DR ETD Summary of New Thermostat, TempNode,

1
DR ETD Summary of New Thermostat, TempNode,
New Meter (UC Berkeley Project)

• New Thermostat, New Temperature Node, New Meter
• Funding 1.81 M
• Period of Performance 3/2003 8/2005
• Multi-disciplinary Collaboration Team
• Paul Wright Mech. Eng. Dept.
• Jan Rabaey Berkeley Wireless Research Center
• Ed Arens Center for Built Environment,
  Architecture Dept.
• David Auslander Mech. Eng. Dept.
• Dick White BSAC, EECS Dept.
• David Culler EECS Dept.
• 20 Graduate Student Researchers (13 are funded)

2
The Big Picture

• Goals include Reduce blackouts at peak usage
  times
• Enabling Technology Development Key to research
  approach
• Meters, thermostats, temperature-nodes In ad hoc
  self organizing wireless networks (low power
  radios, energy scavenging 450 µW, TinyOS)
• Cheaper better faster for a 10x10x10 mission
• Residential focus in CA and affordable for even
  the smallest homes
• One working scenario to demonstrate new
  technologies
• Receives price signals every 15 mins (emergency
  signals treated immediately)
• Users provide preferences to their thermostat
  and major appliances through easy to use GUIs
• Eventually, self-learning systems
• Time stamped usage sent back once a day say at
  3am or immediateresponse in an emergency

3
The project is divided into four Sub teams, each
with a 10x10x10 mission

• 1. Communication - Pico Radio TinyOS for
  networking of devices
• Low-power and low-cost radio platforms, supported
  by appropriate operating systems, for ad hoc
  sensor and actuator network applications.
• 2. Controls - Applications on a prototype of New
  Thermostat
• Easy-to-use thermostat that can act as an
  automatic proxy to optimize energy savings versus
  comfort under varying energy price conditions.
• 3. Sensors - Relation to New Meter
  (Voltage/Current)
• Low-cost wireless, passive and non-intrusive
  current and voltage sensing for application to
  the next generation meter and other devices.
• 4. Power Supply - Energy Scavenging (important
  for Temp Nodes)
• Infinite life power source that scavenges energy
  from the environment. Possible energy sources
  include solar, vibration, air flow, and hybrid.

4
Sub team 1 Communication by Mesh Networking
Our prototype system balances occupant comfort
vs. price preferences with automatic, reactive
short-term load shedding and long-term energy
reduction
Utility
Real-Time Meter
Price Signal
Air Conditioner
Disaggregation of Thermostat into Nodes,
Control, Interface, and Communication
Preference slider
5
Core DR Technology has been createdTinyOS on
Nodes (called Motes)
1
2
Ultra Low Power Node called Telos 16-bit
microcontroller has a sub 1mA sleep state and can
rapidly wakeup from sleep in under 6ms. Telos
operates down to 1.8V to extract as much energy
as possible from the battery source.
6
However!! For further cost reduction we need
much Lower Power Radios

• Why is low power necessary? - Size and cost of
  our motes are today dominated by
• 1) transmission energy
• 2) power supply
• Solar power density (10mW/cm2-10mW/cm2)
• Vibrational power density 450mW/cm3
• Power consumption determines sensor node volume.

For an eventual 1cm3 node, running at a 1 Duty
Cycle, we will need Ptransceiver lt500microwatts
7
RF MEMS More of our Core DR Technology for Low
Power Radio resonators
100mm

• Substantial work in academia in RF MEMS(UC
  Berkeley, Michigan, etc.)
• Heavy industrial research in BAW resonators
  (Infineon, Agilent, ST Microelectronics, etc.)
• Agilent FBAR resonators building block of Tx
  filters and duplexers
• Active resonator area Piezoelectric membrane.
  Isolation chamber bulk-micromachined under
  membrane
• Benefits
• High Quality-Factor passive structures
• Passive RF frequency reference
• MEMS/CMOS co-design

Drive Electrode
Sense Electrode
8
DR Super-regenerative Receiver

• Regenerative -- creates feedback loop around
  amplifier a between incoming frequency and
  amplified frequency
• But delicate potentially unstable so
• SuperRegenerative -- Periodically start-up and
  quench oscillator in presence of RF signal every
  Tquench

jw
s
Tquench

• Sampling of input signal
• Detect area under oscillation envelope to
  demodulate signal

9
DR Receiver Implementation
1.4mm
1.4mm
Fully Integrated no crystal or off-chip
inductors
ST 0.13mm CMOS
10
Fully Integrated Small-scale Rx/Tx
1mm
2mm

• No External Components (inductors, crystals,
  capacitors)
• 0.13mm CMOS
• Full digital SPI control of analog/RF blocks

Total Rx 380mW
11
Communication work in ProgressA sub-100 mW
Integrated Node

• Simplest possible processor
• Dedicated accelerators when needed
• Aggressive power management
• Minimizing supply voltage

Courtesy Mike Sheets
12
Conclusions Sub-team 1

• The combination of RF MEMS/Super-regenerative
  CMOS architecture yields low power consumption
  and high integration
• Arrays of MEMS resonators will ultimately allow
  simultaneous access to many bands for
  interference avoidance, fading immunity
• Using these techniques, a 500 microwatt, fully
  integrated, 1cm2 transceiver for wireless sensor
  networks is feasible and will enable much lower
  power motes

13
Sub-team 2 Controls DREAM Demand Response
Electrical Appliance Manager
Utility
Temperature sensors
Power sensor
Price Indicator
Electricity used
Power actuators
Price
Real-time Meter
Occupancy sensors
Its not just a thermostat, its a DREAM!
14
DREAM Control Code

• Control of simulated, model, or real houses

Sensor/ActuatorModule
Simulation (MZEST)
XML
Controller
Java
Physical Model
RF
Real House
RF
Interface
15
2004-2005 Progress

• Java control code
• Precooling strategy
• Dynamic price generator
• Code documentation
• Java Graphic User Interface (PDA, Tablet PC)
• MZEST (MultiZone Energy Simulation Tool) based on
  California NonResidential Engine
• Modified to interface with Java control code
• Calibration underway
• The Wall testbed and Mica2 mote sensors and
  actuators
• Control strategy hierarchy

16
Controller

• DR Adaptive Precooling Strategy
• Typical precooling strategies use a fixed cooling
  start time which can lead to overcooling or
  undercooling.
• The Adaptive Controller DR precooling strategy
  uses a dynamic cooling starting time using a
  learning algorithm.
• Dynamic Electrical Power Prices Simulator
• Use dynamic prices to evaluate control strategy
• Price response to load
• changes with season (weather)
• changes with customer usage
• Critical peak price
• dispatched during peak periods
• up to 50 hours each year
• flexible notification period

17
MultiZone Energy Simulation Tool(MZEST)
MZEST is a multizone extension of the simulation
code used by CALRES, chosen because it can
predict temperatures in several thermal zones. We
will be using MZEST to test and fine-tune our
controller.

• Climate Zone
• House Construction
• Zone Adjacencies

Control Strategy

• Weather
• Zone Temperatures

DREAM
MZEST
On/Off Control
Reports, including annual energy use and zone
temperatures
18
Interface

• Emulated on PDA and tablet screen in Windows
• Graphic intuitive design
• Various control options based on temperature,
  comfort, and monthly electricity budget
• Information on electricity usage and price
• Next steps
• Use Usability testing
• Implement learning algorithms

19
DREAM mote hardware

• Outlet mote power sensoractuator
• Developed by Sensors group
• TemperatureRHlight sensor
• Signaling motes lights and LCDs
• HVAC relay mote

20
Full-scale testbed
21
Control Strategy Hierarchy

• Based on modular design
• (Demo)

22
Sub-team 2 Conclusions

• The DR system emulator (DREAM) successfully
  integrates hardware and operating system software
  from other sub teams across a range of
  scalesfrom simulations, through physical models,
  to a real occupied house.
• Multizone energy simulation tool (MZEST) acts as
  a simulator within DREAM allowing parametric
  studies of control strategies.
• Graphic user interface emulated on a PDA was
  successful thermostat proxy for both controlling
  the system and obtaining user preferences.
• New price simulator and precooling learning
  algorithms developed.
• Portable Java control code for industry to use in
  developing future DR technology.

23
Sub-team 3 Sensorsfor Metering and Monitoring
Utility
Temperature sensors
Power sensor
Price Indicator
Electricity used
Power actuators
Price
Real-time Meter
Occupancy sensors
Metering and Monitoring
24
Sensors for Metering Wireless Current
SensorMacro scale Prototype
25
Piezo Sensor
26
Wireless Mote
27
Wireless Metering Outlet
28
Operation
29
Sub Metering Data
30
Cheaper Sensors for MeteringCore DR Technology
Our new MEMS current sensor (prototype)

• Electric current (magnetic field) measurement
  techniques
• Inductor
• Hall effect
• GMR sensor
• Magnetic force on MEMS sensor

Magnetic material on MEMS canti-lever


120 Hz
120 Hz
output signal
output signal
I
I
I
I
MEMS cantilever withpiezoelectric film
out
in
out
in
60 Hz
60 Hz

AC current
AC current

AC current sets up time-varying magnetic field
whose gradient exerts force on high-permeability
magnetic material at end of MEMS cantilever
resonator, which vibrates and generates
piezoelectric output voltage
31
MEMS noncontact measurement of AC currents
Proximity Measurement
MEMS Advantage
arrays allow correction for position errors
Microfab
Concept
AC current sets up magnetic field. Gradient in
magnetic field strength exerts force on magnet
located at end of MEMS cantilever. Cantilever
deflection generates piezoelectric output voltage.
single cantilever released
arrays of unreleased sensors
thin films
32
Sub-team 3 Conclusions

• Piezo devices able to measure current and voltage
  for sub metering of devices
• Wireless motes then allow communication between
  central thermostat, and such piezo devices
  mounted on residential appliances. Allows DR
  control.
• New, cheaper MEMS devices have been prototyped
  for current measurements. Future includes more
  robust prototyping
• Electric current, magnetic field measurement
  technique

33
Sub-team 4 The Power Supply for the Devices
Utility
Temperature sensors
Power sensor
Price Indicator
Electricity used
Power actuators
Price
Real-time Meter
Occupancy sensors
34
Sub-Team 4 Power Sources. Critical Enabling
Technology to achieve No replaceable batteries
on the nodes

• Possibilities include
• Photovoltaic (Solar cell)
• Vibrations
• Air Flow
• Temperature Gradients
• Pressure Gradients
• Human Power
• MIT shoe-insert project
• FreePlay wind-up products

35
New Core TechnologyVibrational power runs
motes without replaceable batteries

• Sources
• HVAC ducts
• Raised Floors
• Motors
• Large windows
• Mount under wooden staircase

• Three Rules for Design
• P M
• P A2
• P 1/?
• PZT-shims with W-mass
• Early work 800 µW/cm3 at 5 m/s2 (on a clothes
  dryer!)
• Recent successes
• TinyTemp Node on stairs
• MEMS piezo bender

W
PZT
36
Residential Vibration Sources
37
Design Flow
Frequency and Acceleration of Source
1a
3
Elastic coefficient
Stiffness Model
2
4
Electric Signal
Power
OPTIMUM BENDER DESIGN FOR MAX POWER OUTPUT
Load/Power Circuit
Generator Equation
Preload Effects
1b
Composite Materials
Coupling coefficient
Coupling Model
38
Tunable-resonance device design

• Bracket fashioned from machined aluminum
• Load is applied by a setscrew assembly
• Proof mass is mounted using a tiny screw
• Resonance range can be adjusted by changing proof
  mass or shim thickness
• Proof mass 15 grams
• Piezo shim 32 mm x 13 mm x 0.5 mm
• Unloaded resonance 150 Hz

Proof mass 15 grams Piezo shim 32 mm x 13 mm
x 0.5 mm Unloaded resonance 150 Hz
39
Core DR TechnologyMicrofabrication to address
10x10x10

• Base layer of Si/STO (SrTiO3) provided by
  Motorola
• SRO (SrRuO3) PZT (PbZr0.53Ti0.47O3) deposited via
  PLD

• Si provides ease of fabrication
• STO enables epitaxial growth

50 nm of SRO
500 nm of PZT
50 nm of SRO
Base layer (Silicon Wafer)
40
Deposition and Patterning

• Deposition and patterning
• Initial etching

Edwards electron beam evaporator Ti 10-15
nm (for adhesion purposes) Pt 150-200 nm ?
Pt thickness will be varied to function as
elastic layer.
Ar Ion mill 5 nm/min _at_ 0.25 mA
PhD Research of Eric Carleton, Elizabeth Reilly
41
Schematic vs. Actual Device
View from above of MEMS cantilevers
End-on view of one MEMS cantilever
PhD Research of Eric Carleton, Elizabeth Reilly
Human hair 100 mm
42
Our best MEMS example to date
Human hair 100 mm
Eric Carleton and Beth Reilly (PhDs)
43
Conclusion for Sub-team 4

• Vibration and solar will provide power solutions
  for both outdoor (meter) and indoor devices
• Reduction to MEMS will allow 10x cost reductions
• Miniaturization being demonstrated while
  maintaining sufficient energy scavenging to power
  the next generation low power radios

44
Potential Breakthroughs in Next Phase The
PicoCube
Thermistor

• Early thinking about integrated device 1 cm3
• 3-D wiring
• 4 sides with solar panels
• 2 sides with sensors
• On-board recharg. battery
• On-board piezo electric generation

Solar Panels
Photoresistor
45
Details

• 3 Separate Components
• 1 Bus
• Overall
• Modular Design
• Simplifies Connection
• - Takes up a surface
• - Component packing takes up significant space

Power Bus
Microbattery
46
10x10x10 ?
DR Core Technology Trend

Temp.
Light sensor
Todays prototypes
2007
2006
2005
W
1cm
PZT
MEMS version inprogress for 2007
2 inch
47
PicoCube integrates all MicroScale devices to
satisfy 10x10x10
System automatically balances occupant comfort
vs. price preferences PicoCubes replace motes
Utility
Real-time Meter
Price Signal
Air Conditioner
Disaggregation of Thermostat into Nodes,
Control, UI and communication
Learning algorithmsadjust virtual slider
48
Automatic DR possible scenario1.
TouchPad of New Thermostat shows current /kWhr
expected monthly bill. Automatic adjustment of
HVAC cost/comfort. Appliance nodes glow colors
based on price.2. New Meter conveys real-time
usage, back to supplier 3. PicoCubes throughout
house allow for fine grained comfort/control
Incoming price signals
Appliance lights show price level appliances
powered-down