A) Classification

Capacitive switches are mainly divided into two sub-categories depending on their different functionalities; the RF-MEMS capacitive switch or analog MEMS capacitor, with high isolation and an increased capacitance ratio, Cr (>30) and the MEMS switch capacitor or digital MEMS capacitor with a moderate and usually defined Cr < 5. Analog MEMS capacitors can be used for signal routing purposes, in reconfigurable front-ends, digitized capacitor banks, and reconfigurable antenna orientation or apertures. In order to achieve high isolation, many different configurations have been reported based on shunt and series switches [Reference Rebeiz1–Reference Muldavin and Rebeiz8], low spring constant anchoring [Reference Peroulis, Pacheco, Sarabandi and Katehi3], inductive behavior [Reference Muldavin and Rebeiz8], and/or using high dielectric constant materials [Reference Park, Kim, Chung and Bu4, Reference Ziaei, Dean and Mancuso6, Reference Fernández-Bolaños, Tsamados, Dainesi and Ionescu7].

On the other hand, the digital MEMS capacitor is generally used as part of tunable systems or reconfigurable networks, for example filters, phase shifters, or impedance matching with two known capacitive states. The concept of the digital capacitor was first introduced by Hayden and Rebeiz [Reference Hayden and Rebeiz9] and is defined as the standard MEMS capacitive switch in series with a fixed small capacitance possibly realized with metal–insulator–metal (MIM) capacitor [Reference Fernández-Bolaños, Lisec, Dainesi and Ionescu10], metal–air–metal (MAM) capacitor [Reference Perruisseau-Carrier, Fritschi, Crespo-Valero and Skrivervik11], or interdigitated capacitors. Figure 2 illustrates for comparison purposes the capacitance versus voltage curve (CV-curve) of a standard analog capacitor and a digital capacitor made with MIM fix series capacitances. The high capacitance ratio of the fabricated analog capacitor (Cr = 20–40) and the MIM digital capacitor designed can be observed to have a Cr = 3.

Fig. 2. CV-curves of an analog capacitor and an MIM digital MEMS capacitor.

B) State of the art

Capacitive shunt switches have been realized with a large variety of dielectrics. Silicon nitride (Si₃N₄) with a dielectric constant of ɛ _r = 7.5 is the most popular material for capacitive MEMS [Reference Goldsmith, Zhimin, Eshelman and Denniston2, Reference Peroulis, Pacheco, Sarabandi and Katehi3, Reference Muldavin and Rebeiz8]; however, other high dielectric constant materials such as strontium titanate oxide (ɛ _r = 120, [Reference Park, Kim, Chung and Bu4]), piezoelectric lead zirconate titanate (ɛ _r = 190, [Reference Ziaei, Dean and Mancuso6]), aluminum nitride (AlN, ɛ _r = 9:8, [Reference Fernández-Bolaños, Tsamados, Dainesi and Ionescu7]), and even dielectric-less structures [Reference Blondy5] present excellent RF performance up to 40 GHz. All of them have demonstrated to achieve very low insertion losses (<0.2 dB at 40 GHz) and high isolation (>30 dB) with an elevated capacitance ratio (Cr > 30) and acceptable switching time (3–20 µs) at a relatively low actuation voltage (<30 V) for billions of cycles [Reference Ziaei, Dean and Mancuso6, Reference Fernández-Bolaños, Tsamados, Dainesi and Ionescu7, Reference Goldsmith, Maciel and McKillop12, Reference Lakshminarayanan, Mercier and Rebeiz13].

The RF-MEMS capacitive shunt switches as well as the other MEMS presented in this article use a common fabrication process developed at the Fraunhofer Institute of Silicon Technology (ISiT). The original process includes a combination of thin and thick nickel (Ni) for the suspended membrane and the anchors (and/or stiffening bars), respectively. In addition, AlN is used as a high-lifetime dielectric. The fabricated capacitive switches have performances comparable to state of the art of the technology.

Figure 3 compares the most representative RF capacitive switches regarding the area and the isolation with respect to the applied actuation voltage. In the figure, the selected dark squares represent the desired performances of a switch. The fabricated capacitive switch is in general more compact and achieves higher isolation than their counterpart MEMS switches [Reference Rebeiz1–Reference Ziaei, Dean and Mancuso6], for similar actuation voltages (Fig. 3(a)). In addition, Fig. 3(b) depicts the capacitance ratio (Cr) for each switch, which is in principle a good figure of merit of the RF capacitive switch performance. However, a device with a very high capacitance ratio (Cr ~ 600, [Reference Park, Kim, Chung and Bu4]) could result in poor isolation (<20 dB) due to bad dielectric RF properties.

Fig. 3. Comparison of the actuation voltage of state-of-the-art switches versus (a) the total switch area and (b) the isolation, S ₂₁, at 40 GHz as well as the Cr.

It is noteworthy to remark in Fig. 4 that a trade-off exists between low spring constants (for low actuation voltages purpose [Reference Peroulis, Pacheco, Sarabandi and Katehi3]) and the switching time or speed of the switch following a linear dependency. A known way of reducing the switching time is by device miniaturization [Reference Lakshminarayanan, Mercier and Rebeiz13]. The price to pay is a decrease in the RF performance.

Fig. 4. Comparison of the switching time of the state-of-the-art switches versus the actuation voltage.

The superior characteristics of the fabricated RF-MEMS switch can be attributed to excellent mechanical behavior of Ni membranes, good anchoring designs that reduce the effect of induced stress, and good AlN dielectric properties.

C) Design and fabrication

The RF-MEMS capacitive shunt switches have been developed at ISiT as part of a multi-project wafer consisting of an eight-mask process fabricated on highly resistive Si substrate (>3 Ω cm), presented in detail elsewhere [Reference Fernández-Bolaños, Lisec, Dehollain, Tsamados, Nicole and Ionescu14, Reference Fernández-Bolaños, Tsamados and Ionescu15]. The devices are loaded on a 78 Ω high-impedance line consisting of 2 µm insulated thermal oxide and 3 µm gold (Au) CPW. The total CPW width is 300 µm with S/W/S of 125/50/125 µm. The process combines to achieve a good mechanical and reliability behavior; a thin 1 µm evaporated Au/Ni/Au multi-metal layer for the movable membrane, a thick 14 µm electroplated Ni for the anchors and stiffening bars and 300 nm medium-dielectric constant amorphous AlN dielectric. The release structure presents an approximate air gap of 2 µm. Figure 5 illustrates a cross-sectional view of a capacitive MEMS switch with a fixed–fixed beam configuration showing the main layers of the process and a detailed SEM (scanning electron microscopy) image of a fabricated device.

Fig. 5. (a) Cross-sectional view and (b) SEM image of a detail in a fabricated device showing the different layers of the process.

Different analog and digital RF-MEMS capacitive shunt switches have been specifically designed using mainly three kinds of low spring constant anchoring (spring type, meander type and crabbed-meander type) to achieve different actuation voltages as well as specific inductive components. Each of them was conceived to be used as core devices for applications in phase shifters and tunable filters and their design was optimized for their work at specific frequencies. Figure 6 depicts SEM images of the main switches with different anchoring indicating the application for which they were designed.

Fig. 6. Three different capacitive switches designs used as core devices for the following applications: (a) spring type and (b) crabbed meander type used in combination for analog and digital phase shifter and bandpass filters and (c) standard meander-type for stop band filter applications.

D) Characterization and modeling

The switch is electrostatically actuated directly from the bottom electrode by applying a DC voltage inline with the RF signal of the CPW central conductor. The suspended bridge (top electrode of the capacitor) is anchored to the ground plane of the CPW (Fig. 7) and is actuated, thanks to the resulting electrostatic force.

Fig. 7. Schematic view of a general RF-MEMS capacitive shunt switch electrostatically actuated.

Device characterization from 6 to 40 GHz is carried out on E8361A Vector Network Analyzer and the calibration is performed using an external short-open-line-thru commercial calibration kit. De-embedding corrections are performed by measuring a thru line structure in the wafer to remove the losses and the parasitic pad capacitances of the unloaded line.

A 12 V low actuation RF-MEMS shunt capacitive switch is designed using crabbed-meander-type anchoring (shown in Fig. 6(b)) with an approximate spring constant (K _S) of 9 N/m. The achieved capacitance ratio, Cr, is 27 at pull-in voltage (V _PI), while it increases up to 40 with a higher actuation voltage of 40 V. Figure 8 presents the S-parameter measurements of the shunt switch. The device provides good RF performances with isolation better than 37 dB, an insertion loss of 0.2 dB and a return loss of 12 dB at 40 GHz. In addition, reliability tests have been performed in hot switching conditions (+5 dBm RF power) over 2 × 10⁹ cycles without permanent failure or stiction [Reference Fernández-Bolaños, Tsamados, Dainesi and Ionescu7] and an excellent thermal behavior has been measured from 100 to −130°C with a V _PI shift less than 2 V for the entire temperature range, ΔT = 230°C [Reference Fernández-Bolaños, Lisec, Dainesi and Ionescu10].

Fig. 8. S-parameter measurements of crabbed-meander-type anchoring capacitive shunt switch (shown in Fig. 6(b)) for different applied voltages. The pull-in voltage, V _PI, occurs at 12 V.

The shunt switch is modeled by an accurate equivalent circuit T-model as shown in Fig. 9. Where, C _t, L _t, and R _t correspond to the unloaded CPW and their values for a 460 µm length (d _ref) are: C _t = 35 fF, L _t = 143 pH, and R _t = 0.1 Ω (extracted from the thru de-embedded line). C _MEMS, L _p, and R _p are the components of the parallel shunt impedance (Zp) and correspond to the MEMS membrane. When the switch is not actuated, C _{MEMS Up} is very small and L _p and R _p can be neglected. Finally, ΔL _s and ΔR _s are corrective series elements used to keep the length of the discontinuity close to zero when modeling the device. As a result, ΔL _s can be negative while ΔR _s is expected to be positive [Reference Simons16]. MEMS switch circuit parameters are shown in Table 1 and have been extracted from the S-parameter measurements of the crabbed-meander-type switch and using the equivalent T-circuit model (Fig. 9).

Fig. 9. Equivalent T-model used to extract the circuit parameters of the switch from RF measurements.

Table 1. T-model circuit parameter extraction from S-parameter measurements.

Agilent-Advanced Device System (ADS) software has been used to perform equivalent circuit simulations of the extracted parameters and results are plotted in Fig. 10(a). Moreover, 3D finite-element electromagnetic (EM) full-wave simulations of the crabbed-meander-type capacitive switch are performed by means of Ansoft's HFSS software and depicted in Fig. 10(b). Both simulations are in good agreement with the measurements validating the extracting parameter method and the HFSS 3D EM model.

Fig. 10. (a) ADS circuit T-model simulations for all the actuation voltages shown in Table 1. (b) S-parameters' HFSS Ansoft simulation of the 3D EM model of the crabbed-meander-type anchoring capacitive switch in the up and down states.

E) Capacitive switches with a significant inductive component

In order to achieve higher isolation at specific desired frequencies, one possible solution is to design a switch with a highly dominant inductance component. A highly inductive design can be achieved by varying the anchoring or the switch design over the slots of the CPW as in the case of the meander-type switch (shown in Fig. 6(c)) However, in this kind of design high isolation is only obtained in general in a relatively narrow bandwidth.

It can be observed in Fig. 11 that a minimum of 5 dB improvement in isolation can be achieved compared to the standard crabbed-meander-type anchoring capacitive switch previously presented (Fig. 8). The membrane bridge inductance has been varied from 10 to 324 pH in five different configurations. They result in down-state natural resonant frequencies, f ₀ (or isolation region) at 7.8, 13.3, 17.8, and 27.5 GHz and higher than 40 GHz for the standard switch. A lower than 0.2 dB insertion loss is achieved for all devices with a minimum isolation of 12 dB in the lower frequency band (at 7.8 GHz). Insertion loss and isolation measurements (S ₂₁) of a set of inductive switches in the up and down state, respectively, are depicted in Fig. 11.

Fig. 11. Insertion loss and isolation measurements, (S ₂₁) of the crabbed-meander-type anchoring switch (black line) and highly inductive capacitive shunt switches made specifically to resonate at a certain frequency.

F) Digital MEMS capacitor design

The series fixed capacitance, C _FIX, is designed with a value much smaller than C _{MEMS DOWN} to limit the value of the total down-state digital MEMS capacitance. The idea consists in creating two well-defined capacitive states less dependent on the technology. This device eliminates the uncertainty of the MEMS down-state capacitance, which is much dependent on the fabrication process (due to the roughness and the planarization process of the top and bottom electrodes) and it is very difficult to accurately estimate its value in advance (>50% standard deviation). The total contribution of the MEMS and fixed capacitances placed in series configuration results to be similar to the smallest capacitance value, C _FIX. Consequently, small or medium changes in the value of the C _{MEMS DOWN} (1–1.5 pF, see Table 1) can be neglected for the device performance.

Among different C _FIX approaches considered to design digital MEMS capacitors, interdigitated capacitors seem to be the least dependent on the technology. The capacitance is realized in a single metal photolithography step and the accuracy of its value depends only on the lateral resolution of the equipment. However, it is not a miniaturized solution, overall if the interdigitated metal layer (in our case ~400 nm Ti/Pt/Au/Pt underpath metal) is relatively thin. MAM capacitors mainly depend on the sacrificial layer (which will define the air gap thickness) and on their uniformity around the wafer. Unfortunately, the air gap is highly dependent on the technology and usually has a minimum of 25% uncertainty. Finally, MIM capacitors depend only on the lateral resolution of photolithography and on the dielectric thickness. The dielectric layer can be deposited with high precision and its thickness can differ by a very few percentage of error. In addition, MIM capacitors are very compact and result in the most interesting approach [Reference Fernández-Bolaños, Lisec, Dainesi and Ionescu10].

Figure 2 illustrates the capacitance versus voltage of a standard analog capacitor and an MIM digital capacitor. It can be appreciated how MIM digital capacitors are almost independent on the technology. The total contribution of the up-/down-state capacitances are: C _TotalUP = 33.5 fF and C _TotalDOWN = 99 fF which corresponds almost exactly to the designed Cr of 3 [Reference Fernández-Bolaños, Lisec, Dainesi and Ionescu10]. The designed C _FIX results in a value of 90 fF. Unfortunately, due to this added series fixed capacitance the actuation voltage needed to snap down the MEMS membrane will substantially increase by an approximate factor of 1.3–1.5.

Figure 12 depicts the S ₂₁ phase or argument of the individual digital switch (of 0.5 mm cell length). The figure shows a phase shift between both states of 15° at 17 GHz. These digital capacitors will be used as the core device for phase-shifting applications in the next section. For a phase shift of 180° at 17 GHz, approximately 13 switches periodically loaded in a distributed MEMS transmission line (DMTL) CPW of 6 mm total length would be needed.

Fig. 12. Measurements of the S ₂₁ argument for the up/down states of the digital MEMS capacitor. It represents the phase shift achieved for an individual cell when switching from both states.

A) Analog DMTL phase shifter

Distributed MEMS TTD phase shifters with an analog control [Reference Barker and Rebeiz18] are limited by the MEMS intrinsic instability appearing at a third of the original air gap. This corresponds to a voltage-controllable capacitance ratio (Cr) that is lower than 1.5 during the stable up-state. Analog DMTL phase shifters are actuated below the pull-in voltage of the MEMS switches achieving a continuous and quasi-linear phase shifting. Once the membranes snap down, the Cr can rise up to 20–100. Such high Cr is usually advantageous for switching application where high isolation is needed. However, for other applications such as TTD phase shifters, even if it might seem advantageous for miniaturization purposes, the mismatch produced by such an enormous Cr is prohibitive.

The actuation is similar to that of the individual switches. DC voltage is applied through the central conductor of the CPW inline with the RF signal. Figure 14 illustrates the fabricated analog DMTL phase shifter using crabbed-meander-type switches.

Fig. 14. SEM image of the fabricated distributed MEMS analog phase shifter using 13-loaded RF-MEMS analog capacitive switches which corresponds to the one measured in Fig. 8 (only four switches are shown in this figure).

The main advantage of analog DMTL is its continuous phase shifting with a single voltage command. However, this phase shifter approach is not very compact and the maximum phase shift is limited to Cr < 1.5. The analog phase shift achieves a maximum continuous shifting of 4.36 ps time delay (ΔD), in a wide frequency band (from 5 to 25 GHz). The phase shifter includes 13 crabbed-meander-type anchoring analog capacitors periodically loaded in a CPW transmission line and separated by 400 µm from center to center of each individual MEM cell. The total size of the phase shifter is 5.4 × 0.6 mm².

The insertion loss is on average 2.39 dB at 17 GHz and the return loss is better than 15 dB for all the states over the entire frequency range as shown in Fig. 15.

Fig. 15. (a) S-parameter measurements and (b) continuous TTD of the analog DMTL phase shifter based on analog MEM capacitors.

The analog DMTL phase shifter results in a moderate phase shift of 53°/cm and 6°/dB at 17 GHz. The linearity of the differential phase shift with respect to frequency can be observed up to 25 GHz.

B) Digital DMTL phase shifter

In order to achieve more controllable and miniaturized solutions, digital phase shifters are adopted using periodically loaded digital capacitors with MIM fixed capacitors and are designed to have a higher capacitive ratio (Cr ~ 3) than their analog counterparts. Digital DMTLs result in a larger amount of degrees per centimeter performance than the analog phase shifters, thanks to its greater design compactness related to the higher Cr. However, they are limited to discrete states.

In principle, with a single control voltage, just two phase states are possible in 1-bit DMTL line [Reference Fernández-Bolaños, Tsamados, Dainesi and Ionescu7, Reference Fernández-Bolaños, Lisec, Dainesi and Ionescu10, Reference Perruisseau-Carrier, Fritschi, Crespo-Valero and Skrivervik11]. In order to achieve multi-state phase shifters, the standard solution is to add multi-bit control commands, which substantially increase the complexity of the device [Reference Hayden and Rebeiz9, Reference Hung, Dussopt and Rebeiz19]. However, other approaches could be envisioned. For instance, a three-state digital phase shifter consisting in switches with different spring constant devices would result in switches with different pull-in voltages. Such a phase shifter could achieve at least three phase shifters states with a single actuation command. This seems to be a superior solution in terms of simplicity and miniaturization for phase shifters when only three discrete states are needed.

However, multi-bit actuation phase shifters remain the standard design method by cascading a number N of DMTLs (cascaded-bit, CB-DMTL). Each section provides a delay corresponding to a bit weight (related to the number of switches and corresponding length) [Reference Hung, Dussopt and Rebeiz19] and are independently actuated by N voltage controls. This kind of phase shifter suffers from poor matching in the intermediate states. On the other hand, interlaced-bit configuration (IB-DMTL) has demonstrated theoretically to alleviate this problem [Reference Perruisseau-Carrier, Fritschi and Skrivervik20]. The price to pay is a higher complexity of the optimization design, overall if more than 2 bit are required.

A two-bit DMTL phase shifter has been designed with IB-DMTL configuration and optimized to attain minimum mismatch at 17 GHz (frequency of interest for the beam steering application) in all its phase states. An SEM image of the fabricated phase shifter is shown in Fig. 16. Eighteen spring-type anchoring bridges (with 8 and 10 bridges for bit A and bit B, respectively) are loaded interlaced and spaced by 344 µm. The total size of the 2-bit IB-DMTL phase shifter is 6.1 × 0.9 mm². The bits are individually actuated using high-resistivity bias lines made of 25 nm of tantalum (100 Ω/sq) to prevent the coupling between the RF signals and the bias circuitry.

Fig. 16. SEM image of 2-bit DMTL phase shifters in interlaced (IB) and cascaded (CB) bit configuration loaded in a CPW and actuated through high-resistivity bias lines.

The fabricated 2-bit IB-DMTL phase shifter achieved a de-embedded insertion loss of 1.9 dB on average and an excellent optimized matching better than 21 dB at 17 GHz for all the phase shifter states (Fig. 17(a)). In addition, an acceptable return loss (better than 12 dB) is obtained from 5 to 25 GHz. Constant and equidistant weight-bit time delays of 6.2, 11.9, and 18 ps, which correspond to 01, 10, and 11 states, respectively, were measured (Fig. 17(b)).

Fig. 17. Two-bit IB-DMTL phase shifter with an enhanced optimization method to improve the matching at 17 GHz. (a) S-parameter measurements and (b) constant time delay states at different actuation bits.

The loaded characteristic impedance of the phase shifter in the different states is shown in Fig. 18(a) up to 40 GHz. The real impedance corresponds to 54, 50.9, 48.1, and 45.4 Ω from 00 to 11 phase states. These values demonstrate and successfully validate the IB-DMTL configuration proposed in [Reference Perruisseau-Carrier, Fritschi and Skrivervik20] as well as the frequency matching optimization method. The figures of merit of the phase shifter result in 180°/cm and 20°/dB (without de-embedding the structure) at 17 GHz as shown in Fig. 18(b).

Fig. 18. Two-bit IB-DMTL frequency optimized phase shifter; (a) the real and imaginary impedance and (b) constant time delay states at different actuation bits.

C) 17 GHz fully integrated antenna beam steering concept

A full integration of an antenna beam steering concept has been performed in a single chip for wireless sensor nodes. The system includes a dipole antenna, a balun, and a 2-bit digital DMTL phase shifter. An SEM of the entire fabricated system is depicted in Fig. 19(a). The dipole antenna is patterned in a single thick metal as the CPW lines as well as the balun but with two fixed air bridges to connect both ground planes of the CPW.

Fig. 19. (a) SEM image of the fabricated entire system for applications in antenna beam steering and (b) optical image of the antenna array integration with the PCB showing the wire bonding connections.

The antenna and the phase shifter designs were optimized to have good matching and return loss (>40 dB) at the desired RF frequency (17.2 GHz) using a surface-passivated HR-Si substrate [Reference Fernández-Bolaños, Perruisseau-Carrier, Dainesi and Ionescu21] which was thinned to 300 µm to improve the radiation pattern performance. The balun provides a right-angle transition from the CPW of the phase shifter to the coplanar strip feeding line of the dipole antenna. The Si chip, consisting in an antenna array of two elements in an area of 6 × 7 mm², was successfully fabricated and assembled to a customized PCB with an SMA connector for testing (Fig. 19(b)).

The first measurements of the beam steering concept demonstrate good return loss performance with small deviation versus simulations [Reference Fernández-Bolaños22, Reference Vasylchenko, Fernández-Bolaños, Brebels, De Raedt and Vandenbosch23]. Typical antenna radiation pattern behavior was measured and active beam steering measurements are currently ongoing.

A) Bandpass filter

A compact and simple controllable tunable bandpass filter has been fabricated based on a lumped–distributed transmission line filter concept [Reference Fernánez-Bolaños, Dehollain, Ayoz, Nicole and Ionescu24] to select discretely the specific center frequency (from 20 to 23 GHz) with an independent controllable channel bandwidth (of approximately 0.5–1 GHz).

The electrical length of the filter corresponds approximately to λ/2 at the maximum center frequency of the tuning range in a slow-wave structure (meaning the CPW including the MEMS loaded capacitors in the up-state). The filter combines two kinds of MEMS devices shown in the SEM image of the filter partly represented in Fig. 20. The first kind of devices, the shunt digital capacitors, are used to digitally tune from one frequency band implemented by using the concept of DMTL lines [Reference Fernández-Bolaños, Lisec, Dainesi and Ionescu10]. Changing the state of these capacitors discretely modify the phase velocity and electrical length of the resonator and thus, the resonant frequency. The second part of the system is implemented with MEMS interdigitated capacitors in series configuration (series-analog capacitor). They work in analog continuous mode and its main function is to continuously tune the bandwidth of the bandpass filter by modifying the value of the coupling series-analog capacitor placed at both the input and the output of the filter. In addition, they can be used to fine de-tune the filter center frequency by an extra correcting element originally placed at the center of the filter and first introduced by the authors in [Reference Fernánez-Bolaños, Dehollain, Ayoz, Nicole and Ionescu24].

Fig. 20. SEM images of the fabricated tunable bandpass filter with parallel and series MEMS devices.

S-parameter measurements were taken from 15 to 25 GHz with a center frequency in f ₀ = 21.5 GHz. Actuation of the analog and digital series and shunt MEMS capacitors was accomplished with voltages from 15 to 40 V.

The center frequency of the bandpass filter shifts from 20.45 to 22.6 GHz which corresponds to a 9.53% f ₀ tuning with relative stable fractional bandwidth (15.1 ± 1.9%) and minimum variation of the insertion loss (4.05 ± 0.25 dB) across the complete tuning range (Fig. 21(a)). Tuning of the bandwidth is also demonstrated in a constant center frequency, thanks to action of the correcting element [Reference Fernánez-Bolaños, Dehollain, Ayoz, Nicole and Ionescu24]. A 13.6% of bandwidth tuning is achieved at 20.5 GHz center frequency with a 1 dB fractional bandwidth varying from 17 to 20%. It is the first time that such independent center frequency and bandwidth tuning is demonstrated for filters working in the K-band of frequencies. Similar performance can be found in the literature but for filters working in the lower UHF frequency bands (from 300 MHz to 1.4 GHz) [Reference Carey-Smith, Warr, Beach and Nesimoglu25].

Fig. 21. S-parameter measurements of a distributed-lumped bandpass filter with independent (a) center frequency tuning from 20.5 to 22.5 GHz (9.5%) and (b) a relative bandwidth tuning of 13.6%.

B) Band-stop filter

A wideband high-frequency tunable band-stop filter working in the Ku- to Ka-band frequencies was originally devised [Reference Fernández-Bolaños, Lisec, Dehollain, Tsamados, Nicole and Ionescu14, Reference Fernández-Bolaños, Dehollain, Nicole and Ionescu26]. The filter is based on a miniature single-MEM device combining a central capacitive MEM switch loaded in a CPW line and suspension arms designed as multiple inductive meanders with voltage-controlled coupling to the substrate (Fig. 22).

Fig. 22. SEM images of (a) the fabricated band-stop filter device and (b) the meandering arm from the filter showing the zipping effect of the anchor.

The meander-type anchoring of the suspended structure has been specifically designed to modify the inductance of the LC tank to resonate at the frequency of interest and producing a considerable notch response. The main tuning capability is achieved, thanks to the variable capacitance between the suspended metal membrane and the RF signal bottom electrode via an electrostatic actuation. The tuning range can be increased by reducing the stiffness of the anchors, thus creating a zipping effect after the pull-in, avoiding a perfect contact of the membrane at pull-in. Due to the fabrication-induced stress and the low stiffness of the bridge at zero bias, the central part of the bridge slightly touches the bottom electrode, enabling an enlarged stable tuning region before pull-in (capacitance ratio in the stable region, Cr = 1.5) and a zipping effect after the pull-in. The tuning behavior is well controlled and the capacitance smoothly changes in a repeatable way and in a continuous and quasi-linear fashion, until an improved full contact of both electrodes.

In addition, the intrinsic inductance of the meanders over the slot (between the RF signal and the ground planes of the CPW) plays a significant role in the device operation as their coupling to the substrate additionally tune the capacitive component of the device due to the bending of the metal meander anchors during electrostatic actuation (Fig. 22(b)).

The fabricated device has an air gap of ~3 µm. The width and length of the meanders are 10 and 135 µm, respectively, and the capacitive actuation area is 135 µm × 80 µm. The tunable band-stop filter has been successfully characterized at room temperature and under high vacuum (10⁻⁵ mbar). High tuning range of the filter center frequency of over 55% is achieved. The center frequency tunes from 16.7 to 37 GHz (Fig. 23). A linear capacitive tuning is achieved from 0 to 27.5 V which corresponds to a continuous linear frequency tuning of 20%. The two center frequency tuning mechanisms are visible in Fig. 23. First, the MEMS capacitive tuning and, second, the zipping effect of the inductive arm coupling capacitance can be seen in this figure.

Fig. 23. S-parameters of the band-stop filter showing an outstanding tuning range of over 55%.

The insertion loss in the lower pass-band is better than 0.25 dB and the rejection level is around 20 dB for all the rejection bands. A constant fractional bandwidth of 3.3 and 5.7 GHz has been measured at 50 and 75% of the rejection level of the filter (−3 and −6 dB) which corresponds to a relative bandwidth of 10%. The temperature dependence is 4.5 MHz/°C for the temperature range from 20 to 100°C and gets around the double for the extended range (−100 to 100°C). The fabricated device is very robust and show highly reproducible characteristics at wafer level; the filter characteristics in air and vacuum are quasi-identical and stay stable over 3 month of wafer storage in ambient conditions.