Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Wireless Power Transferfor Medical Devices
Lecture 12
1
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Summary
• History of wireless power transfer
• Wireless power for drug delivery
• Issues with implanted devices
• Applications
2
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
History of Wireless Power
• Depends on the invention of AC power generation
• He also developed a method to transmit power wirelessly
• This could be done over great distances
• Did not fit with theory
3
Nikola Tesla
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Tesla’s Wireless Power
• Tesla started to build Wardenclyffe tower in 1901 but it was never completed
• The plan was to experiment with power transmission across the Atlantic
• Now similar technologiesare being considered byIntel and others
4
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Science Fiction5
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Inductive Coupling
• This is probably the simplest form of “wireless’’ power transfer
• The principle by which transformers work
• Can be used at longer distances and without a magnetic core linking coils
• Power transfer efficiency can be poor and drops off with the cube of distance
6
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Resonant Inductive Coupling
• Transmitter (Tx) and receiver (Rx) coils are inductors - add a capacitor for resonance
• Tune the Tx and Rx circuits to have the same resonant frequency
• This can significantly increase the coupling efficiency of the power transmission
• Commercial development in this area
7
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
WiTricity and WREL
• WREL is a project at Intel
• WiTricity is a spin out from MIT
• Both use resonant energy transfer with the aim of wireless charging of mobile devices
• Near field magnetic coupling with ranges of a few metres
• Difficult to say if it will take off or not
8
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Power Transfer for Drug Delivery
• Large Tx coil driven with RF signal
• Receiver is a SMD inductor
• RF signal is rectified to drive a load
9
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Transmitter and Receiver
10
Rx PCB dimensions:4mm by 5mm
Tx: copper coil with variable capacitor for tuning
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Power Transmission Experiments
• Initial power transmission experiments used a resonant frequency of 7MHz
• Power input to the Tx is chosen to be 1W
• Receiver is connected to a 10kΩ load
• Measurements of induced voltage and power in different media
11
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Transmission in Air12
~4V at 20mm
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Transmission in Various Media
13
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Transmission with Varying Power14
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Heating Effects15
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Summary of Initial RF Power Experiments
• 2V and >4mW of power can be received over a distance of 20mm in air
• Transmission through water or saline does not significantly lower the voltage
• Higher power output from Tx can increase the power received but also cause heating
• 4W for 5 minutes is enough for 1 change
16
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Ex-Vivo Experiments
• Biological model (pig’s eyeball) to test power delivery to implanted device
• Lab in the Princess Alexandra Eye Pavilion normally used for surgical teaching
• Testing performed with receiver, coated with PDMS, implanted into the front and back of the fresh pig eyeball
17
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Implanted Receiver18
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Experimental Setup19
Receiver (implant)
Pig’s eyeball 35mm
Transmitting coil
Receiver outputs source
Dummy skull
power to a 10k loadΩ
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Sufficient power tolight an LED
Power Output20
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Induced Voltage = 2.944V
Power Output21
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Induced Voltage Results22
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Heating from RF Power Transfer
• Careful choice of frequency to avoid too much attenuation by biological materials
• High frequencies will be less efficient and cause heating of tissue. Exposure ∝ 1/f
• Exposing the eye to 1kW m–2 at 2.45GHz for 30 minutes can cause cataracts
• 6.75MHz is a standard “Industrial, Scientific and Medical” radio band
23
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Heating with Ex-Vivo Biological Model
• Transmission of 1W to implanted receiver supplies sufficient power to open cavities
• Temperature of the sclera of the pig eye was measured with an IR thermometer
• Accuracy of this instrument is 1 and measurements were averaged over 1min
• After 5 minutes of applied power there was no significant change in temperature
24
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Standards for RF Exposure
• International Commission on Non-Ionizing Radiation Protection and IEEE standards groups have recommended exposure limits
• Exposure of the eye to frequencies from 3-30MHz should be <
• Maximum exposure for 7MHz is 1980 Wm–2
• Actual exposure for 1W input is <200 Wm–2
25
100000/f2MWm2
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Drug Release Testing
• Testing of one of the drug delivery chips using wireless power delivery
• Uses a 555 timer oscillator circuit to activate the controlled release device
26
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Experimental Setup27
connected to alaptop for readout
ZIF socketSample in
Control circuit
Power amplifierSignal generator
Handheldtransmitter
Container with tap water
USB microscope
Receiver
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Wirelessly Powered Release Video
28
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Wireless Power and Communications
• Communication with an implant can be achieved by modulating the RF power
• Initial development used amplitude modulation of the 7MHz signal
• DTMF - Dual Tone Multiple Frequency control system similar to tone dial phones
• 4-bit address decoded to select one of the 12 cavities on the protoype device
29
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Prototype Wireless Communications
30
LinkWireless
LC tank #1ModulatorAmplitude
AmplifierPower
Keypad EncoderDTMF
SignalGenerator
.
.
.
3V
LC tank #2 Rectifier/Amplitude
Demodulator
Decoder
OscillatorVoltage Demux
DTMF
Regulator(3V)
AddressReservoir
Transmitter Block
Receiver Block
To x12Anodeson DrugDeliveryChip
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Demonstration31
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Alignment Issues
• Alignment between the Tx and Rx coils is very important
• This can be lateral misalignment or angular misalignment
32
D
CRX
CTX
X
Z
Y
dLRX
dLTX
X
Z
Y
O2
O1
b
aA
B
Y
r
Fig. 2. Ideal Coil Configuration for the Tx, Rx Coils.
TABLE ICOIL AND CONFIGURATION PARAMETERS
FACTORS SYMBOLCoil Radii a,bCoil Spacing DLateral Misalignment ∆Angular Misalignment γTx, Rx Coils CT x, CRxTx, Rx Number of Turns NT x, NRxTx Excitation Current IT xOhmic Losses in the Tx, Rx Coils RT x, RRxFree-space Permeability (4π ×10−7H/m) µoMagnetic Permeability of Ferrite Core µr
A. Lateral Misalignment
In the lateral misalignment case we can ignore the xand y components of the magnetic field vector since theyare parallel to the Rx plane and do not contribute to theflux lines cutting through the Rx coil. The z - componentdominates and referring to the geometry depicted in Fig. 3,(11) becomes:
HINTz =! π
0
a2 −a∆cosϕ!"
a2 +D2 +∆2 −2a∆cosϕ#3 dϕ (12)
The integral in expression (12) is a standard elliptic integraland can be evaluated as follows [12], [13]:
HINTz =a√
2m(2a∆)3/2 ·
!∆K +
am− (2−m) ·∆2−2m
·E"
(13)
where K(m), E(m) are the complete elliptic integrals of thefirst and second kind respectively and m is the modulus (0 ≤m ≤ 1) [12], [13].
m =
#4a∆
(a+∆)2 +D2
$(14)
D
CRX
CTX
X
Y
dLRX
dLTX
O2
O1
b
a
∆
r
X
Z
Y
B
A
ϕ
θ
Z Z'
Y
Fig. 3. Lateral Misalignment Configuration of the Tx, Rx Coils.
B. Angular MisalignmentIn the angular misalignment case depicted in Fig. 4 the
x and y components of the magnetic field vector cancel outdue to the circular symmetry at the center of the Rx coil.Therefore, using (11) and substituting for the parameters inFig. 4, the remaining z - component of the field can beexpressed in the standard form [14]:
HINTz =a2 ·π
%&a2 +D2
'3 (15)
D
CRX
CTX
Y
X
dLRX
dLTX
O2
O1
b
a ϕ
r
X
Z
Y
B
A
γ
Z'
θ
Z
Y
Y'
Fig. 4. Angular Misalignment Configuration of the Tx, Rx Coils.
However, as the coil is tilted by an angle γ , the newcomponent of the field which is vertical to the plane of thetilted coil (HINTAMz
) can be determined by the dot product of
!"
Authorized licensed use limited to: Edinburgh University. Downloaded on March 11,2010 at 04:26:01 EST from IEEE Xplore. Restrictions apply.
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Alignment Issues
• Alignment between the Tx and Rx coils is very important
• This can be lateral misalignment or angular misalignment
33
D
CRX
CTX
X
Z
Y
dLRX
dLTX
X
Z
Y
O2
O1
b
aA
B
Y
r
Fig. 2. Ideal Coil Configuration for the Tx, Rx Coils.
TABLE ICOIL AND CONFIGURATION PARAMETERS
FACTORS SYMBOLCoil Radii a,bCoil Spacing DLateral Misalignment ∆Angular Misalignment γTx, Rx Coils CT x, CRxTx, Rx Number of Turns NT x, NRxTx Excitation Current IT xOhmic Losses in the Tx, Rx Coils RT x, RRxFree-space Permeability (4π ×10−7H/m) µoMagnetic Permeability of Ferrite Core µr
A. Lateral Misalignment
In the lateral misalignment case we can ignore the xand y components of the magnetic field vector since theyare parallel to the Rx plane and do not contribute to theflux lines cutting through the Rx coil. The z - componentdominates and referring to the geometry depicted in Fig. 3,(11) becomes:
HINTz =! π
0
a2 −a∆cosϕ!"
a2 +D2 +∆2 −2a∆cosϕ#3 dϕ (12)
The integral in expression (12) is a standard elliptic integraland can be evaluated as follows [12], [13]:
HINTz =a√
2m(2a∆)3/2 ·
!∆K +
am− (2−m) ·∆2−2m
·E"
(13)
where K(m), E(m) are the complete elliptic integrals of thefirst and second kind respectively and m is the modulus (0 ≤m ≤ 1) [12], [13].
m =
#4a∆
(a+∆)2 +D2
$(14)
D
CRX
CTX
X
Y
dLRX
dLTX
O2
O1
b
a
∆
r
X
Z
Y
B
A
ϕ
θ
Z Z'
Y
Fig. 3. Lateral Misalignment Configuration of the Tx, Rx Coils.
B. Angular MisalignmentIn the angular misalignment case depicted in Fig. 4 the
x and y components of the magnetic field vector cancel outdue to the circular symmetry at the center of the Rx coil.Therefore, using (11) and substituting for the parameters inFig. 4, the remaining z - component of the field can beexpressed in the standard form [14]:
HINTz =a2 ·π
%&a2 +D2
'3 (15)
D
CRX
CTX
Y
X
dLRX
dLTX
O2
O1
b
a ϕ
r
X
Z
Y
B
A
γ
Z'
θ
Z
Y
Y'
Fig. 4. Angular Misalignment Configuration of the Tx, Rx Coils.
However, as the coil is tilted by an angle γ , the newcomponent of the field which is vertical to the plane of thetilted coil (HINTAMz
) can be determined by the dot product of
!"
Authorized licensed use limited to: Edinburgh University. Downloaded on March 11,2010 at 04:26:01 EST from IEEE Xplore. Restrictions apply.
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Calculated Lateral Misalignment Results
34
the unit vector (n) vertical to the plane of the rotated coiland the magnetic field vector at the center of the Rx priorto its rotation.
HINTAMz= n•HINTz (16)
As the Rx coil is rotated with respect to the x-axis the unitvector n is defined as n = (0,sinγ,cosγ).
The new component of the magnetic field vertical to theRx that produce flux through the Rx is given as follows:
HINTAMz=
a2 ·π · cosγ!"
a2 +D2#3 (17)
IV. POWER TRANSFER FUNCTION
The power transfer functions developed provide insightto the design of implanted sensor systems for a number ofimplantation depths, radial and axial displacements of thereceiver unit.
Continuing from (10), substituting for the factor HINT assolved for the geometries depicted in Fig. 3- 4, we candevelop a set of equations that express the power transferedfrom the Tx to Rx with respect to coil dimensions, corematerial, frequency, lateral movement and tilt of the Rx withrespect to the Tx.
In the lateral misalignment case the Power Transfer func-tion becomes:
PRx
PT x=
µ2o N2
T x N2Rx b4 ω2 m2
64 a RT x RRx∆3
·!
∆K +am− (2−m) ·∆
2−2m·E
"2(18)
Fig. 5 comprises of a set of graphs for power gain (18)across a typical inductive link for inductive power anddata transfer. Representative Tx and Rx coil dimensionsfor the applications studied in this paper where selectedand for each set of Tx, Rx coils the variation in couplingefficiency with respect to lateral misalignment was plotted.It is evident from the plots that the efficiency of the linkreduces with increasing axial displacement ∆ of the Rx. Thepower efficiency drops significantly as the Rx moves fromthe center of the Tx toward the circumference of the Tx coil.
In the angular misalignment case the power transfer func-tion derived is:
PRxPT x
=µ2
o N2Rx N2
T x ω2 b4 a4 π2 cosγ16RT x RRx (a2 +D2)3 (19)
Fig. 6 is a set of graphs for typical Tx, Rx coil dimensionsfor the applications studied which represent the variation incoupling efficiency with respect to angular misalignment. Itis evident from the plots that the efficiency of the link reduceswith increasing tilt of the Rx and finally it becomes zero at90.
The analysis presented in this paper is not limited toair-cored coils. Expression (10) can be easily extended toaccount for short solenoid Rx coils with a ferromagneticcore. Multiplying (10) by a factor µr we can develop a powertransfer function for air cored Tx and ferrite cored Rx coils.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16−110
−100
−90
−80
−70
−60
−50
−40
−30
−20
−10
0
Lateral Misalignment ∆, in m
Pow
er G
ain,
in d
B
α=45mm, b=10mmα=50mm, b=15mmα=55mm, b=20mmα=60mm, b=25mmα=65mm, b=30mmα=70mm, b=35mmα=80mm, b=40mmα=85mm, b=45mmα=90mm, b=50mmα=95mm, b=55mm
Fig. 5. Plot of Power Gain (in dB) across the Inductive Link for a numberof Tx, Rx Coil Combinations, in the Lateral Misalignment Case.
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
0
Angular Misalignment γ, in deg
Pow
er G
ain,
in d
B
α=45mm, b=10mmα=50mm, b=15mmα=55mm, b=20mmα=60mm, b=25mmα=65mm, b=30mmα=70mm, b=35mmα=75mm, b=40mmα=80mm, b=45mmα=85mm, b=50mmα=90mm, b=55mm
Fig. 6. Plot of Power Gain (in dB) across the Inductive Link for a numberof Tx, Rx Coil Combinations in the Angular Misalignment Case.
V. MODELING COIL GEOMETRIES
The practical coils used in short-range RFID devices fallinto the following categories:
1) Short Solenoids: a short cylindrical coil has a diameterappreciably larger than its length. Each turn of the coilhas the same radius. Short solenoids can have air orferromagnetic cores.
2) Spiral Inductors: the longitudinal thickness of thesecoils is small compared to the radial thickness and coilradius. Each coil is essentially a flat spiral. Spiral in-ductors can take many forms, square/rectangle, circularor polygon. Circular or square printed spirals are theones studied in this paper. Spirals can be made from3D copper enameled or Litz wire or fabricated on PCBor plastic substrates. Printed spirals can be fabricatedon top of a ferromagnetic layer to increase theirinductance but this process is usually expensive and
!"
Authorized licensed use limited to: Edinburgh University. Downloaded on March 11,2010 at 04:26:01 EST from IEEE Xplore. Restrictions apply.
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Calculated Angular Misalignment Results
35
the unit vector (n) vertical to the plane of the rotated coiland the magnetic field vector at the center of the Rx priorto its rotation.
HINTAMz= n•HINTz (16)
As the Rx coil is rotated with respect to the x-axis the unitvector n is defined as n = (0,sinγ,cosγ).
The new component of the magnetic field vertical to theRx that produce flux through the Rx is given as follows:
HINTAMz=
a2 ·π · cosγ!"
a2 +D2#3 (17)
IV. POWER TRANSFER FUNCTION
The power transfer functions developed provide insightto the design of implanted sensor systems for a number ofimplantation depths, radial and axial displacements of thereceiver unit.
Continuing from (10), substituting for the factor HINT assolved for the geometries depicted in Fig. 3- 4, we candevelop a set of equations that express the power transferedfrom the Tx to Rx with respect to coil dimensions, corematerial, frequency, lateral movement and tilt of the Rx withrespect to the Tx.
In the lateral misalignment case the Power Transfer func-tion becomes:
PRx
PT x=
µ2o N2
T x N2Rx b4 ω2 m2
64 a RT x RRx∆3
·!
∆K +am− (2−m) ·∆
2−2m·E
"2(18)
Fig. 5 comprises of a set of graphs for power gain (18)across a typical inductive link for inductive power anddata transfer. Representative Tx and Rx coil dimensionsfor the applications studied in this paper where selectedand for each set of Tx, Rx coils the variation in couplingefficiency with respect to lateral misalignment was plotted.It is evident from the plots that the efficiency of the linkreduces with increasing axial displacement ∆ of the Rx. Thepower efficiency drops significantly as the Rx moves fromthe center of the Tx toward the circumference of the Tx coil.
In the angular misalignment case the power transfer func-tion derived is:
PRxPT x
=µ2
o N2Rx N2
T x ω2 b4 a4 π2 cosγ16RT x RRx (a2 +D2)3 (19)
Fig. 6 is a set of graphs for typical Tx, Rx coil dimensionsfor the applications studied which represent the variation incoupling efficiency with respect to angular misalignment. Itis evident from the plots that the efficiency of the link reduceswith increasing tilt of the Rx and finally it becomes zero at90.
The analysis presented in this paper is not limited toair-cored coils. Expression (10) can be easily extended toaccount for short solenoid Rx coils with a ferromagneticcore. Multiplying (10) by a factor µr we can develop a powertransfer function for air cored Tx and ferrite cored Rx coils.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14 0.15 0.16−110
−100
−90
−80
−70
−60
−50
−40
−30
−20
−10
0
Lateral Misalignment ∆, in m
Pow
er G
ain,
in d
B
α=45mm, b=10mmα=50mm, b=15mmα=55mm, b=20mmα=60mm, b=25mmα=65mm, b=30mmα=70mm, b=35mmα=80mm, b=40mmα=85mm, b=45mmα=90mm, b=50mmα=95mm, b=55mm
Fig. 5. Plot of Power Gain (in dB) across the Inductive Link for a numberof Tx, Rx Coil Combinations, in the Lateral Misalignment Case.
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80−50
−45
−40
−35
−30
−25
−20
−15
−10
−5
0
Angular Misalignment γ, in deg
Pow
er G
ain,
in d
B
α=45mm, b=10mmα=50mm, b=15mmα=55mm, b=20mmα=60mm, b=25mmα=65mm, b=30mmα=70mm, b=35mmα=75mm, b=40mmα=80mm, b=45mmα=85mm, b=50mmα=90mm, b=55mm
Fig. 6. Plot of Power Gain (in dB) across the Inductive Link for a numberof Tx, Rx Coil Combinations in the Angular Misalignment Case.
V. MODELING COIL GEOMETRIES
The practical coils used in short-range RFID devices fallinto the following categories:
1) Short Solenoids: a short cylindrical coil has a diameterappreciably larger than its length. Each turn of the coilhas the same radius. Short solenoids can have air orferromagnetic cores.
2) Spiral Inductors: the longitudinal thickness of thesecoils is small compared to the radial thickness and coilradius. Each coil is essentially a flat spiral. Spiral in-ductors can take many forms, square/rectangle, circularor polygon. Circular or square printed spirals are theones studied in this paper. Spirals can be made from3D copper enameled or Litz wire or fabricated on PCBor plastic substrates. Printed spirals can be fabricatedon top of a ferromagnetic layer to increase theirinductance but this process is usually expensive and
!"
Authorized licensed use limited to: Edinburgh University. Downloaded on March 11,2010 at 04:26:01 EST from IEEE Xplore. Restrictions apply.
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Lab-on-a-Pill
• IDEAS Project - Collaboration between Glasgow and Edinburgh Universities
• Sensor package:
‣ pH sensor (ISFET)
‣ Oxygen sensor (Clark electrode)
‣ Direct conduction sensor
‣ Temperature sensors (p-n diode and NiCr RTD)
36
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Lab-on-a-Pill - Sensors37
TANG et al.: TOWARDS A MINIATURE WIRELESS INTEGRATED MULTISENSOR MICROSYSTEM 629
Fig. 2. An optical micrograph of (a) sensor chip 1 comprising the pH, conductivity, and temperature sensor. Sensor chip 2 (b) comprises the oxygen sensor andan optional NiCr resistance thermometer temperature sensor used to measure the temperature of the DO solution.
real-time signal monitoring of the gastro-intestinal (GI) tract[10]), four different types of microelectronic sensors have beenemployed. The sensor array consists of a dissolved oxygen(DO) sensor [11], [12], a pH-sensitive ion-selective field effecttransistor (ISFET) [13], [14], a standard PN-junction silicondiode temperature sensor [15], and a dual electrode direct con-tact conductivity sensor [16]. The temperature sensor measuresthe body core temperature. However, it is also essential fortemperature compensation of the other three sensors, since therespective parameters are highly temperature-dependent. ThepH sensor measures the acidity of the stomach, although itwill also be used to predict the location of the pill within theGI tract, since various portions of the digestive system havean associated range of pH values. The conductivity sensormeasures the total level of dissolved solids and thus provides anindirect estimation of the intestinal contents. The oxygen sensorwill map the activity of aerobic bacteria within the intestine.The complete sensor array is fabricated on two 5 5 mmsilicon chips.The DO sensor, which includes an optional NiCr resistance
thermometer [Fig. 2(b)], was fabricated on a single chip. Thesensor consists of a three-electrode electrochemical cell of500 m in diameter and a volume of 7.9 nL. The workingelectrode comprises a microelectrode array of 121 circulargold electrodes, each 5 m in diameter, to reduce responsetime and to prevent diffusion-limited transport to the electrodesurface. The interelectrode spacing is 20 m, with a total areaof the working electrode of approximately 2.5 10 m . Anintegrated Ag–AgCl electrode (1.5 10 m ) made from500-nm-thick evaporated silver, of which 50% was oxidizedto AgCl in a 1.0 M KCl solution with the aid of chronoamper-ometry, acts as the reference electrode. The counter electrodeis made of gold, covering an area of 1.0 10 m . In futuredevices, the electrochemical cell will contain a 0.1 M KCl gel(agarose) to promote the constant concentration of chlorideions required to maintain a stable potential of the Ag–AgClreference electrode and to prevent leakage to the external solu-tion. The electrolyte chamber will be covered by a 20- m-thick
layer of oxygen-permeable teflon, forming a barrier to thesurrounding medium in which the DO is to be measured.The second chip [Fig. 2(a)] is based on a commercial ISFET
and silicon diode temperature sensor (ESIEE, France). Theconductivity sensor and a miniaturised Ag–AgCl referenceelectrode are defined by post lithographic pattern integration.The conductivity sensor comprises two gold electrodes with anarea of 5.0 10 m each spaced 400 m apart, which op-erate in a direct contact mode with the surrounding solution inwhich the conductivity is measured. In contrast, the Ag–AgClreference electrode (1.5 10 m ) is fabricated in a similarmanner to the oxygen sensor. Similarly, the ISFET pH sensorand its reference electrode are located in a 7.9-nL electrodechamber designed to contain a 0.1-M KCl gel electrolytesolution to promote a constant concentration of chloride ionsin the presence of the Ag–AgCl electrode. In future devices, afinal 50- m-thick layer of cation-selective Nafion® membranewill cover the electrode chamber and forms a physical barrierbetween the chamber and the solution in which pH is to bemeasured.
B. Sensor Fusion
The third chip in the microsytem is the ASIC. It includesthe analog signal conditioning circuits for each sensor, analogmultiplexers, a 10-b analog-to-digital converter (ADC) anda digital-to-analog converter (DAC), an oscillator, as wellas digital modules for microsystem scheduling, coding andtransmitter control. With minimal compromise in terms of thesensitivity of the sensors, the analog circuits aim to minimizephysical dimensions and power consumption. At first glance,some of our design choices would appear at odds with commonpractice that tends toward far more complex circuits such as thelock-in amplifier or the instrumentation amplifier. However,it is the aim of this project to achieve miniature size and highrobustness while keeping power consumption and componentcount to a minimum—so the common design priorities havebeen changed.
Authorized licensed use limited to: Edinburgh University. Downloaded on March 12,2010 at 04:40:30 EST from IEEE Xplore. Restrictions apply.
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Lab-on-a-Pill - Package38
630 IEEE SENSORS JOURNAL, VOL. 2, NO. 6, DECEMBER 2002
Fig. 3. A schematic illustrating the design of the future microsystem. The components, which comprise the sensors, ASIC, transmitter, and batteries, are attachedto a PCB platform upon integration in a capsule machined from solid PEEK. The capsule has an outer diameter of 16 mm and a length of 55 mm.
In this prototype, the microsystem was designed to workas a real-time multisensor transducer without local signalprocessing. Although a finite state machine was used for taskscheduling in the present configuration, it will be replaced bya small microprocessor for self-reconfiguration in the future.As shown in Fig. 1, this state machine as part of the digitalplatform is timed from a simple RC relaxation oscillatorwithout temperature compensation. The timing precision ofthe oscillator is not high, owing to the fabrication toleranceof on-chip resistors and capacitors, but this had been foreseenand therefore we ensured that it would not have any strongadverse effect on our system specifications. Our approach isto mix a relatively low signal bandwidth of 1.51 kHz withthe transmitter’s reference frequency of 20 MHz so that thereis minimal shift (less than 0.01%) in the transmitted signal’sfrequency. The nominal clock frequency of the RC oscillatoris 24.15 kHz with a duty cycle of 49.8%. This low clockfrequency provides an adequate sampling rate whilst keepingthe dynamic power dissipation acceptably low.For future implementation, an analog stochastic classifier
based on the restricted Boltzmann machine (RBM) [17] willbe introduced to achieve intelligent sensor fusion. Althoughartificial neural networks (ANNs) are well established in manyapplications, most of the data computation is done throughsoft computing [18]–[20] and the rest is achieved througheither microcontroller units (MCU) or digital signal processing(DSP) that require huge computational power. Unlike theMulti-layer Perceptron (MLP) that requires high accuracy(12 b) in weight memory [21] and a complex, supervisedupdate rule, the RBM design can be implemented easily witha simple weight changing circuit [22]. Moreover, the RBMdesigns have demonstrated high reliability in heartbeat analysis[22] and hand-writing recognition [23].
C. Power ManagementTwo smart power management strategies have been imple-
mented. The primary strategy is to disable “idle” function blocks
during different task phases. The second strategy uses a simpleDSP data compression algorithm to achieve low-power serialbitstream transmission. The algorithm decides when transmis-sion is required by comparing the most recent sample with theprevious sampled data. This technique is particularly effectivewhen the measuring environment is at quiescence, a commonlyencountered situation in many applications.The algorithm has been tested using an extended data set,
which comprises 67 cases of 24-h dynamic gastric pH data[24]. The raw data have an average kurtosis of 5.3 2.8 andan autocorrelation of 0.99 0.01. The compression ratio, thatis defined as the ratio of the raw data stream length to thecompressed data stream length, is found to be equal to 2.1 0.1with a distortion degree of 0.008 0.001 with the given dataset. If similar conditions apply to the microsystem discussedin this paper, a total power saving of approximately 50% forthe transmitter can be achieved simply by introducing a smalldigital system overhead ( 10 ).
D. Future Wireless Communications
The wireless communication is proposed to be achieved byimplementing a simple miniature low-power transmitter thatuses a frequency-shift-keying (FSK) modulation method. Thetransmitter is designed out of ultraminiaturized discrete com-ponents, yielding a 5 8 11 mm package. This transceiverhas a crystal oscillator that is bonded onto the substrate of theprinted circuit board (PCB) (Fig. 3). This oscillator providesa highly precise carrier frequency that will allow reliable datacollection at the base station despite the inaccuracy of thetiming of the ASIC. Preliminary experiments have shownthat the transmitter has a cover range of 1 m and the receivedsignal has an SNR of 78 dB at 20 MHz. A fully integratedSoC implementation of a transmitter using standard CMOStechnology is being developed.
Authorized licensed use limited to: Edinburgh University. Downloaded on March 12,2010 at 04:40:30 EST from IEEE Xplore. Restrictions apply.
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Lab-on-a-Pill - Package39
630 IEEE SENSORS JOURNAL, VOL. 2, NO. 6, DECEMBER 2002
Fig. 3. A schematic illustrating the design of the future microsystem. The components, which comprise the sensors, ASIC, transmitter, and batteries, are attachedto a PCB platform upon integration in a capsule machined from solid PEEK. The capsule has an outer diameter of 16 mm and a length of 55 mm.
In this prototype, the microsystem was designed to workas a real-time multisensor transducer without local signalprocessing. Although a finite state machine was used for taskscheduling in the present configuration, it will be replaced bya small microprocessor for self-reconfiguration in the future.As shown in Fig. 1, this state machine as part of the digitalplatform is timed from a simple RC relaxation oscillatorwithout temperature compensation. The timing precision ofthe oscillator is not high, owing to the fabrication toleranceof on-chip resistors and capacitors, but this had been foreseenand therefore we ensured that it would not have any strongadverse effect on our system specifications. Our approach isto mix a relatively low signal bandwidth of 1.51 kHz withthe transmitter’s reference frequency of 20 MHz so that thereis minimal shift (less than 0.01%) in the transmitted signal’sfrequency. The nominal clock frequency of the RC oscillatoris 24.15 kHz with a duty cycle of 49.8%. This low clockfrequency provides an adequate sampling rate whilst keepingthe dynamic power dissipation acceptably low.For future implementation, an analog stochastic classifier
based on the restricted Boltzmann machine (RBM) [17] willbe introduced to achieve intelligent sensor fusion. Althoughartificial neural networks (ANNs) are well established in manyapplications, most of the data computation is done throughsoft computing [18]–[20] and the rest is achieved througheither microcontroller units (MCU) or digital signal processing(DSP) that require huge computational power. Unlike theMulti-layer Perceptron (MLP) that requires high accuracy(12 b) in weight memory [21] and a complex, supervisedupdate rule, the RBM design can be implemented easily witha simple weight changing circuit [22]. Moreover, the RBMdesigns have demonstrated high reliability in heartbeat analysis[22] and hand-writing recognition [23].
C. Power ManagementTwo smart power management strategies have been imple-
mented. The primary strategy is to disable “idle” function blocks
during different task phases. The second strategy uses a simpleDSP data compression algorithm to achieve low-power serialbitstream transmission. The algorithm decides when transmis-sion is required by comparing the most recent sample with theprevious sampled data. This technique is particularly effectivewhen the measuring environment is at quiescence, a commonlyencountered situation in many applications.The algorithm has been tested using an extended data set,
which comprises 67 cases of 24-h dynamic gastric pH data[24]. The raw data have an average kurtosis of 5.3 2.8 andan autocorrelation of 0.99 0.01. The compression ratio, thatis defined as the ratio of the raw data stream length to thecompressed data stream length, is found to be equal to 2.1 0.1with a distortion degree of 0.008 0.001 with the given dataset. If similar conditions apply to the microsystem discussedin this paper, a total power saving of approximately 50% forthe transmitter can be achieved simply by introducing a smalldigital system overhead ( 10 ).
D. Future Wireless Communications
The wireless communication is proposed to be achieved byimplementing a simple miniature low-power transmitter thatuses a frequency-shift-keying (FSK) modulation method. Thetransmitter is designed out of ultraminiaturized discrete com-ponents, yielding a 5 8 11 mm package. This transceiverhas a crystal oscillator that is bonded onto the substrate of theprinted circuit board (PCB) (Fig. 3). This oscillator providesa highly precise carrier frequency that will allow reliable datacollection at the base station despite the inaccuracy of thetiming of the ASIC. Preliminary experiments have shownthat the transmitter has a cover range of 1 m and the receivedsignal has an SNR of 78 dB at 20 MHz. A fully integratedSoC implementation of a transmitter using standard CMOStechnology is being developed.
Authorized licensed use limited to: Edinburgh University. Downloaded on March 12,2010 at 04:40:30 EST from IEEE Xplore. Restrictions apply.
528 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 51, NO. 3, MARCH 2004
is particularly effective when the measuring environment is atquiescent, a condition encountered in many applications [27].The entire design was constructed with a focus on low powerconsumption and immunity from noise interference. The dig-ital module was deliberately clocked at 32 kHz and employeda sleep mode to conserve power from the analog module. Sep-arate on-chip power supply trees and pad-ring segments wereused for the analog and digital electronics sections in order todiscourage noise propagation and interference.
C. Radio Transmitter
The radio transmitter was assembled prior to integration in thecapsule using discrete surface mount components on a single-sided printed circuit board (PCB). The footprint of the standardtransmitter measured mm including the integratedcoil (magnetic) antenna. It was designed to operate at a trans-mission frequency of 40.01 MHz at 20 C generating a signalof 10 kHz bandwidth. A second crystal stabilized transmitterwas also used. This second unit was similar to the free run-ning standard transmitter, apart from having a larger footprint of
mm, and a transmission frequency limited to 20.08MHz at 20 C, due to the crystal used. Pills incorporating thestandard transmitter were denoted Type I, whereas the pills in-corporating the crystal stabilized unit were denoted Type II. Thetransmission range was measured as being 1 meter and the mod-ulation scheme frequency shift keying (FSK), with a data rate of1 .
D. Capsule
The microelectronic pill consisted of a machined biocom-patible (noncytotoxic), chemically resistant polyether-terketone(PEEK) capsule (Victrex, U.K.) and a PCB chip carrier actingas a common platform for attachment of the sensors, ASIC,transmitter and the batteries (Fig. 3). The fabricated sensorswere each attached by wire bonding to a custom made chip car-rier made from a 10-pin, 0.5-mm pitch polyimide ribbon con-nector. The ribbon connector was, in turn, connected to an in-dustrial standard 10-pin flat cable plug (FCP) socket (RadioSpares, U.K.) attached to the PCB chip carrier of the microelec-tronic pill, to facilitate rapid replacement of the sensors whenrequired. The PCB chip carrier was made from two standard1.6-mm-thick fiber glass boards attached back to back by epoxyresin which maximized the distance between the two sensorchips. The sensor chips were connected to both sides of the PCBby separate FCP sockets, with sensor Chip 1 facing the top face,with Chip 2 facing down. Thus, the oxygen sensor on Chip 2 hadto be connected to the top face by three 200- m copper leadssoldered on to the board. The transmitter was integrated in thePCB which also incorporated the power supply rails, the con-nection points to the sensors, as well as the transmitter and theASIC and the supporting slots for the capsule in which the chipcarrier was located.
The ASIC was attached with double-sided copper conductingtape (Agar Scientific, U.K.) prior to wirebonding to the powersupply rails, the sensor inputs, and the transmitter (a processwhich entailed the connection of 64 bonding pads). The unitwas powered by two standard 1.55-V SR44 silver oxidecells with a capacity of 175 mAh. The batteries were serial con-nected and attached to a custom made 3-pin, 1.27-mm pitch plug
Fig. 3. Schematic diagram (top) of the remote mobile analytical microsystemcomprising the electronic pill. The prototype is 16 55 mm, weights 13.5 g.The Type I unit consist of the microelectronic sensors at the front enclosed bythe metal clamp and rubber seal (1) which provide a 3-mm-diameter accesschannel to the sensors (2). The front section of the capsule, physically machinedfrom solid PEEK, is illustrated (3) with the rear section removed to illustratethe internal design. The front and rear section of the capsule is joined by ascrew connection sealed of by a Viton-rubber o-ring (4). The ASIC control chip(5) is integrated on the common PCB chip carrier (6) which incorporate thediscrete component radio transmitter (7), and the silver oxide battery cells (8).The battery is connected on the reverse side of the PCB (9). The Type II unit isidentical to the Type I with exception of an incorporated crystal stabilized radiotransmitter (10) for improved temperature stability.
by electrical conducting epoxy (Chemtronics, Kennesaw, GA).The connection to the matching socket on the PCB carrier pro-vided a three point power supply to the circuit comprising a neg-ative supply rail ( 1.55 V), virtual ground (0 V), and a positivesupply rail (1.55 V). The battery pack was easily replaced duringthe experimental procedures.
The capsule was machined as two separate screw-fitting com-partments. The PCB chip carrier was attached to the front sec-tion of the capsule (Fig. 3). The sensor chips were exposed tothe ambient environment through access ports and were sealedby two sets of stainless steel clamps incorporating a 0.8- m-thick sheet of Viton fluoroelastomer seal. A 3-mm-diameter ac-cess channel in the center of each of the steel clamps (incl. theseal), exposed the sensing regions of the chips. The rear sec-tion of the capsule was attached to the front section by a 13-mmscrew connection incorporating a Viton rubber O-ring (JamesWalker, U.K.). The seals rendered the capsule water proof, as
Authorized licensed use limited to: Edinburgh University. Downloaded on March 11,2010 at 06:55:34 EST from IEEE Xplore. Restrictions apply.
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Pond Skater - EWOD40
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Pond Skater Wireless EWOD
41
Stewart SmithBiosensors and InstrumentationU-Tokyo Special Lectures
Pond Skater Demonstration42