Green Radio - Towards Sustainable Wireless Networks

www.mobilevce.com © 2009 Mobile VCE

Green Radio – Towards Sustainable Wireless Networks 8th July 2010

Steve McLaughlin

University of Edinburgh


Presentation Overview

  Why Green Radio?

  Green Radio as an Enabler

  Metrics and Measures: What and Where?

  Green Radio Research Challenges

  Conclusions


Why Green Radio? Operator & Manufacturer Perspective

  Increasing energy costs with higher base station site density and energy price trends  A typical UK mobile network consumes 40 MW

  Overall this is a small % of total UK energy consumption, but efficient communications can save energy in other industries

 Energy cost and grid availability limit growth in emerging markets (high costs for diesel generators)

 Corporate Responsibility targets set to reduce CO2 emissions and environmental impacts of networks  Vodafone1 - “Group target to reduce CO2 emissions by

50% by 2020, from 2006/07 levels”

 Orange2: “Reduce our greenhouse emissions per customer by 20% between 2006 and 2020”

1. http://www.vodafone.com/etc/medialib/attachments/cr_downloads.Par.25114.File.tmp/CR%20REPORT_UK-FINAL%20ONLINE_180908_V6.pdf 2. http://www.orange.com/en_EN/tools/boxes/documents/att00005072/CSR_report_2007.pdf


Types of Energy

 Embodied energy   Energy used in raw material extraction, transport,

manufacture, assembly, installation of a product or service including disassembly, deconstruction and decomposition

 Operating energy   Energy expended over the operational lifetime of the

product   In VLSI devices the embodied energy can be 2.5 – 3 times

the operating energy but this can be reduced for a recycled handset!


Where is the Energy Used3?

  For the operator, 57% of electricity use is in radio access

  Operating electricity is the dominant energy requirement at base stations

  For mobile user devices, most of the energy is expended during manufacture

9kg CO2

4.3kg CO2

2.6kg CO2

8.1kg CO2

Mobile

CO2 emissions per subscriber per year3

Operation

Embodied energy

Base station

3. T. Edler, Green Base Stations – How to Minimize CO2 Emission in Operator Networks, Ericsson, Bath Base Station Conference 2008


Base Station Power Use @ 20035

5. H. Karl, “An overview of energy-efficiency techniques for mobile communication systems,” Telecommunication Networks Group, Technical University Berlin, Tech. Rep. TKN-03-XXX, September 2003. [Online]. Available: http://www-tkn.ee.tu-berlin.de/∼karl/WG7/

AG7Mobikom-EnergyEfficiency-v1.0.pdf 3.8 kW AC > 120 W RF


Rect- ifier 76% Base

Station 30%

LTE Base Station Efficiency

We make the following assumptions compared to the Nortel GSM model :   Simple climate control – free air cooling only   No feeder cable losses – mast head amplifiers   Base Stn PA uses “Nujira” style power supply controller

100% Antenna

85% Efficient 40% Efficient

Under this model saving X% on Base Station PA power yields ~X% reduction in total power

Cool- ing

90%

90% Efficient


Energy Consumption

  The Base Station (BS) is the most energy–intensive component of a 3G mobile network.

  A typical 3G Base Station consumes about 3-500 W with a output power of ~40 W. This makes the average annual energy consumption of a BS around 4.5 MWh (which is lower than a GSM BS).

  A 3G mobile network with 12,000 BSs will consume over 50 GWh p.a. This not only responsible for a large amount CO2 emission it also increases the system Opex.

  This is worse in developing mobile markets such as India and China which have 10-20 times number of mobile subscribers!


UK Network Consumption 2004-20084

4. CR review, Vodafone UK, Corporate Responsibility 2007/08


What is a Femtocell?

The Femtocell is a low power cellular base station access point for a home or small business.

Residential installations support 2 to 4 mobiles or 8 to 16 for business designs.

Connects to the service network via (DSL or cable) broadband.

WiFi - Femtocells interface to existing mobile handsets via licensed spectrum.

Use GSM, CDMA2000, TD-SCDMA, WiMAX, WCDMA or LTE system.

Key attractions are better coverage and capacity, especially indoors, with superior voice quality and battery life.


Femtocells – further energy reduction!


Low hanging fruit ..

  Optimise rectifier

  1800 MHz Cut-off at night-time

  2G modernisation

  Power Saving Features (software controlled PA bias)

  Remote radio heads (RRH)

  Use of free cooling systems and operate at higher thermostat settings

  Improve measurements of energy usage (Automatic Meter Reading), and upgrade energy inefficient components – e.g. fans, lighting control, ..


Other wireless comparisons

IEEE 802.11 Wi-Fi laptop access is only 1–2% efficient when we compare radiated power to total required drive power6.

(Transmit power of 20 – 100 mW for an AC power drain of 2 – 8 W)

6. WWW.atheros.com/pt/whitepapers/atheros_power_whitepaper.pdf


Energy Consumption – The Challenge

  Since 2006, the growth rate of data traffic on mobile networks has been approximately 400% p.a..

  14% of handset sales in 2009 are smartphones (i-phone, blackberry, android)7.

  With increased deployment of mobile broadband traffic will grow at least at this rate in coming years.

  This growth demands a much higher energy consumption than today.

  The challenge is how to design future mobile networks to be more energy efficient to accommodate the extra traffic.

7. http://www.mobile-computing-news.co.uk/industry-news/2372/gartner-report-shows-smartphones-growth.html


Energy use cannot follow traffic growth without significant increase in energy consumption

  Must reduce energy use per data bit carried Number of base stations increasing

  Operating power per cell must reduce Green radio is a key enabler for cellular growth while guarding against increased environmental impact

Green Radio as an Enabler8

Cos

ts

Time

Voice Data

Revenue

Traffic Diverging expectations for traffic and revenue growth

Trends7:

  Avg annual increase in data traffic is 400%

  Avg annual increase in revenue is 23%

  Number of base stations / area increasing for higher capacity

  Revenue growth constrained and dependent on new services

8. Traffic / revenue curve from “The Mobile Broadband Vision - How to make LTE a success”, F. Meywerk, Senior Vice President Radio Networks, T-Mobile Germany, LTE World Summit, November 2008, London


Green Radio Scenarios Two Market Profiles: 1.  Developed World

  Developed Infrastructure   Saturated Markets   Quality of Service Key Issue   Drive is to Reduce Costs

2.  Emerging Markets   Less Established Infrastructure   Rapidly Expanding Markets   Large Geographical Areas   Often no mains power supply

– power consumption a major issue


Over a year, 1m2 solar panel produces ~400 kWh energy, or about 10% of a 3G macrocell BS requirement (in London, < 5%).

Solar panels in some locations now provide the energy requirement for a small “rural or village” BS see, e.g., the WorldGSMTM 100 W BS from VNL. 10. http://www.vnl.in/technology/sites/


Key Research Questions

What would be an appropriate green network architecture ?   A low power wireless network & backhaul that still provides

good quality of service What are the best radio techniques?

  Need to consider across all layers of the protocol stack that collectively achieve significiant (100x?) power reduction

Another issue: Given that most of the energy associated with mobiles is in the manufacture (~85%) how do we make their use sustainable?


Architecture: Energy Efficiency Analysis

The image cannot be displayed. Your computer may not have enough memory


The image cannot be displayed. Your computer may not have enough memory The image

cannot be displayed. Your computer may not have enough memory The image

cannot be displayed. Your computer may not have enough memory


The image cannot be

The image cannot be

The image cannot be

The image cannot be

The image cannot be

The image cannot be

Macro Micro Pico Femto

RRM

BER/FER vs Eb/No

Link Budget

Mobility/Traffic Models

Packet scheduling, handover, power and load control

Differentiated QoS, fast fading effects, UE speed, MIMO

Energy consumption is proportional to distance

User Equip (UE) movement, traffic types & mixes

Step1: Large vs. small cells applying the energy metrics

Step2: Overlay Source & Network Coding and/or Cooperative Networking

Step3: Evaluate optimum cell size from the following perspectives…….


Where to Measure?

‘Protocols’

Signal Processing RF

Channel

Downlink

RF

Antenna

Antenna Signal Processing RF

RF

Antenna

Antenna

  Two ‘points of measurement’

  Need to evolve to encompass multi-Users in a Cell and multi-cells in the RAN.


BS Efficiency Model

This linear model is a good first approximation, but:   Omits the antenna – not appropriate when relating total

BS power to RF radiated power   Simple, linear, multiplicative model unlikely to match

reality

Total RF Power

Tota

l BS

Pow

er

Ideal

More Realistic ‘Bad News’

‘Urquhart Diagram’


Metrics: RAN Energy Model

  RAN: Whole RAN or RAN segment of fixed geographic area ARAN with Ns cell sites and Nc cells giving variable cell density Cd

RAN of fixed Geographic area ARAN

  Our energy evaluations apply to the RAN  ETotal : Total energy consumed in RAN lifetime  EEmbod : Embodied energy consumed in RAN lifetime  EOpex : Operational energy consumed in RAN lifetime


Energy Performance Metrics

For operational energy, we define a RAN Energy Consumption Rate as

T1 T2

ERAN

ERAN,2

ERAN,1

T

E Energy Consumption

T1 T2

MRAN

MRAN,2

MRAN,1

T

M Data Volume

NB Linearity depicted for illustration purposes only


LTE Transmission Model

For downlink assume signalling overhead is approx 20-40% of the total transmission   Pilot Signals and Cyclic Prefix   Broadcast Channel and Synchronisation Signals   User specific signalling for resource blocks

20% overhead reasonable for low load or small No of users;

Up to 40% for high load and large No of users The remaining 60-80% of resource blocks are available

for supporting data transmission

Signalling: 20-40% Data Transmission: 60-80%


Daily Traffic Variation12

Vertical bars show variability in cell traffic load over 24 hour period   Change by factor of ~2 from lowest load at 7am to highest at 9pm   Significant periods of low network utilisation   Expect major load variations between cells Solid trace shows % data in overall carried traffic (3-6%) in 2009

12. H Holma and A Toskala, “LTE for UMTS - OFDMA and SC-FDMA Based Radio Access” John Wiley 2009

Midnight Midday Midnight

7am 9pm


Sleep Modes: Vodafone Traffic Results

Midnight Midday Midnight

7am

9pm Min is ~25% of the Max value

  Significant periods of low network utilisation – corroborated by Nokia traffic data

  Also expect big load variations between cells

Optimising Sleep Mode Performance: Minimise Energy (J) per frame of data Minimise Energy (J) per transmitted bit or ECR

Comparing Sleep Mode (Energy E2) with Non-Sleep Mode Systems (Energy E1):

Energy Consumption Gain (ECG) = E2/E1 Energy Reduction Gain = 1-ECG (%)

Calibrate gains in RF Energy Consumption


GSM Power Saving with software controlled PA bias switching

Average power saving > 10%


Sleep Mode Control for Power Conservation � Energy savings estimation: 1. Switch off base station/femtocell transmissions in low load conditions 2. Expect that in low load conditions, 20% of the base transmissions are

overheads that could be switched off: 3. For base stations, gating off timeslots in low load transmissions may achieve

energy savings:   With close to zero traffic load with 20% signalling overhead, switch off

50% of idle time slots: Power Reduction = 0.5 or 50%   If we assume this is achieved for 8 hours from 24 hours Power Reduction = 0.5*0.333 = 0.167 or 16.7%   By switching off more slots, higher gains are possible 4.  Femtocells: need a different power model but higher energy

reductions may be possible Future work: 1. implement sleep mode control for multi-link scenario based on OFDMA. 2. For transmit power vs delay tradeoff, define energy-efficiency metrics to find

the optimum sleep mode algorithm design

20 April 2009


Assumptions: 3GPP-LTE (estimated control signaling overhead 43% /PRB) SISO downlink system (20MHz bandwidth) low traffic (information data rate < 1.2 bit/s/Hz) Rayleigh fading channels (1 block length=1 TTI=1ms) Schemes: constant rate transmission: transmit R bits in each block water-filling: optimally allocate 3R bits within every 3 blocks to minimize power consumption (max delay-limit=3ms) awake 1/3: only use the best block in each 3 for transmission

Sleep Mode Control for Power Conservation �

20 April 2009


Relaying: Approaches from Sensor Networks Researchers have investigated network performance

including for energy constraints:   Rodoplu/Meng 2007 investigated the bit per joule capacity

of sensor networks to optimise number of hops   Cui/Goldsmith/Bahai 2004 investigated the most energy

efficient approach to applying MIMO/diversity techniques in sensor networks

  Sankar/Kramer/Manadayan 2008 investigated energy efficiency of different relaying configurations

  Youssef-Massaad/Zheng/Medard 2008 investigated transmission strategies on unit cost basis

Object is to develop these ideas and others towards wireless networks such as LTE


Initial Study of Energy Efficient Routing

Select multihop route to satisfy chosen routing metric, e.g:   Minimum relay transmit power   Minimum energy per bit   Maximise residual energy for mobile relay

Build a portfolio of such metrics Approach could be extended to general network selection

Source

Relays

Destination


System model: Energy efficient relaying

The simple three-node configuration consists of a source node S, a relay node R and a destination D.

(a) NC – BS Transmits with constant power

(b) NFC – BS Transmits in a bursty fashion i.e. instantaneous feedback and variable power

(c) CAC – Relay assists, two phases both constant power

(d) CFC- two phases but instantaneous feedback and thus variable power


Results


Assessing Energy Efficiency

Energy reduction of relay links difficult to estimate:   Wish to identify gains of energy efficient routing

approach   Also interested in gains of multihop per se →

ultimately an architecture issue   Need to specify correct baseline – single hop or

multihop   Need to trade off energy use at source terminal (base

stn or mobile terminal) vs relay terminals What can we say about gains of relaying

  75-90% power reduction on wireless links for cell edge users (30-50% of all users)

  Varying macrocell/relay density gives further architectural energy reductions


Femtocells inside Macrocells?

  Aim to develop operating curves for small vs large cell analysis.   Curves: different user density, RHS all femto i.e. macro switched off.   A lower user density in macrocell gives larger power saving (as energy

“overhead” is shared over smaller number of users).   The trade-off (minima) points for 30, 60, 120 and 240 users/macrocell

imply 80%, 60% , 40% and 20% respectively for femtocell coverage.   Corresponds to energy reduction ratio between 2.5 and 35%.

Embodied (Capex) plus operating (Opex) energy:

power consumption per user for different levels of

macrocell activity & support


  Initial Conclusions:   For femtocell and macrocell shared spectrum networks, the

performance of femtocells are seriously effected, if they are located close to a macrocell base station.

  When femtocells use different spectrum allocation from macrocell, femtocell’s adaptation can improve dramatically system throughput, especially for distant users at the macrocell edge.

  Simulation results examples

Examined area at the macrocell centre, 10 femtocell density scenario, femteocells use different spectral allocation from macrocell.

Examined area in the middle of a macrocell , 10 femtocell density scenario, femtocells use different spectral allocation from macrocell .


Spectrum Balancing

  Switching off networks (or parts of networks) at one or more frequencies, or reducing sectorization in networks, and reallocating users to the frequencies still in use, when experiencing low data traffic load.

  Transmission power saving by re-allocating links to a lower frequency band to improve propagation and reduce link loss, when transmission capacity is available in that band.


  Saving by disabling and changing transmission frequencies

Switching off networks (or parts of networks) at a frequency, or reducing sectorisation in networks, by reallocating users to other frequencies at times of low load

Transmission power saving by reallocating links to a lower frequency band to improve propagation whenever capacity is available at that band


  Performance of “spectrum balancing” against real operator hourly traffic loads over a 24 hour period defining means of Poisson distributions representing numbers of users in cell

  Vodafone 3G, London, and “LTE for UMTS” Chapter 9 traffic   Both show very significant potential power savings

Powering down radio network equipment by moving users between frequency bands

Transmission power saving by reallocating links between frequency bands to improve propagation


Conclusions

  Growth in data transmission requirements for mobile broadband will not bring major revenue increase.

  Every industry has published CO2 reduction targets and the mobile and IT communities are not exempt.

  Power drain in base-station or access point is the major issue in many wireless systems.

  Research challenges exist in system architecture and in techniques to deliver future, more efficient, Green Radio systems.


References

S. Armour, T. O’Farrell, S. Fletcher, A. Jeffries, D. Lister, S. McLaughlin, J.S. Thompson and P.M. Grant, "Green Radio: Sustainable Wireless Networks", Research brief at: http://kn.theiet.org/communications/green-radio-article.cfm, June 2009

Vodafone UK, Corporate Responsibility Review Report 2007/08

T. Edler, “Green Base Stations – How to Minimize CO2 Emission in Operator Networks”, Bath Base Station Conference 2008, kn.theiet.org/magazine/rateit/communications/Green_radio_file.cfm

H. Karl, “An overview of energy-efficiency techniques for mobile communication systems,” Technical University Berlin, Rep. TKN-03-XXX, September 2003. Online at: http://www-tkn.ee.tu-berlin.de/∼karl/WG7/AG7Mobikom-EnergyEfficiency-v1.0.pdf

F. Meywerk, Traffic / revenue curve from “The Mobile Broadband Vision - How to make LTE a success”, LTE World Summit, November 2008, London

S. Chia, “As the Internet takes to the air, do mobile revenue go sky high?,” IEEE Wireless Communications and Networking Conference, April 2008.

Ofcom Report: “Understanding the environmental impacts of communication systems.” http://www.ofcom.org.uk/research/technology/research/sectorstudies/environment/

H. Holma and A, Toskala, “LTE for UMTS - OFDMA and SC-FDMA Based Radio Access”, John Wiley, ISBN 978-0-470-99401-6, 2009


Relevant WWW Pages http://www.vodafone.com/etc/medialib/attachments/cr_downloads.Par.25114.File.tmp/CR

%20REPORT_UK-FINAL%20ONLINE_180908_V6.pdf

http://www.orange.com/en_EN/tools/boxes/documents/att00005072/CSR_report_2007.pdf

WWW.atheros.com/pt/whitepapers/atheros_power_whitepaper.pdf

http://www.mobile-computing-news.co.uk/industry-news/2372/gartner-report-shows-smartphones-growth.html

http://www.vnl.in/technology/sites/

http://re.jrcec.europa.eu/energyefficiency

http://www.femtoforum.com

http://www.cpri.info

http://www.nujira.com

Documents

Green Radio - Towards Sustainable Wireless Networks