1

Energy Harvesting Devices

Davide Brunelli
davide.brunelli@disi.unitn.it
Department of Information Engineering and Computer Science
DISI University of Trento
The nightmare of pervasive embedded computing:
Power avalaibility
Ubiquitous computings dream of pervasive sensors and electronics
everywhere is accompanied by the nightmare of battery replacement y p y g y p
and disposal.
No Moores Law in batteries:
2-3%/year growth
Battery Technology is Stuck!
ES lifetime depends
on battery life!!
2
Limits to Battery Energy Density
Processing power doubles every 2 years, but
Battery capacity doubles every 10 years
We need a more efficient way to enable longer life
Energy Density by Mass (MJ/kg)
2004 - Lithium ion at its
current max
1991 - Lithium Ion battery
released
1899 - NiCd battery created
y g
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
TNT
2012 - Nanowire-based
lithium ion battery
Research in progress
[TI09]
Available Energy is All Around
Light
EM
Technology trend:
Design systems that harvest limited energy from ambient (heat, light, radio,
or vibrations) or scavenge power from human activity
Light
M ti d
EM waves
H t Motion and
vibration
Heat
3
Energy Harvesting Basics
Energy harvesting is the process by which
energy is captured and stored
Energy Harvesting shrinks or replaces batteries or
extents recharge periods
This term often refers to small
autonomous devices micro energy
harvesting
Power output of Energy Harvesting transducers is
linked to their size (area, volume) and thus to their
price
Power addresses matching of loads and of
transducers and aim at the maximum energy output
Setting expectations
[Van Hoof HOLST10]
4
The Good News
WSN
Mobile terminals
Batteries
Todays
Scavengers
scavenger-aware
design
Scavengers
The gap between scavengers energy and requirements of digital
systems is shrinking [Paradiso05]
E l i i d i i
scavenger evolution WSN evolution
Todays WSNs
Exploit energy management strategies and improvements in scavenger
technology
Overcome traditional energy management strategies (battery-driven)
An new unified design methodology is required
Smart adaptation
Design for unreliability
Exploit unpredictable power sources
20W 20W 40W 20W
Avg.
Power
80 Mops 2nJ/b
Sensor Node Evolution
<10kb/s
1%
DSP&storage
Security
MAC
RF Non-E
World
Sensor
CE-ADC
Processor
PicoRadio
p
Power Mgr
Energy Harvester
Objective: 100 W Avg Energy neutrality becomes easy
Power Mgr
Ambient energy
5
Where we are now
1W Harvesters Consumers
Average Power
Cell phone
Zigbee mesh network node
Energy Harvesting Power Generation & Utilization
100 mW
10 mW
1 mW
Large inductive vibe harvesters
1 in
2
TEG on crease beam
TEG stringer clip
1
2
Si PV i bl k
1 cm
2
a-Si PV
in sun lit airplane pax window
Wireless dimming window
AAA LED flash light
Wireless sensor @ 1 Hz Push button harvester
Zigbee mesh network node
(w/ Rx from wireless sensor)
TI MSP430 microprocessor (awake)
Chipcon CC2500 radio (Tx mode)
6 mm
2
TEG on hydraulic line
100 W
10 W
1 W
Small piezo beam vibe harvesters
1 cm
2
a-Si PV in cabin lighting
1 cm
2
a-Si PV in blue sky
Push button transmitter
Sensor @ 2.8 hrs interval
GSE monitoring sensor
(log data every 10sec, Tx 2X per day)
TI MSP430 microprocessor (asleep)
Chipcon CC2500 radio (asleep)
TEG=thermoelectric generator
Energy Harvesting: new design methodology
Hardware Design
Conversion efficiency
Impedance Matching
Natural progression of Energy
Optimization Techniques
Impedance Matching
Maximum power transferred

Software Design
Scheduling algorithm
Adaptive duty cycle
Low Power Design
Power Aware Design
Battery Aware Design
Adaptive duty cycle
Energy prediction algorithm

Energy Harvesting
Aware Design
Why is it different?
6
Energy Source Characteristics Efficiency Harvested Power
Light
Outdoor
Indoor
10~24%
100 mW/cm
2
100 W/cm
2
Human ~0 1% 60 W/cm
2
Energy Harvesting Sources
Thermal
Human
Industrial
~0.1%
~3%
60 W/cm
~1-10 mW/cm
2
Vibration
~Hzhuman
~kHz
machines
25~50%
~4* W/cm
3
~800 W/cm
3
RF
GSM 900 MHz
WiFi
~50%
0.1 W/cm
2
0.001 W/cm
2
1uW 10uW 100uW 1mW 10mW
Seiko watch
~5uW
2 channel EEG
~1mW
100mW 1W+
AdaptivEnergy
~10mW
~30mm
Holst Center
~40uW
BigBelly
~40W
Elastometer
~800mW
Energy from Human daily activity
Thad Starner, Human-Powered Wearable Computing, IBM Systems
Journal 35, pp. 618-629 (1996).
7
Effective, long term, power supplies are limited and/or expensive
Example: At an average power
consumption of 100 mW, you need more
than 1 cm
3
of lithium battery volume for 1
year of operation.
Environmentally powered wireless sensors
Airflow
Goal
Investigate energy harvesting and
management technologies that can
support the operation of a smart
sensor nodeindefinitely
year of operation.
PV
Inductive
PV
Kinetic
RF
Contacts: Telecom Italia, STM
EH powered nodes
philosophy
Input
t ti
ADC
CPU
Wireless
EH-management
Independent
Load
Interface
Mi d A hit t
Energy Harveter and
S t P U it
Sensors
protection
Switch
Supercapacitor Battery
Switch
Ref2 Ref1
Supercapacitor Battery
Mixed Architecture Smart Power Unit
General purpose
Optimized from Ambient Source and storage,
but not for a specific application
Plug-&-play
Analog or with Digital Interface for external
power management (standardization?)
Usually more efficient
Tailored on a specific application
HW /SW dependent
8
Design Methodology
Generic Approach
Dedicated blocks, depending on energy source, ambient
conditions and application
Not all are required in any application and with any source Not all are required in any application and with any source
Rectifier, DC-DC converter and MPPT are the most challenging
and require a very accurate design process
Charger/limiter/protection consumes additional power and are
often to some extent redundant.
Ambient
Energy
Energy
Trans-
ducer
Rectifier
MPPT
DC/DC
Charger/Protection
Storage
DC/DC Load
Ambient Energy
Non-monotone, Unpredictable
Ex: solar power (PV-cells)
Ex: power waveform from
human walk (piezo-scavengers)
Too much
Too little
Aperiodic
18
[Paradiso05]
Aperiodic
9
Challenges for Harvested power
management
Changing polarity input
Low input voltage (e.g some mVs)
AC input with variable frequencies
Several AC inputs
Sources with variable resistance (depending on
temperature and aging)
High dynamic range of input voltage
Rectifier
Energy is usually available with dual polarity voltage
Design choices:
Simple Diode Bridge (Vdrop ~1,2V)
Active mosfet Bridge (Vdrop ~0,4V)
(needs Input Polarity Detector)
Dual circuit topology
(No Vdrop, at the cost of size and complexity)
10
Rectifier
Active Mosfet Bridge
Diodes can be short-circuited by switches to
prevent degrading efficiency from the p g g y
forward voltage drop
Typical values:
Start-up 150mV,
drop 40 mV, later on 5 mV
Diodes are only active during start-up
h till l lt f th where still no supply voltage for the
comparator
Maximum Power Point Tracking
Maximum power from source to load when internal resistances are
matched
Input resistance of a DC-DC converter is influenced with its duty
cycle
Ideal situation:
Load RL and Internal resistance
Ri are naturally matched
Vsupply in the correct range
Typical situation:
DC/DC with MPPT to match Rl and Ri
and /or to adjust Vsupply
11
Maximum Power Point Tracking
- an example-
120
140
I-V chart
0
20
40
60
80
100
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5
V[Volt]
I
[
m
A
]
I

[
u
A
]
V [Volt]
300
P-V chart
1500
2000
January 25, 2007 Ing. Davide Brunelli 23
0
50
100
150
200
250
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5
V[Volt]
P
[
u
W
]
P

[
m
W
]
V [Volt]
0
500
1000
0 150 300 450 600 750 900
T (s)
V
c

(
m
V
)
10.9J 15.7J
Vsolar
Vctrl
Vsolar
Vlow crossing switch off
Vhigh crossing switch on
MPP Regulator
P
Vtransducer
Controlled variable
V
low
, V
high
duty cycle
V
low
V
high
Online control for tracking Transducer curve variations
12
MPPT Techniques
MPPT Techniques depend mainly on the transducer and the ambient
energy
Most common techniques of MPPT employ DSPs or microcontrollers, not
suited for Energy Harvesting suited for Energy Harvesting
Simpler solutions employing only analog circuits sometime have smaller
performance
For e photovoltaic cells with Fractional Open Circuit Voltage: Photovoltaic panels
output voltage that allow to drain the maximum amount of power correspond at
about the 70 % of the open circuit voltage.
KOFCV = VOC/Vmpp ~ 0,71-0,75
MPPT Techniques
Energy storage required
Intelligent, adaptive power management ensures maximum power
output output
Switching frequency is fixed and depends
on circuit parameters and components.
Maximum Power Point Tracker duty cycle
Climb the Hill !!
is controlled and output power
measured
Increasing output power: duty cycle is
changed further in the same direction and
vice versa
13
Buck: V
o
<= V
i
DC/DC
-Typical Power Converter Topologies-
Boost: V
o
>= V
i
Buck-boost: V
o
<=> -V
i
Steady State Transfer Function - Buck
Continuous mode
Discontinuous mode
14
Steady State Transfer Function - Boost
Continuous mode
Discontinuous mode
Steady State Transfer Function Buck-Boost
Continuous mode
Discontinuous mode
15
Microsystemwould not operate when charged from zero voltage.
Microprocessor drew significant amount of power when attempting to
initialise at 0.9V, system locked in perpetual loop.
Start-up problems
Possible Solutions:
To guarantee a charging path even if storage device is depleted.
Voltage level detectors which do not allow the microsystem to boot (or to
start) until supply is above 2V.
controller
With Cold Start circuit
controller
MOSFET
XC61C Supercap
Without Cold Start circuit
Example from Perpetuum Inc. (VIBES project)
Startup
-example-
Vctrl=0V
Charging curve Efficiency
16
Energy Harvesting Storage Required
Scavenged energy is not constant
Power not available on-demand
High peak power not available
An ideal energy storage device:
Infinite shelf life
Negligible leakage
Unlimited capacity p y
Negligible volume
No need for energy conversion
Efficient energy acceptance and delivery
Ideal battery doesnt exist
Energy Storage Technologies
Options
Secondary Batteries
Capacitors
Supercapacitor
Tradeoffs
Configuration
Tiered Capacitor+Battery.
Battery-only, Capacitor-only
p p
THF Batteries
Fuel cell
Tradeoffs
Batteries
Mature technology, high energy density, less efficient, limited to few
hundred full recharging cycles (significantly more shallow cycles)
Ultracapacitors (up to hundreds of Farads)
Virtually infinite recharge cycles, higher leakage current (goes up with size)
Energy Reservoirs will still play an important role
17
Charge Termination Methods
Lead Acid Nicad NiMH Li-Ion
Slow Charge Trickle OK Tolerates Trickle Timer Voltage Limit
Fast Charge 1 Imin NDV dT/dt Imin at Voltage Limit
Recharging Issues
Fast Charge 2 Delta TCO dT/dt dV/dt=0
Back up Termination 1 Timer TCO TCO TCO
Back up Termination 2 DeltaTCO Timer Timer Timer
No general purpose method
E.g. Lithium batteries have:
wide voltage operating range
thi k t d t i th d f h thick range to determine the end-of-charge
and undercharge
Mature Energy Storage Options on the market
Micro-power storage
Li-Ion
Thin Film
Rechargeable
Super Cap
Li-Ion
Capacitor
Recharge Cycles 100s 5k-10k Millions Millions
Self Discharge Moderate Negligible High Moderate
Charge Time Hours Minutes
Sec-
Minutes Minutes Charge Time Hours Minutes Minutes Minutes
SMT & Reflow Poor-None Good Poor Poor
Physical Size Large Small Medium Large
Capacity
0.3-
2500mAHr 12-700uAHr
10-
100uAHr 10-1600mAHr
Environmental
Impact High Minimal Minimal Minimal
18
Looking forward: Fuel Cell
Membrane splits electrons off hydrogen
Electrons recombine with proton on other side in
catalyzed reaction w. oxygen to form water y yg
Photo showing conceptual Motorola/LANL fuel-cell-phone
Fuel in electricity, and exhaust out
Anode Cathode
Fuel Gas Temperature.25(C); Air Breathing;
PCB Mini Fuel Cell
Fuel Gas Pressure.Ambient;
H
2
Flow Rate....0.030(slpm);
Relative Humidity.................100 %;
Max Power Density:
282 mW/cm
2
Power : 1 W (0,52 V @ 1,94 A)
19
Managing harvested energy
It is different from battery energy
Supply varies with time
Need to adapt performance
Supply varies in space
Different nodes get different energy: need load sharing
Supply is repetitive (does not die out)
Opportunity to last forever
Efficiency concerns
Match load to maximize transfer
Supply direct when possible, instead of through battery
Harvesting-Aware Policies
20
Tasking aware of battery status & harvesting opportunities
Richer nodes take more load
Looking at the battery status is not enough
L h i
Harvesting-aware Management
Learn Local Energy
Characteristics
P di t F t
Distributed
Decision
Topology
Control
Learn the energy environment
Predict Future
Energy
Opportunity
Learn
Consumption
Statistics
Decision
for
Scheduling
Routing
Clustering
Energy harvesting Electronic System Design
What is different in Software and Firmware development?
[Sunergy: June 2007]
Conventional energy management: How do we save energy ?
Energy harvesting management: When do we use energy ?
Determine an optimal on-line scheduling of activities:
If the set of activities is schedulable, it determines a feasible schedule.
Determine decisions on the application level that
optimize the long term system behavior
21
System Reconfiguration
Environmental energy is variable (solar power, vibrational
microgenerators, thermal scavengers)
Types of reconfigurations yp g
SW:
Lazy scheduling [Brunelli06], adaptive power management [Kansal06,Moser07],
Game theoretic approach to determine sleep/wake-up schedules [Nihato07]
HW: Reconfiguration through FPGA [Nahapetian07,Susu07]
Concept
Exploit period of light to reconfigure system to execute next
tasks with less power
Statistical energy availability estimation to decide about
reconfiguration
maximize the work done adapting to the available energy profile
energy source S
Lazy Scheduling: Model
Task Ji
can be preempted
energy storage
computing
device
P
S
(t)
P
D
(t)
D
E
C
(t) C
ca be p ee pted
arrives at time ai
has deadline di
needs total energy ei
to complete
can consume power
tasks
a
1
, e
1
, d
1
a
2
, e
2
, d
2
J
1
J
2
therefore, needs time
22
Greedy
When do we use energy ?
2 1
Greedy
scheduling is not
suited.
1 2
ALAP does not
work either.
And what happens if the energy storage is full?
Greedy
When do we use energy ?
2 1
Greedy
scheduling is not
suited.
1 2
ALAP does not
work either.
And what happens if the energy storage is full?
23
Lazy Scheduling Algorithm
Rule 1:
ai di
ei
t si
d
i C
i i
p
t d t E C
d s
)) ( ) ( , min(


Lazy Scheduling Algorithm
Rule 2:
ai di
ei
t
si
24
LSA example
a1 d1
s1 s2 a2 d2
t
Features
Start time Si can be computed once when the task is scheduled
Energy is not wasted on task that cant be finished
Admittance Test
The proof uses
P
max
concepts of network
calculus
and real-time calculus.
25
Performance
* EDF
LSA
X axis = max Capacity
Y axis = time of the first overflows
Capacity savings of ~40% measured for
random task sets for LSA with
l
()
compared to EDF
Conventional energy management: How do we safe energy ?
Energy harvesting: When do we use energy ?
Energy harvesting Software Design
What is different?
If sensor node is not OS equipped:
Determine decisions on the application level
that optimize the long term system behavior
Determine decisions on the application level
that optimize the long term system behavior
sensing rate receive messages
data transmission forward messages
26
Conventional energy management: How do we safe energy ?
Energy harvesting: When do we use energy ?
Energy harvesting Software Design
What is different?
Determine decisions on the application level
that optimize the system behavior
Determine decisions on the application level
that optimize the long term system behavior
sensing rate receive messages
If sensor node is not OS equipped:
minimal sensing rate
reactivity
freshness of data
average throughput
data transmission
forward messages
Principles: Model predictive control
Model predictive control is the class of advanced control techniques
most widely applied in the process industries.
The main idea of MPC is to choose the control action by repeatedly y p y
solving on line an optimal control problem.
MPC is based on iterative, finite horizon optimization of the system
under control Receding Horizon Control
MPC :
M d l A d l f th ( t ) d t l i i d Model A model of the process (system) under control is required.
Predictive Optimization is based on the predicted evolution of the model
Control It is usually adopted for complex systems
(Multi-Input Multi-Output)
27
Principles: Receding
Horizon Control
Two Steps
At time k, solve an open loop optimal control
problem over a predefined horizon and apply problem over a predefined horizon and apply
only the first input (i.e. control law for time
k+1)
At time k+1 repeat the same procedure. (The
previous optimal solution is discarded!)
MPC is like playing chess!! MPC is like playing chess!!
Prediction of opponents moves
Optimization of outcome a few moves ahead
An unexpected move from the opponent =
change of strategy!
Good players thinks several moves ahead =
long prediction horizon!
Principles
Optimization problem: finite horizon control
t
current time t
current state (memory, battery, )
current environment (input power)
28
System Model
environment environment
run-time
platform
Models for application, quality/utility, system behavior ?
Optimization problem ?
Efficient run-time implementation ?
Principles
Optimization problem: finite horizon
control
Example: Linear program for Example: Linear program for
sensing/transmitting optimization
Rate of acquisition
Memory usage
Stored energy
Used memory
Final stored energy
29
Principles
Efficient run time implementation
Solving a linear program in a resource-
constraint sensor node at each time step ?
Efficient run-time implementation
Approach Solve the LP as a parameterized LP and
implement the explicit solution [Morari, Bemporad et al.]:
The optimal controller is a set of controllers with an affine
controller selection rule
Desgin issue: limiting the number of different controllers
Preliminary results on highly constrained CPU are promising
Different control laws in different regions of the state space!
Simulation and Experiments
Example 1
sensing rate control
minimize interval
between samples
Example 2
rate control with
memory buffer
minimize interval
between samples
minimize amount
of stored data
Gain:
56,8 %
30
Distributed energy management
How can a distributed system manage the harvested
energy to maximize performance of system as a
whole?
Energy resources vary across nodes,
Task-load differs at different nodes,
some workload is share-able while some is not
Consider one energy intensive task: routing data
Determine environmental energy aware communication
strategy strategy
Routing paths can change depending on energy
availability
However, how to distribute this information?
Distributed algorithms with low messaging overhead are
required
EH aware routing
EH routing must be able to exploit nodes with
high energy intake and take into account distance high energy intake and take into account distance
between nodes
[Lattanzi06]
31
Case Studies
Electrostatic Electromagnetic Piezoelectric
Vibrations
Case Study
More easily
implemented in
t d d i
Typically output AC
voltages is below 1 volt
i it d
The output voltage is
irregular and depends
th t ti standard micro-
machining processes
Requires a separate
voltage source (such
as a battery) to begin
the conversion cycle.
in magnitude
Not easy to implement
with MEMS technologies
on the constructions
An overvoltage
protection circuits is
required
32
Case Study
-Electromagnetic transducer-
Unique control
signal
modulated by
polarity
detection
Boost Topology for step-up
In-phase sinusoidal current from
a sinusoidal source
Two converter
signal
Source: S. Roundy
Two converter
to eliminate the
need for rectifier
Impedance matching by altering
duty cycle
Not overlapping control signals
Case Study
-Electromagnetic transducer-2-
Seiko Kinetic
Oscillating Weight
Magnetic Rotor
Induced Current
S it Fi l C it
Harvested Energy
Supercapacitor
Boosting Circuit
Final Capacitor
33
Case Study
-seiko kinetic- Boosting Circuit
Charge is initially stored into a supercapacitor
x3
By means of two flying capacitors and
charge is transferred into the final capacitor
where the voltage level rises faster.
Charge is initially stored into a supercapacitor
used as charge tank.
1
C
2
C
3
C
4
C
Case Study
-seiko kinetic- Boosting Circuit
In few seconds, the final capacitor reaches 3x the SuperCap voltage.
34
Case Study
-Electromagnetic transducer-3-
-EnOcean-
Pushing a button causes a Pushing a button causes a
complete inversion of a permanent
Magnetic filed
Voltage and current generate by
Lents law is enough to transmit
16bits
New generation of devices with
self-powered sensors and
bidirectional wireless
communication
Case Study
-ThermoElectric Generator- TEG
Seiko Thermic wristwatch, convert heat from the wrist (body
heat) into electricity.
Thermoelectric conversion
Carnot efficiency : ( TH - TL) / TH = T / TH.
Seebeck-Peltier effect
35
Case Study
-TEG-
TEGs output voltage is very low
TEGs have a maximum power point
(MPP) which change with T
MPP is usually the half of the open
circuit voltage (Vteg-oc)
Problem: Internal resistance of TEG
d d t t d i depends on temperature and aging
Case Study
-TEG-
Essentially a boost converter with auto-generation of the control signal
(regulation loop)
The circuit starts to work with 20 mV due to JFET and L2 (
(Spies et al.)
36
Case Study
-Power Delivery to Bio-Implantable wireless circuits-
Output voltage to
the implanted
(Dondi et al.)
the implanted
device 2,2V
Size: 1 cm
2
WSN HW support a wide voltage
supply range (usually between 1V
and 4V )
Case Study
-Sub-mW PV cells-
Powering sensor nodes with unregulated and variable voltage supply from
the solar cell adaptive Active-Recovery DC
Minimize the energy used for DC/DC or linear regulation
Tmote Sky 2,1 3,6 V
TI Node 1,8 3,6V
TinyNode 584 2,4 3,6 V
[solar scavenger 10mm
2
PV surface: Brunelli, Benini]
Automatically adapt duty-cycle with analog thresholds (comparators)
on voltage supply
Optimize thresholds for MPP in low-lighting condition (no tracking at
high lighting as energy is over-abundant)
Indoor PV powering is feasible!
37
Approach
Select the desired light intensity and find the solar cell MPP
A window (Vth1 , Vth2 ) is defined around the MPP forcing the
senor node to operate in this range of values.
Sub-mW PV cells
-How it works- Inductor-less solar harvester
Design of the energy storage and
conversion circuitry together with
the target platform
Vth2
Vth1
Energy available for the whole Activity time
C, Vth1, Vth2 are evaluated to guarantee
the complete execution of the worst-case task or activity
38
Adaptive duty cycling
Activity time grows with energy intake
PV energy harvesting is usable indoor
Implementation
example
[solar cell for ZigBee Sensor node]
Cmin is evaluated by characterizing the most power-consuming operations,
in order to guarantee the completion of the worst-case task
30 packets each cycle
Cmin = 0.1F
39
EM harvesting Easy
Inductively powerered WSN Node
Energy harvesting exploiting the EM
field fromAC electric current during
+ +
field from AC electric current during
idle (no measurement) times
Research supported by a grant of Telecom Italia
Fully energy-neutral solution
EM Harvesting - Hard
Energy harvested from RF waves,
generated by a transmitter
(wireless power transmission)
St th ith Store the energy with
supercapacitor like energy buffer
Rectenna
RFID
transmitter
868 MHz
40
Power Transfer Efficiency
WISP - 2009
WISP - 2007
Power Cast -
2009
Lessons Learned:
Power levels are low (tens of W)
Advanced RF & Antenna design is needed
Video
Wireless Power over Distance
41
Electrostatic Conversion
Use compression/tension between parallel plates
Use ambient or intentional vibration to cause motion between
plates
Electrostatics tractable only if very small air gaps (microns) due to
Energy filled by internal
source
Net converted energy
field breakdown
U
Q
C
Piezoelectric Effect
Some materials present relations
between deformation and electric
field
1
2
3
T
W
+
_
+
_
Si l d l ti
Unactivated
L
W
Activated
V
+
_
L+L
W+W
T-T
+
_
+
_
S sT dE
Simple model equation:
T Stress S Strain
s Compliance E Electric Field Strength
d Piezoelectric Coefficient
Pi l t i P t t
Size 9,8 x 5,7 x 3 cm
Weight ~120 g
Energy buffer 4,7 F
Mean power (benchmark 2 Hz) 18 W
Energy (1 min.) 1,1 mJ
Piezoelectric Prototype
42
Aluminum chassis mounted on a customized orthopedic knee brace
(1.6Kg)
Donelan et. Al, Science 2008: 5W from leg movement with no extra effort!
BUT
Kinetic Harvester with micro-motors
Mi t
Kinetron (NL)
Micromotors
12,4 mJ Energy per minute
206 W Average power (2 Hz)
4700 F
Storage Capacitance
~80 g Weight
6,5 x 2,5 x 2,5 cm Size
(10x more than piezo!)
Energy from wrist movement
Charge control
circuit
Oscillating weight
Secondar
Proof mass oscillation directly cranks generator rotor
Little intervening mechanics
Drive circuit
Gear train
Rotor
Stator
Coil
Secondary
power supply
Seiko AGS System
Little intervening mechanics
Charge accumulated on capacitor
Power Output:
5 W average when the watch is worn
1 mW or more when the watch is forcibly shaken
43
Energy Harvester Output Power
Transducer mechanisms include electrostatic MEMS,
piezoelectric, and electromagnetic
Output power between 10 W and 1 mW for typical
vibration scenarios
Commercial products
more than 20 mWin the presence of a
significant vibration
Volture
www.perpetuum.co.uk
g
very weak vibration (e.g. microwave oven
0.24 g's, 120 Hz) it is able to harvest 43W.
The Sustainable Dance
Floor
www.enviu.org
44
Thermoelectric conversion
Thermoelectric conversion
Carnot efficiency : ( TH - TL) / TH = T / TH.
Thermo pile (thermolife

)
Seebeck-Peltier effect
Applied Digital Solutions Thermo Life (10 A at 3 V
with only 5 degrees Celsius of temperature
difference ).
Store extra energy produced during periods of
higher T so they can continue to run during
warmer, less efficient ambient temperatures.
p ( )
TEGs Have Variable Polarity
Output polarity of TEG is dependent
on the direction of the temperature p
gradient
Capability to work with positive and
negative input voltages
Rectifier structure:
Polarity switch controlled by a
comparator
diodes are only active during start-up diodes are only active during start-up
where still no supply voltage for
the comparator is present
prototype: Start-up 150mV, drop 40 mV,
later on5 mV
[Fraunhofer]
45
Other issues with TEGs
Vout < 1V when T is low Boost DC/DC
R
out
with T MPTT is needed
out

Solutions similar to PV (indoor) harvester
Thermoelectric example
Seiko Thermic wristwatch, convert heat from the wrist
(body heat) into electricity.
46
Commercial harvesters
PV: quite mature, with many products
Flexible PV materials are interesting e.g. g g
www.powerfilmsolar.com
Solution provides
www.enocean.com (Piezo, kinetic, solar)
www.kinetron.com (EM kinetic)
www.micropelt.com (thermal)
www.powercast.com (RF transmission)
www.microstrain.com (Piezo)
and many others EH forum
www.energyharvesting.net
Enhanced Power Unit Architecture
Energy
Transducers 2
Kinetic Energy,
Photovoltaic
Fuel Cell
Power Unit
Monitor
Measures
Energy
Transducers 1
Kinetic Energy,
Photovoltaic
Conversion Electronics
Take raw electrical signal from
transducer and convert it to a usable
DC voltage
Energy Storage and Delivery
Receives energy from conversion
electronics and stores it (SuperCap,
batteries, etc. ) Regulates the output
voltage and current.
Fuel Cell
Power
Supply
Measures
energy and
battery charging
status,
elaborates
energy
predictions and
provides
information to the
powered system
Energy Delivery
96
Wearable sensing and elaboration platform
47
Power Supply
research branches for next 10 years
Matching
circuit
Power
electronics
Design
optimization
Mechanical
fabrication
HW/SW
co-design
Batteries Energy Scavenging Fuel Cells Etc.
Motion Solar RF
. . .
. . .
Application areas over next 10 years:
smart homes, fatigue monitoring, ubiquitous data access for people, building env. control,
emergency response in commercial buildings, manufacturing monitoring and control, inventory tracking, etc.
Power Supplies
Summary
Energy harvesting systems are promising for many
autonomous and distributed applications
Energy Harvesting and permanent power storage Energy Harvesting and permanent power storage
devices are self-power enablers
All system components need to be Energy Aware
Excellent HW design is the a key factor
but also developing effective power management
l ith l f d t l l algorithms plays a fundamental role.
Distributed energy awareness is the frontier
48
Th k Y Thank You.