RF-MEMS Switching Concepts for High Power Applications
Introduction
Over the past several years, developments in Micro-Electro-Mechanical Systems (MEMS) have promoted exciting advancements in the field of microwave switching. Micromechanical switches were first demonstrated in 1971 [1] as electrostatically actuated cantilever arms usedtoswitchlow-frequencyelectricalsignals.Sincethen,theseswitcheshave demonstratedusefulperformanceatmicrowavefrequencies.Differentswitchtopologieshave been investigated and tested [2, 3], and most of them use electrostatic actuation. The advantageofusingMEMSoverconventionalsolidstateswitchingdevicessuchasFETsorp-
i-n diodes is their low loss performance, low power consumption and lack of measurable intermodulation distortion. There are three main challenging aspects for RF MEMS switches: loweringtheactuationvoltage,increasingtheswitchingspeedandincreasingpowerhandling capabilities. For lowering the switching speed meander spring suspension [4] and push-pull concepts have been investigated [5]. Increasing the switching speed and the power handling capabilities are still a problem. DaimlerChrysler has developed capacitive RF shunt switches with gold metallization lines and gold membranes [6]. Using these capacitive RF shunt switches a 180° phase shifter has been realized [7]. DaimlerChrysler is now developing RF MEMS SPDT switches for RF power applicationbeyond10W.Forthis,newswitching conceptsareevaluated.
Powerhandlingcapabilities
ThepowerhandlingcapabilitiesoftheRFMEMSswitchescanbelimitedeitherbythecurrent densityonthetransmissionlinescausingexcessiveheatingorbytheactuationofthe switches due to the average RF-voltage on the coplanar waveguide (CPW) line (denoted as "self biasing"). Since the electrostatic force acting on the switch can derive either from a negative or positive voltage, and since the relaxation time offree electrons in metals is in the rangeof10-14 s(forAu:Au =2.910-14 s,forAl:Al =0.810-14 s)whichcorrespondsto
However,powerdatahaverarelybeenreportedinliterature,primarilybecausethetestset-up requires a relatively expensive high-power RF generator. The following data are reported: Brown [2] mentions in his review article a private communication from J. Goldsmith (Raytheon)thatself-closingisobservedwithapproximately3WCWpowerat10GHzand with approximately 1 W at 35 GHz. Pacheco [5] reports on capacitiveshuntswitchesno"self biasing"orfailureoftheRFMEMSswitchesforpowerlevelsupto6.6WintheX-band(10
GHz). Schaffner [8] reports on resistive cantilever switches power switching up to1ina50 line.Ourinvestigationsshowthatselfswitchingmayoccur.
These results indicate that for power beyond 10 Watt new switching concepts have to be investigated.
Thenewswitchconcept
The proposed concept is a so called "double anchor switch" (Fig. 2). The double anchor switchconsistsoftwofixedelectrodes:abottomelectrodeandatopelectrode.Theswitching membrane in this case is fixed at both ends and can be actuated either to the bottom or top electrode.ThereforeselfclosingduetoRFpowershouldbestronglyreduced
This switch type is also very suited for SPDT switches, since one can switch from input port1 (membrane)tooutputport2(connectedtobottomelectrode)ortooutputport3(connectedto topelectrode),showninFig.2.
Another type of the double anchor switch consists of one flexible membrane fixed only at one end, which can be switched up and down (Fig.3). In the down position the signal is routed to port3 and in the up position is the signal routed witha 90° bend to port 2, as depicted in Fig.
DifferentSPDTswitchtopologieshavebeensimulated.OnetopologywaswithtwoSerial-Air- Bridge MEMS switches. In comparison to this the scattering capacitance of an SPDT switch with a single-ended "double anchor switch" is lower, because there is no open end stub in case of routing the signal to port 2. This yields better results in return loss and insertion loss if the signal is switched to port 2. The air bridges at the bend (see Fig. 4) are necessary to suppress asymmetrical modes on the coplanar line to port 2. In Fig. 5, the simulated S- parameterfortheon-stateandoff-stateforthedifferentpathsaredepicted.Forthrutoport2 theinsertionlossS21is-0.06at8GHzand-0.7dBat30GHz.Isolationtoport3(S31)is-34.7 dBat8GHzand-17.3dBat30GHz.Forthrutoport3theinsertionlossS31is-0.18dBat8
shown, that metallic losses for the used CPW lines are less than 0.1 dB at 30GHz. Since the isolation decreases with increasing frequency additional shunt capacitive switches may be usedforimprovedSPDTswitches.
Mechanicalsimulation
In addition to the electromagnetic simulation, mechanical simulations of the new structure are performed. The membrane is in the middle of two fixed electrodes and consists of aluminum or gold with a thickness between 0.75 µm and 1 µm. The distance between lower and upper electrodes is 6 µm, and between membrane and electrodes is 3 µm. A thin dielectric film is deposited both over and under the electrodes. A schematic view of the simulated structure is showninFig.6.
Due to the small membrane thickness the mechanical structure was simulated with the simulation program ANSYS. The structure was described with dish elements. With these simulations the actuation voltage and the stress distribution in the actuated membrane (membrane in contact) are calculated. The behavior of the membrane is simulated with the flexurelineasananalyticalfunctionandwiththeenergyatequilibrium.Thegravitationforce
The calculated stress in the membrane is less than 50 MPa at the fix point. The simulated actuationvoltagesareintherangebetween4Vand14V.Theresonancefrequenciesforthis structureareintherangefrom6kHzto21kHz.
Fabricationprocess
ThecapacitiveRFswitchesarefabricatedonhigh-resistivitysiliconwafers( >4000cm) with a wafer thickness of 525 µm. The fabrication process of a capacitive RF switch is depictedinFig.8.
First, the lower electrode (underpass metallization) is defined by a lift-off process with 50 nm Ti and 300 nm Au. Then, the lower electrode is isolated by a 100 nm thick PECVD silicon nitride layer under the membrane region. Next, the transmission lines are defined by a lift-off processwith50nmTiand2500nmAu.Atthispoint,anair-bridgeresistwithaheightof2.5-
3 µm is patterned as sacrificial layer. Afterwards, the membrane metallization is sputtered. Themembranematerialconsistsof0.5-1µmAuorAl.Lastly,themembraneresistisdefined andthemembraneisetched.
After these steps, a third electrode on top must be defined. For this, an air-bridge resist with an height of 2.5-3 µm is patterned as second sacrificial layer. Then, the third upper electrode isisolatedbya100nmPECVDsiliconnitridelayerovertheresistandundertheregionofthe thirdelectrode.Next,themetallizationforthethirdelectrodeissputtered.Theelectrode
Stiction is a difficulty in the fabrication process. When the wafer is pulled out of the rising solutionusedtodissolvethesacrificiallayer(photoresistinourcase)stickingofthestructural elementstothesubstratemayoccur[10].ThiscanbeavoidedbysupercriticalCO2drying[9]
orvaporphaseetching.Inourcase,thesacrificialphotoresistlayersareremovedinO2 plasma.Fig.9showsthelayoutofacapacitiveRFmembraneswitchchipwith120µmsignal lineand90µmspacing(6timesscalingofourstandard50CPWlines).Thewidthofthe membrane is 100 µm, the distance of the membrane to the enforced transmission signal line is15µm.Thetotalchipsizeis1300µmx1100µm,andthemetallizedchipsizeis1200µmx
1000µm.
The membrane metallization thickness was 1 µm, the height above ground was 3.3 µm and thetransmissionlinemetallizationthicknesswas3.6µm.
V, depending on membrane thickness, membrane height and fabrication process. Insertion loss lower than 0.3 dB@ 35GHz in the off-state (0 V) and isolation down to 35 dB@ 35 GHz intheon-state(25V)aremeasured.
Conclusion
A new capacitive SPDT switch for power applications is proposed. The electrical and mechanical properties and behaviors of this so called "double anchor switch" are discussed. Also, technology fabrication aspects of the new switch type are discussed. Capacitive RF- switches have already been fabricated on high resistivity silicon substrates. Measurements demonstrate that these metallic membrane switches have low insertion loss and a good isolation at frequencies into the millimeter-wave bands. These devices offer the potential for building a new generation of low loss high linearity microwave circuits for a variety of phased antennaarraysradarandcommunicationapplications.
No hay comentarios:
Publicar un comentario