CoolBaseStations - Energy efficient base stations of next generation mobile broadband networks Ralf Eickhoff and Frank Ellinger on behalf of the consortium Technische Universitaet Dresden Innovationsforum Software Saxony
Outline 1. Introduction and motivation 2. Project objectives and vision 3. Research tasks and challenges 1. Energy supply and management 2. Energy-efficient hardware systems 3. Lifetime studies and management 4. Network conditions and scenario constraints 4. Conclusions 2
Energy consumption in ICT Fast facts! Today s situation ICT at 3% of total energy (Gartner) (2% of global carbon emissions) EU: ICT at 8% of electrical power (2% of carbon emissions) Wireless ICT: 0.5% of total power Examples for broadband networks GSM network (Germany): energy consumption of a medium-sized city, e.g. Ludwigshafen (160 000 inhabitants) Orange network (France): 280 GWh (81 % data, 12% voice, 2005!) Pollution of ICT > air traffic C0 2 pollution in Gt/year 50 40 30 20 Total Air Traffic 10 Data Centers 0 Source: Die Zeit, 2008 2.0 1.6 1.2 0.8 0.4 0.0 3
What is your expectation? Doubles every 5 years Expert expectations Rise to 1.43 Gt CO 2 in 2020 (5% annual growth rate assumed) 25% telecoms infrastructure 18% data centers 57% PCs and peripherals Potential of 7.4 Gt CO 2 abatements (Smart 2020, GeSI) Cellular Customers (Mio.) 3.000 2.000 1.000 Source: GSM World 1996 2008 4 2007
Base stations: road blockers Today s situation! 1. One half of total power dedicated to power amplifier losses Relaxed linearity of standards 15% average energy efficiency 2. Energy into heat conversion 3. Energy for cooling needed Defined operating conditions Autarkic operation defined by Generator sets Battery packs Performance, efficiency degradation Lifetime cycles Harsh operating conditions RF power (540 W) Input power to 12 power amplifiers (1440 W) Input power to 12 transceivers (1680 W) Input power to base station (1780 W) 2kW Power amplifier losses (900 W) Digital parts (108 W) De-heating (60 W) Antenna network controller (30 W) Station unit (10 W) Source: Alcatel-Lucent study Analogue parts (132 W) 5
Project vision Let s decrease energy consumption of base stations to enable regenerative and autarkic supply in mobile broadband networks! CoolBaseStations, January 2010 6
Partners and research tasks Levers Partners Improve integration of power amplifiers for femto-cells Use energy-aware methodologies during circuit design Device models for harsh operating conditions Investigations of solar cells and panels Alternative supply by fuel cells for less cooling IKTS Optimize power supply switching Optimize system integration and life-cycle mgt. Analysis of network operating conditions 7 CoolBaseStations 23.04.2010
Energy supply by solar cells Vision for ICT in general 8
Energy supply by solar cells The facts! The Challenges! DC power base station: P BS 1.5-2 kw (GSM only) Available solar energy: 115 W/m 2, e.g. in Saxony Efficiency of (industrial) solar-cells: η 16% Available energy for supply: E/A solar-cell =18 W/m 2 Installation area at base stations: A < 15 m 2 Maximum power: P solar-cell < 270 W 20 % of base station supply might be feasible, but Base stations must support UMTS, LTE (advanced), etc. Not many control parameters, e.g. improved efficiency of cells 9
Limits on efficiency, examples Efficiency (crystalline silicon) Mass production: η solar-cell 16% Scientific demonstrations: η solar-cell 25% Reduce major losses Resistance losses Contacts, wires Multiple dimensions Optical losses Shadowing, reflections Anti-reflection Recombination Volume, surface Passivation, amorphous Si 10
Supply power switching Power factor improvements Purely resistive systems: PF 1 possible But wireless networks: P Varying load conditions real Nonlinear effects Device parasitics PF «1, e.g. typical: PF= 0.6 0.8 Development of circuits with PF > 0.95 (active PFC) Adaptively with regard to changing conditions (load, V, I) Highly integrated and low-cost solution Possible savings with PF = 0.95 = 48TWh/a = Required energy of 13.6Mio. households real PF P P reactive 11
Adaptive MPSoCs Energy-efficiency of signal processing platforms Energy management integration in hardware Load balancing together with fine-grained voltage/frequency scaling Runtime manager task scheduler in hardware (energy aspects) Configurable on-chip network regarding high speed and energy efficiency Extended design methodology for energy aspects Integration in 45nm CMOS 12 CDB method compared to standard design flow Area: 58% Power: 70%
Power amplifier integration RF power amplifier aspects Using silicon technologies rather than III/V semiconductors Major challenges due to technology scaling Low supply voltages, low breakdown voltages of devices Nonsufficient P RF ~ V dc ²/R L Device parasitics Power adding by coupled N:1 transformers Voltage added, higher R L and lower matching losses possible Switching of cells feasible to improve efficiency in amplifier back-off R L Amplifier cell V in 13
Life-cycle-management and system aspects Base station life cycle analysis Energy supply and energy management/distribution Parameters of electronic devices can change over time Efficiency degrades versus time, 3-10 years life time expected Life time management to preserve efficiency and minimum costs System integration concepts Enhanced system integration Energy chain management Alternative and hybrid power supply Component reduction and connections Heat transfer and cooling analysis Towards low-power and low-cost femto-cells 14
Conclusions ICT s carbon footprint increases but potential for CO 2 savings Road blockers in base stations for an autarkic supply vision RF circuits, components, and VLSI systems Energy supply and energy management Network infrastructure and base station placement CoolBaseStations energy efficiency leverages Energy-aware circuit design flows Improved components and systems Energy-aware system integration and management Monitoring and network conditions 15
Thank you very much! FH Stralsund HTS Hochtechnologie Systeme GmbH 16