MEMS WING TECHNOLOGY FOR

A BATTERY-POWERED ORNITHOPTER
T.Nick Pornsin-sirirak, S.W. Lee, H. Nassef*, J. Grasmeyer**, Y.C. Tai, C.M. Ho*, M. Keennon**

Caltech Micromachining Laboratory, 136-93, Pasadena, CA 91125

* Mechanical and Aerospace Engineering, University of California, Los Angeles, CA 90095
** MAV Program, AeroVironment Inc., Simi Valley, CA 93063

ABSTRACT
2
10
The objective of this project is to develop a battery-
powered ornithopter (flapping-wing) Micro Aerial
Vehicle (MAV) with MEMS wings. In ht is paper, we MAV
1

Wing length (cm)

present a novel MEMS-based wing technology that we
10
developed using titanium-alloy metal as wingframe and Sphingids Passerifor
parylene C as wing membrane. MEMS technology m
enables systematic research in terms of repeatablility, Hovering Diptera
size control, and weight minimization. We constructed 1
a high quality low-speed wind tunnel with velocity Hummingbirds
uniformity of 0.5% and speeds from 1 m/s to 10 m/s.
We have tested and have studied the unsteady -state
-1
aerodynamics of various types of MEMS wings. 10 -3 -2 -1 1 2 3
Finally, we built lightweight ornithopters with electric- 10 10 10 1 10 10 10
powered transmission system and have demonstrated Weight (g)
successful free flights of with flight duration ranges
from 5 to 18 seconds. Figure 1: Size of natural flyers
1
INTRODUCTION
U = 4. 77 m 6 (1)
We started this project with two difficult constraints.
where U is the flight speed in m/s and m is the mass in
The first constraint was that the flyer must be a MAV grams. From statistical data on wing beat rate vs. wing
and, by definition, MAV must have a total wingspan
length [4], and wing beat rate vs. mass [6] for birds and
less than 15 cm. The second constraint was that the flyer
insects, a relationship of wingtip speed, Uvertical, and
must fly by flapping wings (an ornithopter). mass can be derived and are given by these following
Unfortunately, aerodynamics of flapping-wing flight,
relations:
especially MAV size, is still not a fully-explored
subject. There have been studies of insect flights [1-3];
however, unlike fixed-wing aerodynamics, there have U vertical upper = 11. 7m − 0.065 (2)
not been any available design rules for flapping-wing −0 .043
U vertical lower = 9.6 m (3)
aerodynamics for MAV size. As a result, we believe
that there are two approaches for this project. One is to
learn from natural MAV flyers and try to mimic them.
The other is to study flapping-wing aerodynamics
ourselves and try to improve them.
Speed U (m/s)

From our analysis of natural MAV flyers, we find that

the MAV size falls within the range of small birds, bats,
hummingbirds, and large insects [4]. Figure 1 shows
some samples of natural flyers. We estimate that these
flyers of MAV size weigh about 7-10 grams and we
believe our ornithopter should weigh about the same.
Shown in Figure 2 is statistical data on the speed vs.
size relationship from a wide range of birds [5]. The
general statistical tendency shows that the flight speed Mass m (g)
can be approximately given by: Figure 2: Flight speed of birds [5]
 22.14
100

Unsteady -state regime Quasi -steady -state regime

Speed (m/s) (Wing -tip speed > Wind speed) (Wind speed > Wing-tip speed)

Wingtip speed
vpar ( m )

vvertu48 ( m )
vvertl48 ( m ) 10
Flight speed
MAV regime

2.21404 1 3 4
0.01 0.1 1 10 100 10 10
m 9999.01
Body mass (g)

Figure 3: Flight regime of steady-state and unsteady-state of natural flyers

Combining equations (1) to (3), a plot of wingtip speed dragonfly wings. We developed a MEMS fabrication
and flight speed vs. mass of insects and birds can be process using silicon and parylene C to make
obtained as shown in Figure 3. The flight of flyers can wingframe and membrane, respectively. These
be separated into two regimes: quasi-steady and fabricated wings are shown in Figure 4. The bone width
unsteady states. For larger flyers, their flights can be of the bat wing is 350 µm and the membrane thickness
approximated by quasi-steady -state assumptions for both wings is 15 µm. However, silicon wingframes
because their wings flap at lower frequency during were too fragile. They broke easily. Therefore, we
cruising. This means the wingtip speed is low have developed a new process using titanium-alloy
compared to the flight speed. Thus larger birds, such as metal as wingframe instead.
eagles and seagulls, tend to have a soaring flight. Their
wings behave closely to fixed-wings. On the other
hand, smaller birds and insects fly in an unsteady-state
regime in which their wingtip speed is faster than their
flight speed. From Figure 3, we conclude that our
MAV ornithopter (mass 7-10 grams) operates in an
unsteady-state flow regime. The airflow over the wings 1 cm
is not constant in time and cannot be approximated by
quasi-steady -state assumptions. During the unsteady-
state flight, the airflow is separated from the wing at the a) Silicon bat wing b) Silicon dragonfly wing
leading edge and a separation “bubble” is formed during
Figure 4: MEMS fabricated silicon wings
downstroke to generate a high lift during flight [7]. The
vortex bubble is formed as the stagnation streamline Titanium-alloy Metal Wings
rolls over the leading edge. This bubble continues to
grow during the downstroke and is shed at the start of We have experimented with various materials for
the upstroke. wingframe structure. For example, aluminum metal is
light in weight but it is too soft. Stainless steel is strong
Thus, one of the most difficult and challenging tasks is but its density is twice as high as that of titanium-alloy
to design and develop a highly efficient wing that has an metal. The etchant solution (FeCl 3) is dark brown in
unsteady-state aerodynamic advantage. The wing must color and is impossible to judge when the etching is
be light and strong. In addition, it also has to be able to finished. Besides, the etching process of stainless steel
withstand high flapping frequency without breaking and must be performed at a high temperature in order to
is capable of generating enough lift and thrust to fly the yield a reasonable etching rate.
prototype vehicles.
We have chosen titanium-alloy metal for several
DESIGN AND FABRICATION reasons. First, it is light and strong. Second, it can be
easily tapered to vary the thickness of wingspars. Third,
Since our first approach is to learn from natural flyers since titanium-alloy is ductile, it can be bent to create
and mimic them, we first designed our wings based on wing camber to improve performance. Finally, the
bat wings and other insect wings, such as beetle and etching process of titanium-alloy can be conducted at
room temperature with reasonable etching rate. For
wing membrane, parylene C is selected because it can fabricated using photolithography technology. Figure 6
be deposited directly onto titanium-alloy at any desired shows the fabrication process of titanium-alloy MEMS
thickness. Its adhesion to titanium-alloy is excellent. wing.
Moreover, it is light, strong, and can withstand high
flapping frequency of more than 30 Hz without tearing. A’ Ti-alloy
Lastly, parylene C is deposited at room temperature and
yields a conformal coating. Thus step corners are Resis
uniformly covered. Figure 5 shows various fabricated tParylene
titanium-alloy MEMS wings, ranging from insect wings A
to simple spar wings. Table 1 shows selected
mechanical properties of both titanium-alloy metal and
parylene C. 1) Laminate resist 4) Strip and relaminate
resist; deposit parylene

2) Pattern resist 5) Strip resist

a) simple d) beetle

3) Etch Ti-alloy 6) Deposit backside

b) dragonfly parylene
e) bat Figure 6: Fabrication process of titanium-
alloy MEMS wings
7 cm
First, a 250-µm-thick titanium-alloy substrate is
cleaned in trichloro ethylene (TCE) for 20 minutes.
Later, it is dipped in a diluted HF solution to roughen
c) butterfly f) CIT7x3S20 the surface. Dry film resist is then laminated on both
sides. The resist is patterned and hardbaked at 120 °C
Figure 5: Titanium-alloy M EMS wings for 20 minutes. The substrate is then etched in a
solution of HF:HNO3:H2O = 5:2:100 in volume. We
Table 1: Mechanical Properties of Ti-6Al4V [8] and find that if the concentration is stronger, the resist will
Parylene C [9] peel off before the etching is finished. At this
Properties Ti-6Al4V Parylene C concentration, the etching rate is about 2.5 µm/min.
3 Since this is an isotropic etching, the undercut rate is
Density, g/cm 4.5 1.3
Young’s Modulus, GPa 110 3 about the same. Therefore, undercut must be taken into
Tensile Strength, MPa 100 70 a consideration during the mask design.
Yield Strength, MPa 97 56
After the etching process is finished and the wingframe
Coefficient of Thermal 0.9 3.5
is formed, the resist is stripped from both sides in a
Expansion, x10-5/°C diluted KOH solution. Dry film resist is relaminated on
the backside as a platform for parylene C polymers to
Fabrication process deposit on. Parylence C is then deposited. Afterwards,
dry film resist is stripped. Finally, in order to
We have tried to use a conventional method, i.e. using
strengthen wing membrane, the second parylene C
carbon fiber rod, mylar, and glue, to make wings.
deposition is performed.
However, we found that making wings that way is
cumbersome. Glue also adds significant weight. We find that having the right material as a mask during
Moreover, identical wings are difficult to achieve unless etching is important. A crucial fabrication hurdle is
a mold is made. This is costly, time-consuming, and releasing a large area of wing without damaging the
slow. This method cannot accommodate effectively the membrane itself. The photoresist chosen must have an
study of the design variable changes. ability to withstand HF and HNO3 acids and can be
stripped off without destroying the titanium-alloy metal
For many reasons we claim the new MEMS wing and the parylene membrane. We selected negative dry
technology is necessary because MEMS wings enable film resist and found that its adhesion to titanium-alloy
systematic research in term of repeatability, size control,
substrate was good. It could be stripped off easily in a
weight minimization, mass production, and fast turn- diluted KOH solution if the film was exposed under the
around time. Moreover, complicated structures, such as
UV light beforehand. Both titanium-alloy wingframe
dragonfly, butterfly, and beetle wings, as shown in
and parylene membrane were also not attacked at all.
Figure 5 b), c) and d), respectively, can be easily
 TESTING AND RESULTS x x

Mechanical Testing ll Wing Spar

The wing stiffness test setup is shown in Figure 7. The a

wing is clamped at its root. A blade, connected to a
loadcell and an XYZ stage, is used to probe at various
section of the wings. A plot of spring vs. normalized
distance from wing’s root constant of a butterfly wi ng Figure 9: Transmission system
is shown in Figure 8. We see that MEMS wings can be The wind tunnel has a 30x30x60 cm test section with a
tapered to vary thickness according to natural wing 4:1 contraction. Force measurements were taken using
stiffness distribution. Tapered wings’ stiffness, when low capacity 2D force loadcells. The aerodynamic
normalized with weight, is also comparable to that of performance of natural insect wings, carbon fiber wings,
the natural wing. and MEMS wings has been studied. As shown in
Figure 10, wind tunnel test results show that spanwise
Clamp stiffness is an important factor in lift production in
Wing Blade
wing
flapping flight. For the same size of wings, cicada
wings with rigid leading edge produce larger lift
coefficients compared to our previous design of metal
Loadcell bat wings whose leading edge is flexible.
To XYZ stages

Spanwise Flexible
To force gauge
0.8
Figure 7: Spring constant test setup schematic Quasi -Steady
0.6
Lift coefficient, CL

0.4
0.2
1000 0
-0.2
-0.4
750 -0.6
-0.8 Unsteady
Spring constant

-1
500 Real butterfly 0.1 1 2 3 4 5 6 7 8 910
Metal butterfly
Tapered metal Advance Ratio, J = U/(2Φfb)
250
(N/m)

butterfly a) Titanium-alloy bat wings

0
Spanwise Rigid
Distance L (%)
1.8
Figure 8: Spring constant of butterfly wings 1.6
Unsteady
Lift coefficient, CL

1.4
Transmission Design 1.2
1

We built a lightweight, low-friction transmission 0.8 Quasi -Steady

0.6
mechanism to convert the rotary motion of the driving 0.4
motor into the flapping motion of the wings based on 0.2
0
simplicity, minimal weight, and flapping symmetry. 0 0.5 1 1.5 2 2.5 3 3.5 4
This design restrict the flapping motion in a plane Advance Ratio, J = U/(2Φfb)
perpendicular to the motor shaft and is shown in Figure
b) Cicada wings
9. A small DC motor with gearbox ratio of 22:1 was
used to drive the transmission. MEMS wings were Figure 10: Spanwise stiffness effect
mounted on the transmission system and several
flapping tests were performed. The wings could The lift and thrust coefficients can be expressed as
withstand more than 30 Hz of flapping without follows:
breaking. There were also no tears on the membrane.
2L
Wind Tunnel Test Results CL = (4)
ρAU 2
The MAV aerodynamic study was conducted at UCLA 2T
and CT = (5)
in a high quality low-speed wind tunnel with velocity ρ AU 2
uniformity of 0.5% and speeds from 1 m/s to 10 m/s.
where L, T, U, A, ρ are lift, thrust, flight speed, wing 1.4
A B C
planform area, and air density, respectively. The 1.2

Input Power, P [W]

D
advance ratio J is the ratio of the flight speed to the
1
speed of the wingtip and is given by:
0.8
U
J= (6) 0.6
2Φfb
0.4
where Φ , f, b are stroke angle, flapping frequency, and
wing semi-span, respectively. Typically, unsteady-state 0.2
flight has an advance ratio of less than 1. For example,
0
natural fliers such as bumblebee, black fly, and fruit fly 0 5 10 15 20 25 30 35
have an advance ratio in free flight of 0.66, 0.50, and
Flapping Frequency, f [Hz]
0.33, respectively [10].
Figure 12: Input power
Wind tunnel test results shows that nature-mimic wings
with complicated structure performed poorly compared
PROTOTYPE VEHICLES
to wings with simple designs listed in Table 2. Thus,
Lift and thrust coefficients for these simple-designed
wings are shown in Figure 11. Figure 12 shows the Super Capacitor-Powered Ornithopter
input power required to flap these wings. Our current
MEMS wing design D (CIT7x3S20) with rigid leading- We built a super capacitor-powered electric motor free-
edge shows the best result in terms of lift, thrust, and flight ornithopter, shown in Figure 13, which weighs
power required. It only needs 1 watt to flap at 30 Hz. only 6.5 grams. The system is composed of an electric
motor, a transmission system, two 1-farad super
Table 2: Properties of Various Wing Designs capacitors, MEMS wings, a carbon-fiber-rod fuselage,
Wing types A B C D and tail stabilizers. On the bench test, the flapping
duration was less than a minute before having to
Weight (each), mg 220 220 170 170
recharge the capacitors. This is much shorter compared
Frame material C C Ti Ti
to the NiCd battery’s dischargeme.
ti
Membrane material myl pap par par
Angle of diag. spa, deg 45 n/a 10 20
Planform LxW, cm 7x5 7x3 7x3 7x3
C = carbon fiber; Ti = Titanium-alloy;
myl = mylar; pap = paper; par = parylene C; L =
spanwise; W =chordwise

3
B&D
Lift Coefficient, CL

2.5 C
2 A: Carbon rod
A Figure 13: Super capacitor-powered ornithopter
1.5 + mylar
1 Battery-Powered Ornithopter

0.5 Since our goal is to use a battery to provide a longer

0 B: Carbon rod power source, we built a battery-powered ornithopter
0 0.5 1 1.5 2 2.5 3 + paper MAV as shown in Figure 14. We redesigned the body
Advance Ratio, J = U/(2Φfb) and replace super capacitors with a battery and a dc-to-
dc converter. The mass summary is shown in table 3.
Thrust Coefficient, CT

0.45
0.4
0.35
0.3 C: CIT7x3S10
0.25 D
0.2
0.15 A B&C
0.1
0.05
0 D: CIT7x3S20
-0.05 0.5 1 1.5 2 2.5 3
Advance Ratio, J = U/(2Φfb)

Figure 11: Lift and thrust coefficients of Figure 14: Battery-powered ornithopter MAV
various types of wings
Table 3: Mass Summary for Battery -Powered CONCLUSION
Ornithopter
Components Weight A novel titanium-alloy wingframe technology has been
(g) developed for MEMS wings. Several MEMS wings
were fabricated with parameters, such as chord and spar
MEMS Wings 0.4
widths, membrane thickness, number of spars, and
Motor & Transmission system 3.1
sweep angle, varied. We believe that only MEMS
Battery 3.0
technology can easily and systematically accommodate
DC-to-DC Converter 1.9
these many variable changes with a fast turn-around
Fuselage, tail, switch, wires, etc. 2.2 time. Wind tunnel tests were performed in the high
Total weight 10.6 quality wind tunnel at UCLA. Wings have been tested
under cyclic conditions to assess long-term reliability.
The lightest rechargeable battery available we found is Super capacitor-powered and battery-powered
Sanyo NiCd N-50. It weighs about 3.5 grams. We ornithopters were built. The best free flight duration of
trimmed the casing as thin as possible to reduce the 9 and 18 seconds were achieved by super capacitor-
weight to 3 grams. Since the NiCd battery produces powered and battery-powered ornithopters, respectively.
only 1 volt nominally and the drive motor requires 4 to
6 volts, a DC-to-DC converter is custom-built to step up ACKNOWLEDGEMENTS
the voltage to the necessary 4 to 6 volt level to operate
the electric motor. It weighs only 1.9 grams. The This work is supported under DARPA/TTO MAV
voltage output is adjustable and can be set before each program DABT63-98-C-0005.
flight test. The advantage of the converter and a NiCd
cell power system is that it is light weight and takes REFERENCES
advantage of the good specific power and specific [1] M. Okamoto, K. Yasuda, A. Azuma, “
energy of the 50 mA-hr NiCd cell. We cannot use a Aerodynamic Characteristics of the Wings and
higher quantity of smaller batteries to deliver the same Body of a Dragonfly”, Journal of Experimental
power and performance due to the weight constraint. Biology, vol 199, pp. 281-294 (1996)

. [2] M. Sato, A. Azuma, “ The Flight Performance of a

Damselfly Ceriagrion Melanuram Selys”, Journal
of Experimental Biology, vol 200, pp. 1765-1779
(1997)
[3] M. F. M. Osborne, “ Aerodynamic of Flapping
Flight with Application to Insects”, Journal of
Experimental Biology, vol 28, pp. 221-245 (1951)
[4] C.H. Greenewalt, “ Dimensional Relationships for
Flying Animals”, Smithsonian miscellaneous
collections, vol 144, no. 2, pp. 1-46 (1962)
Figure 15: Recent flight test
[5] C.J. Pennycuick, “The mechanics of bird
Flight Test
migration”, IBIS, vol. 3, pp. 525
-556 (1969).
A recent flight test is shown in Figure 15. We have [6] A. Azuma, “The Biokinetics of Flying an d
tested metal wings and carbon fiber wings with both Swimming”, Springer
-Verlag, Tokyo (1992)
ornithopters. Flight duration of 5 to 18 seconds were
[7] G.R. Spedding, "Advances in Comparative and
achieved. So far, the best flight duration for super Environmental Physiology”, vol 11, pp. 51 -111
capacitor-powered ornithopter was 9 seconds and 18 (1992)
seconds for battery-powered ornithopter. The duration
is mainly limited by the power system and vehicle’s [8] American Society for Metals, “Metal Handbook”,
weight. In both cases, the metal wings did not break. 9th edition, vol 3, pp. 388-391 (1980)
[9] Product Specifications, “ Parylene Conformal
We find that there are several challenges in order to
Coatings Specifications and Properties”, Specialty
achieve a successful sustained flight. First, the wind
Coating Systems, Inc., Indianapolis (1994)
condition must be perfect. Often during flight test, the
nd
wind speed and direction shifted constantly. Second, [10] S. Vogel, “Life in Moving Fluids”, 2 edition.,
the trim of tail stabilizer must be crucial. Finally, each Princeton University Press, Princeton (1994)
launch motion must be the same. We also believe that
our current wings and ornithopter are not optimized,
thus we hope future flight duration can be improved.