Asymmetric Tuning Schemes of MEMS Dual-Shutter VOA - IEEE Xplore

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 5, MARCH 1, 2008

569

Asymmetric Tuning Schemes of MEMS Dual-Shutter VOA X. M. Zhang, Member, IEEE, Q. W. Zhao, A. Q. Liu, Member, IEEE, J. Zhang, John H. Lau, Fellow, IEEE, and C. H. Kam

Abstract—A dual-shutter MEMS variable optical attenuator (VOA) is designed for advanced tuning functions such as linear attenuation relationship and simultaneous coarse and fine tunings. The mechanism behind is to take advantage of the additional shutter to render one more degree of freedom for attenuation adjustment. Although dual-shutter VOAs with asymmetric functionalities have been reported before, these intrinsic capabilities owing to asymmetry have not been extensively investigated. In experiment, the fabricated VOA device has demonstrated a linear tuning over a 20-dB range with respect to the driving voltage of one shutter, and it has also realized simultaneously coarse tuning (2.5 dB/V) and fine tuning (0.1 dB/V) by the two shutters. Ideally, the tuning can start from any available working point, linear to any controlling parameter, at any slope of linearity, and with any tuning resolution. Theoretical attenuation model has also been developed to provide a roadmap for the VOA design and choice of working point. An interesting finding is that over a certain range the linear attenuation can be obtained by moving a fixed aperture rather than by reducing the aperture size, which greatly relaxes the difficulty of shutter position control. The measured results match well with the theoretical data, implying the possibility of developing a look-up table to locate the shutter positions quickly. The dual-shutter VOA accomplishes these features without the need of high-precision control systems and therefore gives a structure-based rather than a control-system-based solution, clearly advantageous over the previously developed VOAs. Index Terms—Linear attenuation, microelectromechanical systems (MEMS), microoptics, ultrafine tuning, variable optical attenuator.

I. INTRODUCTION HE variable optical attenuator (VOA) is a basic component for controlling the optical power in optical systems. It finds wide applications in lab equipment and particularly in optical networks to regulate the signal strength of different channels [1]. Microelectromechanical systems (MEMS) technology has attracted broad interest for developing the VOA devices owing to its potentials of fast control, compact size, low power consumption as well as better optical performance [2], [3]. Over

T

Manuscript received April 3, 2007; revised October 24, 2007. This work was supported by A*STAR under Grant 042 108 0095. The work of X. M. Zhang was supported by the Singapore Millennium Foundation (SMF) of the Postdoctoral Fellowship. X. M. Zhang, Q. W. Zhao, A. Q. Liu, and C. H. Kam are with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798 (e-mail: [email protected]; [email protected]; [email protected]; [email protected]). J. Zhang and J. H. Lau are with the Institute of Microelectronics, MMC, Singapore Science Park II, Singapore 117685 (e-mail: [email protected]; [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JLT.2007.912524

these years, many MEMS VOA designs have emerged based on the mechanisms of partially blocking the light using a shutter [4]–[10] (i.e., shutter-type VOAs) and steering away the light beam using a mirror [11]–[19] (i.e., mirror-type VOAs). Although other physical mechanisms have also been presented such as laser interference [20], [21], evanescent field coupling [22] and photoacoustic interaction [23], the shutter-type and mirror-type VOAs have received most interest of research and commercialization due to easy implementation/integration and superior optical performance [24]. Among many specifications of the VOA, nearly all applica, large attenuation range tions require low insertion loss or even ), fast response speed , low wavelength dependent loss (WDL) and low polarization dependent loss (PDL). These specifications can be considered as the basic requirement. Most of the developed MEMS VOAs meet this requirement very well [4]–[19]. For the convenience of control and adjustment [25], many applications further need linear attenuation relationship and high tuning resolution. This can be regarded as the advanced requirement. In this paper the terms “linear” and “nonlinear” refer to the relationship between the attenuation level (in dB, the unit is most commonly used in VOAs) and a controlling parameter (such as the voltage, mirror displacement, or rotation angle, etc.). If such advanced requirement is imposed, many of the developed VOAs will not perform satisfactorily. Typically, the shutter-type VOA utilizes a moving shutter to block the light propagation between two the fibers, but the attenuation change is nonlinear to the shutter position (i.e., movement) [4]–[8]. When the shutter starts to cut into the light beam from an open state, the attenuation rises up very slowly at the beginning, corresponding to high tuning resolution. But when most of the light beam is blocked, the attenuation goes up sharply with any further movement of the shutter, leading to low tuning resolution and even instability of controlling the shutter position. Some variations of the shutter-type VOAs employ a pair of shutters or even more shutters to block the light [9], [10], [26]. Most of them adopt a symmetric moving scheme like adjusting the aperture of cameras. Such symmetric-shutter VOA can reach high attenuation level quickly, but it has even worse nonlinearity since there is a singularity when the shutters meet to fully block the light path and thus the attenuation goes to infinity. Another design that has very similar configuration to the shutter-type VOA makes use of single or a pair of transparent silicon wedges to deflect the light beam away from the core of the output fiber [27], [28]. It well suppresses the PDL, but still suffers from the nonlinear attenuation and low tuning resolution. In the mirror-type VOAs, most of the early work demonstrates nonlinear attenuation behavior to mirror translation (or rotation) regardless of the types of arrangement of

0733-8724/$25.00 © 2008 IEEE

570

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 5, MARCH 1, 2008

the mirror and fibers [4]–[24]. To linearize the attenuation relationship, our group has recently proposed two VOA designs by changing the mirror shape to be elliptical and parabolic [29], [30]. range has been demonstrated. Linear attenuation over Nevertheless, such designs have complicated configuration and, thus, cause difficulty for packaging and production. Among almost all the developed VOAs, the attenuation is dependent on only one parameter (typically, position or driving voltage of the mirror/shutter) and has only one degree of freedom. In such VOAs, it is contradictory to obtain the linear attenuation relationship and the fine tuning at the same time. Linear attenuation means the tuning resolution is fixed by the gradient of the linear relationship and thus does not allow for higher resolution. In addition, fine tuning means slow increase of attenuation and may not be desirable for sweeping the attenuation over a large range. Two approaches are used in an attempt to solve this problem. First, a more standard solution uses linear tuning over a specific range, with an asymmetric moving scheme alleviating the difficulty in control accuracy. Second, two control voltages are used to obtain a balance between high resolution and fast sweeping. These simple ideas are integrated into an asymmetric dualshutter VOA design based on the control of two degrees of freedom as described in this paper. It is noted that although some earlier VOA designs adopted dual-shutters arrangement [9], [10], [26] and some of them may have asymmetric functionality [26], most designs were used in a symmetric way, whereas the potential capabilities of asymmetric configurations have not been extensively investigated. The purpose of this paper is to analyze theoretically and experimentally how the inclusion of an additional shutter provides one more degree of freedom for the choices of the operating point, attenuation relationship and tuning resolution. This paper is organized to have the theoretical studies elaborated in Section II, covering the optical attenuation model and the tuning analyses. Fabrication and assembly will be discussed in Section III, followed by the experimental verification in Section IV. II. DESIGN AND ANALYSIS A. Attenuation Model of Asymmetric Dual-Shutter VOA The configuration of the asymmetric dual-shutter VOA is shown in Fig. 1. Two independently movable shutters are inserted into the gap between the input and output fibers. The tips of the shutters are deflected by certain degrees (typically 8 or 12 degrees) to allow their engagement of each other and also to reduce the back reflection [5], [31], because the direction of reflected light from the shaped shutters makes it difficult to be coupled back into the input fiber). The facets of input and and , respectively, while output fibers are defined as . In this model, all these planes the plane of the shutters is are parallel and share a common axis. Shutter positions are denoted by and (coordinates on the axis) in Fig. 1, where the signs of and depends on whether the corresponding shutter is higher or lower than the axis. For single-mode fiber, the input beam is assumed to be a normalized Gaussian profile , which has a beam waist radius of in the plane. To calculate the coupling

Fig. 1. Schematic diagram of the dual-shutter VOA for attenuation modeling.

efficiency with respect to two shutter positions, the analytical method similar to our previous work in [32] is adopted. The can be obtained by field pattern in the shutter plane computing the Fresnel–Kirchhoff diffraction integral [33] (1) where is the wavelength, is the wavenumber, is the distance between the points considered, and is the and are not inclination factor. As long as the planes and, thus, the integral can spaced too closely, it has in be worked out analytically. Then the field pattern can be calculated the same way by propagating the plane to the plane at the output fiber facet. Finally, the coupling efficiency can be obtained by coupling into the output fiber by the mode overlap integral [34] as given by

(2) where and are the input power and output power of represents the complex conjuthe attenuator, respectively. gate of the normalized Gaussian profile . Based on (1) and (2), the coupling efficiency can be expressed with respect to the positions of two shutters as given by

(3) where and

and indicate the shutter positions (see Fig. 1); are dimensionless variables defined by and , represents the initial loss given by respectively; ; and are the separation from the shutter plane to the input fiber facet and the output fiber facet, respectively (see and Fig. 1); are the waist radii after the Gaussian beam is propagated by distances of and , respectively; is the

ZHANG et al.: MEMS DUAL-SHUTTER VOA

571

Rayleigh range; and Erf is the complex-variable error function defined by . In the derivation of (3), it is assumed that the two shutters are infinitely long in the direction; however, the shutter 1 is in the lower part of the direction while the shutter 2 in the high part. When the and ), the shutters are fully open (i.e., and 1, respectively, and, thus, error functions converge to is the insertion loss. The form of is exactly the same as the widely used formula of insertion loss in [35] (by taking and in [35, Eq. (34)]). It well validates the effectiveness of (3). Compared with the results in [32], (3) simplifies the insertion loss term and also arrives at a simple form. Finally, the attenuation in dB, by definition, is determined by (4) For simplicity, in the rest of paper the positions of shutter 1 and shutter 2 will be represented by their tips’ coordinates and in Fig. 1, respectively. The calculated attenuation (in as a function of positions of the shutters 1 and dB) 2 is contoured in Fig. 2(a), which exists only in a triangular . This is because the region above the diagonal line light path will be fully blocked and theoretically the attenuaas represented by the tion goes to infinitely large if , shaded region in Fig. 2(a). In the triangular region of the attenuation is kept at the low level for and [top left corner in Fig. 2(b)], i.e., the shutters are almost open to let most of the incident light pass through. In such low level region, the gradient is small, which allows for fine attenuation tuning. At the bottom left and top right corners of the triangular region, the attenuation increases rapidly and, thus, experiences high gradient (low tuning resolution). Such contour provides a roadmap for the choice of attenuation tuning scheme. Here the term attenuation tuning scheme means coordinating the movement of the two shutters ( and ) to obtain a certain tuning relationship. Two common tuning schemes (combination of and ) can be readily presented by curved loci (I and II) in the contour in Fig. 2(a). The scheme I is for single-shutter tuning, i.e., the attenuation is tuned by only the to ) while the shutter 2 is always kept shutter 1 ( from ). This actually represents the tuning reopen ( fixed at lationship in many developed single-shutter VOAs [4]–[8]. The corresponding relation between attenuation and are plotted in Fig. 2(b) by the curve I, which is quite flat at the beginning but then becomes steep. Such undesirable nonlinearity is because the tuning scheme I passes through the low-gradient region and then the high-gradient region as is read in Fig. 2(a). The scheme II is for the symmetric-shutter tuning, i.e., the two shutmoves ters are positioned and moved symmetrically ( from ( 10, 10) to (0, 0)) so as to control the attenuation level by purely adjusting the size of the aperture (always in the center of light path), which is used in some studies [9], [10], [26]. It can be observed in Fig. 2(a) that the scheme II is mostly in the low-gradient region, but experiences drastic gradient change when it moves close to the center of the graph. As verified in Fig. 2(b), (or ) relationship rises slowly the corresponding within a small range when . but shoots up to

Fig. 2. Calculated relationships between the attenuation level, the shutter 1’s position and the shutter 2’s position. (a) Contour of the attenuation level A with respect to the positions of the two shutters  and  . (b) Loci of the two shutters’ positions to realize different nonlinear and linear tuning schemes. The scheme I is for single shutter, i.e., only the shutter 1 is movable while the shutter 2 is kept open; II is for symmetrically shutter design, i.e., the two shutters are moved by the same amount; III is for linear tuning, in which the attenuation is linear to the shutter 1’s displacement; and IV is for another linear tuning, in which the attenuation is linear to the applied voltage of the shutter 1. (c) Relationships between the attenuation level and the shutter 2’s position for the schemes III and IV.

Such nonlinearity results low tuning resolution and control instability. In Fig. 2(b), the shaded region is inaccessible as limited by the single-shutter tuning. In the calculation, the VOA has

572

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 5, MARCH 1, 2008

, , . The parameters are the same as those of the fabricated VOA to be used in the experiment. B. Analysis of Linear Tuning Schemes In the analyses above, a straight locus of versus in the contour of Fig. 2(a) leads to a nonlinear attenuation tuning relationship in Fig. 2(b). To obtain a linear tuning, the locus should be curved. The schemes III and IV represent two types of linear tuning relationships, in which the attenuation is linear to the shutter 1’s position and the driving voltage, respectively. The linear tuning loci are calculated as follows. First, one control variable (shutter position or applied voltage) is selected. Then, the attenuation curve depending on this parameter is computed, assuming the second shutter is taken away from the light path. This curve determines the upper boundary of the attenuation when the second shutter cuts in. After this, on a segment (P-Q) of this attenuation curve computation is carried out for the position of the other shutter required to add extra light blocking to increase the attenuation value to the linear value. This approach works since the single-shutter attenuation curve is monotonically decreasing and strictly convex, which always requires the insertion of a second shutter to make it linear, and of (3) and the computation is nothing but finding a root on and . The choice of these two linear (4) with specified schemes as the examples is because in analysis it is easy to link the attenuation to the shutter 1 position [as given in (3)], while in most real applications the driving voltage is the directly controllable parameter. To compare these tuning schemes, and from the scheme I the same initial and final states are chosen as shown in Fig. 2(b). For the state , it has and , while for it has and . The scheme III is a straight line in Fig. 2(b) as , here the start point given by , the slope , and the shutter 1’s initial position and is implicitly a function to achieve the linear relation. The scheme IV is of , here , and given by and are the driving voltages to the actuators which control the positions of the shutters 1 and 2, respectively. The driving voltage is related to the shutter 1 position by the parabolic rela, here . Such a relation tionship is from the measured result of the comb drive actuator to be used in the experiment. Due to the parabolic relationship and , the scheme IV is a parabolic curve in Fig. 2(b). of The loci of the schemes III and IV are both complicated curves; keeps increasing from to 7 , while moves back and forth so as to maintain a linear increase of the attenuation level. is The actual relationship between the attenuation level and plotted in Fig. 2(c) for both the schemes III and IV. It is interesting to note in Fig. 2(a) that both the loci for the schemes III and IV have linear segments inclined at nearly 45 degrees. It implies that when the attenuation increases, both shutters are practically moving at almost the same pace toward one direction, while the aperture size, which is defined to be , remains constant. To make it clear, the aperture size (i.e., ) as functions of attenuation is drawn in Fig. 3 for both the schemes III and IV. The aperture sizes of III and IV reduce

Fig. 3. Aperture size as a function of the attenuation level for the linear tuning schemes III and IV. Segments of constant aperture size can be observed, which is very useful for reducing the position control difficulty at the high attenuation level. Minimum aperture size can also be read from the curves, which are 0.35 and 0.21  for the schemes III and IV, respectively.

m

with higher attenuation level but the sizes remain over the range of roughly 25–35 dB, before they increase again. This implies that higher attenuation is due to the movement of the aperture formed by the two shutters, rather than the sizes. The segment of constant aperture size is a useful property since the linear attenuation can be achieved by merely moving the aperture, rather than reducing the size. This fundamental feature of the dualshutter design may help eliminate the need for extreme narrow apertures and therefore complicated control systems, even when high attenuation is attempted. To demonstrate this advantage, let us consider achieving 1-dB resolution at 35-dB level, the linear tuning schemes III and IV needs both shutters to move at steps , while the single-shutter scheme I needs 0.18 of 0.36 and the symmetric-shutter scheme needs 0.01 . The accuracy of shutter movement is thus relaxed by 2 to 30 times just by use of the linear tuning scheme. Meanwhile the aperture size is not required to be progressively smaller to obtain higher atfor the tenuation; instead it has a minimum value. It is 0.35 for the scheme IV. scheme III and 0.21 It should be emphasized that linearity here refers to only one , for which the linear attenuation curve gives satshutter is deisfactory resolution performance. The second shutter to achieve this linearity. If strict signed to coordinate with to maintain this linearity, moving accuracy is required for the linear tuning becomes meaningless because the control difto . To prevent this possibility, ficulty is transferred from the moving accuracy (or equivalently tuning resolution) of should be less restrictive compared to . The linear tuning is only found useful in such a case where attenuation changes relatively slowly with respect to and . This criterion is satisfied in the above example, since according to Fig. 3 the aperture size approaches a constant, meaning both shutters share the same moving accuracy. Some diffracted beam patterns of the scheme IV are exemplified in Fig. 4 for three sets of the shutter positions, i.e., (a) , and correspondingly ;

ZHANG et al.: MEMS DUAL-SHUTTER VOA

573

Fig. 4. Diffraction beam patterns for the linear tuning scheme. The intensity distribution after the shutter plane and the intensity and phase distribution in the :  , :  , and correspondingly A : . (b)  :  , output fiber facet plane are presented for different shutter positions. (a)   :  , and A : . (c)  , :  , and A : .

= 02 2 m

= 22 8 dB

=0

=03 m

= 07 0 m = 31 4 dB

(b) , and ; and , and . For each case, (c) the intensity distribution after the shutter plane (right after the light beam being chopped by the two shutters) is shown together with the corresponding intensity and phase distributions in the output plane on the output fiber facet. The phase is calculated with respect to the input fiber facet, where the phase is regarded as zero uniformly on the facet. It can be observed from Fig. 4 that the discontinuous beam chopped by the shutters is restored to a continuous beam pattern after being diffracted to the output plane; the intensity varies quite fast in the direction but the phase distribution is merely flat. With reduction of the aperture size, the field distribution is actually extended in the direction due to the thin slit diffraction. The schemes I–IV serves just as the examples of the many ways to choose the tuning schemes. Since the dual-shutter design has two degrees of freedom to choose the shutters’ positions, it is feasible to make the attenuation be linear to the

= 01 0 m

= 11 7 dB

= 03 5 m

other control parameters, which may have complicated relationship with the shutter movement (generated by other actuation meanings, thermal or piezoelectric actuators [7]). In addition, the linear tuning can start from any point at any slope, only if the start and end points fall in the accessible region in Fig. 2(b). This brings in one more flexibility to set the working condition of the dual-shutter VOAs. C. Analysis of Tuning Resolution Due to the asymmetry of the two shutters, it is possible to obtain coarse tuning and fine tuning in one device simultaneously since, for instance, moving the shutter 1 close to the beam center would obviously give low tuning resolution while keeping the shutter 2 away from the center gives high tuning resolution. Here the main objective is to find one pair of shutter positions (and the aperture position) so that the attenuation changes due to two shutters give an appropriate tuning contrast, such as 1:25. Meanwhile these two resolution values are neither too low nor too

574

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 5, MARCH 1, 2008

realized by two shutters simultaneously. Here half of the plane is ignored as is inaccessible since the shutter overlap and are plotted discussed above. For better visual effect, in logarithmic scale. Due to exchangeability of and , the if . The values of in Fig. 5(b) are replaced by dark band in Fig. 5(a) corresponds to the local maximums in the calculated attenuations and thus has little practical significance. As can be expected from Fig. 2, the resolution can go if the shutters are in the low gradient region up to 0.1 in Fig. 2(a), i.e., high resolution at the low attenuation level. One can also read from Fig. 5(b) that the tuning contrast is 1 (symmetric configuration), but can be easily when increased to the level of 10 if an asymmetric arrangement is algives , lowed. For example, i.e., 11 times difference of tuning effects by two shutters, and hence they can be used for fine and coarse tuning, respectively. can be achieved if both In fact, high tuning contrast the shutters lie in the same side, which corresponds to the high gradient regions in Fig. 2(a). That is to say, an off-axis aperture appears to be critical for satisfactory performance in both the resolution and tuning contrast for the dual-shutter design. In practice, sometimes it is more convenient to use the resolution’s dependence on the driving voltage. In this aspect, the has to be taken into account. If actuator characteristic has the actuator is an electrostatic comb drive for which with rea quadratic form as discussed above, the contrast spect to driving voltage is given by

(7) If

Fig. 5. Contours of the tuning resolution. (a) Tuning resolution of the shutter 1. (b) resolution contrast of the two shutters.

high (i.e., slow tuning) to be impractical. The tuning resolution of the two shutters can be described by (5) and the tuning contrast, which measures the ratio of two resolutions, can be defined as (6) In both definitions the variable is in unit of dB whereas the and are in unit of . Fig. 5(a) illustrates variables for every possible combination of shutter positions ( is similar to and is, thus, not shown) and values of as a function of are plotted in Fig. 5(b). Parameters in the model are the same as Fig. 2. The former contour can be used ) or to find out the operating points for coarse tuning (large fine tuning with shutter 1 (small ), while the second is for locating a pair of shutter positions when coarse and fine tuning are

has typical values of 1–5 and the required value of is assumed to be 10 to 30, the desired contrast value is estimated to be only 2 to 6. As can be seen from Fig. 5(b), this requirement can be met easily by the dual-shutter design. In this aspect, the position control of the shutters is further relaxed. Although the actual feasibility and effectiveness of the VOA device are also strongly dependent on the repeatability of actuators, Fig. 5(b) provides an important roadmap for the selection of working points at which a satisfactory tuning contrast is available. III. FABRICATION AND ASSEMBLY Based on previous analysis, an asymmetric dual-shutter VOA has been designed and fabricated by a two-mask process. Overview of the assembled device as well as the close-up of the shutter part is shown in Fig. 6. In the actuator design, each shutter is driven by a pair of bidirectional comb drives so as to allow it to move both forward and backward for easy adjustment of the initial working condition. The mechanical in both direcstructures are designed for translation of 22 tions with a maximal applied voltage of 10.5 V. The shutter-tips are deflected by 8 degrees, to avoid light being reflected back to the input fiber. The shutter’s thickness and length of the and 46 , respectively. When they deflected part are 5 begin to engage each other, the tips are still separated by 4 in the direction to avoid being stuck to each other.

ZHANG et al.: MEMS DUAL-SHUTTER VOA

575

Fig. 7. Attenuation curve of the single-shutter tuning scheme.

Fig. 6. Fabricated dual-shutter VOA device. (a) Overview of the assembled device. (b) Close-up of the shutter part.

The device is fabricated on an 8-in SOI wafer with a 100device layer and a 2buried oxide layer. In the first step, a PVD aluminum layer is patterned and etched to form the electrodes, after which the structures including the comb drives, the shutters and the fiber trenches are patterned and etched by DRIE method in one step. Movable structures are then released by wet etching. Finally, the shutter part is coated with 200 nm aluminum to block the light. The fiber grooves are designed such that the spacing between the assembled input and output fibers , and the shutter plane is equidistant to are separated to be 57 both fibers facets. The design of fiber gap is to provide enough space to accommodate two freely moving shutters so that they do not stick to the fixed structure. The fiber stoppers on both sides of the gap are designed to avoid breaking the shutters unintentionally during the assembly [see Fig. 6(b)]. Single-mode and beam waist 5.2 ) are asfibers (outer diameter 125 sembled onto the fabricated chip using UV epoxy, and then the whole device is packaged into a butterfly box. Since the depth , the fiber core is below of the fiber grooves is 100 the top surface of the device after the assembly. According to the calculation above, the stray light passing through the top of the shutters is negligible in moderate attenuation levels. IV. EXPERIMENT RESULTS In order to have a direct comparison between theory and experiment, performance of the comb drives is first examined. Since totally four comb drives are involved for the bidirectional

movement of the two shutters, their electrostatic coupling can be serious enough to induce unwanted movement or vibration of the structure, resulting in jittering, leap and mutual attraction of shutters. These phenomena, however, can be suppressed by grounding all suspended electrodes. Static behavior of the comb drives is found to be well fitted by a quadratic relation, with the linear fitting ratio for ship is all the actuators on the device in experiment. The ratio for all the four comb drives. With this movement quality, the actuation behavior of comb drives can be used with confidence for the verification of the attenuation model in Section II. In experiment, it is more convenient to monitor the driving voltage than the actual shutter position. A. Single-Shutter Tuning Scheme The single-shutter tuning (i.e., the scheme I) is measured to examine the validity the physical model. It is realized by withdrawing one shutter far away from the light beam. As shown in Fig. 7, the experimental results agree well with the calculation up to the attenuation level of 43.9 dB. Over this tuning range, measured attenuation is slightly higher than the prediction by an . This discrepancy may almost constant discrepancy of result from some additional losses not included in the calculation model. In particular, insertion loss of the device is 1.2 dB, a bit higher than the predicted value of 1.0 dB. When the attenuation reaches 43.9 dB, however, the measured attenuation begins to deviate from the prediction and is clamped at 44.5 dB. This phenomenon strongly implies that a constant amount of stray light, which comprises approximately 0.01% of the input, can always be coupled into the output fiber and is unaffected by the moving shutter. It is to be noted that this ’saturated’ attenuation value does not represent the largest tuning range achievable of the device. The highest attenuation rate is attained in the symmetric-moving scheme, which is realized by applying the same (the sensitivity voltage on both shutters. It can go up to ). The large discrepancy of the optical power-meter is between the blocking effects of single- and dual-shutter moving scheme is because in this design one single shutter is only able . This distance does not suffice to totally block to move 22 the coupling between the input and output side, considering the whole area around the fiber gap is coated with reflective metal

576

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 5, MARCH 1, 2008

due to the use of shadow mask. The blocking of two shutters is much better since both tips can engage. Since very large attenuation range is not the primary target of this study, this is not further investigated. B. Linear Tuning Scheme The linear tuning performance (i.e., the scheme IV) of the device is also tested. For convenience of explanation, the shutter 1 is chosen as the main shutter, whose driving voltage is to have a linear relationship with the attenuation, while the shutter 2 is as the auxiliary shutter. For easy demonstration, we choose an attenuation range from 1.2 to 20.7 dB, which corresponds to a driving voltage from 0.0 to 7.4 V. In experiment two methods can be used to realize the linear tuning function since three variable are involved (i.e., , and ). The first method is similar to that in calculation. For a targeted linear attenuation-voltage relation, the shutter 1 is positioned according to the calculation while the shutter 2 is moved back and forth until the desired attenuation is obtained. The resultant voltage of the shutter 2 is then recorded and compared with the values from the calculaand to test tion model. In other words, this method gives . The second method is by use of a look-up table of the required voltages calculated from the models. The voltages are then directly applied to the comb drives. The obtained attenuaand tion is then compared with a linear one. That is, it gives to test . The obtained data for these two methods are shown in Fig. 8(a) and (b), respectively. One can see from Fig. 8(a) that the measured voltages of the shutter 2 are larger than the calculated by less than 1 V, which suggests that slightly more light is coupled into the output fiber and an extra movement of the shutter 2 is required to balance it. Since the discrepancy seems to be independent of shutter positions, we believe that the unaccounted light coupling is through the spray light on top or beneath the shutters. The detailed reason is however still unclear. The measured attenuation curve obtained by the look-up table method is shown in Fig. 8(b). It is seen that the attenuation level develops with satisfactory linearity up to the voltage of 4.6 V, after which the attenuation curve is higher than simulation by is the approximately 1.5 dB. It is noticed that point at which shutter 2 begins to move backward, as shown in Fig. 8(a). Therefore, the shutter structure becomes a bit closer to the center than expected when it is drawn back due to the residual stress of the structure, resulting in less effective obstruction of the light path, which explains the attenuation shift . It is noted that this behavior of attenuation beyond after 4.6 V is not evident in Fig. 8(a). Actually from the same argument as above one may have concluded that the measured should be lower than the calculation when so as to reduce the attenuation to a required value as implied by Fig. 8(b). The measured resolution performance of the device obtained by moving two shutters is illustrated in Fig. 9(a) and (b). The attenuation level is chosen to be 11.1 dB, and the shutter posiand , corresponding tions are to the driving voltage of 6.6 and 2.5 V, respectively. Hence, the shutter 1 is for fine tuning [Fig. 9(b)] while the shutter 2 for coarse tuning [Fig. 9(a)]. At this working point, the applied volt) while the attenages are changed over a small range

Fig. 8. Demonstration of linear tuning schemes. (a) Comparison of the measured and calculated values of V when the values of V and A are given. (b) Comparison of the measured and calculated values of A when the values of V and V are given according to a look-up table.

uation shift is examined. In Fig. 9(a) and (b), two different resolutions 2.5 and 0.1 dB/V are realized simultaneously. This represents a tuning resolution contrast of 25 times, which appears to be suitable for practical use. It is seen that the attenuation can be easily tuned by steps as small as 0.1 dB (i.e., fine tuning). In Fig. 9(a), the measured local attenuation curve (and therefore the resolution) is well approximated by the calculated values within the small region considered, while in Fig. 9(b) such fitting is less satisfactory due to the small scales. Since the required working point for a given resolution contrast differs from the attenuation level, the calculation model can be used to choose the appropriate shutter positions. It is worth noting that sensitivity of attenuation value with respect to two shutters’ movement is strongly dependent on their positions, i.e., the working point selected. When the coarse tuning shutter is moved by a large step, however, the attenuation varies steeply and the original working point is lost and a new working point has to be set to maintain a desirable tuning contrast (simultaneous fine and coarse tuning). The range over which the coarse tuning shutter is allowed to move without the need to change to a new working point can be estimated from calculating the change of resolution with respect to . In the current case from and , there are and . Hence, if we stipulate that the tuning contrast in (6) should drift less than two times from

ZHANG et al.: MEMS DUAL-SHUTTER VOA

577

Fig. 10. Influence of the aperture position on the wavelength dependence loss. The nominal attenuation level is kept constant at 25 dB for the wavelength of 1550 nm regardless the aperture position.

Fig. 9. Demonstration of both the fine tuning (0.10 dB/V) and coarse tuning (2.53 dB/V) in one device. The shutter 1 is under 2.5 V while the shutter 2 is under 6.6 V. The starting attenuation is 11.1 dB.

the current value during the coarse tuning, the maximal attenuation change can be estimated as

(8) A tuning range of approximately 3.3 dB can be obtained by coarse tuning at the 11.1-dB level, without resetting the working point. C. WDL, PDL and Dynamic Response Some other properties of the dual-shutter VOA are also characterized, such as the WDL, PDL and response speed. The WDL and PDL cannot be measured in a regular way due to two tuning variables involved here. Their absolute values are strongly dependent on the positions of both shutters. Since the same level of attenuation can be realized by many combinations of two shutter positions (i.e., different aperture position and corresponding aperture size), the WDL could be very different. To investigate this aperture effect, the WDL at a nominal 25-dB level is measured as shown in Fig. 10. In measurement, the size and position of aperture are varied but the attenuation at 1550 nm is kept unchanged. From Fig. 10, it is interesting to see that the attenuation oscillates over the whole wavelength range when the aperture is moved away from the center. For some particular wavelength such as 1580 nm, the attenuation is

Fig. 11. Dynamic responses of the packaged VOA device to the step signals of driving voltage.

simply independent of the aperture position (the independence at 1550 nm is because it is chosen as the reference). In contrast, at several wavelengths such as 1570 and 1600 nm, the attenuation changes sharply from a local maximum to a local minimum with the shift of the aperture. The reason of this phenomenon is still not clear. Regarding to the influence of the aperture position, it can be seen from Fig. 10 that the fluctuation is alleviated when the aperture is close to the beam center, leading to reduced wavelength dependence. The lowest PDL of this device is 0.8 dB at the 25 dB level over the wavelength range of 1520–1620 nm. Polarization dependence of the attenuation is measured with a deterministic method introduced by [36]

578

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 5, MARCH 1, 2008

and also turns out to be a function of shutter positions even at the same attenuation level. However, its actual dependence on aperture position exhibits no special behavior. The attenuation variation on polarization is measured to be 1.0 dB at the 25-dB level. Finally, the response time of the packaged device is also measured. The response of attenuation to a step voltage signal is shown on Fig. 11. The rise time (10%–90%) and fall time (90%–10%) of the optical power are shown to be 34 and 40 , respectively. Though the operating speed is not the key consideration of this VOA design, its performance surpasses many reported structures utilizing rotation or thermal effects [13]. V. CONCLUSION A dual-shutter MEMS variable optical attenuator (VOA) is presented for advanced tuning functions such as linear attenuation relationship and ultra-fine tuning. The key feature is the inclusion of an additional shutter, which provides two degrees of freedom (DOF) for the choice of attenuation relationship in comparison to the previous one-DOF VOAs. In experiment, the VOA fabricated on a SOI wafer has an attenuation range of when two shut43.9 dB if only one shutter is used and ters are used. The insertion loss is 1.2 dB. Linear attenuation with respect to the driving voltage of one shutter is demonstrated over 20-dB range. Meanwhile, two shutters serve as the coarsetuning arm (2.5 dB/V) and the fine-tuning arm (0.1 dB/V) are also exemplified. Reasonable agreement between experiment and theory shows that the calculation model can potentially be developed into a look-up table to set up the shutter configuration quickly. REFERENCES [1] C. R. Giles, V. Aksyuk, B. Barber, R. Ruel, L. Stulz, and D. Bishop, “A silicon MEMS optical switch attenuator and its use in lightwave subsystems,” IEEE J. Sel. Topics Quantum Electron., vol. 5, no. 1, pp. 18–25, Jan./Feb. 1999. [2] S. Cohen, “Novel VOAs provide more speed and utility,” Laser Focus World, vol. 36, no. 11, pp. 139–146, Nov. 2000. [3] X. Ma and G. S. Kuo, “Optical switching technology comparison: Optical MEMS vs. other technologies,” IEEE Opt. Commun., vol. 41, no. 11, pp. S16–S23, Nov. 2003. [4] B. Barber, C. R. Giles, V. Askyuk, R. Ruel, L. Stulz, and D. Bishop, “A fiber connectorized MEMS variable optical attenuator,” IEEE Photon. Technol. Lett., vol. 10, no. 9, pp. 1262–1264, Sep. 1998. [5] C. Marxer, P. Griss, and N. F. de Rooij, “A variable optical attenuator based on silicon micromechanics,” IEEE Photon. Technol. Lett., vol. 11, no. 2, pp. 233–235, Feb. 1999. [6] X. M. Zhang, A. Q. Liu, C. Lu, and D. Y. Tang, “MEMS variable optical attenuator using low driving voltage for DWDM systems,” Electron. Lett., vol. 38, no. 8, pp. 382–383, 2002. [7] H. Debéda, T. v. Freyhold, J. Mohr, U. Wallrabe, and J. Wengelink, “Development of miniaturized piezoelectric actuators for optical applications realized using LIGA technology,” J. Microelectromech. Syst., vol. 8, no. 3, pp. 258–263, Sep. 1999. [8] C. Lee, Y.-J. Lai, C.-Y. Wu, J. A. Yeh, and R.-S. Huang, “Feasibility study of self-assembly mechanism for variable optical attenuator,” J. Micromech. Microeng., vol. 15, no. 1, pp. 55–62, 2005. [9] L. Li and D. Uttamchandani, “Design and evaluation of a MEMS optical chopper for fibre optic applications,” Inst. Elect. Eng. Proc.-Sci. Meas. Technol., vol. 151, no. 2, pp. 77–84, 2004. [10] R. R. A. Syms, H. Zou, J. Stagg, and H. Veladi, “Sliding-blade MEMS IRIS and variable optical attenuator,” J. Micromech. Microeng., vol. 14, no. 12, pp. 1700–1710, 2004. [11] C.-H. Kim and Y.-K. Kim, “MEMS variable optical attenuator using a translation motion of 45 tilted vertical mirror,” J. Micromech. Microeng., vol. 15, pp. 1466–1475, 2004.

[12] K. Isamoto, K. Kato, A. Morosawa, C. Chong, H. Fujita, and H. Toshiyoshi, “A 5-V operated MEMS variable optical attenuator by SOI bulk micromachining,” IEEE J. Sel. Top. Quantum Electron., vol. 10, no. 3, pp. 570–578, May/Jun. 2004. [13] T.-S. Lim, C.-H. Ji, C.-H. Oh, H. Kwon, Y. Yee, and J. U. Bu, “Electrostatic MEMS variable optical attenuator with rotating folded micromirror,” IEEE J. Sel. Topics Quantum Electron., vol. 10, no. 3, pp. 558–562, May/Jun. 2004. [14] C. Chen, C. Lee, and J. A. Yeh, “Retro-reflection type MOEMS VOA,” IEEE Photon. Technol. Lett., vol. 16, no. 10, pp. 2290–2292, Oct. 2004. [15] C. Chen, C. Lee, and Y.-J. Lai, “Novel VOA using in-plane reflective micromirror and off-axis light attenuation,” IEEE Commun. Mag., vol. 41, no. 8, pp. S16–S20, 2003. [16] C. Lee, “Arrayed variable optical attenuator using retro-reflective MEMS mirrors,” IEEE Photon. Technol. Lett., vol. 17, no. 12, pp. 2640–2642, Dec. 2005. [17] R. R. A. Syms, H. Zou, J. Stagg, and D. F. Moore, “Multistate latching MEMS variable optical attenuator,” IEEE Photon. Technol. Lett., vol. 16, no. 1, pp. 191–193, Jan. 2004. [18] B. M. Anderson, S. Fairchild, and N. Thorsten, “MEMS variable optical attenuator for DWDM optical amplifiers,” in Proc. Opt. Fiber Commun. Conf. (OFC), Baltimore, MD, USA, Mar. 7–10, 2000, pp. 260–262. [19] S. Sumriddetchkajorn and N. A. Riza, “Fault-tolerant three-port fiberoptic attenuator using small tilt micromirror device,” Opt. Commun., vol. 205, pp. 77–86, 2002. [20] O. Solgaard, F. S. A. Sandejas, and D. M. Bloom, “Deformable grating optical modulator,” Opt. Lett., vol. 17, no. 9, pp. 688–690, 1992. [21] J. E. Ford, J. A. Walker, D. S. Greywall, and K. W. Goossen, “Micromechanical fiber-optic attenuator with 3 ms response,” J. Lightw. Technol., vol. 16, no. 9, pp. 1663–1670, Sep. 1998. [22] F. Chollet, M. de Labachelerie, and H. Fujita, “Compact evanescent optical switch and attenuator with electromechanical actuation,” IEEE J. Sel. Topics Quantum Electron., vol. 5, no. 1, pp. 52–59, Jan./Feb. 1999. [23] N. A. Riza and Z. Yaqoob, “Submicrosecond speed variable optical attenuator using acoustooptics,” IEEE Photon. Technol. Lett., vol. 13, no. 7, pp. 693–695, Jul. 2001. [24] M. Morimoto, K. Morimoto, K. Sato, and S. Iizuke, “Development of a variable optical attenuator (VOA) using MEMS technology,” Furukawa Rev., no. 23, pp. 26–31, 2003. [25] B. Borovic, A. Q. Liu, D. Popa, H. Cai, and F. L. Lewis, “Open-loop vs closed-loop control of MEMS devices: Choices and issues,” J. Micromech. Microeng., vol. 15, no. 10, pp. 1917–1924, 2005. [26] J. C. Chiou and W. T. Lin, “Variable optical attenuator using a thermal actuator array with dual shutters,” Opt. Commun., vol. 237, no. 4, pp. 341–350, 2004. [27] J.-H. Lee, S.-S. Yun, Y. Y. Kim, and K.-W. Jo, “Optical characteristics of a refractive optical attenuator with respect to the wedge angles of a silicon optical leaker,” Appl. Opt., vol. 43, no. 4, pp. 877–882, 2004. [28] B. Glushko, S. Krylov, M. Medina, and D. Kin, “Insertion type MEMS VOA with two transparent shutters,” in Proc. Asia-Pacific Conf. Transducers Micro-Nano Technol. (APCOT), Singapore, Jun. 25–28, 2006, Paper 95-OMN-A0594. [29] H. Cai, X. M. Zhang, C. Lu, A. Q. Liu, and E. H. Khoo, “Linear MEMS variable optical attenuator using reflective elliptical mirror,” IEEE Photon. Technol. Lett., vol. 17, no. 2, pp. 402–404, Feb. 2005. [30] X. M. Zhang, A. Q. Liu, H. Cai, A. B. Yu, and C. Lu, “Retro-axial VOA using parabolic mirror pair,” IEEE Photon. Technol. Lett., vol. 19, pp. 692–694, 2007. [31] M. Takahashi, “Variable light attenuator of improved air-gap type with extremely low return light,” in Proc. IEEE Instrum. Measure. Technol. Conf. (IMTC ’94), Hamamatsu, Japan, May 10–12, 1994, pp. 947–950. [32] X. M. Zhang, A. Q. Liu, and C. Lu, “New near-field and far-field attenuation models for free-space variable optical attenuators,” J. Lightw. Technol., vol. 21, no. 12, pp. 3417–3426, Dec. 2003. [33] M. Born and E. Wolf, Principles of Optics, 6th ed. Oxford, U.K.: Pergamon Univ. Press, 1999. [34] H. Kogelnik, “Coupling and conversion coefficients for optical modes,” in Proc. Symp. Quasi-Optics, New York, Jun. 8–10, 1964, pp. 333–347. [35] S. Yuan and N. A. Riza, “General formula for coupling-loss characterization of single-mode fiber collimators by use of gradient-index rod lenses,” Appl. Opt., vol. 38, no. 15, pp. 3214–3222, 1999. [36] R. M. Craig, S. L. Gilbert, and P. D. Hale, “High-resolution, nonmechanical approach to polarization-dependent transmission measurements,” J. Lightw. Technol., vol. 16, no. 7, pp. 1285–1294, Jul. 1998.

ZHANG et al.: MEMS DUAL-SHUTTER VOA

X. M. Zhang (S’03–M’05) received the B.Eng. degree in precision mechanical engineering from the University of Science and Technology of China, in 1994, the M.Eng. degree in optical instrumentation from Shanghai Institute of Optics and Fine Mechanics, the Chinese Academia of Science, in 1997, the M.Eng. degree from the Department of Mechanical Engineering, National University of Singapore, and the Ph.D. degree from the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore. His research interests include optical MEMS, optical communication, optical instrumentation, and measurement.

Q. W. Zhao received the B.Sc. degree from the Department of Microelectronics, Peking University, China, in 2006. He is pursuing the Ph.D. degree at the School of Electrical and Electronic Engineering, Nanyang Technological University (NTU), Singapore. His research interest is in photon MEMS design, fabrication processes, and device physics.

A. Q. Liu (M’03) received the B.Eng. degree from Xi’an Jiaotong University in 1982, the M.Sc. degree from Beijing University of Posts and Telecommunications in 1988, and the Ph.D. degree from the National University of Singapore (NUS) in 1994. Currently, he is an Associate Professor with the School of Electrical and Electronic Engineering, Nanyang Technological University (NTU), Singapore. His research interests are MEMS design, simulation, and fabrication processes. Dr. Liu is as an Associate Editor for the IEEE SENSOR JOURNAL and also a Guest Editor for Sensors & Actuators A Physical.

579

J. Zhang received the B.Eng. degree in optoelectronics engineering in 1997 from Huazhong University of Science and Technology (HUST), China, and the M.Eng. degree in optical engineering in 2000, also from HUST. She received the Ph.D. degree in 2005 in fiber optics from the School of Electrical and Electronic Engineering, Nanyang Technological University (NTU), Singapore. Currently, she works as a senior research engineer with the Institute of Microelectronics, Singapore. Her research interests include optical devices for communications, waveguides, and photonics packaging.

John H. Lau (M’88–SM’90–F’94) received three master degrees in structural engineering, engineering physics, and management science, and the Ph.D. degree in theoretical and applied mechanics from the University of Illinois. He has more than 30 years of R&D and manufacturing experience in the electronics, photonics, and automobile industries. Presently, he is the Head of Microsystems, Modules and Components Laboratory, A STAR’s Institute of Microelectronics. Dr. Lau is an elected ASME Fellow.

C. H. Kam received the B.Eng. degree in electrical engineering from University of Singapore in 1974, the M.Sc. degree in physics from Nanyang University, Singapore, in 1976, the M.Sc. degree in electrical engineering from the University of Southern California, Los Angeles, in 1986, and the Ph.D. degree in high energy physics from the National University of Singapore in 1984. He is presently a Professor with the Division of Microelectronics, School of Electrical and Electronic Engineering, Nanyang Technological University (NTU). His research interest includes Sol-Gel photonics, quantum transport, nonlinear optical interactions, and spectroscopy of rare-earth doped glasses.