Ultra Low Power Type

1
Ultra Low Power Type 1 Artefact

• Presented by Dr. Eric Yeatman

2
Ultra Low Power a Necessity for the DC

• The DC idea envisages a world with many devices
  communicating
• Users would find it unacceptable to be constantly
  recharging or replacing batteries
• Therefore, Ultra Low Power Operation is
  Necessary, suggesting
• Low energy communication protocols (Smart-Its)
• Short range communication with lowest energy
  routing
• Ultra low power circuitry and low duty cycle
  electronics
• Ultra high efficiency power processing and power
  management
• Autonomous devices desirable (capable of
  generating own power)
• With potentially hundreds of devices even battery
  lifetimes of years would be inconvenient for a
  user
• sensor artefacts may be difficult/impossible to
  access

3
Imperial College Our Role in ORESTEIA
AT1/AT2
AT4
AT2/AT3
wireless link
AT1/AT2
Imperials ORESTEIA Task - wearable/implantable/
embeddable devices which require autonomy
Battery Powered/Mains - stationary devices
4
Requirements of Artefact 1

• Energy scavenging
• Obtaining energy from the environment
• Power Converter
• Providing a stable voltage from the acquired
  energy
• Radio Link
• Ultra low power (1mW 10mW)
• Sensor
• Consortium should use off the shelf devices for
  prototyping

5
Energy Scavenging

• Vibration
• Electromagnetic
• Electrostatic
• Piezoelectric
• Wireless Power Delivery
• Inductive coupling (such as RFID technologies)
• Other
• Thermal gradient
• Solar energy

6
Initial Work based on Electrostatics

• Vibration as the prime mover allows the device to
  be encapsulated
• Traditionally in MEMS,
• Electromagnetics scale poorly
• Forces scale as L3 (constant current density)
• Electrostatics scale better
• Forces scale with L2, or better (Paschens Curve)
• Piezo materials are very difficult to work with
• We aim to actually fabricate the device

7
A New Approach

• Other groups (MIT, Sheffield, Southampton etc)
  are using high frequency resonant devices.
• Our application is low frequency
• MEMS scaling laws prevent low frequency resonant
  devices
• Resonant devices only work well at the resonant
  frequency
• Solution (pioneered at Imperial College)
• The Parametric Oscillating Generator

8
Parametric Oscillating Generator

• Non-linear device Only move at maximum
  acceleration
• Power output for a double sided device is
• Simulation work has shown b 0.8
• Power is directly proportional to Y0

9
Preliminary Generator Design
Fabrication work has begun
10
So, whats realistic?

• If the device operates at 1.5Hz, with an input
  amplitude of 5mm, then the maximum power we can
  generate is 0.5mW, from a double sided device.
• The movement from a beating heart is not
  sinusoidal we could use the higher frequency
  harmonics.
• At 2Hz, the maximum power output is 2.3mW.

11
Proposed Power Electronics
12
Ideal Circuit Behaviour (single sided)
Power Generated (Green Trace-Red Trace) 0.2mW
13
Findings so far

• Work so far has demonstrated the following are
  important to maximise power output
• Large MEMS movement
• Large mass
• Power Electronics must deal with high voltages
  (gt200V)
• Parasitic elements have a significant effect on
  the generated power. Discrete components are not
  suitable
• A vibration with a higher crest factor than a
  sinusoid is beneficial for this non-linear
  device.

14
Resonant Modelling on Simulink

• Model the system as a (set of) differential
  equations

15
Project Plan for Power Generation

• Oct 01 Jan 02
• literature survey, review of current technology
• Feb 02 May 02
• investigations into trade offs between different
  power generation techniques
• modelling systems in SIMULINK and MATLAB
• Initial practical experiments with prototype MEMS
  device
• June 02 Jan 03
• Implementation at Circuit Level
• Evaluate various power electronics converter
  topologies (MATLAB and SPICE)
• Design of ultra low power control circuitry at
  transistor level
• Design of power electronics circuitry at
  component level
• Jan 03 March 03
• Interface power converter and control electronics
  with MEMS device
• March 03 June 03
• system evaluation and interface with radio link

16
Low Power Radio Today

• We have a power budget of only 1-10uW
• Typical low power radio devices available today
  require 104 to 105 times this power

The above Bluetooth transceiver has an output
power of 1mW but requires 25 times this value in
order to perform the necessary (de)modulation!
Both need minimising!
17
Reducing the Power

• Reducing the frequency and bit rate will reduce
  the power needed to modulate and demodulate the
  signal
• Reducing the bit rate and range will reduce the
  output power

Certain data only needs short bursts of
transmission Operation of the above chip for a
total of 10ms every second (1 duty cycle) would
lead to a transmission power of 50µW!
18
RF Far Field Transmission

• PROBLEM SIZE CONSTRAINT OF APPLICATION
• Aerial length should be equal to a multiple of a
  quarter of a wavelength to be 100 efficient
• At 50kHz this is 1500m
• Trade Off
• Low Frequency means Low Power Consumption, Larger
  Range BUT Large Aerial
• High Frequency means HIGH Power Consumption,
  SHORTER range BUT smaller aerial
• Possible SOLUTION
• Use magnetic induction in the near field at low
  frequency

19
Magnetic Induction

• Current Research
• To ascertain whether magnetic induction
  techniques allow smaller aerials at lower
  frequency for an acceptable power transfer
• Consider the set-up opposite
• The magnetic field set up by the
• current iTX induces a voltage in
• Coil RX
• Can be modelled as a
• poorly-coupled transformer

x iTX
iRX Coil TX
Coil RX (radius rTX) (radius
rRX)
20
Magnetic Induction vs RF

• Initial theoretical analysis has led to the
  following
• The power coupling efficiency for magnetic
  induction can be approximated as (for xgtgtrTX,
  rRX)
• Magnetic Induction transfers more power than RF
  at short range

• Magnetic induction
• RF Transmission

21
Ultra Low Power Techniques

• High Q inductors are a necessity for ultra low
  power RF circuits
• Historically it has been difficult to produce
  high Q integrated inductors (due to high
  substrate losses)
• Many circuits use discrete inductors
• Unsuitable for this application due to large size
• Latest technology allows high Q integrated
  inductors
• silicon micromaching such as is shown below
• silicon on sapphire

Work done at Imperial College by the Optical and
Semiconductor Devices Group Out of
plane inductor minimises the eddy current
losses
22
Ultra Low Power TechniquesSub Threshold CMOS
Nano watts power consumption! Major research
area of the IC Circuits and Systems
Group UCLA have demonstrated ultra low power
sub-threshold CMOS RF circuits (VCO 300?W and
a mixer at 45?W)
23
Related Academic Work

• Important to know what research is being done
  else where
• can advance on new discoveries rather than
  reinventing the wheel
• fill gaps/weaknesses in other research
• Similar DC projects focusing on ultra low power
  operation exist at
• UC Berkeley (Picoradio project)
• nodes aim to be smaller than 1 cubic centimetre,
  weigh less than 100g, cost less than a dollar,
  have a power dissipation level lower than 100uW,
  self powered.
• UCLA (WINS)
• particularly for military applications. Average
  power consumption 100uW. Not necessarily self
  powered.
• MIT (uAMPS)
• MEMs powered. MEMs power generator device using
  vibration driven variable capacitor.

24
Project Plan - UPLR

• Oct 01 Jan 02
• carried out literature survey, review of off the
  shelf devices and competing techniques
• began review/theoretical analysis of magnetic
  induction techniques
• Feb 02 May 02
• investigations into trade offs at system level
  (MATLAB)
• magnetic induction versus RF transmission
  techniques
• modulation schemes
• error correction
• addressing
• June 02 Jan 03
• Implementation at Circuit Level
• Evaluate various architectures (MATLAB)
• Transistor-level design (CADENCE
• Jan 03 March 03
• fabrication via Europractice