SWITCHES
MEMS switches are surface-micromachined devices which use a mechanical movement to achieve a short circuit or an open circuit in the RF transmission-line (Figs. 1-2). RF MEMS switches are the specific micromechanical switches which are designed to operate at RF to mm-wave frequencies (0.1 to 100 GHz). The advantages of MEMS switches over PIN diode or FET switches are [1]: Near-Zero Power Consumption: Electrostatic actuation requires 30-80 V, but does not consume any current, leading to a very low power dissipation (10-100 nJ per switching cycles). On the other hand, thermal magnetic switches consume a lot of current unless they are made to latch in the down-state position once actuated. Very High Isolation: RF MEMS metal-contact switches are fabricated with air gaps, and therefore, have very low off-state capacitances (2-4 fF) resultingin excellent isolation at 0.1-60GHz. Also, capacitive switches with a capacitance ratio of 60-160 provide excellent isolation from 8-100Hz. Very Low Insertion Loss: RF MEMS metal-contact and capacitive switches have an insertion loss of 0.1dB up to 100GHz. Linearity and Intermodulation Products: MEMS switches are extremely linear devices and therefore re-
sult in very low intermodulation products in switching and tuning operations. Their performance is 30-50 dB better than PIN or FET switches. Potential for Low Cost: RF MEMS switches are fabricated using surface micromachining techniques and can be built on quartz, Pyrex, LTCC, mechanicalgrade high-resistivity silicon or GaAs substrates. RF MEMS switches also have their share of problems, and these are: Relatively Low Speeds: The switching speed of most electrostatic MEMS switches is 2-40 μs, and
High Voltage or High Current Drive: Electrostatic MEMS switches require 30-80 V for reliable operation, and this requires a voltage up-converter chip when used in portable telecommunication systems. Thermal magnetic switches can be actuated using 2-5 V, but require 10-100 mA of actuation current. Power Handling: Most MEMS switches cannot handle more than 200 mW although some switches have shown up to 500 mW power handling (Terravicta and Raytheon). MEMS switches that handle 1-10 W with high reliability simply do not exist today. Reliability: The reliability of mature MEMS switches is 0.1-40 Billion cycles. However, many systems require switches with 20-200 Billion cycles. Also, the long term reliability (years) has not yet been addressed. It is now well known that the capacitive switches are limited by the dielectric charging which occurs in the actuation electrode, while the metalcontact switches are limited by the interface problems between the contact metals, which could be severe under low contact forces (in electrostatic designs, the contact forces are around 40-100 μN per contact).
It is important to note that the reliability and packaging issues have been the limiting factors to the quick deployment of RF MEMS switches, and they are currently under intense investigations. DARPA has initiated two programs in 2002 and 2003 to address these problems, the RF MEMS Improvement program (Dr. Larry Corey), and the HERMIT program (Dr. Clark Nguyen), and it is expected that some of these problems will be solved in the coming 2-3 years. Packaging: MEMS switches need to be packaged in inert atmospheres (Nitrogen, Argon, etc..) and in very low humidity, resulting in hermetic or nearhermetic seals. Hermetic packaging costs are currently relatively high, and the packaging technique itself may adversely affect the reliability of the MEMS switch. Microassembly (Fig. 3) and Analog Devices have both developed excellent packages for RF MEMS switches. The Microassembly package is based on gold-to-gold thermo-compression at 250◦C while the Analog Devices package is based on glass-to-glass seal at 400−450◦C. Other companies which have packaged switches are Terravicta (ceramic package) and Omron (glass-to-glass). Cost: While MEMS switches have the potential of very low cost manufacturing, one must add the cost of the packaging and the high-voltage drive chip. It is therefore hard to beat a $0.3-0.6 single-pole doublethrow 3 V PIN or FET switch, tested, packaged and delivered. It is for this reason that Prof. Rebeiz believes that RF MEMS switches will be first used in defense and high-value commercial applications and not in cellular phones.
DETAILED DISCUSSION OF MEMS
SWITCHES
Actuation Mechanisms: The actuation forces required for the mechanical movement can be obtained using electrostatic, magneto-static, piezoelectric or thermal designs. To date, only electrostatic-type switches have been demonstrated at 0.1-100GHz with high reliability at low RF powers for metal contact and medium power levels for capacitive contacts (100 4A1.4 Million to 50 Billion cycles depending on the manufacturer) and wafer-scale manufacturing techniques. Other switches which have demonstrated excellent performance are the Microlab Latching switch (up to 100 Million cycles) using magnetic actuation, and the thermal switches developed independently by Cronos Microsystems and the Univ. of California, Davis. It
is hard to test thermal switches for long cycle times
The near-ideal electrical response of RF MEMS witches (both metal-contact and capacitive) have allowed many designers to build state-of-the-art switching circuits from 0.1GHz all the way to 120GHz. In the past 4 years, these applications concentrated on the replacement of GaAs phase shifters which are commonly used in phased arrays by the thousands of units. A comparison between 3-bit GaAs phase shifters and MEMS phase shifters is shown in Table I and it is seen that MEMS switches provide an immense performance benefit especially at Ka-Band to W-band applications.
Fig. 4 presents a 4-bit miniature RF MEMS phase shifter developed jointly by the Univ. of Michigan and Rockwell Scientific. It is based on the Rockwell metalcontact switch and on CLC delay lines for miniaturization. The phase shifter results in an average loss of 1.4dB at 10GHz, a ±3◦ phase error, and is matched to −13 dB at the input and output ports from 6-16GHz. This phase shifter represents the smaller design using RF MEMS to-date, and with excellent response. Fig. 5 presents an 885-986MHz 5-pole tunable
filter using switched MEMS capacitors developed by Raytheon Systems Co. In this case, capacitive switches are used to switch fixed-value metalinsulator- metal capacitors in the transmission line. The filter employs 18 switches and is a very complicated circuit with variable resonators and impedance inverters. Its measured response is nearly ideal, with excellent frequency tuning capabilities, very high linearity (in terms of measured IIP3) and a loss of 5- 6 dB due to the finite Q of the planar inductors used (Q = 30 at 0.9GHz). Fig. 6 presents a W-band 3-bit phase shifter developed at the Univ. of Michigan using MEMS capacitive switches [3]. This is the highest frequency MEMS phase shifter to-date and results in an average loss of 2.7-2.9 dB at 77-94GHz with an associated phase error of ±3◦. The results are about 8 dB better than GaAs designs.
Other circuits, which are not shown due to space constraints, are very wideband SP4T switches, highisolation series/shunt switches covering 0.1-50GHz, double-pole double-throw transfer switches, and a whole range of phase shifters from 8GHz to 120GHz. Also, tunable filters covering 200MHz to 23GHz have been developed by various groups. In general, RF MEMS circuits outperform GaAs FET and PIN diode circuits by a large margin at all frequencies of interest
circuits developed in the world can be found in [1].
It is now clear that we understand RF MEMS switches well, both from the mechanical and electrical/ electromagnetic point of view. We can design complicated circuits using MEMS switches or varactors, and we can accurately predict their performance all the way to 120 GHz. They are still not accepted in the commercial and defense arena due to their need of a hermetic package, and their reliability under medium to high-power conditions. There is currently an intense effort to solve these problems, and the author
believes that RF MEMS switches and varactors will play an essential role in future high-value commercial and defense systems.
When it comes to organizing the file structure for your mask layout, it will be dependent on the type of MEMS device. For example, a MEMS device that has a large number of metal bondpads that connect to a doped region in the silicon substrate through an oxide layer, it may be helpful to create a cell that contains the pattern of a contact etch through the oxide and of a metal bond pad. Inserting and aligning a cell instance will take much less time than drawing the feature at every location, especially after the twentieth bondpad. But the power of using these hierarchical cells to create instances goes further when it is determined that the bondpad size needs to be decreased or increased. If you have used the cell hierarchy, this change is made once and then automatically propagated through the rest of the instances in the entire layout. The cell hierarchy is also useful when dealing with layout variations. If you have a device design that uses cantilevers of varying widths, you can create individual cells with each of these cantilevers and then insert each as a sub-cell into a top-level cell as needed without drawing each cantilever over and over again. synonym for thought leadership
ResponderEliminar