Electromagnetic MEMS drive large-scale crossconnects
A myriad of actuation mechanisms are feasible using microelectromechanical systems (MEMS), from thermal to electrostatic to thermal-fluidic. Electromagnetic actuation in MEMS optical components remains relatively underused although there are significant benefits. Some approaches use a global magnetic actuation over an array of two-dimensional (2-D) mirrors coupled with individual electrostatic actuators to select or deselect specific mirrors in the array. For large-port-count optical crossconnects, three-dimensional (3-D) mirrors are a more preferable architecture, minimizing the number of mirrors and loosening the mechanical alignment tolerances of the lenses and mirrors normally required in a 2-D array.
The electromagnetic drives in electromagnetically actuated 3-D mirrors provide relatively high forces that enable large low-loss mirrors with robust suspensions. The low drive voltages (typically less than 2 V) allow for integrated local control electronics via an ASIC. The magnetic design allows the integration of inductive sensors for mirror position and full three-degree-of-freedom closed-loop control in tip, tilt, and piston modes. These local sensors work autonomously with the local ASIC to control the mirror motion, freeing the main CPU to direct switching and global control. Such a system can compensate for any disturbances induced from vibration, shock, or long-term drift. Switch times of less than 10 ms are possible with typical connector-to-connector losses of 1.5 dB.
The fundamental actuation mechanism for the 3-D mirror is similar to that of a permanent magnet DC motor without commutators and bearings. A permanent magnet provides a static magnetic field, B, with a relatively high flux density. When a current is sent through a winding coil, a local magnetic moment, m, is generated normal to the plane of the coil. The resulting torque, T, will tend to align m to B, via:
The torque is essentially the sum of the differential torques generated by the Lorentz force along each elemental section, dl, of the coil. This differential torque is the cross product of the position vector with respect to the center of rotation, r, and the Lorentz force, F:
where F = I × B, and I represents the current vector in the coil. As in a conventional DC motor, an increase in the current will induce a proportional increase in the generated torque. Similarly, a relatively strong magnetic field, B, will require a smaller amount of current, reducing power consumption.
In the case of a single-axis actuator, a magnetic field with a uniform parallel field is the most beneficial. Current passing through a single coil will cause the device to tip about the axis m × B. In theory, this can be achieved with two magnets of opposite polarity facing each other with the device in the air gap. In practice, this hinders an effective package design or interferes with other optical components in the system. An alternate approximation to this uniaxial field is the use of a dual-poled magnet (see Fig. 1).
The transition of this basic drive concept to a multi-axis mirror begins with the static magnetic field. Whereas a unidirectional field is well suited to a single-axis actuator, a multi-axis device would require field lines that reversed or changed direction. In this case, one cylindrical magnet mounted beneath the device provides a radial field at its periphery. With the proper design of multiple coils, the Lorentz forces generated at the outer circumferential arcs of the coils will generate tip and tilt. The radial segments of the coil contribute little to the device actuation.
For example, a counterclockwise current in the north coil will move the north portion of the mirror downward and a simultaneous clockwise current in the south coil will push that end of the mirror upward, causing the mirror to tip about the east-west axis. Similarly, the east and west coils induce a tilt about the north-south axis.
An additional degree of freedom, critical to the pointing stability of the mirror, is the "piston" mode-the independent actuation of the device in the vertical direction. A clockwise current sent through all four quadrants simultaneously moves the mirror upward. Conversely, a counterclockwise current will pull the mirror downward. Superimposing the drive currents for tip, tilt, and piston to the four independent coils allows full three-degree-of-freedom control of the mirror. Other designs incorporate three independent coils to function similarly.
Placing the magnets beneath each device also allows the magnetic fields to interact constructively, further reducing the drive currents. For example, with the magnets arranged in a rectilinear checkerboard array with alternating north- and south-pole faces, the radial component of the B field in the cardinal directions will increase. This creates higher field lobes at the working portions of the coil segments. Other designs use additional ferromagnetic yokes or hexagonal packing patterns to tailor the field pattern, which with specifically designed coil geometries minimizes power consumption.
The relatively high forces obtainable with this electromagnetic drive enable designs with large mirrors and robust suspensions. The mirror diameter is intentionally oversized to 2 mm or more. This loosens the optical subassembly alignment tolerances and eliminates losses due to beam clipping. It also provides sufficient area, on the obverse side, for the drive-coil layout.
The mirror is fabricated from single-crystal silicon of considerable thickness to retain an rms (root-mean-square) surface flatness of less than 15 nm across the diameter. A dry-etch process defines the suspension from an epitaxial layer. The width of the flexures allows adequate space for routing of the drive coils to the frame, yet is compact enough to accommodate a high fill-factor in an array, which maximizes the mirror diameter-to-pitch ratio.
Various designs have been implemented, but each has common properties. For example, the higher undesirable mode shapes, such as in-plane motion or rotation about the axis normal to the mirror ("flying disk" motion), are kept higher than 3 kHz. Because these modes are neither sensed nor is the actuator configured to act in those degrees of freedom, these modes must be high enough to allow stable closed-loop control of the mirror. Generally, the high-piston-mode constraint (consistent with conventional gimbaled-type suspensions) is relaxed. Because the drive mechanism can push and pull (as opposed to pull only), and the closed-loop feedback control uses position sensing, the piston mode is "electronically stiffened." In other words, the suspension is aided by a virtual magnetic-levitation spring.
The drive coils occupy nearly the entire moving surface of the opposing face of the mirror. The MEMS die is flip-chip bonded to a ceramic substrate with tin-lead solder balls acting as both mechanical spacers and electrical interconnects (see photo, above). The top surface of the ceramic substrate is fabricated with thick-film RF transmitters that work as part of the local position-sensing scheme. An RF current passed through the transmitter produces an RF voltage drop on each of the mirror's four drive coils. As the mirror moves, the mutual inductance changes between each of the drive coils and the transmitter coil. Taking the difference between the RF voltage magnitudes on the north and south coils produces a signal proportional to the tip angle of the mirror. This difference is divided by the sum of the signals to normalize the signal and reduce any sensitivity to changes in the transmitter output strength.
Tilt is sensed similarly from the delta between the east and west RF voltages. The third-degree-of-freedom sensor signal for the piston mode is proportional to the sum of all four quadrant's voltage drops. Using electromagnetically actuated mirrors, the drive coils function both as the actuators with low-frequency drive currents and as position sensors with high-frequency signals for collocated sensors.
This sensing scheme eliminates sensor crosstalk within the array. The motion of adjacent mirrors does not influence the high-frequency magnetic field produced by the stationary transmitters. As a test for sensor crosstalk, a primary mirror is commanded to hold a signal where nonoptimal coupling is achieved, for example, at -4.5 dB. The neighboring mirrors are then actuated to ±4°. The maximum observed change in the optical signal for the primary mirror is -0.02 dB, corresponding to a sensor signal change of less than 100 µ°. This small crosstalk effect is further reduced when the coupling of the primary mirror is optimal (at 1 to 1.5 dB) because the coupling at that point is less sensitive to change in position.
Fundamental to the crossconnect design is the use of an ASIC to implement local closed-loop position control for each mirror. Each linear array of integrated mirrors is mounted to an electronics board with all the local control electronics and ASICs. The cards are then stacked to form a 2-D array of the mirrors. Each mirror within the array has a unique digital address, which greatly simplifies wiring issues for large arrays. The local loop can sense and compensate for disturbances induced from vibration, shock, or any drift. The main CPU is then free to occupy itself only with switching and monitoring global feedback via optical power monitoring (see Fig. 2).
The local feedback controller takes the digital command for mirror position from the CPU and converts that to appropriate drive currents for the four engine coils. The local feedback loop allows for quick slewing and vibration damping. Switch times of less than 10 ms are needed to reach peak coupling, with 3 dB achieved in less than half that time (see Fig. 3).
Viktoria Temesváry-Díaz is principal design engineer at Integrated Micromachines, 1400 South Shamrock Ave., Monrovia, CA 91016. She can be reached at email@example.com.