Prototype feedback-controlled bidirectional actuation ... - IEEE Xplore

4 downloads 7 Views 455KB Size Report
Actuation System for MEMS Applications. Shekhar Bhansali, Member, IEEE, Andy Lei Zhang, Ronald B. Zmood, Member, IEEE, Paul E. Jones, Member, IEEE,.



Prototype Feedback-Controlled Bidirectional Actuation System for MEMS Applications Shekhar Bhansali, Member, IEEE, Andy Lei Zhang, Ronald B. Zmood, Member, IEEE, Paul E. Jones, Member, IEEE, and Dinesh K. Sood

Abstract—We have successfully developed a one-degree-of-freedom microsuspension system, with active position control, as a paradigm of a micromagnetic bearing. This system integrates an electromagnetic actuator, a position sensor, and a feedback control system that provides active position control. This paper discusses the design and fabrication details of the microelectromechanical system (MEMS) components: the beam mass structure integrated with a drive coil and metallized targets, spacer plate, and sensor coils. It also discusses their integration with millimagnets and electronics. Noncontact magnetic bearings based on this principle have the potential of overcoming the tribo-physical issues associated with active MEMS devices. [450] Index Terms—Beam mass structure, bidirectional actuation and control, electroplating, etching, feedback control, inductive sensor, magnetic actuator, magnetic microsuspension system, micromagnetic bearing, position sensor.



RICTION and wear of components have been a major impediment in the development of active microelectromechanical systems (MEMS) devices like micromotors. The coefficient of dynamic friction between the small smooth surfaces of the micromachined components has been estimated to be between 0.2 to 0.5 [1], a value similar to the coefficient of static friction between brake pads and cast iron. The effort to minimize friction in MEMS components is as old as MEMS itself. One of the early ideas on overcoming friction and wear has been to develop side-drive actuation mechanisms [2, pp. 168–173]. Studies on friction and wear in bearings of the early side-drive motors confirmed significant wear even in these designs. These effects resulted in the gear ratios of the motor changing by as much as 20% toward the end of the test cycle. Magnetically suspended levitated noncontact devices are highly desirable for active MEMS components, as they overcome the tribo-physical issues associated with friction and wear. Unlike superconducting actuators that use the Meissner Manuscript received May 19, 1999; revised December 20, 1999. Subject Editor, R. T. Howe. S. Bhansali is with the Center for Microelectronic Sensors and Microelectromechanical Systems, Department of Electrical and Computer Engineering and Computer Science, University of Cincinnati, Cincinnati OH 45221-0030 USA. A. L. Zhang is with the Cooperative Research Centre for Cochlear Implants, Speech and Hearing Research, East Melbourne, Vic. 3002, Australia. R. B. Zmood and P. E. Jones are with the Department of Electrical Engineering, Royal Melbourne Institute of Technology, Melbourne, Vic. 3000, Australia. D. K. Sood is with the Department of Electronic and Communication Engineering, Royal Melbourne Institute of Technology, Melbourne, Vic. 3000, Australia. Publisher Item Identifier S 1057-7157(00)04862-9.

effect for levitation [2, pp. 314–319] and produce a downward force, the electromagnetically actuated devices and systems have the advantage that the levitated structure can be actively controlled in both upward and downward directions. In this paper, we discuss the fabrication and actuation of the one-degree-of-freedom (1-DOF) actuator, which has been developed at the Royal Melbourne Institute of Technology (RMIT), Melbourne, Vic., Australia. The 1-DOF actuator is a paradigm of a micromagnetic bearing, as it exhibits the main attributes of a five-degree-of-freedom (5-DOF) magnetic bearing and includes the key elements of the electromagnetic actuator, position sensor, and feedback control system. II. PRINCIPLE OF OPERATION Fig. 1 schematically illustrates the MEMS component of the 1-DOF microsuspension system. It is comprised of three subsystems, namely, a platform plate, sensor plate, and feedback control system. Micromachined components in this system are placed on the platform plate and the sensor plate, both of which are made from h100i Si. On the platform plate, the drive coil is electroplated around a highly compliant beam-mass structure, which is designed to allow large deflections along the vertical direction and to constrain movement in all other directions. A bulk NdFeB permanent magnet is mounted on the top of the proof mass. Additionally, two identical thin metallic structures are fabricated on the bottom of the wafer. These metallic structures act as sensing targets for the inductive sensing coils. Two planar coils are fabricated on the sensor plate. They work as the reference and sensor coils, respectively. Glass spacer plates of different thickness (30–50 m) are used to control the gap between the targets and sensor coils. Fig. 2 illustrates the block diagram of the principle of operation of the 1-DOF microsuspension system. It operates along a line similar to that of a conventional magnetic bearing. The basic principle is: 1) a shift in the position of the magnet is detected by a noncontact inductive position sensor and 2) the signal from the position sensor is fed to the signal conditioning circuit, which applies a controlling current to the drive coil to restore the position of the magnet to the desired position. The system can also work in a force balance mode, where the control current to the drive coil interacting with the field of the permanent magnet generates a force to balance the applied force. When the external force is applied to the platform, its motion is detected by the position sensor and fed to the signal conditioning and control circuit. The signal conditioning and control circuit regulates the movement of the magnet and restores it to its original position.

1057–7157/00$10.00 © 2000 IEEE



Fig. 1. Schematic illustration of the MEMS components of the 1-DOF system.

Fig. 2. Block diagram for the principle of operation of the 1-DOF microsuspension system.

III. STRUCTURAL DETAIL The design concept of the platform plate with a moving magnet is based on the earlier work of Wagner and Benecke [3]. However, there are major differences in our design. The two most important differences are: 1) longer folded cantilever beams and 2) metallized sensor targets. The introduction of the metallic sensor targets increases the processing difficulties many fold. Fig. 3 schematically illustrates the cross-sectional view of the platform plate. An NdFeB permanent magnet with vertical magnetization Mz is mounted on the silicon boss structure suspended by four meander-shaped silicon beams. A planar gold coil is electroplated around the folded silicon beam mass structure. Gold was chosen as the preferred material for the coils because it has an electrical conductivity comparable to Cu, is chemically inert (does not oxidize in atmosphere), has excellent thermal conductivity, and is easy to wire bond. Its excellent thermal conductivity ensures uniform heat flow from the coil to the underlying Si substrate. The metallic targets for position sensing are made of Ni, a soft magnetic material. The moving metallic target is placed on the backside of the proof

mass, while the stationary target is attached directly to the platform plate. This stationary target is positioned sufficiently far from the beam mass structure to ensure that it is not influenced by the platform motion. The position sensor is a noncontact inductive displacement transducer, in which the sensor coil is placed beneath the moving target and the reference coil is positioned facing the stationary target. Thus, the motion of the platform in the vertical direction is detected by comparing the impedance change of the sensor coil with that of the reference coil using a hybrid signal conditioning circuit. The block diagram of the noncontact inductive position transducer, shown in Fig. 4, includes the reference and sensor coils Lr and Lx , as well as the balanced differential bridge signal conditioning circuit. The presence of the coaxial cable stray capacitances and circuit capacitances is indicated in Fig. 4 by Cs and Ci . The high-frequency oscillator excites the sensor coils, which are linked to the inputs of a pair of trans-impedance amplifiers by coaxial cables. It also supplies the carrier signal, whose phase is shifted by 90 , to the synchronous demodulator to improve the signal conditioner sensitivity. This forms a phase sensitive detector (PSD) with output low-pass filters to remove the unwanted high-frequency demodulation signal components. The output amplifier is used to amplify the low-frequency signals and for impedance matching. As expected, the impedance variations of sensor coil in micromachine applications are very small. This results in the transducer having low sensitivity. One approach to alleviating this problem is to increase the operating frequency of the signal conditioner as the coil dimensions are reduced. However, in this case, the parasitic capacitances from the connecting cables and signal conditioning circuitry will then begin to influence the operation of



Fig. 3. Cross-sectional view of the silicon platform plate for the 1-DOF microsuspension system.

Fig. 4. Block diagram of the noncontact position transducer.

the transducer. A promising approach, which minimizes the effects of stray capacitances as well as the effect of the circuit input capacitances, is to use trans-impedance amplifiers having ultra-low input impedance and a low output impedances. This approach results in the capacitances CS being essentially bypassed by the input impedance of the trans-impedance amplifier. The input capacitance Ci of the signal conditioner also becomes of little consequence, as it is being driven by the output voltage source of the trans-impedance amplifiers. This design concept has been shown to be robust in applications and appears to have considerable scope for further development. The output of the position transducer is compared with the input position command signal in the signal conditioning circuit. The error signal is then fed to a proportional plus derivative (PD) controller that controls a power amplifier that supplies current to the actuator drive coil and, thus, controls the position of the magnetic platform. Our design objective was to achieve large static deflection of several tens of micrometers in the z -direction while restraining deflections along the other degrees of freedom. To achieve this objective, we designed a silicon spring configuration with suspending beams parallel to the frame. The outer dimension of the suspension structure was 2360 m 2 2360 m. The proof mass was designed to be 1100 m 2 1100 m (sufficient to accommodate a 800 m 2 800 m 2 1200 m permanent

magnet). The beam structure with four L-shaped beams was also designed to increase the mean beam length. The beams were 200-m wide. The gaps between the frame and beams, between the beams, and between the beam and the platform were all chosen to be 100 m. Corner compensation structures were incorporated in the etching masks for etching both the beams and backside features to ensure satisfactory fabrication. The inner track of the planar drive coil is at a distance of 70 m from the edge of the beams and has 15 turns. The coil tracks are 25-m wide with distances between them of 10 m. The coil is typically 12-m thick. This coil generates its maximum field in the region of the magnet. The planar coils on the sensor plate have 17 turns. These coils are typically 11-m thick and the track width and spacing is 10 m. The assembled structure (with magnet) was designed to have a high resonant frequency to ensure that, at operating conditions, resonance effects were eliminated. The theoretical resonance value for this structure was estimated to be around 250 Hz. Due to the manufacturing tolerances, there are variations in the sizes of the beams and magnet, and it has been estimated that the resonant frequency can change substantially (10%) from device to device. The beams were designed to have maximum working stress at a deflection of 150 m. IV. FABRICATION Fig. 5(a) schematically illustrates the fabrication process for the beam mass structure. Fabrication of the beam mass structure was carried out on commercially available 100-mm doublesided polished (100) epi-wafers with a buried p+ layer. The handle wafer is 540-m thick, on which a 20-m-thick epi-layer is grown. A 7-m-thick p+ layer is sandwiched between the epi-layer and handle wafer, and is used as an etch-stop layer for the EDP etching processes. Thermal oxidation of the epi-wafer is carried out to form a 400-nm-thick SiO2 layer on both surfaces of the epi-wafer. The back etching mask is first patterned onto the backside of the wafer. EDP etching is performed on the backside and it terminates at the p+ etch-stop layer. Next, the beam structure





(c) Fig. 5. (a) Schematic of the platform plate fabrication process: a) clean wafer, b) oxidize wafer, c) pattern backside oxide and back etch wafer, d) pattern front oxide and etch beams, e) spin resist, metallize and liftoff back side to create targets, f) deposit seed layer, spin expose and develop resist, electroplate coils, strip resist and etch seed layer, g) release beams. (b) Photograph of the top side of the micromachined platform plates. The beams, beam-mass structure and the gold coils are visible. (c) Photograph of the metallized target on the moving target.

on the front is patterned and etched in EDP. Again, the p+ layer terminates the etching process and the beam structure is held together by the p+ layer.

Next, the gold coils are electroplated on the front side of the wafer using Ti/Ni as a seed layer. AZ 4562 photo resist is patterned to act as a mould. A commercial cyanide-based plating solution with neutral pH was used for plating the soft gold coils. Soft Au was preferred to standard hard gold conductors, as it is easy to wire bond to it. After plating, the resist is stripped off and the seed layer etched. The metallic targets on the backside of the wafer are formed using a metallization and etching process. A 300-Å Ti adhesion layer and a 2000-Å Ni layer is sputtered on the backside of the wafer. The resist is then patterned using a micromanipulated dispenser to apply a measured quantity of photo resist (AZ 4620) to the backside of the platform structure (moving target) and the corresponding reference target area. The resist is then cured and the exposed Ti/Ni layer is etched. The beam mass structure is released by an isotropic etching step using a standard HNA etching solution, consisting of HF, HNO3 , and CH3 COOH (10 : 30 : 80 in volume). HNA has an etch selectivity of 150 for p+ silicon (>1018 cm03 ) over lightly doped Si at room temperature. Fig. 5(b) is a photograph of the top side of the platform plates after etching and plating. The photograph shows the three different beam mass structures that were made. Fig. 5(c) is the photograph of the backside of a typical mesa structure. The bright region on the top of the mesa structure is the metallized target (target for the sensing coil). The magnet is now mounted on the silicon platform fabricated on these plates. Two approaches have been used for positioning and attaching the small magnet to the moving platform. In the first approach, a small magnet was glued to a movable substrate, which we will term a “magnet manoeuvring plate,” and this plate was manipulated below the platform. In the magnet mounting procedure on the platform, the magnet was manoeuvred into the correct position on the platform using the magnet manoeuvring plate before applying a “superglue” adhesive. In the second approach, the magnet was held onto a nonmagnetic vacuum chuck attached to a micromanipulator. The micromanipulator was used to position the magnet over the platform before being bonded to its surface using an epoxy resin adhesive. As this method has been found to be quite flexible and reliable, it is currently the preferred approach for work of this kind. Both Si and glass were studied as potential substrate material for the sensor coils. Si was preferred to glass as the substrate for the sensor coils because of its higher thermal conductivity. Higher thermal conductivity of the substrate resulted in better heat dissipation from the coils and improved performance (because of lower temperature-related impedance drifts). Fig. 6(a) schematically illustrates the sensor coil plating process. The sensor coils are plated on a separate Si wafer. After the deposition of the seed layer, photoresist was spun on the wafer and patterned. The sensor coils were then electroplated using the gold electroplating process. A thieving ring was incorporated around the coils, in the mask, to ensure that the coil tracks are of uniform thickness. The thieving rings draw the stronger edge current toward them and ensure a uniform current density across the coil windings being electroplated. To fabricate planar interconnects, the coils were planarized with Hitachi PIX 3400, an insulating polyimide. The polyimide





Fig. 6. (a) Schematic of the sensor plate fabrication process: (A) Clean wafer and deposit electroplating seed layer. (B) Spin and expose photoresist. (C) Develop photoresist and electroplate. (D) Strip resist and etch seed layer. (E) Spin polyimide, spin, and expose resist. (F) Develop resist, etch polyimide, strip resist, and cure polyimide (to get tapered sidewalls). (G) Deposit seed layer, spin, expose, and develop resist. (H) Electroplate via and interconnect, strip resist, and etch seed layer. (b) Photograph of the electroplated sensor coil with planar interconnect.

was patterned to open the vias and bonding pads. A seed layer was deposited, patterned, and electroplated to fabricate the planar interconnects and complete the electrical circuit. Fig. 6(b) is the photograph of one of the sensor coils with planar interconnects. The array of sensor coils is diced in pairs and for assembly. The metallic lines around the set of coils in Fig. 7 are the electroplated thieving rings. The spacers are made of borosilicate glass. The spacers are diced from a substrate that has been thinned to between 30–50 m, using buffered HF. They are positioned and bonded with silver epoxy resin onto the sensor plate. The platform plate, with the NdFeB magnet, is now aligned to the sensor plate to ensure that target areas were directly above the sensor coils. The platform plate is then silver epoxied in place. Fig. 7 shows the photograph of a pair of completed platform and sensor plates before component assembly. Each of the platforms was assembled in an exactly similar fashion. The thieving rings, plated around the sensor coils, to ensure uniformly thickness of the coil tracks are clearly visible. Fig. 8 shows a photograph of the assembled 1-DOF microsuspension system under test. The dark piece in the center (on the top) is the magnet,

Fig. 7. Photograph of sensor and platform plates, diced accurately for alignment, before assembly. The electroplated thieving rings on the sensor plate are visible.

while the light rectangular region around it is the control coil winding, which is not clearly visible. The meandering light regions between the coil and magnet is empty space between the



Fig. 8. Photograph of the fully assembled and bonded 1-DOF actuator, with magnet on top, under testing. The system is connected to the feedback control circuit and tested on an antivibration table.


beams (the dark region sandwiched between the light regions) that provide a mechanical support for the platform. V. PERFORMANCE As discussed earlier, our 1-DOF system also includes a feedback control system incorporating a proportional-plus-derivative controller. In this controller, the position sensor output is compared with the input position command signal, and the error signal is fed to a power amplifier stage. This stage supplies the current to the drive coils. The micromachined structure is mounted on the alumina substrate carrying the signal conditioning circuit. The amplifier and sensor circuits are wire bonded to the drive and sensor coils, respectively. Fig. 8 shows the experimental setup with the device pads bonded to the electronics. The system is being tested on an antivibration table to ensure isolation from stray vibrations. It has been found that the magnet and flexible beam structure can be elevated and maintained at a predetermined level by a PD control system. We were successful in arbitrarily fixing the position of the magnet (and the platform) in a range of 620 m around the central position. The damping and response time of the system can be varied over a wide range by adjusting the PD controller coefficients. Fig. 9(a) shows an underdamped response of the platform position sensor (Ch1) to a square wave command signal (Ch2) for a set of typical controller coefficients. The total optically measured deflection of the platform in this case is 10 m and the output from the platform position sensor is 100 mV (102 probe on CRO). Therefore, the sensitivity is seen to be 10 mV/m. Fig. 9(b) shows the response of the platform position sensor (Ch1) and drive coil voltage (Ch2) of the closed loop control system for a similar step input command, as given in Fig. 9(a). The transient waveforms of the step-input voltage, the drive coil voltage, and the platform position for typical system parameter values have been studied. VI. CONCLUSION We have successfully designed, fabricated, and tested the 1-DOF micromachine with an active z –direction control. The

(b) Fig. 9. Transient performance of the signal conditioning circuit. (a) Response of the platform position (Ch1) to the command signal (Ch2). (b) Response of the platform position (Ch1) to the drive coil voltage (Ch2).

design, fabrication, and the study of the operation of this 1-DOF system demonstrates that it is feasible to actively control the position and amount of vertical displacement of a magnet over a range of 620 m. These results provide a basis for the development of a range of bidirectional and multidirectional actuators. The results are being used for the development of a 5-DOF micromagnetic bearing, which is currently under fabrication.

ACKNOWLEDGMENT This work was conducted in the Micromachine Technology Program consigned to the Micromachine Center from the New Energy and Industrial Technology Development Organization, which is carried out under the Industrial Science and Technology Frontier Program enforced by the Agency of Industrial Science and Technology, Ministry of International Trade and Industry, Japan. The authors would like to thank Prof. M. Parameswaram, Simon Fraser University, Burnaby, B.C., Canada, for his help during fabrication of the device and P. L. Yu for his help in the development of the proportional plus derivative (P-D) control for the noncontact position sensor.


REFERENCES [1] Y. C. Tai and R. S. Mueller, “Frictional study of IC-processed micromotors,” in Proc. 5th Int. Solid-State Sens. Actuators, Eurosensors III Conf., Montreux, Switzerland, June 1989, pp. 108–110. [2] W. S. Trimmer, Ed., Micromechanics and MEMS: Classic and Seminal Papers to 1990. Piscataway, NJ: IEEE Press, 1996. [3] B. Wagner and W. Benecke, “Microfabricated actuator with moving permanent magnet,” in Proc. IEEE MEMS Nara, Japan, 1991, pp. 27–32. [4] P. L. Yu, C. Y. Zhang, and R. B. Zmood, “Wide-band low-input-impedence amplifiers for instrumentation applications,” Meas. Sci. Technol., vol. 8, no. 11, pp. 1351–1355, 1997.

Shekhar Bhansali (M’98) received the B.E. degree in metallurgical engineering (with honors) from the Malaviya Regional Engineering College (MREC), Jaipur, India, in 1987, the M.Tech. degree in aircraft production technology from the Indian Institute of Technology (IIT), Kanpur, India, in 1991, and the Ph.D. degree in electrical engineering from the Royal Melbourne Institute of Technology (RMIT), Melbourne, Vic., Australia, in 1997. He is currently Research Assistant Professor at the Center for Microelectronic Sensors and MEMS, Department of Electrical and Computer Engineering and Computer Science, University of Cincinnati, Cincinnati, OH. His professional experience includes aeronautics (Hindustan Aeronautics Limited 1988–1992), teaching (RMIT, 1996) and research (STA Fellow, National Research Laboratory of Metrology Japan, 1996–1998). His research interests are in the area of development and characterization of materials and technologies for MEMS components and systems, microfluidic “labs on a chip,” bioengineering applications of MEMS devices, microscale heat transfer, and system integration.

Andy Lei Zhang received the B.S. degree in semiconductor physics and microelectronics from Fudan University, Shanghai, China, in 1988, and the M.Eng. degree from the Royal Melbourne Institute of Technology, Melbourne, Vic., Australia, in 1996. Following graduation, he was a Design Engineer at the Information Research Centre, Hudong Shipyard, Shanghai, China. He is currently a Research Fellow and Project Officer at the Cooperative Research Center for Cochlear Implant and Hearing Aid Innovation, East Melbourne, Vic., Australia. His current research interests focus on cochlear implants, hearing aids and implantable transducers.


Ronald B. Zmood (M’74) was born in Melbourne, Australia, in 1942. He received the B.S. degree in electrical engineering and the Master of Engineering Science degree from the University of Melbourne, Melbourne, Vic., Australia, in 1963 and 1967, respectively, and the Doctor of Philosophy degree from The University of Michigan at Ann Arbor, in 1971. His professional and research experience has included working at Telecom Research Laboratories (formerly APO Research Laboratories), where he was involved with adaptive echo cancellers for satellite communication systems, and Aeronautical Research Laboratories, where he was involved with the computer simulation of aircraft and guided missiles, as well as manned aircraft simulators. Over the last 20 years, he has consulted widely to Australian industry on applications of control systems. Since joining the Department of Electrical Engineering, Royal Melbourne Institute of Technology (RMIT), in 1980, he has led the teaching of automatic control systems. Since 1992, he has been instrumental in developing micromachine research at RMIT, and is currently the Leader of the Micro Machine Research Group. He has recently built up a Micro Machining Laboratory, which is currently conducting research on a number of MEMS devices for medical and industrial applications. His research interests are in the fields of electromagnetic actuators and other devices, flywheel energy storage, power electronics and nonlinear control. His interests in control are mainly in the application of control techniques to magnetic bearings and to microelectromagnetic systems.

Paul E. Jones (M’89) received the Associate diploma in electronics engineering from the Royal Melbourne Institute of Technology (RMIT), Melbourne, Vic., Australia. He has been involved in the area of MEMS for the last six years, and is currently a Technician involved on the NEDO Micromachine Project at RMIT.

Dinesh K. Sood received the Ph.D. degree in physics from the Indian Institute of Technology, Kanpur, India, in 1969. He is currently a Professor of materials science and engineering at the Royal Melbourne Institute of Technology (RMIT), Melbourne, Vic., Australia. He has held research and teaching positions at Bhabha Atomic Research Centre, Bombay, India, Atomic Energy Research Establishment, Harwell, U.K., IBM Research, Yorktown Heights, NY, and the University of Padua, Padua, Italy. He has authored or co-authored over 100 research papers in international journals, and he holds a few patents. He has supervised several students for Masters and Ph.D. degrees. His research is focused on advanced materials processing, development, and characterization. He is currently engaged in interdisciplinary research and development in diverse areas, such as fabrication, processing, and materials development for micromachines and MEMS, and protective coatings suitable for use in high-temperature environment of solid oxide fuel cells. His research is in close collaboration with industry in Australia and overseas. Dr. Sood is a Fellow of the Institution of Engineers, Australia, a member of the Bohmisch Physical Society, Materials Research Society, and the Institute of Metals and Materials Australia. He is an Australian representative member of the Asia–Pacific Society for Advanced Materials.

Suggest Documents