NORTHWESTERN UNIVERSITY Electrostatic Force on a Human Fingertip A DISSERTATION SUBMITTED TO THE GRADUATE SCHOOL IN PARTIAL FULFILLMENT OF THE REQUIREMENTS for the degree MASTER OF SCIENCE Field of Mechanical Engineering By David John Meyer EVANSTON, ILLINOIS December 2012
53
Embed
Electrostatic Force on a Human Fingertip Force on a Human Fingertip ... I modeled a dependence of the electrostatic force on actuation frequency, ... Chapter 4. Measuring Electrostatic
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
)The electrostatic force on the fingertip is the derivative of the potential energy with respect to
the air gap. When the finger is in contact with the screen, this expression is evaluated as dg
approaches zero, given in equation 3.4.
(3.4) Fe = − ∂U∂dg
∣∣∣∣dg→0
=ε2sc(−Qfdi + VFAε0εi)(Qfdi + VFAε0εi)
2Aε0(dscεi + diεsc)2
18
Defining capacitances for the stratum corneum and insulator as shown in 3.5, it is much easier
to understand the terms, shown in 3.6.
(3.5) Ci =ε0εiA
diCsc =
ε0εscA
dsc
(3.6) Fe =
C2i C
2sc
(V 2F −
Q2f
C2i
)2Aε0(Ci + Csc)2
By inspection, this equation reveals the square law dependence of electrostatic force on
voltage, as well as the dependence on capacitances of the stratum corneum and insulating layer
atop the ITO. More interestingly, in the case that the free charge on the surface of the glass
builds up to match that on the ITO (Qf → CiV ), the electrostatic force on the finger completely
disappears. This is because the charge on the surface of glass reduces the electric field in the
fingertip. Charge can flow to the surface through any resistive path, which in this case can be
the fingertip itself.
Assuming that no charge flows to the surface of the glass, however, the electrostatic force
on the finger is the same as the force on a capacitor with two dielectrics in series. In terms of
only non-derived quantities, this force is:
(3.7) Fe =ε0AV
2F
2
(diεi
+dscεsc
)2
Equation 3.7 agrees with previously published work by Strong and Troxel[13]. However, this
has been more recently contested by Kaczmarek et al. [17], who noticed that if the insulating
19
layer is zero thickness, the model predicts:
(3.8) F1 =ε0AV
2F
2
(dscεsc
)2
which does not agree with the standard textbook definition of force on a parallel plate capacitor,
given by:
(3.9) F2 =ε0AV
2F
2
(dscεsc
)dsc
The difference between the two models, F1 and F2, is a multiplication factor of relative permit-
tivity of the dielectric in the capacitor. I argue that equation 3.9 is an incorrect application of
electrostatic theory to the fingertip-surface system, and that F2 is not the proper expression for
force on the fingertip.
This discrepancy can be more easily analyzed with the Lorentz force law, relating force on a
sheet of charge to the electric field. Assuming zero insulator thickness in both models, Q = CVF ,
and C = ε0εscAdsc
, equations 3.8 and 3.9 become:
(3.10) F1 =Q2
2ε0A
(3.11) F2 =Q2
2ε0εscA
Let us consider the two conducting surfaces as sheets of constant charge. The electric field in a
vacuum near a large sheet of charge (the ITO) is derived using Gauss’ law as E = Q2ε0A
. The
force on a parallel sheet of equal charge (the conductive finger tissue) is F = QE, which agrees
with F1. If we then assume the world is filled with a dielectric fluid, the electric field is reduced
by a factor of the relative permittivity. Thus, the force on a sheet of charge inside the dielectric
20
also reduced, as predicted by the model F2. This is only true, though, if the sheets of charge
are inside the dielectric and experience the reduced electric field.
In the case where a slab of dielectric is introduced between two sheets of constant charge, I
argue that the force between the two sheets does not change, because the electric field is reduced
only inside the dielectric. The slab instead reduces the voltage difference between the sheets.
The model F2 is only correct for a world filled with a liquid dielectric, in which all electric
field strengths are reduced by a factor of the relative permittivity. F1 is the proper model for
a capacitor with a slab of solid dielectric between two conductors, which is the case for the
stratum corneum and insulator on the touch screen. Therefore, the electrostatic force on the
fingertip is correctly described by equation 3.6.
3.2. Dynamic Analysis
Electrostatic theory predicts that the force on the fingertip is determined by the capacitive
capabilities of the skin-surface interface, as well as the free charge on the surface of the glass. It
is therefore crucial to understand the dynamics of the entire electrical system in order to create a
known electrostatic force on the finger. A dynamic analysis also explains how a finite resistivity
in the stratum corneum allows charge to flow to the surface and reduce the force on the finger.
To begin the analysis, I developed a simple linear circuit model for the finger and 3M screen.
In the figure 3.3, Csc and Ci are the capacitances of the stratum corneum and insulator
respectively as defined in section 3.1. Additional elements which don’t directly contribute to
electrostatic force on the finger include the series resistance in the finger, Rf , and finite resistance
of the stratum corneum, Rsc. A series capacitance, Cs, is also modeled because the touch screen
has a capacitor in the cable between the input and ITO.
Referring to equation 3.6, it is clear that the only two dynamic elements which affect the
electrostatic force on the fingertip are VF , the voltage drop from the ITO layer to the finger
21
Glass
Indium Tin Oxide
1µm SiO2
Copper Foil
VA
Figure 3.2. Circuit element manifestation of finger and screen
Cs
VF
Ci
Csc RscQf
Rf
VA
Figure 3.3. Circuit diagram of the system
tissue, and Qf , the charge which leaks through the stratum corneum and accumulates on the
surface. A dynamic analysis of this developed linear model shown above predicts the electrostatic
force on the finger for any given actuation signal, VA, and reveal frequencies at which a force
will be produced.
Using Kirchhoff’s current and voltage laws, two transfer functions describing the behavior
of Qf and VF with respect to the supply voltage are written. These transfer functions, derived
22
in Appendix A, are shown in equation 3.12.
VF (s)
VA(s)=
Cs +[CscRsc(Ci + Cs)
]s[
Ci + Cs
]+[CscRsc(Ci + Cs) + CiCs(Rf +Rsc)
]s+
[CiCsCscRfRsc
]s2
Qf (s)
VA(s)=
CiCs[Ci + Cs
]+[CscRsc(Ci + Cs) + CiCs(Rf +Rsc)
]s+
[CiCsCscRfRsc
]s2
(3.12)
To then calculate the force produced by a certain supply voltage signal, the equations in 3.12
are simulated in the time domain using MATLAB. Plugging the resulting time-varying signals
VF and Qf into equation 3.6 results in a time-varying electrostatic force. Averaging this force
over time gives the electrostatic force felt by the fingertip.
To establish a starting point for experimental work, I completed this analysis using estimated
circuit parameters. Using skin properties previously published[18], the resistivity and dielectric
constant of the stratum corneum at 1 kHz are ρsc ≈ 33 kΩm and εsc ≈ 1000 respectively.
Assuming the contact area of the finger is a circle with radius 4mm, the electrical parameters
are:
Rsc =ρscdscA
=(33kΩm)(200µm)
π(4mm)2= 130kΩ
Csc =ε0εscA
dsc=
(8.85× 10−12 Fm)(1000π)(4mm)2
200µm= 2.2nF
Ci =ε0εiA
di=
(8.85× 10−12 Fm)(3.9π)(4mm)2
1µm= 1.7nF
(3.13)
The estimated series resistance is 500 Ω, the capacitor in the screen is 140 pF , and the
resulting predicted electrostatic force for 140 V actuation signals of varying frequencies is shown
in figure 3.4. At low frequencies, the resistivity of the stratum corneum allows charge to flow
from the body and effectively nullify the force on the fingertip. At high frequencies, charge
cannot build up on the capacitors in the circuit due to the finger resistance, Rf . Because the
23
Electrostatic
Force
Frequency (Hz)
101 102 103 104 105 106 107 108 1090
0.05
0.1
0.15
0.2
Figure 3.4. Predicted Force over Frequency
electrical properties of fingertips are highly variant, the values of these knee frequencies are quite
uncertain. Nevertheless, the model predicts that DC voltages create no electrostatic force, and
suggests that force is produced at a wide range of frequencies.
24
CHAPTER 4
Measuring Electrostatic Force
The goal of my experimental work is to verify the models presented in the previous chap-
ter. In an ideal world, electrostatic attraction would be measured under different actuation
signals and compared with theory. Unfortunately, measuring the electrostatic attraction is very
problematic. Typically, a force is measured with a transducer placed in the load path, but this
method is not possible at the finger-screen interface. The electrostatic force relies on an electric
field being present at the fingertip, and any force transducer between the finger and glass would
disrupt the electric field substantially. For this reason, I measured the friction modulation due
to electrostatic attraction, referred to as electrostatic friction.
25
4.1. Measurement Considerations
Before I started measuring friction, I considered two other methods to directly measure
electrostatic attraction. The first method involved measuring energy in the capacitive coupling
formed by the finger and screen. The second focused on imaging optically the compression of
the fingertip under electrostatic force. Figure 4.1 is a simplified diagram showing forces and
stiffnesses in the finger-screen interface. Spring k1 represents the soft tissue between the bone
Electrostatic Force
Pressing Force
Finger
3M screen
Soft Tissue Stratum Corneum
k1 k2 k3
Figure 4.1. Force diagram of finger-screen interface
and fingertip surface, k2 models the stiffness of the stratum corneum, and k3 is a load cell. There
is an important distinction to be made between the “pressing force” and “electrostatic force”.
Both are normal forces at the fingertip-screen contact, but the former is applied through the
bone, whereas the latter is only present in the shallow fingertip tissue, represented by sping k2.
The relative stiffnesses of these springs provide useful insight to the behavior of the fingertip
under electrostatic force. The elastic modulus of the stratum corneum has been previously
recorded around 107 − 108 Pa[19]. Calculating k = EAt for a finger in contact with the screen,
26
the stiffness of the stratum corneum is around k2 ≈ 107 N/m. Compared with the stiffness of
the soft tissue, recorded previously around k1 ≈ 103 N/m[20], the stratum corneum is effectively
incompressible. Because all the springs are in series, each one is loaded by the pressing force.
The stratum corneum experiences the addition of pressing force and electrostatic force, as is
shown in the diagram. Because k2 is relatively incompressible, an electrostatic force will not
cause much deflection, and will not be measurable by the load cell, k3. I instead attempted an
energy-based approach to measure electrostatic attraction.
The energy in a capacitor is closely related to the force on the two capacitor plates. In
the case of the finger-screen interface there is potential energy in the capacitive bond when an
electrostatic force is present (equation 3.3). One way to measure this energy is to push the
finger into the glass, turn the actuation voltage on, then pull the finger away from the glass.
The energy used to push the finger into the glass is stored in the three springs shown in figure
4.1. With the electrostatic force applied, however, the energy recovered by pulling the finger
away from the glass is reduced. The difference in the energy during the push in phase and pull
away phase is the energy in the electrostatic bond.
Unfortunately, the energy-based approach also suffers from the softness of the finger tissue.
Based on data from my preliminary experiments, pushing the finger into the screen with 0.5 N
of pressing force uses about 150 µJ of energy. Electrostatic theory derived in section 3.1 predicts
the energy in the capacitive bond at 0.2 µJ. The force sensor cannot reliably distinguish this
small of an energy difference. Because I couldn’t successfully measure the electrostatic attraction
with a separate force transducer, I considered the fingerprint itself a transducer, and explored
the possibility of optically assessing fingerprint compression.
A finger in contact with a flat surface rests on the epidermal ridges of the fingerprint. The
real contact area encapsulates only the tips of the ridges that are in direct contact with the
glass. The apparent contact area is roughly an ellipse which encloses the region of real contact.
27
As the finger is pressed into the surface with greater force, the fingertip flattens out and the
contact area increases. An electrostatic attraction on the stratum corneum should increase the
real contact area of the finger by compressing the dermal ridges. By observing this change in
real contact area, the electrostatic force can be measured optically.
Fingerprints can be imaged through a prism using the principle of frustrated total internal
reflection (FTIR) in a method similar to that of Andre et al[21] and Soneda and Nakana[22].
These studies showed that not only was real contact area measurable with FTIR, but that it
changes dramatically with respect to pressing force. An undergraduate student in the lab, Paul
Barnes, conducted preliminary experiments to assess whether electrostatic force would also cause
a change in contact area, but the results showed no conclusive evidence of that effect. These
results are not completely shocking though, since the pressing force is different in nature from
the electrostatic force. When the bone presses on the fingertip into the screen, a large amount
of tissue distributes the pressing force over the whole area of the finger. Electrostatic force, on
the other hand, only substantially acts on tissue that is within 100 microns of the surface, and
can only compress the epidermal ridges themselves. Since the electrostatic force acts on such a
stiff material, deflection should be very minimal, and therefore difficult to see optically.
FTIR may still be a valid method for observing fingertip compression due to electrostatic
attraction. Improving optics and image processing could yield much finer resolution than was
obtained in preliminary experiments. The work presented here, though, focuses on measuring
the effect of electrostatics on fingertip friction. Electrostatic friction is the change in friction
due to the electric field, and can be both measured and perceived by touch over a wide range of
conditions.
28
4.2. Tribometer Construction
I designed and built a tribometer, shown in figure 4.2 to measure fingertip friction force
under different conditions. Fingertip friction in general is not very constant in time, largely due
to moisture variations, and can vary significantly in just a few seconds[23]. To extract useful
information from otherwise noisy friction data, electrostatic actuation is pulsed on and off while
the finger is traversing the glass. Consequently, to capture rapid changes in electrostatic friction,
a high-bandwidth sensor is needed. Estimations from chapter 3 show the electrostatic force will
be on the order of 100 mN, so a very high-sensitivity transducer is needed to accurately measure
friction changes due to electrostatic attraction. The ATI Nano17 force/torque sensor chosen for
cable-drivenslider
3M MicroTouchscreen
6-axisforce sensor
controlledDC motor
counterweight andmagnetic damper
Figure 4.2. Tribometer drawing
this work offers 3 mN resolution and has a resonant frequency of 7200 Hz. The sensor itself is
mounted just below the 3M screen on a balance beam which loads the finger with a pressing force
controllable by counterweights. To prevent vibratory motion, the balance beam is magnetically
damped.
The finger is held by Velcro straps in a sliding carriage, which is driven translationally by
a motor and capstan drive to keep finger velocity constant over a large portion of the screen.
To control the motor’s velocity, a shaft encoder is used in a velocity feedback loop. The control
29
system is implemented on a PC104 stack running Matlab’s xPC real-time operating system. I
designed and built an RS232-controlled function generator to create the waveforms for electro-
static actuation. Its output feeds into a high-voltage piezo amplifier to create the high voltages
necessary for electrostatic attraction. The xPC machine records all experimental data at 1kHz
and transmits it via TCP/IP to a PC running Matlab.
PC104 (xPC)
Sensoray 526PC
(MATLAB)FunctionGenerator
Piezo Amp Motor Amps
Tribometer
F/T
Controller
Touch Screen
F/T
Sensor
Motors Encoders
RS232TCP/IP
Figure 4.3. Block diagram of experimental setup
A friction measurement begins by raising the screen up to the finger, and swiping the finger
back and forth across the screen with a trapezoidal velocity pattern. Beginning on the left side
of the screen, the finger is dragged back and forth several times. During each even-numbered
swipe, the electrostatic effect is switched on when the finger is on the right-hand side of the
30
glass; during an odd-numbered swipe, the left-hand side is the active portion. Data observed
when the electrostatic actuation is off is used to establish baseline friction.
Each subject washed his or her hands prior to beginning the experiment, and talcum powder
was used to dry the fingertip and eliminate friction effects due to moisture. Every forty five
seconds, the finger was removed from the surface for twenty seconds to prevent build up of
moisture due to occlusion of the sweat glands.
4.3. Electrostatic Friction
To observe electrostatic friction due to DC actuation, a finger was swiped across the screen
while a 140 volt signal used for the electrostatic actuation was switched on and off during travel.
The results of this experiment are shown in figure 4.4. The gray portions of the plot indicate
times during which the actuation voltage is applied. As is clearly shown by the dark blue line,
a DC actuation voltage does not create a measurable change in friction. Two example swipes of
the finger are shown, as indicated by the finger velocity profile in green. The first swipe shown
is even-numbered, during which the electrostatic actuation is on in the right half of the screen.
The odd-numbered swipe shows the opposite behavior, with the actuation on in the left half.
Normal force, shown in light blue, is held constant with the damped balance beam.
Finger VelocityFriction ForceNormal Force
Time (s)
Square Wave, 140 Volts DCVelocity
(mm/s)
Force
(N)
20 22 24 26 28 30
0
−0.5
0
0.5
Figure 4.4. Overview of raw friction data taken with tribometer
31
Next, a 10 Hz sine wave was applied to the screen’s input, the results of which are shown in
figure 4.5. In figure 4.5a, the overview of one example swipe data is shown in the top plot. The
corresponding electrostatic actuation voltage is shown in the bottom subplot. Close inspection
of the normal and friction forces reveals time-varying electrostatic friction occurring at the
surface, shown in figure 4.5b. The electrostatic friction peaks at the same time the actuation
voltage reaches an extremum, and is therefore twice the frequency of the actuation signal. It
is important to notice that the measured normal force does not change. This is because the
electrostatic normal force is not measured by the load cell. What is measured is only the
pressing force on the finger, controlled by the counterweight applied to the balance beam. These
data support the force model in figure 4.1 and verify that electrostatic friction is measurable by
the tribometer.
Finger VelocityFriction ForceNormal Force
Voltage
(V)
Time (s)
Sine Wave, 140 Volts, 10 Hz
Velocity
(mm/s)
Force
(N)
14 16 18
−200
0
200
−40
0
40
−1.0
−0.5
0
0.5
1.0
(a) Overview of raw data with a 10 Hz sine waveelectrostatic effect.
Friction ForceNormal Force
Voltage
(V)
Time (s)
Force
(N)
Sine Wave, 140 Volts, 10 Hz
14 14.5 15
14 14.5 15
−200
0
200
0
0.1
0.2
0.3
0.4
0.5
0.6
−50
0
(b) Close-up of raw data during electrostaticeffect switch time. Supply voltage is 10 Hz,friction is oscillatory at 20 Hz.
Figure 4.5. Electrostatic friction under low-frequency AC actuation
32
Finger VelocityFriction ForceNormal Force
Time (s)
Square Wave, 140 Volts, 10 kHz
Velocity
(mm/s)
Force
(N)
16 17 18 19 20 21 22 23 24 25
−50
0
50
−0.4
−0.2
0
0.2
0.4
Figure 4.6. Overview of raw data taken for 10 kHz
In an attempt to maximize electrostatic friction, I experimented with high-frequency square
waves as actuation signals. Figure 4.6 shows raw measured forces for a 140 volt 10 kHz square
wave actuation signal. When the electrostatic actuation is turned on, during the gray portions
of the plot, the electrostatic friction force remains continuous. Because high-voltage square
waves yield consistently high electrostatic friction force, these signals are used to investigate
relationship between voltage and force.
4.4. Friction Analysis
The data obtained with the tribometer show that electrostatic friction is a clear increase in
friction force due to electrostatic actuation. With regard to haptics, this shows that electrostatics
is a valid actuation method for modulating friction force on a surface. However, the transduc-
tion between electric field and increased friction is not well understood. A friction model is a
mathematical relationship between the normal and lateral forces at the fingertip-glass contact.
I created a friction model for each subject based on data taken during the experiment. Using
these models, electrostatic friction can be attributed to an electrostatic attraction force between
the finger and screen.
33
To reduce the effect of measurement noise, I averaged the measured normal and friction
forces across constant velocity segments. Each swipe of the finger yields 4 data points, two in
each direction,which were used to extract electrostatic friction and pressing force. A sample
plot with the averaged data points is shown in figure 4.7. By repeating these measurements
Finger VelocityFriction ForceNormal Force
Time (s)
Square Wave, 140 Volts, 10 kHz
Velocity
(mm/s)
Force
(N)
16 17 18 19 20 21 22 23 24 25
−50
0
50
−0.4
−0.2
0
0.2
0.4
Figure 4.7. Raw data with averaged data points shown
many times with different pressing forces applied, data points can be plotted on a friction versus
normal force plot, as shown in 4.8. This example shows one subject’s data for the 140 volt
10kHz square wave input tests. Two data groups appear, one for data points during which the
electrostatic force was present, and one for when it was not. The data taken when electrostatic
actuation is off provide a model for fingertip-surface friction, and the vertical difference between
data points represents electrostatic friction.
The fingertip-surface friction model shows a linear relationship between friction and normal
force, which follows the Coulombic model of dry friction. Performing a linear regression on these
data gives a friction coefficient, µ. Assuming µ itself does not change due to electric field, the
increase in friction force can be attributed to an electrostatic attraction force, inferred by the
34
Electrostatics Off
Electrostatics On
∆F
Fe
FrictionForce
(N)
Normal Force (N)
Square Wave, 140 Volts, 10 kHz
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Figure 4.8. Sample lateral versus normal force plot of averaged data
equation:
(4.1) Fe =∆F
µ
where Fe is the electrostatic force, µ is the slope determined by regression, and ∆F is the average
electrostatic friction force. For each half-swipe of the finger, ∆F was calculated. Gathering
these values gives an average and a standard error, which are propagated through equation 4.1
resulting in an inferred electrostatic force and confidence interval. In the example figure 4.8, the
inferred electrostatic force is 250 mN, with a standard error of 70 mN.
35
As can be seen in the example figure 4.8, electrostatic friction increases slightly with increas-
ing pressing force at the contact. This could be due to the area of contact increasing with the
pressing force. At higher pressing force measurements, the greater contact area contributes to
a bigger capacitive coupling between the finger and screen. More capacitance creates a greater
attraction force and a higher measured electrostatic friction.
36
CHAPTER 5
Analyzing Force and Electronics
The friction data have proven to be a very useful method for observing the effectiveness of an
electrostatic tactile display. Through friction analysis, I obtained a measure of the electrostatic
attraction force on the fingertip. Using the same technique in this chapter, I explore how that
force changes with respect to various actuation signals. The model derived in chapter 3 predicts
a square law relationship between force and actuation voltage, as well as a distinct dependence
on actuation frequency. I conducted two experiments, one to investigate actuation voltage and
one to investigate actuation frequency, to determine the validity of the model’s predictions.
37
5.1. Inferred Electrostatic Force Data
To compare the difference between electrostatic force at varying actuation voltages, I used
10 kHz square waves. I tested a total of seven subjects all using five different actuation voltages
at each of five different pressing forces. For each subject, I generated the plot shown in figure
4.8 for every actuation voltage for a total of thirty-five plots. Each plot provided an inferred
electrostatic force. A log-log plot of voltage versus electrostatic force for each subject is shown
in Figure 5.1a.
Subject Data
Fe ∝ V 2
Square Wave, 10 kHz
Inferred
ElectrostaticForce(N
)
Actuation Voltage (VA)
60 80 100 120 140
0.01
0.02
0.03
0.05
0.10
0.20
0.30
(a) Log-log plot of inferred electrostatic normalforce versus supply voltage at 10kHz square wave.Subject data points have average standard errorof 45 mN. Mean exponent of the data 1.92, thedashed line shows an exponent of 2.
Subject Data
RC model prediction
Sine Wave, 140 Volts
Inferred
ElectrostaticForce(N
)
Actuation Frequency (Hz)
102 103 1040
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
(b) Inferred electrostatic normal force versus sup-ply voltage frequency. Prediction does not modelthe relationship properly. Subject data points haveaverage standard error of 37 mN.
Figure 5.1. Electrostatic force results
38
The actuation voltages shown in the plot are not equal to the force generation voltage, VF ,
because of the series capacitance in the cable of the 3M screen. Since all tests were done using
the same actuation signal, the impedance of the series capacitor in the cable does not change,
and the relationship between VA and VF is constant. While all subjects were able to detect
the transient change when electrostatics was switched on and off, the measured electrostatic
friction varied significantly across subjects. All subjects showed increasing force with voltage
relationship, the exponent of which was determined by fitting a line on a log scale. These
exponents are shown in figure 5.1a, and have mean 1.92. The data show very consistent square
law behavior over all seven subjects, and support the force model derived in chapter 3.
To assess the relationship between force and frequency, I used sine waves of 140 volt peak
amplitude. Because of the high-frequency components of square wave signals, sine waves must
be used when studying the frequency dependence of electrostatic force. All seven subjects were
tested under five different frequencies, equally spaced on a log scale from 100 Hz to 10 kHz.
The resulting experimental relationship between actuation frequency and electrostatic force is
shown in figure 5.1b. As the frequency of the signal increases, the force exhibited on the finger
increases as well, but it does not match the shape predicted by the simple dynamic RC model. To
examine this further, I measured the electrical impedance of each subject’s finger as a function
of frequency.
5.2. Fingertip Electric Characterization
The circuit model from chapter 3 predicted a first-order cutoff shown by the dashed line in
figure 5.1b. This prediction was based off the assumption that the stratum corneum acts as
a resistor and capacitor in parallel. I tested the validity of this assumption by measuring the
impedance of the finger part of the circuit. The finger was placed in direct contact with a flat
piece of copper connected to the source of an HP 35665A spectrum analyzer, as shown in 5.2a.
The electrical schematic for this setup is shown in figure 5.2b. The spectrum analyzer measures
impedance vs. frequency by performing a sine sweep input and measuring the voltage across a
10 Ω current sensing resistor, labeled Rsense in the diagrams.
The impedance results on a bode plot are shown in figure 5.3. The dashed line represents the
theoretical impedance based on the model from chapter 3. The subjects displayed a wide variety
of impedance, ranging over two orders of magnitude. These data also show the shortcomings of
the first order RC model. All subject’s fingers again show a dependence on frequency, but at a
significantly more shallow slope than that of a first-order system, very similar to the behavior
observed with the electrostatic force. The phase information is also very important for analysis.
The dashed line shows the theoretical phase predicted by the simple linear model. The subject
data is very noisy, but shows that phase information does not follow that of the prediction. Both
the magnitude and phase plots resemble those of fractional-order models, such as those associated
40
RC ModelSubject Data
Frequency (Hz)
Phase(D
eg)
Impedan
ce(Ω
)
101 102 103 104 105
101 102 103 104 105
−100
−80
−60
−40
−20
0
20
102
103
104
105
106
107
Figure 5.3. Finger impedance data
fractal combinations of first-order systems[24]. It is possible that the stratum corneum exhibits
fractional order behavior due to multi-scale effects, and so this must be considered in future
modeling work. All we can conclude from these impedance data now, though, is that a first-
order assumption for the electronics of the stratum corneum is insufficient.
41
CHAPTER 6
Conclusion
Electrostatic friction shows promise in the field of haptics with the capability of producing
lateral forces as high as 100 mN. This is similar to previously reported ultrasonically actuated
surface haptic displays. The ShiverPaD generated 100 mN of lateral force using low frequency
(∼ 800 Hz) in-plane vibrations[6], and the LateralPaD generated 70 mN of lateral force with
ultrasonic vibrations both in and out of plane[7]. These comparisons indicate that electrostatic
friction has potential applications not only in passive devices, but also in active devices, similar
to those described in [6] and [7].
Past studies[16] have called into question the model described in section 3.1. However, these
efforts operated under the assumption that human detection threshold was constant. My work
eliminates the human factor by directly measuring the interaction force at the surface-finger
interface. The data show a clear proportional relationship between electrostatic force and the
square of actuation voltage, strong evidence supporting equation 3.6. However, the dynamic RC
model from section 3.2 does not correctly model the relationship between electrostatic force and
actuation frequency. This inaccuracy is supported by both electrostatic force data and electrical
measurements of the finger.
6.1. Future Work
During the course of this work, a few issues arose that could be handled by improved exper-
iment design. The tribometer uses a damped balance to control the pressing force. Since the
original objective was simply to keep the pressing force constant, the only way it can be changed
42
is by adding or removing counterweight to the balance. Because of this limitation, all electro-
static test cases are run for one pressing force, then the weight is changed and experiments are
repeated. This means that for each actuation signal test case, the data is taken at five different
times during the experiment, during which environmental conditions may have changed.
After analyzing my data, it’s clear that a friction model (lateral vs. normal force) should
be measured before each test case. A more ideal experimental procedure would be to sample
points on the friction curve without electrostatic actuation, then run a specific actuation signal
test case. This can be done by implementing pressing force control with an actuator and force
feedback. A friction model could be developed with more evenly sampled pressing force cases,
and be more relevant to finding the friction model at the time of a specific test case.
Improving the electrical modeling and measurement of the finger is a key direction for future
research. The first step is to remove the additional skin-electrode interface at the copper band
around the finger. By constructing a screen with two leads of opposite polarity under the finger,
the copper foil for grounding is not needed. The finger effectively closes the circuit between
the two different leads, and the body plays a significantly smaller role. With this approach,
measuring finger impedance becomes much simpler. A new electrical model is, of course, needed
to account for these changes. In addition, since electrostatic force is a direct result of charge,
an amplifier that controls charge rather than voltage would be better suited for actuation. A
current drive amplifier would generate a more consistent electrostatic force across subjects.
43
References
[1] G. Robles-De-La-Torre and V. Hayward, “Force can overcome object geometry in the per-ception of shape through active touch,” 2001.
[2] T. Nara, M. Takasaki, S. Tachi, and T. Higuchi, “An application of SAW to a tactile displayin virtual reality,” in Ultrasonics Symposium, 2000 IEEE, vol. 1, pp. 1 –4 vol.1, Oct. 2000.
[3] T. Watanabe and S. Fukui, “A method for controlling tactile sensation of surface roughnessusing ultrasonic vibration,” in Robotics and Automation, 1995. Proceedings., 1995 IEEEInternational Conference on, vol. 1, pp. 1134 –1139 vol.1, May 1995.
[4] N. Marchuk, J. Colgate, and M. Peshkin, “Friction measurements on a large area TPaD,”in Haptics Symposium, 2010 IEEE, pp. 317 –320, Mar. 2010.
[5] L. Winfield, J. Glassmire, J. E. Colgate, and M. Peshkin, “T-PaD: tactile pattern displaythrough variable friction reduction,” in EuroHaptics Conference, 2007 and Symposium onHaptic Interfaces for Virtual Environment and Teleoperator Systems. World Haptics 2007.Second Joint, pp. 421 –426, Mar. 2007.
[6] E. Chubb, J. Colgate, and M. Peshkin, “ShiverPaD: a glass haptic surface that producesshear force on a bare finger,” IEEE Transactions on Haptics, vol. 3, pp. 189 –198, Sept.2010.
[7] X. Dai, J. Colgate, and M. Peshkin, “LateralPaD: a surface-haptic device that produceslateral forces on a bare finger,” in Haptics Symposium (HAPTICS), 2012 IEEE, pp. 7 –14,Mar. 2012.
[8] J. Mullenbach, D. Johnson, J. Colgate, and M. Peshkin, “ActivePaD surface haptic device,”in 2012 IEEE Haptics Symposium (HAPTICS), pp. 407 –414, Mar. 2012.
[9] O. Bau, I. Poupyrev, A. Israr, and C. Harrison, “TeslaTouch: electrovibration for touchsurfaces,” in Proceedings of the 23nd annual ACM symposium on User interface softwareand technology, UIST ’10, (New York, NY, USA), p. 283–292, ACM, 2010.
[10] J. Linjama and V. Makinen, “E-sense screen: Novel haptic display with capacitive elec-trosensory interface,” in HAID 2009, 4th Workshop for Haptic and Audio Interaction De-sign, (Dresden, Germany), 2009.
44
[11] E. Mallinckrodt, A. L. Hughes, and W. Sleator, “Perception by the skin of electricallyinduced vibrations,” Science, vol. 118, pp. 277–278, Sept. 1953.
[12] S. Grimnes, “Electrovibration, cutaneous sensation of microampere current,” Acta Physio-logica Scandinavica, vol. 118, no. 1, p. 19–25, 1983.
[13] R. Strong and D. Troxel, “An electrotactile display,” Man-Machine Systems, IEEE Trans-actions on, vol. 11, pp. 72 –79, Mar. 1970.
[14] D. Beebe, C. Hymel, K. Kaczmarek, and M. Tyler, “A polyimide-on-silicon electrostaticfingertip tactile display,” in Engineering in Medicine and Biology Society, 1995., IEEE 17thAnnual Conference, vol. 2, pp. 1545 –1546 vol.2, Sept. 1995.
[15] H. Tang and D. Beebe, “A microfabricated electrostatic haptic display for persons withvisual impairments,” Rehabilitation Engineering, IEEE Transactions on, vol. 6, pp. 241–248, Sept. 1998.
[16] A. Agarwal, K. Nammi, K. Kaczmarek, M. Tyler, and D. Beebe, “A hybrid natural/artificialelectrostatic actuator for tactile stimulation,” in Microtechnologies in Medicine amp; Bi-ology 2nd Annual International IEEE-EMB Special Topic Conference on, pp. 341 –345,2002.
[17] K. Kaczmarek, K. Nammi, A. Agarwal, M. Tyler, S. Haase, and D. Beebe, “Polarity effect inelectrovibration for tactile display,” Biomedical Engineering, IEEE Transactions on, vol. 53,pp. 2047 –2054, Oct. 2006.
[18] T. Yamamoto and Y. Yamamoto, “Electrical properties of the epidermal stratum corneum,”Medical and Biological Engineering and Computing, vol. 14, no. 2, pp. 151–158, 1976.
[19] Y. Yuan and R. Verma, “Mechanical properties of stratum corneum studied by nano-indentation,” MRS Online Proceedings Library, vol. 738, pp. null–null, 2002.
[20] K. Shima, Y. Tamura, T. Tsuji, A. Kandori, M. Yokoe, and S. Sakoda, “Estimation ofhuman finger tapping forces based on a fingerpad-stiffness model,” in Annual InternationalConference of the IEEE Engineering in Medicine and Biology Society, 2009. EMBC 2009,pp. 2663 –2667, Sept. 2009.
[21] T. Andre, V. Levesque, V. Hayward, P. Lefevre, and J.-L. Thonnard, “Effect of skin hydra-tion on the dynamics of fingertip gripping contact,” Journal of The Royal Society Interface,vol. 8, pp. 1574–1583, Nov. 2011.
[22] T. Soneda and K. Nakano, “Investigation of vibrotactile sensation of human fingerpads byobservation of contact zones,” Tribology International, vol. 43, pp. 210–217, Jan. 2010.
45
[23] S. Pasumarty, S. Johnson, S. Watson, and M. Adams, “Friction of the human finger pad:Influence of moisture, occlusion and velocity,” Tribology Letters, vol. 44, no. 2, pp. 117–137,2011. 10.1007/s11249-011-9828-0.
[24] H. Schiessel and A. Blumen, “Hierarchical analogues to fractional relaxation equations,”Journal of Physics A: Mathematical and General, vol. 26, pp. 5057–5069, Oct. 1993.
46
APPENDIX A
Dynamic Analysis and Simulation of Touch Screen Circuit
The circuit diagram for the 3M MicroTouch screen and grounded fingertip is given in figure
A.1. The electrostatic force on the finger is static with respect to the voltage, VF , and charge
on the surface, Qf , but these values are dynamic with respect to the actuation voltage signal,
VA. A dynamic analysis is needed to predict an electrostatic force due to an arbitrary actuation
signal. Writing the Laplace transforms of the voltage and current laws gives:
Cs
VF
Ci
Csc RscQf
Rf
VA
Figure A.1. Circuit diagram of the system
VA(s) =Qt(s)
Cs+Qt(s)
Ci+ sQf (s)Rsc + sQt(s)Rf
0 =Qsc(s)
Csc− sQf (s)Rsc
VF (s) =Qt(s)
Ci+Qsc(s)
Csc
Qf (s) = Qt(s)−Qsc(s)
(A.1)
47
Where Qf is the charge leakage through the stratum corneum, Qt is the total charge flowing
from the voltage source, and Qsc is the charge across the stratum corneum. Solving for those
charges yields the output relationships for the free charge Qf and Vf :
VF (s)
VA(s)=
Cs +[CscRsc(Ci + Cs)
]s[
Ci + Cs
]+[CscRsc(Ci + Cs) + CiCs(Rf +Rsc)
]s+
[CiCsCscRfRsc
]s2
Qf (s)
VA(s)=
CiCs[Ci + Cs
]+[CscRsc(Ci + Cs) + CiCs(Rf +Rsc)
]s+
[CiCsCscRfRsc
]s2
(A.2)
To predict the resulting force on the finger, these systems are simulated in Matlab using the
function lsim. The signal Vs(t) is passed as an input to the system, and lsim returns VF (t) and
Qf (t). With these two signals determined, the force as a function of time is:
(A.3) Fe(t) =
C2i C
2sc
((VF (t))2 − (Qf (t))
2
C2i
)2Aε0(Ci + Csc)2
This force is, of course, periodic with the same frequency as the input, VA(t). The average of
this force over one period provides the prediction for the normal force exhibited by the finger.
48
APPENDIX B
Tribometer Control System Design and Electronics
All finger friction experiments were conducted using my custom-built tribometer. The tri-
bometer is capable of electronically controlling finger velocity and haptic surface actuation signal.
A six-axis force torque sensor measures all contact forces, which are converted to analog voltages
by the ATI F/T Controller box. A PC104 stack controls all the peripheral devices to run the
experiment. All resources used for these experiments can be found on the LIMS server.
PC104 (xPC)
Sensoray 526PC
(MATLAB)FunctionGenerator
Piezo Amp Motor Amps
Tribometer
F/T
Controller
Touch Screen
F/T
Sensor
Motors Encoders
RS232TCP/IP
Figure B.1. Tribometer control system schematic
49
B.1. PC104 Stack Control System
The PC104 Stack controls all real-time signals in the system, and records them to its memory
card. The table below shows the connections made to both the PC104 mainboard and the
Sensoray 526 data acquisition board. The RS232 connection controls both the function generator
and ATI controller. The force signals are carried by a shielded 4-wire cable to a D-Sub connector
on the back of the ATI controller. The colors of the signals are also reported in the table. A
ground connection between the function generator board and PC104 stack must also be made.
Table B.1. Connections to the PC104 stack
MainboardRS232-1 Function GeneratorTCP/IP Local Router
Sensoray 526AnalogIn-6 Z Force (Yellow)AnalogIn-7 Y Force (Red)AnalogIn-8 X Force (Green)Analog Ground Analog Ground (Gray)DigitalOut-1 Function Generator AnSwitchAnalogOut-1 Function Generator AmpCtrlAnalogOut-2 Finger Motor AmpAnalogOut-3 Screen Stop Motor Amp
The program for the xPC machine is written in the Simulink graphical language. The model
file is called FrictionMeasure.mdl, and can be run directly from Simulink by first connecting
to target and then executing real-time code. A graphical interface was written to simplify the
experimentation process, called FrictionMeasureGUI. Running this script automatically loads
the real-time code onto the xPC stack loads a control panel for running experiments. A screen
shot of the panel is shown in figure B.2.
Before starting the code, the screen should be in the level position, ready for calibration.
The large button in the upper left should be pressed first to start the real-time code. The three
F/T Sensor buttons reset the sensor, set the analog voltage ranges, and calibrate (set the zero)
50
Figure B.2. Graphical control interface for experiments
the sensor respectively. Lowering and raising the screen using the screen stop motor can be
done manually by pressing the respective buttons. The set wave and pulse ES selections are
used to generate a pulsing effect outside of an experiment. This is useful for demonstrating a
certain waveform and voltage before running an experiment. The waveform that will be tried is
determined by the appropriate voltage, frequency, and type selected in the experiment panel.
All experiments are run from the experiment panel. The subject name should read “TSXX”,
where XX is the subject number. The number of trials refers to the number of swipes recorded
for each experimental condition. Temperature and humidity are recorded as they are typed.
The three boxes below take either scalars or vectors, and all combinations of entered values will
51
be tested. For example, if the desired experiment is 140 volt sine waves of various frequencies
at 50 mm/s, then the boxes would read 50 mm/s, 140 V, and [100 1000 10000] Hz respectively.
B.2. Motor Amplifiers
The two motor amplifiers are linear current amplifiers. Each amplifier contains two Darling-
ton pair transistors, one to control the current from the positive voltage rail, and one for the
negative. They are powered each by two floating DC power supplies, the negative terminal of
one connected to the positive terminal of the other to create positive and negative rails. The
common rail is connected to ground. An op-amp is used to control for current through the
sensing resistor, Rsense. A schematic and image of one of these amplifiers is provided below.
V-
V+
Rbe
330Cb
0.1uF
−
+
OPA177
Cshunt
1uF
Cshunt
1uF
Rin
10kVin
22
RmotorCmotor
.22uF
0.33
Rsense
Rifdbk
1k
Figure B.3. Linear motor amplifier schematic
52
Figure B.4. Linear motor amplifier
B.3. RS232 Function Generator
The function generator board is run by a PIC microcontroller that listens to RS232 com-
mands from the PC104 stack. Depending on the content of the message, it either passes it along
to the ATI force sensor controller, or parses it to set waveform properties. The waveform is
created locally on the board using 4 integrated circuits. Using the Analog Devices AD9833 SPI
function generator chip, arbitrary sine, square, or triangle waves can be created with a command
from the PIC. Because the voltage out from the chip is unipolar and different amplitude for each
type of wave, a controllable amplitude correction circuit shown in Figure B.5 was developed. The
pairing of the Maxim MAX518 dual digital to analog converter(DAC) and the Analog Devices
AD633 linear multiplier is used for amplitude correction. The signal is first sent through a 10x
53
PIC32
SPI
I2C
AD9833
FuncOut
MAX518A0
A1AD633
(Vin)(A0)10 +A1
Vin
1kΩ−
+
10kΩ
+15V
-15V
Figure B.5. Function generator amplitude correction schematic
inverting amplifier before feeding into the multiplier chip. The multiplier chip then uses analog
values set by the DAC to properly adjust the signal levels. The values used are shown below
in table B.2. The output of the final AD633 chip is fed though another multiplier chip. The
Table B.2. Function generator amplitude correction settings (All values are voltages)