference to the macro layer and/or within the femto layer. For example an approach based
on the WCDMA approach described in [22] section 7.4.4 could be adopted (perhaps on a per-
subcarrier or block of sub-carriers basis). This describes how the DL power setting is a function of
femto cell coverage, macrocell interference (as measured by the UE) and other-femto interference.
Given a suitable DL operating power range, the femto BS could set its total Tx power in the
following way:
Such that the SNIR at the expected edge of coverage of the femto is of sufficient quality to
provide adequate coverage/performance for the femto.
Such that the macro layer SNIR at the expected edge of coverage of the femto is not
severely degraded.
20.8.1.1 Macro Layer (LTE) Downlink Power Control
In the 3GPP RAN4 coexistence study assumption, no power control is assumed in the DL, i.e., the BS
transmits equal power in all sub-carriers. The same assumption is made here.

.
20.8.1.2 Uplink
20.8.1.2.1 Femto UE Uplink Power Control
The baseline approach does take account of interference to the macro layer or within the femto
layer, and can be either:
a fixed UE power (for initial simulations - not likely a realistic option in practice)
a variable power based on the desired femto coverage area (assuming no interference), up
to the maximum UE power (e.g. based on fractional power control as defined for the macro
uplink).
More sophisticated approaches aimed at protecting the macro may be evaluated taking account of
interference to the macro layer and/or within the femto layer. For example a similar approach to
that described in [22] section 7.5.1 for WCDMA could be adopted on a per sub-carrier or block of
sub-carriers basis, where the pathloss from the vicinity of the femto BS to the neighbouring macro
BSs is estimated from measurements, and based on this and other related parameters, the femto
BS can then determine a maximum allowed UE transmit power, such that the noise rise
experienced at the neighbour cells is constrained to be within an acceptable limit.
20.8.1.2.2 Macro UE (LTE) Uplink Power Control
Fractional power control [16] may be used for each UE in the UL simulation. The transmit power is
controlled to compensate a fraction of the path loss as per the following equation:
(
(
|
|
.
|
\
|
=
ile x
t
PL
PL
R P P , max , 1 min
min max (7)
where P
max
is the maximum transmit power, R
min
is the minimum power reduction ratio to prevent
UEs with good channels to transmit at very low power level, PL is the path loss including shadowing,
antenna and MCL for the UE and PL
x-ile
is the x-percentile path loss value. If x set to 5, then
statistically 5 percent of UEs with bad channels will transmit at P
max
. Finally, is the balancing
factor for UEs with bad channel and UEs with good channel.
Next, we discuss how to select power control parameters for simulations and present a general rule
governing the selection of PC parameters. Since P
max
is agreed to be 24dBm and R
min
can be derived
as P
max
/ P
min
, the parameters we are left with are PL
x-ile
and . It can be easily seen from Equation
(7) that for most UEs, their power is likely to be adjusted as
.

PL
PL
P
PL
PL
P P
ile x ile x
t

|
|
.
|
\
|
=
|
|
.
|
\
|
=

1
max max (8)

As a result of discussions in 3GPP RAN4, it was agreed to use the following two sets of power
control parameters for simulation study as listed in Table 20-9.
Table 20-9. Power control algorithm parameter
Parameter set Gamma () PLx-ile
10 MHz bandwidth 5 MHz bandwidth
Set 1 1 112 115
Set 2 0,8 129 133

It is worth mentioning that parameter Set 1 will lead to higher LTE single system performance in
terms of throughput because it enables higher user transmit power, where parameter Set 2 will
target better coexistence by tuning down user transmit power to some extent without sacrificing
LTE single system significantly. Note the value of PLxile for 10MHz bandwidth is 3dB or 4dB smaller
than that for 5MHz bandwidth. The reason is to compensate the higher noise floor in the 10MHz
bandwidth case where one user uses 16 RBs as compared to the 5MHz bandwidth case where one
user uses 8 RBs.
The Fractional Power Control is one example of a power control scheme. Other power control
schemes may be assumed.
20.8.2 Interference Modelling
Full modelling of frequency selective channel for every mobile and base station pair can increase
simulation complexity significantly. To reduce complexity, two strongest interferers (either macro
BS or femto BS) to the mobile can be determined based on pathloss and shadowing. The fast fading
channels between the mobile and the two strongest interferers can be modelled as multipath (i.e.,
frequency selective) as Section 20.4.4. The channels between the mobile and other weaker
interferers will be modelled as one-path Rayleigh fading (i.e., frequency flat).
In case of MIMO configuration, the MIMO channel correlations should be fully modelled between
the mobile and the two strongest interferers. For other weaker interferers, the MIMO channel can
be modelled as multiple independent fading processes.
.
To further simplify the interference modelling, frequency flat channel model can be assumed
between all the BSs and UEs. And single antennas can be assumed in all the BSs and UEs.
Furthermore in the case of the RAN4 based approach to link to system mapping, it is possible to
model no fast fading and no frequency selective fading on the interference links.
20.8.3 Traffic Models
A full buffer traffic model shall be the baseline for calibration.
A bursy traffic model is also defined. This is taken from [24]. This is a simple model with a fixed
packet size and with an exponential distribution for packet inter-arrival times. The model further
assumes that a new packet is not generated until the previous packet has been successfully
delivered, which ensures that the buffer occupancy is stable. The inter-arrival time is therefore
defined as the time between the successful delivery of a packet and the generation of the next
packet.
The packet size and mean inter-arrival time are both parameters. To approximate WEB browsing
traffic, these could be set to 16.25KB and 200ms respectively.

.
20.9 Simulation Procedure and Flow
20.9.1 Cell layout
The cell layout is wrap-around to form a toroidal surface in order to remove edge effect. Totally 57
cells with 19 BSs or 21 cells with 7 BSs will be simulated. The cell layout in the 19 BS case is shown
in the figure below, where BSs are numbered from 1 to 19.

Figure 20-4. LTE Cell layout
To illustrate the wrap-around cell structure, 7 sets of 57-cell structure are shown in the figure
below, where the original cell set is located in the center and the other 6 sets are evenly
surrounding the center set. Each cell structure has 19 BSs.

.

Figure 20-5. Illustration of wrap-around for the center cell set

.
The coordinate system is defined as in the figure and the cell radium is assumed to be r. Then the
distance from a UE at (x,y) to a BS at (a,b) can be obtained by choosing the minimum of the
following:
Dist( (x,y), (a,b))
Dist( (x,y), (a+10.5r,b+4.5 sqrt(3) r))
Dist( (x,y), (a-10.5r,b-4.5 sqrt(3) r))
Dist( (x,y), (a+1.5r, b-7.5 sqrt(3) r))
Dist( (x,y), (a-1.5r,b+7.5 sqrt(3) r))
Dist( (x,y), (a-12r,b+3 sqrt(3) r))
Dist( (x,y), (a+12r,b-3 sqrt(3) r))
Here Dist(A, B) denotes the function that calculates the distance between point A and point B.
The distance between UE and BS is used to compute the path loss and antenna gain of a UE at (x,y)
to a BS at (a,b).
For DL simulation, either centre-cell or wrap around approach can be used. For UL simulation,
wrap-around approach is recommended to appropriately model interference. Details of centre-cell
approach can be found in [9] while details of wrap-around in [5, 9]. To reduce the complexity when
simulating both macrocell and femtocell, it is recommended to use 7-cell/21-sector wrap around
instead of the 19-cell/57-sector wrap around typically employed in macrocell system level
simulation.
20.9.2 3GPP RAN4 based Monte-Carlo Static Simulation Methodology
For interference analysis and coexistence studies, typically 3GPP RAN4 based Monte-Carlo Static
simulation is used to derive the impact due to other system interference.
This simulation methodology is based upon a Monte Carlo Technique, i.e. a statistical technique
that functions by considering many independent instants in time. For each instant, or simulation
trial, a scenario is built up using a number of different random variables. In the case of a
communication system, the random variables will be mainly the positions of the users in the
system, the channels and services required by the users. If a sufficient number of trials are
considered then the probability of a certain event occurring (such as for example the probability
that a user is interfered) can be calculated with a high level of accuracy.
The overall architecture of the static methodology is illustrated in Figure 20-6 . At the beginning of
each snapshot, Macro cell users are uniformly distributed in the system area and then attached to
.
specific base stations according to path loss and log-normal fading. Femto cells are also deployed
with femto cell BS and UE dropped uniformly into each cell.
Next, the Macro cell users will be scheduled according to some scheduling algorithms such as
Round-Robin or maximum C/I. For maximum C/I, the CQI (Channel Quality Indicator) for each user
is calculated, which also includes the inter-system interference. After scheduling, all the users will
be assigned a priority. In other words, an ordered list of users will be generated. Then, each user
will be allocated a certain amount of resource according to their C/I and QoS requirement. Note
that it might be the case that not all the users can be allocated resource. For the time being, we can
use Round-Robin scheduling, which is used in 3GPP LTE coexistence study.
For the moment, we can assume one femto cell users per each femto cell. In this case, if the femto
cell is active, the femto cell user is always scheduled with adequate resources assigned.
After resource allocation, transmission power control could be exercised for both macro cell and
femto cell, depending on the technology of interest. Currently, it is assumed that for Macro LTE
downlink, all the sub-carriers are allocated fixed transmission power; therefore no power control is
needed. For LTE uplink, power control is necessary in order to ensure the performance of users that
are located at cell edge or experience severe fading. For femto cell downlink, the femto cell BS
downlink power needs to be adjusted depending on the interference mitigation techniques.

.
Simulation initialization
* Cellular pattern
* Masks
Generation of a new
snapshot
* user distribution
* Femto cell deployment
Scheduling
Power control
Calculate SIR and
obtain throughput/
capacity
#of snapshots >
predefined value
Collecting statistics
* Throughput/capacity
Yes
No
#of snapshots ++

Figure 20-6. Monte-Carlo simulation flowchart
Finally, the C/I for each user will be calculated. Depending on the system evaluation criteria, the
corresponding values will be calculated. For example, if the system throughput is considered, for
LTE the actual data rate of each user can be obtained through SNR-to-throughput mapping based
on the obtained C/I. If HARQ is used, say for LTE, its effect is taken into account in the mapping.
The detailed general assumptions for the LTE system are described in the appendix.
.
20.10 Performance Metrics
The following performance metrics shall be obtained:
Macro layer average cell throughput
Macro layer 5 percentile user throughput (gives an indication of Macro layer coverage)
Macro layer average Interference over Thermal (UL)
Femto layer average cell throughput
Femto layer 5 percentile user throughput
Femto layer average Interference over Thermal (UL)

.
20.11 References
[1] Femtoforum WG2 OFDMA Interference Study Candidate Scenarios and Priorities, Sept 2008
[2] 3GPP RAN1 document R1-070674
[3] Femtoforum WG2 Recommended Simulation Parameters, FF_WG2_(08)_001_v1.0
[4] 3GPP2, C30-20080616-016R1 Femtocell Interference Management Methods and
Performance Results, Qualcomm Inc.
[5] IEEE 802.16m Evaluation Methodology Document (EMD), IEEE 802.16m-08/004r3
[6] 3GPP TR 36.942 V8.0.0
[7] WiMAX System Evaluation Methodology, Version 2.1, WiMAX Forum, July, 2008
http://www.wimaxforum.org/documents/documents/documents/WiMAX_System_Eval
uation_Methodology_V2_1.pdf
[8] 3GPP2 C30-20040920-017, Enhanced BCMCS, accuracy of PER predictions, Qualcomm, Sept.
2004
[9] 3GPP2, cdma 2000 Evaluation Methodology
[10] 3GPP2, C30-20071203-024, Channel estimation backoff for UMB calibration, Qualcomm,
Dec. 2007
[11] 3GPP R1-051335, Simulation methodology for EUTRA UL: IFDMA and DFT-Spread-OFDMA,
Motorola, Nov 2005
[12] Path Loss Measurements for a Non-Line-of-Sight Mobile-to-Mobile Environment, J. Turkka,
M. Renfors, Proc. 8th international conference on Intelligent Transport System
Telecommunications (ITST 2008), Phuket, Thailand, October 2008.
[13] Statistical Peer-to-Peer Channel Models for Outdoor Urban Environments at 2GHz and 5GHz
Zhenyu Wang, Eustace K. Tameh and Andrew R. Nix, IEEE VTC 2004 Fall Conference.
[14] 3GPP TR 36.942 V8.0.0, Evolved Universal Terrestrial Radio Access (E-UTRA); Radio
Frequency (RF) system scenarios
[15] 3GPP R4-081876 HSDPA Macro and Home NodeB downlink performance in co-channel
deployment, Motorola
[16] 3GPP RAN1 R1-060026, Power Control and FDM Resource Allocation for LTE Uplink and
TP.
.
[17] 3GPP RAN4 R4-071941, Simulation results for Home NodeB to Home NodeB downlink co-
existence considering the impact of HNB HS utilization
[18] 3GPP TR 25.814, Physical layer aspects for evolved UTRA, v7.1.0, 2006
[19] R4-071263, System simulation results for Home NodeB interference scenario #2, Ericsson,
RAN4#44, Aug. 2007
[20] R4-071617, HNB and HNB-macro propagation models, Qualcomm, RAN4 #44-bis, Oct 2007
[21] Evaluation Methodology for Femto Interference Scenarios, ALU contribution to OFDMA
Interference Evaluation Methodology, Dec 2008.
[22] 3GPP TR 25.967
[23] R4-071151, Antenna Coupling Loss Measurement in Indoor Environment, Orange.
[24] 3GPP TR 36.814

012 OFDMA Interference Management

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

012 OFDMA Interference Management

Uploaded by

Copyright:

Available Formats

www.femtoforum.

, where N is the number of data symbols in the coded

. Appropriate values for various MCS levels, channel

You might also like