Design and Nonlinear Servo Control of MEMS Mirrors ... - IEEE Xplore

JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 14, NO. 2, APRIL 2005

261

Design and Nonlinear Servo Control of MEMS Mirrors and Their Performance in a Large Port-Count Optical Switch Patrick B. Chu, Member, IEEE, Igal Brener, Member, IEEE, Chuan Pu, Member, IEEE, Shi-Sheng Lee, Senior Member, IEEE, Jerry I. Dadap, Sangtae Park, Member, IEEE, Keren Bergman, Senior Member, IEEE, Nicolas H. Bonadeo, Tai Chau, Ming Chou, Robert A. Doran, Rick Gibson, Roey Harel, John J. Johnson, C. Daniel Lee, David R. Peale, Bo Tang, Dennis T. K. Tong, Ming-Ju Tsai, Qi Wu, William Zhong, Evan L. Goldstein, Fellow, IEEE, Lih Y. Lin, Senior Member, IEEE, and James A. Walker

Abstract—In this paper, we demonstrate full closed-loop control of electrostatically actuated double-gimbaled MEMS mirrors and use them in an optical cross-connect. We show switching times of less than 10 ms and optical power stability of better than 0.2 dB. The mirrors, made from 10- m-thick single-crystal silicon and with a radius of 400–450 m, are able to tilt to 8 corresponding to 80% of touchdown angle. This is achieved using a nonlinear closed-loop control algorithm, which also results in a maximum actuation voltage of 85 V, and a pointing accuracy of less than 150 rad. This paper will describe the MEMS mirror and actuator design, modeling, servo design, and measurement results. [1247] Index Terms—Electrostatic actuator, microelectromechanical systems (MEMS), nonlinear closed-loop control, optical cross-connects, snap-down, torsional spring.

Manuscript received January 5, 2004; revised April 14, 2004. Subject Editor H. Fujita. P. B. Chu is with Seagate Technology, Pittsburgh, PA 15222 USA (e-mail: [email protected]). I. Brener was with Amersham Biosciences/GE Healthcare, Piscataway, NJ 08855 USA. He is now with Sandia National Laboratories, Albuquerque, NM 87185 USA. C. Pu is with the Center for Optical Technologies, Lehigh University, Bethlehem, PA 18015 USA. S.-S. Lee is with the Institute for Cell Mimetic Space Exploration (CMISE), University of California, Los Angeles, CA 90032 USA. J. I. Dadap is with Columbia University, New York, NY 10027 USA. S. Park is with Applied MEMS, Stafford, TX 77477 USA. K. Bergman is with Columbia University, New York, NY 10027 USA. N. H. Bonadeo, M. Chou, R. Gibson, J. J. Johnson, and B. Tang were with Tellium, Inc., Oceanport, NJ 07757 USA. T. Chau is with Raytheon, Space and Airborne Systems, El Segundo, CA 90245 USA. R. A. Doran is with Symbol Technologies, Inc., Holtsville, NY 11742 USA. R. Harel is with Kodeos Communications, South Plainfield, NJ 07080 USA. C. D. Lee is with Inplane Photonics, Inc., South Plainfield, NJ 07080 USA. D. R. Peale is with KLA-Tencor, San Diego, CA 92121 USA. D. T. K. Tong is with Hong Kong University of Science and Technology, Hong Kong. M.-J. Tsai is with Tessera, San Jose, CA 95134 USA. Q. Wu is with Corning, Inc., Corning, NY 14831 USA. W. Zhong is with Tellium, Inc., Oceanport, NJ 07757 USA. E. L. Goldstein and L. Y. Lin are with the Electrical Engineering Department, University of Washington, Seattle, WA 98195 USA. J. A. Walker is with JayWalker Technical Consulting, LLC, Freehold, NJ 07728 USA. Digital Object Identifier 10.1109/JMEMS.2004.839827

I. INTRODUCTION

I

N high port-count 3D-MEMS optical cross-connects (OXC) [1], one is typically faced with the challenge of achieving as large a mirror angular swing as possible with a minimum separation between the ports to achieve high port-count and low optical loss. Most MEMS-based OXCs demonstrated to date use an open-loop architecture in mirror position control [2]–[5]. Optical power monitoring through each connection can provide certain feedback for mirror reference positions, but this is a slow process and does not provide direct mirror position control. The physics of an electrostatically actuated MEMS gimbal mirror dictates that the tilting mirror becomes unstable at a certain angle commonly referred to as the “snap-down” angle [6], [7], which lies between one-third and one-half of the mirror “touchdown” angle (i.e., when the mirror plate physically contacts the underlying electrodes). This instability phenomenon is similar to nontilting electrostatic actuators used in MEMS sensors and microrobotics [8], [9]. Under open-loop control, the mirror tilt angle is usually limited to less than a third of the touchdown angle. Due to this unstable behavior of electrostatic actuators, the physical air gap employed may need to be more than three times larger than the minimum gap needed to achieve the desired physical swing angle, thus resulting in unnecessarily high drive voltages. One method used to extend the stable operating range is to reduce the electrode radius toward the axis of rotation [10], but this method inevitably leads to lower torque (and a higher drive voltage). Using vertical electrodes working in conjunction with traditional electrodes is an effective way to increase snap-down angle [11], [12]. Depending on the implementation, this approach may require complex fabrication process, asymmetric devices [11], and potentially higher device cost. Additional complexity in OXC systems arises from the extremely high pointing accuracy required, regardless of the technology used. Typically, an angular misalignment of the order of is enough to increase the optical loss by a fraction of 100 a dB. Maintaining such a pointing accuracy of MEMS mirrors over long periods of time is a big challenge. Also, stochastic perturbations such as shock and vibration can introduce mechanical misalignment into the mirror position leading to severe penalties in optical performance.

1057-7157/$20.00 © 2005 IEEE

262


Fig. 1. Schematic cross-section of an electrostatic micromirror.

All of these issues can be eliminated through the use of active position control of the MEMS mirrors. Servo controlled mirrors can be actuated far beyond the snap-down angle [13]–[15], thereby reducing the required air gap and thus drive voltage. By employing the MEMS mirror design and servo controller described in this paper, we are able to extend the operation range to a little over of the mirror from a snap down tilt angle of of the mirror “touchdown” angle (of 10.3 ), 8 , which is with less than 100 V. Servo control also provides corrective action needed to eliminate position drifts that arise from either internal mechanisms, such as charge buildup, or external perturbations, such as vibration or shock. In addition to the servo considerations mentioned above, an ideal MEMS mirror for optical switching applications would have the following merits: light and rigid—for vibration and shock immunity; flat, large and high-reflective surface—for low optical loss; simple and repeatable motion—for high manufacturing yield and long-term stability; and linear and uncoupled motion—for simple control. These criteria govern the design parameters for our MEMS mirrors needed to achieve high-performance MEMS-based OXCs. The Section II describes the fabrication and design of the MEMS mirrors used in this work. Details of the different components of the mirror (mirror structure, actuator, and springs) will be given. In the subsequent sections, we describe the nonlinear electromechanical modeling of the mirror, followed by the nonlinear controller algorithm and the performance of a single MEMS mirror under servo control. Finally the switching and operation performance of servo-controlled MEMS mirrors inside an OXC will be presented. II. MEMS MIRROR DESIGN AND SIMULATION A. Mirror Fabrication and Design The MEMS mirrors used in this work are fabricated using Analog Devices’ proprietary optical iMEMS process [16], [17]. Optical iMEMS is an SOI (Silicon-on-insulator) based threelayer process with the sandwiched silicon spacer as the sacrificial layer. This process provides the advantages of both bulk and surface silicon micromachining technologies. An illustration of the cross-section of a MEMS mirror designed for this process is shown in Fig. 1. single-crystal silicon (SCS) is the material of choice to achieve mechanically robust springs in order to meet rigorous reliability requirements [18]. SCS also yields a thin but nearly stress-free and sufficiently rigid bulk substrate for the mirror.

Fig. 2. SEM micrograph of a fabricated MEMS mirror.

Typical mirrors used in this work have a thickness of 10 , . This same SCS layer can with radius from 400 to 450 be used to fabricate precision high- and low-voltage BiCMOS circuitry [17]. A scanning electron microscopy (SEM) micrograph of a fabricated MEMS mirror is shown in Fig. 2. One pair of springs (gimbal axis) connects the gimbal ring to the fixed surrounding structure and the other pair (mirror axis) connects the mirror plate to the gimbal ring. The mirror axis springs penetrate into the mirror structure in order to yield a gimbal with a simple circular shape and to reduce the overall mirror weight. The additional optical loss due to the penetrating mirror axis springs is only 0.3 dB by optical beam propagation analysis. Surrounding the mirror is a trench realized with the same conductive polysilicon as the vias connecting the top circuitry to the bottom electrodes. The trench, which is initially covered with oxide, forms a critical chemical barrier for the bulk silicon sub[19] sacrificial etch of the material unstrate during the derneath the MEMS mirror. The trench also serves several other purposes during operation: i) It is used to isolate individual mirrors from each other when an array is fabricated; and ii) it is also used to define the boundary of the volume of the air pocket underneath the mirror, thus controlling the mirror quality factor, Q [20]. In order to achieve the desired optical reflectivity, a thin 700Angstrom-thick layer of gold is evaporated on the top surface of the mirrors. When coated with gold on one side, these mirrors were found to have a typical radius of curvature greater than 30 cm under normal operating conditions, and with a typical , reflectivity of 97% for wavelengths of interest ( , and ). The torsional mirrors show typical resonant frequencies at 0 deflection angle between 300 and

CHU et al.: SERVO CONTROL OF MEMS MIRRORS AND THEIR PERFORMANCE IN A LARGE PORT-COUNT OPTICAL SWITCH

Fig. 3.

263

Schematic diagram of a MEMS mirror with design dimensions.

400 Hz (both axes), and a quality factor (Q) of 3–5 in air. The low Q value of the mirror is realized primarily by the damping force of the air under the mirror. This low Q value improves the mirror dynamics significantly, as will be discussed later. B. Actuator Design In theory, three electrodes are adequate for controlling a two-axis gimbaled mirror. However, the symmetry of a four-electrode design helps to decouple the motion of the two axes and thus simplifies the control algorithm. Therefore, the MEMS mirror presented here is actuated through the application of voltages to four electrodes located 80 underneath the mirror with an illustrative geometry shown in Fig. 3. Every quadrant electrode can be further divided in an optimal fashion between driving and sensing electrode if capacitive sensing is required. The sensing portion may be used for angular sensing through differential-capacitance measurements using on-chip circuitry [21]. The optimization of the electrode design is based on tradeoffs between several primary considerations, including linearity and resolution for drive torque and sense capacitance as functions of voltages and tilt angles, and maximum drive voltage needed to move the mirror between any two arbitrary angles. Other considerations include switching time (governed by magnitude of electrostatic torques and mechanical properties like spring constant and Q), position resolution and stability (dependent on actuator linearity, torque and capacitance look-up table implementation, A/D and D/A resolution, and so forth), and electronics reliability (maximum drive voltage). The desired maximum actuation voltage for our mirror designs is 85 V. This voltage value is selected for electronics circuitry reliability, MEMS device reliability, electronics drift, and noise considerations. Depending on the specific mirror design, this maximum voltage takes into account an additional 10% to 25% margin for servo performance optimization in addition to the maximum open-loop snap-down voltage. In the segmented electrode design shown in Fig. 3, both inner and outer electrodes can serve for either driving the mirror or sensing the mirror angle. Inner electrodes typically offer better linearity in terms of capacitance and rate of change in capacitance as a function of tilt angle compared to outer electrodes. One of the main reasons for this effect is a slower change of the average gap between the mirror and the electrodes as a function of tilt angle for inner electrodes. Linearity in the capacitance is desirable for capacitive sensing so that the tilt angle can be readily estimated from the measured

capacitance without complex look-up tables or conversion functions, which may introduce error in the angle estimate for feedback. On the other hand, linearity in the rate of change in capacitance is desirable for actuation because electrostatic torque is a function of the derivative of the capacitance with respect to the MEMS angle instead of the capacitance. For this reason, the inner electrodes seem suitable for both actuation and sensing. For actuation, however, additional considerations are refrom the center creates a level arm quired. The distance effect: larger could lead to larger torque for the same area of electrode. For this reason, the outside electrodes are particularly suitable for actuation, especially if drive voltage is to be minimized. The tradeoff for large torque is a less optimized linearity in terms of torque as a function of tilt angle and reduced open-loop stability. “Outside-drive” tends to yield smaller snap-down angles. In fact, it is possible to design an “inside-drive” configuration with no snap-down angle (i.e., the mirror touches down before it can snap down). In this case, several hundred volts may be needed to tilt the mirror to a large angle. In our design, we make use of “outside-drive” to ensure the maximum voltage requirement of less than 85 V. Simulations, employing finite element analysis (FEA), of the relation between the applied voltage and tilt angle for mirror and gimbal axes are shown in Fig. 4 for one of our mirror designs. This particular mirror design has a 900- -diameter mirror with 24- -wide gimbal ring. It has an axial spring constant of for both mirror and gimbal axes. 3.314 The simulations show that the angles as functions of the applied voltages are highly nonlinear. For each axis, the required voltage is monotonically increasing for angles smaller than the pull-in angle (about 4 ). For larger angles, monotonically decreasing amplitudes of voltages are needed. C. Spring Design As described earlier, there are two pairs of springs for each MEMS mirror, the mirror and the gimbal axes. Spring design impacts all aspects of mirror performance. Both stiffness and mode frequencies of a spring are important design factors in addition to physical form factor. Precision engineering of spring stiffness improves mirror controllability. Spring size may also affect optical reflectivity. All the mirror structures are fabricated on the top 10-thick SCS layer. Therefore, the springs must have the same . Restriction in this one geometric dimension thickness of 10 imposes a great challenge to design a soft torsional spring. For example, a straight tether spring would be several millimeters

264


Fig. 4. FEA simulation of open-loop actuation voltage versus tilt angle for a 900 Note that less voltage is required for the gimbal axis.

m

diameter MEMS mirror about the mirror (a) and gimbal (b) axes.

Fig. 5. Two configurations of serpentine springs with long spring members (a) orthogonal and (b) parallel to axis of rotation.

long in order to maintain the maximum operating voltage under 85 V. Such a design is impractical since the springs would be longer than the diameter of mirror itself. The only way to overcome this limitation is to fold the long straight spring into a serpentine shape. 1) Two Configurations of Serpentine Springs: There are two possible ways to fold a string spring into a serpentine shape: long elements can be either aligned to or perpendicular to the axis of rotation. A spring with the long elements normal to the axis of rotation is called orthogonal serpentine torsional spring (OSTS), and a spring with its long elements arranged parallel to the axis of rotation is called parallel serpentine torsional spring (PSTS). Fig. 5 shows serpentine springs with two different configurations. The two spring configurations offer distinct mirror performance. The OSTS and PSTS in the figure may be designed to obtain the same torsional mode frequency; however there would be a large difference in their sizes. In fact, the size of an OSTS is twice as large as a PSTS with the same fundamental mode frequency. In addition to the size advantage, the fundamental mode of a PSTS is always the torsional mode, while this is not the case for an OSTS. This is due to the difference in their fundamental working principle. In the OSTS case, the bending mode of long spring elements contributes to the majority of its torsional mode and the bending stiffness is proportional to the cubic power of its , where is the beam thickthickness (stiffness

TABLE I MODE FREQUENCY SIMULATION RESULTS OF OSTS AND PSTS CONFIGURATIONS HAVING THE SAME FUNDAMENTAL FREQUENCY. THE x AND y AXES ARE IN THE PLANE OF THE MIRROR AND z IS THE OUT-OF-PLANE AXIS, ORTHOGONAL TO THE MIRROR PLANE

ness and is the beam width). In the PSTS case, the torsional mode of individual long spring elements contributes to the torsional mode and torsional stiffness is directly proportional to . In our design, and its thickness . Typical mode frequencies (ANSYS simulation results) of OSTS and PSTS designs are tabulated in Table I. In both cases, an 800- -diameter, 10- -thick mirror structure is used in the simulation. Even though both designs have the same fundamental mode, unlike the OSTS design, the fundamental mode of the PSTS design is indeed the torsional mode as desired. 2) Enhanced PSTS: A typical PSTS is found to have a relatively high mode frequency ratio. The mode frequency ratio is defined as the ratio of the first higher mode frequency to the


fundamental mode frequency. Since only the torsional modes of the MEMS mirrors are useful in optical beam steering, it is undesirable to have torsional motions exciting the higher modes. Therefore, the mode frequency ratio ideally should be as high as possible. ANSYS modeling shows that in the PSTS design these short spring elements have insignificant contribution to the torsional resonant frequency but have more effect on the higher modes. By stiffening the short spring elements, the torsional mode frequency is unchanged, while the other mode frequencies are increased. Therefore, the mode frequency ratio is further improved. When widening the short spring elements from , the mode frequency ratio increased by 8% without 2 to 4 any other penalty. However, no additional benefit is found for . widths beyond 4 To further improve the mode frequency ratio, one can enhance the PSTS design by stiffening the short spring elements as shown in Fig. 6(a). As in the case of PSTS design, spring elements perpendicular to the axis of rotation have very small contribution to the fundamental mode (torsional mode). 3) Enhanced Diamond-Shaped PSTS: The spring designs described so far have long elements with equal length, yielding an overall rectangular shape (RPSTS). The enhanced diamond-shaped parallel serpentine torsional spring (DPSTS) [see Fig. 6(b)], which has an outline of a diamond shape, provides the best mechanical performance. Extensive ANSYS simulations show that the outer elements of a PSTS contribute more to higher mode frequencies and have less effect on the fundamental frequency. Therefore, tapering (gradually shortening) the lengths of the outer elements of the spring yields a higher mode frequency ratio while maintaining a similar value of the fundamental mode. Typical improvement in mode frequency ratio is 5–10%. This improvement, however, comes with a penalty. Since the outer spring elements have nonzero contribution to the resonant frequency, the center spring element has to be lengthened in order to compensate the tapering of the outer spring elements. Therefore, a DPSTS is longer than an RPSTS design even though they have comparable overall area. This could cause an additional 0.2 dB optical loss. To minimize optical loss we chose an enhanced RPSTS as the preferred springs for our final mirror design. III. ELECTROMECHANICAL MODELING As already mentioned, the electrostatically actuated gimbaled mirror is inherently nonlinear in its actuation. Moreover, the angular sensing scheme can also have a nonlinear response, especially when differential capacitance is used for measuring angles. In order to make the control problem manageable, a number of simplifying assumptions are made regarding the plant to be controlled. The assumptions are the following: i) Only torsional motions are significant; all other motions from higher-order modes are neglected; ii) the torsional springs have a linear angle and torque relationship; iii) all interaxis mechanical coupling are negligible (but not the coupling in the electrostatic torque); iv) all electrostatic action on the mirror springs is small; v) the

265

Fig. 6. SEM micrographs of enhanced (a) rectangular PSTS and (b) diamond PSTS.

mirror damping is independent of angle; and vi) the moments of inertia of the mirror are independent of angles. We expect that the deviations from the ideal cases above, i.e., higher-order effects and nonlinearities, are minimal such that they can be bundled as phase margin that a good closed-loop servo design can account for. We now consider the electro-mechanics of our system. Let and denote the mirror and gimbal axes, respectively; while and denote the angles around the gimbal and mirror axes respectively. The equations of motion can then be expressed as

(1) Here,

are the electrostatic torques, the damping parameters, the moments of inertia, and the torsional spring stiffness. We calculate purely from geometrical factors and also by modeling software. and from experimental measurements, We extract by driving each axis with a low-amplitude, chirped, sinusoidal , and by subsequently fitting these voltage around two parameters. Equation (1) describe a classical two-dimensional, secondorder linear system, for which well-known linear control techniques can be used to design a controller using torque as the control variable [22]. The torques depend in a nonlinear fashion on the two angles and four electrode voltages, which are the actual control inputs to the electrostatic system. If one can establish a good knowledge of this nonlinearity, then the necessary applied voltages can then be determined using an inverse torque function. From symmetry considerations, we can approximate the electrostatic torques with the following formulas:

(2) where and , and is the driving electrode capacitance with respect to the mirror node (it is assumed to be the same for all quadrants).

266

Under


servo

operation, the torques are calculated by the controller. Then one needs to find a combination of voltages that will provide these torques. Equation (2) represents two , and thus the equations with four variables to solve system is underdetermined. A possible solution to (2) is to require that at any given time, at most two voltages are nonzero. This solution reduces the number of analog-to-digital converters (ADCs) that are required to drive each mirror from four to just two, but adds analog switches. This approach shifts most of the complexity of the servo design to the torque inversion problem. However, there is no , and thus simple way of measuring the capacitances some modeling simplifications have to be made. We can treat and as empirical tables, the torque functions and use experimental techniques under static conditions to measure them. We assume that the quadrant capacitance has the following functional form:

Fig. 7. Experimental torque functions f and g .

IV. SERVO CONTROL DESIGN (3)

This form yields torques that behave in a physically intuitive way, i.e., they are monotonic with angles and diverge when the angles approach “touchdown”. The parameters , , , , and can be found by performing a nonlinear fit of meaand the angles calcusured static angles for given voltages lated from (1) and (2) as shown in (4) at the bottom of the page. The only caveat is that for this procedure to be effective, data have to be measured for sufficiently of the form large angles, in particular, larger than the snapdown angles. This regime can only be accessed with the servo controller in place. Therefore, we first carry out a numerical estimate of the capaciand use the derived torques to servo the mirror to tance some angle past snapdown. This servo operation has degraded dynamical properties, but it provides enough stability to mea. When the final controller is sure the required data up to operational with the correct torques, the mirror can be moved to with the designed dynamics and stability. Fig. 7 shows a three-dimensional (3-D) representation of a typical quadrant torque found using the procedure described before. These torques are finally calculated at a finite grid of 0.25–0.5 and stored into appropriate matrices in the controller code. Torques for angles not on the grid are calculated by linear interpolation.

The main performance parameters required from the servo(with the controlled mirror are i) the switching time is mirror angle within of target angle), ii) the maximum per axis, iii) noise angular deviation due to noise is rejection and vibration immunity, and iv) a finite available control voltage of less than 85 V. We use state space techniques to design a digital controller with full-state feedback. The choice of sampling frequency is not such a simple task. As a rule of thumb, this value is comtimes higher than the desired system monly chosen to be response frequency. At first sight, it would seem that a sampling frequency of the order of a few kHz could be sufficient for this task, providing enough phase margin to account for system nonidealities. The issues of torque inversion and nonlinear behavior under finite sampling time, however, impose stricter restrictions on the stability of the system. Because such issues are difficult to quantify, some experimentation needs to be carried out within the usual controller design guidelines. From our measurements, we obtain satisfactory results for sampling frequencies between 10 and 20 kHz. A schematic representation of the controller design is shown in Fig. 8. A current estimator is used to estimate the angular position and velocity of each mirror axis. In addition, we include an integrator and a feed-forward term to compensate for the extra pole introduced by the integrator [22]. We choose the controller poles at 200 Hz (leading to switching times of the order of a few milliseconds) with estimator and integrator poles

(4)


Fig. 8.

267

Servo Controller block diagram.

Fig. 9. Comparison of switching under open loop and closed loop operation. The closed loop angular trajectory can exceed the snapdown angle and shows no overshoot.

four to six times faster, and a damping ratio of 1, which provides no angular overshoot. and , are calculated by the servo After the torques, controller, the torque inversion procedure outlined in the previous section is carried out, leading to the instantaneous values of , , , and . These values are then applied to the corresponding electrodes through the DAC. V. RESULTS AND DISCUSSIONS A. Single Mirror Closed-Loop Mirror Performance 1) MEMS Test Setup: We first evaluate the performance of a single MEMS mirror in an optical tabletop setup to verify and refine the controller design, and to obtain the electro-mechanical characteristics of the individual mirror, such as the anglevoltage relationship and electrostatic torque functions and , needed for building the optical switching system. We implement an optical method for detecting the mirror angular position. A collimated diode laser beam is incident on the mirror surface at 10 with respect to the fixed substrate normal, and is deflected by the mirror onto a 20-mm-wide position-sensitive detector (PSD). The mirror tilt angles along each axis are, therefore, calculated from the PSD readouts. The PSD is placed about 30 mm away from the MEMS mirror, which permits a maximum of 9 mirror tilt angle to be measured. The PSD output has a very low noise that is equivalent to tens of precision for the detected angle. A Matlab xPC-target control program is used to implement the close-loop control algorithm and interfaces with a

four-channel high-voltage amplifier with a 12-bit DAC for driving the mirror, and interfaces with the PSD output with a 16-bit ADC for mirror angle detection. The MEMS open-loop characteristics are also obtained with the same setup. 2) Closed-Loop Servo Results: A comparison between open and closed loop switching is shown in Fig. 9 for single axis switching. It is evident that the servo controller removes the angular limitation imposed by the instability past snapdown. The touchdown angle for this axis is 10.3 , and thus we were able to servo successfully up to 80% of the angular range. The maximum angle that can be achieved with such controller is limited by the reduction of phase margin caused by the finite sampling time combined with the mirror moving in the unstable regime. Other factors, such as modeling imperfections, neglected resonances, etc., can contribute too. Despite the short switching time, the maximum voltage under servo operation never exceeds 85 V throughout the servoed angle range. Steady-state voltages for each electrode for different tilt angles within a 9-degree cone are also measured. The maximum steady state voltage is about 81 V. Note that the values of the required steady-state voltages for this large range of angles can be experimentally obtained only with the aid of the closed-loop servo. The most stringent measure of success is the steady state angular noise. We record the peak angular noise around the target angle over a period of several minutes, and for a full mesh of for both axes. This noise is recorded using angles spanning the same angular sense circuitry used for servo control, which

268


Fig. 10. A colormap representation of the peak to peak noise under servo operation and for both axes.

provides an additional 25–50 noise coming from this circuitry alone. The peak-to-peak noise map is shown in Fig. 10. mandates The optical power stability requirement that the angle stability for a single mirror must be better than around the target angle. This corresponds to a max150 peak-to-peak noise requirement, a number imum of 300easily met by our controller and plant. We also tested the robustness of the servo controller against the variations of mirror resonance frequencies and torque tavariations bles. We find that for the tested mirrors having in these values, the controller functions essentially the same by using only one set of parameters, which included torsional frequencies, Q, spring stiffnesses, moment of inertia of the mirror , and . The robustness of the conand gimbal, troller significantly facilitates its implementation by eliminating the necessity of adjusting controller parameters for each specific mirror within a certain range. B. Modular Optical Crossconnect With MEMS Mirrors 1) Optical Crossconnect Setup: We use these closed-loop controlled mirrors to build a free-space modular optical switch [23]. A schematic diagram of the optical path consisting of optical heads is shown in Fig. 11, representing the architecture for modular optical cross-connect. Each optical head a is composed of a fiber with a fused lens, a fixed folding mirror, a MEMS mirror, a collimating lens, and an optical angle-sensing unit (not shown in the figure) measuring up to 7 . The total optical path between the transmitting and receiving ports is and the collimated beam diameter is . Details of this optical architecture are described in [23]. An out-of-band beam is employed for sensing the MEMS tilt angles. The optical angle-sensing unit in each optical head is composed of a dichroic beam splitter located between the MEMS mirror and the collimating lens, a set of focusing lenses, a position-sensitive detector (PSD), and supporting electronics. For an input port, the co-propagating data and angle-sensing beams at wavelengths 1310 nm and 980 nm, respectively, exit the input fiber and are reflected by the fixed and MEMS mirrors. These beams are then separated by the dichroic filter, which passes the data beam with negligible loss and reflects fully the out-of-band beam into a PSD through a set of focusing optics. A similar setup is used in a receiving port, except the data and the angle-sensing beams are propagating in opposite directions [23].

Fig. 11. A schematic diagram of the optical path consisting of two optical heads (optical angle-sensing unit is not shown).

2) Hardware Setup for Servo Control: The servo control coding is implemented using a 100 MHz 600FLOPS (32-bit floating point) digital signal processor (ADSP-21 161N) with interfaces to multiple ADC and DACs (16/12 bit resolution). After code optimization, a single DSP could servo 8 mirrors with a sampling rate of 20 KHz (or 16 mirrors at 10 KHz). The signal from the PSD in the optical angle-sensing unit for each MEMS mirror is preamplified and manipulated in an analog fashion to obtain meaningful angles [24]. A quad-voltage amplifier takes the low voltage output of the DAC’s and converts them to the appropriate level required to drive the mirrors. As mentioned before, the use of a closed-loop controlled mirror allows the use of a smaller gap between mirror and actuators, thus reducing necessary driving voltage considerably. Therefore, voltage amplifiers with a maximum output of 85 V (as opposed to 200–300 V usually required by open loop designs [3]) satisfy our requirement. 3) Optical Alignment: An optical connection is established by accurately aligning the input and output port MEMS mirrors. The optical alignment procedure is as follows: For each input port and output port combination, the theoretical pointing angles for each pair of MEMS mirrors are computed based on the ideal geometric locations and optical parameters of the various relevant optical components. Mirror curvature, focal length of lenses, and optical path lengths are all taken into account. These calculated angles serve as starting angles for a four-dimensional peak-search algorithm that determines the best pointing angles based on optimizing the coupling power for a given connection [23]. The four dimensions are from the two tilting angles of both input and output mirrors. The optimized angles are then stored in a look-up table for each connection. 4) Optical Cross-Connect Performance: Fig. 12(a) shows a fiber-to-fiber switching event commanded to a pair of mirrors. The switching time, as determined from the fiber-coupled optical power, is of the order of 5 ms, and was fairly independent of the angular swing (this is expected in a closed-loop design). When both mirrors are under closed loop control and establishing a connection, the fluctuations in the optical power are


269

the deposition temperature of the reflective coating and the nominal operation temperature of the MEMS mirror. Retargeting of the MEMS mirrors, however, cannot compensate for coupling loss contributed by mirror curvature change due to temperature. Fortunately, optimization of the mirror-coating process as well as temperature control of the switch fabric can minimize this loss mechanism. VI. CONCLUSION We demonstrate full closed-loop control of electrostatically actuated MEMS mirrors and their performance in a free space optical crossconnect. The mirror fabrication process, actuation method, and spring designs are presented. A state space controller is developed to implement the nonlinear control of the mirrors. The mirror angular operation range is extended to 80% of the mirror touchdown angle, well beyond its snapdown angle. is also achieved. Long-term angular noise of less than We use these closed loop controlled mirrors in a free space optical crossconnect, and achieve switching times less than 10 ms and optical power stability of better than 0.2 dB. REFERENCES Fig. 12. (a) A fiber-to-fiber switching event as observed through the fiber-coupled optical power when two ports establish a connection. (b) Optical power stability under a stable connection. The angles used for these measurements were close to 5 , which is beyond snapdown.

of the order of 0.1–0.2 dB, which are very close to the target values. This is shown in Fig. 12(b). Results as shown in Fig. 12(a) is normalized to the optimal steady-state transmitted power of that optical connection. Note that the best total optical loss for a given connection is independent of the performance of the servo system. Instead, it is dependent on the optical components’ quality, packaging, and assembly, path length, and mirror angles. The theoretical opfor all tical loss for our optical configuration is possible connections, taking into consideration of various loss mechanisms including MEMS mirror etch holes, spring area, and variation in MEMS mirror radius of curvature (20–60 cm). Using prototype mirrors with gold coating on only 85% of the , close mirror surfaces, the measured losses were [23]. to the predicted value of While the PSD optical feedback systems could achieve excellent performance in pointing each mirror to a target angle, a given pair of target angles does not always yield optimal coupling power due to thermal mechanical drift of the whole optical train. Electronics drift (in PSD sensor or amplifier) may also invalidate the target angle values. To compensate for slow thermally-induced drift, power measurements taken at 1-kHz rate are used to trigger the peak-search algorithm whenever the optical loss is greater than a selected threshold (such as 0.4 dB) in order to determine new optimal pointing angles. With this “dual” feedback system, long term optical power stability of is possible by properly setting better than 0.5 dB at the peak-search threshold and step-size. The optical power sensitivity to temperature change is also strongly dependent on the MEMS mirror curvature, and thus, to

[1] P. B. Chu, S. S. Lee, and S. Park, “MEMS: The path to large optical crossconnects,” IEEE Commun. Mag., vol. 40, no. 3, pp. 80–87, Mar. 2002. [2] A. S. Dewa, J. W. Orcutt, M. Hundson, D. Krozier, A. Richards, and H. Laor, “Development of a silicon two-axis micromirror for an optical cross-connect,” in Tech. Dig. IEEE Solid-State Sensor Actuator Workshop, Hilton Head Island, SC, 2000, pp. 93–96. [3] R. Ryf et al., “1296-port MEMS transparent optical crossconnect with 2.07 petabit switch capacity,” in Tech. Dig. Optical Fiber Conference 2001, Anaheim, CA, 2001. [4] D. J. Bishop, C. R. Giles, and G. P. Austin, “The Lucent LambdaRouter: MEMS technology of the future here today,” IEEE Commun. Mag., vol. 40, no. 3, pp. 75–79, Mar. 2002. [5] J. H. Smith, S. S. Nasiri, J. Bryzek, M. Novack, J. B. Starr, H. Kwon, A. F. Flannery, D. L. Marx, Z. Chen, and E. Sigari, “1200 mirror array integrated with CMOS for photonic switching: Application of mechanical leveraging and torsional electrostatic actuation to reduce drive voltage requirements and increase angular tilt,” in Tech. Dig. IEEE Solid-State Sensor Actuator Workshop, Hilton Head Island, SC, 2002, pp. 378–379. [6] O. Degani et al., “Pull-in study of an electrostatic torsion actuator,” J. Microelectromech. Syst., vol. 7, no. 4, pp. 373–378, Dec. 1998. [7] Y. Nemirovsky and O. Bochobza-Degani, “A methodology and model for the pull-in parameters of electrostatic actuators,” J. Microelectromech. Syst., vol. 10, no. 4, pp. 601–615, Dec. 1998. [8] P. B. Chu and K. S. J. Pister, “Analysis of closed-loop control of parallelplate electrostatic microgrippers,” in Proc. IEEE Int. Conf. on Robotics & Automation, San Diego, CA, 1994, pp. 820–825. [9] M. S.-C. Lu and G. K. Fedder, “Closed-loop control of parallel-plate microactuator beyond the pull-in limit,” in Tech. Dig. IEEE Solid-State Sensor Actuator Workshop, Hilton Head Island, SC, 2002, pp. 255–258. [10] E. S. Hung and S. D. Senturia, “Extending the travel range of analogtuned electrostatic actuators,” J. Microelectromech. Syst., vol. 8, no. 4, pp. 497–505, Dec. 1999. [11] T. D. Kudrle et al., “Pull-in suppression and torque magnification in parallel plate electrostatic actuators with side electrodes,” in Proc. 12th Int. Conf. on Solid State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 360–363. [12] C. Pu, S. Park, P. Chu, S.-S. Lee, M. Tsai, D. Peale, N. Bonadeo, and I. Brener, “Electrostatic actuation of 3-D MEMS mirrors by sidewall electrodes,” in Proc. Int. Conf. on Optical MEMS and Their Applications, Waikoloa, HI, 2003. [13] I. Brener et al., “Nonlinear servo control of MEMS mirrors and their performance in a large port-count optical switch,” in Proc. OFC, vol. 1, 2003, pp. 385–386.

270


[14] T. Juneau, K. Unterkofler, T. Seliverstov, S. Zhang, and M. Judy, “Dualaxis optical mirror positioning using a nonlinear closed-loop controller,” in Proc. 12th Int. Conf. on Solid State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 560–563. [15] N. Yazdi, H. Sane, T. D. Kudrle, and C. H. Mastrangelo, “Robust slidingmode control of electrostatic torsional micromirrors beyond the pull-in limit,” in Proc. 12th Int. Conf. on Solid State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 1450–1453. [16] S. Blackstone and T. J. Brosnihan, Proc. IEEE/LEOS Int. Conf. Optical MEMS ’01, Okinawa, Japan, 2001, pp. 35–36. [17] T. J. Brosnihan et al., “Optical IMEMS—A fabrication process for MEMS optical switches with integrated on-chip electronics,” in Proc. 12th Int. Conf. on Solid State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 1638–1642. [18] K. E. Petersen, “Silicon as a mechanical material,” Proc. IEEE, vol. 70, pp. 420–457, 1982. [19] P. B. Chu et al., “Controlled pulse-etching with xenon difluoride,” in Proc. 9th Int. Conf. on Solid State Sensors, Actuators and Microsystems, Chicago, IL, 1997, pp. 665–668. [20] X. Wang, M. Judy, and J. White, “Validating fast simulation of air damping in micromachined devices,” in Tech. Digest of 15th IEEE Int. Conf. on MEMS, 1999, pp. 210–213. [21] T. Roessig. (2002) Mirrors With Integrated Position-Sense Electronics for Optical-Switching Applications. Analog Dialogue by Analog Devices Inc.. [Online]. Available: www.analog.com [22] G. F. Franklin et al., Digital Control of Dynamic Systems. Menlo Park, CA: Addison Wesley, 1998, pp. 323–325. [23] J. I. Dadap, P. B. Chu, I. Brener, C. Pu, C. D. Lee, K. Bergman, N. Bonadeo, T. Chau, M. Chou, R. Doran, R. Gibson, R. Harel, J. J. Johnson, S. S. Lee, S. Park, D. R. Peale, R. Rodriguez, D. Tong, M. Tsai, C. Wu, W. Zhong, E. L. Goldstein, L. Y. Lin, and J. A. Walker, “Modular MEMS-based optical cross-connect with large port-count,” IEEE Photon. Technol. Lett., vol. 15, pp. 1773–1775, 2003. [24] Y. Reznichenko, M. Judy, and S. Zhang, “Measuring the optical and electromechanical properties of MEMS mirrors,” in Proc. 12th Int. Conf. on Solid State Sensors, Actuators and Microsystems, Boston, MA, 2003, pp. 1466–1469.

Patrick B. Chu (M’91) was born in Hong Kong in 1970. He received the B.S. degree in electrical engineering from Massachusetts Institute of Technology (MIT), Cambridge, in 1992 and the M.S. and Ph.D. degrees in electrical engineering from University of California at Los Angeles in 1994 and 1998, respectively. Since 2002, he has worked at Seagate Research of Seagate Technology, Pittsburgh, PA, as a Research Staff Member in the Servo Dynamics group. Prior to Seagate, he worked at Tellium, Inc., Oceanport, NJ, as a Senior Member of Technical Staff from 2000 to 2002. Prior to Tellium, he worked at Tanner Research, Pasadena, CA, and Jet Propulsion Laboratory, Pasadena. He has published about 20 refereed papers, conferences and invited papers and served as principal investigator in contracts from DARPA and ARMY. His past work included MEMS accelerometers, optical devices, MEMS fuze, human–machine interface, and MEMS-based optical switch. His current interests include MEMS design and fabrication, control systems, data storage, and optical devices. Dr. Chu was a recipient of Allied Signal Fellowship in 1993 and Rand Fellowship in 1994. He is a member of Eta Kappa Nu Honor society.

Igal Brener (M’99) was born in Uruguay. He received the B.Sc. degree in electrical engineering, the B.A. degree in physics, and the Ph.D. degree in physics from the Technion—Israel Institute of Technology, Haifa, Israel, in 1983, 1983, and 1991, respectively. From 1983 to 1986, he worked for National Semiconductors in microprocessor VLSI design and as head of the NS32332 testing team. He was with Bell Laboratories from 1991 to 2000, where his research dealt with ultrafast measurements in semiconductor quantum wells, wavelength conversion and novel nonlinear-optical devices for optical communication, Terahertz research, near-field imaging, coherent phenomena in semiconductor microcavities, and GaN optical phenomena. He was with Tellium, Inc., Oceanport, NJ, from 2000 to 2002, where he worked in all-optical switching and MEMS servo-control. From 2003 to 2004, he was a Senior Scientist for Amersham Biosciences/GE Healthcare, performing research in single-molecule DNA sequencing, ultrasensitive optical detection for cellular assays and high-throughput drug discovery. He joined Sandia National Laboratories, Albuquerque, NM, in 2004, where he leads an effort in biophotonics and optical biosensing. He has authored more than 80 publications and has received 12 patents. Dr. Brener has served in the CLEO and IEEE LEOS conference committees several times. He is a Member of the Optical Society of America (OSA) and the IEEE Laser and Electro-Optics Society.

Chuan Pu (M’98) was born in Taiyuan, China, in 1972. He received the undergraduate degree in physics from Peking University, Beijing, China, and the M.S. and Ph.D. degrees in applied physics from Cornell University, Ithaca, NY, in 1998 and 2000, respectively. From 1999 to 2000, he was with AT&T Research Laboratories as a Senior Member of the Technical Staff. He worked on optical MEMS technologies with applications in optical switching, optical polarization controlling, and PMD compensation. From 2000 to 2003, he was with Tellium, Inc., Oceanport, NJ, as a Senior Member of the Technical Staff. At Tellium, he was involved in the research and development of MEMS-based optical cross-connects. Since 2004, he has been with the Center of Optical Technologies, Lehigh University, Bethlehem, PA, where he has been working on optical tweezers and biophotonics. He has authored over 25 technical publications and has received four patents. Dr. Pu is a Member of the Optical Society of America (OSA) and the IEEE Laser and Electro-Optics Society (LEOS).

Shi-Sheng Lee (M’95–SM’99) received the B.S., M.S. and Ph.D. degrees in electrical engineering from University of California, Los Angeles, in 1993, 1995 and 1998, respectively. Since 2004, he has worked at SiWave, Inc., Arcadia, CA, as a Senior MEMS Engineer. Prior to SiWave, he worked at Brion Technologies, Santa Clara, CA as a Staff Engineer involved in areas of MEMS packaging. Prior to Brion Technologies, he worked at Tellium, Inc., Oceanport, NJ, as a Senior Member of Technical Staff involved in the development of MEMS-based optical cross-connect switches. Prior to Tellium, he worked at Rockwell Science Center, Thousand Oaks, CA, as a Research Scientist. He has published more about 50 referred journal papers, conference papers and invited papers. His research interests are in the areas of microfabrication, MEMS sensors, MEMS transducers and optical MEMS. Dr. Lee was a recipient of RAND Fellowship in 1995 and 1996. He is a member of Eta Kappa Nu Honor society and IEEE/LEOS.


Jerry I. Dadap received the Ph.D. degree in physics at the University of Texas at Austin in 1995, where he studied surface nonlinear optics and ultrafast spectroscopy. In 1996, he became a postdoctoral scientist at Columbia University with the Physics and Chemistry departments where he carried out a wide range of projects from nonlinear optics of molecules and nanoparticles to optoelectronic probing of high-speed circuits and devices and generation of far-IR terahertz radiation. In 2000, he joined Tellium, Inc., Oceanport, NJ, as a Senior Member of Technical Staff and collaborated with a team of scientists and engineers to develop a large port-count MEMS cross-connect. At Tellium, he performed optical design, simulation, and MEMS and optical testing. He also developed the four-dimensional peak-search algorithm for optimizing the connection power in a MEMS switch fabric as demonstrated in this paper. In 2003, he became a research scientist at Columbia University where he now works on studies of electronic structure on nanostructured surfaces and on novel photonics materials and their applications, such as the development of silicon-on-insulator nanowire waveguides as active photonics devices.

Sangtae Park (M’94) received the B.S. and M.Eng. degrees in electrical engineering from Cornell University, Ithaca, NY, in 1993 and 1994, respectively. From 1994 to 2000, he worked at Rockwell Science Center as Member of Technical Staff, where he was engaged in research and development work of various MEMS devices. Some of his major works were on 2-D optical scanner, current sensor, and tunable capacitor for RF applications. From 2000 to 2002, he worked at Tellium, Inc., Oceanport, NJ, as Senior Member of Technical Staff, where he was involved in development work of large port count optical cross-connect switch, focusing on MEMS design and control. In 2002, he joined Molecular Reflections as Senior MEMS Engineer, working on acoustic resonance biosensor. He is presently with Applied MEMS, Stafford, TX, as Senior MEMS Development Engineer, working on various foundry projects.

271

Tau Chau received the B.S., M.S., and Ph.D. degrees in electrical engineering from the University of California at Los Angeles (UCLA), in 1994, 1996, and 2000, respectively. From 2000 to 2002, he worked at Tellium, Inc., Oceanport, NJ, as a Member of Technical Staff, where he was involved in the development of MEMS-based all-optical switch fabric for optical cross-connect. Since 2003, he has been with Raytheon, Space and Airborne Systems, El Segundo, CA, where he works on Advanced-Tactical FLIR systems. He has coauthored about 35 archived journals and refereed conferences. Dr. Chau is a member of Eta Kappa Nu Electrical Engineering Honor Society.

Ming Chou, photograph and biography not available at the time of publication.

Robert A. Doran received the B.S. degree in mechanical engineering from Lowell University, Lowell, MA, in 1984. In 1984 he joined Allied Signal, Teterboro, NJ, where he was a Senior Engineer working on space-based inertial guidance systems. His research and development efforts focused on ring-laser gyroscope technology. From 1989 to 1996, he worked at Symbol Technologies, Inc., Holtsville, NY, as a Senior Mechanical Engineer working on 1-D and 2-D laser-based bar code scanning systems, and rugged hand-held computers. From 1995 to 2001, he worked at Ericsson Inc., Research Triangle Park, NC, as a Consulting Engineer. His work included developing TDMA and PDC cellular phone systems. In 2001, he joined Tellium, Inc., Oceanport, NJ, as a Principal Mechanical Engineer where he led the mechanical engineering development of an all-optical MEMS-based cross-connect switching system. He is currently a Principal Mechanical Engineer at Symbol Technologies Inc., Holtsville, NY, working on hand-held laser scanning systems. He holds 15 U.S. patents and three European patents.

Rick Gibson, photograph and biography not available at the time of publication.

computing. Dr. Bergman is a Fellow of the Optical Society of America (OSA). She currently serves as Chair of the Optical Networks and Systems technical committee for IEEE-LEOS and as an Associate Editor for IEEE PHOTONIC TECHNOLOGY LETTERS and for the OSA Journal of Optical Networking.

Roey Harel received the B.Sc degree in physics from Tel-Aviv University, Tel-Aviv, Israel in 1989 and the M.Sc. and Ph.D. degrees in physics from the Technion—Israel Institute of Technology, Haifa, in 1996 and 2000, respectively. He is currently an Optics Engineer with Kodeos Communications, working on the development of advanced modulation format transponders. From 2000 to 2002, he served as a Senior Member of Technical Staff at Tellium, Inc., Oceanport, NJ. At Tellium, he participated in a team effort to develop an optical cross connect. In particular, he led the optical design effort to develop the world’s first modular MEMS cross connect. Prior to Tellium, he was a Member of Technical Staff at Bell-Labs, Lucent Technology, Murray Hill, NJ, conducting basic research in the fields on periodically poled LiNbO3 and coherent THz radiation from semiconductor microcavities. He is a coauthor on 15 papers and holds one U.S. patent.

Nicolas H. Bonadeo was born in Buenos Aires, Argentina, in 1967. He received a Title of “Licenciado en Fisica” in 1992 from the Universidad de Buenos Aires and the Ph.D. degree in applied physics from the University of Michigan, Ann Arbor, in 1998, where he worked on nanostructures and nonlinear optics. He also received the M.S.E. degree in systems from the Department of Electrical Engineering and Computer Science (EECS) at the University of Michigan. From 1999 to 2000, he worked on mode-locked fiber lasers, and quantum-dot dynamics as a Postdoctoral Member of Technical Staff in the Advanced Photonics Research Unit at Bell Labs, Lucent Technologies, Holmdel, NJ. During 2000, he moved to the optical division of Tellium, Inc., Oceanport, NJ, and worked as Senior Member Of Technical Staff on a MEMS-based 3-D optical cross-connect. In September 2001, he moved back to Argentina where he is an external contractor for a laser company doing research and development on semiconductor lasers. Dr. Bonadeo received the Distinguished Achievement Award 1998 from EECS and Horace H. Rackham Distinguished Thesis Award 1999. He is also member of the Argentine Consejo Nacional de Investigaciones Cientificas y Tecnologicas.

John J. Johnson received the A.A.S. degree from Brookdale Community College, Lincroft, NJ, in 1985. His degree was augmented by numerous additional credits earned in Computer Science and Physics from Brookdale and Monmouth University, West Long Branch, NJ. He has worked as a Member of Technical Staff at Bellcore, Red Bank, NJ, in Lightwave Device Research and Network Integrity/Disaster Prevention and Recovery (1985-1996); Lucent Technologies, Holmdel, NJ, in Lightwave Technology Development (1998-2000); and Tellium, Oceanport, NJ, in Advanced Technologies (2000-2002). During this time his projects included integrated photonics, audits of network integrity for the RBOCs, multistage optical amplifiers, device characterization, and automated data acquisition. He periodically and currently acts as an electronics and electronics’ contamination consultant. He is a coauthor on over 20 publications related to integrated lightwave devices fabricated on lithium niobate, indium phosphide, and glass. He co-patented an acoustooptic polarization converter with apodized acoustic waveguides and is a recipient of an R&D 100 award. He has been recognized as an Expert Witness by the Brooklyn Supreme Court in the area of contamination and remediation of electronic equipment.

Keren Bergman (M’87–SM’93) received the B.S. degree in electrical engineering from Bucknell University, Lewisburg, PA, in 1988 and the M.S. and Ph.D. degrees in electrical engineering from the Massachusetts Institute of Technology (MIT), Cambridge, MA, in 1991 and 1994, respectively. She is an Associate Professor of electrical engineering at Columbia University, New York, NY. Her research interests include ultrafast optical signal processing, WDM optical networking, and optical packet interconnection for high-performance

272


C. Daniel Lee was born in Chiayi, Taiwan, in 1969. He received the B.S. degree in Physics from National Tsing Hua University, Taiwan, in 1991 and the M.S. and Ph.D. degrees from the University of Maryland, Baltimore County, in 1997 and 2002, respectively. He has worked at the Photonics Packaging Automation group of Newport Corporation, Irvine, CA, in 1997 and later joined the division of Optical Internetworking Research of Telcordia Technology, Red Bank, NJ, in 1997, as a Research Engineer for Next Generation Internet project. Since 2000, he was a Member of Technical Staff in the CTO organization at Tellium, Inc., Oceanport, NJ, where he worked particularly on the design, testing, and system prototyping with a team of scientists and engineers for the development of large port count MEM-based all-optical cross-connect switch. He is currently a Senior Optical Designer responsible for the development of integrated planar lightwave circuits at Inplane Photonics, Inc., South Plainfield, NJ.

David R. Peale received the M.S. and Ph.D. degrees in solid-state experimental physics from Cornell University, Ithaca, NY, in 1992 and the B.S. degree in physics from Dartmouth College, Hanover, NH, in 1983. He is currently a Physicist and Principle Research Scientist for KLA-Tencor, San Diego, CA. The work for this paper was done while he was a Senior Member of Technical Staff at Tellium Inc., Oceanport, NJ, where he worked with a world-class team of scientists and engineers to create a MEMS-based optical switch. He was the first to propose closed-loop servo control of a tilting MEMS mirror beyond the mirror’s electrostatic “snapdown” limit, and led the team designing the electronics and signal processing methods needed for concurrent capacitance-based position sensing in the presence of high voltage actuation. He was also instrumental in improving the electromechanical performance of the MEMS device through improved methods for electrostatic actuation, improved compact torsion spring designs, and improved first-principles understanding of gas damping mechanisms within the mirror cavity. His previous experience includes two years as a Senior Scientist at Phase Metrics, San Diego CA, five years as a Postdoctoral and Member of Technical Staff at Bell Labs, Murray Hill, NJ, and two years as a Postdoctoral at IBM Thomas J. Watson Labs, Yorktown Heights, NY. He has authored numerous papers, and holds 13 patents.

Ming-Ju Tsai was born in Keelung, Taiwan, in 1966. He received the B.S.E.E. degree from National Taiwan University, Taiwan, in 1989 and the M.S.E.E. and Ph.D. degrees from University of California, Los Angeles (UCLA) in 1993 and 1996, respectively. From 1996 to 2000, he was a Member of Technical Staff in Wireless Components and Packaging Research Department of Lucent Technologies, Murray Hill, NJ. He worked on antenna design for WLAN and cellular basestation, and high-frequency packaging design for optoelectronics. From 2000 to 2002, he was a Senior Member of Technical Staff at Tellium, Inc., Oceanport, NJ, and worked on MEMS mirror and servo control design. He joined Tessera, San Jose, CA, in September 2002 as a Senior RF Engineer and currently is working on embedded passives and chip-scale packaging design for RF/microwave applications.

Qi Wu received the Ph.D. degree in applied physics at Rensselaer Polytechnic Institute, Troy, NY, in 1997. From 1997 to 2000, he was a Senior Scientist at the Science and Technology Division of Corning Incorporated, Corning, NY, where he developed photonic technologies for fiber optic communications. From 2000 to 2002, he was engaged in the development of MEMS-based optical cross-connect switches at Tellium, Inc., Oceanport, NJ. He worked on cavity ring down spectroscopy and trace gas sensors at Tiger Optics, Warrington, PA, from 2002 to 2003. He rejoined Corning in 2003 as a Senior Research Scientist. His current research interest includes high throughput screening, label free affinity sensor, and biophotonics. He holds 11 patents. Dr. Wu is a member of the Association for Laboratory Automation (ALA).

William Zhong, photograph and biography not available at the time of publication.

Bo Tang, photograph and biography not available at the time of publication.

Dennis T. K. Tong received the Ph.D. degree from University of California in 1998. He is current an Assistant Professor in the Electrical and Electronic Engineering Department at the Hong Kong University of Science and Technology (HKUST) in Hong Kong. Prior to HKUST, he worked at Tellium, Inc., Oceanport, NJ, as a Senior Member of Technical Staff. His research interest is in optical communication, microwave photonics, and siliconbased integrated photonics.

Evan L. Goldstein (F’04) received the Ph.D. degree in electrical engineering from Columbia University, New York, NY, in 1989, where he also taught academic philosophy. He has since worked at Bellcore, Bell Labs, AT&T Labs, and Tellium, Inc., Oceanport, NJ, where he was Director of Optical Networking Systems until terminated with cause, in 2002, for his association with fiscally unproductive activities involving lightwave micromachines. He is currently an Affiliate Professor of Electrical Engineering at the University of Washington in Seattle. He has coauthored about 120 papers on optical amplification, noise, switching, and architecture in lightwave-communications networks. Dr. Goldstein is a Fellow of the Optical Society of America (OSA).


Lih Y. Lin (M’93–SM’02) received the Ph.D. degree in electrical engineering from University of California, Los Angeles in 1996. In 1996, she joined AT&T Labs-Research, where she conducted research on micromachined technologies for optical switching and lightwave systems. In March 2000, she joined Tellium, Inc. as Director of Optical Technologies. She has been an Associate Professor at the Electrical Engineering Department of University of Washington since June 2003. She has over 120 publications in refereed journals and conferences, holds 21 U.S. patents, and has seven patents pending. Dr. Lin has served on the technical program committee and co-chairs of various technical conferences, including International Optical MEMS Conference, CLEO Pacific Rim, IEEE LEOS Annual Meeting, OSA Annual Meeting, and OSA Photonics in Switching Topical Meeting. She was the editor for Journal of Lightwave Technology: Special Issue on Optical MEMS and its Future Trends. Currently, she is on the steering committee of the International Optical MEMS Conference. She was the Finalist of the 2001 IEEE Eta Kappa Nu Outstanding Young Electrical Engineer Award, and she received the MIT Technology Review 100 Award in 2003.

273

James A. Walker received the B.S. and M.S. degrees in electrical engineering from Rutgers University, New Brunswick, NJ, in 1984 and 1989, respectively. In 1984, he joined Bell Labs, where he was a Member of Technical Staff and rose to the position of Technical Manager of the MEMS Network Element Subsystems Group in Bell Labs Research, Holmdel, NJ. In 2000, he became the Director of Advanced Technologies at Tellium, Inc., Oceanport, NJ, where he jointly led an effort to develop all-optical MEMS-based cross-connect switching systems. He now heads a team of MEMS and telecommunications consultants as the President of JayWalker Technical Consulting, LLC, Freehold, NJ. He is also a registered Patent Agent with the U.S. Patent and Trademark Office, and provides patent prosecution expertise in the areas of MEMS, Nanotechnology, Optical Telecom, and Opto-mechanical Packaging. He holds 53 U.S. patents, has authored over 100 journal articles and conference publications, presented numerous invited and submitted conference papers in the areas of photonics and MEMS, and chaired several international conferences in the area of optical MEMS technology.