The present invention relates to a tuneable guided-wave electromagnetic filter device.

Background

Tuneable filters are presently realised at radio, microwave, and millimetre-/sub- millimetre-wave frequencies by using either lumped element filters or half- wavelength-long resonant waveguide structures. As the frequency of operation increases, component size reduces, and the performance of lumped element filters is degraded by the increased dominance of parasitic effects and fringing fields. Consequently, the circuit design becomes more complicated. These problems can be overcome to a certain extent by using half- wavelength resonant structures and hollow waveguide-based components. However, it is difficult to implement tuneability and the cost of manufacturing these complicated components with sufficient precision increases prohibitively with frequency.

Another difficulty with implementing practical radio wave or microwave tunable filters is that they must have a characteristic impedance of 50 ohms in order to be integrated into conventional microwave systems. If the characteristic impedance of the device is not 50 ohms, the system suffers power losses through two mechanisms. Firstly, the impedance mismatch between device and system results in less-than- optimal overall power transfer and secondly, unwanted reflections are created that interfere to form standing waves. Thus, it is impractical to use devices that do not have a characteristic impedance of 50 ohms in a microwave system.

Summary of invention

According to one aspect of the present invention, there is provided a guided electromagnetic radiation filter or switch comprising a waveguide, a movable member/device that is operable to modify one or more electrical properties of the waveguide by movement into or out of the waveguide, and an actuator for causing movement of the mechanical member/device.

Preferably, the waveguide is a co-planar waveguide having two ground planes and a

signal conductor.

The characteristic impedance of a co-planar waveguide is determined primarily by the ground-plane to signal-line spacing and the width of the signal and ground planes, but also by other parameters such as the substrate permittivity, and the metallisation thickness. By using a co-planar waveguide structure, it is possible to optimise the filter to have a characteristic impedance of 50 ohms, the same impedance as microwave systems. This impedance match is an essential prerequisite to integration into microwave systems because it allows for optimum power transfer. Thus, the present invention provides a device that can be integrated into microwave systems, for example microwave systems created with standard multi-user MEMS processes.

The signal conductor may be discontinuous. For example, the signal conductor may be in two parts, wherein the discontinuity is the gap between those parts. The mechanical device may be movable towards and/or away from the discontinuity. The mechanical device may be movable into and/or out of the discontinuous region. The mechanical device may be movable to a position in which it straddles the discontinuity and connects the two parts of the signal conductor on either side of that discontinuity.

The movable member/device may be movable in substantially the same plane as the co-planar waveguide or a plane parallel thereto. Additionally or alternatively, the movable member/device may be movable in a different plane to that of the co-planar waveguide, for example an orthogonal plane.

The signal conductor may be split into two parts and the movable mechanical member/device may be movable into and/or out of the gap between the two parts of the signal conductor. The mechanical member/device may have a plurality of protrusions extending from it, which protrusions mesh with corresponding protrusions that extend from the signal conductor. The protrusions may be elongate members or fingers. The movable member may be movable in a plane perpendicular to the plane of the signal conductor.

Where the waveguide is co-planar and has ground planes and a signal conductor, one of the ground planes may be split. The movable member may extend between the parts of the split ground plane. The actuator may be positioned outside the ground planes. The parts of the split ground plane may be connected by a bridge connection that extends over but does not touch the movable member. The bridge connection may be a wire bond.

The actuator may be provided externally of the waveguide. The actuator may include a latch that may be used to maintain at least one tuning position in the absence of application of a current or voltage to the actuator.

The actuator may include at least one micro mechanical actuator. The micro mechanical actuator may be one or more of an electrostatic actuator, a magnetic actuator, a magnetostrictive actuator, a shape memory alloy actuator, a piezoelectric actuator, an ultrasonic actuator, a pneumatic actuator, an actuator driven by optical energy, an electrothermal actuator.

The co-planar waveguide may include at least one supporting substrate that is formed of a material selected from Silicon (Si), a III-V semiconductor material, e.g. Gallium Arsenide (GaAs) or Indium Phosphide (InP), quartz, silica, sapphire or a dielectric medium such as a ceramic.

The movable device may be operable to move between positions in which a particular frequency or range of frequencies incident on the filter device is passed or rejected, so that the filter behaves as an all-pass or an all-reject filter.

The filter may be operable to be tuned continuously across a frequency range. Additionally or alternatively, the filter device may be operable to be tuned discretely across a frequency range. The filter may be operable to tune the electromagnetic spectrum between 10 MHz and 300 GHz.

BRIEF DESCRIPTION OF DRAWINGS

Embodiments of the present invention will now be described by way of example only with reference to the accompanying drawings of which:

Figure 1 is a schematic top view of a co-planar waveguide (prior art); Figure 2 is a schematic side view of a co-planar waveguide (prior art);

Figure 3 is a schematic side view of a co-planar waveguide with part of the supporting substrate removed (prior art);

Figure 4 is a schematic top view of a co-planar waveguide transition for making probe contact (prior art); Figure 5 is a schematic top view of a radiofrequency section of an electromagnetic filter device in which the invention is embodied in a first tuning position;

Figure 6 is a schematic top view of the radiofrequency section of Figure 5 in a second tuning position; Figure 7 is a schematic top view of an electromagnetic filter device that includes the radiofrequency section of Figures 5 and 6;

Figure 8 is a schematic side view of the material layers used in the fabrication of an electromagnetic filter device of Figure 7;

Figure 9 is a schematic top view of individual mask layers used in the fabrication of the electromagnetic filter device of Figure 7;

Figure 10 is a schematic top view of an electromagnetic filter device on which specific device dimensions are shown;

Figure 11 is a scanning electron micrographic representation of an electromagnetic filter device according to a first embodiment of the first aspect of the present invention;

Figure 12 is a plot of actuator deflection as a function of electrical power for the electromagnetic filter device of Figure 11 ;

Figure 13 is a plot of transmission as a function of power supplied to the actuator for the electromagnetic filter device of Figure 11; Figure 14 is a plot of transmission as a function of frequency of the electromagnetic filter device of Figure 11;

Figure 15 is a scanning electron micrographic representation of a modified version of Figure 11 ;

Figure 16 is a plot of transmission as a function of signal frequency of the electromagnetic filter device of Figure 15;

Figure 17 is a schematic top view of a wire bond applied to an electromagnetic filter device in which the invention is embodied; Figure 18 is a schematic top view of an air bridge applied to the electromagnetic filter device in which the invention is embodied;

Figure 19(a) is a schematic side view of the electromagnetic filter device of Figure 17;

Figure 19(b) is a schematic side view of the electromagnetic filter device of Figure 18;

Figure 20 is a schematic top view of the radiofrequency section of another electromagnetic filter device in which the invention is embodied;

Figure 21 is an expanded top view of the interdigitated finger structure of Figure 20; Figure 22 is a top view of an electromagnetic filter device that includes the radiofrequency/interdigitated finger structure of Figures 20 and 21;

Figure 23 is a schematic side view of the material layers used in the fabrication of an electromagnetic filter device of Figure 22;

Figure 24 is a schematic top view of the individual mask layers used in the fabrication of the electromagnetic filter device of Figure 22;

Figure 25 is a schematic top view of an electromagnetic filter device of the same general form as the device of Figure 22, but on which dimensions are shown;

Figure 26 is a perspective view of the electromagnetic filter device of Figures 22 and 25; Figure 27 is a plot of actuator deflection as a function of electrical power applied to the actuator for the device of Figure 25;

Figure 28 is a plot of transmission as a function of frequency for the device of Figure 25;

Figure 29 is a schematic top view of the electromagnetic filter device of Figure 25 on which is an air bridge;

Figure 30 is plot of transmission as a function of frequency for the electromagnetic filter device of Figure 29;

Figure 31 is a plot of transmission as a function of frequency for the electromagnetic filter device of Figure 29;

Figure 32 is a schematic top view of a modified version of the electromagnetic filter device of Figure 22; Figure 33 is a perspective view of the device of Figure 32;

Figure 34 is a schematic top view of the individual mask layers used in the fabrication of the device of Figure 32;

Figure 35 is a schematic top view of the device of Figure 32, but on which dimensions are shown; Figure 36 is a plot of actuator deflection as a function of electrical power applied to the actuator for the device of Figure 35;

Figure 37 is a plot of transmission as a function of frequency for the device of Figure 35;

Figure 38 is a schematic top view of a modification of the electromagnetic filter device of Figure 22;

Figure 39 is a plot of transmission as a function of frequency for the device of Figure 38;

Figure 40 is a perspective view of a flip chip bonded structure;

Figure 41 is a section through the flip chip bonded structure of Figure 40 in a first operating position;

Figure 42 is a section through the flip chip bonded structure of Figure 40 in a second operating position;

Figure 43 is a top view of a first chip for use in the flip chip bonded structure of Figure 38; Figure 44 is a top view of another chip for use in the flip chip bonded structure of Figure 40, and

Figure 45 is a schematic top view of a radio-frequency section of a discretely tunable stub filter.

DETAILED DESCRIPTION OF THE DRAWINGS

Figure 1 shows a conventional co-planar waveguide. This has two ground planes Ia and Ib on opposite sides of a central signal conductor 2. Electromagnetic signals are transmitted through this structure by guiding one or more propagating electromagnetic

modes in the gap between the ground planes Ia, Ib and the central signal conductor 2. The coplanar waveguide can be constructed by patterning a single conducting layer of material to form the ground planes Ia, Ib and conductor 2 on top of a suitable dielectric substrate 3 such as gallium arsenide, quartz, alumina or silicon, as shown in Figure 2.

The characteristic impedance Zo of the waveguide of Figures 1 and 2 is determined primarily by the ratio of the width s of the central conductor 2 to the distance (2g + s) between the ground planes Ia, Ib. Thus, the co-planar waveguide has the advantage that its characteristic impedance can be readily tailored to be of the order of 50 ohms, the same impedance as microwave systems. Furthermore, its overall size can be tapered whilst maintaining the same Zo, whereas other waveguide designs such as striplines require the substrate thickness to be tapered as well. This is useful when layout constraints require the waveguide to pass through narrow gaps, e.g. between pins of other components or flip-chip bond-pads or when it is necessary to create pads for making contact with test station probes or bond pads for connecting adjacent substrates in a multi-chip RF sub-system.

Other design parameters such as the metal thickness, substrate permittivity, substrate conductivity, substrate thickness, width of the groundplanes, overall width of the signal conductor and gaps to the ground planes, and the presence or absence of ground planes above or below the waveguide (such as on the opposite side of the substrate) also influence the characteristic impedance Zo. There are no design equations that produce waveguide designs for a particular Zo, or attenuation. However, given the waveguide dimensions, elliptical integrals or any other suitable technique can be employed to determine the resulting Z ₀ . Of course, fabrication parameters such as surface roughness, line edge roughness, variation in the metal thickness and tolerances in the lateral dimensions (for example due to pattern shrinkage or expansion due to over or under exposure during photolithograpy) may serve to degrade the waveguide performance slightly, depending on the severity of the departure of the fabricated structure from its design parameters.

Figure 3 shows a modified version of the co-planar waveguides of Figures 1 and 2. In this case, the substrate is partly removed from underneath a fraction of the waveguide to leave an air-gap 4. An advantage of this is that transmission losses of the waveguide are reduced compared to the case where the substrate material has a high conductivity resulting in losses due to signal absorption in the substrate. For example, this allows the integration of co-planar waveguide designs into low-resistivity (typically resistivity of 1 Ohm-cm, corresponding to conductivity of 100 S/m) silicon substrates such as those used for RF CMOS integrated circuits. The section of the waveguide over the resulting air gap 4 is typically redesigned to have new dimensions s, w & g, as indicated 10a, 10b & 11, in order to maintain the same characteristic impedance Zo. In order to connect the main part of the waveguide Ia, Ib and 2 with the sections 10a, 10b & 11 over the air gap, a transition section 12a, 12b & 13 is needed, as shown in Figure 4.

The device in which the present invention is embodied is based on a modified version of the co-planar waveguides of Figures 1 to 4. Figure 5 shows the radiofrequency section of an electromagnetic filter device in which the invention is embodied. This has a first continuous ground plane Ia. Opposite this is a split ground plane having two parts 7a and 7b. Between the split ground planes 7a and 7b is a separate, independently movable conductor 8. Extending between the ground plane Ia and the split ground planes 7a and 7b is a central conductor that has a break 5 along its length so that it defines two separate signal conductors 6a and 6b. The break 5 capacitively couples the two signal conductors 6a and 6b but the coupling is sufficiently weak that signal transmission through the device is prevented. In use, the conductor 8 is movable from a first position, in which its end closest to the break in the conductor 5 is flush with the edge of the ground planes 7a and 7b to a second position in which, as shown in Figure 6, its end is in contact with both parts of the central conductor 6a and 6b. Movement of the conductor 8 controls the frequency of operation of the device. When its end is flush with the edges of the ground planes 7a and 7b, the device acts as an all reject filter. In contrast, when its end is in contact with both parts of the central conductor, the device acts as an all pass filter. Filters of this nature can of course be considered as switches. The conductor 8 will be referred to herein as a shuttle.

Figure 7 shows an inclined beam electrothermal actuator 15 for moving the shuttle 8 between its two operating positions. This is connected to the shuttle 8 by a rod 14. This type of actuator 15 is sometimes known as a chevron beam actuator. The inclined beam electrothermal actuator 15 has an inclined beam array formation 16a and 16b. To make the actuator move, an electrical power source supplies an electrical current at an anchor electrode 17. This current enters the electrode 17 and exits from another anchor electrode 18. Alternatively, the current could be applied to anchor electrode 18. In this case, it would then exit via electrode 17. The polarity of the current does not affect the direction of movement of the actuator.

The current flowing in the beams 16a and 16b experiences a total resistance R = 1.5 Ohms, causing P = I ² *R (Watts) of heat to be dissipated in the beams. The resulting rise in temperature causes the beams to expand. This causes the actuator beams 16a and 16b to push against the rod 14 and so move it, and the shuttle 8, forwards towards the central conductor 6. In actuator materials like polysilicon, the maximum temperature rise is limited to 600 - 800 degrees centigrade in order to avoid irreversible damage due to self-annealing. Typical values of the angular displacement of the actuator beams 16a and 16b is in the range from 2 degrees to 6 degrees, with a value of 3 degrees having been selected for a prototype device. This prototype device has been tested and will be described in more detail later. Typically, actuation was achieved by supplying 0 - 0.6 V, resulting in up to 0.4 A of current flow.

Various techniques can be employed to make the device of Figure 7. For the sake of illustration only, prototype devices have been fabricated using the METALMUMPs process, which is an electroplated nickel surface micromachining process obtained via the commercial foundry MEMSCAP, Inc. Figure 8 shows specific materials that can be used to make the device of Figure 7.

With reference to Figure 8, a silicon substrate 20, for example 400nm thick, is coated with a layer of silicon oxide 21. Then, electroplated nickel 22 is used as the primary structural material and electrical interconnect layer. Doped polysilicon 23 can be used for resistors, additional mechanical structures, and/or crossover electrical routing. Silicon nitride 24 is used as an electrical isolation layer. Deposited oxide (PSG) may

be used for sacrificial layers (not shown). A trench layer or 'recess' 25 may be included in the silicon substrate. This may be done, for example, for additional thermal and electrical isolation. In order to coat the sidewalls of nickel structures 22 with a low contact resistance material, a gold over-plate 26 can be used.

Figure 9 shows the individual mask levels of the MET ALMUMPs process that were used to fabricate the device of Figures 7 and 8. The individual mask level layouts are the only aspect of the fabrication METALMUMPS process that the user can control. The first mask 27 controls where the oxide layer 21 is preserved. The second mask 28 is for the substrate recess 25. The third mask 29 is for the anchors 23. The fourth mask 30 is for the metal layer 22.

Figure 10 shows the design of a prototype of the device of Figure 7 on which specific device dimensions are shown. Inset 31 shows the width of the beams in the inclined beam actuator 15. Inset 32 shows the dimensions of the gap 5 in the central conductor 6a and 6b. Figure 11 shows a scanning electron micrograph of the device of Figure 10. In this, the shuttle 8 is visible, as is the break 5 in the central conductor. The electrothermal actuator 15 is also visible.

Figure 12 is a plot of the amount of deflection of the shuttle as a function of the amount of electrical power applied to the electrodes 17 and 18 of the device of Figures 10 and 11. The data of Figure 12 was determined by experiment. Power of up to at least 0.25W can be applied to the device for shuttle deflection in excess of 25 microns. Figure 13 shows the transmission coefficient of the device as a function of the applied current, at selected frequencies in the range 40 MHz - 54 GHz. From this it can be seen that at bias currents below about 0.355 A, this particular prototype device has transmission of less than -25 dB (or 'isolation' in excess of 25dB). At bias currents above 0.360 A, the device has transmission greater than about -10 dB (or insertion loss better than about 10 dB). Thus, the bias current needs to be changed by only approximately 5 niA (from 0.355A to 0.360A) in order for the filter to change from all-reject to all-pass. In other words, the transition can be made 'sharply', allowing the filter to be tuned relatively quickly.

Figure 14 shows the transmission coefficient of the device of Figures 10 and 11 as a function of frequency in the range 0 - 60 GHz, for two selected bias currents. The first bias current produces an 'off or 'all-reject' state 35, whilst the other produces an 'on' or 'all-pass' state 36. The transmission characteristic of this device in the first position 35 (the so-called 'off state) shows a high-pass filter response. The magnitude of the transmission (S ₂₁ ) in the off-state 35 is never greater than -25dB, so it may be said that the off-state 'isolation' of the switch is better than 25 dB. In the second position, the so-called 'on' state, the S ₂₁ transmission characteristic 36 shows a magnitude in the range of -2 to -10 dB, so it may be said that the on-state insertion loss of the switch is 2 - 1OdB. The performance is comparatively better at lower frequencies (for example 1 GHz) where the 'off -state isolation is about 50 dB and the 'on' -state loss is about 2 dB.

Figure 15 shows a modified version of the electromagnetic filter device of Figures 10 and 11. In this case, the shuttle has a wedge shaped contact 8a and the signal conductors 6a and 6b have tapered ends that define a substantially triangular shaped gap 5. The wedge 8a is shaped to plug the gap 5 between the conductors 6a and 6b, so that the wedge 8a is in physical contact with the conductors over a substantial part of its perimeter. This provides better contact between the signal conductors 6a and 6 and the shuttle. The device of Figure 15 also has a latch 8b that is used to maintain at least one tuning position of the shuttle 8 in the absence of application of a current or voltage to the actuator.

Figure 16 shows the transmission coefficient of the device of Figure 15 as a function of frequency in the range 0 - 40 GHz, for two selected bias currents. The first bias current produces an 'off or 'all-reject' state 35a, whilst the other produces an 'on' or 'bandstop' state 36a. The transmission characteristic of this device in the first, off- state position 35a shows a high-pass filter response. The magnitude of the transmission (S21) in the off-state 35a is never greater than -35dB, so it may be said that the off-state 'isolation' of the switch is better than 35dB. In the second position, the so-called 'on' state, there is a stopband with an insertion loss of 22dB, centred at 14.8 GHz. It is believed that the position of the centre value is determined by the dimensions of the shuttle 8, for example the length. Above 20GHz, the device acts

essentially as an all pass switch.

In a modification to the device of Figures 7, 10 and 15, an electrical connection may be made between the two separated sections 7a and 7b of the ground plane that lie either side of the shuttle 8. The purpose of this electrical connection is to improve the on-state insertion loss. This is achieved by reducing the magnitude of the step change in the electrical impedance 'seen' by the guided electromagnetic wave at the break in the ground plane. This connection may be realized using several methods. A first method is shown in Figure 17, where the connection is made by means of a wire-bond 37, typically after the rest of the circuit has been fabricated. The wire-bond 15 can be made of, for example, gold-wire with the aid of an ultrasonic or thermal bonding machine as widely used to make connections between CMOS silicon dies and the appropriate pins in their ceramic SIP, DIP and SMT packages. A second method shown in Figure 18 where the connection is made with an air-bridge, typically at the same time as and as part of fabrication of the rest of the circuit. These methods are further illustrated in the side view schematics of Figure 19, with the wire-bond method being shown in Figure 19(a) and the air-bridge method being shown in Figure 19Cb).

Figure 20 shows the radiofrequency portion of another device in which the invention is embodied. This is the radiofrequency portion of a filter that has a stop-band that can be turned on and off. When the stop-band is turned on, the centre frequency may be tuned. The device of Figure 20 has a central conductor that has two parts 40 and 43 separated by a conducting polygon 42. The central conductor 40 terminates in fingers 41 and is separated from the conducting polygon 42 by an air-gap. The finger pattern is repeated on the part of the central conductor with part 43 also terminating in fingers 44 and being separated from the conducting polygon 42 by an air-gap. The conducting polygon 42 has two sets of complementary fingers, with the first set 45 meshing with the fingers 41 and with the second set 46 meshing with the fingers 44. A section of the first set of intermeshed fingers 41 and 45, representative also of the second set 44 and 46, is shown schematically in Figure 21. The individual fingers have length ti and width t _w ; the gap between the fingers is t _g and the gap between the end of the fingers and the neighbouring conducting polygon is t _e . Typically t _w = t _g = t _e

and ti = AR.t _w where the aspect ratio AR is in the range 1 < AR < 100 and typically AR = 20. A typical finger therefore would be 10 microns wide by 200 microns long.

In order to turn the stopband on and off and tune its centre frequency, the conducting polygon 42 can be lifted out of plane by a vertical movement MEMS thermal actuator.

It should be understood that any reference to movement of the further conducting polygon 42 includes the constituent sets of fingers 45 and 46. Varying the power applied to the MEMS actuator and so the amount of deflection of the actuator allows the stopband to be turned on and off and tuned. This will be described in more detail later.

Figure 22 shows an example of a MEMS thermal actuator combined with the device of Figure 20. In this case, the MEMS thermal actuator has a plurality of interconnected beams 50 that consist only of a semiconductor material, such as silicon, and two anchor electrodes 51 and 52. The physical connection 53 joins the vertical movement actuator to the conducting polygon 42. Current entering at electrode 51 (or 52) and exiting from electrode 52 (or 51) causes the central two beams 54 and 55 of the beam cluster 50 to heat and expand much more than the outer two beams 56 and 57 of the beam cluster 50.

The beam cluster 50 has an initial small vertical ('out of page' with respect to Figure 22) offset produced during the fabrication process due to stress in the fabrication materials. This initial offset promotes movement in the vertical direction thereby lifting the conducting polygon 42 further out of plane when current is applied. To reduce the overall loss of the RF filter and to avoid 'stiction' (where moving parts bind strongly to non-moving parts, preventing operation) of the MEMS actuator, the substrate beneath the device is etched away and hence completely removed leaving air in the etched regions as previously described in connection with Figure 3. The areas enclosed by the broken line 58 shown in Figure 22 indicate the areas of the supporting substrate that are removed by etching.

While the device of Figure 22 can be made using any suitable technique, it is preferred that micromachining is used. In practice, prototype devices (which will be described in more detail later) were fabricated using a SOIMUMPs process, which is a silicon-on-insulator micromachining process obtained from the commercial foundry MEMSCAP, Inc. The main features of the process used are described with reference to Figure 23. A silicon-on-insulator (SOI) wafer is used as the starting substrate. For the prototype device, this wafer had the following layer thicknesses: silicon layer 60, of thickness 10 ± 1 μm; silicon oxide layer 62 of thickness 1 ± 0.05 μm, and handle wafer 62 (or 'Substrate') of thickness 400 ± 5 μm. The silicon layer 60 was doped, patterned and etched down to the oxide layer 61. The silicon layer 60 was used for mechanical structures, resistor structures, and/or electrical routing. The substrate 62 was then patterned and etched from the "bottom" side to the Oxide layer to create a recess 4. This allowed for through-hole structures. A shadow-masked metal process was then used to provide coarse metal features such as bond pads 63. A second pad- metal feature was provided for allowing finer metal features and precision alignment in areas not etched in the silicon device layer 64.

The individual SOIMUMPs mask levels used in the production of the example device are shown schematically in Figure 24. The first mask level 65 is not used. The second mask level 66 controls where the doped silicon layer 60 is preserved. This is the level that defines the MEMS actuator, the central conductors 40 and 43 and the conducting polygon 42. The third mask level 67 controls where the substrate 62 is etched away to form a recess 4. The fourth level mask controls where the blanket metal layer 63 is deposited. This is designed so that all parts of the device are coated in metal, except the beams of the beam cluster 50. No further specifications are required beyond the mask level layouts because the individual mask level layouts are the only aspect of the fabrication SOIMUMPS process, which the user can control.

Figure 25 shows an example of the device of Figure 22 on which feature dimensions are provided. The first inset 71 shows the cross section of the outer two beams 56 and 57. The second inset 72 shows the cross section of the inner two beams 54 and 55. The third inset 73 shows the dimensions of the fingers 41 and 45. The fourth inset 74 shows the dimensions of the physical connection 53 between the further conducting

polygon 42 and the micro-actuator formed from the beams 50 and electrodes 51 and 52. Figure 26 shows a perspective view of the device of Figure 25.

Figure 27 is a plot of experimentally determined vertical deflection of the MEMs thermal actuator as a function of the applied power, for the part of the filter shown in

Figures 25 and 26 and fabricated using the SOIMUMPs process described. From this it can be seen that 30OmW of applied power causes a vertical movement of 100 microns. The relationship between applied power and vertical deflection is nearly linear except for a slight deviation from the ideal response at powers less than 5OmW. The initial or 'power off deflection is approximately 10 microns.

Figure 28 shows the transmission coefficient of the device across the frequency range 0 - 60 GHz, for selected levels of power applied to the actuator. With no applied power (0 V), the transmission coefficient 75 represents a high-pass filter response with a transfer function H(s) that can be mathematically described as H(s) = s ² / [s ² + (s.2.pi.f _c )/Q + (2.pi.f _c ) ² ] where s = j.w = j.2.pi.f and w is the frequency in radians, f is the frequency in Hertz, f _c is the -3dB cutoff frequency of the filter and Q is the quality factor. The fitted values of the two parameters are f _c = 13 GHz and Q = 0.063. With a small amount of power applied (5 V, 15 mW) there is little change in the response 76 and 77. However as the power is increased to (15 V, 100 mW) a stop band 78 emerges, centred around 20 GHz. As the power is further increased, the stop band shifts up in frequency, reaching 25 GHz at (25 V, 230 mW) 79.

Figure 29 shows a variation to the device of Figures 18, 20 and 23. In this case, a bond wire 37 is added over the break in the ground plane 7a and 7b. As before, this is done to improve the transmission in the pass band. Figure 30 shows the transmission coefficient of the device with wirebond. From this is can be seen that the tuning performance of the device is not impaired by the addition of the wirebond. Figure 31 shows the transmission coefficient 80 of the device with wirebond compared to the transmission coefficient 75 of the device without the wirebond, in two representative tuning positions (0 mW and 230 mW). From this it can be seen that the transmission in the passband is improved by up to 3 dB by the addition of the wirebond.

Figures 32 and 33 show schematics of another guided electromagnetic filter in which the invention is embodied. This device exhibits a tunable bandpass filter response. This tunable filter has all the features of the filter described with reference to Figures 20, 21, 22 and 26, but modifications are made to the beams of the beam cluster. In this case, rather than being made of a semiconductor material alone, the beams 56a and 57a are coated with a layer of metal. Beam cluster 50a therefore comprises two central beams 54 and 55 that are made only from a semiconductor material and two metal-coated outer beams 56a and 57a. Hence, the entire surface of the device of Figure 32 is coated with a metal, such as gold, with the exception of beams 54 and 55.

While the device of Figure 32 can be made using any suitable technique, it is preferred that micromachining is used. In practice, prototype devices (which will be described in more detail later) were fabricated using a SOIMUMPs process, which is a silicon-on-insulator micromachining process obtained from the commercial foundry MEMSCAP, Inc. This process has previously been described in connection with Figures 23 and 24. Figure 34 shows the individual SOIMUMPs mask levels used. The first mask level 65a is used for coating metal to the beams 56a and 57a of Figure 32. The other levels are essentially the same as those shown in Figure 24.

Figure 35 shows an example of the device of Figures 32 and 33, and made using the masks of Figure 34, on which feature dimensions are provided. The first inset 71a shows the cross section of the outer two beams 56a and 57a where the black layer in the inset represents the metal coating. An experimental plot is shown in Figure 36 of the vertical deflection of the MEMs thermal actuator, as a function of the applied power, for the part of the filter shown in Figure 35. From this it can be seen that 330 mW of applied power causes a vertical movement of 50 microns. The relationship between applied power and vertical deflection is nearly linear except for a slight deviation from the ideal response at powers less than 50 mW. The initial, or 'power off, deflection is approximately 50 microns.

Figure 37 is a graphical representation of the transmission coefficient of the device across the frequency range 0 - 60 GHz, for selected levels of power applied to the actuator. With no power applied to the actuator, there is a passband with an insertion

loss of 17dB, centred at 18,8 GHz. As power is applied to the actuator, the passband shifts up in frequency. For an applied power of (24 V, 250 mW), the insertion loss becomes slightly greater (22 dB) but the centre frequency has shifted to 20.7 GHz. This represents an absolute tuning shift of 1.9 GHz, giving a fractional tuning range of 9.6%. The pass band quality factor Q is in the range 7 - 8.

Figure 38 shows a schematic of another guided electromagnetic filter in which the invention is embodied. This device exhibits a tunable bandpass filter response. This tunable filter has all the features of the filter described with reference to Figures 20, 21, 22, 25 and 26, but modifications are made to the surface of the whole structure. In this case, the whole surface is coated with a thick layer of highly conducting material, such as gold, but in such a manner as to avoid a short circuit between the electrodes and surrounding layers.

Figure 39 is a graphical representation of the transmission coefficient of the device in Figure 38 across the frequency range 0 - 40 GHz, for selected levels of power applied to the actuator. With no power applied to the actuator, there is a passband with an insertion loss of 6dB, centred at 18 GHz. As power is applied to the actuator, the passband shifts down in frequency. For an applied power of (0.36 V, 22 mW), the insertion loss becomes slightly lower (4 dB) but the centre frequency has shifted to 13.5 GHz. This represents an absolute tuning shift of 4.5 GHz, giving a fractional tuning range of 28 % around the mid-band frequency. The pass band quality factor Q is in the range 8-13.

Figures 40 and 41 show another device in which the invention is embodied. This has two chips that are aligned and joined together by a technique known as flip-chip bonding. Both chips have a "top" surface on which electrical, electromagnetic or micro-electro-mechanical devices have been fabricated. In flip-chip bonding, one of the chips is inverted and aligned to the second chip with the two chips separated by a distance g. The separation g can vary from a few microns to around one hundred microns. This means that both "top" surfaces of the chips are now positioned top-face to top-face. The two aligned surfaces are next bonded by use of electrically conducting solder bumps and other adhesive materials that further hold the two chips

together. Flip chip bonding enables electrical signals to be applied to chip 1 and delivered to chip 2 via the electrically conducting solder bumps. Electrical devices in chip 2 thus receive electrical signals e.g. from electrical power sources via chip 1.

Chip 1 consists of co-planar waveguides 85 of designs such as those shown schematically in Figures 1, 2 and 3. Chip 2 consists of vertical movement micro- actuators 86 such as the actuator types shown in Figures 22, 25 and 26. The purpose of the flip chip bonding is explained with reference to Figures 41 and 42. In Figure 41, features 95, 96 and 97 represent structures on chip 2, while features 99, 100, 101 represent structures on chip 1. Vertical movement actuator 97 has attached to its tip a polygon 96 having a highly conducting face 95 made from metal such as gold. Ground planes 99 and 101 and signal line 100 constitute a coplanar waveguide formed on a low-loss substrate 98. Polygon 96 with highly conducting face 95 has such dimensions so as to overlap ground planes 99 and 101 and signal line 100. Figure 41 is described as the "open" state because there is a gap g between the features on chips 1 and 2.

Figure 42 illustrates what happens when electrical power reaches the vertical movement microactuator on chip 2. The electrical power causes the vertical movement actuator 97 to deflect towards the coplanar waveguide formed by ground planes 99 and 101 and signal line 100 fabricated on the low-loss substrate 98 of chip 1. The purpose of the deflection is to allow the polygon 96, which has a highly conducting face 95, to make contact with both the central signal conductor 100 and ground planes 99 and 101. Figure 42 depicts the "closed" state because the finite gap g of Figure 41 is no longer present in Figure 42. In the "closed" state when electrical contact is made between the central signal conductor 100 and the ground planes 99 and 101, an electrical short is said to occur. Specifically this is described as a DC short circuit as all signals from zero frequency (DC) upwards are directed from the signal line to the ground. A switchable short circuit of this type in a waveguide can be used to form a guided electromagnetic wave filter device. It will be appreciated that more than one actuator can be formed on chip 2 such that short circuit action can be made to occur at different but fixed locations on the coplanar waveguide of chip 1.

Figure 43 depicts an embodiment of chip 1. Feature 93 is a coplanar waveguide; feature 92 represents a solder pad on which solder bumps can be located during the flip chip bonding process while feature 94 is a wire bonding pad which allows electrical signals to be delivered to chip 1 from the outside world. Figure 44 depicts one possible embodiment of chip 2. Feature 91 is a vertical movement actuator; feature 90 is the polygon with a highly conductive surface, while feature 89 is a solder pad on chip 2 for the flip chip bonding process.

A skilled person will appreciate that variations of the disclosed arrangements are possible without departing from the invention. For example, the actuator circuit may include a latch that may be used to maintain at least one tuning position in the absence of application of a current or voltage to the actuator. Also, although the device of Figure 7 operates essentially as a switch and the device of Figures 22 and 32 allow the operating frequency to be continuously varied within a particular frequency range, it is possible to have a discretely tunable device. An example of this is shown in Figure 45. This is based on the well-known stub filter. However, whereas the conventional stub filter has a continuous central conductor, the signal conductor 2 of Figure 45 has a plurality of gaps. Associated with each gap is a movable conducting member 8, which can be moved into and out of contact with the signal conductor 2 as described previously in connection with Figure 7. By selectively moving one or more of the conductors 8 of Figure 45 into contact with the signal conductor, its length can be varied, and so the frequency of operation. Thus, the device stop band is tuneable, but only discretely.

Furthermore, although specific embodiments of the invention use an electrothermal actuator, any other suitable micro mechanical actuator could be used, such as an electrostatic actuator, a magnetic actuator, a magnetostrictive actuator, a shape memory alloy actuator, a piezoelectric actuator, an ultrasonic actuator, a pneumatic actuator, an actuator driven by optical energy, and an electrothermal actuator. Accordingly, the above description of a specific embodiment is made by way of example only and not for the purposes of limitations. It will be clear to the skilled person that minor modifications may be made without significant changes to the operation described.