Jan M' Rabaey

1
Short Distance Wireless A Short Perspective

• Jan M. Rabaey
• Fred Burghardt, Yuen-Hui Chee, David Chen, Luca
  De Nardis, Simone Gambini,, Davide Guermandi,
  Michael Mark, Nathan Pletcher, and many others
• EECS Dept.
• Univ. of California, Berkeley

2
Short-Distance Wireless
Advanced sensing (e.g. in automotive)
Wireless bio-monitoringand actuation
3
Example Tire Pressure Monitoring
on a dime
First-generation PicoCube
uC board
sensor board
fit into slots around periphery of PCB
Average power of complete node 6 mW!
F. Burghardt and PicoCube group
4
A Next Step Intelligent Tires
Sensors embedded in liner of tire collect and
transmit informationabout tire deformation,
temperature gradients, etc to assist engine
control and braking systems.
Challenges weight and size of sensor nodes (lt
5g), high data rate (gt 100 kbs), reliability
With A. Sangiovanni-Vincentelli and P. Wright
5
Going One Step Further Dense Networks
6
An Huge Window of Opportunity

• Below the radar screen in the wireless world
• Short distance currently means around 10m
  (Bluetooth, 802.15.4, 802.15.4a)
• RF-ID the most visible member so far
• What exists is mostly at hoc nothing really in
  place in terms of standards
• The only constraints are the power levels in the
  different spectrum bands

Crucial challenges Power, Energy per useful bit,
Size!
7
Why at BWRC?
A natural extension of our low-power wireless
activities
Ultra Low Power Receivers
0 dbM Transmitters
Courtesy Y.H Chee, B. Otis
8
Pushing the Envelope Ever Further
Taped Out May 06
Wake-up Receiver
Courtesy N. Pletcher
9
Why at BWRC?
Building on and extending our miniaturization
technologies
Integrated Sensor Node Courtesy Y.H. Chee
Back (with solar module)
Front
Antenna
Solar module
38mm
TX
Reg.
PicoCube Courtesy F. Burghardt andA. Redfern
L.C.
Sens.
Reg.
?C
Sens.
25mm
10
Classifications / Design Choices

• Radiative versus Reactive
• Wideband (pulse-based) versus Narrowband
  (sinusoidal)
• Passive versus Active
• Power source

11
Power versus Size

• Circuits Perspective
• Lower frequency
• ? Lower power
• ? Smaller energy scavenging/storage devices

• Radiations Perspective
• Large antennas (?/4 - ?/2)
• ? Efficient radiators

Tradeoff between size and power
Can we operate with small antennas and low
frequency circuits?
12
Electrically Small Antennas
CourtesyY.H. Chee

• N turns, radius a

Electrically small antennas threshold

• Electrically small loops
• No significant phase shift in electrical path
  length
• 2?aN lt ?/10 or equivalently

N 3, a 5mm
For frequency lt 100 MHz, size lt 1 cm3 and free
space propagation ? Electrically small antennas
13
Field Properties of Electrically Small Antennas
Poynting Vector (Power density)
Number of turns N1 Radius a Width w ltlt
a Current amplitude I0 Wave number k
2?/? Wave impedance ??(?/?)

• Radiative Far Field
• kr gtgt 1 ? (r/?) gtgt 1/2?
• Wr is real (radiative), ? 1/r2
• Radiative Near Field
• kr 1 ? (r/?) 1/2?
• Wr have both Re and Im terms
• Reactive Near Field
• kr ltlt 1 ? (r/?) ltlt 1/2?
• Wr is imaginary (reactive), ? 1/r5

Field equations for an electrically small loop
14
Near Field vs Far Field Communications
a 5mm, I0 1mA, ? 90?

• Electrically small antennas are poor radiators.
• However, they are efficient coupling structures
  via a high Q medium for short distance, low
  frequency communication systems

15
Near Field vs Far Field Communications
16
Wireless Assembly (chip/package)From mm to mm
distances
Approach 1 Capacitive (Sutherland et al,
ISSCC04, HotInterconnect05)
17
Capacitive Interconnect
Using 10 of a 150mm2 chip for Proximity
Communication with a gap misalignment of under 5
microns yields a bandwidth per square millimeter
of about 3.1 Tb/s and a chip bandwidth of about
46 Tb/s.
18
Wireless Assembly (chip/package)From mm to mm
distances
Approach 2 Inductive (Kuroda, ISSCC 05 06)
19
Inductive inter-chip interconnect
20
Board-to-Board Wireless?
? 1 mm (40 mils) side ? 125 um (5 mils) width L ?
1.8 nH Terminated with 1 MW resistor at receiver
side
ChallengeCoupling Factor Small
21
Opportunity Resonance at the receiver

• 1 turn for TX (1.8 nH)
• 3 turns for RX (8.2 nH)
• 2 pF load cap at the receiver (f0 1.24 GHz)
• 80 mils (2 mm) distance (exp k ? 13.4e-3)

Advantage Gain! Disadvantage Narrow Band
Eq. Circuit
ADS
22
Extending to multiple cm distances
Applications Body-area networking
(biomedical) Flexible wireless backplanes
23
Extending to multiple cm distances
5 cm
8 mm
Frequency Content
3 nsec
Transmitted pulse
24
Inductive Transceiver
Precise Timing the Main Challenge in Reducing
Power Consumption How to avoid precise
synchronization components (Xtals)?
25
Frequency Locking to Transmitter
26
Prototype Inductive Transceiver
Receiver (analog part) ? 500 pJ per bit 10 µA _at_
20 kbps, 160 µA _at_ 320 kbps 3.2 mA (always on _at_
full bandwidth) Transmitter ? 30 pJ per Bit 0.6
µA _at_ 20 kbps, 9.6 µA _at_ 320 kBps 6.0 mA _at_ full
speed (200 Mbps) PLL (analog part, including
references) Ring Oscillator VCO ? 20 uA Loop
filter CP ? 40 uA
In Fab (May 06)
Courtesy D. Guermandi
27
Avoiding Accurate Timing Elements or Expensive
Synchronizers
Through local, collaborative strategies

• Nodes synchronize by overhearing neighbors
• A small number of precise timing elements
  (anchors)
• Anchors synchronize to global beacon

Courtesy L. De Nardis
Example 400 nodes, 4 anchors
28
Design Exploration of Inductive Wireless for
Dense Networks (1-5 cm distance)
For 20 kbits/sec PTX ? 20 mW PRX ? 200 mW
Narrow-band Resonant
D. Guermandi and S. Gambini
29
Providing Power

• Most applications for short-distance wireless are
  battery-averse (not accessible, high density, )
• Scavenging of power for data acquisition,
  storage, and transmission hence a necessity

Magnetic shaker
Piezo-electricbender
Capacitive vibrator
Challenges mass, size, reliability
Courtesy P. Wright, S. Roundy, M. Koplow
30
ElectroMagnetic Scavenging
31
Electro-Magnetic Scavenging - Options
Secondary power source(battery or scavenger)
32
Regulations
Largest amount of power available at the 13.5 MHz
band, but the least amount of bandwidth (much
better at 400MHz and 800 MHz) Hybrid solutions
maybe desirable
33
Projecting Forwards Stochastic Networks on a
Chip?

• Maybe not the most efficient communication
  mechanism, but
• Generic architectural approach for dealing with
  failure not dependent upon fault mode
• Reliability transparent to the algorithm
• Graceful degradation of performance

Sources K. Ramchandran (UCB), D. Jones (UIUC)
34
Distributed Collaborative Systems on a Chip
An interesting case study A collaborative
configurableradio architecture based on
collaborative autonomous entities

• Array of locally-coupled cheaplow-power
  oscillator-based units
• Known to exhibit complex, spontaneous pattern
  formation (somewhat related to cellular
  non-linear networks)
• Operational model selectedthrough choice of
  coupling factors and operational nodes

Source J. Roychowdhury, J. Rabaey
35
The Mechanical Radio
The Ultimate ULP Tunable Wireless Transceiver?
Input Electrode
Wine-Glass Disk
Support Beams
Output Electrode
R 32 ?m
Coupling Beam
Anchor
9 wine-glass disc oscillator-based GSMcompliant
oscillator
Courtesy C. Nguyen, UC Michigan
36
Summary

• Short distance wireless presents huge window of
  opportunity
• Needs clear metrics to allow for classification
  of different approaches in terms of energy and
  size efficiency
• Combining energy and data transmission very
  attractive, but somewhat contradictory
• May ultimately lead to novel computation and
  communication models