T Seminar on Embedded Systems. Internet of Things Ambient energy harvesting Mikko Lampi

Transcription

1 T Seminar on Embedded Systems Internet of Things Ambient energy harvesting Mikko Lampi 1

2 Internet of Things Early precursors from -90 by IBM and Motorola Nebulous term, many interpretations As a term first coined by Kevin Ashton 1999 Sea of addressable nodes Both powered and passive 2

3 Taxonomy of different IoT installations Trivial 1-10 sensors or actuators Up to 10 meters of cable, Fits into a briefcase Power from the central unit Data from the power cable 3

4 Taxonomy of different IoT installations Classic sensors or actuators 80 m cable reel Can be carried, but not far Power can come from a central system or from local sources Data can still use copper, but it is getting hard. 4

5 Taxonomy of different IoT installations True IoT M sensors or actuators A truckload of cable Even pulling power from each AC socket would fill the rooms with transformers Probability of critical cabling failure almost a certainty Probability of crossing potential wells is almost guaranteed. 5

6 Power requirements of IoT A hedge of compromises Powerful local processing means less data traffic High speed data traffic means processing can be done elsewhere If the nodes have to shout long distances the energy cost goes up If the nodes can be chained or used as a mesh network the energy needs can be distributed between multiple units 6

7 Processing per Watt Intel Core Intel Atom OMAP3430 Microchip i MHz 95 W D MHz 10 W ARMCortex8 650 MHz W PIC18F86J60 42 MHz W Active sleep mode W 7

8 Communication per Watt 3G 384 kbps2 W GPRS 1 W WiFi W Bluetooth W Zigbee W 8

9 How can we power IoT Power source can't be large AA-AAA Classic battery V 4.2 Wh Self discharge negligible NiMh battery V 0.96 Wh Self discharge 0.03 Wh per month Empty in 3 years. Usually much faster Li-Ion battery is too dangerous Micro Fuel cell is still not available 9

10 Expected lifetimes Best options ARM processor and 8-bit microcontroller both using Zigbee ARM 383 mw with 1 mw Zigbee 4.2/0.384= 11 h 0.5 d PIC18F86J60 32 mw with 1 mw Zigbee 4.2/0.033= 127 h 5 d PIC18F86J60-Timer mw with 0.1 mw 4.2/ = h = 1411 d 3-4 y 10

11 Expected lifetimes Changing one million batteries, or even ten thousand batteries every three years is infeasible. No matter the processor technology, no matter the algorithmic solutions and no matter the wireless solution true IoT can not be built using batteries. 11

12 Ambient energy harvesting The only road forwards is that the True IoT nodes have to be able to feed themselves. Nodes have to be able to gather sufficient energy for operations and data communications from around themselves. From the earlier calculations we can conclude that the low end of power usage is at 124 µw. Need to find ambient power sources that can fulfill at least that. 12

13 Solar cell Small cell outdoors 150 mw Needs a bigger battery to help survive nighttime or cloudy days Shiny. Tends to get stolen Has to withstand cold, rain and dust Hard to find a shadow free spot 13

14 Small indoor solar cell Small cell indoors 51 µw. Three of these and we hit our target Very small power indoors Depends on windows or lamps to illuminate it Can run out of energy during summer breaks or other times when there is no indoors activity 14

15 Magnetic coupling Magnetic coupling with a nearby power line High power on top of a power line Can easily feed a high power processor Even 10 cm distance from the cable diminishes the power radically Unless a device is not using the power line there is no magnetic field to harvest 15

16 Harvesting radio signals Usual case is 10 m from antenna gives about 4 µw with a smallish (10x5 cm antenna). Not enough unless we either use a stronger signal or larger antenna With large antenna we start creating radio shadows behind the antenna Very hard to create a design that won't create problems for other nodes by damping their transmissions 16

17 Vibration magnetic capture 0.85 mm movement produced 180 µw Excellent and reliable energy source Energy can be collected from trucks passing over a bridge Or from a building swaying Usually includes moving parts that tend to wear out over time 17

18 Strain piezo capture A Human walking over a large piezoelectric plate can easily produce 8 W If we expect the human to pass the plate in half a second producing 4 J and our energy needs are J per hour. We will need one human every 9 hours to pass the plate to keep the IoT device running. This is an excellent source of power in human rich areas 18

19 Thermoelectric (Peltier) Seiko wristwatch produced C from human body heat Scaling up by 10x surface or 10x temperature difference we can easily get 220 µw Both buildings and clothes usually designed to minimize large temperature differences 19

20 Air movement Piezo based wind turbines Can be inside ventilation ducts Can use natural currents inside buildings A small device can easily get 200 µw from random air movements. 20

21 Conclusions True IoT can not be built using wires True IoT can not be built using batteries Processing and communicating at µw levels is possible and practical Powering from ambient is possible Most likely sources are vibration and light 21

22 Thank you Any questions? 22