Design and fabrication of MEMS-based thermally-actuated image stabilizer for cell phone camera

(1)

Design and fabrication of MEMS-based thermally-actuated image stabilizer

for cell phone camera

Chun-Ying Lin

a

_{, Jin-Chern Chiou}

a,b,⇑ a

Department of Electrical and Control Engineering, National Chiao Tung University, Hsin-Chu, Taiwan, ROC

b

School of Medicine, China Medical University, Taichung, Taiwan, ROC

a r t i c l e

i n f o

Article history:

Available online 29 June 2012 Keywords: Component XY stage Image stabilizer MEMS Thermal actuator

a b s t r a c t

A micro–electro-mechanical system (MEMS)-based image stabilizer is proposed to counteracting shaking in cell phone cameras. The proposed stabilizer (dimensions, 8.8 8.8 0.2 mm3_{) includes a two-axis}

decoupling XY stage and has sufﬁcient strength to suspend an image sensor (IS) used for anti-shaking function. The XY stage is designed to send electrical signals from the suspended IS by using eight signal springs and 24 signal outputs. The maximum actuating distance of the stage is larger than 25lm, which is sufﬁcient to resolve the shaking problem. Accordingly, the applied voltage for the 25lm moving dis-tance is lower than 20 V; the dynamic resonant frequency of the actuating device is 4485 Hz, and the ris-ing time is 21 ms.

1. Introduction

Advances in device miniaturization and high precision systems now enable micro fabrication of XY stages[1]. One of the most common actuators used to drive MEMS-based XY stages is the elec-trostatic actuator, which is often used in optical applications[2–7]

and probe position systems[8]because it has good controllability and low power consumption; applications of the electrostatic XY stage include SPM (Scanning Probe Microscopy) probe [9,10,12], data storage[11–15], and nano-position control[1,16–18]. How-ever, since electrostatic actuators often require a high driving volt-age, the step-up circuit must be very large. To decrease driving voltage and to increase actuating force, thermal actuators are used to increase actuating displacement[25].

The camera function has a growing role in cell phone applica-tions. In addition to meeting the consumer demand for increased pixels, digital single lens reﬂex (DSLR) camera functions are now being embedded in commercial cell phones. Anti-shake technology is among the many new technologies developed for mobile phones. Since increasing the number of camera pixels exacerbates the problem of image blurring caused by hand shaking, an effective image stabilization method is needed[19]. The most common im-age stabilization methods include lens shifting[20], CCD shifting

[21], and signal processing [22]. Signal processing, which is the anti-shaking technique conventionally used in DSLRs, is now used in mobile phones. Although signal processing requires no

addi-tional hardware and does not affect miniaturization of the system module, performance and reliability depend on the algorithm used, thus signal processing method usually require large memory. Gi-ven the demand for device miniaturization, the lens shifting anti-shaking approach is inadequate since adding a movable lens causes nonlinearity that requires compensation with a complex control algorithm. Although it requires an actuating system, IS shifting is less disruptive of miniaturization; the overall size of the system is also easier to reduce compared to lens shifting method[19]. The MEMS device enables precise movement of suspended micro-structures and monolithic integration with microelectronic circuits on a chip[23]. Therefore, an effective image stabilizer design is needed to miniaturize cell phone camera modules.

The MEMS-based two dimensional decoupling thermal actuat-ing XY stage proposed in this study has several advantages: 1. Small chip size: since the presented XY stage is fabricated by MEMS process, the dimensions are smaller than those of other commer-cial anti-shaking devices. 2. Low driving voltage: the thermal actu-ating mechanism requires a lower driving voltage and smaller step-up circuit compared to conventional actuators. 3. Simple con-trol method: the CCD shifting approach directly compensates for blurred pixels, so complex control algorithms and optical path de-signs are not needed.

The image stabilizer design is suitable for a cell phone equipped with a 3 optical zoom, an actuator that can move at least 25

l

m

[19], and a sufﬁciently strong structure to withstand a 3.5 3.5 0.1 mm3IS load, from which it must be decoupled in two dimensions when driven. Therefore, the proposed design has a high aspect ratio and 50

l

m-thick silicon springs. Four decou-pling springs for the main suspended and decoudecou-pling structures

http://dx.doi.org/10.1016/j.sse.2012.05.013

⇑Corresponding author at: School of Medicine, China Medical University, Taichung, Taiwan, ROC

E-mail address:[email protected](J.-C. Chiou).

Contents lists available atSciVerse ScienceDirect

Solid-State Electronics

(2)

connect with the shuttle. The fabrication process used to manufac-ture the device includes double-side lithography, two ICP etching processes, and ﬂip–chip bonding process. The output routing for electrical connection of the IS includes 24 signal outputs from eight signal springs. The required 25

l

m displacement is achieved when driven in a moving direction lower than 20 DC voltages, and the decoupling effect in the opposite direction is excellent.

2. Concept and structure design

The novel feature of the proposed MEMS-based image stabilizer is a 2D decoupling actuator that can carry an IS to compensate for blurring caused by shaking. The design of the XY stage is described below.

2.1. Design of XY stage

The proposed image stabilizer is designed to provide excellent decoupling performance and to minimize the size of the device used for suspending the IS. The anti-shaking function is provided by a decoupling structure that compensates for shaking with no additional moving deviation in another direction.Fig. 1a shows a simple schematic diagram of the proposed decoupling main struc-ture. When a single direction force Fxis applied to the decoupling

structure, displacementDx occurs only in the direction parallel to the driving force, and displacement in the opposite direction is negligible.Fig. 1b illustrates a simpliﬁed operating mechanism of the proposed decoupling XY stage when different forces are applied in the x- and y-directions. The decoupling design assumes perfect displacement in the x- and y-directions (Dx andDy, respectively) and assumes perfect mechanical isolation between the two orthog-onal driving directions.

The proposed for suspending the IS has a thickness of 50

l

m to provide sufﬁcient stiffness in the z-direction. Circuit routing of sig-nal outputs must be patterned on the device layer. The sigsig-nal springs and pads must be designed to maintain a working electrical connection. Flip–chip bonding, wire bonding, routing springs, and electrode pads are used to obtain an electrical connection between cell phone circuits in the IS pads and the electrode pads in the de-vice structure and also between the electrode pads in the dede-vice structure and the output of the PCB pads.Fig. 2schematically

de-picts the entire device. Separate decoupling beams are used in each direction to satisfy the kinematic design requirements shown in

Fig. 1. Each thermal-driven actuator has a push-only design, and each operates in only one direction, i.e., X+, X, Y+, or Y. In the symmetrical spring design, the electrical routings are arranged in a balanced layout with two signal output springs and pads in each direction. The signal output springs are designed to be as thin as possible so that stiffness does not affect the overall spring constant system. The isolation layers and the isolator SU-8 ensure that the electric potential between each structure and signal output is independent.

2.2. Operating principles of electro-thermal chevron actuator The proposed image stabilizer is driven by electro-thermal mechanical force.Fig. 3shows a simple schematic diagram of the electro-thermal chevron actuator of the proposed image stabilizer. An electro-thermal mechanical force generated by applying a volt-age to the pads changes the relative displacement between the XY stage and pads.

When estimating driving displacement, several boundary con-ditions are assumed: 1. The actuator is heated uniformly, 2. Pad temperatures are fixed at 300 K, and 3. The expansion coefficient is a constant equal to Si expansion coefficient, 2.5 ppm/K. Under these conditions, the driving displacement is linearly related to temperature. The equation for predicting driving displacement is given below[26]:

EI@

2_y

@x2¼ M ¼ MA Fy Fox=2 ð1Þ where boundary conditions are

yj_x¼0¼ 0;dy dx _x¼0¼dydx _x¼L=2¼ tan hA;

r

¼

D

TE

a

¼E L

D

L þ FL EWT ð2Þ where EI is the ﬂexural rigidity, F is the reaction force along x-axis used to model the effects of thermal stress, and Fois the output

force. The L, W and T inFig. 3are the length, width and thickness of the actuator, respectively. The DT is the temperature change, DL is the change in L due to F, and

a

is the thermal expansion

Fig. 1. Illustration of driving modes of decoupling XY stage: (a) in X-direction only and (b) in both X- and Y-directions.

(3)

coefficient. Peak deflection is obtained by setting output force to zero, and peak output force is obtained by setting deflection to zero. The equation used to derive the relationship between the lateral force for peak deflection and peak output force is

Fomax¼ 2Fmaxtan hA ð3Þ

The equation also gives peak output force according to peak deﬂec-tion. The output force and displacement can be calculated by mod-ifying the actuator beam dimensions and gradients.

2.3. Designing and estimating spring stiffness

Fig. 4shows the three ﬂexure beams used in this work.Fig. 4a is a chevron thermal actuator. The small pre-bending angle can be simpliﬁed as parallel clamped–clamped beams[10,12]. According to Ref. [27], the decoupling ratio for the decoupling beam in

Fig. 4b must be larger than 10 when the pixel size and the maxi-mum actuating displacement are 2.5

l

m and 25

l

m, respectively.

Fig. 2. Structure of XY stage, (a) Signal output beam (b) thermal actuator and SU8 isolator (c) electrode pads (d) decouple beam.

Fig. 3. Force acting on the chevron thermal actuator.

Fig. 4. Illustrations of (a) chevron thermal actuator beams, (b) decoupling beams and (c) folded ﬂexure beams.

(4)

Fig. 4c is an IS signal output beam.Fig. 5shows the spring design of the structure[11,13], and the ﬂexure beam speciﬁcations ofFig. 4

in this work are shown inTable 1.

2.4. Motion range analysis of decoupling XY stage

Most commercially available cell phones have multi-megapixel cameras but lack an anti-shaking function. Because shaking is dif-ﬁcult to avoid when using a cell phone camera, blurring is com-mon. Fig. 6 displays the relationship between the number of blurred pixels and the angle of hand shaking. When the camera moves through an angle dh in the horizontal plane, the number of blurred pixels can be represented as

BP ¼dh hL

PH ð4Þ

where hL denotes the horizontal angle and PH denotes the total

number of horizontal pixels. Cell phone cameras typically have an angular ﬁeld of view ranging from 35° to 45°. For an IS with a 45° ﬁeld of view, experimental data[24]show that, within a 2 s period, hand shaking can cause horizontal drift of up to 0.08°. The device proposed in this work contains a three-megapixel (2056 1544 pixels) IS. According to Eq.(4), each pixel corresponds to a horizontal angle of 0.022°. Therefore, the number of blurred horizontal pixels is approximately 3.7. Since pixel size is 2.2 2.2

l

m2_{, the blurred image can only be adjusted if the range}

of motion in the XY stage is larger than 8.14

l

m. Since many com-mercially available cameras have a 3 zoom function, the range of motion must be threefold higher than this value, i.e., 24.42

l

m.

3. System modeling and FEA simulation

The decoupling effect, stiffness and the natural frequency of the designed image stabilizer is elucidated in a simulated actuator con-structed using Intellisuite ﬁnite elements analysis (FEA) software. The software is used for static simulations of in-plane distance and the coupled mechanical interference of the image stabilizer in each orthogonal direction. In dynamic simulation, this software simulates the resonant frequency.

3.1. Simulated decoupling of main structure

This section describes the design and simulation of the decou-pling structure, and Fig. 7 shows the simulation results for the decoupling effect. Notably, non-decoupling relative structures are excluded. Four major decoupling springs are designed to connect with the chevron thermal actuator. The simulated decoupling ef-fect of the main structure is based on the following actual bound-ary conditions: 1. The anchor is ignored since it is bonded with a substrate and does not affect the simulation results. 2. The shuttle, decoupling and flexure beams are suspended above the substrate. If the moldings of the designed moving spring and a frame-loaded spring are the same size as the image stabilizer, a significant decoupling effect occurs in the x and y directions (Fig. 7a and b, respectively). In the simulation, a specific force Fyin the y-direction

causes a 31.27

l

m displacement in the y-direction. Simulation re-sults indicate that the displacement in the x-direction is only 0.33

l

m. The simulated decoupling ratio in the x and y directions conﬁrms the excellent decoupling effect in the design of the mova-ble springs, which clearly meets the decoupling requirement. 3.2. Simulated structural stiffness and actuator force output

The stiffness and force output of thermal actuators in the XY stage are simulated in three steps. First, establish a reference beam where the width, length, thickness and folding number are 20

l

m, 320

l

m, 50

l

m and 8

l

m, respectively, and apply a force of 1000

l

N to the reference beam Then use Hooke law to calculate dis-placement and stiffness.Fig. 8a shows the simulation results. Sec-ond, apply 1000

l

N to the actuator, and observe the displacement to determine actuator stiffness for different series of numbers

Table 1

Flexure beam speciﬁcations. Spring

type

Fig. 4a Fig. 4b Fig. 4c

Length La1:630lm La2:300lm L:2000lm La1:900lm La2:790lm Width 30lm 20lm 25lm Height 50lm 50lm 50lm

Fig. 6. Relationship between blurred pixels and hand shaking angle.

Fig. 7. Schematic of decoupling simulation in (a) the Y-direction and (b) the X-direction when the main structure of image stabilizer is driven with an outside force in the Y-direction.

(5)

(Fig. 8b). Third, connect the reference beam with thermal actuators of several series numbers, and apply a voltage to simulate the force output (Fig. 8c).

Table 2shows the stiffness, displacement and force output of series numbers 1–7, where series number is thermal actuator number, Disp. is actuating displacement, Kais actuator stiffness,

and Kr is reference beam stiffness. Force outputs are calculated

using Eq.(5), andFig. 8d plots the results. The simulation results show a linear relationship between series number and force out-put. However, power consumption correlates positively with series

F ¼ ðKaþ KrÞ Disp: ð5Þ

3.3. Simulated dynamic frequency

To determine the resonant frequency and shapes of the de-signed structure, an instantaneous analysis is performed using Intellisuite FEM software is performed to verify the vibration char-acteristics. The simulated ﬂexure structures are completely re-served and modeled, and the calculated weights of the shuttle

and IS indicate that they are equivalent to those of the other struc-tures. The following equation gives the resonant frequency. Fresonant¼ 1 2

p

ffiffiffiffiffi k M r ð6Þ where k is the system stiffness of the image stabilizer, M is the over-all weight of the image stabilizer, and Fresonantis the resonant

fre-quency of the XY stage. Because the structure is symmetrical, the simulated resonant frequency of the ﬁrst mode is 4660 Hz with movement in x-direction and in y-direction.

4. Fabrication

4.1. Fabrication processes

The fabrication steps are device layer formation, substrate for-mation, and the ﬁnal combination of IS and XY stage.Fig. 9depicts the device layer fabrication process. This segment performs the steps required for device layer formation, including structure def-inition, heat isolation, circuit routing and electric isolation. Fabrica-tion begins with a (1 0 0) SOI wafer, where the thicknesses of the device layer, isolator and substrate are 50

l

m, 2

l

m and 350

l

m, respectively. The fabrication process is performed in the following steps:

a. After basic RCA cleaning, use CVD process to deposit a 0.5

l

m-thick nitride layer on the topside and a 2

l

m-thick silicon oxide layer on the backside of the SOI wafer (Fig. 9a). b. During photolithography, pattern the nitride as the electric isolation between the shuttle and circuit on the SOI wafer topside (Fig. 9b).

Fig. 8. (a) Reference beam displacement; (b) actuator beam displacement; (c) actuator displacement when apply 1000lN. (d) Force output versus series number of thermal actuators.

Table 2 Simulation results.

Series number Disp. (lm) Ka Kr F (lN)

1 27.16 137.36 91.74 6222 2 34.62 283.29 91.74 12,984 3 38.03 425.53 91.74 19,672 4 40.00 568.18 91.74 26,397 5 41.3 714.28 91.74 33,289 6 42.25 854.7 91.74 39,987 7 42.96 1000 91.74 46,901

(6)

c. Deposit a 1

l

m-thick Al layer by DC sputter and pattern as the circuits and pads (Fig. 9c).

d. Pattern the device structure using the ﬁrst ICP process to a depth of 50

l

m (Fig. 9d).

e. Coat and pattern SU8 to isolate heat and electricity in the actuators from the shuttle (Fig. 9e).

f. Use RIE process to pattern the backside oxide layer to be hard mask (Fig. 9f).

g. Use the second ICP process to open the suspending area on the back of the SOI wafer (Fig. 9g).

h. Release the structure by RIE process (Fig. 9h). 4.2. Flip–chip bonding

Fig. 10shows an IS combined with the proposed image stabi-lizer by ﬂip–chip approach.Fig. 11depicts the packaging process captured on the ﬂip–chip bonder monitor. Fig. 11a–c shows the

Fig. 9. Fabrication process for the XY stage: (a) backside oxide and nitride deposition (oxide: 2lm, nitride: 0.5lm), (b) nitride patterning, (c) Al patterning (1lm), (d) device layer patterning, (e) SU8 patterning, (f) backside oxide patterning, (g) backside layer patterning (350lm), and (h) release.

Fig. 10. The XY stage after ﬂip–chip bonding.

Fig. 11. Flip–chip bonding process of the image stabilizer.

Fig. 12. Reflow profile of flip–chip process.

Fig. 13. MMA system.

(7)

IS absorbing process. Before absorption, the spectroscope is used to align the pick-up head and sensor (Fig. 11b).Fig. 11d–f shows how the sensor bumps and pads of XY stage are aligned by the com-bined processes of the XY stage and IS.Fig. 11e and f shows how 3 N force is applied to the IS before the bumps are reflowed to 235 °C, andFig. 12shows the reflow profile.

5. Results and discussion

Experiments performed to test the effectiveness of the proposed image stabilizer included a static driving test with the actuator dri-ven by a DC voltage and actuator displacement measured with a MEMS motion analyzer (MMA) (Fig. 13). When a 14-V square wave was applied to the proposed device in the y-direction, moving dis-placement in the y-direction was 27

l

m, and displacement in the x-direction was only 1.42

l

m (Fig. 14). The experimental decou-pling ratio in the two directions was 19, which conforms to the minimum ratio of 10 required by the system.Fig. 15shows how displacement varies with driving voltage. The MMA system was also used in the dynamic characterization to evaluate the resonant frequency of the proposed image stabilizer. Fig. 16 shows that although system resonant frequency approximates 4485 Hz, the operating bandwidth is determined by the thermal time constant. In this case,Fig. 17shows the step wave test results for the thermal actuating XY stage, and the device has a rising time of 21 ms.

Fig. 18shows the measurement results for thermal imaging. The red1_{and blue triangles represent the maximum and the}

mini-mum temperatures of the actuator, respectively, when 10 V is ap-plied at 1 Hz. The ﬁgure shows that the maximum temperature

occurs in the middle of the actuator and that the pads approach room temperature. Additionally, comparison ofFig. 18a and b shows that the maximum temperature of the thermal actuator is about

Fig. 14. Displacement of x- and y-direction on 14 V.

Fig. 15. Static measurement of XY stage in x- and y-directions.

Fig. 16. Measurement results for resonant frequency without IS.

Fig. 17. Step wave test in 10 V.

Fig. 18. Measurement results for thermal imaging in 10 volts 1 Hz. (a) Thermal actuator without isolator. (b) Thermal actuator with isolator.

1

For interpretation of color in Figs. 1–18, the reader is referred to the web version of this article.

(8)

290 °C when it is not connected to the isolator and about 360 °C when it is connected to the isolator. Therefore the isolator design en-ables a lower driving voltage to achieve the same actuating displacement.

6. Conclusion

This work simulated, fabricated, and tested a novel design for an integrated micro decoupling XY stage. The integrated XY stage is designed to load an IS. The IS was accurately bonded to the shuttle by ﬂip–chip bonding method. The electrical signals of the IS were also connected to the output circuits according to the signal spring design. The proposed XY stage has potential applications in com-mercially available cell phone cameras owing to its anti-shaking function. The main components of the proposed device are a sili-con-based XY stage and chevron thermal actuators fabricated and packaged with two ICP etching processes by ﬂip–chip bonding technique. Experimental results indicate that a driving voltage of 14 V produces a 27

l

m displacement in the driving direction and a 1.42

l

m displacement in the vertical direction, which are sufﬁ-cient for effective anti-shaking. The results for the perpendicular direction are similar since the design is symmetrical. Some varia-tion between the simulavaria-tion and the experimental results from the undercutting of thermal actuators during ICP etching as well as room temperature variations. The natural frequency of the pro-posed XY stage design approximates 4485 Hz, and the response time approximates 21 ms when measured by MMA system. The isolator SU8 not only insulates the heat ﬂux and driving current, it also increases actuating displacement given a similar driving voltage.

Acknowledgments

This work was supported in part by National Science Council, Taiwan, under Contract 032, 100-2220-E-009-019, and in part by Taiwan Department of Health Clinical Trial and Research Center of Excellence under Contract Nos. DOH99-TD-B-111-004 and DOH99-TD-C-111-005. This work was also sup-ported in part by the UST-UCSD International Center of Excellence in Advanced Bio-engineering sponsored by the Taiwan National Science Council I-RiCE Program under Grant No. NSC-100-2911-I-009-101. This paper is particularly supported by ‘‘Aim for the Top University Plan’’ of the National Chiao Tung University and Ministry of Education, Taiwan, ROC.

References

[1] Sun Lining, Wang Jiachou, Rong Weibin, Li Xinxin, Bao Haifei. A silicon integrated micro nano-positioning XY-stage for nano-manipulation. J Micromech Microeng 2008;1(8):125004.

[2] Takahashi Kazuhiro, Mita Makoto, Toshiyoshi Hiroyuki Fujitaand Hiroshi. A high ﬁll-factor comb-driven XY-stage with topological layer switch architecture. IEICE Electron Express 2006;3:197–202.

[3] Kwon HN, Lee J-H, Takahashi K, Toshiyoshi H. Micro XY stages with spider-leg actuators for two-dimensional optical scanning. Sens Actuator A Phys 2006;130–131:468–77.

[4] Kwon S, Lee LP. Stacked two dimensional micro-lens scanner for micro confocal imaging array. In: Proceedings IEEE MEMS ’02; 2002. pp. 483–6. [5] Laszczyk K, Bargiel S, Gorecki C, Krezel J. Towards integration of glass

microlens with silicon comb-drive X–Y microstage. In: IEEE/LEOS international conference on optical MEMS and nanophotonics; 2008. pp. 168–9.

[6] Yasumura KY, Grade JD, Jerman H. Fluid damping of an electrostatic actuator for optical switching applications. In: Proceedings solid-state sensor, actuator and microsystems workshop; 2002. pp. 358–61.

[7] Takahashi K, Kwon HN, Mita M, Saruta K, Lee J-H, Fujita H, et al. A silicon micromachined f–h microlens scanner array by double-deck device design technique. IEEE J Sel Top Quantum Electron 2007;13:277–82.

[8] Yang JP, Mou JQ, Chong NB, Lu Y, Zhu H, Jiang Q, et al. Probe recording technology using novel MEMS devices. Microsyst. Technol 2007;13:733–40. [9] Ando Y. Development of three-dimensional electrostatic stages for scanning

probe microscope. Sens Actuator A 2004;114:285–91.

[10] Indermuhle PF, Linder C, Brugger J, Jaecklin VP, de Rooij NF. Design and fabrication of an overhanging XY-microactuator with integrated tip for scanning surface proﬁling. Sens Actuator 1994;43:285–91.

[11] Kawai Y, Ono T, Esashi M, Meyer E, Gerber C. Resonator combined with a piezoelectric actuator for chemical analysis by force microscopy. Rev Sci Instrum 2007;7(8):063709.

[12] Vancura C, Ruegg M, Li Y, Lange D, Hagleitner C, Brand O, Hierlemann A, Baltes H. Magnetically actuated CMOS resonant cantilever gas sensor for volatile organic compounds. In: Proceedings 12th international conference on solid state sensors, actuators and microsystems; 2003. pp. 1355–8.

[13] Sasaki M, Bono F, Hane K. XY-stage for scanning media for optical data storage. In: Proceedings IEEE optical MEMS and their applications; 2006. pp. 36–7. [14] Kim CH, Jeong HM, Jeon JU, Kim YK. Silicon micro XY-stage with a large area

shuttle and no-etching holes for SPM-based data storage. J Microelelctrmech Syst 2003;12:470–8.

[15] Lu MS-C, Fedder GK. Position control of parallel-plate microactuators for probe-based data storage. J Microelectromech Syst 2004;13:759–69. [16] Liu X, Kim K, Sun Y. A MEMS stage for 3-axis nanopositioning. J Micromech

Microeng 2007;17:1796–802.

[17] Mukhopadhyay D, Dong J, Pengwang E, Ferreira PM. A SOI MEMS based 3-DOF planar parallel-kinematics nanopositioning stage. Sens Actuator A 2008;147:340–51.

[18] Gu Lei, Li Xinxin, Bao Haifei, Liu Bin, Wang Yuelin, Liu Min, et al. Single-wafer-processed nano-positioning XY-stages with trench-sidewall micromachining technology. J Micromech Microeng 2006;16:1349–57.

[19] Yeom DH, Park NJ, Jung SY. Digital controller of novel voice coil motor actuator for optical image stabilizer. In: International conference on control, automation and systems; 2007. pp. 2201.

[20] Sato K, Ishizuka S, Nikami A, Sato M. Control techniques for optical image stabilizing system. IEEE Trans Consum Electron 2005;39:461–6.

[21] Okamoto Yasuhiro, Yoshida Ryuichi. Development of linear actuators using piezoelectric elements. Electron Commun Jpn Part III 1998;81:11–7. [22] Chen Guan-Rong, Yeh Yeou-Min, Wang Sheng-Jyh, Chiang Huang-Cheng. A

novel structure for digital image stabilizer. In: IEEE Asia–Paciﬁc conference on circuits and systems; 2000. pp. 101–4.

[23] Mohamed A, Elsimar H, Ismail M. Analysis, and optimization of a CMOS vertical thermal actuator. In: Proceedings symposium design, test, integration and packaging of MEMS/MOEMS; 2003. pp. 214–7.

[24] Sachs D, Nasiri S, Goehl D. Image stabilization technology overview. <http:// www.invensense.com/shared/pdf/ImageStabilizationWhitepaper_051606.pdf>. [25] Lee Chih-Chun, Chen Wen-Chih, Lee Sz-Yuan, Fang Weileun. Design and implementation of a novel polymer joint for thermal actuator current and thermal isolation. IEEE MEMS Conf 2010:156–9.

[26] Que L, Park J-S, Gianchandani YB. Bent-beam electro-thermal actuators for high force applications. J Micro Electro Mech Syst 1999:31–6.

[27] Lin Chun-Ying, Chiou Jin-Chern. Design, fabrication and actuation of 4-axis thermal actuating image stabilizer. Micro Nano Lett 2011;6(7):549–52. C.-Y. Lin, J.-C. Chiou / Solid-State Electronics 77 (2012) 64–71 71