Institut für Telematik | Universität zu Lübeck
Wireless Sensor Networks
Chapter 3: Hardware Stefan Fischer Dennis Pfisterer
Motes
4
Rene (1999) Mica (2001)
Telos (2004)
Developed at UC Berkeley Many incremental developments: Rene, MICA, MICADot, MICAz, Telos, ...
Many variants worldwide BTnode, Eyes nodes, MANTIS, XYZ, Scatterweb, iSense, ...
BTnode (2003)
Motes: Components
Processor CPU (few MhZ), RAM (1-100 KB), I/O
Communication 10-500 kbps, 10-1000m range
Sensors
Energy supply Battery, Energy scavenging (Solar, Vibrations) Stabilization of energy supply (e.g., to 2.4V)
External Storage Flash (100-1000 kByte)
Real-Time clock 20-500ppm
5
Selection of individual components
Huge variety of options for each component Selection criteria?
Important factors
Lifetime (energy consumption) Performance / Accuracy Robustness Form factor Cost
Depending on the application
6
FleGSens
7
FleGSens: „Sichere und flexible Grenz- und Liegenschaftsüberwachung durch drahtlose Sensornetze“ Funded by the Federal Office for Information Security (BSI) Secure trespasser detection using WSNs
Heiligendamm, Germany, G8 Summit 2007
Secure trespasser detection using WSNs
Wireless Sensor Node
8
Coverage of the passive infrared sensor (PIR)
FleGSens Project Goals
Goal: Provide IT-Security despite the scarce resources in WSNs Monitor an area of 500m x 30m with 200 nodes
Report trespassers at a base station within 5s and a precision of 10m Network lifetime > 1 week on batteries
Tolerate 10% node failure and up to 5% compromised nodes Demonstrate scalability using simulations (>2000 nodes, ~2km x 60m)
Partners Institute of Telematics, University of [Lübeck, Karlsruhe] coalesenses GmbH
9
Processor
Alternatives Microcontroller - Integrates CPU, Program memory, RAM, I/O-ports, serial/i2c/spi interfaces, analog/digital-converters, …
DSP (Digital Signal Processor) - Optimized for signal processing, e.g., as a co-processor
FPGA (Field-Programmable Gate Array) - Programmable logic gates, e.g., as co-processor
ASIC (Application-Specific Integrated Circuit) - Customized for a particular use, rather than for general-purpose use
Examples of Microcontrollers Texas Instruments MSP430 - 16-bit RISC core, up to 8 MHz, up to 10 KB RAM, multiple ADCs
Atmel ATMega 128 - 8-bit controller, up to 8 MHz, 4 KB RAM, multiple ADCs
Jennic 5139 (used for the iSense devices)
10
Example: Jennic 5139 microcontroller
3-11
Communication
Medium Radio, Light, Sound
Properties
Frequency range? Wide or Narrow band communication? Multiple channels? Data rate? Range? Interface: Bit, Byte, or Packet? Indication of received signal strength?
Energy Energy consumption for sending, receiving, and listening? Duration and energy consumption for state changes? Adjustable sending power?
12
Examples: Radio Modules
RFM TR1000
Chipcon CC1000
Frequency range: 916 / 868 MHz Data rate: up to 115,2 kbps (typ.: 19.2 kbps) Sending power: 36 mW Range: 10 – 200 m Interface: Bits Frequency range: 300 - 1000 MHz, selectable in in 250 Hz steps Data rate: up to 76 kbps Sending power: 42 mW Range: 10 – 100 m Interface: Bytes
Chipcon CC 2400
802.15.4 (“Zigbee”) Frequency range: 2.4 GHz (amongst others) Data rate: up to 250 kbps Sending power: 38 / 35 mW Range: 50 – 125 m Interface: Pakets
13
Sensors
Categories Passive vs. Active (Emitting a measuring signal) Directed vs. omnidirectional
Examples Passive, omnidirectional - Temperature, Microphone, Humidity, … - Chemical properties, Gas
Passive, directed - Light, Movement detection, Camera
Active, directed - Radar
Additional properties Analog vs. digital Calibrated vs. uncalibrated Coverage: Which region is covered by a sensor?
14
Energy Supply
Options (Rechargable) Batteries Goldcaps, Supercaps: Capacitors with high capacity (1-1000 F)
Energy Scavenging (Movement, Sun, …) Hybrid Systems
Requirements Low self-discharge - LiIon: 2-3%, nickel cadmium15-20%, nickel metal hydride 30% (per month).
Rechargeable (Time, Max. number of cycles) Energy Scavenging
Voltage Stability (under load, discharge curve) - Try to avoid voltage stabilization circuits (degree of efficiency)
15
Energy
16
1 J = 1 Nm = 1 kg m2 / s2 1 J = 1 Ws 1 cal = 4.2 J The amount of energy required to increase the temperature of 1g of water by 1 Kelvin
James Prescott Joule 1818 - 1889
Energy Density
17
Energy per Volume (Joule per cm3): Primary batteries Chemistry
Zinc-air
Lithium
Alkaline
Energy (J/cm3)
3780
2880
1200
Secondary batteries Chemistry
Lithium
NiMH
NiCd
Energy (J/cm3)
1080
860
650
Supercaps: up to 17 J/cm3
Energy Scavenging
Gain energy from the environment by converting energy to electricity
Examples Fuel Cell: 10 – 100 mW / cm2 Light (Solar Cells): 10 W/cm2 - 15 mW/cm2 Temperature difference: 80 W/cm2 @ 1 V with 5K Difference Vibrations: 0.1 - 10000 W/cm3 Pressure (piezoelectricity): 330 W/cm2 (Shoe) Radioactive sources?
18
Solar Power Harvesting System Lithium-ion rechargeable battery
Power source (solar panel)
iSense Power Management Module
Sensor Node
Average delivered current [mA] 80
Average delivered current [mA] Lüb…
70
80
60
60
50
50
40
40
30
30
20
20
10
10
0
Bueno Aires Iguacu
70
0 Jan Feb Mar Apr May Jun
Jul Aug Sep Oct Nov Dec
Jan Feb Mar Apr May Jun
Estimation based upon average daily hours of sun
Jul
Aug Sep Oct Nov Dec
Energy consumption
Typical energy consumption Energy per instruction: 1 nJ (ca. 100ns) Transmit one byte: 1 uJ (ca. 30 us) One byte in flash: 3 uJ (ca. 80 us)
Calculate vs. communicate 1000 Instructions ~ 1 Byte Calculate locally (reduce amount of data) instead of communication
Expected lifetime LiIon Battery (few cm3): 10000 J Processor: ca. 12 days Communication: ca. 4 days Flash: ca. 4 days
This is not very long…
20
Solution: Duty cycling
Duty cycling Long sleep phases (typically > 99%) - Many components are switched off or in a sleep mode
Fast wake up - Reactivate components (costs time and energy)
Short wake phases - Do some work, quickly go back to work
Minimize the integral over energy over time Radio, Processor, ...
Achievable life time: several 100 days
21
Wake up Mikrocontroller
22
Radio
292 ns
10ns – 4ms typical
2.5 ms 1– 10 ms typical
Example: Optimized Duty Cycling [1]
3-23
[1] CUPID - Communication pattern informed Duty Cycling in Sensor Networks, Daniela Krüger, Dennis Pfisterer, Stefan Fischer, Submitted to IPSN 2010, ACM/IEEE International Conference on Information Processing in Sensor Networks
Example: Telos Tmote Sky
Mikrocontroller Texas Instruments MSP 430 2 KB RAM, 60 KB Flash
Radio Chip Chipcon CC2420 (IEEE 802.15.4, „Zigbee“) 250 kbps, 50-125 m
Sensors Temperature, Humidity, Light Extension connector
24
Example: Telos Tmote Sky
25
Example: Telos Tmote Sky
26
Motes Compared
27
Miniaturisation
Until now: Standard components „COTS Dust“
Vision „Smart Dust“ Sensor nodes in the size of a dust particle?
Strategies Efficient packaging System-on-Chip Micro-Electro-Mechanical Systems (MEMS)
28
System-on-Chip
All functions on one chip Minimal external circuitry
29
SoC = Smart Dust?
Antennae Simple Monopol: Length /4 At 1Ghz -> ca. 10cm Higher frequency -> more energy
Batteries Energy density does not scale
Largest issue: radio chip Complex signal processing High output power
30
Alternatives to radio waves?
Laser Very focused: large distances with only few output power - Several km with 5 mW output power
Simple optical receiver: few energy “Antennae” are very small - Photodiode (integrated in SoC)
31
Smart Dust
32
Mobile Smart Dust?
33
Whats up for the future?
34
Moore‘s Law: Number of transistors doubles almost every two years It is believed that this trend lasts for more than 10-15 years Impact - Less costs, less energy consumption
Fundamental technological improvements New energy sources - Tiny fuel cells
More efficient radios - Goal: < 100 uW Piezo
V
