Transcript Smart Dust: Communicating with a Cubic

Smart Dust: Communicating with
a Cubic-Millimeter Computer
Presentation by
Hörður Mar Tómasson
13. October 2006
The Smart Dust project
• Went on at UC Berkeley 1998-2001
• Primary investigator: Kristofer S.J. Pister
The goal
•
•
•
•
•
1 mm³ motes
with onboard sensors,
CPUs
and wireless communications facilities
forming the basis of a sensor network
Fundamental goal
• To explore the limitations of
microfabrication technology
Ideas for uses for smart dust
•
•
•
•
•
Surveillance networks for defense
Monitoring environmental conditions
Human-computer interfaces
Inventory and product quality control
Tracking movements of animals
Power considerations
• Batteries: 1 J/mm³ storage
• Capacitors: 10 mJ/mm³ usable storage
• Solar cells: 1 J/(mm² · day) in sunlight or
1-10 mJ/(mm² · day) indoors
• Optical receiver: 0.1 nJ/bit
• Optical transmitter: 1 nJ/bit
• A/D converter: 1 nJ/sample
• Computation: 1 pJ/instruction
Power considerations
• 1000 8-bit operations per sample will not
make a big difference in power used.
• 1 mJ per day from a solar cell indoors will
be sufficient for making a measurement
every second, processing the result and
transmitting it.
Low-energy computation
• Smaller transistors with less parasitic capacitance
consume less dynamic power.
• Reduced supply voltage also means less dynamic
power.
• Leakage currents can be decreased by reverse
biasing the channel-to-source junction.
• Clock rates of 1-100 kHz are sufficient for
working with some important types of physical
signals.
Wireless communication
• Radio communication currently requires
several mW of power and preferably
antennas longer than a millimeter.
• Semiconductor lasers and diode receivers
can use less power and are more directional.
• The Smart Dust project explored optical
communication.
Passive reflective systems
• A MEMS corner cube reflector (CCR) with
a side that can be tilted
• Less than 1 nJ used per transition
• The mote can use the CCR to communicate
with a base station equipped with a light
source.
Active steered laser systems
• Semiconductor laser
• Collimating lens
• MEMS steerable
micromirror
Optical receiver
• An imaging receiver has several benefits.
• Only one pixel receives the signal but the
ambient light is divided between the pixels.
• Several signals can be received in parallel.
• The authors did an experiment with a laser
and a video camera.
• A smart pixel has an integrated receiver.
Ad hoc mote networks
• If the motes can communicate directly with
each other, they can form ad hoc multihop
networks to carry the data around.
• This is an interesting problem for network
algorithm design.
An Ultra-Low Energy
Microcontroller for Smart Dust
Wireless Sensor Networks
Presentation by
Hörður Mar Tómasson
13. October 2006
Creators of the microcontroller
• Brett A. Warneke
• Kristofer S.J. Pister
Application
• The microcontroller
was developed for this
prototype smart dust
mote.
Architectural features
•
•
•
•
•
•
•
Highly independent subsystems
Component-level clock gating in decoder
Processor halt mode
Guarded ALU inputs
Multiple busses
Harvard architecture
Load-store RISC
Main oscillator
•
•
•
•
•
•
Runs continuously at a few kHz
Operates real time clock and five timers
One timer for each sensor sampling period
One timer for invoking the transmitter
One timer for invoking the receiver
One timer for waking up the datapath
Other oscillators
• 100 kHz for driving the sensor ADC
• 8 MHz for sampling a 1 Mb/s optical signal
ADC automation
• The ADC is configurable to different levels
of automation.
• At the minimum level, the sensor and
sample and hold are activated.
• At the maximum level, the voltage is
compared to a threshold and, if the
threshold is exceeded, converted and stored
in the SRAM along with a time stamp.
Transmitter
• The processor core uses two registers to
specify what memory blocks contain data to
be transmitted.
• The transmitter formats the data into
packets and transmits them asynchronously
to the CCR.
Four types of received packets
• Short sync packets trigger the transmitter.
• Immidiate packets contain an instruction
that is immediately executed.
• Program packets are streamed to the
program memory.
• Data packets are streamed to the data
memory.
Trimmable oscillator
Specifications
RF Telemetry System for an
Implantable Bio-MEMS Sensor
Presentation by
Hörður Mar Tómasson
13. October 2006
The long range goal
• NASA wants to develop implantable
sensors to monitor physiological parameters
of humans during space flights.
• It would be of great benefit to have
contactless powering and data readout for
the implants.
Advantages of contactless powering
and telemetry
• The inductor/antenna is small in size.
• There is no need to implant batteries.
• The circuit only operates when interrogated,
avoiding heating of the surrounding tissue
and extending the life span of the sensor.
• Feed-through wires not needed, enhancing
mobility and reducing risk of infection.
This paper
• A system for contactless powering and RF
telemetry from an implantable bio-MEMS
sensor.
• A square spiral inductor/antenna
• A MEMS capacitive pressure sensor
• A pick-up antenna
Spiral inductor/antenna
MEMS pressure sensor
Pick-up system
•
•
•
•
A printed circuit with mounted components
Spiral inductor/antenna, printed
MMIC low noise amplifier, mounted on
Antenna matching network, mounted
discrete components, Π-network
• Output connector
Operating principle
• The idea is to send pulses down into the
implant and detect the decaying sine
response.
Parameters
• Desired frequency range: 200 - 700 MHz
• Expected capacitance of pressure sensor:
0.3 - 4 pF
• Expected required parameters for square
inductor: 150 nH and Q=10
• Several inductors with different geometries
were tried.
Fabrication of the inductor
•
•
•
•
High resistance silicon wafer
Spin-on glass coating
Chrome/gold metallization
The goal is to have high Q.
The next step
• Exploring coupling between the inductor
and the pick-up antenna through stratified
media