lte/lte-a interference coordination for femtocells

Copyright © 2012 DOCOMO Communications Laboratories Europe GmbH Infrastructure Research Group

LTE/LTE‐A Interference Coordination for Femtocells

BeFEMTO Winter SchoolFebruary 6‐10 2012Zubin BharuchaDOCOMO Euro‐LabsMunich, Germany

Copyright © 2012DOCOMO Communications Laboratories Europe GmbH

Acknowledgements: Serkan Uygungelen; Nobuhiko Miki (NTT DOCOMO)

Copyright © 2012 DOCOMO Communications Laboratories Europe GmbH Infrastructure Research Group 22

In a nutshell

• Part 1: Refresh your memory!– LTE and LTE‐A– The road to the future– An overview of ICIC techniques

• Part 2: Femto‐macro interference– Relevant details of the LTE air interface– Performance comparison of existing techniques– Introduction of a novel technique to protect non‐CSG users

• Part 3: Femto‐femto interference– Network „densification“ and its effects– Centralized interference mitigation– Distributed interference mitigation

• Conclusion


Part 1: Know your LTE‐A (B,Cs)


What’s so great about LTE?

• LTE– Long‐term evolution of 3G using 3G

spectrum– Smooth introduction of 4G

• LTE‐Advanced– Evolution of LTE: Targets

achievement of sufficiently highersystem performance than that forLTE• Bandwidth: 100 MHz• Peak throughput: 1 Gbps

– Backward compatible with LTE to enable continuous enhancement and deployment

– Meet or exceed IMT‐Advanced requirements within the ITU‐R time plan

5~20 MHz bandwidth

~100 MHz bandwidth

System performance

2000’s 2010’s

HSUPA

HSDPA

WCDMA Release 99

LTE

Smooth introduction of 4G

Long‐term evolution of 3G

LTE‐Advanced


The old and the new

• LTE‐Advanced shall be deployed as an evolution of LTE Rel. 8 with new bands available

• LTE‐Advanced shall be backwards compatible with LTE Rel. 8 Smooth and flexible system migration from LTE Rel. 8 to LTE‐Advanced

An LTE‐A UE works in an LTE cell An LTE UE works in an LTE‐A cell

• LTE‐Advanced contains all features of LTE Rel. 8&9 and additional features for further evolution


LTE Rel. 8 LTE‐Advanced

Peak data rateDL 300 Mbps 1 GbpsUL 75 Mbps 500 Mbps

Peak spectrum efficiency [bps/Hz]

DL 15 30UL 3.75 15

* Target peak data rate of 1 Gbps for nomadic/local areas is specified in Circular Letter (CL)*1 See TR25.912 (Case 1 scenario) *2 See TR36.913 (Case 1 scenario) *3 See ITU‐R M.2135 (Base Coverage Urban scenario)

Target Performance for LTE‐Advanced

Cell‐edge user throughput [bps/Hz/cell

/user]

DL

2‐by‐2 0.05 0.07

4‐by‐2 0.06 0.09

4‐by‐4 0.08 0.12

UL1‐by‐2 0.024 0.04

2‐by‐4 – 0.07

Ant. Config. LTE Rel. 8*1 LTE‐Advanced*2

Capacity[bps/Hz/cell]

DL

2‐by‐2 1.69 2.4

4‐by‐2 1.87 2.6

4‐by‐4 2.67 3.7

UL1‐by‐2 0.74 1.2

2‐by‐4 – 2.0x 1.4‐1.7


What’s new in LTE‐A?

• Wider bandwidth (carrier aggregation)– Improves peak data rate and spectrum flexibility– Meets ITU‐R requirements for bandwidth (>=40

MHz)– Spectrum/carrier aggregation based on

component carrier (CC) concept to maintain backward compatibility and allow smooth network migration

• Advanced MIMO techniques (covered yesterday)– Improves peak data rate and cell/cell‐edge

spectrum efficiency– Meets ITU‐R requirements for DL cell spectrum

efficiency– SU‐MIMO with up to 8‐layers for DL and 4‐layers

for UL– MU‐MIMO with enhanced CSI feedback


What’s new in LTE‐A?

• Enhanced inter‐cell interference coordination (eICIC)– Improves cell‐edge user throughput, coverage, and

deployment flexibility– Interference coordination for layered cell deployment with

different transmit power levels– Carrier aggregation can be used for frequency domain

coordination– Time domain coordination and power control are also to be

introduced• Relaying

– Improves coverage and cost effective deployment– Type 1 relay node which can be seen as a Rel. 8 eNB from a

Release 8 LTE terminal• Coordinated multipoint (CoMP) transmission and reception

– Scope is limited to intra‐eNB CoMP (implementation issue)– LTE Self Optimizing Network (SON) enhancements– HNB and HeNB mobility enhancements


HeteroGenius Networks

Characteristics• Wired backhaul• Closed access• User‐deployedMajor Issues• Mitigating femto‐to‐macrointerference• Mitigating interferencebetween nearby femto‐cells

Characteristics• Wireless backhaul• Open access• Operator‐deployedMajor Issues• Effective backhaul design• Mitigating relay to macro‐cell interference

Characteristics• Wired backhaul• Open access• Operator‐deployedMajor Issues• Effectively offloadingtraffic from macro‐cell• Mitigating interferencecaused to macro‐cellusers

Motivation•4G networks will be characterized by a high‐densitydeployment of low‐power nodes• It is essential for these nodes to operate without negativelyaffecting the overall performance


Why do we need interference management with femtocell deployment?

Significant femto interference for nearby macro UEs!


Overview of ICIC in LTE/LTE‐A

• LTE (Rel‐8/9)– Only one CC is available– Make do with what you have and devise interference management

techniques assuming that macro and femtocells use the same CC– Frequency‐domain ICIC ?– Time‐domain ICIC within one CC?

• LTE‐Advanced (Rel‐10/11)– Multiple CCs available in the system– Frequency‐domain ICIC over multiple CCs is possible– Time‐domain ICIC within one CC is also possible– Much greater flexibility for interference management


Sharing is caring

• Fractional frequency reuse (FFR) improves the throughput for UEs close to the cell boarder– Protecting UEs close to cell boarder employing frequency reuse


Sharing is caring, but keep us informed

• Relative Narrowband Transmit Power (RNTP) is exchanged between macroeNBs via a backhaul interface (X2 interface)– The bitmap indicates whether transmission power of respective RB

exceed the predetermined threshold or not


Rel‐10 ICIC in heterogeneous networks

• To support femtocell deployment effectively, ICIC is necessary• Different from homogeneous network (macrocell deployments),

– Low power nodes (femto eNBs) must mute (or reduce transmission power) Named as “Protected resources” here

– High power nodes (macro eNBs) need not mute Named as “Non‐protected resources” here

• Protected/Non‐protected resources are multiplexed in frequency or time‐domain Both ICIC techniques are effectively supported in Rel‐10

Cell layer

Time

Frequency

Femto layerMacro layer

Frequency-domain ICIC

Car

rier

#1C

arrie

r#2

Frequency

Time

Cell layer

Time-domain ICIC

Car

rier #

1


Frequency‐domain ICIC for LTE

• Frequency‐domain ICIC for data channel is already supported from Rel‐8/9employing RNTP, although frequency‐domain ICIC for control channel isnot supported– Data channel is multiplexed in limited bandwidth, i.e., at RB‐level to

obtain multi‐user diversity in the frequency‐domain– Control channel is multiplexed in the entire bandwidth to obtain

frequency‐diversity• Here, control channel means downlink shared control channel (PDCCH)which sends the assignment information of UEs, and must be decodedcorrectly before decoding data channel (more on that later!)


Frequency‐domain ICIC for LTE‐A

• Multiple CCs are employed to perform ICIC for control channel• In order to indicate the assignment for different carriers, additional bits

(CIF: Carrier Indicator Field) is introduced


Time‐domain ICIC

• In order to apply time‐domain ICIC, femto eNBs must mute specific subframes to protect UEs connected to macro eNBs

• However, cell‐specific reference signal (CRS) needs to be sent for handover measurements, etc. Known in the 3GPP community as “Almost blank subframes (ABSs)”

• There are issues with CSI measurements on protected and non‐protected subframes at the macro layer


What else?

• Cell‐specific reference symbol (CRS) interference is a major issue• Additional mechanisms to cope with the CRS interference are under

discussion– Non‐zero transmit power ABS– CRS cancelation at UE– Transmitter side processing (sending interfering cell lists)– Etc.


Part 2: A comparison of state‐of‐the‐art ICIC techniques


The almighty grid – the LTE frame structure

• A lot of work has been done on data region interference mitigation• In this work, we focus on the control region because if it cannot be

decoded, the data region (and therefore the whole subframe) is anyway lost


Introducing the control channels: PCFICH

The control channel is 1/2/3 OFDM symbols long!


Introducing the control channels: PHICH

OK Mr. UE, I’ve received your UL transmissions!


Introducing the control channels: PDCCH

Hey you UE! Here are your DL and UL grants: x/y/z RBs!


What the control region really looks like

• The control region contains 3 control channels:– PCFICH: occurs only on first OFDM symbol; scattered in frequency

domain; indicates size of control region– PDCCH: spread in time and frequency; carries scheduling information– PHICH: spread in time and frequency; contains HARQ information

• We focus on the performance of the first two because of differences in their distribution patterns – the PCFICH has restricted positions in the time domain, whereas the PDCCH is dispersed in the time and frequency domains


What is already done

(a)•No coordinationHeavy

interference on 2OFDM symbols

(b)• Femto controlchannel sparsenessInterference to

first OFDM symbolis lowered

(c)•Almost blank subframeOnly interference from

reference symbolFemto data transmission

is not allowed


Enter my apartment at your own peril!

• 5x5 grid model• Macro users uniformly distributed• Trapped macro UEs are the focus of attention


System setup (simulation parameters)

Parameter ValueAvg. 5x5 blocks per sector 4Avg. macro UEs per sector 10Inter-site distance 500 mHeNB activation probability 10%System bandwidth 10 MHzeNB transmit power 46 dBmHeNB transmit power 20 dBmWall penetration loss 20 dB


Results (1/3): PDCCH performance for trapped macro UEs

• Significant improvement over benchmark• Sparseness also degrades femto‐to‐femto performance (not seen here)


Results (2/3): PHICH performance for trapped macro UEs

• Macro performance improves• Femto performance degrades (not seen here)


Results (3/3): PCFICH performance for trapped macro UEs

• Macro performance improves, but is still not good enough• Femto performance degrades, but is acceptable (not seen here)


Discussion

• The backward compatible macro‐to‐femto interference mitigation techniques are good for PDCCH

• However, their performance for the PCFICH is poor• The next section specifically deals with PCFICH protection for trapped

macro UEs• Once again, backward compatibility is key!


• Closed Subscriber Group (CSG) ID manipulation [3GPP TR 36.921].– The HeNB changes between a default CSG ID (assigned at deployment

time) and a dedicated (operator configured) CSG ID.– When there is a nearby macro UE, the HeNB uses the dedicated CSG ID

so that the UE can access the HeNB, otherwise it uses the default.The HeNB needs to be aware of when a macro UE is near it to trigger CSG ID selection.Centralized controller is required to ensure that no HeNB uses either CSG ID for a long time.Heavy signaling burden.

• Physical Cell Identity (PCI) reservation– It is possible to reserve a subset of available PCIs for HeNB use

No interference coordination through this approach

Things others are doing

We actively change the PCI of the HeNB at startup so that it causes the lowest collision with the PCFICH of the trapped macro UE!


• The PCFICH is important to protect because– Our past work has shown that it exhibits the worst SINR performance

compared to the other control channels.– So far it has not been possible to satisfactorily protect the PCFICH from femto‐

cell interference.– If the PCFICH is incorrectly decoded by the trapped macro UE, the subframe is

lost.• Further advantages:

Since HeNBs serve a small number of users (with typically a low PDCCHaggregation level), the control channel is sparse enough to allow for therearrangement of PCFICH, PHICH and PDCCH on the femto layer.This proposal can easily handle PCFICH protection for macro UEs trappedwithin the coverage of multiple HeNBs.

Why is the PCFICH so important?


How are PCFICH elements physically mapped?

• The 16 PCFICH resource elements are distributed over the entire frequency spectrum.• The PCFICH always occurs on the first OFDM symbol.• The location of the PCFICH resource elements undergoes an offset depending on thephysical cell identity (PCI).

x is an integer


And what about the PDCCH?

• The PDCCH search space (which CCEs are used for the PDCCH) of a UE depends onthe C‐RNTI assigned to that UE.

• The order of the CCEs is interleaved – the interleaving pattern is fixed.• The CCE interleaved order is cyclically shifted, depending on the PCI of the H/eNB.• This leads to the PDCCH locations being randomized, depending on the PCI.

Illustration only


So we propose…

• The proposal advocates carefully selecting the PCI of HeNBs at start‐up, such thatany interference caused by their control channels to the PCFICH of any trappedmacro UEs is avoided.– In order for this to be possible, the HeNB needs to identify the eNB that it is

closest to.• Identifying the eNB means that the HeNB must be aware of the PCI of the eNB

(decoded using synchronization procedure).

Illustration only


What needs to be done

• This procedure can not only protect all the control channels but also the CRSs

Identify•HeNB identifies most dominant macro eNB

Decode•HeNB decodes dominant eNB’s PCI

Adjust•HeNB adjusts its own PCI to reduce interference


Co‐channel deployment of macro and femto‐cells

• Stripe model used• Not all UEs are allowed to connect to a HeNB

For UEs having no access to HeNBs, downlink interference is significant

• Since the control channel is very important for proper functionality, how do we protect the control channel of trapped macro UEs?


Overall macro UE performance

• Compared to sparseness, this proposal results in an improvement of approximately 2 dB – especially at the low percentiles. This corresponds to the trapped macro UEs.

• Better performance than ABS configuration (due to better collision avoidance).

Deceivingly smallImprovement!


Overall macro UE performance with power control

• All curves shift to the right due to power control• Femtocell performance is still acceptable (not seen here)


Improvements/advantages

Enables the aggressor HeNB to continue to transmit data. Not possible with almost blank subframes

The proposed technology results in a significant improvement over introducing sparseness to the control channel.

• Therefore this technology incorporates the benefits of both sparseness and almost blank subframes.

• Multiple macro UEs can be protected simultaneously.• No additional hardware is needed.• No additional signaling is needed.• This procedure is backwards compliant with Rel.‐8/9 UEs.• Can be seamlessly combined with power control to boost performance even

further.


Lessons learned

• First study dedicated to control channel performance for LTE• Impact on vulnerable trapped macro UEs assessed• Two backward compatible techniques analyzed• Results show significant performance improvements for PDCCH but not for

PCFICH• PCFICH protection is further analyzed• A novel technique employing only PCI manipulation is shown to

significantly improve PCFICH performance without losing the femto subframe

• A few topics for further work would involve data channel interference mitigation, power consumption analysis and handover improvements for legacy systems; new control channel designs for future releases.


Part 3: Femto‐to‐Femto interference


Femtocells ‐ Overview

Increase in coverage Increase in data rate

Increase in interference

macro‐BS

FBS‐2

FBS‐1

FUE‐2

FUE‐1

MUE

2

3

1

1. Between FUE and MBS2. Between MUE and FBS3. Between FUE and FBS


Femtocells ‐ Overview

Increase in coverage Increase in data rate

Increase in interference

macro‐BS

FBS‐2

FBS‐1

FUE‐2

FUE‐1

MUE

2

3

1

1. Between FUE and MBS2. Between MUE and FBS3. Between FUE and FBS

How can we maintain acceptable user experience in dense femtocell networks?


Carrier Aggregation for LTE‐A

freq.CC1 CC2 CC3 CC4 CC5

100 MHz

• LTE-A makes use of carrier aggregation via the use of component carriers (CCs)

• Improves peak data rate and spectrum flexibility• Meets ITU-R requirements for bandwidth (>=40 MHz)• Backward compatibility is maintained• Smooth network migration is possible with minimal loss of

service for legacy terminals


How should the cake be eaten?

Interference between femtocells is a severe problem in densely deployed networks Desired quality of service cannot be achieved for cell edge users

Resource partitioning is widely used to enhance the performance of cell edge users interfering neighbors transmit data on different CCs the drawback is that it decreases the network’s overall resource efficiency

Vast variations of the interference conditions experienced by a BS during its operation Dynamic environment

BSs should use as many resources as possible depending on their interference environment flexibility in the amount of assigned resources

B

C

A

2

Component Carrier

1 3 freq.

pow.

Interference


Aim

Interference mitigation techniques should:

1. Be dynamic in nature resource assignment should be updated according to changes in the radioenvironment

2. Achieve high resource utilization

21 3 freq.

pow.B

C

AInterferenceDesired Signal


Aim




3. Be suitable formulti‐user deployments Each user in the same cell experiences different interference conditions CC allocation should be done according to the UE measurements

Primary CC (PCC)

21 3 freq.

pow.

A

B

C

PCC

3

2

3

11

3B

C

A

2

InterferenceDesired Signal

1


Aim




3. Be suitable formulti‐user deployments Each user in the same cell experiences different interference conditions CC allocation should be done according to the UE measurements

Primary CC (PCC) Secondary CCs (SCC)

4. Be applicable to the networks with a central controller ‐ central approach without a central controller ‐ distributed approach

5. Be compatible with the LTE‐A systems

21 3 freq.

pow.

A

B

C

PCC

3

2

3

1 3 SCC1

3B

C

A

2

31

InterferenceDesired Signal


Two different approaches

Central ApproachResources are assigned by a centralcontroller

More efficient resource utilization thanthe distributed approach

Needs extra signaling between the BSsand the controllerHigh computational complexity at the

controller

Distributed Approach Resources are assigned autonomously byBSs

Less complexity

High signaling overheadRequires long time period to reach a stable

resource allocationLow resource efficiency

Dynamic interference mitigation by resource partitioning


Central brain

• Interfering neighbor discovery:

C

A

Central controller

How does the controller assign resources to the BSs?

B Interference


Central brain

• Interfering neighbor discovery: UE makes measurement Identifies its interfering neighbors according to a predefined SINR threshold

C

A

Central controller


AA,C

B

B InterferenceFeedback


A centrally controlled graph based scheme


• BSs send cell IDs of the interfering neighbors to the central controller

C

A

Central controller


AA,C

B

A

A, C

B


Backhaul


A centrally controlled graph based scheme


• BSs send cell IDs of the interfering neighbors to the central controller• The central controller maps this information into an interference graph where

Each node corresponds a BS An edge connecting two nodes represents the interference between two BSs

C

A

Central controller

B C

A


AA,C

B

A

A, C

B


Backhaul


So what is graph coloring?

Graph coloring is a way of coloring the vertices of a graph such thatno two adjacent vertices share the same color here, Node BS; color CC

−20 −10 0 10 20−25

−20

−15

−10

−5

0

5

10

15

20

25

distance (m)

dis

tan

ce (

m)


So what is graph coloring?

Graph coloring is a way of coloring the vertices of a graph such thatno two adjacent vertices share the same color here, Node BS; color CC

−20 −10 0 10 20−25

−20

−15

−10

−5

0

5

10

15

20

25

distance (m)

dis

tan

ce (

m)

−20 −10 0 10 20−25

−20

−15

−10

−5

0

5

10

15

20

25

distance (m)d

ista

nce

(m

)

4

21

3

313

4

234

1

3

5

6

2

1

3

Resources can be assigned dynamicallyOne CC per BS is inefficient, as, when the number of CCs increases, a lot ofbandwidth tends to be wastedInefficiencies in terms of resource utilization


How can we improve upon this?

A BD

E

F

pow.

freq.

C


The recursive step

Applying the graph coloring algorithm multiple times Identify CCs that can be assigned to BSs without causing undue

interference

A BD

E

F

A BD

E

F

pow.

freq.

C C


Being clever helps too

A BD

E

C

Resource efficiency : 5/15

pow.

freq.



Identify the CC‐BS pairing whichmaximizes the resource efficiency

A BD

E

C

A BD

E

C

Resource efficiency : 5/15Resource efficiency : 6/15

pow.

freq.



Identify the CC‐BS pairing whichmaximizes the resource efficiency CCs are assigned to BSs by using a cost function

A BD

E

C

A BD

E

C

A BD

E

C

Resource efficiency : 5/15Resource efficiency: 6/15 Resource efficiency : 9/15

pow.

freq.


• 1st Step: Apply the graph coloring algorithm smin times

– Where smin is the minimum number of CCs that must be allocated toeach BS

– Using the cost function, assign one CC to every BS in each iteration(gains seen especially when the number of available CCs is high) –doing so increases the reuse efficiency of the system

• 2nd Step: For each CC:

– Using the cost function again, identify the combination of BSs whichmaximizes the utilization of this CC (example on slide 65)

• Advantages: Dynamic adaptation according to prevailing interference conditions Number of assigned CCs per BS is automatically adjusted depending

on the interference conditions Very low wastage of resources Low complexity and computational cost

Graph based dynamic frequency reuse (GB‐DFR)


Simulation Parameters

• 5x5 grid case

• Downlink only

• Only femto‐femto interference isconsidered

HeNB

UE

Parameter ValueSystem bandwidth 20 MHz

Traffic model Full buffer

max BS power 10 dBm

Antenna gain 0 dBi

Fading model No fast fading

Activation ratio 0.5

Number of UEs per BS 1

Number of CCs 6

SINR threshold 5 dB


Performance Evaluation – CDF of SINR

-10 0 10 20 30 40 5050

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

SINR [dB]

CD

F

Reuse-1Conv. Graph Col. (S=6)GB-DFR (S=6)


Performance Evaluation – CDF of User Capacity

0 5 10 15 20 25 30 35 400

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

User capacity [Mbps]

CD

F

Reuse-1Conv. Graph Col. (S=6)GB-DFR (S=6)


BS activation probability versus resource utilization

0 0.2 0.4 0.6 0.8 110

20

30

40

50

60

70

80

90

100Pe

rcen

tage

of A

ssig

ned

Subb

ands

BS Activation Probability p

Conv. Graph Col. (S = 6)GB-DFR (S=6)

Probability that an apartment contains an active femto BS


Effect of SINR threshold on performance

14 16 18 20 22 24 260

2

4

6

8

10

12

Average User Capacity [Mbps]

5th a

nd 1

0th P

erce

ntile

Use

r Cap

acity

5th percentile user capacity

10th percentile user capacity

th= 0dB

th= 5dB

th= 20dB

th= 15dB

th= 10dB

Sweet spotSweet spots


• Femto‐femto interference is a severe problem in femtocell networks

• Dynamic assignment of resources– Decreases coverage holes– Results in high resource utilization

• GB‐DFR attains a significant capacity improvement for cell‐edge UEs, at theexpense of a modest decrease for cell‐centre users

• Next section:– Extending the GB‐DFR to the networks where BSs serve multiple UEs– Fully distributed/autonomous approach

Lessons learned


Two different approaches (recap)

Central ApproachResources are assigned by a centralcontroller

More efficient resource utilization thanthe distributed approach

Needs extra signaling between the BSsand the controllerHigh computational complexity at the

controller

Distributed Approach Resources are assigned autonomously byBSs

Less complexity

High signaling overheadRequires long time period to reach a stable

resource allocationLow resource efficiency

Dynamic interference mitigation by resource partitioning


• Aim:– Autonomously assign resources in unplanned wireless networks– Balance high spatial reuse of radio resources with interference

protection for cell‐edge users• The proposed method relies on UE measurements

– Dynamic adaptation to the interference conditions faced in randomdeployments

• Less signaling overhead compared to existing LTE and LTE‐A signalingprocedures

• Can easily be adapted to work in either the time or the frequency domain

The decentralized technique – a summary


Resource assignment – who gets what?

21 3 freq.

pow.

A B

C



• Dynamic interference environment Number and position of neighbors change during the

operation Fixed frequency planning is sub‐optimal

21 3 freq.

pow.

Potential interference path

A B

C




operation Fixed frequency planning is sub‐optimal Dynamic assignment of resources!

21 3 freq.

pow.

A

B

C3

2

3

13


A B

C

21




operation Fixed frequency planning is sub‐optimal Dynamic assignment of resources!

• Multi‐user deployment Users in the same cell experience different interference

conditions Resource assignment should depend on UE

measurements to maximize resource utilization Classify resources according to their foreseen usages

21 3 freq.

pow.

A

B

C3

2

3

1 33


A B

C

21

3


Not all CCs are created equal

• Reserved CC (RCC):– Allocated to cell edge UEs– Protected region

2

3

A

B

C

1

A B

C

Potential interference path3

21

1




• Banned CC: – Interfering neighbors are restricted to use

the RCC allocated to the victim UE– This guarantees desired SINR at cell edge

UEs

2

3

A

B

C

1

A B

C


21

1

XXX X




• Banned CC: – Interfering neighbors are restricted to use

the RCC allocated to the victim UE– This guarantees desired SINR at cell edge

UEs

• Auxiliary CC (ACC):– Allocated to the UEs facing less interference– Neighbors are not restricted– Increases resource efficiency, especially, for

the multi‐user deployments 2

3

A

B

C

1

A B

C


21

1

XXX X

3

3


1. IDs of interfering BSs (UE Serving BS)– Each UE can measure the received

power from the BSs in its vicinity– It identifies interfering BS IDs according

to the predefined SINR threshold

What is needed to get this to work?

A

C


3

1

2B

1

2

3

A

B

C

1


1. IDs of interfering BSs (UE Serving BS)– Each UE can measure the received

power from the BSs in its vicinity– It identifies interfering BS IDs according

to the predefined SINR threshold


C

BA

1, 32, 3


Feedback from UE

2

3

A

B

C

1


2. RCC Indicator (BS Interfering BS)– Used for preventing interfering

BSs to use the RCC allocated to the victim UE


A

C


2

3

A

B

C

1

3

1

2B

1to B & C: Don’t use 1

XX

RCC indicator

X

X

to A & C: Don’t use 2


3. SINR over each CC (UE Serving BS)– Each UE observes different SINR over each CC– These measurements are used to find out which

CCs are available for transmission (as a RCC or ACC) depending on the predefined SINR threshold value


A

C


2

3

A

B

C

1 XXX X

3

1

2B

1


1 2 3- + -

1 2 3+ + +


A

C


B

Feedback from UE

1 2 3+ - -+ + +

Received SINR on each CC (cell A):

2

3

A

B

C

1 XXX X

Received SINR on each CC (cell B):

Received SINR on each CC (cell C):

+ = over threshold‐ = below threshold= banned CC

3. SINR over each CC (UE Serving BS)


1 2 3- + -

1 2 3+ + +


A

C


B

Feedback from UE

1 2 3+ - -+ + +

Received SINR on each CC (cell A):

2

3

A

B

C

1 XXX X

Received SINR on each CC (cell B):

Received SINR on each CC (cell C):

+ = over threshold‐ = below threshold= banned CC

3. SINR over each CC (UE Serving BS)

2

3

1 3XXX X

next time slot


Our latest acronym: Dynamic Autonomous CC Assignment – DACCA


Our latest acronym: Dynamic Autonomous CC Assignment – DACCA

Event triggered

CCs configuration is updated only if there is a change in the interference environment

All BSs are synchronized with a time duration equal to that of a so‐called ‘time slot’ Between the starting instances of two time slots, the CC configuration remains undisturbed


Simulation parameters

• 5x5 grid case and downlink direction is investigated

• Only interference between femto BSs is considered

• Statistics are taken at the end of 10th slot

• Three methods are compared: BS sniffing 1/4 and 2/4

DACCA

Femto BS

UE

Parameter ValueSystem bandwidth 40 MHz (4 x 10 MHz)

Traffic model Full buffer

Max. Tx Power per CC 20 dBm

Antenna gain 0 dBi

Shadowing std. dev. 10 dB

Activation ratio 0.2

Number of UEs per BS 4 (closed access)

SINR threshold 5 dB

-20 -10 0 10 20-25

-20

-15

-10

-5

0

5

10

15

20

25


CDF of SINR

-20 -10 0 10 20 30 40 50 60 70 8050

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

SINR [dB]

CD

F

BS Sniffing (1/4)BS Sniffing (2/4)DACCA


CDF of user capacity

0 5 10 15 20 25 30 35 40 45 500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

User capacity [Mbps]

CD

F



35 40 45 50 55 60 65 70 750

1

2

3

4

5

6

7

8

9

Mean Cell Capacity[Mbps]

Use

r Cap

acity

[Mbp

s]


Mean cell capacity versus user capacity

20%

20%

10%

10%

10%

5%5%

5%

20%


1 2 3 4 5 6 7 8 9 100

10

20

30

40

50

60

70

80

Time Slot

Perc

enta

ge

Percentage of Assigned ResourcesPercentage of Collisions (SINR<-10dB)

Convergence of the algorithm

Allocated RBs / All RBs

RBs Facing SINR below ‐10dB / Allocated RBs


Effect of SINR threshold

50 55 60 65 70 75 801.5

2

2.5

3

3.5

4

Average Cell Capacity [Mbps]

Cel

l Edg

e C

apac

ity [M

bps]

-5 dB

0 dB

15 dB

10 dB5 dB


Wrap up

• We have had a look at some fairly simple and backward‐compatible femto‐macro interference mitigation techniques and studied their performance

• We have identified that the control channel is particularly susceptible to interference – especially since it is so inflexible

• In particular, the most important control channel exhibits the worst performance

• We have addressed this issue by proposing a clever interference mitigation technique

• We then consider the case of femto‐femto interference• We have had a look at an interference mitigation technique which relies

on a central controller• We have then attempted to remove the central controller and see if that

works (it does)


Where do we go from here?

• Lots of interesting areas for further research• Femtocells are not going anywhere• Design of special air interfaces to deal especially with the interference

problem• New ways of handling handovers• Clever scheduling strategies with tight macro‐femto cooperation• Femtocells with cognitive radio?• MIMO?• Etc.

Copyright © 2012 DOCOMO Communications Laboratories Europe GmbH Infrastructure Research Group

DOCOMO Communications Laboratories Europe GmbHLandsberger Strasse 312 – 80687 Munich, GermanyPhone: +49 (89) 56824‐0 | www.docomolab‐euro.com

Zubin Bharuchabharucha@docomolab‐euro.com

lte/lte-a interference coordination for femtocells

Technology

design mitigating relay

effective deployment

relay node

bandwidth lteadvanced5

bandwidth evolution

lte terminal

power control

throughput 4by40