PUBLICATIONS. Radio Science. MEMS-based LC tank with extended tuning range for multiband applications RESEARCH ARTICLE 10.

Transcription

1 PUBLICATIONS RESEARCH ARTICLE Special Section: Innovative Microwave Devices, Methods and Applications Key Points: The paper focuses on the implementation of the RF-MEMS in tunable oscillators RF-MEMS varicap and spiral inductor have been used for the design of the LC resonator By using ANSYS Multiphysics environment, the LC resonator was modeled and performance-optimized range Correspondence to: A. Cazzorla, Citation: Cazzorla, A., P. Farinelli, L. Urbani, R. Sorrentino, and B. Margesin (2016), MEMS-based LC tank with extended tuning range for multiband applications, Radio Sci., 51, , doi:. Received 15 MAR 2016 Accepted 9 AUG 2016 Accepted article online 11 AUG 2016 Published online 22 SEP American Geophysical Union. All Rights Reserved. MEMS-based LC tank with extended tuning range for multiband applications A. Cazzorla 1, P. Farinelli 2, L. Urbani 2, R. Sorrentino 1, and B. Margesin 3 1 Department of Engineering, University of Perugia, Perugia, Italy, 2 RF Microtech, Perugia, Italy, 3 Fondazione Bruno Kessler, Trento, Italy Abstract This paper presents the modeling, simulations, and measurements of a compact multiband microelectromechanical (MEMS)-based LC tank resonator suitable for low phase noise voltage-controlled oscillators (VCOs). The resonator is based on a high-q spiral inductor and high capacitance ratio varicap fully integrated in FBK-irst (Fondazione Bruno Kessler) MEMS manufacturing process. The design of the varicap is based on double-actuation mechanism with a mechanical central bond that inhibits the pull-in allowing for a theoretically infinite tuning ratio. The measurements have shown a total not continuous capacitance ratio (C r ) of 5.2 with a continuous variation of the capacitance values in the range 225 ff 600 ff which corresponds to a continuous capacitance ratio (C r *) of 2.6. The performance repeatability, the power-handling capability, and the stability over time were tested on 10 samples showing a negligible variation of the capacitance values. The spiral inductor consists of a suspended gold membrane thick 5 μm in a circular shape which was modeled in order to optimize the quality factor (Q) in the frequency range 2 4 GHz. The measurement results show a Q of about 55 in the 2 4 GHz frequency band. The LC tank measurements show an overall tuning range better than of 45% in the GHz frequency band, consisting of two continuous tuning ranges of 7.5% and 25%. The LC tank allowed the design of MEMS-based voltage-controlled oscillators (VCOs) with an overall tuning better than 60% in the frequency range 2.15 GHz 3.85 GHz and two separate regions of continuous tuning range. The VCO prototype will be fabricated on Surface Mount Technology on RO4350 laminate. The main figures of merit are presented in comparison with the state of the art. 1. Introduction With the recent developments in the wireless communication industry, the demands for radios covering the whole frequency spectrum of fixed and mobile WiMAX, and IEEE a/b/g/n, is desired together with minimum hardware resources and costs [Sadhwani et al., 2010; Guo and Xu, 2014]. In many RF transceiver systems, the voltage-controlled oscillators (VCOs) are key components since they are the sources of the reference oscillation frequency. In conventional harmonic VCO, oscillation frequency is basically determined by the resonant circuit (typically an LC circuit called LC tank) and the frequency tuning can be achieved by varying the voltage-dependent capacitance of the varicap element in the LC tank. Low phase noise and wide tuning range are the main requirements, imposed by the market, for the design of a VCO to be used in new generation front-end circuits. As reported in literature [Steer, 2009], the phase noise of a VCO depends on many factors like oscillator output power, output oscillation frequency, operating temperature, and quality factor (Q) of the LC resonant circuit. The latter is considerably limited by the loss of the passive elements, such as on-chip spiral inductors [Ibrahim and Kuhn, 2000], as well as substrate and conductor losses. In addition, the use of on-chip spiral inductors is limited in frequency by their self-resonance frequency (SRF). At self-resonance, the coupling capacitance between inductor eddy current and the substrate forms a parallel resonance [Bahl, 2003]. This is a limiting factor when high inductance values are required. In a recent work, [Jeong et al., 2015] the modeling and measurements of an integrated spiral inductor on 90 nm CMOS (Complementary Metal-Oxide Semiconductor) technology have been presented, showing quality factor lower than 14 at 10 GHz. Regarding the tunable capacitors, many approaches have been recently proposed for increasing the Q factor and the capacitance ratio of CMOS varactors that still remain pretty poor. In Oh and Rieh [2011] a high-q island-gate varactor suitable for millimeter-wave frequencies has been presented. The device was fabricated in standard CMOS 200 nm process, and it shows a Q factor of 12 at 24 GHz together with a capacitance ratio lower than 2. In Quemerais et al. [2015] a MOS varactor on the low-power CMOS 28 nm fully depleted silicon on insulator has been presented. The measured Q factor is about 16 at 24 GHz and the capacitance ratio lower than 2. CAZZORLA ET AL. MEMS-BASED LC TANK 1519

2 Figure 1. MEMS-based LC resonator: (a) 3-D model and (b) cross section. An alternative to conventional CMOS fabrication process is the microelectromechanical (MEMS) technology [Rebeiz, 2003] which allows one to achieve highly miniaturized LC resonator with high-q performance. In this case, the LC resonators can be realized by combining a MEMS capacitor (varactor [McFeetors and Okoniewski, 2006; Hui Pu et al., 2010; Mahameed and Rebeiz, 2010] or switch [Kaynak, 2014]) and a planar [Heves et al., 2008] or micromachined inductor [Rahimi et al., 2010] or fixed MIM capacitor with switched inductance [Gaddi, 2005]. For high-frequency applications, the micromachined inductor quality factor is the bottleneck of the overall quality factor (Q) of the LC tank [Ketterl et al., 2001]. The inductive transmission line [Yu et al., 2005] or cavity resonators [Mercier et al., 2003; Stefanini et al., 2011] are preferable for the high-q factors with respect to spiral or lumped inductors. This paper presents a deep analysis of a compact multiband MEMS LC series resonator based on a MEMS spiral inductor and a MEMS varicap reported in Cazzorla et al. [2015a]. In order to increase the inductor Q factor, the HR Silicon substrate was partially removed reducing the substrate loss and parasitic couplings. The LC tank allows one to design VCOs (voltage-controlled oscillators) that can simultaneously operate in different frequency bands, i.e., IEEE b/g (WLAN) and IEEE d (WiMAX) standards. The modeling and measurement results of the MEMS varicap and of high-q MEMS spiral inductor are presented. The performance repeatability and preliminary reliability tests of the MEMS varicap are reported together with details of the fabrication and packaging FBK (Fondazione Bruno Kessler) processes. The measurement results of the MEMS-based LC tank are presented showing an overall tuning range of about Figure 2. MEMS varicap: (a) 3-D model, (b) cross section, and (c) prototype. CAZZORLA ET AL. MEMS-BASED LC TANK 1520

3 Figure 3. Principle of operation: (a) initial state, (b) upstate, and (c) downstate. 45% (not continuous) and two continuous tuning of about 7.5% and 25%. A preliminary characterization of the MEMS-based voltage-controlled oscillator (VCO) accounting for the additional bonding wire connections is also presented. The VCO prototype is under fabrication in Surface Mount Technology (SMT) on RO4350 laminate. Finally, the main figures of merit (FOM) are also compared with the state of the art. 2. Resonator Geometry Figure 1 shows the proposed LC resonator. It consists of the series of a spiral inductor and a varactor, both designed to be fabricated by using the FBK-irst MEMS process [Giacomozzi et al., 2011]. The bonding wire interconnections with the active part of the VCO have been also included in the model since they are the main responsible of the shift of the resonance frequency of the circuit. Such kind of circuit resonates when the magnetic energy stored in the inductance (L r ) equals the electric energy stored in the capacitance (C r ), i.e., at the frequency 1 F 0 ¼ p 2π ffiffiffiffiffiffiffiffi (1) L r C r The quality factor of a series LC resonator is defined as Q ¼ 2πF 0L r 1 ¼ (2) R r 2πF 0 R r C r where R r represents the total resonator loss, consisting of the conductor loss, the substrate loss, and the inductor and varactor loss. Figure 4. Simulated (a) static deflection as function of the applied voltage on L.E and C.E, (b) static deflection in the threestate mode, and (c) average pressure on mechanical stoppers close to the C.E. electrodes. CAZZORLA ET AL. MEMS-BASED LC TANK 1521

4 2.1. MEMS Varicap Figure 2 shows the layout, cross section, and prototype of the MEMS varicap that has been used in the LC tank. It consists of a gold membrane (1250 μm long and 200 μm wide) mechanically anchored at the CPW (Coplanar Waveguide) ground (by six identical lever springs, long L a and wide w a ) and at the substrate. Four identical lever springs (long L b and wide w b ) were designed in the central part of the membrane to prevent the membrane snapdown as explained in the following. The movable gold membrane is 2 μm Figure 5. RF-MEMS varicap: equivalent RLC circuit. thick, locally reinforced with a second deposition of 3.5 μm thick gold layer ensuring high beam flatness. The proposed approach, as proposed in Cazzorla et al. [2015b], is based on a double-actuation mechanism. To this purpose, the DC electrodes have been split into two separate electrodes, called L.E. (lateral electrodes) and C.E. (central electrodes). When the device is in its initial state (Figure 3 a), the residual gap (g 0 ) between movable membrane and RF electrode is 2.7 μm. When the L.E. are polarized, the bridge lateral parts snap on the L.E., reducing the gap from g 0 to g up ( upstate, Figure 3b). During this first actuation, the mechanical bonds in the center inhibit the snapdown of the membrane central part, which remains suspended above the RF line (g up ). Afterward, the central electrodes (C.E.) are polarized to continuously reduce the gap up to g down,( downstate, Figure 3c) allowing for a continuous capacitance variation. Mechanical stoppers were integrated in the actuation pads first to avoid the contact between movable membrane and the dielectricless actuation electrodes and second to avoid the membrane snapdown when C.E. are polarized. They are made by the stuck of the thin metal and insulating layers available in the standard process [Giacomozzi et al., 2011], such as polysilicon, TiN/Ti/Al/Ti/TiN multimetal, and gold for a total height of about 1.2 μm. The varactor was modeled in the FEM ANSYS Multiphysics ( accounting for residual stresses and nonlinear behavior of the lever springs at standard temperature (T = 25 C). First, the static deflection of the movable membrane was analyzed, taking into account the initial deformation of the membrane due to residual stress, Figure 4a. At 0 V the membrane is downward bent of about 200 nm due to the residual stress in the gold layer. Then, by applying DC control voltage (up to 40 V) on the L.E., the initial gap (g 0 ) decreases down to g up ( 1.1 μm). Note that as in conventional parallel plate capacitors, after one third of the initial gap the Figure 6. RF-MEMS varicap: comparison between measured and simulated performance. (a) Return loss and (b) insertion loss as a function of the applied voltage on the DC electrodes. CAZZORLA ET AL. MEMS-BASED LC TANK 1522

5 Table 1. Comparison Between Simulated and Measured RLC Parameters Simulated Measured C var (ff) initial state C var (ff) upstate C var (ff) downstate R s (Ω) L s (ph) C sub (ff) R sub (kω) membrane collapses due to the pullin effect (simulated V pull-in = 37.5 V). The complete membrane snapdown above the RF line is avoided thanks to additional mechanical bond in the central part of movable membrane. Then, by applying a DC signal (up to 60 V) on the C.E., the gap g up is continuously decreased up to g down (0.45 μm) allowing for continuous and high tuning of the MEMS capacitance. In this case, the movable membrane touches the mechanical stoppers (placed close to the central electrodes) before starting pull-in behavior (DC signal on central electrodes equal to 30 V). Then by increasing the voltage, the membrane continues to displace assuming a zipping motion. The varactor displacement along the z axis is shown in Figure 4 b for the three states (initial state, upstate, and downstate). Figure 4c shows the simulated average pressure on mechanical stoppers placed closed to the central electrodes. The device was modeled in ANSYS HFSS ( It is placed in shunt configuration with respect to a 2 mm long RF CPW line and analyzed in the 0 30 GHz frequency range. In its initial state, the device presents a shunt capacitance C var = ff. In its upstate and downstate, the simulated shunt capacitances are ff and ff, respectively. The resulting theoretical noncontinuous capacitance ratio (C r ) is about 6, whereas the continuous tuning ratio (C r *) is about 2.6. These values were obtained by fitting the full-wave simulations with the equivalent circuit of Figure 5. The on wafer RF characterization was performed by using Agilent N5230A VNA, CPW probes, and SOLT (Short-Open-Load-Thru) calibration in the 0 30 GHz frequency band. Return loss and insertion loss as a function of the applied voltage are shown in Figure 6. In its initial state (0 V), the varicap shows good transmission properties, featuring return loss better than 25 db and insertion loss better than 0.15 db up to 5 GHz. Table 1 shows the comparison between the simulated and measured values of the varactor equivalent RLC circuit. The results are in a good agreement with the simulations, showing a measured not continuous capacitance ratio (C r ) of 5.2 and continuous tuning ratio of 2.6. The continuous tuning range was measured by increasing the applied voltage at fixed steps of 5 V on L.E. and 1.5 V on the C.E. The return loss as a function of the applied voltage is shown in Figure 7a. As expected, the varactor shows two different continuous tuning ranges. The first one, before the pull-in (V pull-in = 35 V), allows a continuous tuning ratio of 1.6. The second one, after the pull-in (L.E. = 40 V), allows a continuous variation of the signal up to 2.6. The capacitance values as a function of the applied voltage are reported in Figure 7b. Note that after the first actuation (L.E. are polarized), the membrane is very close to the CPW RF line and small gap variations allow for increased tuning. Figure 7. RF-MEMS varicap: (a) return loss as function of the voltage applied on DC electrodes and (b) capacitance as function of the voltage applied on DC electrodes, when it is increased and decreased. CAZZORLA ET AL. MEMS-BASED LC TANK 1523

6 Table 2. Repeatability of the Performance on 10 (#) Samples Measurement Results #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 C r C r * V pull-in V pull-out The performance repeatability has been also verified on 10 samples. Measured pull-in, pull-out voltages as well as the capacitance ratios are reported in Table 2. The results show good RF and mechanical performance reproducibility, resulting in a dispersion of about ±5% and ±4% for the noncontinuous capacitance ratio and continuous capacitance ratio, respectively. In order to verify the pull-in/pull-out voltage shift and the capacitance variation after large number of actuations, three samples were cycled up to 100 million cycles. In the first case, the test was performed by using the Agilent 33120a (plus commercial DC-DC converter) function generator, generating a 1 KHz, 50% duty cycle, square waveform with a peak value equal to the measured pull-in voltage of the samples which was applied at the lateral electrodes (L.E.), as reported in Figure 8a. The pull-in and pull-out voltages at steps of 10 million cycles are shown in Figure 8b. A negligible variation of the voltage pull-in and pull-out values was measured resulting in a voltage shift of about +3.5 V up to 100 million cycles. The cycling test was repeated by using the same duty cycle, square waveform with a peak value equal to 60 V applied to the C.E. while the L.E. were set to 55 V, as shown in Figure 9a. Figure 9b shows the capacitance (C var ) as a function of the applied voltage at steps of 10 million cycles. A negligible variation of few tens of ff was recorded, resulting in a variation of about 3% and 1% for the noncontinuous and continuous capacitance ratio. The comparison between measured capacitances before and after 100 million cycles is thus shown in Figure 10. Figure 8. RF-MEMS varactor: (a) applied waveform and (b) measured pull-in and pull-out voltages t up to 100 million cycles. Figure 9. RF-MEMS varactor: (a) applied waveform and (b) measured down (in blue) and upstate (in red) capacitance over time (up to 100 million cycles). CAZZORLA ET AL. MEMS-BASED LC TANK 1524

7 Figure 10. MEMS varicap: comparison between measured capacitance values before and after 100 million of cycles. The varactor robustness to self-biasing was then preliminary tested by applying an equivalent DC voltage V RMS on the central RF line. The proposed test does not account for the dissipated RF power that can increase the temperature of the movable membrane, relaxing the stress and further decreasing the biasing voltage. Still, the test is useful to preliminary evaluate the varactor robustness to self-actuation. Table 3 shows the measured capacitance values and capacitance ratios as function of the input power and root-mean-square voltages (V RMS ). The device shows negligible variation in the continuous capacitance ratio (C r *) up to 35 dbm. For higher input power, the device collapses onto the RF line. The device can operate as a classical MEMS capacitance switches up to 40 dbm. In this case, the noncontinuous capacitance ratio (C r ) increases up to 10, since the membrane better relaxes onto the fixed electrode MEMS Inductor Figure 11 shows the proposed spiral inductor (3-D model and cross section) in series configuration with respect to a 50 Ω coplanar line. It consists of a 5 μm thick gold suspended membrane in circular shape (external diameter D ext of about 440 μm). D int is the internal diameter of the structure. S and w are the spacing among Table 3. Measured Capacitance Values and Capacitance Ratio for Increased Values of V RMS V RMS (V) P in (dbm) V RMS (V) C var a (ff) Cvar b (ff) Cvar c (ff) Cr C r * dbm dbm dbm dbm dbm dbm a Initial state. b Upstate. c Downstate. Figure 11. MEMS inductor: (a) top view and (b) cross section. CAZZORLA ET AL. MEMS-BASED LC TANK 1525

8 turns and the width of the wire conductor, and they are 15 μm and 10 μm, respectively. The spiral inductor is monolithically integrated in the FBK CMOS-MEMS process together with the MEMS varicap described above. Its equivalent π model circuit is shown in Figure 12. L ind is the inductance value, and R ind and C coup account for the losses in the conductor and the coupling capacitance among spirals. C sub Figure 12. MEMS inductor: equivalent π model circuit. and R sub are the parasitic capacitance and the resistivity of the silicon substrate, respectively. C air is the parasitic capacitance between suspended structure and the substrate (g 0 = 2.7 μm). Mechanical analysis was performed in ANSYS Multiphysics environment in order to predict the suspended inductor sensitivity to residual manufacturing stresses. Figure 13a shows the delta displacement (Δ-z) along z axis just after the release of the structure. The suspended conductor is upwarped of 0.1 μm and downwarped of 6 μm, resulting in an unpredictable inductance value, due to the changes in the coupling capacitance (C coup ) between spirals. In order to ensure high conductor flatness, mechanical square pillars (15 μm 15μm), equally spaced at the distance of about 60 μm, were inserted in the design. Figure 13b shows the resulting delta displacement (Δ-z) along z axis in case of mechanical pillars. Afterward, the inductor with mechanical pillars was modeled in ANSYS HFSS full-wave environment in the 0 30 GHz frequency range. The main responsible of the low Q performance was found to be the loss in the thin metal below suspended conductor, named underpass in Figure 11b. In order to maximize the quality factor (Q) and the self-resonance frequency (SRF) of the device, two different solutions were implemented: the first one was to increase the thickness of the underpass metal from the standard value of 0.6 μm upto 1.2 μm and the second one was to partially remove the silicon substrate reducing the substrate loss and the parasitic coupling with the inductor. The substrate removal will be done by deep reactive-ion etching (DRIE) that allows one to create deep penetration with high aspect ratio in the substrate. Simulated S parameters and inductance (L) are shown in Figure 14a, whereas Figure 14b presents the simulated equivalent series resistance (R) and the quality factor (Q). These parameters were calculated by using the following equations: L ind ðfþ ¼ R ind ðfþ ¼ re im 1 Y 12 2πF 1 Y 12 (3) (4) Figure 13. MEMS inductor: simulated delta displacement just after the release of the structure (a) without and (b) with pillars. CAZZORLA ET AL. MEMS-BASED LC TANK 1526

9 Figure 14. MEMS-based LC resonator: simulated (a) S 12, L ind and (b) Q, R ind of the RF-MEMS spiral inductor. QF ð Þ ¼ L indðfþ *2πF (5) R ind ðfþ The device exhibits a theoretical inductance of about 5.6 nh up to 10 GHz and a self-resonance frequency (SRF) of about 24 GHz. Up to 5 GHz, the resistance is in the range 1 Ω 2.7 Ω, resulting in a Q higher than 60 at 3 GHz. By using the equivalent π model circuit in Figure 12, the S parameters were fitted in Advanced Design System (ADS) circuit environment and equivalent parameters were extracted resulting in an Figure 15. MEMS inductor: (a) prototype and (b) comparison between measured and simulated S 12. Figure 16. Simulated full-wave (line) and circuital (dash line) admittance (real part) of the MEMS-based LC tank. CAZZORLA ET AL. MEMS-BASED LC TANK 1527

10 Table 4. Simulated RLC Parameters Initial State Upstate Downstate C r (ff) L r (nh) R r (Ω) inductance (L ind ) of about 5.8 nh, an equivalent series resistance (R ind ) of 1.6 Ω and a coupling capacitance (C coup ) of about 8.8 ff. The device has been manufactured and tested in the 0 30 GHz frequency band. A photo of the prototype and the comparison between measured and simulated S 12 are shown in Figure 15. The measurements are in a good agreement with the simulated ones showing an inductance of about 5.8 nh ± 0.3 nh in the 0 10 GHz frequency band and a calculated Q of about 55 in the 2 4 GHz frequency band. 3. Fabrication The devices were fabricated by using the standard FBK RF-MEMS process with an added bulk etching module for the realization of the thin silicon membranes. The baseline process combines the standard CMOS process elements of the planar technology, as used for CMOS devices, with surface micromachining techniques to build suspended mobile structures. The technology, originally developed for high-reliability RF-MEMS switches for space applications, is able to produce complex RF circuits including MEMS switches and highquality passive components in large quantities and at low cost. The main structural elements that characterize the process are a 2 μm and a 3.5 μm thick galvanic gold layer that provides the high-conductivity transmission lines and, in combination with the sacrificial layer, the mobile suspended structures used to build the electrostatically actuated switches [Giacomozzi et al., 2011]. For the construction of the silicon membranes a process module has been added that consists in an aluminum hard mask on the wafer backside followed by a timed silicon etch with a DRIE process that has been carefully tuned in order to leave a 50 μm thick silicon membrane. Figure 17. MEMS LC tank: (a) prototype and (b) measured admittance (real part) as function of the capacitance tuning of the varicap. The devices presented in this paper are part of a Multi Project Wafer run. On each wafer 32 repetitions of the MEMS resonator were present. In order to reduce the stress gradient in the structural gold layers, the devices were released in an oxygen plasma at 80 C in 6 h. This keeps the out of plane movements of highly nonconstrained structures like the toggle-type varactor to a minimum [Mulloni et al., 2010]. About 10 devices were individually covered with a quartz cap. The caps were fabricated on a 400 μm thick double side polished quartz wafer. To this purpose, the quartz wafers were first laminated with a 55 μm thick dry film, ORDYL SY 355. The wafers so prepared were then exposed with a dedicated lithography mask that was designed with many different sized sealing rings in order to optimize the process. A typical sealing ring thickness is 100 μm. From such wafers individual caps were obtained by dicing. The caps have then been applied to the dies using a Semi auto TRESKY T3000 FC3 die bonder using a force of 550 g at 100 C for 30 min and 150 C for 30 min in dry nitrogen flow. CAZZORLA ET AL. MEMS-BASED LC TANK 1528

11 Table 5. Comparison With the State of the Art Yu et al. [2005] Stefanini et al. [2011] Our Work TR (%) 75 a 15 a a Q Central frequency (GHz) Device size (mm 2 ) Year a Noncontinuous tuning ratio. Shear tests performed on a few devices measured adhesion forces of 14.2 MPa, indicating a good adhesion to the substrate. More details on the process can be found in Giacomozzi et al. [2014]. 4. MEMS-Based LC Tank Finally, the MEMS-based series LC resonator was modeled in ANSYS HFSS full-wave environment in the 1 5 GHz frequency range. The bonding wires were modeled to account for the interconnection with the voltage-controlled oscillator (VCO) active circuitry. The simulated admittance (real part), as a function of the capacitive tuning of the MEMS varicap, is shown in Figure 16. The LC resonator plus bonding wire connection shows an overall tuning range higher than 60% in the frequency range GHz. Two separate regions of continuous tuning are obtained thanks to the two separate continuous tuning regions of the MEMS varicap. The first one is centered at ~3.65 GHz (WiMAX applications) and allows for a continuous tuning up to 9.5% ( GHz). The second one allows for a continuous tuning between 2.15 and 2.95 GHz, (30%) ensuring to cover all the frequency bands of WLAN and Bluetooth applications. Table 4 shows equivalent RLC parameters obtained by fitting the simulation results with its equivalent circuit. The values refer to the three-state mode of the MEMS varicap. Note that the resulting inductance (L r ) is slightly higher than the simulated value of the simple MEMS inductor (7 nh and 5.6 nh) due to the additional inductance of the bonding wires. A photo of the prototype as well as the measured admittance (real part) as a function of the capacitive tuning of the MEMS varicap is shown in Figure 17. Note that since the bonding wires are not included in the measurements, the resonance frequency is slightly higher than the final one. The measurements show that the resonator allows for an overall tuning range higher than 45% in the frequency range GHz and the two separate regions of continuous tuning are GHz with tuning ratio of 7.5% and GHz, i.e., tuning ratio of about 25%. The quality factor for the proposed prototype was calculated to be about 50. The slight reduction of the tuning range is most likely due to 0-level cap which has limited the tuning capability of the MEMS varicap with respect to the unpackaged device. Table 5 shows the comparison between the proposed MEMS-based LC tank and the state of the art. Figure 18. Simulated substrates distance for tolerances analysis Tolerances Analysis on Bonding Connection Tolerance analysis was performed in order to define a robust design of the LC tank by varying the distance (called S in Figure 1a) between substrates (HR-Si for resonator and Roger T-Duroid 4350) as function of the wire bonding. CAZZORLA ET AL. MEMS-BASED LC TANK 1529

12 S was swept from 50 μm to 300 μm and the LC resonator simulated in HFSS environment. The wires introduce a series of inductance, which can affect the frequency range of the resonator, Figure 18. Their length was considered as a function of the distance between substrates. The 80 MHz frequency shift was simulated considering the resonator in its initial state. Theoretically, the design is very robust and less sensitive with respect to the wire bonding connections. Figure 19. MEMS-based VCO: (a) simplified circuit and (b) final board layout with chip-on-board wire bonding connection. 5. Preliminary Analysis of MEMS-Based VCO A preliminary design of MEMS-based voltage-controlled oscillator (VCO) has been performed in Advanced Design System (ADS ) ( environment. The proposed oscillator is a single-ended Clapp topology and was designed to be manufactured in Surface Mount Technology (SMT) on RO4350 laminate (dielectric permittivity ε r = 3.66, substrate thickness t = 762 μm, and metal thickness t m =35μm). The simplified schematic is reported in Figure 19a. The circuit consists of the series MEMS-based LC tank, an RF transistor NPN Bjt (Infineon BFR740L3RH), and the associated feedback capacitors C1 and C2. Chip-on-board wire bonding technique will be used to connect the MEMS-based LC tank to the active part of the proposed VCO. Additional resistors, inductors, and capacitors have been inserted in the final board design in order to realize the required biasing and DC coupling/decoupling networks, Figure 19b. For the circuital analysis, SPICE models of the RF transistor and SMD passive devices were used. The MEMSbased LC tank was substituted by its equivalent RLC circuit accounting for the wire bonding connection. The simulated VCO's phase noise (PN) and the output power across the tuning range are reported in Figure 20 resulting in a PN lower than 126 dbc/hz at 1 MHz and an output power of about 0 d Bm when the resonator is in its initial state (0 V applied on DC electrodes). Note that in Figure 20b a harmonic at 6.25 GHz was recorded with an output power lower than 25 dbm. In Table 6 is reported a comparison of the performance with the state of the art, considering MEMS-based and CMOS solutions. Figure 20. MEMS-based VCO: simulated (a) PN and (b) output power at initial state. CAZZORLA ET AL. MEMS-BASED LC TANK 1530

13 Table 6. Comparison With the State of the Art VCO Bhattacharya et al. [2015] Tseng et al. [2007] Guo et al. [2011] Aqeeli and Hu [2013] Heves et al. [2008] Our Work MEMS a Technology CMOS 0.13 μm + MEMS L CMOS 0.18 μm CMOS 0.18 μm CMOS 0.35 μm Bi-CMOS 0.35 μm SMT + MEMS LC + MEMS L ONLY ONLY + MEMS C PN at 1 MHz (dbc/hz) Tuning range (%) % (9.5% and 30%) b Output power (dbm) F 0 (GHz) Supply voltage (V) Power consumption (mw) FOM (dbc/hz) Transistor technology On Die On Die On Die On Die On Die SMD Year a Simulation results. b Continuous tuning. 6. Conclusions This paper presented the modeling, simulations, and measurements of a compact multiband MEMS-based LC tank resonator suitable for low phase noise voltage-controlled oscillators (VCOs). The resonator is based on high-q spiral inductor and high capacitance ratio varicap fully integrated in FBK-irst MEMS process. The varicap is based on double-actuation mechanism and shows measured not continuous capacitance ratio (C r )of 5.2 and continuous ratio (C r *) of 2.6. The performance repeatability and the power-handling capability were reported showing good robustness and negligible variation of the capacitance up to 35 dbm equivalent input power. In addition, the stability over time ( 23 h) has been tested, showing negligible variation of the capacitance values. The measurement results of the MEMS spiral inductor have been presented, and they are in a good agreement with the simulated ones, showing Q factor of about 55 in the 2 GHz 4 GHz frequency band. The proposed LC tank resonators demonstrated an overall tuning range better than of 45% in the GHz frequency band and two continuous tuning ranges of 7.5% and 25%. A preliminary design of the MEMSbased voltage-controlled oscillator (VCO) has been reported. The LC tank allows for an overall tuning better than 60% in the frequency range 2.15 GHz 3.85 GHz and two separate regions of continuous tuning which could be used to cover the whole frequency spectrum of WiMAX and IEEE802.11a/b/g/n. The VCO prototype will be fabricated on Surface Mount Technology (SMT) on RO4350 laminate. In the meanwhile, the main figures of merit (FOM) have been presented in comparison with the state of the art. Acknowledgments The work has been carried out in the framework of the ESA Project AO/1-6730/11/NL/GLC. The authors thank Francois Deborgies from ESA/ESTEC for their support and suggestions. The data used in this paper can be made available upon request to the following author (alessandro.cazzorla@studenti.unipg.it). References Aqeeli, M., and H. Hu (2013), Design of a High Performance 5.2 GHz Low Phase Noise Voltage Controlled Oscillator Using 90nm CMOS Technology, in International Conference on Computer Science and Electronics Engineering (ICCSEE), Hangzhou, P. R. China, March. Bahl, I. (2003), Lumped Elements for RF and Microwave Circuits, edited by Artech House, Norwood, Mass. Bhattacharya, A., D. Mandal, and T. K. Bhattacharyya (2015), A GHz 3.1-mW VCO using electro-thermo- mechanically tunable self-assembled MEMS inductor on HR substrate, IEEE Trans. Microwave Theory Tech., 62(2), Cazzorla, A., R. Sorrentino, and P. Farinelli (2015a), MEMS based LC tank with extended tuning range for multi-band applications, in Microwave Symposium (MMS), 2015 IEEE 15th Mediterranean, Lecce, 2015, pp Cazzorla, A., P. Farinelli, and R. Sorrentino (2015b), Double-actuation extended tuning range RF MEMS varactor, in Microwave Conference (EuMC), 2015 European, Paris, 2015, pp Gaddi, R. (2005), Reconfigurable MEMS-enabled LC-tank for multi-band CMOS oscillator, in IEEE MTT-S International Microwave Symposium Digest, 2005, pp. 4. Giacomozzi, F., V. Mulloni, S. Colpo, J. Iannacci, B. Margesin, and A. Faes (2011), A flexible technology platform for the fabrication of RF-MEMS devices, in International Semiconductor Conference (CAS), vol.1, pp , Sinaia, Romania, Oct. Giacomozzi, F., F. Mattedi, P. Farinelli, B. Margesin, G. Resta, and V. Mulloni (2014), Assessment of ORDYL SY 355 dry film for RF MEMS 0-level packaging, in 15th International Symposium on RF-MEMS and RF-Microsystems MEMSWAVE 2014, La Rochelle, France, 30 June-2 July. Guo, C., J. Hu, S. Zhu, H. Sun, and H. Lv (2011), A 5 GHz low-phase-noise CMOS LC-VCO for China ETC applications, in Microwave Technology & Computational Electromagnetics (ICMTCE), 2011 IEEE International Conference on, pp , Beijing. Guo, J., and Z. Xu (2014), Design of miniature tri-band monopole antenna for WLAN and WiMAX applications, in Electronic Packaging Technology (ICEPT), th International Conference on, Chengdu, 2014, pp Heves, E., I. Tekin, and Y. Gurbuz (2008), A MEM-varactor tuned, 7.8 GHz differential LC voltage-controlled oscillator, Sens. Actuators, A, 144(2), Hui Pu, S., A. S. Holmes, E. M. Yeatman, C. Papavassiliou, and S. Lucyszyn (2010), Stable zipping RF MEMS varactors, J. Micromech. Microeng., 20, Ibrahim, W. B., and N. M. Kuhn (2000), Analysis of current crowding effects on multiturn spiral inductors, IEEE Trans. Microwave Theory Tech., 49, CAZZORLA ET AL. MEMS-BASED LC TANK 1531

14 Jeong, J. W., S. K. Kwon, J. N. Yu, S. Y. Jang, S. H. Oh, C. Y. Kim, G. W. Lee, and H. D. Lee (2015), Modeling of T-model equivalent circuit for spiral inductors in 90 nm CMOS technology, in Proceedings of the 2015 International Conference on Microelectronic Test Structures, pp , Tempe, Ariz. Kaynak, M. (2014), RF-MEMS Switch Module in a 0.25 μm SiGe:C BiCMOS Process, PhD dissertation. Ketterl, T., T. Weller, and D. Fries (2001), A micromachined tunable CPW resonator, in Microwave Symposium Digest, 2001 IEEE MTT-S International, vol.1, pp , Phoenix, Ariz. Mahameed, R., and G. M. Rebeiz (2010), Electrostatic RF MEMS tunable capacitors with analog tunability and low temperature sensitivity, in Microwave Symposium Digest (MTT), 2010 IEEE MTT-S International, pp , Anaheim, Calif. McFeetors, G., and M. Okoniewski (2006), Performance and operation of stressed dual-gap RF MEMS varactors, in 2006 European Microwave Conference, pp , Manchester, U. K. Mercier, D., et al. (2003), A micromachined tunable cavity resonator, in Microwave Conference, rd European, pp , Munich, Germany. Mulloni, V., F. Giacomozzi, and B. Margesin (2010), Controlling stress and stress gradient during the release process in gold suspended micro-structures, Sens. Actuators, A, 162(1), Oh, J., and J. S. Rieh (2011), A comprehensive study of high- Q Island-Gate Varactors (IGVs) for CMOS millimeter-wave applications, IEEE Trans. Microwave Theory Tech., 59(6), Quemerais, T., D. Gloria, D. Golanski, and S. Bouvot (2015), High-Q MOS varactors for millimeter-wave applications in CMOS 28-nm FDSOI, IEEE Electron Device Lett., 36(2), Rahimi, M., S. S. Jamuar, M. N. Hamidon, M. R. Ahmad, S. A. Mousavi, and M. Bayat (2010), An optimized 2.4 GHz VCO circuit design and simulation with high-q MEMS LC-tank, Int. J. Electron. Commun., 64(5), Rebeiz, G. M. (2003), RF MEMS: Theory, Design, and Technology, Wiley, Hoboken, N. J. Sadhwani, R., A. B. Bassat, A. A. Kidwai, and S. Rivel (2010), Multi-band multi-standard local oscillator generation for direct up/down conversion transceiver architectures supporting WiFi and WiMax bands in standard 45nm CMOS process, in 2010 IEEE Radio Frequency Integrated Circuits Symposium, pp , Anaheim, Calif. Steer, M. (2009), Microwave and RF Design: A Systems Approach, SciTech Inc., Raleigh. Stefanini, R., J. D. Martinez, M. Chatras, A. Pothier, V. E. Boria, and P. Blondy (2011), Ku Band high-q tunable surface-mounted cavity resonator using RF MEMS varactors, IEEE Microwave Wireless Compon. Lett., 21(5), Tseng, S. H., Y. Hung, Y. Z. Juang, and M. Lu (2007), A 5.8 GHz VCO with CMOS-compatible MEMS inductors, Sens. Actuators, A, 139(1-2), Yu, A. B., A. Q. Liu, and Q. X. Zhang (2005), Tunable MEMS LC resonator with large tuning range, IEEE Electron. Lett., 21, CAZZORLA ET AL. MEMS-BASED LC TANK 1532