This invention relates to electronic filters, and more particularly to tunable filters of the type used in radio-frequency (RF) systems such as RF communications and test equipment, for example, in RF front-end sections of receivers and transmitters.
The continuous rapid development of wireless communication technologies has led to significant growth of RF and microwave devices and systems for both civilian and military applications. This is particularly true for mobile applications with increased demands for high efficiency, small volume, low weight and low cost. Furthermore, there is a defining trend to combine multiple wireless communication functionalities into a single frequency-agile platform for multi-band and multi-standard operation with a lower overall system cost. While digital technologies and the associated signal processing have enabled the implementation of frequency agility at base-band, the design and implementation of the RF front-end components remain a significant challenge for mobile systems. Tunable filters suitable for band selection and image rejection constitute typical examples.
Most of today's practical tunable filters are based on the ferromagnetic tunability Yttrium-Iron-Garnet (YIG) resonators. YIG based tunable filters exhibit very wide tuning range over multiple octaves and very high quality factors (Q) of 10,000 at 10 GHz. Nevertheless, the large volume and high power consumption (0.75-3 W) of YIG based tunable filters hinder their integration into mobile communication systems. Many alternative approaches have been proposed to make miniaturized tunable RF/Microwave filters. These approaches are predominantly based on planar transmission line resonators loaded with solid-state varactors, ferroelectric-tuned varactors and MEMS varactors, switches and switched capacitors, as described, for example, in the following papers:
While achieving significant progresses in terms of miniaturization and tuning, planar resonator based tunable filters have relatively low quality factor (Q) of less than 400-500. This is due to either the low-Q lumped elements, as in the case of solid-state varactors, and/or the low-Q of the resonator itself. Inherently high-Q resonators loaded with tuners that result in a graceful Q degradation while tuning are need for high-Q tunable filters. Several approaches have attempted to accomplish this by employing dielectric resonators, cavity resonators and high temperature superconductivity (HTS) resonators. However, their limited tuning range and the requirement for cryogenic cooling (in the case of HTS tunable resonators) makes them unsuitable for mobile systems in the near future.
Evanescent-mode waveguide filters have recently attracted interest for realizing low-loss, highly-selective tunable filters for reconfigurable RF front-ends. Reference is had to the following papers:
Compared to half-wave cavity resonators, evanescent mode resonators offer several advantages including substantially smaller volume and weight, larger spurious-free region and feasibility for monolithic integration while maintaining a very high-Q. It is well known that waveguides below cut off can be used to create microwave filters by introducing obstacles inside the guide, as described by G. F. Craven et al. in “The Design of Evanescent Mode Waveguide Bandpass Filters for a Prescribed Insertion Loss Characteristic,” IEEE Trans. Microwave Theory & Tech., Vol. 19, No. 3, pp. 295-308, March 1971, and by R. V. Snyder in “New Application of Evanescent Mode Wave-Guide to Filter Design,” IEEE Trans. Microwave Theory & Tech., Vol. 25, No. 12, pp. 1013-1021, December 1977. This concept finds wide applications in the form of evanescent mode filters, ridge waveguide filters and combline filters. See, e.g., R. Levy et al., “Transitional Combline/Evanescent-Mode Microwave Filters,” IEEE Trans. Microwave Theory & Tech., Vol. 45, No. 12, pp. 2094-2099, December 1997; A. Kirilenko et al., “Evanescent-Mode Ridged Waveguide Bandpass Filters With Improved Performance,” IEEE Trans. Microwave Theory & Tech., Vol. 50, No. 5, pp. 1324-1327, May 2002; and R. V. Snyder et al., “Bandstop Filters Using Dielectric Loaded Evanescent Mode Resonators,” 2006 IEEE MTT-S Int. Microwave Symp. Dig., Vol. 2, No., pp. 599-602 Vol. 2, 6-11 Jun. 2004.
The simplest and most practical type of waveguide obstacle is a conductive re-entrant post, which represents an effective shunt capacitance and is usually realized by a tuning screw. By changing this capacitance, i.e., the gap between the post and waveguide wall (
Another shortcoming of current filters in front-end receivers is that they have very limited tuning. Generally multiple filters are needed in a multi-band environment. Also, the filter bandwidth tends to vary linearly with frequency in evanescent-mode filters due to the coupling iris, which can be quite significant over tuning ranges of more than an octave.
The present invention generally provides improvements in tuning filters. According to one aspect of the invention, a tunable filter has an electronically tunable center frequency and dynamic bandwidth control over a large tuning range.
Another aspect of the invention is an electrostatically actuated tunable evanescent-mode cavity filter with a high quality factor (Q) and a wide tuning range.
The objects and advantages of the present invention will be more apparent upon reading the following detailed description in conjunction with the accompanying drawings.
For the purpose of promoting an understanding of the principles of the invention, reference will now be made to the embodiments illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended, such alterations and further modifications in the illustrated device and such further applications of the principles of the invention as illustrated therein being contemplated as would normally occur to one skilled in the art to which the invention relates. The following description is divided into three parts each having its own series of numbered equations.
One preferred embodiment of the present invention is a substrate-integrated, tunable evanescent-mode cavity filter with controllable bandwidth using surface mount varactors. The filter has a bandwidth compensation network that allows reduction/control over the bandwidth of the filter without affecting the filter's performance. The bandwidth compensation network also does not take up any additional volume. A 3D model of the concept is shown in
This bandwidth compensation network can be further improved by using tunable lumped components to allow bandwidth control at one frequency, which would be particularly helpful in further reducing the bandwidth variation to come close to an almost constant bandwidth. It will also help reduce the insertion loss of the filter by increasing the bandwidth, depending on the relaxation in rejection requirements, interferer strength, etc, as a function of time. An embodiment of a bandwidth compensation network utilizing lumped capacitors is shown in
By varying the capacitance value, the coupling through the bandwidth compensation network can be changed thus changing the bandwidth. A larger vale of C corresponds to a longer length of line 1 as can be seen from equation 2, which would result in more reduction in bandwidth. This is because the phase variation due to the longer length of the line will increase the coupling through the bandwidth compensation section at higher frequencies. The important factor that is to be considered is the Q of the capacitor, and its effect on the pass band of the filter. Based on full wave simulations it was observed that for a 4.7 pF capacitor, a Q of 10 is sufficient to achieve a good pass band response for a filter at about 1 GHz as shown in
The concept has been verified through measured results as shown in
A major advantage of this concept is that it allows a significant control over bandwidth while dynamically tuning the filter, without lowering the quality factor. An illustration of this is that this concept allows an almost constant bandwidth over large tuning ranges, since the bandwidth can be reduced quite significantly as shown in
This filter is capable of dynamic band selection for intelligent or cognitive radios which can adapt to the current electrical environment and operate over a wide range of frequency bands. It also allows dynamic bandwidth selection to adapt to changing interferers. One embodiment of the filter has two evanescent-mode cavity resonators that are generated by forming via cages in a PCB substrate, for example, in a 125 mil thick Rogers® TMM3 microwave substrate with double-sided copper. The cavities are separated by a coupling iris which is formed by removing vias between the two resonators; the exact number of vias removed and the spacing between the resonators determines the bandwidth of the filter. The bandwidth increases almost linearly with frequency, and this variation is quite significant for large tuning ranges. To reduce this bandwidth variation, a transmission line bandwidth compensation network is designed which comes in parallel with the coupling iris. The transmission line compensation network can also incorporate surface mount varactors to allow tunable bandwidth. The design is done so that the low-Q of the varactors does not affect the filter performance. The patterns on both sides of the board can be generated using conventional PC board manufacturing processes, either a milling machine or photolithography. The tunable membrane is generated by laminating a thin copper foil onto the substrate where the bonding layer has a void where the capacitive post is situated. A piezoelectric disc actuator is then mounted on top of the cooper foil in order to move the membrane, which causes changes in the capacitance of the loading post and enables the frequency tuning. This capacitance change can also be achieved by using a microelectromechanical systems (MEMS) variable capacitor.
The new filter is a suitable component for tunable front end receivers and transmitters among other applications, and is especially suitable for enabling software-defined radio.
Further information about fabrication methods for use with the present invention may be found in the above-referenced paper by Gong et al. and in the following paper: P. Ferrand et al., “LTCC Reduced-Size Bandpass Filters Based on Capacitively Loaded Cavities for Q Band Application,” Microwave Symposium Digest, 2005 IEEE MIT-S International, pp. 2083-2086, June 2005. Both papers are hereby incorporated herein by reference along with all references cited therein.
Referring now to
The resonant characteristics of capacitive-post-loaded evanescent-mode cavity resonator are well-studied. The resonant frequency and quality factor are both functions of the cavity size, post size and the gap between the post top and cavity ceiling. The evanescent mode resonator is the cavity equivalent of a lumped element resonator and can be approximated by a lumped element L-C tank. The resonant frequency f0 is given by (1) in which Cpost represents the capacitance between the post and cavity ceiling.
Because the post top area Apost dimensions are far greater than the gap d (see
Since C0 is typically very small compared to Cpost, the resonant frequency f0 is approximately inversely proportional to the square root of the gap d. As a result, the frequency tuning range is dependent on the initial gap d0 and the maximum distance the diaphragm can deflect.
We use a simple 1-D spring-mass model to model the actuation of the diaphragm (
The analytical solution to (4) is
The relationship between the resonant frequency f0 and V is then given by (8)
Since pull-in occurs at about g0/3 deflection, the maximum tuning range Rmax is approximatively given by (9)
In order to achieve a higher tuning range, a small d and a large g0 are desired. As shown in the next subsection though, careful balancing is needed between the desired tuning range and the required actuation voltage.
Although a larger g0 results in a higher tuning range, it also yields a higher actuation voltage. The actuation voltage is also dependent on the actuation area A and the spring constant k. While g0 is primarily determined by the tuning range requirement, A can be chosen to reduce the actuation voltage by increasing the size of the bias electrode.
The spring constant of a circular diaphragm is given by (10)
where a and t are the diaphragm side length and thickness, E is the Young's modulus, σ the residual stress and v the Poisson's ratio. In the presence of practical residual stress (>5 MPa), the k_term dominates the spring constant value particularly for large diaphragms with radius over 1 mm Therefore, it is critical to achieve low residual stress in the sputtered Au film for low voltage operation.
The final design parameters are listed in Table. I. Ansoft HFSS (High Frequency Structure Simulator), available from Ansoft Cooperation (http://www.ansoft.com/products/hf/hfss/), was used for simulation, and the results for resonant frequency and quality factor of the design are shown in
The gold skin depth is 1.24 μm at 4 GHz, assuming the conductivity of Au thin film is 4.1×107 Siemens/m. The deposited Au film on the diaphragm is only 0.5 μm. The thickness of the Au film is primarily determined by pull-in voltage considerations and by the desire to keep the single crystal SOI thickness much greater than the gold thickness. The penalty paid is a slightly reduced quality factor by approximately 200-300 as shown in
The fabrication process of the tunable resonator involves standard microfabrication steps and conventional machining techniques. It consists of three parts: a) the SOI diaphragm piece, b) the bias electrode piece, and c) the cavity piece.
The fabrication of the SOI piece starts with patterning an AZ9260 photoresist layer on the handle layer side as an etching mask for deep reactive ion etching (DRIE). The buried oxide layer has very high selectivity (>200:1) to silicon in the DRIE process and serves as an etch stop. After etching through the handle layer by DRIE, the oxide is removed by diluted HF solution. The device silicon layer is released upon the removal of the oxide layer. The released diaphragm is nearly perfectly flat due to the extremely low residual stress in the device layer. A 0.5 μm thick Au layer is then sputtered on top of the released silicon diaphragm. The sputtering condition is carefully controlled to achieve a low tensile stress in the metal layer.
The bias electrode consists of two pieces of silicon bonded together. The thickness of the smaller piece h is controlled by wet etching in a 25% TMAH solution at 80° C. The etching condition ensures <0.1 μm thickness control and a very smooth surface finish. The two pieces are then metalized with Au on both sides and bonded together by Au—Au thermal compression bonding at 350° C. and 50 MPa pressure. A layer of 2 μm Parylene-C is deposited on the smaller piece side to create an insulation layer for biasing.
The cavity with the evanescent mode post is machined from a Rogers TMM substrate. The process starts with drilling vias that form the boundary of the resonant cavity and the feeding coplanar waveguide structure. The vias are electroplated with copper. Then a milling-machine is used to create the evanescent post by removing the substrate material around it. The removal of the substrate material also helps to increase the quality factor of the resonator by reducing dielectric loss. The vias and the post are then metalized by electroplating a thick layer of copper. The top copper layer of the cavity and the evanescent post is polished to reduce the surface roughness.
The SOI diaphragm piece is attached to the cavity by silver epoxy cured for 20 min at 125° C. The bias electrode is fixed on the SOI diaphragm by non-conductive epoxy cured at room temperature. After the assembly is done, two SMA connectors are soldered to both ends of the resonator to characterize the RF performance.
The measured quality factor of the tunable resonator at each resonant frequency is shown in
A new electrostatically-tunable MEMS cavity resonator has been demonstrated which has a very high quality factor of 460-530 that is maintained thought a continuous tuning range of 3.4-6.2 GHz (1.8:1). The required electrostatic voltage ranges from 0-130 V. The biasing scheme and the fabrication process contribute to ensure a robust mechanical structure that preserves the high quality factor of the cavity. The primary material of the tuning membrane is single-crystal silicon that is coated with a thin gold layer. The resonator is useful as a key building block of a pre-select filter in a reconfigurable RF front-end.
Referring to
The discussion begins with the modeling of a single post evanescent-mode cavity resonator and the filter design can be accomplished by properly coupling multiple resonators. The resonant characteristics of capacitive-post-loaded evanescent-mode cavity resonator have been studied. See, e.g., the above-referenced paper by Gong et al., and the paper by R. G. Carter et al. in “Calculation of the Properties of Reentrant Cylindrical Cavity Resonators”, IEEE Trans. Microwave Theory & Tech., vol. 55, no. 12, pp. 2531-2538, December 2007. The latter paper shows that, for cavities with simple and symmetrical geometries, such as a cylindrical re-entrant cavity (inset of
It is worth noting that in the highly-loaded case, the resonant wavelength is much larger than the size of the post and the resonant cavity. In other words, the size of the resonant cavity is significantly reduced by loading the cavity with a capacitive post.
At resonance,
where f is the resonant frequency, l the post height, c0 the free space speed of light, Cpost the effective capacitance of the post and Z0 the characteristic impedance of the coaxial line, which is determined by the post diameter a and cavity diameter b.
Eqn. (3) can be solved numerically once the post capacitance Cpost is known. Cpost has been evaluated by using the parallel-plate capacitance Cpp as a first-order approximation, as described, e.g., by H. Joshi et al. in “Highly Loaded Evanescent Cavities for Widely Tunable High-Q Filters,” 2007 IEEE MTT-S Int. Microwave Symp. Dig., pp. 2133-2136, June 2007. Preferably, the effect of fringing field capacitance Cff is taken into consideration. Cff has been estimated for a circular peripheral. L. Lewin, Advanced Theory of Waveguides, London, U.K.: Iliffe, 1951; and W. C. Chew et al., “Effects of Fringing Fields on the Capacitance of Circular Microstrip Disk,” IEEE Trans. Microwave Theory & Tech., Vol. 28, No. 2, pp. 98-104, February 1980. However, in the highly-loaded case, the fringing field capacitance due to the post side wall should be taken into account as well (
With the knowledge of Cpost, Eqn. (3) can be solved.
It can be seen from
An evanescent-mode cavity resonator is designed for 2-6 GHz range as vehicle to demonstrate the concepts. For size reduction and better integration with other circuit components, the evanescent-mode resonant cavity is integrated into a Rogers TMM microwave substrate.
The above designed substrate-integrated cavity is enclosed by a Au-coated thin diaphragm to form the evanescent-mode cavity resonator. The reason for using a thin diaphragm is twofold: 1) the diaphragm preserves the natural current path and avoids any current distortion; and 2) the diaphragm completely isolates the biasing circuit from the resonant field inside the cavity, minimizing any energy leak to the bias circuits. Therefore, the diaphragm design inherently preserves the high Q of the resonant cavity.
As described in above, the diaphragm consists of a thin layer of sputtered Au on top of the released device layer of an SOI wafer. The single crystal silicon device layer is a defect free and stress-free material and provides reliable mechanical support for the Au layer. The buried-oxide layer of the SOI wafer also enables a highly tolerant fabrication process, details of which are discussed below.
Frequency tuning of the resonator is realized by electrostatically pulling the diaphragm away from the capacitive post. The mechanical actuation of the diaphragm can be effectively modeled by a 1-D spring-mass system (
The spring constant of a circular diaphragm is given by
where w and t are the diaphragm radius and thickness, E is the Young's modulus, σ the residual stress and v the Poisson's ratio. As previously noted, the single crystal silicon device layer is a near stress free material and the residual stress is mainly due to the thin Au layer. In the presence of practical residual stress (>5 MPa), the k″ term dominates the spring constant value particularly for large diaphragms with radii over 1 mm. This dominance of the stress induced stiffness is true for rectangular thin diaphragm as well.
The spring constant k for thin diaphragm is quite high (500 N/m for 10 MPa residual stress) compared with conventional MEMS devices. For low voltage operation, it is desirable to lower the spring constant k by reducing either the residual stress or the thickness of the Au film. However, the reduction of residual stress in the Au film is limited by fabrication process tolerances and extremely low stress (<5 MPa) is very difficult to achieve. Reduction of the thickness of Au film also comes with a penalty in excessive RF loss. The Au skin depth at 2-6 GHz range is 1.76-1.02 μm. It is desirable to have a Au layer thickness larger than the skin depth at this frequency range. Careful compromise must be made in choosing of Au film thickness from the mechanical point of view.
While a high spring constant k is limited by process tolerance and quality factor requirements, the size of the diaphragm can be increased to reduce the actuation voltage. Although a larger diaphragm has relatively insignificant impact on k (diaphragm size only comes into play in the k′ term, which is dominated by the k″ term), it can accommodate a larger bias electrode, therefore reducing the required actuation voltage.
Mechanical stability is an additional benefit of using a larger diaphragm. It is well known to the MEMS community that material creep results in performance drift over time. It has been shown that the creep rates in microscale devices are directly related to the induced stress/strain in the device materials.
The fabrication process of the tunable resonator involves standard microfabrication steps and conventional machining techniques. It consists of three parts: 1) the SOI diaphragm piece, 2) the bias electrode piece, and 3) the cavity piece.
The fabrication of the MEMS diaphragm starts with patterning an AZ9260 photoresist layer on the handle layer side as an etching mask for deep reactive ion etching (DRIE). The buried oxide layer has very high selectivity (>200:1) to silicon in the DRIE process and serves as an etch stop layer. The oxide layer is etched in Buffered-Oxide Etchant (BOE) after the handle layer is etched by DRIE. The device silicon layer is released after the removal of the oxide layer. The released diaphragm is flat due to the extremely low residual stress in the device silicon layer. A 0.5-μm thick Au layer is then deposited on top of the released silicon diaphragm by DC sputtering. The sputtering condition is carefully controlled to achieve a low tensile stress in the metal layer.
The bias electrode consists of two pieces of silicon bonded together. The thickness of the smaller piece h is controlled by timed wet etching in a 25% TMAH solution at 80° C. The etching condition ensures <0.1 μm thickness control and a very smooth surface finish. The two pieces are then metalized with Au on both sides and bonded together by Au—Au thermal compression bonding at 350° C. and 50 MPa pressure. A layer of 2 μm Parylene-C is deposited on the smaller piece side to create an insulation layer for biasing.
The cavity with the evanescent mode post is machined from a Rogers TMM substrate. The process starts with drilling vias that form the boundary of the resonant cavity and the feeding coplanar waveguide structure. The vias are electroplated with copper. Then a milling machine is used to create the evanescent post by removing the substrate material around it. The vias and the post are then metalized by electroplating a thick layer of copper. A second milling removes more substrate material to increase the quality factor of the resonator by reducing dielectric loss. The top copper layer of the cavity and the evanescent post is polished to reduce the surface roughness.
The SOI diaphragm piece is attached to the cavity by conductive silver epoxy. To accurately control the gap between the post and the diaphragm, the assembly is performed while the resonator is connected to a network analyzer. The resonant frequency is monitored in real time as the diaphragm piece is mounted on the resonator. The position of the diaphragm piece can be adjusted until the desired frequency is achieved. The assembled sample is then cured at room temperate until the solvent content of epoxy precursor fully evaporates. After the assembly is completed, two SMA connectors are soldered to both ends of the resonator to characterize the RF performance.
From a production point view, it is desirable to have accurate and repeatable control over the gap. However, the current technology using TMM substrate is limited by the surface roughness of substrate and the copper laminate. Bonding process with precise vertical alignment is being developed using Si micromachined evanescent cavities and holds great promise in achieving a low-cost reliable assembly of the tunable resonators.
RF measurements were taken with an Agilent 8722ES network analyzer. The tuning response of two resonators with different post sizes are shown in
The measured quality factors of the tunable resonators are 300-650 at 1.9-5.0 GHz for resonator 1 and 460-530 at 3.4-6.2 GHz for resonator 2. The measured quality factors are lower than the simulated ones (
As previously discussed, material creep causes material property change over time when constant stress is applied. In MEMS structures this is typically seen when a stimulus is applied over longs periods of time. For example, the resonator may show a frequency drift if biased under a constant voltage for several hours. It is therefore important to consider this parameter in the filter design so the filter control becomes as simple as possible. The frequency stability of the fabricated resonators is tested by applying a constant bias voltage on the resonator and continuously recording the resonant frequency. Every two minutes ten samples were averaged to result in a recording.
A 2-pole tunable bandpass filter is designed by coupling two evanescent-mode tunable resonators.
The designed filter is fabricated using process similar to that discussed above, and RF measurements were again taken with an Agilent 8722ES network analyzer. The two MEMS diaphragm tuners are biased individually with two 200 V voltage supplies. Because of assembly tolerances, the two resonators are not aligned at the same frequency initially and have to be adjusted individually.
The filter is continuously tunable from 3.76-5.17 GHz with less than 120 V bias voltage. The insertion loss is <4 dB at 5.17 GHz for a bandwidth of 35.3 MHz. The extracted equivalent quality factor is 410 at 5 GHz. The value is a bit lower than the measured results from the individual resonators. This is mainly due to the increased RF radiation loss from the coupling slots.
Table III lists the recorded center frequency f0, bandwidth BW, the bias voltages (V1 and V2) and the equivalent unloaded Q (Qu).
The modeling, design and fabrication techniques of high-Q, high-tunability evanescent-mode cavity based RF MEMS tunable resonators and filters have been described herein. Electrostatic actuation of thin-film MEMS diaphragm based on SOI wafers enables highly reliable operation with no hysteresis and excellent mechanical stability. Continuously tunable resonators from 1.9 to 5.0 GHz (2.8:1) with measured Q of 300-650 and from 3.4 to 6.2 GHz with Q of 460-530 have been demonstrated. The required electrostatic voltage ranges from 0-120 V. Stability measurements under constant bias shows frequency drift less than 0.2%/hour. The low frequency drift is attributed to the use of single crystal silicon as a mechanical support layer and the inherently low stress design. A 2-pole bandpass tunable filter is also designed and measured covering 3.76-5.17 GHz with insertion loss less than 4 dB for a bandwidth of 0.7%.
While the invention has been illustrated and described in detail in the drawings and foregoing description, the same is to be considered as illustrative and not restrictive in character, it being understood that only preferred embodiments have been shown and described and that all changes and modifications that come within the spirit of the invention are desired to be protected. It should be noted that, as used herein, the term “RF” in intended to comprehend microwave frequencies as a particular portion of the RF spectrum, which also includes lower frequencies. All papers and other reference cited herein are incorporated herein in their entireties along with all references cited therein.
This application is related to U.S. Provisional Patent Application No. 61/102,717, filed Oct. 3, 2009, which application is hereby incorporated by reference along with all references cited therein.
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/US2009/059466 | 10/3/2009 | WO | 00 | 6/15/2011 |
Number | Date | Country | |
---|---|---|---|
61102717 | Oct 2008 | US |