earth summer school - oulu · entering new markets –densification of mobile networks /additional...

Copyright © 2011 EARTH – Energy Aware Radio NeTwork TecHnologies. All Rights reserved.

EARTH Summer School Oulu, 29 June 2011

Gunther Auer

DOCOMO Euro-Labs

Munich, Germany

http://cordis.europa.eu/fp7/ict/future-networks/programme_en.html

2Copyright © 2011 EARTH – Energy Aware Radio NeTwork TecHnologies

FP7 Project EARTH

15 Partners from 10

European countries

Funding Scheme: IP

Total Cost: € 14.8 m

EC Contribution: € 9.5 m

Duration:

January 2010 - June 2012

Project Coordinator:

Dietrich Zeller, Alcatel- Lucent

Technical Manager:

Ylva Jading, Ericsson

http://cordis.europa.eu/fp7/ict/future-networks/programme_en.html

http://upload.wikimedia.org/wikipedia/commons/b/b7/Flag_of_Europe.svg


Motivation of EARTH

• Today ICT CO2 contribution equals global air traffic contribution.Wireless communication ¼15% of ICT.

• Mobile Broadband traffic is “exploding” and entering new markets

– Densification of mobile networks /additional roll-out

– High costs and long lead time for power infrastructure

– Higher ratio of off-grid and weak grid BS

– Increasing network operational costs due to increasing energy costs


EARTH Outline

• EARTH Energy Efficiency Analysis

– Life Cycle Analyis

– EARTH Carbon Footprint Model

– Energy Efficiency Evaluation Framework (E3F)

– Case Study: Performance Evaluation of a 3GPP LTE Rel-8 Network

• Fundamental Challenges

– Application of EARTH E3F to Elaborate Fundamental Trade-offs

• Green Radio Key Enablers

Scaling Network Power Consumption with System Load

• Energy Efficiency Trade-offs

– Power Control vs Discontinuous Transmission (DTx)


EARTH Outline













Carbon Footprint Model

Objective:• Assess the impact of future wireless systems

on the global carbon footprint

Methodology:

• Historical trends are the basis for a forecast on the future growth of wireless communications until 2020

– Assess power consumption per site

o Consider various RAN technologies (GSM, WCDMA, LTE)

– Anticipate deployment trends

Number of required sites (for mix of RAN technologies)

Total power consumption of the network

Carbon footprint

* A. Fehske, et al., “The Global Carbon Foot-print of Mobile Communications − The Ecological and Economic Perspective”, IEEE Communications Magazine, to appear 2011


Key Drivers in Grows in Wireless Communications

• „Moore‘s Law“: Processing power and data storage double every 18 month

• Requires powerful information and communication systems that are able to transport the created information in acceptable time

Carbon footprint of the mobile communications industry may increase substantially

8Copyright © 2011 EARTH – Energy Aware Radio NeTwork TecHnologies 8

0

2

4

6

8

10

12

14

16

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

RBS operation (average) RBS operation (new)

RBS "sites" taken out of service RBS op. (new, no improv.)

RBS op. (average, no improv.)

MW

h /

RBS s

ite

RBS ”basic” unit

1.1 kW

1.5 kW

0.9 kW

Whole site (incl. cooling, power, transmission etc.)

1.0 kW (no improvements after 2010)

≈1.4 kW (no improvements after 2010)

RBSs taken out of service orswaped will be important

Capture product mix and deploymentfactors and the current trends gets extrapolatedCapacity also increases per site (/standard)

Carbon footprint forecast – Base Station Sites


Cellular Networks

Global Carbon Footprint 2007-2020

0

50

100

150

200

250

300

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

45%

30%

39%

RAN operation focus of EARTH

Mobile devicesoperation andmanufacturing

Data centers &data transport

2020:0.6% of global direct CO2

0.4% of total global CO2e

2008:0.3% of global direct CO2

0.2% of total global CO2e

Mto

nn

es C

O2e

Worst case: No improvements

34%

Operator activities

RAN construction

19%

Continousimprovementof 8% per year

1.6% of global economy [GSMA]


0

20

40

60

80

100

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

Diesel Area 2 Non-diesel

xxx Worst case

Radio Access Network (RAN)

Global Carbon Footprint 2007-2020M

ton

nes

CO

2e New base stations

2012 - 2020 (-8%/year)

No improvements

Installed baseup to 2012

Installed basediesel consumption

New diesel consumptionInstalled base of all diesel sites will be

modernized with hybrid operation until 2020

20

07

20

200.14% global direct CO2 0.18%

0.09% total global CO2e 0.12%


0

20

40

60

80

100

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

Diesel Area 2 Area 3

Non-diesel xxx Area 6

Worst case


Global Carbon Footprint 2007-2020M

ton

nes

CO

2e New base stations

2012 - 2020 (incl. EARTH)

No improvements


New diesel consumption

Impact of EARTH on newly installed base stations


Alternative energy modules (solar, etc.)

savings

20

07

20

20Green Mobile Communications

• 1000x total traffic• 3x base station sites• 1.5x RAN energy consumption


0

20

40

60

80

100

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

Diesel Area 2 Area 3

Non-diesel xxx Area 6

Worst case


Global Carbon Footprint 2007-2020

Energy savings on already installed base stations (e.g. through software updates) may yield further CO2e reduction

Potential for reduction in carbon footprint

Mto

nn

es C

O2e

Impact of EARTH on all (new & installed) base stations

savings

New base stations2012 - 2020 (incl. EARTH)

No improvements


New diesel consumption

Alternative energy modules (solar, etc.)



Carbon Footprint Analysis

Major findings:

• Served data rates have grown by about a factor of 1000 over the last 15 years

• The carbon footprint of mobile communications has only increased by a factor of 3 in this period

• Potential impact of EARTH on reducing the carbon footprint :

– New sites that consume only 50% energy (EARTH target):

No further increase in the RAN carbon footprint

– EARTH innovation implemented in installed sites:

Potential for significant carbon footprint reduction

• To further reduce RAN carbon footprint, diesel generators for off-grid sites may be replaced by alternative energy modules

Indirect EARTH impact


EARTH Energy Efficiency Evaluation Framework (E3F)

E3F

Power Model MetricsC

ell-

ed

ge r

ate

Aggregation to Global ScaleReference System & Scenarios

Base Station

PinPout


Reference System & Scenarios

E3F

Basis for traditional (small-scale, short-term)System Level Simulations

• Define system parameters and scenarios so to ensure comparable results among project partners

• Based on LTE

o 3GPP and ITU compliant evaluation methods

• Dynamic system simulations for different deployments

o Dense urban, urban, suburban and rural

Reference System & Scenarios


Reference Scenarios

• Baseline is hexagonal cell structure with 19 sites (57 sectors)

• Macro-cells are complemented by low power nodes

– Relays / repeaters

– Micro-, pico-, femto-cells

– Various types of base station cooperation

– Multiple radio access technologies (multi-RAT)

backhaulmicro/pico-cell

femto-cell

relaymacro-cell



E3F


ell-

ed

ge r

ate


Base Station

PinPout


The EARTH Power Model

E3F

Power Model

Realistic Power Model

Base Station

PinPout


Base Station (BS) Power Model

Power model

• Maps RF output power Pout to the total BS supply power Pin

Benefits: Key enabler for holistic energy efficiency analysis

• Allows to assess improvements on node and network level with realistic assumptions on the BS power consumption

• Improvements on components (e.g. power amplifier) can be quantified on system level

Base StationPin

Pout


Base Station (BS) Power Model

EARTH Power model

• Maps RF output power Pout to the total BS supply power Pin

• Power breakdown of different components for various BS types

(macro, micro, pico, femto BS)

Includes power amplifier (PA), RF, base band (BB) processing, DC-DC unit, cooling, mains supply

• Includes (basic) BS sleep (standby mode)

Base StationPin

PoutPARF

BB

PARF

DC

-DC

Co

olin

g

Mai

ns

Sup

ply


Power Model for Various BS Types

Power model

• PA dominant for macro BS, while BB dominates for small BS types

• Modest load dependency for macro BS, but to a much lesser extend for small BS types

Poor efficiency at low loads

• Power Savings through sleep mode at zero load

Linear approximation of power model is justified

MS – Main Supply

CO – Cooling

BB – Baseband

RF – Radio Frequency incl tranceiver

PA – Power Amplifier

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

500

1000

1500MacroCell - 46dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

50

100

150MicroCell - 38dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

5

10

15PicoCell - 21dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

2

4

6

8

10

12FemtoCell - 17dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]


Power Model for Various BS TypesTechnology Scaling

Power model

EARTH reference 2010:

o Base stations available on the market in 2010

EARTH reference 2012:

o BSs with components available 2010

o It takes about 2 years until these components are integrated into BSs

Significant improvements through enhanced PAs and silicon technology scaling

MS – Main Supply

CO – Cooling

BB – Baseband

RF – Radio Frequency incl tranceiver

PA – Power Amplifier

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

500

1000

1500MacroCell - 46dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

50

100

150MicroCell - 38dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

5

10

15PicoCell - 21dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]

PA

RF

BB

DC

CO

MS

20 40 60 80 1000

2

4

6

8

10

12FemtoCell - 17dBm Max

RF Output Power [%]

BS

Po

wer

Co

nsu

mp

tio

n [

W]

38%42%


Linear Power Model

• The power cosumption breakdown suggests that linear approximation of power model is justified*

NTRX: No. of transceiver chains

P0

Pin

Supply Power

RF Power

Pout

Pmax

Benefits: Key enabler for holistic energy efficiency analysis

o Allows to assess improvements on node and network level with realistic assumptions on the BS power consumption

o Improvements on components (e.g. power amplifier) can be quantified on system level

Ps

Pmax

* G. Auer, et al., “Cellular Energy Efficiency Evaluation Framework”, GreeNet Workshop, Budpest, Hungary, May 2011



E3F


ell-

ed

ge r

ate


Base Station

PinPout


EARTH Energy Efficiency Metrics

E3FPerformance Metrics for Concept Evaluations

• [J/bit] – reflecting traditional capacity and data rate improvements

o Increasing peak data rates quickly decreases [J/bit]

• [W/m2] – reflecting crucial coverage and service area aspects

o High fixed cost for coverage makes [W/m2] very important

• Metric communicated to ETSI TC EE

• Performance metrics follow 3GPP recommendations [TR36.814]

• Served cell throughput

• Mean, 5th, 10th, and 50th

percentile user throughput

Cell-edge throughputMetrics

Cell-

ed

ge r

ate

Recommended use of Energy Consumption Metrics


Energy Effinciency Metric vs Energy Consumption (Cost) Metric

• At low powers, small improvements in EE result in a large increase of the metric

• At high output powers, large improvements in EE result in a comparably small increase of the metric

May lead to false conclusions

Energy Efficiency (EE) Metric [bit/Joule]EE

Power

¢ EE

¢ EE


Energy Effinciency Metric vs Energy Consumption (Cost) Metric

• Linear increase of energy consumption index with increasing power P

Energy Consumption (Cost) Metric is selected for EARTH studies

Energy Consumption Metric [Joule/bit]

Power

¢ ECI

¢ ECI



E3F


ell-

ed

ge r

ate


Base Station

PinPout


Aggregation to Global Scale

E3F

Country wide 24 hour Traffic & Deployment Model

Aggregation to Global Scale


From Small-Scale to Large-Scale

• Aggregation to large-scale by weighted summing

Based on Geographical data determine deployment mix and population densities

Dense UrbanUrban

Suburban

Rural

• Small-scale, short-term system level simulations provide snapshots of the respective deployments


Large-Scale Deployment Model

Determine number of active subscribers for a given deployment

• High correlation of deployment densities within Europe

– Exception are nordic countries and Russia (excluded here)

Suburban

Urban

Dense urban

Population Areas in Europe

Rural Sparsely populated

Population Densities in Europe[citizens/km2]


Large-Scale Deployment Model

Determine number of active subscribers for a given deployment

• High correlation of deployment densities within Europe

– Exception are nordic countries and Russia (excluded here)

• Sparsely populated areas are only covered by 2G networks

not considered in evaluations

Sparsely populated

Suburban

Urban

Dense urban


Population Areas in Europe Population Densities in Europe[citizens/km2]

Dense urban

Urban

Suburban



Aggregation to Global ScaleMethodology

1. Define the average traffic demand per active subscriber for different terminal types

2. Define relevant scenarios of terminal/subscriber mixes

3. Daily traffic profile determines traffic volume per subscriber

4. Determine the number of active users per unit area for the considered terminal/subscriber mixes

5. Given the population densities for the respective deployments, the scenario specific network traffic per unit area in [Mbps/km2] can be derived;

6. the total network traffic of an average European country is obtained by weighted summing of the scenario specific network traffic


Long-Term Large-scale Traffic ModelsAverage number of subscribers sk [%] of the entire population for »2015

– anticipated traffic demands are based on UMTS forum forecast*

• 3 terminal types k are considered:

– PC: sPC=20%

– Tablet: stab=5%

– Smartphone: sfon=50%

• Reference PC of 2010: sPC=20%

*UMTS Forum, “Mobile Traffic Forecasts 2010-2020”, Report no. 44, May 2011 http://www.umts-forum.org/component/option,com_docman/task,cat_view/gid,485/Itemid,213/

%


10 100 1.000 10.000 100.000

PC (2010)

Smartphone

Tablet

PC

Tarffic demand [MB]

Long-Term Large-scale Traffic ModelsAverage traffic demand per active subscriber for »2015

• Usage of mobile services strongly depends on operator policies

– anticipated traffic demands are based on UMTS forum forecast*

• 3 terminal types are considered:

– PC data demand: [0.25, 2] Mbit/s

2Mbit/s provides HDTV

[8,64] GB/month/subscriber

– Tablet demands 1/2 of PC:

[0.125, 1] Mbit/s

– Smartphone: demands 1/8 of PC:

[31.25,250] kbit/s

• Reference PC of 2010: [31.25,125] kbit/s

High-end 3G traffic in 2010

[1,4] GB/month/subscriber


day month


10 100 1.000 10.000 100.000

PC (2010)

Smartphone

Tablet

PC

Tarffic demand [MB]

Long-Term Large-scale Traffic ModelsAssumptions for terminal & subscriber mixes in »2015

• assumptions are based on UMTS forum forecast*

• Heavy users request an average data rate of:

– Heavy PC: rhPC=2Mbit/s

– Heavy Tablet: rhtab= 1Mbit/s

– Heavy Smartphone: rhfon= 0.25Mbit/s

• Ordinary users request

1/4 the rate of heavy users:

– Ordinary PC: roPC=500kbit/s

– Ordinary Tablet: rotab=250kbit/s

– Ordinary Smartphone: rofon= 62.5kbit/s

• Reference PC of 2010: rrPC=62.5 kbit/s


day month

day month


Average Daily Data Traffic Profile

Actual traffic demand:

• Peak traffic demands are scaled by the daily traffic profile

• The EARTH daily traffic profile is extracted from operator data, and represents an exemplary European country

Daily traffic profile determines the actual traffic demand for the considered terminal & subscriber mixes for a given deployment

Traf

fic

pro

file


Traf

fic

pro

file

Long-Term Large-scale Traffic ModelsTotal Traffic demand per operator per unit area

pd: population density for deployment d

Nop: number of operators

Traffic demand per active subscriber rk

10 100 1.000 10.000 100.000

PC (2010)

Smartphone

Tablet

PC day month

MB

Number of active subscribers ®(t)


Long-Term Large-scale Traffic Models

0

10

20

30

40

50

60

70

80

90

100

Dense urban Urban Suburban Rural

low

mid

high

Deployment Scenario

Pe

ak T

raff

ic [

Mb

it/s

/km

2] 20% of subscribers are heavy users*

50% of subscribers are heavy users

all subscribers are heavy users

3.1 Mbit/s/km20.9

* most relevant according to UMTS Forum, “Mobile Traffic Forecasts 2010-2020”, Report no. 44, May 2011 http://www.umts-forum.org/component/option,com_docman/task,cat_view/gid,485/Itemid,213/

1.7

Average Peak traffic demand Rd per unit area is determined for:• Area served by Nop=3 operators• Various deployments • Various terminal & subscriber mixes


Global

Energy Efficiency Evaluation Framework (E3F)

Small-scale, short-term system

levelevaluations

Global Metric(long term, large scale)

Metric(short term, scenario specific)

BS

channel

powermodelPout

Pin

system performance

mobile

Large scale area & Long term traffic load

Aggregation to global

scale

E3F Block Diagram


Case Study: Energy Efficiency of LTE

Short-term, small-scale evaluations:

• Macro-cellular network with regular hexagonal cell layout is implemented 19 sites, each with 3 sectors

• Inter-site distance (ISD)

– Dense urban and urban environments: 500 m

– Suburban and rural areas: 1732 m

• The users are uniformly distributed, with user densities corresponding to the respective deployment scenarios

• The simulation parameters are taken from 3GPP specifications [TR36.814], [TR25.814]

– 10 MHz bandwidth

– 2.1 GHz carrier frequency

– 2£2 MIMO transmission with adaptive rank adaption


Energy per bit [Joule/bit]

Short-term, small-scale evaluations

• Energy efficiency [Joule/bit] severely degrades at low loads

• Suburban/rural scenarios are more energy efficient at low loads

Larger cell size is more energy efficient

• However, high system throughputs may only be achieved in urban scenarios

Small cell size boosts capacity


Power Consumption per Unit Area (P/A)

• Urban scenario achieves more than 10 times higher system throughput than suburban/rural scenarios due to higher site density (urban ISD=500m, compared to suburban/rural ISD=1732m)

• However, suburban/rural scenarios consume more than 10 times less power

• Only a modest dependency of the system throughput on the network power consumption Due to macro BS power model, which exhibits poor efficiency at low loads



Power Consumption per Unit Area (P/A)


• Long-term, large-scale evaluations:

o average power per area unit is 0.6 kW/km2, independent of data usage scenario

o System operates a very low loads: on average less than 10 % of resources are utilized, even for the high data usage scenario

• Only a modest dependency of the system throughput on the network power consumption Due to macro BS power model, which exhibits poor efficiency at low loads


49 49 50

16,1 16,3 16,414,4

13,2 13,0

0

10

20

30

40

50

60

2013 2017 2020

kWh

/ ye

ar

/ su

bsc

rip

tion

Traffic Scenario

LTE System 30% traffic share LTE System, 100% traffic share Global statistics & trends

Compare E3F with Carbon Footprint Model

• Assuming that all deployments are served by 3 operators for the E3F leads to grossly overestimated power consumption

The network assumed for the E3F appears to be over provisioned

• The following measures allow to scale the E3F estimates with the carbon footprint model:

– Base station sharing: 3 times less base stations required

– Inter-site distance (ISD): the E3F uses 3GPP assumptions; increasing the ISD also results in a significant reduction in power consumption

– Energy efficiency improvements in future base stations

E3F: 3 networks

E3F: 1 network

Carbon footprint model


Traffic in Mbps

Fraction of cells in % (sorted by throughput)

Fra

ction o

f tim

e in %

What about Real Networks?

For comparison: measurements of system load for a real 3G network*

* P. Frenger, P. Moberg, Y. Jading, and I. Godor, „Reducing energy consumption in LTE with cell DTX“, VTC-Spring2011

• HSDPA traffic data

• more than 300 cells in a bigEuropean city

• measured for several weeks.

• Median traffic: 15 kbps

83% of samples below 1 Mbps


Energy Efficiency Analysis Main Achievements

• Carbon footprint model

– Quantify the impact of future wireless network on the carbon footprint until the year 2020

– Assess the potential impact of energy savings provided by EARTH

• The EARTH Energy Efficiency Evaluation Framework (E3F)

– The EARTH E3F facilitates a holistic performance analysis at network level, comprising:

o Realistic base station power model for various BS types

o Long-term traffic model & large-scale deployment model that facilitate aggregation to global scale

o Energy efficiency metrics that complement existing performance metrics

• Evaluate the energy consumption of a LTE Rel-8 cellular network

• Identify the fundamental challenges for energy savings

Vast potential at low traffic loads


EARTH Outline













Fundamental ChallengesLessons Learned from EARTH Energy Efficiency Analysis

High loads sustainable growth challenge

• Squeeze more bits through an already busy network without an increase in energy consumption

Improve energy efficiency [Joule/bit]

Already thoroughly investigated

Improvement ¼300 times over the last 15 years

Challenge to keep up with this trend in a subject well understood

Low loads scale power consumption to network load

• On average more than 90% of radio resources are idle

• Yet only modest reduction in energy consumption of ¼20% with respect to maximum load

Energy efficiency severely degrades at low loads

Vast potential for energy savings in particular at low loads


Application of EARTH E3FMIMO Energy Efficiency in LTE

• Compare energy efficiency of LTE Rel-8 with different number of Tx antennas

• Number of Rx antennas is kept constant

MIMO benefits from enhanced directivity & spatial multiplexing

In particular cell-centre users benefit from MIMO

XX

X

ISD: inter-site distance

MTX = 2

MTX = 1

10

20

30

40

50

60

70

Us

er

bit

rate

pe

rce

nti

les

(5

5

0 9

5)

Suburban/Rural Scenario, ISD=1730m

2 4 6 8 100

System throughput [Mbps/km2]

10

20

30

40

50

60

70

Urban Scenario, ISD=500m

0 20 40 60 80 100 120 140

System throughput [Mbps/km2]Us

er

bit

rate

pe

rce

nti

les

(5

5

0 9

5)


Application of EARTH E3FMIMO Energy Efficiency in LTE

MIMO leads to higher power consumption due to additional radio unit

Trade-off between spectral efficiency and energy efficiency

Results suggest to only activate additional Tx antennas on demand

Note: MIMO may greatly benefit from improved hardware and sleep modes

XX

X

Sup

ply

Po

wer

RF Power0

120 W

694 W

1340 W

410 W

776 W

Power Model

ISD: inter-site distance

0 20 40 60 80 100 120 140


1

2

3

4

5

6

P/A

[kW

/km

2]

MTX = 2

MTX = 1

Urban Scenario, ISD=500m

2 4 6 8 100

0.1

0.2

0.3

0.4

Suburban/Rural Scenario, ISD=1730m


P/A

[kW

/km

2]

MTX = 2

MTX = 1


Power per Area [W]

0

1000

2000

500

1500

1500500 1000

Application of EARTH E3FFinding the Optimal Cell Size

• Small cells: less power per site is needed to transport data

• Large cells: less sites consume less offset power

• Coverage and capacity conditions

– Min bit-rate = 2 Mbps

– Min capacity 20 Mbps/km2

– Limits the feasible range of ISDs

Results:

The offset power at zero load drives the power consumption

A minimum appears only when the slope of the power model is large with respect to the zero load offset

BS load [%]

Sup

ply

Po

wer

[W

]

0

1000

2000

3000Power Model

100%

Inter-site distance (ISD) [m]


Application of EARTH E3FEnergy Efficiency of HetNets

Geometry

• Macro-cells: hexagonal cell layout with 3 sectors per site

• Micro-cells: Nmic nodes per site randomly placed

• Users are uniformly distributed

• Urban scenario: inter-site distance ISD=500m

BS operation:

• 2x2 MIMO, 10MHz bandwidth

• EARTH 2010 reference power model

Macro only

HetNet, 10 micros per site

Snapshot of simulatedSINR distributions

0 20 40 60 80 1000

50

100

150

200

250

300

350

RF load in %

To

tal B

S p

ow

er

in W

Macro BS

Micro BS

At idle times BS power consumption is reduced by ¼50%

by exploiting sleep mode


Application of EARTH E3FEnergy Efficiency of HetNets

0 5 10 15 20 25 300

2

4

6

8

10

Av. Number of micro BS deployed per site

Are

a p

ow

er

in k

W/k

m2

2 Mbit/s/site

47 Mbit/s/site

92 Mbit/s/site

136 Mbit/s/site

0 50 100 1500

20

40

60

80

100

120

140

160

Offered traffic in Mbit/s/site

Se

rve

d tra

ffic

Mb

it/s

/site

Nmic = 0

Nmic = 3

Nmic = 10

Nmic = 20

Nmic = 30

Results

• Complimentary deployment of small cells boosts capacity

• Additional small cells deployment can reduce network energy consumption for certain load conditions


EARTH Outline













Key Enablers for Green Wireless Networks*Scaling Power Consumption to Network Load

Components

• Improve power efficiency at low loads of power amplifier (macro-cell) and baseband engine (small BS types)

Adaptive network reconfiguration

• Capacity gains of MIMO, carrier aggregation (CA) and heterogeneous networks (HetNets) do not come for free

*L. Correia, et al., “Challenges and Enabling Technologies for Energy Aware Mobile Radio Networks,” IEEE Communications Magazine, vol. 48, no. 11, pp. 66 –72, 2010.

00.00hrs 24.00hrs12.00hrs

Telecom trafficSaved energy

Freq.

Freq .

Switch off sites or RATs

Switch off radio units

HetNetto macro

Bandwidth adaptation

Time.

Time .

BS sleep modes

Hig

hLo

w

Traffic

Reconfigure system to long and/or short-term traffic variations


Key Enablers for Green Wireless Networks Discontinuous transmission (DTX) in LTE

Deactivate radio components when not transmitting

Micro DTX (<1 ms)

• Possible in LTE Rel-8

Short DTX (<10 ms)

• UEs perform mobility measurementson synchronization signals

• Standardization impact:

Remove CRS; Only PSS/SSS and PBCHtransmissions in empty cells

Long DTX (>10 ms) Low activity mode:

• Active period: Transmission of SSS/PSS and PBCH (short DTX)

• Idle period: No transmissions at all

• Standardization impact:

cell search alternatives needed

72 center

sub-carriers

1 radio frame (10ms)

1 subframe (1ms)

1 slot (0,5ms)

SSS

CRS

PSS

BCH

DTX

72 center

sub-carriers

Normal mode

(Only short DTX)

Normal mode

(Only short DTX)

Low activity mode

Active Period

(e.g. 100 ms)

Long DTX

(e.g. 900 ms)

Micro DTX

Short DTX

Long DTX


Key Enablers for Green Wireless Networks DTX in LTE: Technology Potential

• Explore technology potential for cell DTX in LTE

– Assume very low sleep power consumption

Without cell DTX the power usage scales badly due to high power consumption in idle mode

Significant energy savings of DTX at low load, due to high probability of empty (sub)frames

Sup

ply

Po

wer

RF Power0

120 W

1310 W

720 W

Power Model

30W

By not transmitting CRS (short DTX) further energy savings are achieved

Very high saving potential as according to the EARTH E3F, more than 90% of resources carry no data

Inter-site distance ISD=1200m

5 10 15 20 250

100

200

300

400

500

600

System load [Mbps/km2 ]

No DTX

Micro DTX

Short DTX

P/A

[W/k

m2]


Key Enablers for Green Wireless Networks Adaptive Transceiver For Macro BS

o Stepwise adaptation of PA output power to required bandwidth

• Operating point adjustment

Usage of variable bias and adapted supply voltages

• Deactivation of Components –Switch off certain PA stages when no signal is transmitted


Key Enablers for Green Wireless Networks Transceiver for Small Base Station Types

BB-DLBB-UL

RF-DLRF-UL

PA

Main supply

DC-DC

Pico/Femto-cell Base station

LLL

BB-DLBB-UL

RF-DLRF-UL

PA

L

Digital baseband engine

RF transceiver

Power amplifier

Adaptive energy efficient PA

- Operating point adjustment

and PA deactivation

- Tunable matching for PA

load modulation

Tx/Rx antenna port

Common component tracks related to small-cell base-station→ Energy adaptive small-cell transceiver for dynamic signal load

Expected power savings:

• Up to 40% of power reduction at low signal level

• 60% power savings in sleep mode


EARTH Outline













• Investigate the theoretical limits of energy efficient resource allocation algorithms*

• Find lower limit of base station power consumption by combining

• Adhering to QoS requirements (minimum rate)

Energy Efficient Resource Allocation Objective

Resource Sharing (TDMA)


Power Control (PC)

Power Control (PC)

Sleep mode (DTX)Sleep mode (DTX)

Link 1

Link 2

*H. Holtkamp, G. Auer, H. Haas, “On Minimizing Base Station Power Consumption,” IEEE Vehicular Technology Conference (VTC), 2011 fall, San Francisco, USA.


• Investigate the theoretical limits of energy efficient resource allocation algorithms

• Find lower limit of base station power consumption by combining

• Adhering to QoS requirements (minimum rate)

• Take into account BS power model

Energy Efficient Resource Allocation Objective



Power Control (PC)

Power Control (PC)

Sleep mode (DTX)Sleep mode (DTX)

Link 1

Link 2

Pin Base Station

PoutPARF

BB

PARF

DC

-DC

Co

olin

gM

ain

s Su

pp

ly


Energy Efficient Resource Allocation Methodology

Establish analytical solution for PC + TDMA on two linksEstablish analytical solution for PC + TDMA on two links

Add a power model to translate transmit power to supply powerAdd a power model to translate transmit power to supply power

Add the option of sleep mode (DTX)Add the option of sleep mode (DTX)

Evaluation in simulationEvaluation in simulation

From the Shannon capacity, find the sum transmit power as a function of rate, noise power and channel gain on both links.

Transmit power: PTx

Resource share: µRate target: ³min

Noise power: NChannel gain on link i: Gi

(1)







(2)

A base station’s supply power is composed of the standby consumption P0

and a load dependent part with load factor m and transmit power PTx.In sleep mode, the base station consumes Ps.







Determine numerically, which combination of transmit power per user and sleep duration minimizes the supply power consumption.

3GPP model• 1 cell with 250m radius• 8dB shadowing SD• 10MHz BW• 10 users


Energy Efficient Resource Allocation Considered Power Models

Examine performance with different power models

1. LTE macro base station as deployed in 2010

– Including sleep mode

– Only one sector considered

0 10 20 30 400

50

100

150

200

PTx

[W]

Ps

up

ply

[W

]

EARTH Reference 2010

0 10 20 30 400

50

100

150

200

PTx

[W]

Ps

up

ply

[W

]

Power Model 2014

1.

2.2. Base stations expected to be available on the

market in 2014*

– Improved power efficiency

– Reduced sleep mode consumption

*L. Correia, et al., “Challenges and Enabling Technologies for Energy Aware Mobile Radio Networks,” IEEE Communications Magazine, vol. 48, no. 11, pp. 66 –72, 2010.


Energy Efficient Resource Allocation Results

0 10 20 30 400

50

100

150

200

PTx

[W]

Ps

up

ply

[W

]


0 10 20 30 400

50

100

150

200

PTx

[W]

Ps

up

ply

[W

]

Power Model 2014

1.

2.

0 5 10 1560

100

140

180

220

Required link rate [Mbps]

Ps

up

ply

[W

]


0 5 10 1520

60

100

120

Required link rate [Mbps]

Ps

up

ply

[W

]

Power model 2014

30%

30%

35%

47%

18%

23%

Pmax

reference

bandwidth adapt.

PC only

binary DTX

PRAISDTX + PC

Low traffic loads

• Discontinuous transmission (DTx) is most efficient to save energy

High traffic loads

• Power control (PC) yields significant gains

Mid traffic loads

• Both DTx and PC attain comparable gains

• Cross-over point depends on used power model

Combination of DTx and power control (DTx + PC) provides additional gains

Compared to bandwidth adaptation (Psupply is scaled by amount of used resources), DTx + PC achieves significant gains (30—50%) for all loads

BUT, in LTE

• Gains of DTX are compromised by control signaling

• No downlink power control


Thank you

earth summer school - oulu · entering new markets –densification of mobile networks /additional...

Documents