By partnering with an established semiconductor foundry, a developer of MEMS switches has found a practical path to integrating MEMS with CMOS devices.
Mobile-telephone designs continue to integrate more functions, but with goals of cutting supply voltage and current. Because of these conflicting design goals, RF microelectromechanical-system (MEMS) technology has attracted the attention of designers as a viable solution for certain cell-phone functions, such as filtering and switching, while operating from limited power supplies often as low as 2.7 V or less.
Still, many factors limit the use of MEMS technology in mobile-telephone applications, including the difficulty of providing MEMS-compatible control voltages within the mobile-telephone design. Considering that modern microelectronic systems are migrating to voltages below 3 V and eventually as low as 1 V leaves little room for widespread adoption of higher-voltage RF MEMS components. RF MEMS devices are designed as either low-voltage/high-current electromagnetic components or high-voltage/low-current electrostatic components. The problems in using RF MEMS within a mobile telephone is the need to provide either the high current or high voltage needed for either type of device (see table).
An RF MEMS can be thought of as a transistor-scale mechanical system with moving parts, shunt or serial switching elements. Designed to replace traditional electrical RF active or passive components, a MEMS capacitor, for example, is built from a micro scale, singly or doubly anchored beam that is displaced or actuated either electrostatically or electromagnetically. A dielectric material rests under the cantilever and is comprised of an insulating material (Fig. 1).
RF MEMS can be fabricated in much the same way as a semiconductor where layers of materials are deposited, patterned, and etched to form the basic architecture of the target devices whether they are capacitors, inductors, or switches. Standard semiconductor fabrication plants can manufacture these devices in high volumes, although the process is not trivial because of precise etching requirements. For the MEMS capacitor, for example, a typical dielectric material such as silicon dioxide must be layered in the right amounts and in the right places for the structure to be useful. The process involves encasing the dielectric in materials that wash away during chemical etching. Although difficult at first, the process yields good results with repeated efforts and so almost any semiconductor fabrication facility can adapt their processing recipes to form these structures from layers of conductors and insulators.
RF MEMS components are activated or controlled either with electrostatic or electromagnetic fields. Both methods are viable, but each has drawbacks. Electromagnetic MEMS devices draw current through its control line into a micrometer sized coil to generate a magnetic field that pulls or releases the cantilever. These devices often use a ferromagnetic plate in the construction of the beams so as to have a constant magnetic pole for latching the device. The added bulk of a magnet opposing an electromagnetic circuit can result in a larger-than-desired device. The electrostatic version uses a charged field to attract (close) or release (open) the MEMS beam (Fig. 2). In contrast to electromagnetic devices, the electrostatic versions have a thin layer of metal within the beam structure, which makes them susceptible to damage from heat created by excess current.
Electrostatic MEMS actuation or control voltages have changed from 100 V in early designs to less than 40 V in more current designs. This type of capacitor can translate the control voltage into mechanical movement and so a change in the actuation voltage can effectively change the capacitance of the device. This is done by altering the cantilever's electrostatic field, which in turn alters the distance between the capacitor plates, forming a variable or trimmer capacitor that can be used in electromechanically tunable circuits.
Unfortunately, wireless terminals do not have the high voltages (or currents) needed for MEMS actuation in order to take advantage of tunable RF MEMS technology. As a result, MEMS developers must provide an intermediary step that enables the device to operate at lower power levels. In an electrostatic device, DC-to-DC voltage conversion can be used for this purpose. Ultimately, increased integration allows for a voltage converter and logic controller to be designed within these high-voltage devices to create a lowervoltage, monolithic solution.
A variety of DC-to-DC voltage conversion methods are available for integration with MEMS that include transformers and multistage amplifiers. Possible implementations of voltage conversion with RF MEMS switches leverages the packaging techniques of monolithic microwave integrated circuits (MMICs) and multi-chip modules (MCMs) where multiple die are integrated into one package. Ultimately, the goal is to reach a monolithic or fully integrated switch solution (Fig. 3). With the availability of fabrication facilities, there will be no problems in providing low-voltage RF MEMS switches. Currently, several semiconductor suppliers in CMOS, silicononinsulator (SOI), and GaAs technologies are investigating this path toward integration.
Today, a voltage of roughly 40 V is needed for electrostatic MEMS actuation. Such a high voltage is not available in a mobile terminal. In a monolithically integrated system, these voltages can be generated and controlled on chip in order to fully leverage the advantages of MEMS designs. The integration of MEMS and CMOS technologies, for example, can increase the flexibility of a tunable filter. The MEMS structures can be placed without consideration of external bond pads and allow for some interesting circuit designs. The more notable benefit is that the integration greatly decreases system or printed-circuit-board (PCB) complexity by containing the high actuation voltages to the internals of the integrated circuit.
Until now, a great deal of attention has been paid to developing methods and circuits to generate high actuation voltages. These approaches included having separate controllers for MEMS and CMOS circuits, a method that draws too much power and potentially increased the costs related to PCB space, reflow, and long-term quality of a mobile telephone.
To realize a practical path to MEMS integration, WiSpry has partnered with Jazz Semiconductor (Irvine, CA) to develop the first MEMS device with an integrated CMOS controller. The resulting active CMOS controller is a multifunction IC with the capability of converting 3 V or less to the high voltages needed to control WiSpry's electrostatic RF MEMS switch. The controller includes a charge pump, high-voltage switches to multiplex or steer the high voltages from the charge pump to the MEMS, and digital CMOS processor (Fig. 4).
Designing RF MEMS technology into a low-power application such as a mobile terminal requires a small monolithic device with low power consumption in order to set it apart from a dual-chip, MEMS-plus-controller approach. The WiSpry solution features an integrated MEMS/CMOS controller IC with little impact on battery life due to power consumption of a mere hundreds of microwatts. The controller features an adjustable power profile that sets custom levels of power consumption, giving the controller the ability to consume less than 100 µW. The adjustable efficiency is the result of its on-chip charge pump and switch multiplexer with use more or less power depending on changes to the output drive.
The charge pump features a highefficiency "current-reuse" boost mode, enabling quick circuit startup with reduced power consumption. It also includes a low-power, efficient tracking/lock mode for output voltage regulation. The charge pump can source up to 60 µA of load current, enabling the integration of large numbers of MEMS devices on a single chip (Fig. 5). The example of Fig. 5 shows the on-chip reference from which the charge pump generates a high-voltage output that feeds a FET switch multiplexer in a simple 1:2 demultiplexer which can be easily expanded into a 1:8, 1:16, or 1:32 configuration. The key aspect of the design is that it utilizes CMOS transistors to steer the high voltages to the MEMS beams. With the exception of an external high-voltage monitor pin, which enables visibility into the part, all MEMS actuation voltages are contained safely on chip. A CMOS digital signal processor (DSP) is used to digitally control the output of the demultiplexer. A simple three-wire-control interface (such as SPI) can be used to generate all of the driver states for a 32-output device. The outputs of the DSP toggle the highvoltage translator switches in a single or simultaneous output configuration.
The active circuitry for the MEMS CMOS integrated controller resides beneath the MEMS switches. Jazz Semiconductor developed a process that allows this physical integration without noise and spurious coupling from the on-chip active circuitry. This " coupling immunity" allows for the construction of on-chip RF structures, which previously could not be integrated into a cellular telephone.
The full integration of RF MEMS in a mobile terminal results in a miniature system that consumes low current without sacrificing functionality. Tunable MEMS devices stand to replace passive devices designed to filter, balance, drive, or switch discrete frequency bands with single-chip solutions that cover multiple bands. The devices themselves will allow mobile terminal designers to integrate a wider band antenna to cover all the frequencies used in an optimized "world phone."
The resulting RF MEMS design transforms an all-band "world phone" into a mobile product with a reduced bill of materials (BOM). At the head-end, a multi-band matching network will fine tune and balance the line found between the transceivers and the antenna. That transmission line often suffers from mismatched impedances, which can negatively impact the efficiency of the antenna. Discrete matching circuits are often used to bring these differences as close as possible to 50 ohms, but often time the circuits are unable to match the loads for different frequency bands. This results in a higher VSWR performance and reduces the maximum efficiency of the antenna. An RF MEMS tunable matching network can dial-in coverage for both the uplink or downlink of a given frequency band with low insertion and return losses. When coupled with an RF MEMS band switch, these line losses can be kept to a minimum while maximizing performance.
With RF MEMS technology, essential mobile-telephone passive devices such as bandpass filters and duplexers can be replaced with a few highquality-factor (high-Q) RF MEMS devices covering multiple frequency bands. The MEMS device switching selects the required frequency band while in-band loss and out-of-band rejection can be optimized by finetuning. RF MEMS duplexers can replace isolators or circulators because they can be designed to switch off and isolate the power amplifiers from the antenna in case the transmission line to the antenna breaks.
Ultimately, the most attractive potential use of RF MEMS would be in a cellular-handset power amplifier (PA). The PA is the only active device between the transceiver and the antenna. A typical "world phone" can use seven or more PAs, resulting in high power consumption as well as PCB realestate consumption. A tunable PA could possibly yield higher efficiencies, but a band-selectable amplifier could do the job of three or more such units. The increased efficiency stemming from using RF MEMS devices between the transceiver and antenna greatly improves the potential for this new and exciting technology in mobile terminals.