EM Design of Broadband RF Multiport Toggle Switches

EM Design of Broadband RF Multiport Toggle Switches W. Simon 1, B. Schauwecker 2, A. Lauer 1, A. Wien 1 and I. Wolff, Fellow IEEE 1 1 IMST GmbH, Carl-Friedrich-Gauss-Str. 2, 47475 Kamp Lintfort, Germany 2 DaimlerChrysler AG, Research & Technology, Wilhelm-Runge-Strasse 11, 8981 Ulm, Germany ABSTRACT Radio Frequency (RF) MEMS is an emerging sub-area of MEMS technology that is revolutionizing RF and microwave applications. RF MEMS devices have a broad range of optional applications in military and commercial wireless communication, navigation and sensor systems. This paper presents the EM design of different multiport Toggle Switches. Such a multiport switch can be used in compact designs of switching matrices, routing networks or phase shifters. One application range is the creation of electronically steerable antenna arrays which can be used for radar applications and satellite communication. The miniaturized switches are based on the SPST Toggle Switch [1-4] and in addition to the small size they have an increased RF performance regarding losses and operation bandwidth (DC GHz). A 3D FDTD field solver has been used for the electromagnetic design of all switches. I. EM DESIGN OF RF MEMS COMPONENTS For designing MEMS structures it is recommended to use a full 3D field solver if coupling effects and radiation should be considered. By using a thin sheet modeling technique, the drawback that thin sheets and thin metallisations slow down the simulation has been solved. A. Thin Sheet FDTD modeling If thin sheets, e.g. nitride sheets of 1 nm, are treated with FDTD, due to the stability criterion [6, Chapter 4] the time step has to be chosen very small. This leads to a very long simulation time. Considering a graded mesh FDTD setup with one single small cell where the nitride is located, the stability criterion [6, Equation 4.27] yields t z, t is the time step, z the smallest cell size. The nitride cell size is e.g. 1 times smaller than the smallest cell size otherwise, the CPU time is multiplied by a factor of 1 compared to structures without thin sheets. Using the equivalent circuit formulation and stability considerations [7], in the case of only one single thin cell, it is observed numerically that t z. This yields a speedup factor of 3-7 in the case discussed above. B. Design considerations for MEMS switches The Toggle Switch designs presented in this paper and also the Shunt Switch designs [8] imply a parasitic capacitance which must be compensated to achieve a good broadband RF performance. The Shunt Air bridges cause this capacitance in open state as they connect both ground metallisations in a small distance to the center conductor of the coplanar line. The parasitic capacitance of the Toggle Switch is caused by the small distance of about 3µm between the cantilever and the grounded DC switching electrodes. A broadband compensation of the capacitance is achieved by using a LC matching network (Fig. 1). s11 L/2 L/2 C Fig. 1 : L/C matching network Z=5Ω The maximum capacitance which could be compensated while fulfilling the specifications of a match of 2 db at 3 GHz has been investigated (Fig. 2). s 11 [db].e+ 1.E+1 2.E+1 3.E+1 4.E+1 5.E+1-6 1 2 3 4 5 Frequency [GHz] C/fF L/pH 1 25 1 3 1 35 118 4 118 45 118 5 Fig. 2: Return loss at feeding port dependent on switch capacitance and inductive matching A capacitance of 1 ff could be compensated by a total inductance of 35 ph up to a frequency of 34.5 GHz. Assuming that a match of 2 db at 3 GHz is needed, a maximal capacitance of 118 ff could be -6

compensated. For that an inductance of 45 ph is required. A higher compensating inductance decreases the performance in the lower frequency range. If a Shunt Switch is designed, there is always the tradeoff between decreasing the capacity in open state and increasing the capacity in closed state. The broader the switch is, the higher is the achieved isolation in closed state and the higher is the capacity which must be compensated in open state. The capacity of the Toggle Switches is proportional to the size of the cantilever. A smaller cantilever has a lower capacitance which leads to a good RF performance. The drawback of a smaller cantilever is the higher actuation voltage as well as the reduced power capability. II. MINIATURIZED SPST-TOGGLE SWITCH The SPST Toggle Switch (Fig. 3) consists of a thin metal cantilever (~9 nm), which is fixed sideways by a torsion spring. The torsion spring is built of a thin metal sheet. At the connecting edges of the torsion spring, the small radius prevents a breakaway at mechanical load. Thin electrodes on the substrate allow electrostatic switching of the cantilever utilizing a push pull concept, so no DC potential on the signal line is needed for switching. The cantilever contacts the inner conductor of the coplanar line directly (ohmic contact). An air-bridge connecting both ground metallisations of the coplanar line acts as a upper mechanical limit for the cantilever in open position. A flexible metal band creates the contact on the other side of the cantilever. The size of the flexible metal band is varied between 3 µm and 5 µm to ensure adequate mechanical flexibility of the cantilever. a) SPST Toggle Cantilever DC electrodes flexible metal band large bandwidth of operation can be achieved. This is a great advantage compared to the well-known Shunt- Air-Bridge switches [8] where a capacitive shunt connection can be achieved. Here the serier capacitance limits the lowest frequency range of usage if certain isolation must be obtained. The Toggle Switch is used in a 5 Ω coplanar line environment on a 525 µm thick silicon substrate (center conductor width 144µm, gap 78 µm) to create an open circuit in the center conductor. In closed position, the signal is routed from the cantilever to the center conductor of the coplanar line via the direct metal contact and via the flexible metal band. The 3D FDTD field solver EMPIRE TM [5] has been used to optimize the switch. The thin nitride sheet (1 nm) and the thin cantilever (9 nm) have been resolved in the simulation. The simulation model has a graded mesh with a grid spacing between 1 nm and 1 µm. With a total amount of 125 x 8 x 4 cells the simulation time on a 2.6 GHz Pentium 4 PC is 1 min. A total memory of 2 MB is needed. Due to the usage of the thin sheet algorithm (Chapter I) the time step could be increased from.3 fs to 1.4 fs which leads to a threefold reduction of the simulation time. As optimal implementation of the compensating inductors for the scaled Toggle Switch compensation lines with a width of 25 µm (Z L =135 Ω) and a length of 4 µm on the left side of the switch and 3 µm on the right side were found. A good RF performance up to 5 GHz could be achieved as the capacity of the scaled toggle is in the range of 8 ff. The current density distribution of the closed Toggle Switch at 2 GHz is shown in Fig. 4. The current density value is visualized by color and height. A difference of 5 db lies between the maximum amplitude and the minimum amplitude of the current density values. The highest current density values are on the toggle cantilever and on the flexible metal band. The current distribution at the surrounding coplanar line corresponds to the standard behavior of high current values at the line edges near the gap. db metal torsion spring db nitride isolation sheet Fig. 3: SPST Toggle Switch : a) Schematic 3D view of simulation model The metal-metal connection in closed position allows a transmission starting at DC and yields ideal isolation for DC in the open position of the switch. For this reason and by the small length of the switch (15 µm) a Fig. 4: Current density at 2 GHz for the closed switch The high current density on the cantilever is one limiting factor for the power capability of the switch. Compared to the large Toggle Switch [2,3] which has been tested up to 2.5 W RF power, the power capability

of this switch is expected due to the reduced cantilever width (25 µm instead of 4 µm) to be in the range of 1.5 W. The current density distribution of the open switch at 2 GHz (Fig. 5) shows high current density values at the edges of the open ended coplanar center conductor as well as at the edges of the surrounding ground metallisation. A small signal part is still coupled to the switch membrane, but the current density values decrease strongly to about 3 db at the other side of the switch. db Fig. 5: Current density at 2 GHz for the open switch db A frequency dependent loss model, which considers the skin effect, is used in the FDTD simulation to get accurate results for the insertion loss [7]. The simulation results of the scaled Toggle Switch (Fig. 6) in closed position display a return loss above 18 db up to 5 GHz while the insertion loss is below.6 db. Compared to the large Toggle Switch formerly described [3] this is a great improvement. In open position of the switch, an isolation greater than 5 db at 1 GHz, 28 db at 1 GHz, 22 db at 3 GHz and 2 db up to 5 GHz is achieved. Fig. 7: SEM photo of the fabricated scaled Toggle Switch A voltage between 8 V and 15 V is needed to close the Toggle Switch with a 3 µm long flexible metal band. The return loss of the closed switch is in the measurements up to 5 GHz better than 15 db while the insertion loss of the switch in the measurements is up to 3GHz below.5 db and up to 5 GHz below.95 db (Fig. 8). Compared to the simulations a slightly higher insertion loss is observed due to a smaller distance between the crossing air bridge and the Toggle s cantilever. If the switch is in open position an isolation of at least 48 db at 1 GHz, 26 db at 1 GHz, 2 db at 3 GHz and 16 db at 5 GHz is measured. This corresponds very good with the simulated isolation of 5 db at 1 GHz and 22 db at 3 GHz.. s11 thru s21 open s21 thru s11 open -.2 -.4 -.6 -.8-1. Transmission,Reflection /db Fig. 8: Measurement results of the scaled Toggle-Switch in open and closed position Fig. 6: -6-1.2 1 2 3 4 5 Simulation results of the SPST Toggle Switch. The Daimler Chrysler Research & Technology Center in Ulm fabricated the Toggle Switches on highresistivity silicon wafers (ρ > 4 Ωcm) with a wafer thickness of 525 µm (Fig. 7). III. MINIATURIZED SPDT-TOGGLE SWITCH The miniaturized SPDT Toggle Switch is build of two SPST Toggle Switches and two capacitive Shunt Air Bridge switches [8]. The small size of the scaled Toggle Switch and the short inductive compensation lines allow the design of a SPDT Toggle Switch including 2 Shunt switches with a size of only 1 mm by 1 mm (Fig.9). This requires a narrow routing of the DC electrodes for the Toggle Switches. The capacitive Shunt s, which are used in serial to the Toggle Switch to increase the isolation to the non switched port, can not be actuated as usual.

a) T T: Toggle Switch S: Shunt Switch SPDT DC Electrodes S S Inverse Shunt Toggle Switches Nitride Isolation Layer The current density distribution for the switch state that the signal is routed around the corner from port 1 to port 3 is shown for 2 GHz in Fig. 1. The current density value is visualized by color and height. A difference of 5 db lies between the maximum amplitude and the minimum amplitude of the current density values. The highest current density values are on the cantilever from the Toggle Switch, on the small s and on the ground signal air-bridges. Even the Shunt switches, which are positioned in a distance of about 2 µm from the corner, have both a quite high current density (~ -25 db). The cantilever from the open Toggle Switch in the straight signal path has a high value of induced current close to the bend, but the current density values decrease strongly on the subsequent coplanar center conductor. Behind the closed Shunt switch in the straight signal path the maximum current density value decreased more than 4 db. DC Electrodes db Fig. 9: SPDT Toggle Switch : a) Schematic 3D view of simulation model If a DC voltage would be applied to the center conductor of the coplanar line to actuate the Shunt Air Bridge, the Toggle Switch actuation would be disturbed. By using a common DC ground on the center conductor and applying the switching voltage directly to the membrane from the shunt switch the problem is solved. Therefore the DC connections for these Inverse Shunt switches must be isolated against the ground plane. They are placed above a nitride isolation sheet on top of the ground metallisation to prevent an overlapping with the DC electrodes of the Toggle Switches, which are routed below the ground metallisation. The ground to ground distance of the coplanar line has been reduced in the area of the Toggle Switches to allow a short air bridge connection between the different ground metallisations. Hereby a strong suppression of unwanted higher modes on the coplanar line if the signal is routed around the corner is achieved. A smooth taper in the ground metallisation ensures a good transmission if the signal is routed straight on. The simulation model of this SPDT switch has 14 cells in x-direction, 12 cells in y-direction and with 4 cells the same z-discretisation as the model from the SPST Toggle Switch. Thereby the time step is not changed compared to the SPST simulation model. A total memory of 3 MB is needed and the simulation time is 15 min on a 2.6 GHz Pentium 4 PC. The optimized SPDT structure needs an inductive with a width of 25 µm (Z L =135 Ω) and a length of 18 µm at the input port. The increased length compared to the SPST toggle switch design is necessary because the additional capacity of the crossing air bridge must be compensated. Due to the taper in the ground metallisation no matching is required from the side of port two. A with the length of 7 µm is necessary at the side of port 3. Matching lines Toggle cantilever Matching line Fig. 1: Shunt Switch db Current density at 2 GHz for the SPDT Toggle Switch. The simulation results of the SPDT Toggle Switch (Fig.11) point out excellent performance. The return loss is better than 3 db for frequencies up to 3 GHz while the insertion loss, which is considered as a fixed conductivity at 15 GHz, is below.25 db. The isolation to the off-state port is in the case that the signal is routed around the corner above 45 db for frequencies up to 3 GHz, and in the case that the signal is routed straight on, above 38 db. s11 thru port3 s21 thru port3 s11 thru port2 s31 thru port2-6 -1.2 1 2 3 4. -.2 -.4 -.6 -.8-1. Fig. 11: Simulation results of the SPDT Toggle Switch with actuated Shunt switches.

Substrate modes and parasitic couplings are responsible for the reduction of the isolation for higher frequencies. A thinner substrate of 256 µm would prevent these substrate modes, reduce the couplings and increase the total performance. The inner structure of the manufactured SPDT Toggle Switch is shown in Fig.12. The two inverse serial Shunt switches are not visible. a) s11 thru port3 s11 thru port2. -.5-1. -1.5-2. -2.5 1 2 3 4. Fig. 12: SEM photo of the fabricated SPDT Toggle Switch. Due to a misalignment in the positioning of the nitride sheet in the manufacturing process, a short circuit was created in the DC routing of the Inverse Shunt Air Bridge switches and the Toggle Switches. Nevertheless the Toggle Switches could be actuated by applying a DC switching voltage between 8V and 15 V to the center conductor of the coplanar line. This actuation voltage also attracts the membrane of the Inverse Shunt Air Bride switches, which increases the capacitance in the signal path and reduces the performance of the SPDT switch. The simulation results of this configuration and the corresponding measurement results (Fig. 13) show that even in this not perfect configuration for both signal paths the insertion loss is below.6 db for frequencies up to 3 GHz. Due to the usage of a frequency dependent loss model in the simulation a good agreement is achieved in comparison to the measurements. For both switching states (port 1 to port 2 and port 1 to port 3) the isolation to the non switched port is 2 db at 3 GHz and 3 db at 1 GHz. If the inverse shunt air bridge switches could be closed, the isolation in particular for the higher frequency range would be increased (Fig. 11). The return loss in closed position for both signal paths in simulation and measurement for frequencies up to 3 GHz is below 17 db. The performance of the optimized SPDT Toggle Switch regarding return loss, isolation and insertion loss is comparable with the performance of a SPST Toggle Switch. This is founded in the fact that the same Toggle structure regarding cantilever length and contact area layout was used. The SPDT only needs a longer inductive due to the additional which connects the ground metallisations. Fig. 13: s11 thru port3 s11 thru port2 s21 thru port3 s31 thru port2-2.5 1 2 3 4 SPDT Toggle Switch: a) simulation results measurement results IV. MULTIPORT TOGGLE SWITCH -.5-1. -1.5-2. The multiport Toggle Switch is based on the design of the SPST and SPDT switches and built out of four miniaturized Toggle Switches. All Toggle Switches are rotated 9 against each other and can contact a pad in the crossing of the four coplanar lines (Fig. 14). The ground to ground spacing of the coplanar line is tapered in the region of the crossing from 3 µm to 1 µm. This is possible due to the small dimension of the scaled Toggle Switches (length 15 µm, width 25 µm). Four air-bridges are used as upper limit for the small Toggle Switches and as connection between the ground signal lines. The DC signal lines for the switch pads of each Toggle Switch are routed below the ground signal line of the coplanar line. The 3D FDTD simulator EMPIRE has been used to optimize the taper in the ground metallisation as well as the short inductive compensation lines. For ports 3 and 4, which are routed from the side, an inductive with a width of 25 µm and a length of 7 µm is the optimal solution. The total dimension of this 4 port switch is 5 µm x 5 µm. This enables an easy embedding in switching matrices or in a routing network.

a) Port 4 Signal router parallel to the signal path and route the ground signal. A strong decay in the current density values is achieved on the signal path to the non switched ports. The current density values decrease after a short distance (2 µm) about 5 db. Contact pad Air bridge metal suspension Nitride suspension Toggle Cantilever Nitride suspension Air bridge metal suspension Port 4 DC electrodes Toggle Cantilever Port 4 closed Toggle closed Toggle db db Fig. 14: Multiport Switch: a) Schematic 3D view of simulation model. The current density distribution, for the switch case that the signal is routed around the corner from port 1 to port 3, is shown for 2 GHz in Fig. 15. The current density value is visualized by color and height. A difference of 5 db lies in between the maximum amplitude and the minimum amplitude of the current density values. The highest current density values are on the cantilevers from the closed Toggle Switches while the open Toggle Switches have a low induced current. The current from the outer ground metallisation at the corner causes high current density values on the two connecting air bridges. The two air bridges which connect the two sides of the ground metallisation only have a minor current density, as no strong odd mode has to be shortened on the coplanar line. On the small in the signal path a high current density exists. Fig. 15: closed Toggle Port 4 Current density at 2 GHz for the multiport switch. db db The current density distribution at 2 GHz, for the switch case that the signal is routed straight on from port 3 to port 4, is shown in Fig. 16. The highest current density values exist on the closed Toggle Switches and on the air bridges which are placed in Fig. 16: Current density at 2 GHz for the multiport switch. The simulated insertion loss including frequency dependant metal losses for all switch cases for frequencies up to 5 GHz is below.8 db (Fig. 17). A return loss above 23 db to the routed ports could be achieved while the non switched ports have an isolation above 2 db. This corresponds good with the measurements and simulations of the SPDT switch, where the contact area layout is similar. Due to this and due to the not changed layout of the four single Toggle Switches can be assumed that this Multiport Switch can be realized similar to the step from SPST Toggle Switch to the SPDT Toggle Switch. s31 thru port 3->4 s21 thru port 1->2 s33 thru port3->4 s21 thru port3->4 s11 thru port1->2 s31 thru port1->2-6 -1.2 1 2 3 4 5 Fig. 17: Simulation results of the multiport switch.. This four port switch can be used to reduce the size of large switching matrices where one input port is routed to one output port. The small size of the switch allows an easy integration into the design. In addition to the coplanar design these switch can be converted easily for use in a microstrip environment. Another application is the creation of a switchable delay line array. Due to the small size of the four port switch the additional loss from the switches in the delay line array is very low.. -.2 -.4 -.6 -.8-1.

V. CONCLUSIONS In simulation and measurement the new miniaturized design of the SPST Toggle Switch shows an excellent electrical performance. A good match and low insertion loss for the closed switch state as well as a high isolation for the open switch state could be achieved from DC up to 5 GHz. The DC switch voltages are in the range from 8 V to 15 V for different samples. The small size of only 146 µm length and 25 µm width enabled the design of the compact SPDT switch and of the small four port switch. This four port switch with a size of only 5 µm x 5 µm can easily be used in large switching matrices or routing networks. Advanced FDTD techniques for efficient thin sheet modeling of conductors and dielectrics have been used for MEMS switch EM simulations in good agreement to the measurements. [8] Goldsmith C, Lin T-H, Powers B, Wu W-R and Norvell B.: "Micromechanical membrane switches for microwave applications", Tech. Digest, IEEE Microwave Theory and Techniques Symp., pp. 91-94, 1995 [9] V. K. Varadan, K.J. Vinoy, K.A. Jose: RF MEMS and their applications, ISBN: -47-8438-X, 23 [1] Gabriel M. Rebeiz: RF MEMS Theory, Design, and Technology, ISBN: -471169-3, 23 VI. ACKNOWLEDGEMENT The authors would like to thank K.M. Strohm, T. Mack and J.-F. Luy (Daimler Chrysler AG, Research & Technology Ulm) as well as F. Deborgies and L. Marchand from ESA/ESTEC for many fruitful discussions and J. Mehner (FEM-ware GmbH) for the mechanical simulations. This projects is partially funded by ESA/ESTEC, Contract No. 14547//NL/CK. Reference: [1] B. Schauwecker, K. M. Strohm, W. Simon, J. Mehner, J.-F. Luy: A new type of high bandwidth RF MEMS switch Toggle Switch, Journal of Semiconductor Technology and Science, Special Issue on MEMS, Vol. 2, No. 4, pp. 237 245; December 22 [2] B. Schauwecker, K. M. Strohm, W. Simon, J. Mehner, J.-F. Luy: Toggle-Switch - A new type of RF MEMS switch for power applications ; IMS 22, Vol. 1, pp.: 219-222, Seattle, Washington, USA, 4.6. 6.6. 22 [3] W. Simon, B. Schauwecker, A. Lauer, A. Wien: Designing a novel RF MEMS switch for broadband power applications, Vol. 2, pp. 519 522, European Microwave Conference, Milan, Italy, 24.9. 26.9. 22 [4] B. Schauwecker, K. M. Strohm, W. Simon, J.-F. Luy: RF-MEMS Components for Broadband Applications ; IV. Topical Meeting on Silicon Monolithic Integrated Circuits in RF systems, near Garmisch, Germany, 9.-11. April 23 [5] IMST GmbH, "User and Reference Manual for the 3D EM Time Domain Simulator Empire", http://www.empire.de/empire.pdf, November 23 [6] A. Taflove, Computational Electromagnetics - The Finite Difference Time-Domain Method, Artech House, 1995 [7] Lauer, A.; Wolff, I.: "A conducting sheet model for efficient wide band FDTD analysis of planar waveguides and circuits", IEEE MTT-S, Int. Microwave Symp. Digest, pp. 1589-1592, 1999