Transcript Techniques for Miniaturization of Circuits and Systems for Wireless

Techniques for miniaturization of
circuits and systems for
wireless sensing
Brian Otis
Wireless Sensing Lab
Seattle, WA, USA
[email protected]
• Vision
• Existing technologies
• How do we get there?
–
–
–
Circuit techniques
Energy harvesting techniques
Integration techniques
Vision: autonomous sensing
• Miniaturized devices (a few mm3)
• Extremely inexpensive
• Frequent radio contact with peers
and with basestation
• Periodic sensing of environmental
parameters (temperature, light,
pressure, acceleration etc.)
• Flexible deployment in wide variety of
biological, manufacturing, or
environmental monitoring applications
Vision: autonomous sensing
• Miniaturized devices (a few mm3)
• Extremely inexpensive
• Frequent radio contact with peers
and with basestation
• Periodic sensing of environmental
parameters (temperature, light,
pressure,
acceleration
etc.)miniaturization of
Critical
challenges:
- RF Link
• Flexible
deploymentclock
in wide
variety of
- Reference
generation
biological,
manufacturing,
- Power
sources or
environmental monitoring applications
RF Link: existing designs won’t work – why?
1.They are too large. Traditional architectures require multiple
off-chip components, high die area, and a large quartz crystal
resonator.
2.They consume too much power. Bluetooth & Zigbee (the “low
power” standards) consume > 20mW. This eliminates the
possibility of energy harvesting.
3. They require high-end processes and high transistor counts.
What about RFID?
• Case study: Hitachi m-chip
• (150x150x7.5)mm3 (168e-6 mm3)
• Si Density r=2330kg/m3
 mass of one chip = 0.393 mg
(small)
• Millions of die/wafer
• < $0.10 US (cheap)
• Interrogator output power: 0.3W
• Range: 450mm (limited capabilities)
M. Usami et. al, ISSCC 2006
Case Study: Hitachi RFID chip
Power harvesting
Frequency reference harvesting
(100kHz clock)
• Power is extracted from
incoming RF energy
• External antenna
(few cm)
• Ideal for embedding in
secure documentation
M. Usami et. al, ISSCC 2006
RFID Interrogators
Power dissipation >1W
Cost >$100 US
Provides two critical
functions that are currently
impossible to generate
on-chip:
1. Accurate quartz-based
frequency reference
2. Power source
RFID summary
1.
RFID chips can be made extremely small and
cheap
2.
These are radios that harvest their power from an
incoming RF signal. RF power falls off
quadratically (at best) with distance, resulting in
high interrogator power and very short range.
3.
There is little energy available for sensing or
computation.
4.
They cannot form peer-to-peer networks.
Research Goal
Self-contained wireless sensing systems that can be
fabricated exclusively with thin-film processing techniques.
This should include:
Peer-to-peer Wireless links
Computation/Data Storage
Chemical/biological Sensors
Electrical Sensor Interfaces
Energy/Power Source
Three steps to autonomy
1.
Generate accurate frequency reference locally
2.
Generate power locally
3.
Develop circuit design techniques for reducing
computing/sensing/communication power consumption
RF MEMS: path to ultra-small radios?
On-Chip Inductors (Q ~10)
MEMS Resonators
(Q~1000)
100mm
~300mm
• MEMS resonators have significantly higher Q than on-chip
inductors
• Possibility for elimination of quartz resonators
• MEMS sensing capabilities
System proof-of-concept
Can we design an entire low-power radio link using MEMS
resonators as a frequency reference?
Case Study: 2GHz transceiver for wireless sensors
Goal: Use matching RF MEMS resonators on the transmit and
receive paths to define carrier frequency
1mm3, 2GHz super-regenerative transceiver
1mm
BAW
CMOS
2mm
• No external components
(inductors, crystals, capacitors)
• 0.13um CMOS
Total Rx: 380uW
Range: 30m
Datarate: 50kbps
• Operates above transistor fT
B. Otis et al., IEEE ISSCC 2005
Three steps to autonomy
1.
Generate accurate frequency reference locally
2.
Generate power locally
3.
Develop circuit design techniques for reducing
computing/sensing/communication power consumption
Energy Harvesting
antenna
Extracting energy from the environment to power the electronics
reduces maintenance costs and increases capabilities
PV cell
Bottom line:
-Approximately 100uW/cm3 available
(but efficiency decreases as volume shrinks)
-Power consumption of electronics determines wireless sensor
volume and capabilities
Thermoelectric energy harvesting
Why thermoelectric?
• Large, stable temperature gradients
often exist in ubiquitous sensing
applications
• Monolithic, solid state, possible
to integrate with circuitry
How does it work?
• Converts thermal gradient to electric
potential via Seebeck effect
• Thermocouples connected in series
as a thermopile increases voltage
(and resistance)
• Radioisotope powered TEGs widely
used in space missions
Work-in-progress:
• SOI-based mTEG
• p,n silicon thermoelements
• Floating membrane increases
thermal isolation
Three steps to autonomy
1.
Generate accurate frequency reference locally
2.
Generate power locally
3.
Develop circuit design techniques for reducing
computing/sensing/communication power consumption
-> example: sensor ID generation
Inexpensive, low power sensor identification
10101111
00110101
0111001
•
•
•
•
Wireless sensor network addressing
Object identification for Radio Frequency ID (RFID) tags
Wafer and process tracking of individual chips for failure analysis
Tracking for implantable electronics devices
Can we extract a unique digital fingerprint from process variations?
ID Generating Circuit Requirements
• ID circuit must generate a digital output
• ID code must be repeatable and reliable over supply,
temperature, aging and thermal noise
• The ID code length and stability must allow positive unique
identification of each die
• Low power consumption, no calibration
Proposed Idea: positive feedback ID generation
B
A
voltage (V)
• Each ID cell: cross-coupled gates used to amplify transistor
mismatch
A
B
time(s)
– Evaluation period  Node A and B will split due to transistor
mismatch
– Readout period  Digital-level output will be obtained directly
at ID node
Chip Implementation
• 128 ID generators – 140nW @ 1V
• Technology: 0.13mm CMOS
• Provides stable fingerprint with extremely high probability of correct
chip identification
Su, Holleman, Otis, IEEE ISSCC 2007
Conclusions
1. Wireless sensor scaling is constrained
by energy source, antenna
dimensions, and frequency reference
2. Self-contained wireless
sensors less than 1mm3
are on the horizon
3. Future chips will include circuitry, EM elements,
MEMS structures, sensors, and power generation
4. Interdisciplinary collaboration is critical to focus our efforts on
relevant sensing problems
500um