Department of Mechanical Engineering Northwestern UniversityEvanston, IL 60208
Project Summary This paper explores the growing segments for potential use of MEMS devices. In particular, we focus on space-based applications in which MEMS have yet to play a significant role. If the performance of particular MEMS devices can match those of macro-sized components, the same system objectives can be achieved while significantly reducing the overall weight – usually the driving factor in spacecraft design. Of course, modifications in spacecraft architecture will hopefully evolve as MEMS components prove their performance in the space environment. Examples are given for future distributed satellite missions showing that a greater quantity of smaller satellites with individualized capabilities can help reduce production cost as well as add robustness to the entire mission. Control of position and velocity are natural requirements of the distributed architecture. In order to trade sensor information effectively, remote communication must play a key role. Therefore, the remainder of this report focuses on MEMS RF communication devices. Examples of phase shifters, signal filters, switches and antennas are given. The modeled performance of the switch and antenna are derived as well as the processes used for fabrication. Finally, an example is given of a MEMS RF-switch tested in orbit. The results indicate that MEMS structures can be cheaper to place in orbit and achieve improved performance over macro-sized devices.Keywords: MEMS, RF Communication, satellites, space-based communication systems, MEMS fabrication processes, MEMS modeling
Introduction When NASA Administrator Daniel Goldin introduced the new catch phrase, "Faster, Better, Cheaper" he may not have envisioned where spacecraft technology would head over the next decade. The defining parameter in most spacecraft is payload weight for launching from the earth's surface to outer space [1, 2]. As expected there is a direct relationship between launch cost and vehicle weight. If the overall system weight can be lowered, or distributed, the potential for reducing satellite launching costs can be significantly improved. Naturally, the development of Micro-Electro-Mechanical Systems (MEMS) has expanded into the space sector. By cutting the mass of components onboard the space vehicle, the launch costs and hopefully the overall budget for production can be reduced. Furthermore, other features of MEMS devices, aside from their small size, can benefit system components, such as power. By converting from solid-state electronics to mechanical systems, the power consumption of a device can be significantly lowered. This increase in efficiency could help reduce the battery power, size, and charge necessary to operate the satellite. Similarly, solar panel sizes could be smaller again bringing down the overall mass and power consumption of the satellite. Two key advantages arise when considering MEMS devices in space applications. The first advantage is realized by lowering launch cost. Currently, launching a spacecraft into low Earth orbit (LEO) costs about $10,000 per kilogram, and placing a craft into a higher geosynchronous Earth orbit (GEO) costs about $50,000 per kilogram . Obviously, by reducing mass, designers stand to gain in reducing total project cost. The second advantage is the devices' resistance to radiation and vibration. Cosmic radiation can upset the operation of solid state components, but MEMS structures can withstand radiation. In addition, because MEMS devices have such a low mass, potentially damaging inertial and vibration forces are minimized. A rocket launch can induce very high accelerations, but with a robust MEMS design the possibility of damage is low . Nanosatellites, roughly classified as satellites weighing 1-10 kg, are some of the most promising spacecraft being designed today. Below 1 kg, satellites are classified as picosatellites. MEMS can revolutionize their design when applied to the communications, data processing, navigation, and propulsion systems.
Satellite Missions and MEMS: Past, Present and Future One must consider what scientific objectives could be accomplished on such a small platform. One of the driving factors of programs such as NASA's New Millennium Program (NMP) is distributing satellite capability and costs across separate vehicles rather than placing all focus on a single, monolithic device. In shrinking the size, the cost of replacing a damaged or failed component can be drastically altered. Budgets for failed single satellite missions have cost billions of dollars without any capability for rectifying the problem. Two missions that can be compared and contrasted are the Hubble Space Telescope, and the Chandra X-Ray Observatory . Hubble's original failures due to an out of focus lens have been well documented. Fortunately, with a significant investment in time and manpower, the problem was fixed thereby allowing the space telescope to operate with its original purpose. It is interesting to note that the entire Chandra X-Ray Observatory Satellite project was nearly cancelled due to worries of similar problems experienced with Hubble. With an orbital perigee of approximately 6,000 statute miles, Chandra is well beyond the orbital range that can be reached by the Space Shuttle. Conversely, Hubble, with a closer orbit of only 350 miles could be reached and repaired by a human crew. Any component failure aboard Chandra, from mirror lenses having a speck of dust or a failed switch on the camera, could have caused the $1.5 billion project to be a complete failure without any means to rectify. If either of these satellite systems could have been distributed over a network of smaller satellites, the potential repair (and possibly production) costs could be more reasonable for a successful space mission. Hopefully with the added robustness of distributed satellite systems, NASA and other space agencies can avoid budget catastrophes such as those experienced with the Mars Surveyor in 1999. One outcome from initial NMP research has shown that Formation Flying of multiple satellites can help distribute the single satellite payload capability. Many future NASA and military missions are basing their design on such capabilities, for example the Terrestrial Planet Finder  and TechSat 21  shown in figure 1. By expanding the number of satellites in a formation to more than one, the total aperture size can be increased allowing for improved resolution without adding cost for connecting the components with structural hardware. Without a rigid connection, the relative motions of payload components must be maintained with tight navigation and control tolerances. Previous missions such as Earth Observing 1 have shown the capability to maintain relative orbital motion or to conduct autonomous maneuvers with standard solid state electronic hardware and macro-size propulsion and control systems . These types of proof of concept experiments are even planned for microsatellite missions. Future NASA missions like Space Technology 5 & 6 (figure 2) hope to be one of the first satellites to use actual MEMS devices in design and production . The Air Force and DARPA are supporting expansion into miniaturization with their University Nanosatellite Program . While some microsatellites will use standard components, with one component necessary for the entire mission payload, the overall size of these satellites will begin to decrease in size, making MEMS more important features for spacecraft production.
Satellite Systems System components can roughly be segregated into three major categories: propulsion, navigation and communication. Micro-thrusters have been explored to help reduce the size of propulsion components on the vehicle making attitude control cheaper by using smaller devices and less fuel. Furthermore, navigational aids such as GPS receivers and gyroscopes are able to reduce the overall size of the navigation payload with MEMS technology. Another important aspect that must be included in all motion control is feedback from sensors. Naturally on a monolithic satellite, each component will be directly hardwired to an overall communication bus allowing information to be exchanged quickly. The new architecture of distributing satellite capabilities across a fleet of separated vehicles presents a new problem for overall communication design. With the new architecture, reliable and adaptable communication systems are going to need to be developed for remote communication between devices. Algorithms and communication standards are being developed and shared to enhance capabilities of remote satellites not only with the ground, but also amongst their own fleet . For example, if only one satellite in a multi-vehicle fleet has a star sensor to determine orientation, it must communicate the measurement to the other members of the fleet in order for all pieces to move as a cohesive unit. Autonomous capability to reconfigure networks and remote communication will be necessary for fleets to maintain continuity for operation. Occlusions to communication could happen as satellites cross paths or distances become too large. Potentially, some satellites' primary purpose would be simple relays of information between vehicles separated over too great a distance. If the remote communication devices were based on MEMS technologies, many different communication capabilities could be carried on a single microsatellite allowing for potential to overcome single-point failure or allow the overall system to re-configure relatively easily. The remainder of this report will not explore the details of specific communication protocol or information, which are generally mission specific, but rather indicate devices that have potential in space applications for remote communication between separated vehicles.Nanosatellite Communication Systems and MEMS The communication system on any satellite consists of two basic mechanisms: a transmitter and a receiver. Satellites usually deal with signals in the microwave range, which are high frequency, short wavelength signals that can carry a large amount of information. Often, the signals are in the gigahertz (109 Hz) radio frequency (RF) range. In order to effectively transmit and receive RF signals, satellites must have extremely sensitive signal processing equipment. Also, the signal must be able to transmit over a large distance with very high fidelity and fast data rate. Currently, most microwave signal handling is done with solid-state electronic components. Electrical performance of microwave components is determined almost entirely by the mechanical dimensions of the devices, and precision in the manufacturing of these components is extremely important . MEMS devices are very attractive to use in these applications in this respect because of the ability to accurately control their dimensions. The design of conventional solid-state microwave and RF devices is hindered by three constraints, which are power consumption, sensitivity, and size . MEMS can improve on conventional designs in all three of these areas, as well as offering a lower production cost. As frequencies are driven higher and higher by increased data transmission requirements, the shortcomings of conventional designs are compounded by the lethargic response of macro-sized components . Also, in an increasingly crowded radio spectrum, high sensitivity is crucial. By integrating a MEMS device directly on a silicon chip and semiconductor circuit, energy consumption and signal noise are reduced even further. MEMS devices in a nanosatellite can be used as signal filters, micro-switches, and antenna components such as phase shifters. Some examples of each of these components will be presented as well as some of the fabrication and performance parameters.
Signal Filters An RF circuit requires at least one filter to pull out a desired signal from a receiving antenna or to insert one to be transmitted. Currently, surface acoustic wave (SAW) filters are used to do this job. However, these filters are relatively large and do not work well at very high frequencies . Digital signal processing can be done on the back-end of the process, but this consumes valuable electrical power from batteries or solar panels. MEMS designs are well suited to perform front-end analog frequency filtering, taking up less space and using less power because the device is passive. By increasing filter sensitivity, potential exists to improve communications systems. A filter could not only preselect a communications band, but also perform channel selection within that band, enabling significant improvement over existing technology. In one such design, a passive microwave filter is deposited on a GaAs membrane and displays very low power loss and good performance . First, a 2.2 m thick membrane is machined utilizing RIE techniques on the GaAs wafer. To fabricate the filter structure, a 0.7 m gold layer is deposited on the membrane and conventional contact lithography, e-gun evaporation, and lift-off techniques are used (figure 3). This device could easily be coupled with other components such as antennas, capacitors, and inductors on the same substrate.
RF-MEMS switches Switches are vital to the successful operation of an antenna array. They are used for beam shaping and steering by individually selecting elements in the array . Conventional switches suffer from losses in efficiency and have higher insertion losses, due to parasitic effects of the system. MEMS switches, on the other hand, do not suffer from these problems. These switches have properties that are ideal for MEMS space-based applications, including the following: low insertion losses, rapid response, improved power handling, wide bandwidth, good electromagnetic isolation, and high open position isolation. Drawbacks are minor, but include stiction and a comparatively slower response time with active components. Several switch designs have been explored. Cantilever and membrane (bridge) types are the most popular, however there are also seesaw bar, rotary, and derrick-type switches that are available for specialized applications . Selection of the specific design may include parameters such as insertion loss and isolation criteria. For membrane switches, both insertion loss and isolation are a function of the change in capacitance during the on and off states. To decrease the insertion loss, contact should be as close as possible. Conversely, improved isolation requires the conductors be as far from each other as they can in the off position. An electrostatically actuated MEMS coplanar waveguide (CPW) shunt switch is shown in figure 4. It is comprised of a metallic bridge between two off-center pull-down electrodes. A supplied bias voltage moves the bridge into contact with the ground. The in-line configuration of the switch results in a compact design with high isolation at 0.1-18 GHz and specifically, with regard to inductance to ground, the isolation at mm-wavelengths is –20 dB at 18GHz . This can be compared with capacitive contact shunt switches with high isolation ranging between 10 and 120 GHz.
Figure 4. RF-MEMS CPW shunt switch schematic  Force balance is used to calculate the electrostatic force as follows, (1)where , A, V, d, and x are the permittivity of the free space, projected area of the electrodes, applied voltage, gap between the line and bridge, and the deflection of the bridge respectively. A restoring force, Fs, is given as (2)
where the spring constant, k, can be found with the following equation, (3)where E, t, w, L, and are the effective Young's modulus of the bridge material, thickness, width, and length of the bridge, residual stress, and Poisson's ratio respectively . If the deflection is greater than 1/3 d, the system becomes unstable and pull-in occurs. The voltage required for this may be calculated by (4) The resonant frequency of the cantilevered beam is another important design consideration. It controls the maximum speed under which the switch can reliably perform. Attenuation also varies as √f. This frequency can be determined by considering the beam to be supported at both ends, where (5)with a given system mass, m. For the aforementioned shunt switch, an attenuation of 0.4 dB/cm occurs at 10 GHz. Given the two states of the MEMS switch, isolation and insertion loss, the following values may be calculated from impedance and capacitance. When in the down-state position, the switch can be modeled as an RsL circuit. Switch resistance, Rs, is defined by half the sum for the resistance at the contact points and at the short transmission line, Rc and Rl respectively. With impedance given as Z0, the isolation can be found with the following equation, (6)The insertion loss for the up-state position can be calculated by modeling the switch with two sections of transmission lines, with two capacitors, to represent switch and electrode coupling, and two bias resistors. Therefore, by using an RC model, the loss can be found by (7)where Cb and Rb are the capacitors and resistors respectively. Trends for the insertion losses for this switch design can be found in figure 5. Here, the strong dependence on bias resistance is shown, while the bias capacitance has a variable effect on the loss .
Antennas This section will now shift focus from individual switch design to another communication component, the RF antenna. General considerations will be introduced followed by specific examples, including a Hilbert Curve Fractal Antenna (HCFA). Planar antennas are important in both the transmission and reception of signals in nanosatellites. With a MEMS-based design, an antenna can be integrated with other passive or active components. Also, many antennas can be arranged on the same substrate, allowing a phased array configuration . In a specific design by IMT-Bucharest, an antenna is fabricated on a 1.5 m thick silicon oxide/silicon nitride dielectric membrane micro-machined from a silicon substrate, called a double-folded slot antenna. Gold is then electrochemically deposited on the wafer to form the antenna of thickness 2.5 m, shown in figure 8. The two cross members of the antenna are placed one half wavelength apart for good radiation performance . The device operates at 77 GHz and could be easily be included in a nanosatellite design. Another design of a MEMS antenna, called a V-antenna (figure 9), has moving parts and can be reconfigured for adaptation in new environments. The arms of the V-shaped antenna can be moved independently with comb drive micro-actuators and a forward- or backward-moving bias. When both arms are moved in the same direction with a fixed angle, the antenna can be used to steer the radiation beam and focus its reception or transmission of data . The shape of the beam can also be adjusted by changing the angle. This technology has been demonstrated for a 17.5 GHz MEMS V-antenna.
HCFA Fractal antennas are also employed for MEMS space-based communication. Similar to V-antennas with regard to their incorporation of MEMS devices, fractal systems are capable of adjusting their characteristics to suit specific needs for reconfiguration in changing environments . Hilbert curve fractal antennas (HCFAs) are well suited to satellite broadcasting. These can maximize the amount of space available for RF communication.MEMS technology eliminates the need for the more expensive, bulkier, and less efficient conventional phase shifters. Furthermore, high frequency array batch fabrication can be carried out on a single chip. HCFAs utilize switches and phase shifter as integral parts of their design. Their multiband nature, useful for multiplexing, and size to weight ratios make them much more attractive than their conventional macro-sized counterparts. Design of and experimental results for HCFAs will now be addressed. HCFA take advantage of the relatively self-similar pattern of the fractal, as seen in figure 11. This figure shows 4 iterations of Hilbert curves. The additional line segments used between each iteration are small compared to the overall geometry, and can be considered negligible. For each iteration order, the line length of the segments dramatically grows while the area covered by the antenna footprint increases only moderately. This shows how the curve almost fills the plane. Resonant frequency can be significantly reduced for a given area by increasing the number of iterations, thereby averting a problem many small-scale antennas encounter. This approach is similar to that of Euclidian based geometrical systems; however, the use of fractals can help achieve higher degrees of freedom. Figure 11 provides a more detailed look at the individual segments of the fractal. The total inductance of the system can be calculated by determining the inductance of the turns of the meander line. The inductance of a regular half-wavelength dipole can then be compared to this total inductance.
Assuming a constant dipole configuration capacitance, the induction of the system can be calculated with the following equations. First, line segment length, d, can be defined as (11)where l and n are the outer dimension and iteration order of the fractal. The number of short-circuited parallel segments of length d can then be defined by (12)The total length of wire not forming parallel sections can also be defined as (13)Now, the characteristic impedance of the transmission line can be calculated by the following equation, (14)where the diameter, b, and spacing, d, are obtained from the fractal geometry. is the intrinsic impedance of the free space. Input impedance at the end of the line can then be found, (15)The impedance for the straight-line segments is therefore, (16)where there are m sections of input impedance, is the permeability of free space and is a function of the system. Resonant frequency can be found by equating the total inductance with that of a half-wave dipole, where l=/2 . Since regular dipole antennas have resonance at quarter wavelength multiples, the equation for the first several HCFA resonant frequencies is given as (17)where k is an odd integer. This equation only considers low order effects and is not recommended for high order modes. Figure 12 shows the measured input impedance compared with calculated values for a third iteration fractal antenna at various frequencies. For the corresponding radiation patterns, there are fewer nulls present than would be expected for a dipole antenna due to the smaller overall size of the radiator.
Reconfigurability of the HCFA is achieved by incorporating RF-MEMS switches in series with the length of the antenna, thereby facilitating frequency tuning. Such agility would be required if ambient operating conditions were to change, altering frequency characteristics. The frequency tuning characteristics of the antenna are shown in figure 13 by the voltage standing wave ratio (VSWR) as a function of the various switching configurations and frequency, when two switches are located at adjacent nodes. Furthermore, radiation patterns are also presented in figure 14. Activating switches to enable additional line segments have generated these patterns. Changes in the xy-plane are provided due to their dominance in the system.
Figure 14. Radiation patterns for HCFA with additional segments, shown in the xy-plane 
Table 2. Phase shift in each arm and resulting peak direction of the beam Case Elem. 1 phase Elem. 2 phase Elem. 3 phase Elem. 4 phase Beam dir1 0 0 0 0 0˚2 0 20 40 60 6˚3 0 40 80 120 13˚4 0 60 120 180 19˚5 0 90 180 270 29˚6 0 120 240 360 38˚
Picosatellite Experiment Initial experiments for MEMS components have been conducted with low earth orbit (LEO) satellite payloads. As previously mentioned, cost savings achieved though reducing payload weight can be realized through the use of such technology. A Stanford University designed satellite, the Orbiting Picosatellite Automated Launcher (OPAL), has released multiple picosatellites for testing individual MEMS devices . The Aerospace Corporation with funding from DARPA engineered one of the picosatellites with an array of RF-MEMS switches onboard developed at Rockwell Science Center (RSC) . Specifications of the switches are as follows: low insertion losses at 40GHz of 0.2 dB, isolation >60 dB at dc and ~25 dB at 40 GHz, actuation voltage ~80 V, on/off response within ~10 s, mechanical response reliability (no modulus reduction) after 60 billion cycles, and hot-switched lifetimes 7 orders of magnitude at ~1mA. The RSC microrelay is approximately 250 m 250 m. The microrelay is fabricated in a similar manner as shown in figure 7. The surface micromachining is performed under 250 C. High conductivity evaporated Au is used for all RF signal and dc lines. The sacrificial layer is removed, in this instance, by dry release etching in an oxygen plasma. This small system consumes very little power as a result of its electrostatic actuation. Shifts in the elastic constant resulting from plasma enhances chemical vapor deposition (PECVD) SiO2 are negligible. The switch experiments conducted in the picosats employed four MEMS switches in series with resistances of 3, 10, 30, and 300 k. Batteries rated at 3 V were used to power the satellites, and inductive charge pump circuits were used to generate the 100 V (25% over voltage) for switch activation. On February 6th, 2000, two picosatellites were released from OPAL and began orbiting the earth in a sun-synchronous 750 km polar orbit. A 30.5 m tether was used to maintain a constant separation distance between satellites because they did not have any control thrusters to adjust their relative positions. Each satellite had dimensions of 10.16 7.62 2.54 cm3. The RF MEMS switches were switched on and off repeatedly as requested by programming uploaded from a base station on earth. A standard radio communication and networking module was used to as the means for remote communication. Photographs of the system can be found in figure 15. The picosatellites were only in visible communication range of the base station two times each day for approximately 300-600 seconds during each pass. During such contact, information regarding switching status and temperature would be transmitted. While this picosatellite experiment did not test a complete MEMS communication system, it has shown how individual MEMS devices can be launched and tested in space without significant experimental costs. Indirectly, the mission showed how MEMS communication components would help improve mission robustness in the future. Unfortunately, the picosatellite mission was shortened by nearly 75% while the standard radio modem was broadcasting in the "transmitting" state attempting to form a link to the ground station. The continuous operation of the radio module drained the limited battery supply during the initial stages of the mission. If for future missions, the module is replaced by a MEM system, the power drained from initialization can be reduced allowing longer life for fixed power storage devices.Conclusions and Recommendations There are many opportunities for MEMS communication devices in the space industry. With an increasing emphasis on designing spacecraft with lower mass and cost, MEMS structures are ideally suited for a nanosatellite platform. While these devices can be incorporated in the satellite propulsion and navigation systems, some of the most practical and beneficial applications of MEMS devices exist in the satellite communication system.
Patrick Schubel received his BS in mechanical engineering at the University of Dayton in 1999 and MSME from Northwestern University in 2002, where he is now a PhD student. His interest lies in experimental mechanics and composite behavior. His research has included studying the fracture mechanics response of tires and quantifying effects of porosity on the strength of aerospace fiberglass. Besides being a dedicated scholar, he enjoys playing cards, driving his 1990 Mercury Grand Marquis, and rock and roll.
Nick A. Pohlman received a B.S. in Mechanical Engineering from the University of Dayton in 2000 and a S.M. in Aeronautics and Astronautics from the Massachusetts Institute of Technology in 2002. He is currently conducting research and working to complete a Ph.D. in Mechanical Engineering from Northwestern University. His future interests include obtaining tenure at an academic institution and having a guest appearance with Kermit the Frog or a cameo in the next Superman film.
Jeremy G. Opperer received a B.S. in Mechanical Engineering from Lawrence Technological University in 1998 and a M.S. in Mechanical Engineering from Northwestern University in 2002. He is currently conducting top secret research and working to complete a Ph.D. in Mechanical Engineering from Northwestern University. His interests include long walks on the beach and a warm fire to cuddle up to. He also enjoys tennis and the ancient art of the caber toss.