THINK AHEAD STAY AHEAD

The Performance of Open AccessFemtocells in 4G MacrocellularNetworksAndy Jeffries

[email protected]

WWRF 20, Ottawa 22.04.2008

THINK AHEADSTAY AHEAD

STRATEGIC VISIONSon future research directions

in the wireless field

Page 2Performance of Open Access Femtocells in 4G Macrocellular Networks WWRF#20 · Andy Jeffries · [email protected]

Introduction

• Femtocells have the potential to offer significant performance benefits to future cellular networks

• RF interference issues are key risk area in the deployment of femtocells in macrocellular networks

• Paper addresses SNIR and resulting data rates that can be achieved- Downlink only RF performance using a generic 4G air interface protocol- Open Access femtocell operation

• Work conducted in collaboration with Vodafone and Multiple Access Communications under the MONOTAS* Programme with UK Department of Trade & Industry funding

- Public deliverable available at http://www.macltd.com/monotas/

* MONOTAS - MObile Network Optimisation Through Advanced Simulation


What are Femtocells?

ControlEntity

Femtocellsin homes

Femtocells in commercial/

industrial organisations

Macrocellboundary

Operator’s Network Management centre. May set some of the

femtocell parameters

Broadband (wired) backhaul to operator’s

network • Femtocells are low power basestations deployed in domestic and commercial environments to improve coverage (known as Home-eNodeB in 3GPP).

• These access points will be connected to the operator’s network through their wired broadband connections

• Operator may set power levels, carriers etc. on the femtocells• May use same carrier for all macro- & femto-cells in a network• Can operate in Open Access or Closed Access mode

• Femtocells are low power basestations deployed in domestic and commercial environments to improve coverage (known as Home-eNodeB in 3GPP).

• These access points will be connected to the operator’s network through their wired broadband connections

• Operator may set power levels, carriers etc. on the femtocells• May use same carrier for all macro- & femto-cells in a network• Can operate in Open Access or Closed Access mode

Femtocells extending range of the existing

macrocell


-2000

-1500

-1000

-500

0

500

1000

1500

2000

-2000 -1500 -1000 -500 0 500 1000 1500 2000x-coordinate (m)

y-co

ordi

nate

(m)

MacrocellsFemtocellsFemtocell BoundaryCentral Cell Boundary

Simulation Scenario - 1

Macrocellular network• 19 tri-sectored

basestations• Spacing 1km• TX power +43dBm• Beam pointing angles

at 30°, 150° and 270°(referenced to due E)

• Antenna peak gain ~14dB, backlobe level -6dB

Femtocells• TX power +24dBm• Omnidirectional

antenna pattern gain 0dB

• Located on a grid spaced by 15m in x-and y- directions covering an area of one tri-sectored macrocell BTS.

Users• Randomly distributed,

200 per macrocellsector

Macrocellular network• 19 tri-sectored

basestations• Spacing 1km• TX power +43dBm• Beam pointing angles

at 30°, 150° and 270°(referenced to due E)

• Antenna peak gain ~14dB, backlobe level -6dB

Femtocells• TX power +24dBm• Omnidirectional

antenna pattern gain 0dB

• Located on a grid spaced by 15m in x-and y- directions covering an area of one tri-sectored macrocell BTS.

Users• Randomly distributed,

200 per macrocellsector

Tri-cellular arrangement of

Macrocell sectors

1km Macrocellspacing

Baseline simulations use 20% Femtocell

density

Femtocellsonly in 1 house in N: typically

10-50%

Femtocell layout


Simulation Scenario - 2

0

1

2

3

4

5

6

-20 -10 0 10 20 30 40SNIR (dB)

Thro

ughp

ut (b

its/s

/Hz)

Implemented ThroughputCharacteristic

Shannon Throughput

Throughput Conversion Characteristic

-220

-200

-180

-160

-140

-120

-100

-80

-60

-40

-20

1 10 100 1000 10000Range (m)

Path

Los

s (d

B)

Macrocell to User Path Loss

Femtocell to User Path Loss

Downlink Path Loss

Minimum SNIR -10dB before any data can get through

• Because they are unplanned, femtocellscan be located close to windows, in an upper floor etc. This has a significant impact upon the propagation to users

• A single slope propagation model, with a significant loss, to reflect transmission from outside to inside, has been used to calculate initial system performance

• 2GHz• 10dB Log-Normal σ• Macrocell-to-user:

• COST Hata single slope path loss model with correction for UE height ~r3.5

• Femtocell-to-user:• Higher path loss to reflect wall

penetration losses• r3.5 slope

Maximum throughput capped at 4.4b/s/Hz

(ref. 3GPP)


Performance Evaluation Technique

• C/I for a user calculated from the set of RXed signals: best/ ∑rest

• Maximum RX power from best server calculated per user

Serving Basestation(Macro or Femtocell)

Basestation(Macro or Femtocell)

User

Users

Identify #users served by basestation

Raw data rate calculated from SNIR from serving

basestation and Shannon capacity equation

Effective data rate calculated from raw data rate scaled by #users for

serving basestationCDFs

calculated of SNIRs and

Effective Data Rates

• Path losses & power levels for Macro- & Femtocell basestations combined to determine user SNIR• User data rate determined by dividing raw data rate by number of users on a base station

• Path losses & power levels for Macro- & Femtocell basestations combined to determine user SNIR• User data rate determined by dividing raw data rate by number of users on a base station


SNIR Distribution (scale in dB) EDR Distribution (scale in bits/s/Hz/User)

• Maps show open access operation of femtocells in a macrocell network. One dot = one user; • Approximately 60% of users are served by femtocells, 40% by macrocells. • Users served by femtocells enjoy very high SNIRs• Macrocell users in the femtocell region often see lower SNIRs because of the increased

interference from the femtocells.

• Maps show open access operation of femtocells in a macrocell network. One dot = one user; • Approximately 60% of users are served by femtocells, 40% by macrocells. • Users served by femtocells enjoy very high SNIRs• Macrocell users in the femtocell region often see lower SNIRs because of the increased

interference from the femtocells.

For clarity, log normal shadowing is omitted

Open Access Performance - 1 SNIR-Based Cell Selection


Open Access Performance - 2 SNIR-Based Cell Selection

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

-10 -5 0 5 10 15 20 25 30 35 40 45

SNIR (dB)

Cum

ulat

ive

Prob

abili

ty

Macrocell only network - all users

Open Access Macrocell served users

Open Access Femtocell served users

• Femtocell served users enjoy >100x increase in effective data rates because they are the only user on that femtocell

• Macrocell served users see a 1.23x increase in data rates because of the lower macrocell loading despite the lower SNIR

• Femtocell served users enjoy >100x increase in effective data rates because they are the only user on that femtocell

• Macrocell served users see a 1.23x increase in data rates because of the lower macrocell loading despite the lower SNIR

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.0001 0.001 0.01 0.1 1 10

EDR (bits/s/Hz/User)

Cum

ulat

ive

Prob

abili

ty

Macrocell onlynetwork - all users

Open AccessMacrocell servedusersOpen AccessFemtocell servedusers

SNIR Distribution EDR Distribution


Improving Data Rates Using EDR Based Cell Selection

With SNIR-based cell selection, users very close to femtocells can remain served by the macrocell

EDR Distribution

• Preferentially associating users with Femtocells benefits all• Remaining macrocell users benefit by reduced cell

loading• Users moved over to the femtocells benefit from

higher data rates due to the low femtocell loading -even if their resulting SNIR is lower

• Best EDR cell selection technique proposed for improving the data rate distribution, rather than the highest SNIR cell selection as traditionally used

• Preferentially associating users with Femtocells benefits all• Remaining macrocell users benefit by reduced cell

loading• Users moved over to the femtocells benefit from

higher data rates due to the low femtocell loading -even if their resulting SNIR is lower

• Best EDR cell selection technique proposed for improving the data rate distribution, rather than the highest SNIR cell selection as traditionally used

• Femtocells are small due to low Txpower and building penetration losses

• Cell loading thus generally low enabling very high data rates to be offered to users

• Femtocells are small due to low Txpower and building penetration losses

• Cell loading thus generally low enabling very high data rates to be offered to users


Comparison of SNIR and EDR Based Cell Selection

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.0001 0.001 0.01 0.1 1 10

Effective Data Rate (bits/s/Hz/User)

Cum

ulat

ive

Prob

abili

ty

Best SNIR cell selection Macro onlyBest SNIR cell selection Macro servedBest SNIR cell selection Femto servedBest EDR cell selection Macro servedBest EDR cell selection Femto served

EDR Distributions with users split by server type

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.0001 0.001 0.01 0.1 1 10Effective Data Rate (bits/s/Hz/User)

Cum

ulat

ive

Prob

abili

ty

Best SNIR cell selection Macro only

Best SNIR cell selection all servers

Best EDR cell selection all servers

EDR Distributions across all users in scenario

Best EDR cell selection improves data rates by ~3.7 times for users at the 10% level of the

distribution. These are macrocell served users seeing the benefit of lower macrocell loading

• Femtocell TX power +24dBm

• 20% Femtocelldensity

• Femtocell TX power +24dBm

• 20% Femtocelldensity

Macrocell users see improved EDRs

Femtocell users see slightly degraded EDRs, but still a substantial improvement

over macrocell users

Macrocellonly

baseline

EDR Cell selection increases proportion of femtocell users from 60% (SNIR-based) to 80%


Performance Impact of Femtocell Density

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 10 20 30 40

Active Femtocell Density (%)

Prop

ortio

n of

Use

rs s

erve

d (%

)

Femtocell Users Best SNIR cell selection LN 10dB

Macrocell Users Best SNIR cell selection LN 10dB

Femtocell Users Best EDR cell selection LN 10dB

Macrocell Users Best EDR cell selection LN 10dB

Proportion of Users Served for scenario with varying femtocell density

To off-load 50% of users to femtocells,

require ~10% active femtocelldensity, or ~7% if ‘Best EDR’ cell selection used

• A relatively small femtocell penetration is sufficient to reduce macrocell loading by 50%• EDR-based cell selection off-loads users onto femtocells faster than SNIR-based cell selection

• A relatively small femtocell penetration is sufficient to reduce macrocell loading by 50%• EDR-based cell selection off-loads users onto femtocells faster than SNIR-based cell selection

FemtocellTX power +24dBm

FemtocellTX power +24dBm


Impact of EDR-based Cell Selection in Real DeploymentsEDR distribution for real user distribution in a major city,

10dB LN hexagon network data for comparison

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.0001 0.001 0.01 0.1 1

Effective Data Rate (bits/s/Hz/User)

Cum

ulat

ive

Prob

abili

ty

Best SNIR cell selected Macrocell hexagon network 10dB LN

Best SNIR cell selected data from major city

Best EDR cell selected data from major city

The use of ‘Best EDR’ cell selection increases the EDRs for users in the lowest percentiles of the EDR distribution

Higher EDRs than regular hexagon network reflect the

higher SNIRs seen in the traffic distribution for this city

• Use of an EDR-based cell selection mechanism is effective at improving performance in current deployments as well as in macro- / femtocell networks

• Use of an EDR-based cell selection mechanism is effective at improving performance in current deployments as well as in macro- / femtocell networks

Best EDR-based cell selection improves data rates by 30% at 10% level c.f. best SNIR-based Cell Selection


• Introducing open-access Femtocells into Macrocell networks can significantly improve network capacity and user data rates

- Data rates for users attached to femtocells can be improved by 2-3 orders of magnitude

- Macrocell users also benefit due to the reduced loading on the cell despite a general increase in interference

• Femtocells can provide significant network performance improvements with relatively low penetration rates

- a femtocell density of ~10% of households can displace ~50% of the macrocellload

• A Best-EDR based cell selection technique is very effective at promoting macrocell users onto femtocells

- It is particularly beneficial to poorer served users, with improvements in data rates ~3.7x seen by poorest 10% of users

• Best EDR-based cell selection also offers benefits in macrocell-only networks- ~30% improvement in EDRs for the worst 10% of users in a real network

Conclusions

Documents

THINK AHEAD STAY AHEAD