A Shape-Memory Alloy Actuator

FIELD

The present invention relates to a shape memory alloy actuator. The present invention also relates to a method of operating an SMA actuator.

BACKGROUND

A Shape Memory Alloy (SMA) is an alloy that can be deformed when cold but which returns to its pre-deformed or 'remembered' shape when heated. One well-known form of SMA material is NiTi (-50:50 alloy of Nickel and titanium). This material undergoes a phase change from Martensite to Austenite when heated through its transition region, and in so doing contracts by a substantial amount, typically between 2% and 8%. The transition temperature is achieved by Joule heating of the SMA material, by the passage of electric current through it.

Actuators are components that are responsible for moving and controlling mechanisms or systems within an apparatus. Shape Memory Alloys can be used to form actuators, temperature changes in the alloy allowing them to change shape to operate, which is usually achieved by passing an electric current through the material.

Various types of SMA actuators are known in the art, such as the straight-wire, bow string or“Vee”-type actuators, these usually being used as linear actuators

(actuation along a line).

A convenient form of this material for actuators is SMA wire, which has been appropriately mechanically and thermally processed to produce a reliable and long lasting contraction during phase change. Simple on/off control of the heating current to the SMA wire is used to control the actuator. However, this type of control generally results in very low position precision, unless the position is defined by fixed mechanical end-stops.

Much more precise positioning can be achieved by the use of sophisticated electronic control, where the SMA wire resistance is measured and used as a proxy for wire- length. Such techniques can give around 1 % positional precision in the short term, but over the longer term the wire tends to suffer from long-term drift with repeated mechanical load and as the wire material ages. The wire is also affected over the longer term by changes in ambient conditions such as draughts and other air temperature changes. Precision SMA linear actuators of the type just described also require continuous power (and control) to maintain a (more or less) fixed position. Where such SMA actuators are controlled via self-heating electric current through the SMA wire, then the average power requirement is directly related to the length of actuator stroke, since longer strokes require longer SMA wires (all else being equal). This also means that a very long stroke actuator will require a very long SMA wire.

Another issue that can occur with SMA wire-actuators is overload, where the force applied to the SMA-wire due to abnormal conditions at the load (e.g. out of specification load conditions) is sufficiently large to damage or even break the SMA- wire. With some actuators, e.g. simple single straight-wire actuators, it is possible to prevent certain forms of overload by the provision of end-stops that limit the stretching of the SMA-wire by load forces. However, end-stops alone cannot prevent overload in the fully actuated position - that is, with the SMA mainly or wholly in the austenite phase - since it is possible for excessive force to be applied within the designed stroke range (i.e. within travel between the end-stops).

With more complex actuators such as the double bowstring-actuator, this problem is exacerbated as it is possible to have heating-drive (e.g. electrical Joule-heating drive) that alone can overstress the SMA-wire, when for example both wires in such an actuator are heated above the austenite-start temperature, and the total contraction in the load direction is greater than the designed stroke.

It is well known in the art that the cycle time of operation of all types of SMA actuators is limited (at least in part) by the rate of cooling of the SMA material. Most SMA actuators use passive cooling (heat loss to ambient via conduction and convection and to a lesser extent, radiation). Shaping the SMA material into long thin wires assists this cooling process as it provides a large surface-area-to-volume ratio, but thermal time constants of even the thinnest available wires (~25um) are still on the order of 0.1 sec. Active cooling can considerably improve cooling times but effective and consistent active cooling is difficult and expensive to achieve. For example, liquid-cooling (as opposed to air-cooling) considerably improves cooling times but is complex and expensive in practise, especially for small actuators. Even with active cooling such as liquid cooling, the cooling acts to speed up only the expansion part of the SMA actuation/de-actuation cycle and simply causes additional unwanted power consumption during the contraction part of the cycle. US9,234,509 describes and shows an apparatus and method intended to produce a“quick-return” actuator, in an attempt to overcome the slow cooling rate problem. Furthermore, the rate of heating (which determines the rate of contraction during the actuation cycle) can be arbitrarily high. Although this is easily controlled if electrical Joule-heating is utilised, there are upper limits to the safe rate of heating.

Miniature cameras have a requirement for a zoom-lens facility, especially in cameras inside cellphones. Unlike adding optical image stabilisation (OIS) or auto-focus (AF), both of which require movement of optical elements a distance only a fraction of the lens focal length, and both of which require at best, short-term position stability, an adjustable focal-length zoom facility (e.g. 5:1 or 10:1) generally requires a much larger movement distance for one or more optical elements because large zoom implies long focal length, and it is desirable that the element(s) once moved, stay put indefinitely, until subsequently commanded to move elsewhere.

While shape-memory-alloy (SMA) actuators of the simple straight-wire type are generally ideal for both the OIS and AF functions, they can be problematic for controlling the zoom function. Several problems arise: firstly, an SMA wire of length L is able to reliably and continually move a load a distance of perhaps L/30 to L/20, if it’s useful lifetime is to be satisfactory, for, for example, a cellphone camera actuator. Thus a miniature camera requiring an optical element for zoom function even as short a distance as 4mm would need an SMA wire to be between 80mm and 120mm in length, far too big to fit conveniently within the confined space of a cellphone body; secondly, such a long SMA wire is relatively costly; thirdly, to hold the optical element in place the SMA wire would have to be continuously powered (to operate within the SMA temperature phase-change zone), and making power consumption even worse, the power requirements of such an SMA actuator are directly proportional to the length of the SMA wire, which as said, is necessarily long in this application.

Small electromagnetic motors could in principle be used to move optical elements for zoom function in miniature cameras, but they have not found favour, and cellphone electromagnetic zoom cameras do not exist in mass-production.

The industry has instead settled for a combination of multiple cameras in a cellphone, each with a different (fixed) focal length, to provide a pseudo-zoom function - with a severely restricted number of optical zoom factors (e.g. 1X, 3X, 10X). Relying on the very high resolution of modern image sensors in these cameras, digital zoom may then be used to interpolate or extend the zoom factors available but this provides a somewhat unsatisfactory user experience and is poor compared to zoom functions on“conventional” cameras such as 35mm handhelds. RF/Microwave (hereafter just RF) cavities and filters are known in the art, as are tuned filters, tuneable filters and adjustable cavities. Such a filter will have an input port and an output port (generically l/O-ports), sometimes coincident resulting in a single-port filter, and sometimes forming diplexers or multiplexers by having a common input port and two or more output ports. Recently, multiband filters are being developed employing multiple resonances per resonator, or multiple resonators per cavity. Electromagnetically-coupled multi-cavity filters with a plurality of stages are also known in the art wherein a sequence of cavities are electromagnetically coupled one to the next (in sequence) with possibly additional couplings between non-sequential cavities, and wherein the electromagnetic properties of one or more of the cavities are individually tuned and sometimes dynamically tuneable (i.e. when in service), resulting in a multi-cavity filter which has much higher performance than that of any of its individual cavities. Filter performance may be measured as some combination of factors including pass-band insertion-loss (ideally zero), stop-band rejection (ideally infinite), phase-linearity (usually ideally perfectly linear with signal frequency), ripple (ideally zero), group delay and power handling (PIM, breakdown due to discharge in air , multipaction breakdown). Such filters are typically of the low- pass, band-pass, band-stop, high-pass, phase-shift or all-pass configuration.

Where the filter is of bandpass or bandstop configuration it is characterised by a lower and upper cut-off frequency, the difference between which is called the bandwidth of the filter, and a“centre-frequency” between these two. A tuneable bandpass or bandstop filter will in general have a tuneable centre frequency and sometimes also a tuneable bandwidth. This may in practice be implemented by separately tunable lower and upper cut-off frequencies.

Such an RF filter can also be designed to produce primarily a phase shift, and in this case the useful effect of the cavity is to produce a phase difference dphi between that of the input signal and the output signal where dphi may be a tuneable quantity: i.e. in this case it is primarily the magnitude of phase-shift that is tuneable, and not the frequency at which the phase-shift occurs. Such a device is called a phase-shifting filter or just phase-shifter and an ideal phase-shifter would produce no change in insertion loss throughout the passband, and no change in insertion loss over the full range of tuneable dphi.

In some RF filters both a phase-shift function and a frequency-filtering function may be combined and both may be tuneable or specifically optimised. We use the term filter to encompass all of these types of device. RF cavities comprising chambers with solid conductive walls entirely surrounding the cavity are known in the art, as are RF cavities comprising a pair of often but not necessarily parallel, conductive ground-planes, and in the latter case individual cavities between the ground- planes may be delineated either by: i) solid conductive walls; or ii) by one or a plurality of conductive vias appropriately spaced and connecting between the ground-planes; or iii) by a combination of i) and ii). The input and output ports either: i) pierce the solid conductive cavity walls; or ii) are positioned between conductive vias connecting the ground-planes.

Combinations of all of these types of RF cavities within the same filter are possible.

Physically adjacent cavities which are not necessarily sequentially adjacent cavities as far as the direct signal path through the filter is concerned, may be

electromagnetically-coupled by couplings. Each such coupling may be comprised of: i) one or more holes or“irises” passing right through the solid conductive walls separating the physically adjacent cavities; or, ii) where the cavities are physically separated by one or more conductive vias connecting between the ground planes then by careful and appropriate positioning and spacing of the vias and again these couplings are said to be formed by irises - gaps between vias; or iii) by coupling wires protruding into both cavities to be capacitively coupled when the wires are electrically isolated, or inductively coupled when the wires are connected to the inside of the cavity; or iv) by conductive tracks placed on an insulating layer on the inside or outside wall of one or both ground-planes protruding into each of the cavities to be coupled capacitively or inductively. Combinations of these types of coupling within the same filter and even between the same cavities are possible.

The cavities may or may not contain one or more resonators. Typically a frequency- tuneable filter will have at least one cavity containing one or more resonators whereas a tuneable phase-shifter filter cavity may contain zero one or more resonators, and in the case where there is no resonator then instead a reflector is needed, and the phase-shifter may be realized with at least one side of the cavity being opened to a waveguide carrying the RF signal to be phase-shifted. The resonators may be either conductive or dielectric or some combination of these. Each of any conductive resonators within a cavity may be either connected to ground at one end only or connected to ground at each end or connected to ground in the middle or connected to ground at one or more half-wavelength intervals or not be connected to ground at all resulting in a floating conductive resonator. The one or more cavities of a tunable phase- shifter filter may contain non-resonant components with one or more tuneable parameters the changing of which primarily effects the phase between input and output. In a simple phase- shifter filter the parameter might be the position of the non-resonant component in the cavity.

Typical of current filter state of the art is patent WO2016/202687.

Completely electronic tuning of filters (i.e. involving no mechanical change of components) as described above is possible using semiconductor components (e.g. varactors, PIN diodes, FET switches). However, the RF losses (and associated RF noise production) associated with such electronic tuning components (such as varactor diodes which also introduce waveform distortion, intermodulation and cross modulation products, and reduce power breakdown threshold) increase with RF frequency and are often unacceptable above the low GHz region, especially where the RF power budget is critical or where very low receiver noise is required. In these cases very high-Q mechanically tuned filters are necessary.

In this specification where reference has been made to patent specifications, other external documents, or other sources of information, this is generally for the purpose of providing a context for discussing the features of the invention. Unless specifically stated otherwise, reference to such external documents is not to be construed as an admission that such documents, or such sources of information, in any jurisdiction, are prior art, or form part of the common general knowledge in the art.

SUMMARY

It is an object of the present invention to provide a shape-memory alloy actuator which goes some way to overcoming the abovementioned disadvantages or which at least provides the public or industry with a useful choice.

It is a further object of the invention to provide a method of operating a shape- memory alloy actuator which goes some way to overcoming the abovementioned disadvantages or which at least provides the public or industry with a useful choice.

The term“comprising” as used in this specification and indicative independent claims means“consisting at least in part of”. When interpreting each statement in this specification and indicative independent claims that includes the term“comprising”, features other than that or those prefaced by the term may also be present. Related terms such as“comprise” and“comprises” are to be interpreted in the same manner.

As used herein the term“and/or” means“and” or“or”, or both. As used herein“(s)” following a noun means the plural and/or singular forms of the noun.

Accordingly, in a first aspect the present invention may broadly be said to consist in

In a first aspect, the invention may broadly be said to consist in a shape- memory alloy actuator, comprising:

a stator body;

a shuttle, the stator body configured to receive the shuttle, the shuttle configured to move relative to the stator body;

a movement and locking mechanism connecting between the stator body and shuttle, the movement and locking mechanism configured to engage with the shuttle to move the shuttle relative to the stator body;

the movement and locking mechanism comprising first and second movement portions configured to engage and disengage from the shuttle in use, the first movement portion biased away from engagement with the shuttle, the second movement portion biased into engagement with the shuttle, and an SMA element actuator configured so that actuation of the SMA element actuator causes the first movement portion to engage with the shuttle to move the shuttle a part- step in a first direction, the shuttle and second movement portion configured so that de-actuation of the SMA element actuator causes the shuttle to move the remainder of the step.

In an embodiment, the shuttle comprises an elongate element configured to slide linearly in relation to the stator, with first and second long sides arranged in parallel with each other and the direction of travel, the first and second sides toothed, the first and second movement portions comprising first and second toothed pawls, the pawls and sides configured for mutual engagement.

In an embodiment, the teeth of the first pawl are moved into engagement with the teeth on the first side of the shuttle by actuation of the SMA element actuator to move the shuttle a part-step, the first pawl and teeth on the second side configured so that once the part step has been completed, the first pawl teeth and teeth on the second side engage to cause the shuttle to move the remainder of the step.

In an embodiment, the teeth of the pawls and the teeth on the first and second edges are configured with slopes of substantially 45 degrees each side. In an embodiment, the first movement portion comprises a leaf spring configured to bias the first movement portion away from engagement with the shuttle.

In an embodiment, the second movement portion comprises a leaf spring configured to bias the second movement portion into engagement with the shuttle.

In an embodiment, the leaf spring or springs comprise multiple leaves.

In an embodiment, the shape-memory alloy actuator further comprises a second SMA element actuator, connected and configured so that when actuated, the second SMA element actuator moves the second movement portion out of engagement with the shuttle.

In an embodiment, the shape-memory alloy actuator further comprises an overload spring configured to engage with the first SMA element actuator to at least partly absorb any overload force on the first SMA element actuator.

In an embodiment, the stator body and movement and locking mechanisms are formed as a unitary or one-piece element.

In an embodiment, the shuttle comprises a toothed cog, the first and second movement portions comprising first and second toothed pawls, the pawls and cog configured for mutual engagement.

In an embodiment, the teeth of the first pawl are moved into engagement with the teeth on the cog by actuation of the SMA element actuator to move the cog a part- rotation in a first direction, the first pawl and teeth on the cog configured so that once the part step has been completed, the first pawl teeth and cog teeth engage to cause the cog to move the remainder of the step.

In an embodiment, the teeth of the pawls and the cog teeth are configured with slopes of substantially 45 degrees each side.

In an embodiment, the first movement portion comprises a leaf spring configured to bias the first movement portion away from engagement with the cog.

In an embodiment, the second movement portion comprises a leaf spring configured to bias the second movement portion into engagement with the cog.

In an embodiment, the leaf spring or springs comprise multiple leaves.

In an embodiment, the shape-memory alloy actuator further comprises a second SMA element actuator, connected and configured so that when actuated, the second SMA element actuator moves the second movement portion out of engagement with the cog.

In an embodiment, the shape-memory alloy actuator further comprises a third pawl and third SMA element actuator, actuation of the third SMA element actuator causing the third pawl to engage with cog to move the cog a part-rotation in a second direction opposed to the first direction, the first pawl and teeth on the cog configured so that once the part step has been completed, the first pawl teeth and cog teeth engage to cause the cog to move the remainder of the step.

In a second aspect the invention may broadly be said to consist in a shape- memory alloy actuator, comprising: a stator body; a shuttle; the stator body and shuttle mutually configured for connection in use such that the shuttle can move in relation to the stator body; a movement and locking mechanism connecting between the stator body and shuttle, the shuttle and stator body mutually

configured such that the locking portion of the mechanism can engage with the shuttle to hold the shuttle in position relative to the locking portion, and disengage to allow the shuttle to move relative to the stator body; the locking portion of the movement and locking mechanism comprising an SMA element actuator, configured so that actuation of the SMA element actuator causes the locking portion to disengage, and de-actuation of the SMA element actuator causes the locking portion to re-engage.

In an embodiment, the shuttle comprises a linear element, the movement and locking mechanism comprising a first stop-block, the first stop-block configured to engage with one long edge of the shuttle so that so that the shuttle is held in position relative to the first stop-block, actuation of the SMA element actuator causing the first stop-block to disengage from the long edge.

In an embodiment, the shape-memory alloy actuator further comprises a second stop-block and second SMA element, the second stop-block configured to engage with the second long edge so that the shuttle is held in position relative to the second stop-block, the second SMA element configured so that activation of the second SMA element causes linear movement of the second stop block in a first direction parallel to the axis of the shuttle body.

In an embodiment, the first and second long edges comprise ratchet teeth, the stop blocks comprising pawls configured to engage with the teeth on their associated long edge, the teeth of the first long edge and first stop block configured to prevent movement in a first direction when engaged, the teeth of the second long edge and second stop block configured to prevent movement in the second, opposite direction when engaged.

In an embodiment, the shape-memory alloy actuator further comprises a first spring wire configured to bias the first pawl into engagement with the shuttle, and a second spring wire configured to bias the second pawl into engagement with the shuttle, the actuator further comprising a third SMA element configured so that activation of the third SMA element causes the second pawl to disengage.

In an embodiment, the shape-memory alloy actuator further comprises a fourth SMA element, configured so that activation of the fourth SMA element causes linear movement of the first pawl in the second direction.

In an embodiment, wherein the pawls and shuttle are configured such that in use, activation of the SMA elements allows movement of the shuttle for one tooth length only.

In an embodiment, each of the pawls comprises a pair of pawls, each pair comprising a stepper pawl and a passive pawl, on each side of the stator the stepper pawl trailing the passive pawl for linear movement of the shuttle caused by engagement of the stepper pawl on that side.

In an embodiment, each of the pawls comprises a pair of pawls, each pair comprising a stepper pawl and a passive pawl, on each side of the stator the stepper pawl leading the passive pawl for linear movement of the shuttle caused by engagement of the stepper pawl on that side.

In an embodiment, the pawl pairs are located in channels on the stator body, movement of the pawls in the direction of movement of the shuttle constrained by the edges of the channels, the pawls biased towards the edges of the channels and separated by a distance of substantially one tooth length.

In an embodiment, the pawls comprise locking pawls, configured to lock into the engaged and disengaged positions.

In an embodiment, the shape-memory alloy actuator further comprises locking SMA actuators configured to actively pull the pawls into an engaged state.

In an embodiment, the shape-memory alloy actuator further comprises a further pair of stepper pawls arranged one on each side of the shuttle, the movement and locking mechanism configured to activate a further stepper pawl substantially halfway through the activation cycle of the primary stepper pawl on that side. In an embodiment, the shape-memory alloy actuator further comprises opposed- stepper SMA actuators connected to the stepper pawls and configured to pull the stepper pawls into engagement with the shuttle.

In an embodiment, the stator and shuttle are configured to have a zero-stroke starting position, the actuator further comprising a spring mechanism configured to reset the stator body and shuttle to the zero-stroke position.

In an embodiment, the shape-memory alloy actuator further comprises a pair of slew-back pawls, the spring mechanism connecting between at least one of the slew-back pawls and the stator so that movement of the shuttle relative to the stator activates the spring mechanism, the slew-back pawls and shuttle configured so that the slew-back pawls are carried with the shuttle in normal use,

disengagement of the stepper and passive pawls causing the activated spring mechanism to return the shuttle to the zero stroke position.

In an embodiment, the shape-memory alloy actuator further comprises wings extending from each side of the shuttle, and the stator further comprises a groove configured to receive the wings.

In an embodiment, the wings are wedge-or triangular-shaped.

In an embodiment, the slew-back pawls are configured to extend at least partly around the shuttle.

In an embodiment, the shape-memory alloy actuator further comprises a reverse slew-back mechanism, comprising a reverse slew-back spring, and reverse slew- back pawls, the shuttle configured to extend through each side of the stator, the shuttle engaging with the reverse-slew back pawls at the opposed side of the stator from the slew-back pawls, the reverse slew-back spring configured to operate in the opposite direction to the spring mechanism.

In an embodiment, the wire portions of the SMA element actuators are embedded in a heat-sink compound.

In an embodiment, the movement and locking mechanism comprises a pair of pawls, the shuttle comprising a linear body with teeth on one long edge, the pawl teeth and the shuttle teeth symmetrical and configured to mesh with one another, each pawl independently engagable with the shuttle.

In an embodiment, the teeth have a slope angle of substantially 83 degrees.

In an embodiment, teeth on one or both of the shuttle and pawls are rounded. In an embodiment, one of the pawls comprises an active pawl, the movement and locking mechanism configured to allow the active pawl to move linearly along the same axis as the shuttle by between one and two tooth lengths, movement of the active pawl in the direction of travel of the shuttle limited by the configuration of the stator.

In an embodiment, the shape memory alloy actuator further comprises SMA actuators configured to independently lift the pawls out of engagement with the shuttle, springs configured to bias the pawls into engagement with the shuttle, and linear SMA actuators configured to move the pawls in parallel with the shuttle.

In an embodiment, the shape memory alloy actuator further comprises a cage, the pawls located within the cage, the cage and pawls configured to allow the pawls to slide freely substantially perpendicularly to the axis of the shuttle, the stator and cage configured so that the cage can move freely parallel to the axis of the shuttle but is prevented from moving perpendicularly to the axis of the shuttle, the linear SMA actuators connected to the cage.

In an embodiment, the shape memory alloy actuator further comprises a balancing shuttle and balancing movement mechanism, the balancing movement mechanism configured so that movement of the first shuttle causes movement of the balancing shuttle in parallel but in the opposite direction to the movement of the first shuttle.

In an embodiment, the movement mechanism comprises a swing arm pivoted at a point substantially midway between the shuttles and rotatably connected at each side to the shuttles.

In an embodiment, the movement mechanism comprises pulleys and cables connected between the shuttles.

In an embodiment, the movement mechanism comprises at least one cog, the shuttles aligned in parallel with the at least one cog between their inner edges, the inner edges comprising teeth configured to engage with the teeth of the cog, movement of one shuttle causing the at least one cog to rotate and move the other shuttle in the opposite direction.

In an embodiment, the shuttle comprises a ratchet wheel and the movement mechanism comprises a verge.

In an embodiment, the movement mechanism further comprises a spring connected to the verge to act to return the verge to an initial position once the SMA element actuator is disengaged. In a third aspect, the invention may broadly be said to consist in an SMA wire mechanical overload protection mechanism, comprising: a wire mount;

a pre-stressed tension spring, the mount and spring configured so that the spring can be located substantially within the mount with one end of the spring connected within the mount and the spring extending within the mount so that the free end is at or towards one end of the mount;

a wire terminal having a first end configured for connection to the spring free end, and a second end formed as a crimp end configured to receive the end of an SMA wire, the wire mount, terminal, and spring configured so that the terminal first end is in contact with and pulled onto mount by the spring, the spring exerting a pulling force so that the terminal first end remains in contact with the mount for any opposed pulling force exerted on the wire terminal that is under the level of the pulling force.

In an embodiment, the spring comprises a serpentine tension spring.

In an embodiment, the spring further comprises first and second pins that extend laterally from the body of the spring, the pins configured for insertion into a PCB, the pins and spring formed as a one-piece item from a conductive material, the first pin located at the mount connection end.

In an embodiment, the SMA wire mechanical overload protection mechanism further comprises a PCB, the PCB having a hole configured to receive the first pin, and a slot configured to receive the second pin, the slot aligned with the axis of the spring and allowing movement of the pin within the slot as the spring stretches.

In a fourth aspect, the invention may broadly be said to consist in a method of using a shape-memory alloy actuator as claimed in any one of claims 26 to 32, comprising the steps of:

i) de-actuating all shape-memory alloy actuator so that all pawls are in engagement with the shuttle;

ii) actuating the first shape-memory alloy actuator to disengage the first pawl pair from the shuttle;

iii) actuating the second shape-memory alloy actuator to move the second pawl pair in the first direction;

iv) de-actuating the first shape-memory alloy actuator to release the first pawl pair, the first pawl pair re-engaging with the shuttle; v) de-actuating the second shape-memory alloy actuator.

vi) repeating steps i) to v) as required to move the shuttle in the first direction to the required position.

In an embodiment, the method of using a shape-memory alloy actuator comprises the further steps of:

v) actuating the third shape-memory alloy actuator to disengage the second pawl pair from the shuttle;

vi) actuating the fourth shape-memory alloy actuator to move the first pawl pair in the second direction;

vii) de-actuating the third shape-memory alloy actuator to release the second pawl pair, the second pawl pair re-engaging with the shuttle;

viii) de-actuating the fourth shape-memory alloy actuator.

ix) repeating steps v) to viii) as required to move the shuttle in the second direction to the required position.

In an embodiment, the method uses a shape memory alloy actuator of the type that further comprises a spring mechanism configured to reset the stator body and shuttle to the zero-stroke position, the method comprising the further step of:

i) activating the first and third shape-memory alloy actuator to disengage the first and second pawl pairs from the shuttle so that activated spring mechanism returns the shuttle to the zero stroke position.

In an embodiment, the method uses a shape memory alloy actuator of the type that further comprises a spring mechanism configured to reset the stator body and shuttle to the zero-stroke position, and a pair of slew-back pawls, the spring mechanism connecting between at least one of the slew-back pawls and the stator so that movement of the shuttle relative to the stator activates the spring mechanism, the slew-back pawls and shuttle configured so that the slew-back pawls are carried with the shuttle in normal use, the method comprising the further steps of:

i) activating the first and third shape-memory alloy actuator to disengage the first and second pawl pairs from the shuttle;

ii) disengaging the stepper and passive pawls from the shuttle so that activated spring mechanism returns the shuttle to the zero stroke position. In an embodiment, the method uses a shape memory alloy actuator of the type that further comprises a second spring mechanism configured to reset the stator body and shuttle to a zero-stroke position, and a pair of slew-back pawls, the spring mechanism connecting between at least one of the slew-back pawls and the stator so that movement of the shuttle relative to the stator activates the spring mechanism, the slew-back pawls and shuttle configured so that the slew-back pawls are carried with the shuttle in normal use, the method comprising the further steps of:

i) activating the first and third shape-memory alloy actuator to disengage the first and second pawl pairs from the shuttle;

ii) disengaging the stepper and passive pawls from the shuttle so that activated spring mechanism returns the shuttle to the zero stroke position.

In a fifth aspect, the invention may broadly be said to consist in a miniature camera, comprising:

at least one optical element capable of movement substantially along the optical axis of the camera, the optical element mechanically directly or indirectly connected to a moving shuttle or output node that is connected to an SMA linear stepper actuator;

the body of the actuator fixed relative to the body of the camera, the direction of motion of the shuttle associated with the actuator substantially aligned with the optical axis of the camera, such that actuation of the actuator causes the at least one optical element to move along the optical axis of the camera;

the SMA linear stepper actuator as claimed in any one of claims 19 to 49.

In a sixth aspect, the invention may broadly be said to consist in a miniature camera, comprising:

at least one optical element capable of movement substantially along the optical axis of the camera, the optical element mechanically directly or indirectly connected via a linkage to a moving shuttle or output node that is connected to an SMA linear stepper actuator; the linkage configured so that that movement of the actuator shuttle along the actuator movement axis causes movement of the at least one optical element along the camera axis via the linkage;

the SMA linear stepper actuator as in any of the preceding statements.

In a seventh aspect, the invention may broadly be said to consist in a miniature camera, comprising: at least two optical elements or groups of elements configured for movement substantially along the optical axis of the camera;

at least two independently controlled SMA linear stepper actuator, one for each optical element or group of elements;

each optical element or group mechanically connected to the moving shuttle or output node of the associated independently controlled SMA linear stepper actuator; the body of each of the actuators fixed relative to the body of the camera, the directions of motion of the shuttles associated with the actuators substantially aligned with the optical axis of the camera, such that independent actuations of the actuators causes the optical element or groups thereof connected to each actuator to move independently along the optical axis of the camera;

each individual one of the SMA linear stepper actuators configured as in any of the preceding statements.

In an eighth aspect, the invention may broadly be said to consist in a miniature camera, comprising:

at least one optical element capable of movement substantially along the optical axis of the camera, the optical element mechanically directly or indirectly connected to a moving shuttle or output node that is connected to an SMA stepper actuator;

the body of the actuator fixed relative to the body of the camera, the direction of motion of the shuttle associated with the actuator substantially aligned with the optical axis of the camera, such that actuation of the actuator causes the at least one optical element to move along the optical axis of the camera;

the SMA stepper actuator as in any of the preceding statements.

In a ninth aspect, the invention may broadly be said to consist in a miniature camera, comprising:

at least one optical element capable of movement substantially along the optical axis of the camera, the optical element mechanically directly or indirectly connected via a linkage to a moving shuttle or output node that is connected to an SMA stepper actuator; the linkage configured so that that movement of the actuator shuttle along the actuator movement axis causes movement of the at least one optical element along the camera axis via the linkage;

the SMA stepper actuator as in any of the preceding statements. In a tenth aspect, the invention may broadly be said to consist in a miniature camera, comprising:

at least two optical elements or groups of elements configured for movement substantially along the optical axis of the camera;

at least two independently controlled SMA stepper actuators, one for each optical element or group of elements;

each optical element or group mechanically connected to the moving shuttle or output node of the associated independently controlled SMA stepper actuator;

the body of each of the actuators fixed relative to the body of the camera, the directions of motion of the shuttles associated with the actuators substantially aligned with the optical axis of the camera, such that independent actuations of the actuators causes the optical element or groups thereof connected to each actuator to move independently along the optical axis of the camera;

each individual one of the SMA stepper actuators configured as in any of the preceding statements.

In an embodiment, the SMA stepper actuator is connected directly or indirectly to a leadscrew so as to rotate the leadscrew when the stepper actuator rotates, the leadscrew fitted with a nut that moves axially along the leadscrew when the leadscrew rotates,

the at least one optical element or group thereof connected directly or via a linkage to the nut.

In an embodiment, the stepper actuator is connected to the leadscrew via gearing or belt.

In an eleventh aspect, the invention may broadly be said to consist in a phase or frequency tuneable device comprising:

a walled RF cavity;

at least one SMA actuator connected to the RF cavity, activation of the SMA actuator causing deformation and/or movement of the walls of the RF cavity;

The RF cavity and SMA actuator connected so that deformation and/or movement of the walls of the RF cavity is achieved in a controlled manner, so as to affect the electric and/or magnetic components of one or several of the eigenmodes of electromagnetic field supported by the RF cavity. In an embodiment the phase or frequency tuneable device further comprises additional electromagnetic structures in the vicinity of, or inside, the RF cavity, activation of the SMA actuator further causing deformation and/or movement of the additional electromagnetic structures so as to affect the electric and/or magnetic components of one or several of the eigenmodes of electromagnetic field supported by the RF cavity.

In a twelfth aspect the invention may broadly be said to consist in an RF tuneable filter device comprising a plurality of devices as in any of the preceding statements, each of the devices electromagnetically coupled to at least one other of the plurality of devices.

In an embodiment, at least one of the electromagnetic couplings is in the form of an iris penetrating the solid walls or ground planes separating the devices

In an embodiment, at least one of the electromagnetic couplings is in the form of an iris formed by a gap in a wall of conductive vias separating the Claim 1 devices.

In an embodiment, at least one of the electromagnetic couplings between the devices is formed by non-grounded cross- coupling wires protruding into both of the cavities of the adjacent device or devices through an iris.

In an embodiment, at least one of the electromagnetic couplings is formed by the provision of additional non-grounded conductive tracks formed on an insulating layer formed on the inside and/or outside of one or both of the ground planes sandwiching the cavities to be coupled, and wherein the conductive tracks are configured to protrude into both of the adjacent cavities of the devices.

In an embodiment, the RF tuneable filter device further comprises a tuning device comprising SMA material in the shape of wires or ribbons or sheets applied in such a way as to achieve controllable deformation or controllable movement of a conductive or dielectric tuning element in the vicinity of the electromagnetic coupling, and wherein at least one of the electromagnetic couplings between the RF cavities of the devices is configured to be tuneable and is connected to the tuning device.

In an embodiment, the RF tuneable filter device further comprises one or more stages, the filter either being of the low-pass, band-pass, band-stop, high- pass or phase-shifting configuration comprising:

two or more spaced conductive ground planes with joining walls connecting between the conductive planes and/or conductive vias positioned between the conductive ground planes, having inside between the ground planes one or a plurality of separate RF cavities separated by solid conductive partitions and/or by a plurality of conductive vias positioned between the conductive planes and in each of those cavities is zero, one or a plurality of resonators or electromagnetic reflectors, and where there is a plurality of cavities each cavity is electromagnetically coupled to at least one other cavity by an iris penetrating the solid walls or ground planes or by an iris formed by a gap in a wall of conductive vias between ground planes, and wherein one or more of the RF cavities has each one or more tuning elements penetrating into or wholly contained within the RF cavity and wherein each such tuning element is either wholly moveable or is deformable in such a way that the movement or deformation thereof changes the electromagnetic characteristics of the RF cavity so as to satisfy the tuneability requirement of the filter and wherein the movement or deformation of at least one of the tuning elements is caused by the expansion and contraction of one or more associated SMA structures each under the heating influence of a controlled electric current passing through said SMA structure and where each SMA structure is located outside of the RF cavity or within the walls of the RF cavity or located wholly within the RF cavity.

In an embodiment, the RF tuneable filter device further comprises at least one resonator wherein the or each of the resonators are made of one of: a conductive material; a low-loss dielectric material; a non-conductive material coated or plated with conductive material; a combination of the preceding elements.

In an embodiment, the RF tuneable filter device further comprises at least one dielectric resonator formed from high permittivity low loss RF ceramic.

In an embodiment, one or more of the resonators are in the form of one or more of: strips; T-shaped strips; rings; spirals or any other suitable shape that resonates at the required frequency.

In an embodiment, when using any resonator having geometry wit more than one eigenmode, the concurrent modes in the resonator are suppressed by shorting to ground the corresponding ends of the branches of the resonator structure.

In an embodiment, the resonator comprises a dual-mode or triple-mode resonator with a minimum of two or three mutually orthogonal branches with a single common point.

In an embodiment, the cavity is configured to support two or three orthogonal modes. In an embodiment, the tuneable RF filter comprises a plurality of RF cavities, external signal connections provided in the form of spaced input and output tapping points to a first or input cavity and a last or output cavity.

In an embodiment, the tuneable RF filter comprises a single RF cavity, and wherein an external signal connection is provided in the form of an input/output tapping point to the cavity.

In an embodiment, the tuneable RF filter comprises at least one resonator and at least two RF cavities wherein one or more of the resonators and zero, one or more of the inter-cavity couplings and zero, one or both of the input and output tapping points has each a tuning element penetrating into the resonator’s RF cavity or sited wholly within the cavity such that the movement or deformation thereof changes the capacitive loading or inductive loading or both of the associated resonator or coupling or tapping point, and wherein said movement or deformation of the tuning element is controllably caused by the controlled heating of one or more SMA-wires sited outside or partially or wholly within the RF cavity.

In an embodiment, the tuneable RF filter further comprises at least one resonator, wherein one or more of the resonators is sited within an RF cavity and is caused controllably to change shape or mechanically deform by the controlled contraction of at least one controllably heated SMA-wire such that the movement or deformation thereof changes the self capacitance and/or inductance of the associated resonator in such a way as to controllably tune the resonator.

In an embodiment, the tuneable RF filter further comprises at least one tuning element wherein each tuning element has the shape of a thin strip, or a rod, or a bar, or a tube, or more generally a long prismatic section with flat or curved surfaces.

In an embodiment, the tuning elements are made of one or more of: conductive material; low-loss dielectric material; non- conductive material coated or plated with conductive material; a combination of these elements.

In an embodiment, the dielectric tuning elements are formed from high permittivity low loss RF ceramic or a glass wafer.

In an embodiment, one or more tuning elements that is tuning a resonator is aligned in the same direction of greatest extension as the resonator that they are tuning so that the gap between the tuning element and the resonator is also aligned with the resonator. In an embodiment, the tuneable RF filter comprises at least one resonator, the resonator or resonators having a longitudinal slot configured to receive a tuning element without this touching the resonator.

In an embodiment, the tuneable RF filter comprises at least one conductive tuning element extending outside the RF cavity wall, wherein RF isolation for the portion of the tuning element protruding outside the cavity is provided by integrating an RF choke into the structure of the tuning element at least around the region where the tuning element exits the cavity and enters the cavity wall.

In an embodiment, the choke comprises a capacitive load at the external end of the tuning element sufficiently large to be considered an RF short.

In an embodiment, the choke comprises a series of one or more inductive sections each followed by a parallel capacitive section positioned down the length of the tuning element from the cavity to the external end of the element, the inductive sections being narrower than the capacitive sections.

In an embodiment, the tuneable RF filter comprises a plurality of tuning elements wherein each of the tuning elements are each caused to move by one or more actuators, with one or more of the tuning elements sharing an actuator.

In an embodiment, each actuator comprises an SMA-wire actuator and wherein the length of one or more sections of SMA wire are caused controllably to change by controllably changing the SMA wire temperatures.

In an embodiment, the temperature of an SMA wire is changed by controlling the RMS electric current passing through the SMA wire, the filter further comprising a programmable device configured to control the current.

In an embodiment, in one or more of the actuators are mechanically connected either directly or indirectly between the tuneable filter body and the moveable or deformable elements of the tuneable filter so causing the moveable filter elements to move relative to the filter body or to deform.

In an embodiment, one or more of the actuators are mechanically connected only to the filter component that is designed to deform with no mechanical connection required between the actuator(s) and the RF filter body.

In an embodiment, one or more of the actuators are mechanically connected only between two of the filter components that are required to move relative to each other. In an embodiment, the mechanical linkage of a tuneable element to its respective actuator is direct such that part of the tuneable element forms part of the actuator structure.

In an embodiment, at least one tuning element is formed form a dielectric material and the associated actuator is not wholly separated from the inside of a respective cavity by the solid conductive wall of the cavity is RF electrically isolated by the suitably close positioning to the tuning element of one or more conductive vias connecting between the conductive walls of the cavity.

In an embodiment, at least one tuning element is formed of a conductive material and wherein TEM mode propagation along the tuning element of RF energy from within the cavity to the outside of the cavity and towards its associated actuator is prevented by two or more buried vias located adjacent to and across the longitudinal line of the tuning pin and separated by substantially a half-wavelength corrected for the reactance introduced by the adjacent vias, for the propagation at this wavelength to be blocked and which capacitively loads the leaking TEM mode to stop the leakage.

In an embodiment, the tuneable RF filter comprises multiple actuators controlling the movement of the totality of tuning elements wherein the synchronisation of the movements of all of the tuning elements is electrically controlled by the

synchronisation of the appropriate control signals to the plurality of actuators.

In an embodiment, the control is by way of a pre-computed look-up table stored in the memory of a controller.

In an embodiment, the control is by way of an algorithm generating in real-time the actual required positions of all tuning elements to achieve the required state of the filter.

In an embodiment, the tuneable RF filter comprises at least one moveable tuning element wherein each moveable tuning element is movably supported by a tuning support structure, the support structure partially or fully dielectric, and/or partially or fully conductive.

In an embodiment, each SMA wire is enclosed within a dedicated void in one of the one or more Supports to ensure free movement of the SMA wire relative to the Support.

In an embodiment, the tuneable RF filter comprises at least one moveable tuning element wherein each moveable tuning element is positioned slidably in a channel through the support, the channel configured to ensure free movement of the tuning element while maintaining a precise gap between and accurate distance from the tuning pin to the corresponding resonator, coupling or tapping point for all positions of the tuning element controlled by the associated Actuator.

In an embodiment, each actuator is fully integrated into the filter.

In an embodiment, the tuneable RF filter comprises at least one resonator, wherein the one or more resonators are movably mounted within a cavity and caused to so move by direct or indirect mechanical connection to one or more SMA-wires.

In an embodiment, the tuneable RF filter comprises at least one resonator, wherein one or more resonators are configured so as to be deformable and are caused to so deform by direct or indirect mechanical connection to one or more SMA-wires.

In an embodiment, one or more of the deformable resonators has the form of one of: a thin strip; a flat-section spiral; a flat-section helix; a shape which has at least one direction of low force deformation.

In an embodiment, one or more of the deformable resonators is formed from elastic material such that it returns to its original shape after deformation and is capable of providing the necessary restoring force to stretch to its original cold length the SMA- wire that upon heating caused its deformation, once the SMA-wire has cooled again.

In an embodiment, the shape of the deformable resonator is such as to effectively enclose the SMA wire in a Faraday cage and so isolate it from the RF energy in the surrounding RF cavity.

In an embodiment, the SMA wire or wires provided to cause motion of the tuning element are attached to the low-impedance capacitive sections of the tuning element so as to maximally isolate them from any RF energy transmitted from within the cavity.

In an embodiment, the tuneable RF filter comprises an SMA-wire positioned partly or wholly within an RF cavity wherein strong coupling of the SMA-wires to the RF field in the cavity is prevented by the use of <100micron diameter straight SMA wire located entirely on or within the electric wall of the cavity.

In an embodiment, the tuneable RF filter comprises an SMA-wire positioned partly or wholly within an RF cavity wherein strong coupling of the SMA-wires to the RF field in the cavity is prevented by positioning the line of the wire orthogonal to and symmetrical to the magnetic walls of the cavity and parallel to the electric walls of the cavity.

In an embodiment, the tuneable RF filter comprises an SMA-wire positioned partly or wholly within an RF cavity wherein the SMA-wire or the SMA-wire together with its electrical connections are constrained to lie in a plane and that plane is positioned orthogonal to and symmetrical to the magnetic walls of the cavity and within the electric walls of the cavity

In an embodiment, one or more resonant or reflective elements are positioned in a waveguide with conductive walls and wherein the one or more resonant or reflective elements are caused to move axially along the waveguide each by an actuator.

In an embodiment, the waveguide conductive walls are formed from alternate metal and dielectric layers with adjacent metal layers joined together by rectangular arrays of conductive vias through the dielectric layers the rectangular arrays forming the walls of the waveguide whose axis is orthogonal to the metal and dielectric layers, and the waveguide cavity is formed by the removal of the dielectric and metal layers within and between the waveguide walls.

In an embodiment, one or more resonant or reflective elements are constructed so as to reflect as perfectly as practically possible all of the RF energy incident at one end of the waveguide back to that same end of the waveguide with a phase directly proportional to the axial position of the moveable elements along the waveguide thus providing a single-port reflective tuneable phase- shifting filter.

In an embodiment, two resonant or reflective elements are constructed so as to reflect as little as practically possible of the RF energy incident at one end of the waveguide back to that same end of the waveguide such that nearly all of the RF energy emerges from the other end of the waveguide with a phase directly proportional to the axial positions of the moveable elements along the waveguide and wherein a second Actuator is used to control the axial separation of the two resonant or reflective elements to optimise the input return loss with operating frequency thus providing a dual-port tuneable phase-shifting filter.

In an embodiment, the tuneable RF filter comprises two separate sets of one or more resonant elements each set independently of the other moveable axially along the waveguide by independently controllable Actuators, wherein each set of resonant elements is responsive to only one of two different polarisations of waves incident on one end of the waveguide, for plane polarisation waves the different polarisations being orthogonal to each other, and for circular polarisation the different polarisations being of opposite sign.

With respect to the above description then, it is to be realised that the optimum dimensional relationships for the parts of the invention, to include variations in size, materials, shape, form, function and manner of operation, assembly and use, are deemed readily apparent and obvious to one skilled in the art, and all equivalent relationships to those illustrated in the drawings and described in the specification are intended to be encompassed by the present invention.

This invention may also be said broadly to consist in the parts, elements and features referred to or indicated in the specification of the application, individually or collectively, and any or all combinations of any two or more said parts, elements or features, and where specific integers are mentioned herein which have known equivalents in the art to which this invention relates, such known equivalents are deemed to be incorporated herein as if individually set forth.

Therefore, the foregoing is considered as illustrative only of the principles of the invention. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the invention to the exact construction and operation shown and described, and accordingly, all suitable modifications and equivalents may be resorted to, falling within the scope of the invention.

BRIEF DESCRIPTION OF THE FIGURES

Further aspects of the invention will become apparent from the following description which is given by way of example only and with reference to the accompanying drawings which show an embodiment of the device by way of example, and in which:

Figure 1 shows a side view of an embodiment of a latching bi-directional linear SMA stepper-actuator, or BSA, of the present invention, the BSA comprising a stator top half, a stator bottom half, and a shuttle, the stator bottom half rigidly attached to the stator top half, the shuttle sliding freely within and moving relative to the top and bottom stator halves and coupled in use to an external load.

Figure 2 shows a side close-up view of a portion of the stepper mechanism of the BSA of figure 1 , showing detail of the shuttle which has ratchet teeth on each of it's two long edges, the ratchets facing in opposite directions so that those on the top edge allow ratchet sliding in the left-hand direction, and the ratchets on the bottom edge allowing ratchet sliding in the right-hand direction, SMA wires, springs and pawls for controlling movement of the shuttle within the stators also shown. Figure 3 shows a perspective side view of a variation of the linear actuator of figures 1 and 2, this variation providing a fast-slew mechanism.

Figure 4 shows a close-up perspective view of the linear actuator of figure 3 at the point where the shuttle exits the structure formed by the upper and lower stators, showing detail of the connection between the stator top and bottom halves via a base component that rigidly connects the stator halves, and two pawls located against the right-hand edges of the stator halves that in the embodiment shown in this figure are held against and interlocking with the ratchet teeth on the shuttle by a heated and contracted SMA wire, but which are otherwise held apart and off the shuttle by a bent wire spring which applies a separating force to these two pawls.

Figure 5 shows a perspective end view of the linear actuator of figures 3 and 4, showing detail of the fast-slew mechanism, the mechanism having an SMA wire actuator that straddles a pair of pawls and which when actuated pulls these pawls towards each other and into teeth-meshing contact with the shuttle, and further showing detail of the chamfered entry points in the upper and lower stators which help to ensure clean re-entry of the pawls at the end of a fast reverse slew back of the shuttle, so that the pawls locate properly against their reference edges in the stators.

Figure 6 shows a view of the linear actuator of figures 3 to 5 part way through an indexed movement where the fast reverse slew back mechanism has been engaged, the pawls shown locked to the shuttle through the action of the actuated SMA wire actuator and carried along with the shuttle to the right, this movement stretching the long tension spring and priming it for action on the fast reverse slew-back at the end of the stepped/indexed stroke.

Figure 7 shows a variation of linear actuator that includes side wings on the shuttle, the wings sitting below the moveable pawls and supporting them against movement parallel to the ratchet teeth, the wings fitting into a V-groove in the base element that is formed by the stators.

Figure 8 shows detail of an alternative support arrangement for the moveable pawls, the pawls in this embodiment having wings which wrap around the top and bottom sides of the shuttle, the wings holding the pawls firmly in place to assist with eliminating movement when the pawls are pulled together and locked onto the ratchet teeth of the shuttle, Figure 9 shows a rear view of a variation of linear actuator provided with

components that allow fast reverse slew-back for both directions of motion of the stepping linear actuator, the linear actuator having a first set of reverse-slew-back components as in the linear actuator of figure 1 , and a second set directed in the reverse sense relative to the ratchet teeth sets on the shuttle, the second set on the rear of the actuator and at the other end of the shuttle from the end used for the first set of such components.

Figure 10 shows a size comparison example of an actuator of an embodiment of the present invention configured with 100pm step-size and bidirectional travel of ±11mm (22mm total stroke), next to a US dime coin, both items shown to scale.

Figure 11 shows a close-up side perspective view of a stepper-pawl arrangement suitable for use in embodiments of the actuator of the present invention, with the stepper-pawl arrangement moved to the opposite side of its adjacent passive-pawl so that the SMA-wire actuator does not require routing through the passive-pawl, the pawls held in contact with the shuttle by compression springs, the pawls liftable together off the shuttle teeth by the action of an SMA-wire actuator.

Figure 12 shows a side view of an embodiment of linear stepper-actuator that uses a rack-and-pinion rack, the assembly comprising a grooved stator, a stepper-cage, a shuttle and two stepper-pawl stepper SMA-wire actuators.

Figure 13 shows a perspective close-up view of the linear-stepper embodiment of Figure 12 with the stepper-cage and shuttle removed to show detail of the two stepper-pawl stepper SMA-wire actuators, set into two grooves in the stator, and detail of a rod coupling the lift wire to the pawl.

Figure 14 shows a graph/chart of a performance example for a capacitor-driven SMA element using bursts for actuation, the element having a 25pm SMA wire with a roughly 130hm cold resistance surrounded by medium-thermal-conductivity

(0.2W/m.K) silicone rubber, the waveforms showing the effect of driving the SMA element from a 5V pulse burst source of 50 pulses of duration 100ps each and 200ps apart via a 10pF capacitor, the top chart showing the pulse burst drive voltage waveform, the second-top chart showing the voltage across the 10pf capacitor, the third chart showing the wire current, and the bottom chart showing the wire temperature, the x-axis showing time elapsed. Figure 15 shows a perspective view to one side and from above of a dual-cog arrangement suitable for use with any of the embodiments of figures 1 to 14 that can be used to balance the shuttle arrangement.

Figure 16 shows a side view of an embodiment of rotary inverse escapement suitable for use with the embodiments of linear actuator of the present invention, the escapement shown in two possible positions in the left- and right-hand views, the escapement comprising a verge that pivots about a pin fixed to a stator that rocks back and forth about an axis, so that teeth distal from the axis can interface with the teeth of a ratchet wheel, and SMA wires and their actuating mechanisms.

Figure 17 shows a perspective view from above and to one side of a first embodiment of SMA-wire mechanical overload prevention mechanism suitable for use with the actuators of the present invention, the mechanism having a wire mount mechanically coupled to one end of a pre-stressed tension spring via a first pin, the spring stretched so that the other end can be coupled to a second pin which rests in slots in the mount, a wire terminal with a crimp-end mechanically attached to the pin.

Figure 18 shows a perspective view from above and to one side of a second embodiment of SMA-wire mechanical overload prevention mechanism suitable for use with the actuators of the present invention, the spring in this embodiment having a terminal extended away from its crimp-section where the SMA-wire is attached via an extended section that is bent to form a serpentine tension spring, the extended section allowing for extension of the spring portion in the direction of the wire.

Figure 19 shows a perspective detail view of the spring of figure 18, the spring optimised for insertion directly into a printed circuit board (PCB) and having a single piece of flat metal formed into a crimp at one end, a pair of PCB insertion-pins and a serpentine tension spring, the tension spring section stretched in use into restraining mounts and held in place by the pins, the SMA-wire crimped into the crimp-end.

Figure 20 shows the spring of figure 19 inserted into a PCB, with a first pin inserted into a tight-fitting hole and soldered to a PCB track (not shown) to provide electrical connection, the other pin inserted into a slot which provides an end-stop for the pre stress tension in the spring.

Figure 21 shows an underside view of the PCB and spring of figure 20 and the tight fitting location of the first pin in the PCB, and a view of the slot giving clear space for the other pin to move freely under overload conditions, constrained only by the force of the spring. Figure 22 shows a close-up perspective view of the SMA-wire mechanical overload prevention mechanism of figure 18 applied to a bowstring actuator, an SMA-wire crimped at both ends to mechanical and electrical connections engaging with a load- attach-pin at or near the wire’s centre, the load attach pin mechanically connected to the push-rod of the actuator via a connecting member and pin which in turn engage with the spring whose other end is mechanically connected to the push-rod.

Figure 23 shows a perspective schematic view of a double bowstring SMA-wire actuator with first and second SMA-wire mechanical overload prevention

mechanisms as shown in figure 17, one for each of the bowstrings to protect against overload.

Figure 24 shows a perspective schematic view of an alternative embodiment of actuator using a double bowstring SMA-wire, with first and second SMA-wire mechanical overload prevention mechanisms as shown in figure 17, one for each of the bowstrings to protect against overload, but sharing a common overload spring.

Figure 25 shows a perspective view of a fast cycle-time dual SMA actuator.

Figure 26 shows a perspective view from above and to one side of a double bowstring SMA-wire actuator fitted with a fast cycle-time dual SMA actuator as shown in figure 25.

Figure 27 shows a perspective view from above and to one side of an alternative form of double bowstring SMA-wire actuator fitted with a fast cycle-time dual SMA actuator as shown in figure 25, sharing a common overload spring.

Figure 28 shows the analytically modelled results for an embodiment of a simple straight-wire SMA actuator optimised for high-speed drive according to an embodiment of the present invention, the SMA actuator comprised of 17mm of 25micron diameter SMA wire, and pushing a load of mass of 225mg and a frictional force of 0.011 N, the vertical scale shows output stroke in mm (0.0 to ~ 0.7mm) while the horizontal scale is time from initiation of stroke, in milliseconds.

Figure 29 shows further analytically modelled results for an embodiment of a simple straight-wire SMA actuator optimised for high-speed drive according to an embodiment of the present invention, the SMA actuator comprised of 17mm of 25micron diameter SMA wire, and pushing a load of mass of 225mg and a frictional force of 0.011 N, the vertical scale shows wire heating current (~50mA to nearly 300mA) while the horizontal scale is time from initiation of stroke, in milliseconds. Figure 30 shows further analytically modelled results for an embodiment of a simple straight-wire SMA actuator optimised for high-speed drive according to an embodiment of the present invention, the SMA actuator comprised of 17mm of 25micron diameter SMA wire, and pushing a load of mass of 225mg and a frictional force of 0.011 N, wire resistance [ohm] shown on the vertical scale against stroke [mm] on the horizontal scale is shown for maximum safe actuation speed.

Figure 31 shows the analytically modelled results for an embodiment of a single bowstring SMA-wire actuator optimised for high-speed drive as in one aspect of the present invention, comprised of 17mm of 25micron diameter SMA wire, with a crimp separation of 16mm, and pushing a load of mass of 225mg and a frictional force of 0.011 N, the vertical scale showing output stroke (or distance moved by the actuator load) in mm (0.0 to ~ 1.2mm) while the horizontal scale is time from initiation of stroke, in milliseconds.

Figure 32 shows the analytically modelled results for a single bowstring SMA-wire actuator having the same arrangement as described in figures 28 to 31 , showing drive current [mA] on the vertical scale against time [ms] on the horizontal scale is shown for maximum safe actuation speed.

Figure 33 shows the analytically modelled results for a single bowstring SMA-wire actuator having the same arrangement as described in figures 28 to 31 , the load acceleration [m/s/s] plotted against time [ms] to show how the available acceleration falls as stroke increases.

Figure 34 shows the analytically modelled results for a single bowstring SMA-wire actuator having the same arrangement as described in figures 28 to 31 , the drive voltage (for optimum safe high speed drive) plotted in volts against time in milliseconds.

Figure 35 shows an embodiment of a tuning element of the present invention relative to a resonator in a conventional solid-wall cavity.

Figure 36 shows an embodiment of a tuning element of the present invention relative to a resonator in a cavity constructed between a pair of ground planes.

Figure 37 shows an embodiment of two coupled and tuned resonators sandwiched between ground planes, with one plane removed for clarity.

Figure 38 shows an embodiment of the components needed for a complete tuneable two- resonator filter with I/O ports. Figure 39 shows an embodiment of an alternative arrangement of the components of the tuneable filter of figure 38.

Figure 40 shows an embodiment of a 3D folded tuneable six resonator filter with three ground planes.

Figure 41 shows an embodiment of an integral tuner, Actuator and resonator.

Figure 42 shows an embodiment of an SMA tuneable-phase reflective waveguide phase-shifter filter.

Figure 43 shows an embodiment of a two-Actuator tuneable through-waveguide phase-shifter filter.

Figure 44 shows an embodiment of an alternative SMA integrated capacitive tuning element.

Figure 45 shows an embodiment of a twisting-mode SMA magnetic I/O coupling mode tuning element.

Figure 46 shows an embodiment of a bending mode SMA tuned resonator.

Figure 47 shows an embodiment of an alternative Faraday cage bending mode SMA tuned resonator.

Figure 48 shows an embodiment of a form of waveguide-phase-shifter with phase adjusted by an Actuator.

Figure 49 shows an embodiment of a modified form of waveguide phase-shifter capable of independently varying the phase of each of two orthogonal polarisations, and capable of modification to independently tune each of two opposite circular polarisations,

Figure 50 shows an embodiment of a phased-array antenna comprising an array of SMA-tuneable phase-shifters with an offset feed and a schematic beam produced by the array.

Figure 51 shows an outside view of an SMA linear stepper actuator

Figure 52 shows the as-manufactured (pre-assembly) components of the main functional layer of the actuator of Figure 51

Figure 53 shows the components of Figure 52 as assembled, and the two SMA actuator wires.

Figure 54 shows the functional components of an SMA rotary stepper actuator. DETAILED DESCRIPTION

In this specification, the generic term SMA (Shape Memory Alloy) is used to refer to any or all forms of shape-memory-alloy material (e.g. metals such as NiTi or Nitinol nickel-titanium plus additives) as well as, where appropriate, any and all polymer equivalent materials which have similar thermo-mechanical properties (a thermally induced phase-change accompanied by a significant change in length in at least one direction). For SMA materials that are not electrically conductive then any reference to electrical Joule heating should be ignored for that particular material, and the defined temperature profiles for that material used instead.

References to SMA wire should be taken to refer to SMA strips or foil as well as wire.

Directional references, such as left, right, up, down, etc should be taken as referring to directions as shown in the embodiments on the figures. That is, as relative, and not absolute.

Embodiments of the invention, and variations thereof, will now be described in detail with reference to the figures.

Latching Bi-Directional Linear SMA Stepper-Actuator

An embodiment of a latching bi-directional linear SMA stepper-actuator, or BSA, is shown in figure 1. The BSA 100 can be categorised generally as a load-shifting SMA actuator (LSSA). The BSA 100 comprises three main components: a stator top half 1 , a stator bottom half 2, and a shuttle 3. The top and bottom stator halves comprise a pair of bodies that are rigidly connected or attached to one another, with a central channel or aperture 101 between the two. The shuttle 3 comprises an elongate member that is configured to locate into and move within and along the central channel 101 in use, the fit between the walls of the channel 101 and the shuttle 3 being snug or close, but allowing free movement of the shuttle 3. The shuttle 3 slides freely within and moves relative to the top and bottom stator halves 1 , 2, and is coupled in use to an external load.

A close-up view of a portion of the stepper mechanism is shown in Figure 2. The shuttle 3 comprises ratchet teeth 102 on its two long edges, with the ratchets on opposite edges facing in opposite directions, the teeth 102a on the top edge shaped, positioned, and aligned to allowing ratchet sliding in the left-hand direction, and the teeth 102b on the bottom edge shaped, positioned, and aligned to allow ratchet sliding in the right-hand direction. Upper and lower channels 103, 104 are formed in stators 1 and 2 respectively, aligned perpendicularly to the shuttle channel 101.

Four toothed-pawls 4, 5, 24, 26 are located in upper and lower channels 103, 104 that are formed within the upper and lower stator halves 1. 2 - upper pawls 4 and 5, and lower pawls 24, 26. The pawls 4, 5, 24, 26 move within the upper and lower channels 103, 104 in parallel with the long axis of the channels, to engage with and disengage from the shuttle ratchet teeth 102a, 102b. When engaged with the teeth 102a, the top pawls 4, 5 block or prevent movement of the shuttle 3 to the right, and the bottom pawls 24, 26 prevent movement to the left when engaged with the teeth 102b. The pawls can be lifted together off the shuttle teeth via activation of upper and lower SMA actuator wires 8 and 27 respectively. The pawls are slidingly connected to the wires 8, 27 by rigid links - upper link 10 and lower link 25

respectively. Links 10 and 25 are configured to fit loosely into apertures formed through the pawls, so that when not actively lifted by the SMA wires, the pawls may themselves be freely lifted by motion of the shuttle teeth engaging on the pawl teeth in the appropriate unblocked direction.

Note that with a linear stepper actuator as shown in figure 1 , the pawls may have multiple teeth arranged in a line so as to mesh with multiple successive teeth on the shuttle. This has the benefit of reducing wear on any individual pawl tooth, as well as reducing the interface forces between pawl and shuttle which in turn reduces the rate of wear, both for pawl and for shuttle.

The teeth of upper pawl 4 are held in contact with the shuttle teeth by a long bent- wire spring 9 anchored at the other end to base 1 , which exerts a force directed substantially towards the shuttle sufficient to retain pawl engagement under normal operating conditions, but small enough to be overcome by the lift-off force applied by SMA wire 8 when actuated. The same spring wire 9, aligned generally in parallel with the channel 101 also holds pawl 4 in contact with the left-hand edge of the upper channel 103, which acts as a leftwards mechanical reference position for this pawl. Pawl 4 is a passive-pawl that serves only to prevent reverse motion of the shuttle, when engaged with it.

A spring wire 6 holds the teeth of pawl 5 in contact with the shuttle teeth 102a.

Spring wire 6 is entirely above the stator and channels in the stator, except for the portion which enters pawl 5. The left hand end of spring wire 6 is clamped to base 1 so as to apply a downward and a rightward force to pawl 5, locating it against the teeth 102a of the shuttle and against the right-hand reference wall of channel 103. Pawl 5 is also attached to the end of SMA wire actuator 11 whose other end is connected within stator 1 in a cavity 12 formed in stator 1 and opening onto the upper side of the channel 101. When the SMA wire of wire actuator 11 is actuated (e.g. by heat or electric current) and when both pawls 24, 26 are lifted off the shuttle by SMA wire actuator 27, the combined effect is to pull pawl 5 to the left by just more than one ratchet/pawl tooth length. When this occurs (and when both pawls 4, 5, are in teeth-mesh contact with the shuttle 3), the motion to the left of pawl 5 also pulls the shuttle 3 to the left by the same distance, the shuttle 3 sliding under pawl 4, which is lifted on the sloping edges of the teeth 102a, until after more than one tooth distance has passed, pawl 4 then drops back down into full mesh contact locking the shuttle 3 into its new position. At this point, SMA wire 11 is de-actuated, and pawl 5 is then pushed back to the right by the force of spring wire 6, slipping over the teeth 102a on the shuttle 3, until it contacts the right-hand side of the channel 103 again. The gap between pawls 4, 5 is greater than one tooth length but smaller than two tooth lengths to allow the actuated movement of pawl 5 and to ensure that the shuttle moves no more than one tooth length per actuation cycle. Pawl 5 which is an integral part of the stepper-mechanism is called a stepper-pawl.

A mirror arrangement is present on the opposite side of the shuttle 3, with a lower channel 104 opposite and parallel to the channel 103, with pawls 24 and 26 corresponding to pawls 5 and 4 respectively, springs 28 and 105 corresponding to springs 6 and 9 respectively, SMA wire actuator 27 corresponding to SMA wire actuator 8, and SMA wire actuator 23 corresponding to SMA wire actuator 11 (with differences outlined below) connected within stator 2 in a cavity 50 formed in stator 2 and opening onto the lower side of the channel 101. The teeth 102b on the lower side of the shuttle 3 are arranged in the opposite direction to teeth 102a on the upper side of the shuttle 3.

However, on this lower side of the channel, SMA wire 23 now has one end fixed in stator 2 at location 20, to the right of the lower channel 104, with the other end attached to a coupling link 21 located within cavity 50 which in turn rigidly couples to the first or left-hand end of push-rod 22 whose second end pushes to the right on pawl 24 (passing through a hole in pawl 26 en route), against the restoring force of bent-wire spring 28, which in this lower half of the actuator pulls pawl 24 to the left, and pushes it upwards so that its teeth mesh with those on the shuttle 3. Otherwise than reversing the direction of motion of the shuttle this lower half of the mechanism operates in the same way as that just described for the top half. An alternative arrangement that avoids the use of push-rod 22 is to move the mechanism (the step-right mechanics described in the paragraph above) to the left away from the right-hand edge of base 2, so that there is space for the SMA actuator to be placed to the right of pawl 24 and to be mechanically attached thereto at one end of the SMA wire 23 and have the other end of the wire rooted in base 2 to the right of the right-hand edge of the channel 104. This would enable the SMA wire to pull pawl 24 to the right directly without the use of any link or push-rod.

In figures 1 and 2, the bent-wire springs are shown as a series of straight wire sections with right-angled joints for simplicity. However, in practice each spring would be made from a single piece of wire with smooth bends rather than the right- angled joints shown. Alternative spring forms such as coil springs and flexures can equally well be used and the illustrations should not be taken as limiting in this respect. Particularly when flexure-type springs are used these may be integrated with the base components 1 , 2 if these base components are made of suitable material (e.g. spring steel or stainless steel), by for example, etching or laser-cutting or ion- beam-cutting, or punching them out of the solid. Similarly, the pawls may then be also be integrated with the springs by a similar process in the same manufacturing step.

The SMA-wire actuators in the figures are in this embodiment electrically operated. However, for clarity, the electrical connections to the SMA wires are not shown in these figures. It is however convenient, where the body 1 ,2 material and/or the spring material is electrically conductive to use the body or the springs or both as at least one electrical connection to one or more of the SMA wires, and this is conveniently achieved by crimping those SMA wire ends to the body or spring to which they are required to be mechanically attached.

Where an end of an SMA wire is mechanically fixed directly or indirectly to the stator, then no special connection problems arise, and for example, an electrical connection may be made via a wire-crimp soldered to a printed conductive (e.g. copper) track as per standard PCB technology. However, for those SMA wire ends mechanically attached to moveable components (primarily the pawls in these embodiments), a preferred solution when those components are not electrically conductive (and thus cannot themselves be used as an electrical return path) is to co-crimp such an SMA- wire Free-End with a flexible (and preferably insulated) conductive wire, preferably stranded copper wire, whose gauge and flexibility are consistent with the electric current to be supplied to the SMA-wire, and the forces the SMA-wire is capable of supporting in actuator mode. Alternatively, a conductive track may be laid over the surface of the nonconductive component from the SMA wire-end to a remote electrical connection point, to avoid the use of separate wires.

The BSA stepping cycle for this embodiment is as outlined below, with individual times for each of the sequence events shown in [square brackets] at the end of each event.

In operation, t _on is the time required to heat and actuate an SMA wire, and t _off is the time required for the SMA wire to cool and be ready for another actuation.

1. SMA wire actuator 27 is actuated to lift pawls 24, 26 in a pre-step-left

operation [t _on ];

2. Cycle starting condition - at step position N, pawls 4, 5 are engaged, all SMA- wires (other than 27) de-actuated, pawl 5 at its RH end stop [0];

3. SMA wire actuator 1 1 is actuated to pull pawl 5 left. This pulls the shuttle 3 with it, and pawl 4 is prevented from moving left by the LH wall of cavity 103 and so is lifted up by shuttle 3’s ratchet teeth, while shuttle 3 is shifted more than one tooth left beneath it [t _on ];

4. SMA wire actuator 11 is de-actuated to release pawl 5. This allows spring 6 to re-stretch SMA wire 1 1 and to push pawl 5 over the shuttle 3 ratchet teeth to its right end stop [t _off ];

5. Return to step 2. Or if at end of sequence of shift-left operations goto step 6.

6. De-actuate SMA wire 27 allowing springs 105 and 28 to re-stretch it and to push pawls 24, 26 into engagement with ratchet teeth 102b of shuttle 3, thus locking the shuttle into place.

So the total cycle time [t _cyc ] for one step in a series of N steps to the left, is t _cyc = t _on + t _0ff which is very efficient in terms of operational steps/cycle, and about as fast as feasible for such an SMA stepper device, with a single stepper-pawl per direction.

This step-left actuation cycle as described assumes that the actuator has to do work against the shuttle to move it to the left, e.g. if it is pulling against a spring, friction, or gravity. However if the external load is pushing the shuttle to the left with sufficient force F _r to overcome the braking effect of the ratchet pawls, then when both pawls 24, 26 are lifted off the shuttle by SMA wire actuator 27, the external load can move the shuttle 3 leftwards in an uncontrolled manner, if it pushes sufficiently hard, as the shuttle’s ratchet teeth 102a will simply slip underneath the engaged pawls 4, 5. The value of F _r can be adjusted by modifying the slope of the ratchet pawl teeth (as described below). Increasing the slope angle of the teeth increases F _r but also requires more of the stepper SMA-wire actuator’s maximum load force capability F _w [wire maximum tension] to lift the passive pawl over the shuttle’s teeth. If F _w is sufficiently greater than the maximum rated BSA output force F _| then the tooth slope angle can be made steep enough to hold the load in check against reverse load forces, and the stepping cycle scheme described above may then be used successfully.

Where there is insufficient wire tension/force F _w to provide both BSA output force F _| as well as pawl-lifting force, the BSA structure can be modified so that each pawl now has its own lifting SMA-wire, enabling all pawls to be independently engaged and disengaged. The stepping cycle for this variation is as follows:

1 . Starting condition - at step position N, all pawls engaged, all SMA-wires de- actuated;

2. Lift pawl 26 off the shuttle teeth and simultaneously actuate SMA-wire 11 to pull pawl 5 to the left bringing the shuttle 3 with it [t _on ];

3. Movement of the shuttle 3 to the left, even if the external load attempts to overdrive the shuttle 3, is limited to -1.5 tooth lengths, because pawl 24 is still engaged with the shuttle and can move to the left only as far as coming into contact with the right side of pawl 26 [0];

4. Meanwhile the shuttle 3 has potentially moved leftwards beneath pawl 4 by more than one tooth-length, lifting it on its teeth in the process, so pawl 4 is now sitting over the next set of shuttle teeth [0];

5. Drop pawl 26 back onto the shuttle teeth, which is now sitting over the next set of shuttle teeth [t _off ];

6. De-actuate SMA-wire 11 and actuate SMA-wire 22 to pull the shuttle 3 to the right (about half a tooth length) until pawl 5 stops further motion by engaging with stator 1 on its RH side [max(t _off , t _on )];

7. The return force of spring 6 acting on pawl 5 pushes pawl 5 back to the right slipping over the shuttle teeth until stopping in contact with stator 1 on its right-hand side [0];

8. De-actuate all SMA-wires; [t _0ff ]

9. Final condition - at step position N+1 , all pawls engaged, all SMA-wires de- actuated [0];

The total cycle time has now increased to t _cyc = t _on + t _off +max(t _off , t _on ) +t _0ff

Assuming that t _off > t _on then t _cyc = t _on + 3 ^* t _off , which is significantly longer.

Under certain conditions, it is possible for the external load to push the shuttle 3 to the left immediately as pawl 26 is lifted and before SMA-wire 11 has actuated, so that pawl 5 slips over the shuttle teeth one step to the right prior to actuation of 11. As a consequence the full actuation of SMA-wire 11 is blocked and only—1/2 a step of movement is possible. Therefore, in the preferred embodiment, SMA-wire 11 (and similarly 22) are provided with force overload protection, as described in detail below.

Actuation of the shuttle 3 to the right is similar to the description above, but now using pawls 26 (passive-pawl), 24 (stepper-pawl), and SMA wire 23, and with the other two pawls 4, 5 lifted off the shuttle by means of SMA actuator wire 8.

When all SMA actuator wires are de-actuated the shuttle 3 is locked in place by the four pawls, 4, 5, 24, 26 providing a zero-power hold facility. Because each of the SMA-wires has to move its load by a small distance only (< 2 tooth lengths or < 2 ratchet tooth heights) then the SMA-wires can be short and thus consume very little power. The SMA-wires all operate directly in tension on their loads and so can apply their full tensile force capability (minus the opposing spring forces) to their loads.

The bidirectional stepping linear actuator is almost planar. A typical ratchet tooth size may be as small as 100mpi long by 50 mhi high, in which case the SMA wires can be as small as ~2.5mm long, and 25 mhi in diameter, and consume only a few mJ of electrical energy to heat them to the Austenite state from ambient.

The electrical connections for applying heating current to the various SMA actuator wires are not shown (for clarity), and the ends of the SMA wires are electrically insulated from any metal parts (only) as required to avoid short circuits and to suitably isolate them (again, not shown for clarity).

BSA With Latching Pawl Lifts

When operating in any one of the two directions, a load-shifting SMA actuator (LSSA) designed to step in a particular direction operates completely autonomously, as for example in the embodiments of LSSAs described above. The other LSSA is taken out of play by lifting all of its one or more pawls out of engagement with the shuttle, using SMA wire actuator(s). In many applications there will be a sustained period of stepping in one direction, followed by a possibly also sustained period of stepping in the other direction. In this scenario drive power used to continuously power the SMA wire pawl-lifts of the out of play LSSA is wasted.

This can be avoided by using latching pawl-lift mechanisms on each of the pawls. These have two stable unpowered positions: pawl-down and pawl-up. This can be achieved by any of the many latching mechanisms known in the art. These latching mechanisms are then controlled by at least one but preferably two SMA wire actuators each (one SMA wire actuator can be used if a pull-on/pull-off type of latch is used. Two SMA wire actuators are used if a separate latch-wire and unlatch-wire are provided). A two-SMA-wire driven latching mechanism is preferred, as these provide a definitely-known state (definitely on, or definitely off), whereas if using a single-wire pull-on/pull-off system, the initial state of the system cannot be known at first power- on without using a separate sensor subsystem. With latching pawl lifts, one of the LSSAs can be mechanically decoupled from the shuttle by a single short electrical pulse to its latched pawl-lift SMA-wires and thereafter will consume zero power, until it subsequently is needed to drive the shuttle in the opposite direction which is achieved by one more short electrical pulse to the pawl-lift SMA-wires (whereafter the other LSSA is similarly put out of action until next required).

Increased Stepping Time Rate

The maximum rate of stepping of a linear actuator as described (and thus maximum speed of output motion) is limited by the thermal time constant of the SMA wire actuators used to control the various movements of the pawls. In general this can be maximised by using the thinnest possible SMA wire (shortest thermal time constant) consistent with a given force output.

More force at the same speed can be achieved by using parallel pairs or multiples of thin SMA wires in place of each single SMA wire actuator. Shorter stepping times may also be achieved by improved electronic control which can deliver a high power pulse for a short time at the initial heating phase of an SMA wire, the pulse delivering enough energy to match the thermal energy capacity and latent heat of phase- change of the wire to very rapidly raise the temperature of the SMA material and convert the Martensite material of the SMA to Austenite, without over-raising the temperature of the material. Such techniques give very fast 'rise-times' for the phase-change and thus pulling force of the SMA wires. However, these techniques cannot assist with the 'fall-time', which is limited by natural cooling rates. Assisted or forced cooling is possible, but this is complex and often costly. Thermally heat-sinking the SMA wires to larger components (e.g. the stator) will cool the wires more quickly (as described in detail later) and lead to shorter actuation cycle times, but at the cost of increased power requirements to actuate.

However, the type of indexed linear stepper - the BSA as described above - can be used to increase actuator step-speed (and thus actuator output speed) by providing additional stepper-pawls operating in parallel with each other along the length of the shuttle 3, each with an associated mechanism including SMA wire actuator and drive control.

The method is as follows:

After the first stepper pawl SMA actuator (as already described above) is rapidly heated, stepped and then turned off all in time t _h , this takes time t _c (a time much greater than t _h ) to cool enough to be ready for re-actuation.

A second additional stepper-pawl SMA actuator is actuated at a time t _c /2 after activation of the first stepper pawl. This then drives another step in the same direction while the first actuator is still cooling, and because of the ratchet operation of the pawls, the two actuator cycles do not interfere with each other.

Thus, adding a second stepper-pawl mechanism doubles the speed of stepping and thus of actuator output (and also doubles the average maximum power

consumption). This technique is easily extended to three or more such stepper-pawl mechanisms, three giving a speed three times as great as one, etc.

If there is insufficient space along the shuttle 3 to fit such multiple pawls along the length of the same set of ratchet teeth, then a second (or third, or more) row of ratchet teeth may be provided in parallel with the first on the shuttle, and the multiple ratchet pawls are divided amongst the parallel sets of shuttle teeth as best suits the mechanical layout. These will operate in parallel with significant integral speed up factors. Alternatively, the shuttle teeth width may be made greater than (multiples of) the pawl teeth width, and a pair of pawls (or more) may then share the same position along the shuttle but sit on different portions of the shuttle’s teeth width.

For example, using the thinnest easily available commercial 25pm SMA wire for the actuators limits the stepping rate of a single SMA wire actuator to roughly five to fifty steps per second (one step per 200ms to 20ms) depending on the ambient conditions, wire parameters, any wire-heatsink methods or apparatus used, and mode of electronic drive used. Adding a second parallel stepper pawl mechanism will increase this step rate to ten to one hundred steps per second (one step per 100ms to 10ms), and adding a third can achieve fifteen to one-hundred-and-fifty steps per second (one step per 67ms to 6.7ms). Still more parallel steppers may be used if necessary. Such a speed-up technique is not available with simple (non stepping) SMA wire linear actuators.

Modification For Higher Force Output

Some of the available force output of the SMA wire actuators used to move the stepper pawls (e.g. pawl 5) is needed to overcome the required return force provided by their associated springs (e.g. spring 6) and so is not available to move the output load via the shuttle (SMA wires have to be physically stretched back to their original length after they have been heated through the Austenite phase, otherwise they will remain at least partly contracted even when cooled and e.g. spring 6 serves this function, amongst others). Depending on the precise operating mode of the wire, this can be as much as 40% of the rated load of the SMA wire.

Where this is more‘lost’ force than an application can withstand, and where the use of thicker SMA wires is ruled out for speed of operation or power consumption considerations, the device can be modified by removing the return spring of the stepper pawl’s SMA wire, and replacing this with another length of SMA wire connected in its place between the pawl and the stator /base, so as to act in opposition to the stepper pawl’s first SMA wire actuator. The stepper pawl can now be pulled in two opposite directions, one for each SMA-wire attached.

These are actuated only one at a time as follows:

• To step, the first (original) SMA wire is actuated by heating and thus contracts which pulls the pawl in the direction of contraction of that wire, while the second SMA wire is unheated and remains at or near ambient temperature, and the actuator steps.

• At the end of the step, and once the non-stepper pawl for the same direction has latched in place, the first SMA actuator wire is unheated and begins cooling. With no restoring force it will not expand very much.

• The second SMA wire actuator is heated and thus contracts, pulling the pawl in the opposite direction and in so doing stretching the first SMA wire back to its original unheated length. Once that point is reached, the second SMA wire actuator is unheated and begins cooling, to complete a stepper cycle. This method of operating the stepper pawl SMA wire actuators results in significantly more of the wire’s rated force loading to be used to move the external load of the linear stepper actuator, at the cost of additional SMA wire, and higher electrical power consumption as well as a slight increase in complexity of the electronic controller driving the actuator, but can be useful in certain applications.

A variation of the actuator 100 is shown in figures 3 and 4. The actuator 3000 of figures 3 and 4 is broadly similar to the actuator 100 described above, with additional components to allow fast slewing back to a previous indexed output position, in one of the possible actuator directions. The actuator 3000 is configured so as to provide a fast reverse slew-back operation to an indexed position.

A fast reverse slew-back to an indexed position is defined as follows: the linear stepping actuator is being used to produce a stepped, indexed movement of the shuttle from left to right from starting position step N1 to ending position step N2, and it is desired that at the end of this stepped linear indexed movement the shuttle (in this embodiment shuttle 3003 should return as quickly as possible (and much more quickly than can be achieved with a reverse stepping action) to a third position step N3, where N1 <=N3 <= N2 (i.e. position N3 is passed through in going from N1 to N2). This sort of movement scenario is important for certain applications, and can be referred to as a reverse slew-back to indexed position.

In this embodiment, the upper and lower stators 3001 , 3002 are the same as the previously described stators 1, 2, and in this variation are shown connected by a further base component 3041 which supports the stators 3001 , 3002 and rigidly connects them. A cover piece 3040 forms a rigid base extension over the top of and guiding the shuttle 3003.

As well as pawls 3004, 3005, 3024, 3026 similar to those described for the first embodiment above, in this embodiment an additional pair of pawls 3046, 3047 are slidably positioned on the shuttle, 3003, locating against the right-hand edges of stators 3001 , 3002. The pawls 3046, 3047 are normally (in normal bidirectional stepping operation of the linear actuator) held apart and off shuttle 3003 by the additional bent wire spring 3048 which applies a separating force to pawls 3046, 3047.

Pawl 3046 meshes with the upper ratchet teeth 3102a of shuttle 3003, and is also mechanically coupled to one end of a long tension spring 3045 whose other end is rooted in the base at 3046. This tension spring 3045 provides a pulling force on pawl 3046 in the direction along the line of the shuttle, to the left in the drawing, and in normal bidirectional stepping operation, ensures that pawl 3046 is held tight against the adjacent edge of stator 3001 which is a mechanical reference point for this pawl. The connection between spring 3045 and pawl 3046 is not completely rigid, but does provide some guidance to the alignment of the pawl and maintains its orientation such that its row of ratchet teeth are approximately parallel to those of shuttle 3003.

The ratchet teeth on pawls 3046, 3047 may each fully mesh with the nearest ratchet teeth on the sides of shuttle 3003. A compression spring may be used instead of tension spring 3045 by suitable rearrangement of attachment points and positioning.

A compression spring has the advantage of most likely providing a more compact arrangement, given the same force and spring-constant requirements. However, a compression spring will also likely require additional mechanical guidance to prevent it buckling, which may be achieved simply by, e.g. embedding it into a loose-fitting slot in the base 3041 or indeed in the shuttle 3003.

Spring 3048 acts to hold apart the two pawls 3046, 3047 so that they have no contact with the teeth of the shuttle 3003. However, they may also be pulled towards each other against the separating force of spring 3048 by the action of an additional SMA wire actuator 3049, which mechanically links these two pawls, and when heated to the Austenite phase, pulls both pawls into contact with shuttle 3003, as shown in figure 3. The normal position of these two pawls (with the SMA actuator wire 3049 unpowered) is away from and off the shuttle, and the position shown in Figure 3 is the position they occupy when SMA wire 3049 is actuated. For subsequent fast slew- back left to position N3, then SMA wire actuator 3049 is actuated and thereafter held actuated once the shuttle is right-stepped into position N3 during the prior right-step sequence from N1 to N2. Pawls 3046, 3047 can be referred to as 'floating-pawls' because they are not anchored directly to either the shuttle or the stator and may move significantly relative to both of these primary components.

Another view of the fast-slew mechanism of the actuator 3000 is shown in figure 5, where can be seen a clearer view of SMA wire actuator 3049 that straddles pawls 3046, 3047 and which when actuated pulls those pawls towards each other and into teeth-meshing contact with the shuttle 3003. As shown, the entry points 3050, 3052 in the stators 3001 , 3002 are chamfered, which helps to ensure clean re-entry of the pawls 3046, 3047 at the end of a fast reverse slew back of the shuttle, so that the pawls locate properly against their reference edges in the stators 3001 , 3002.

Figure 6 shows a view of the linear actuator part way through an indexed movement to the right, the fast reverse slew back mechanism engaged at some position along the stepped movement. It will be seen that the pawls 3046, 3047 have now been carried along with the shuttle 3003 also to the right, as they are locked to the shuttle through the action of the actuated SMA wire actuator 3049, and that this movement has in turn stretched the long tension spring 3045, priming it for action on the fast reverse slew-back at then end of the stepped/indexed stroke.

It should be noted that it is not necessary to fast-slew-back at the end of the stroke - priming the mechanism simply allows for this possibility, but at the end of the stroke it is still possible to simply step back in the opposite direction in a controlled indexed manner. However, if the fast reverse slew-back is activated at stroke end, then the locked position of the pawls 3046, 3047 ensure that return to the selected step (N3) is guaranteed as these pawls will re-engage with base components 3001 , 3002 at the end of the fast reverse slew at their position reference edges.

For a left-to-right stepped motion, the steps to accomplish this are as follows:

1 . Initially SMA wire actuator 3049 is unheated / de-actuated and spring 3048 holds pawls 3046, 3047 off the shuttle teeth. Tension spring 3045 pulls pawl 3046 back to its reference position against base 3001 , and the linking action of spring 3048 tends to pull pawl 3047 also back against base 3002 although this is not critical for this operation.

2. The stepping motion of the actuator commences as described above. The shuttle begins its indexed motion starting from position N1 , moving to the right towards N2.

3. When position N3 is reached, SMA wire actuator 3049 is heated / actuated pulling pawls 3046, 3047 into tight contact with the shuttle 3003 and the ratchet teeth of pawl 3046 intermesh with those of the shuttle and lock this pawl into position N3 on the shuttle.

4. On the next step of the shuttle to the right (on its journey to position N2) the shuttle takes pawl 3046 with it, and this pawl in turn pulls on tension spring 3045 and extends it by one step-length. Meanwhile SMA actuator wire 3049 remains actuated until the end of the indexed motion to N2. Pawl 3047 is moved by the shuttle, along with pawl 3046, but its ratchet teeth are sloped in the opposite direction so it is incapable of applying significant force to the tension spring - that is all done by pawl 3046.

5. On successive steps of the shuttle to the right, the tension spring 3045 is further extended by the shuttle, via pawl 3046. 6. At the end of the indexed movement to N2 (i.e. once N2 has been reached), as soon as the controller is given the command to fast slew-back, then pawls 3004, 3005, 3024, 3026 (which are part of the indexed stepping mechanism previously described) are all rapidly lifted off the shuttle by SMA wire actuators 3008, 3027, and the shuttle 3003 is then free to slide within the channel formed by the stators 3002, 3003, and the base member 3041. However, the long tension spring 3045 is extended to the right by (N2-N3) steps (teeth-lengths) and applies a restoring force to the left, to the shuttle via pawl 3046 which is still locked to the shuttle. This causes the shuttle to rapidly accelerate to the left, until pawls 3046, 3047 come into contact with the stators 3001 , 3002 which occurs when the shuttle reaches position N3 (the position where pawls 3046, 3047 were initially locked to the shuttle). Overshoot of the shuttle due to its inertia (and inertia of the external load) is prevented primarily by pawl 3047 which prevents further leftward relative movement of the shuttle.

7. Once position N3 has been reached, then pawls 3004, 3005, 3024, 3026 are released by de-actuating SMA wire actuators 3008, 3027, and re engage with the shuttle 3003.

8. Finally, SMA actuator wire 3049 is de-actuated and pawls 3046, 3047 are pulled off the shuttle 3003 by the spring action of spring 3048.

The fast reverse slew is now complete and the actuator may be stepped normally in either direction from this new starting position. It is preferable to provide some additional mechanical buffering and/or damping to reduce the impact force when the shuttle does a rapid relatively uncontrolled return under the action of the long tension spring. A compliant end-stop may be adequate. More velocity control may be achieved by e.g. by a gas-damper connected between the shuttle and the stator, which can conveniently and cheaply be incorporated into the body of the stator (e.g. a long blind hole in the stator parallel to the shuttle movement direction, open at the end of the actuator, and a long piston fitting into that hole attached to the shuttle - a small leak-hole then providing the damping rate). In a moulded plastic

implementation in particular, this need hardly increase the cost of the BSA. The shuttle itself may be used as such a piston if the base/stator is arranged to enclose the end of the shuttle near the end of its stroke in a nearly airtight manner - e.g. by assuring close clearance. For reliable operation where slew-back position N3 is close to the left-most position attainable by the actuator, then the slew-back spring 3045 must be suitably pre-tensioned at this left-most position to ensure adequate return force against external loads and internal actuator forces.

Additional support structure can be provided to maintain the pawls 3046, 3047 in their locations on the shuttle during a stepped stroke when fast reverse slew-back is engaged.

An example of this additional support is shown in the embodiment of Figure 7, where the actuator 701 is substantially similar to that described above and shown in figures 3, 4, and 5, except with the addition of side wings 761 , 762 added to the shuttle 703. The wings 761 , 762 are located below the two moveable pawls 746, 747 to support them against movement parallel to the ratchet teeth when they are extended beyond the base/stator. The V-shape of the wings fits into a V-groove in the base 741 , acting as a guide to the shuttle clear of its ratchet teeth.

An alternative means of supporting the moveable pawls 846, 847 is illustrated in the embodiment of figure 8. Here the pawls have wings 870 and 871 which wrap around the top and bottom sides of the shuttle 803. With the pair of pawls 846, 847 pulled together by their SMA actuator wire, and thus locked onto the ratchet teeth of shuttle 803, the wings 870, 871 hold the pawls 846, 847 firmly in place, eliminating any remaining degrees of freedom

As no modification to the shuttle 803 is required in this variation, the wings can be used with the bidirectional version of the fast reverse slew-back mechanism without interfering with it. A small notch in base 841 around the shuttle entry point gives clearance for the pawl wings when the pawls are docked in their reference position against the base.

Electrical connections (not shown in the figures) to the SMA actuator wires connecting the moveable pawl pairs can be provided by a pair of lightweight flexible insulated and preferably stranded conductive cable, one wire to each end of such an actuator wire, with enough of a free length of this flexible wire to avoid any restriction to movement of the pawls on and along with the shuttle. Alternatively a flexible printed circuit can be used. The fast slew-back mechanism as described, when in action, requires the SMA wire actuator 3049 holding the two floating-pawls to be continuously activated for the duration of the stepping motion from step N3 to step N2 and then until the slew-back is complete. If a slow release of the floating-pawls’ SMA wire actuator 3049 is acceptable, this SMA wire can be air-cooled and therefore little static power is required. However, for applications which demand a very fast release of the floating pawls then a modified scheme may be employed. Here the floating pawls are held together by a latching actuator which is pulled ON (i.e. pulling the floating pawls into meshed contact with the shuttle teeth) by a first SMA wire actuator becoming actuated, and which is then held in this position by a mechanical latching mechanism (of any suitable form of the very many known in the art) which incorporates a spring stressed (compressed or expanded) by the latching event and which in turn continues to apply meshing force between the floating-pawls and the shuttle, even after the first SMA actuator is released (de-actuated). The floating pawls then continue in this meshed state until at a later time a second SMA wire actuator upon actuation unlatches this latched actuator, and may then be immediately released again. Because both the first and second SMA actuators just described need be actuated only transiently (to latch, or unlatch the floating pawls) their static actuation power is unimportant and only the transient actuation energy is of concern. Thus these two SMA wire actuators for latching/unlatching may preferably be heat-sinked to the surrounding mechanical structures (as described hereafter), thus ensuring fast de-actuation times.

It should be noted that some latch mechanisms require just one actuation device (e.g. SMA-wire) and operate in a push-on/push-off manner. Such a device could be used here. However, it is not preferred, as it is not possible to know the state of such a latch after first power-on, which is undesirable. A dual actuated latch as described is highly preferable.

Simplified Fast Unidirectional Reverse Slew-Back To Fixed Position

If the requirement is for a fast reverse slew-back to zero position (i.e. to the beginning of the stroke of the stepper actuator) then the slew-back mechanism as described above can be modified as follows:

The additional two pawls, their associated SMA-wire actuator, and separator spring are not present in this variation. The long tension spring is connected directly to the shuttle such that it provides a return force to the shuttle in all stepper actuator positions. An end-stop is rigidly fixed to the shuttle (or the shuttle is formed with an integral end-stop) that limits the return stroke of the shuttle to the zero position by coming into contact with reference location(s) on the base components, in the same or a similar manner to the pawls in the variation described above.

In operation, at any position along a stroke of the stepper actuator, all of the pawls may be lifted off the shuttle by their associated SMA wires, and the long tension spring will then apply a restoring force to the shuttle which will very rapidly return to the zero position when the end-stops will hit the“buffers” on the base.

This simplification has two effects:

Firstly, the tension spring always loads the stepper actuators and reduces the maximum available output force (whether or not slew-back is to be used), and;

Secondly, the total stepper actuator power consumption is lower because the slew- back mechanism no longer has an SMA-wire actuator for pawl-clamping to be powered during the stroke.

Simplified Unidirectional Stepper Motor With Fast Reverse Slew-Back

For applications where stepped indexed motion is required in only one direction (a unidirectional stepper motor) the provision of the fast reverse slew-back mechanism just described allows for a simplified motor. The stepper motor mechanism for one of the two directions can be removed completely (e.g. the two pawls 24, 26, and their associated springs, plus the SMA wire actuator 22 and its connecting mechanism 21 , 23) and the slew-back mechanism alone is used for the return stroke.

This can save weight, size, operating power and cost.

Fast Bidirectional Reverse Slew-Back

It is also possible to provide fast reverse slew-back for both directions of motion of the stepping linear actuator. To achieve this it is necessary to add a second set of reverse-slew-back components, with this second set directed in the reverse sense relative to the ratchet teeth sets on the shuttle. One convenient way to do this without having these components interfere with the others already described is to place them on the rear of the actuator and at the other end of the shuttle from the end used for the first set of such components. Such an arrangement is shown in Figure 9.

In figure 9, the shuttle 903 extends from both sides of the base / stator components 901 , 902, 941 and two additional pawls 960, 961 are provided, each meshing with one set of ratchet teeth on shuttle 903, these pawls being sprung apart by a bent wire spring 962, and capable of being pulled towards each other to grip the shuttle by another SMA wire actuator 964. The fast release version of the floating pawls using two SMA wire actuators may similarly be used for both directions of fast slew-back if desired. Very Compact Actuator

Figure 10 shows a size comparison of a practical example with 100pm step-size, and a bidirectional travel of ±11mm (i.e. 22mm total stroke). An American dime is shown drawn to scale. In this embodiment, dimension W is 3.9cm, dimension X is 1 2cm, dimension Y is 1.1cm, and dimension Z is 2mm.

Adding Heatsinking To Accelerate SMA-Wire Cooling

With an SMA actuator wire positioned in air the rate of cooling after actuation ends (and with heating turned off) is determined by the thermal conductivity and thermal capacity of the surrounding air, and the proximity of other significant heat-sinks (e.g. the stator body).

Convection can play a significant part in cooling (particularly vertical) linear-actuator wires if their on-time is large (e.g. multiple seconds). However, for short transient actuation as is possible for stepper-actuators (e.g. the stepper-pawl SMA wire actuators described herein) there is very little time for convection to establish itself if the heating current is pulsed on and off quickly (on time-scales of milliseconds instead of seconds) and then conduction dominates. Radiation effects are generally less than 1%.

One effective and simple way to greatly increase the cooling rate without significantly impeding the motion of the SMA-wires, is to embed them in a heatsink-compound such as a standard silicone grease or thermally-conductive but electrically insulating grease or silicone polymer. This can reduce SMA wire cooling-times by a factor of ten, or more. The thermal conductivity of air is ~0.02W/m/K, for standard (lubricating) silicone grease is ~ 0.18 W/m/K and for these types of thermal heatsink compounds it is ~1 to >2W/m/K.

In a normal linear SMA-wire actuator this heatsinking comes at the cost of increasing the steady-state power required to keep the wire hot (contracted) also by the same factor (e.g. ten or more) which is generally highly undesirable, and so is rarely used in practice. However, in an SMA stepper-motor, the heating-up plus on-time can be kept very short, since once the SMA wire has reached an Austenite state, it will have contracted, its pawl will have irrevocably moved along the ratchet (or lifted off the ratchet), and it may be immediately turned off again, to await cooling time, and the arrival of the next 'step' command. Even with“continuous” stepping the SMA wire's 'on-time' can be just a small fraction of a cycle-time (i.e. on+off-time), since the heating can be done very quickly by suitably raising the drive voltage for a short time, the primary limit being the inertial forces then experienced by the SMA-wire due to the rapid acceleration of its load (e.g. the BSA’s external load mass and the internal moving mass of pawl(s) and shuttle). The cooling time which represents the minimum off-time will generally be much longer than this.

If active heat-sinking of the SMA-wire is used, as described above, then it is essential that the entire length of the SMA-wire is heat-sinked as uniformly as possible, since otherwise hot-spots will occur in regions of the wire where the heat-sinking is reduced, which will eventually result in wire failure. One advantage of using this form of active heat-sinking of the SMA wires is that the cooling performance of the SMA wires is then much the same whether the devices are operated in normal

atmospheric conditions or in vacuum, because they no longer rely on the air to take away the heat from the wires. This is particularly important for high-altitude or space use where the air is either thinner or totally absent than at ground level. Such active cooled SMA wires enables this technology to be effectively used for space -borne equipment, such as moving lenses and mirrors, tuning RF filters for communications, and for phase-shifters for microwave antennas.

Alternative Pawl Layout To Facilitate Heatsink-Compound Cooling Of Pawl SMA-Wires

Figure 11 shows an alternative arrangement where the stepper-pawl 1105 is on the opposite side of its adjacent passive-pawl 1104. The effect of this is that the SMA- wire actuator 1111 for stepper-pawl 1105 no longer needs to pass through the passive-pawl 1104. The pawls 1104, 1105 are now held in contact with the shuttle by forces provided by springs 1106, 1109 as before but in this implementation these are compression springs for mechanical convenience (although it should be noted that any suitable form of spring would suffice). The pair of pawls are liftable together (off the shuttle teeth 1102a) by the action of SMA-wire actuator 1120, in a similar manner to that already described above. The movement of these pawls is limited by the adjacent sections of stator/base as well as by blocks 1101 , 1103 rigidly attached to (or part of) the stator. The left/right movement of passive pawl 1104 is limited to —1/2 a tooth length, and that of drive pawl 1105 to -1.5 tooth lengths. Otherwise, operation is as described before.

However, SMA-wire actuator 1111 is now mounted in a groove 1120 in stator 1102 and this groove surrounding the SMA-wire is filled with heatsink-compound (not shown, for clarity) which tightly thermally couples actuator 1111 to the thermal compound itself, and then to stator 1102, enabling it to cool to ambient temperature very quickly and thus facilitating fast step-rates. Groove 1120 is sufficiently deep to give adequate mechanical clearance between the SMA wire 1111 and stator 1102 when the stepper pawl is in the lifted state (i.e. off the shuttle teeth). Similarly pawl- lift SMA-wire actuator 1120 also now is sited within a narrow groove 1102 in the stator (between the position reference blocks 1101) and so similarly may also be thermally coupled to the stator with heatsink compound, allowing fast cycling.

In this embodiment, the actuator is constructed so that the heatsink compound is contained in the grooves around the SMA-wires, and egress of the compound is prevented.

With a high viscosity (thermal) grease surrounding the SMA wire the forces on the wire from its contact with the grease when the wire contracts and expands will be relatively small. If instead a setting/curing elastomer (thermal) compound is used then the forces can be very much higher, especially if the clearance between the SMA wire and its static surroundings (e.g. the stator) are small (for example in the range of 200pm to 400pm). In this case the elastomer will adhere to the

surroundings (and the wire) and the effective length of elastomer being stretched by SMA wire movements is approximately the same as these small clearance sizes (e.g. 200pm to 400pm). It is nonetheless desirable to use such a setting elastomer compound as a thermal heatsink if possible, because then a permanent and known bond between SMA wire and compound is produced, and furthermore, there is no danger of leakage of the elastomer from the actuator (or just away from the wire). A convenient way to resolve this dichotomy is to prevent the elastomer bonding to the side walls of the surrounding cavity, either by choice of material, or by applying a suitable coating, or by loosely line the cavity surrounding the SMA wire with a very thin elastomer sheet of high compliance material, and then to fill the remaining space to and around the SMA wire with the elastomer compound and allow it to set/cure. Then the slug of cured elastomer wrapped on one or more sides by the thin sheet, is able to contract and expand longitudinally (parallel to the wire contraction direction) without being impeded by the local cavity wall (e.g. parts of the stator) in which it is contact, but without any adhesion.

Alternative Electrical Drive To Protect SMA Wires From Stuck-On Drive Signals

As stated above, a simple way to decrease the SMA wire turn-on time (time from deactivated to activated, or from sub-Austenite-start temperature to post Austenite- finish temperature) is to drive it with an initially high voltage to increase the initial rate of heating and then to lower the voltage to a level enough to merely maintain the temperature, or indeed with this stepper actuator, to then simply turn off the drive voltage.

However, if the microprocessor (or other pulse source) driving the SMA element should“freeze” /“stick-high” / crash or suffer a software error that can allow the drive line to hold in the high state for much longer than the intended short pulse drive time, then the SMA element can overheat, and in extreme cases suffer irreversible thermal and possibly mechanical damage, rendering the actuator useless. This situation may be avoided by driving the SMA element via a capacitor (ac drive) rather than with direct dc drive.

In the variation/embodiment described below, one or more of the SMA active elements of the SMA-stepper-actuator is electrically driven by an electrical pulse via a series capacitor, Cs, whose capacitance value is chosen such that the energy passed through the capacitor to the SMA element when the pulse is applied is sufficient to heat the SMA element through its Martensite/Austenite transition.

Direct application of this series-capacitor technique, for typical SMA element resistances in the ~10ohm range, can require large values of capacitors, unless the drive voltage is high (say >5V) which itself is inconvenient in digital systems where supply voltages of 5V, and nowadays, 3.3V, is much more common. For example, for a ~120hm SMA element to run off a 3.3V supply, this requires approximately a 250pF series capacitor if it is to be driven to Austenite temperature by a single pulse in a few milliseconds. Such capacitors are either bulky, or expensive, or both.

This problem can be solved in the following two ways:

1) by driving the SMA element via a smaller series capacitor with a burst of closely spaced pulses, each pulse of shorter duration than the single pulse drive-method requires. In general, for a number Np of pulses in the drive burst the series capacitor can be very approximately Np times smaller in value than for a single pulse, with the same supply voltage. Using such a method allows the use of a very small 10pF capacitor.

2) Instead of driving the SMA wire directly from a microprocessor output port, it is instead driven via a small transistor (e.g. bipolar or field-effect /FET) which having significant gain, will have a much higher input impedance Zin than the SMA wire resistance Rs. If now the capacitor is inserted between the microprocessor output line and the transistor input, and the SMA wire is driven directly by the transistor, then the capacitor may now have a value ~Zin/Rs times smaller than previously, for the same time-constant. Especially with an FET, where an input resistance of at least, for example, 1 Mohm may be easily achieved with e.g. a 1Mohm gate- source resistor, this allows the use of a very small and low-cost capacitor (in the example above instead of a 250pF series capacitor, a value of 250x12/10 ⁶ = 3nF may be used, typically a tiny <1cent ceramic device). Given that a transistor drive for the SMA wire will frequently be needed in any case due to output-current limitations of many microprocessor output ports, this technique can be

implemented with very little additional cost in practice.

Direct application of the burst technique or 1) above, can put inconveniently high processing and timing loads on a control processor (as it now has to determine and control not only the overall burst duration or single pulse length, but also carefully time each individual pulse of the burst to make optimum use of the smaller series capacitor). For pulses around 1 ms or less this is problematic for most processors, especially where many SMA elements have to be controlled independently at the same time.

It should be noted that the pulse-mode of driving the SMA element is relevant to stepper actuators. In non-stepper actuators it is usually the case that wire temperature has to be maintained continuously for as long as actuator output position is to be maintained, in which case short-pulse drive is inappropriate, and the use of high thermal-conductivity material surrounding the wire (to increase speed of de actuation) with continuous drive current, causes unacceptably high quasi-continuous power loss.

A performance example is shown in Fig 14. In this example the SMA element is a 25pm SMA wire (not shown) with a roughly 130hm cold resistance surrounded by medium-thermal-conductivity (0.2W/m.K) silicone rubber. The waveforms show the effect of driving the SMA element from a 5V pulse burst source of 50 pulses of duration100ps each and 200ps apart, via a 10pF capacitor.

In Fig.14 the top chart shows the pulse burst drive voltage waveform, with 'time elapsed' shown in the x-axis in ms. Below that is the voltage across the 10pf capacitor. The third chart in the figure is the wire current which will be seen to quickly become -symmetrical about zero. The bottom chart in the figure shows the wire temperature initially rising steeply, then at a slower rate as the Martensite/Austenite transition occurs, and after ~8ms reaching the full Austenite temperature and shortly thereafter beginning to cool quite rapidly (-40C within about 30ms of start of heating in this example) due to the silicone rubber surrounding the wire, ready for another actuation cycle.

Single Bi-Directional Stepper-Pawl Implementation

In the embodiments of linear stepper-actuators described above, a directional ratchet and pawl combination is used, one for each direction of travel.

In the variation/embodiment described below, and as shown in figure 12, the ratchet is replaced by a rack-and-pinion rack, which is inherently symmetrical and

bidirectional.

The shuttle 1203 has on one edge one set of teeth 1203a which are symmetric and bidirectional so that an engaged pawl with meshing teeth will block relative motion in both directions, in contrast with the double-edged shuttle described in the previous embodiments.

In this variation, the teeth on both pawl and shuttle preferably have a primary tooth slope of angle alpha, where this angle is chosen to be steep enough that an engaged pawl blocks shuttle motion with a small pawl-shuttle contact force, but not so steep that meshing of teeth is difficult. In this embodiment an angle of ~83degrees is used, which gives a roughly 10:1 ratio of pawl-shuttle force to shuttle blocking-force assuming moderate inter-tooth friction.

The shuttle teeth can have vertical sides, but these are difficult to mesh, so the arrangement described above is preferred.

Other tooth angles are of course useable and will be easily calculated by those proficient in the art. One or both of the pawl teeth and the shuttle teeth preferably have rounded tips, as shown in figure 12. With rounded teeth, on an attempted meshing of pawl to shuttle, the teeth slide easily together even when there is misalignment. The entire tooth profile may be curved.

In this embodiment, there are two pawls, each independently liftable off the shuttle by their own SMA wire actuators. The stepper-pawl has a pair of SMA wire actuators for stepping the shuttle (one for each direction). The other pawl is a passive pawl which lifts on and off the shuttle to lock it in place. Both pawls have springs which push the pawls into mesh with the shuttle (but which can be overcome by the SMA-wire actuators that lift the pawls). However, these springs are omitted in the drawings for clarity. The passive pawl is free to move towards and away from the shuttle but has essentially no movement parallel to the shuttle’s travel direction, being limited in this direction by adjacent walls of stator acting as end-stops. The active pawl is similar except that it may move in the latter direction by more than one tooth length (and less than two, and preferably less than 1.5, or even less still). The passive pawl is much as previously described apart from its tooth profile and is not shown in Figure 12.

In use, a step-right cycle comprises the following steps:

1. Both pawls engaged with shuttle, all SMA-wires de-actuated; [0]

2. Lift stepper-pawl by actuating the stepper-pawl’s lift SMA-wire actuator); [t _on ]

3. Move stepper-pawl left; (i.e. actuate the SMA wire actuator that pulls pawl to left); [t _on ]

4. Drop stepper-pawl (i.e. de-actuate stepper-pawl’s lift SMA-wire actuator); [t _off ]

5. De-actuate step-left SMA wire and simultaneously lift passive-pawl (i.e.

actuate passive-pawl’s lift SMA-wire actuator); [max(t _off , t _on )]

6. Move stepper-pawl right; (i.e. actuate the SMA wire actuator that pulls pawl to right); [t _on ]

7. Drop passive-pawl (i.e. de-actuate passive-pawl’s lift SMA-wire actuator); [t _0ff ]

8. De-actuate step-right SMA wire; [t _off ]

The cycle time is t _cyc = t _on +t _on ⁺ t _off +max(t _off , t _on ) +t _on ⁺ t _off +t _off = 3t _0n +4t _0ff

The power-stroke (i.e. when the shuttle is moved to the right) is step 6, where the stepper-pawl is meshed with the shuttle and pulled to the right by one of its SMA wires. This mechanism works whether the load is passive, or is actively pulling in either direction, as long as the load’s pull is small enough to be overcome by the SMA-wire actuator’s pull.

For example, if the load is pulling to the right for the duration of this cycle then no shuttle movement will occur until step 5, due to the fact that prior to that the shuttle is locked by the passive pawl. Thereafter, the load might pull the shuttle (and the meshed (dropped) stepper pawl) to the right before step 6 has started or completed, but either way it cannot move much further than a single step because of the meshed stepper pawl hitting its end-stops. This has no consequence for the stepper SMA- wire(s) as they can be pulled/pushed either way when in the de-actuated state. To always maintain the SMA wires in tension, then an additional action is to simultaneously in step 5 partially actuate the SMA wire actuator that pulls the pawl to the right by enough to shorten it under zero load. This will prevent this wire from buckling if the load over-drives the actuator.

If on the other hand the load is pulling the shuttle to the left throughout the cycle then this simply puts additional load on the stepper- pawl’s power stroke.

In use, a step-left cycle comprises the following steps:

1. Both pawls engaged with shuttle, all SMA-wires de-actuated;

2. Lift stepper-pawl (i.e. actuate stepper-pawl’s lift SMA-wire actuator);

3. Move stepper-pawl right; (i.e. actuate the SMA wire actuator that pulls pawl to right);

4. Drop stepper-pawl (i.e. de-actuate stepper-pawl’s lift SMA-wire actuator);

5. De-actuate step-right SMA wire and simultaneously lift passive-pawl (i.e. actuate passive-pawl’s lift SMA-wire actuator);

6. Move stepper-pawl left; (i.e. actuate the SMA wire actuator that pulls pawl to left);

7. Drop passive-pawl ((i.e. de-actuate passive-pawl’s lift SMA-wire actuator);

8. De-actuate step-left SMA wire;

As each pawl is now capable of applying significant blocking force to the shuttle in either direction, this arrangement provides full bi-directionality as well as guaranteed one-tooth step under external loads, all with a simplified structure and only two pawls and four SMA-wires. The electronic controller is also simpler than for the previously described designs. However, it cannot step as quickly, all things being equal.

The embodiment as shown in figure 12 has additional features to allow heatsink- compound cooling of the SMA wires to be used without applying potentially damaging forces to the SMA wires.

These two stepper SMA wires 1210, 1211 are connected to a cage 1213 surrounding pawl 1212 on 4 sides. Cage 1213 can, if made of suitable conductive material, e.g. phosphor bronze, brass or stainless steel, serve as a common electrical termination for these SMA wire ends.

Cage 1213 is just wide enough to allow pawl 1212 to slide freely vertically (to and away from the shuttle) but eliminates any significant lateral motion of the pawl within the cage. The cage 1213 in turn may slide freely laterally (parallel to the direction of motion of the shuttle) but is prevented from moving vertically by the pair of extended wings 1216, 1217 which slide in grooves in stator 1202. The effect of this

arrangement is that when SMA lift wire 1214 is actuated and lifts pawl 1212, the pawl slides within cage 1213 away from the shuttle, but crucially does not pull the stepper SMA-wires 1210, 121 1 with it since these are attached to cage 1213, not pawl 1212. When there is thermal-compound present around the stepper wires (1210, 121 1), any such vertical movement could apply shear forces to these wires. The cage 1213 decouples these wires from such movements and protects them.

Similarly, SMA wire 1214 - the pawl lift wire - pulls directly on rod 1219 which passes freely through a hole in pawl 1212, and whose length is such that the ends of the rod just clear the walls of stator 1202 at either end, but are close enough to prevent any significant sideways movement of the rod (and thus of the SMA wire 1214). When the pawl 1212 is pulled by either of the stepper wires 1210, 121 1 pulling on the cage, the pawl slides along rod 1219 without transmitting its lateral motion to SMA wire 1214, which again might otherwise risk shearing the wire in the presence of surrounding thermal-compound. These protection measures are particularly important when a semi-setting or fully-setting silicone thermal compound is used even if this has high compliance.

A different view of the embodiment shown in figure 12 and described above is shown in figure 13. In the view of figure 13 the stepper-cage 1213 and shuttle 1203 are removed (in the figure, but not in the operational device) to show more clearly some of the internal arrangements, in particular the two (stepper-pawl) stepper SMA-wire actuators 1210, 121 1 can be seen to be deeply set into two grooves 1221 , 1220 in stator 1202. Rod 1219 coupling the lift wire to pawl 1212 can be seen to be closely fitting (with clearance) between the adjacent walls of stator 1202, to prevent substantial movement in the direction parallel to the shuttle movement.

The passive pawl for this version of the actuator is essentially the same as for the previously described actuators except for the modified tooth shape (similar to that of pawl 1212 in figures 12, 13). This pawl may be fitted either side of the stepper-pawl, preferably far enough along the shuttle not to interfere with the stepper pawl’s SMA stepper SMA wires. The shuttle 1203 now needs teeth on only one edge, as only two pawls in total are needed and they both interface with the same set of teeth on the shuttle. Alternatively, the lock-pawl may be placed on the opposite side of the shuttle to engage with a second set of rack teeth on the opposite edge of the shuttle from those shown in fig.12.

A fast return mechanism (or fast slew-back) as previously described can be added to this version of the stepping linear actuator as before, except that one of the moving pawls (the one on the edge of the shuttle with no teeth) is replaced by a smooth slider. Alternatively, and preferably, teeth as shown can be added to both edges of the shuttle, and then both fast-slew-back pawls can mesh with the shuttle teeth and apply bidirectional locking forces. In this case the long- tension spring can attach to both pawls, as neither will now slip when pulled by the spring.

For all of the variations of stepper actuators described above that use a thermal- compound to enhance cooling time and thus increase the stepping speed, it is not always necessary to add one or more additional parallel stepper pawls (to be actuated sequentially and overlapping as described previously) to speed up the actuator, since the fast cooling times are usually adequate to produce the desired output shuttle speed.

An additional parallel stepper pawl assembly may still be added, as this allows the electronic controller to leave the second stepper-pawl assembly de-actuated in normal use, and to monitor the currents in the SMA-wires of the first stepper-pawl assembly. Should the electronic controller detect an open circuit (suggesting a broken wire) or a short circuit (suggesting some other kind of fault) then it can seamlessly cease driving the first pawl’s SMA wires and instead switch over to the second stepper assembly and use that instead, and carry on functioning without a pause. The controller can also then report back to a higher-level controller that a possible fault has been detected, so that the stepper motor can be repaired at the next service opportunity. In some implementations the same measure of

redundancy/enhanced reliability may be achieved by adding a second SMA-wire in parallel to the existing SMA-wire on the stepper-pawl, with at least one end of each of the two wires electrically insulated from each other. In this case an additional pawl is not required - instead sensing a break in the current drive wire will cause the controller to switch over to driving the second wire instead.

Where the external load connected to the shuttle in any of the above variants of the BSA is uncontrolled, and capable of exerting damaging forces on the internal mechanism of the BSA via the shuttle, some form of overload protection is desirable. This is achieved in all the embodiment by appropriate selection of the angles of the teeth on the shuttle (and the matching meshing teeth of the pawls). For example, where one or both edges of a tooth are vertical, or 90degrees (normal to shuttle travel direction) then external forces on the shuttle can be transmitted directly and at full magnitude to each and any pawl that is engaged, and so may be capable of causing damage to the teeth or surrounding structures.

If the maximum slope of either edge of a tooth is limited to A (0 < A <90 deg) then the force Fp transferred to a shuttle tooth by a pawl tooth held against the shuttle by a normal force Fs (from a pawl spring for example) can be shown to be

Fp = Fs(mu+ tan A)

where mu is the friction coefficient between the mating teeth. Thus if the shuttle tooth exerts a force greater than or equal to Fpmax on the pawl tooth, where

Fpmax / (mu+ tan A) = Fs

then the pawl tooth will begin to lift off the shuttle tooth, against the force of its pawl spring (e.g.) and no additional force above Fpmax will be experienced by the pawl tooth, thus providing an overload limit capability. Typical values of tan A to be used can be in the range of 5 to 10, whereas mu is limited to a maximum value of one.

Since mu is a poorly controlled variable the overload safety factor can be designed in almost entirely by a suitable value for A since for A >45 degrees the tan A factor will dominate. That is for example if A ~83degrees then tan A ~8, which is very much greater than an average friction coefficient of 0.5. Thus in all of the BAS variants above this form of overload protection may be added by ensuring that the steepest part of any tooth profile (shuttle or pawl) is restricted to a maximum slope angle A < 90 degree chosen as appropriate (in conjunction with the pawl restoring force Fs) for the degree of protection required.

Conversion To Rotary Stepper Motor Any of the variants of the BSA described above may be converted easily from a linear stepper motor to a rotary stepper motor by engaging a pinion gear with the teeth on one edge of the shuttle. Where the teeth shape are ratchet teeth (asymmetrical) then it is preferable to engage a pair of pinions, one to each set of teeth on either shuttle edge, these two pinions then linked together by a further pair of passive pinions inter-meshing with each other and one to each of the first pinion pair, to guarantee non-slipping drive in both directions.

Reduction Of Friction

In the SMA stepper actuator designs of the present invention described above, various components slide over various others and are therefore subject to friction forces, e.g. the shuttle slides within the body, the pawls slide within the body and/or within a cage. Friction forces in sliding components are difficult to control and are subject to dependence on environmental conditions (particularly temperature and humidity) and to accumulation of dirt and dust. Any such changes in operational friction can modify actuator behaviour and are best avoided where possible. Many of the components in the actuators described also have springs attached to them to give a default return position when not otherwise driven, e.g. many of the pawls described have springs to engage or disengage then from the shuttle teeth, and in several instances the shuttle has a return spring to cause it to return to a given position when not otherwise driven by the pawls.

T o address this, one or more of the sliding components of any of the stepper actuators described herein can be replaced with a similar component that instead of sliding on or against or within the body of the actuator or within another component thereof, is instead supported by one or more flexures capable of supporting it and guiding it in its direction of travel without direct supportive contact from the other surrounding components. For sliding components that also require springs to return them to a reference position, it is also possible to combine the support function of the one or more flexures with a spring function. In practice multiple parallel leaf flexures are useful for all spring functions in the stepper actuators described herein, in that they may have slender flexures allowing large deflections without plastic deformation of the flexure material and yet substantial return force because of the additive force effect of the multiple parallel leaves.

For example, a pawl may be supported by a single leaf spring which is pre-stressed such as to push the pawl into contact with the shuttle (or away from it against an end- stop) with a certain required force, in such a way as there is no friction produced when the pawl moves. Similarly, for short-stroke actuators, the shuttle may be supported within/adjacent to the body of the actuator by one or more flexures, but more typically two or more flexures, which also provide a returning force, holding the shuttle at (typically) one end of its range of travel against an end-stop (although the reference point might in some cases be an intermediate position which would then be less precisely defined). During actuation via the stepper and lock pawls the shuttle would then be coerced to move away from this reference position against the spring force of the flexure(s) but substantially without friction. Such elimination of friction not only improves the reliability of this type of actuator, but also substantially eliminates wear, and the associated creation of particulates. Where a moving component supported by one or more flexures is also the mechanical terminating point for one or more SMA actuation wires, then if one or more of the flexures is made of conductive material (e.g. metal) or carries a conductive track along its body from end to end, then each such flexure can be used as an electrical connection to that end of the one or more SMA wires, eliminating the need for“flying-lead” connections to moving components. Where the flexures are made of suitable material, especially metals and more especially phosphor bronze or stainless steel, then a part of the flexure material may be formed into a crimp structure to allow direct mechanical and electrical connection to the SMA wire and the moving component, greatly simplifying the overall structure of the actuator. The fixed end of the flexure may then be mounted to the body of the actuator directly, or if the body is electrically conductive and the flexure connection needs to be electrically isolated from the body then instead it may be mechanically connected to the body via an insulated component, typically polymer and the electrical connection made at this fixed end by any convenient means. One convenient means is for the body of the actuator to be a printed circuit board (PCB) in which case at least part of the fixed end (root) of each flexure may be mechanically and electrically attached to conductive pads on the PCB that connect it appropriately to the rest of the circuit, providing electrical and some mechanical support. The majority of mechanical support may be provided directly between the root of the flexure and the PCB itself. Such techniques can eliminate the need for any discrete wiring (other than the SMA wire itself, which is a mechanical-electrical component).

Reduction Of Sensitivity To External G-Forces

The structure of the stepper actuators described herein makes them highly resistant to the effects of g-forces directed at right-angles to the direction of shuttle motion (lateral g-forces). As long as the gross structure of the actuator is not damaged by them then even very high lateral g-forces have essentially zero effect on the actuator position because the pawls and their springs may be made to have very little mass, and so inertial forces induced by stator accelerations will be small compared to the related spring forces holding those components in place. However, external g-forces parallel to the direction of shuttle motion (longitudinal g-forces) can seriously effect the positional stability of the stepper-actuator because such forces magnify the apparent load mass.

For example, if the load mass is 1gm and the actuator has a load rating of 2gm, then the actuator can accurately move and hold the load in a stable position. If the body of the actuator is then subject to e.g. ~3g acceleration (i.e. ~30m/s ² ), then the effective mass of the load becomes anywhere between 2gm and 4gm (depending on the orientation of the actuator in the 1g Earth’s gravitational field) and the actuator is now“out of spec”. So the load (relative to the actuator) is likely to move and positional accuracy is lost.

This effect is particularly severe if the external g-forces are applied while the actuator is stepping because in some of the stepper actuator designs the lock-pawl(s) lift off during stepping, reducing the hold on the load by the actuator body.

One way to almost completely eliminate this effect and to make the actuator nearly immune to external shock and vibration is to balance the actuator, so that at any time there is motion of the shuttle (and load) relative to the body/stator, there is simultaneously motion in the opposite direction with the same or very similar momentum mechanically locked to the shuttle/load motion.

This is achieved by mounting a second shuttle parallel to the first, capable of freely sliding parallel to the first, and caused to move in the opposite direction and by the same amount as the first by some sort of mechanical linkage. Such a linkage could be for example a swing arm pivoted at a point midway between the shuttles and rotatably connected to each at either side of the pivot (e.g. via a pin protruding from each of the shuttles at the same distance from the pivot point, passing through slotted holes in the arm) to act like a see saw. In this way as one shuttle moves, the other shuttle is forced to move equally in the opposite direction.

Another mechanism would be to use inextensible cords or cables passing around pulleys beyond each end of the shuttles, the pulley diameters being the same as the separation of the shuttle centres, and the ends of the cables being connected to the ends of the shuttles. So when one shuttle moves one of the cords/cables would be placed in tension and would pull around its pulley onto the closest end of the other shuttle causing it to move the same amount in the opposite direction.

One type of dual-cog, balanced shuttle arrangement is to place a cog between a set of identical rack teeth on each facing edge of the two shuttles and engaging with those teeth. The cog rotates about a central bearing fixed to the actuator body midway between the two shuttles. Movement of one shuttle causes the meshed cog to rotate and it in turn then causes the other shuttle to move the same distance in the opposite direction to the first. It is most preferred that one of a pair of such cogs is placed near to both ends of the shuttles (but always remaining meshed with the rack teeth over the entire range of allowable shuttle movement), to better balance the forces on the shuttles and prevent jamming. With this arrangement, linear motion of the shuttle is simultaneously converted to rotary motion of the cog(s). It then becomes possible to move the position of any fast-slew-back springs from the shuttle to the cog(s), and to replace linear tension or compression springs attached between the stator and shuttle with spiral clock-springs attached between the stator and non central portion of the cog(s), which may provide a more compact arrangement. That is, the fast slew-back springs for each direction can be fitted one to each of two such cogs to provide a more compact arrangement.

In each of these three cases there is an identical magnitude of movement (in opposite directions) of the two shuttles. If the second shuttle, plus any attached load, is arranged to have the same mass as the sum of the first shuttle mass plus the mass of its external load, then the momentum of each will be the same when they move. Such a balanced actuator will suffer no net driving force along the direction of shuttle motion when its body is subjected to axial g-forces.

Figure 15 shows schematically an implementation of a dual-cog, balanced shuttle arrangement. For the purposes of explanation and clarity the remainder of the SMA stepper actuator (which could be any of the variations described herein), is not shown in Fig.15, where 1503a is a shuttle of an SMA stepper actuator as already described, bearing a set of teeth along one long edge. Cogs 1510, 1511 , and a balancing rack 1503b (the mass of which is arranged to be identical to the mass of the shuttle 1501a and any load connected thereto) are also shown.

The cogs have stub shafts 1505, 1506 which bear in the stator (not shown) and upon which they rotate freely. The intermeshing of the cogs with the shuttle teeth and balancing rack teeth cause the balancing rack to move an equal and opposite distance to the shuttle, whenever the shuttle moves along the direction of its gear teeth. Thus if an external force causes the stator (not shown) to accelerate (i.e. an external g-force), then this will cause the cog shafts to accelerate similarly and the balancing inertial forces of the shuttle and rack on opposite sides of the cogs will cancel out producing no net rotation of the cogs and thus no movement of the shuttle relative to stator. This arrangement thus renders the actuator virtually immune to external g- forces.

It is possible to arrange for the same balanced condition in other ways. For example by arranging the second shuttle’s own mass to be the same as that of the first shuttle+load instead of adding a separate load mass to the second shuttle,. This allows it to be made more compact. If“gearing” is used then the balancing mass need not be the same as that of the first shuttle+load. For example if a 4:1 gear ratio is used by interspersing an odd number of cogs/gears in series, with a 4:1 overall ratio between the shuttles, then the 2nd shuttle mass need be only one quarter that of the first shuttle+load, and can be slightly less still because of the inertial effects of the gears.

For balance it is only necessary to maintain the momentum balance as stated above.

Escapements

An escapement is a mechanism well known in the art, that controls the movement of a carriage or wheel by a small fixed amount, when the carriage or wheel is being urged to move by some external force. Rotary forms are found in clocks and watches. Linear forms are found on typewriter carriages.

An escapement may be used in another variant of the SMA stepper actuator, instead of the arrangement of one or more pawls described above. The advantage of using an escapement over multiple pawls is that movement of the same single mechanical component (the verge) acts to position the shuttle at the end of a step, as well as to guarantee that only a single step occurs, under all within-specification load conditions. If the verge is spring-loaded in one direction of its movement, and SMA- wire driven in the other, then a step cycle may be controlled in exactly one SMA heat/cool cycle, allowing for relatively fast stepping compared with more complex multi-pawl-movement mechanisms.

An inverse escapement mechanism is also well known in the art. This type of mechanism not only controls the movement of a carriage or wheel by a small fixed amount as per an escapement, but in addition drives the motion as well. That is, there need not be any additional driving forces on the moving member.

When such a mechanism is used as part of the linear actuators of the present invention, the moving member is the shuttle. The verge of the inverse escapement is arranged to mesh with the shuttle in such a way as to always limit the motion of the shuttle to a single step at a time, under the control of the position of the verge, and to also push or pull the shuttle along its direction of travel by means of the meshed contact points, by precisely one step during one cycle of movement of the inverse escapement mechanism.

An example of this mechanism suitable for use with the embodiments of linear actuators of the present invention is shown in figure 16. The basic mechanism of rotary inverse escapement is already known in the art, but the example shown is modified and improved for the purposes of the present invention. The rotary inverse escapement of figure 16 is shown in two of many possible positions.

The verge 1611 pivots about a pin 1612 fixed to a stator 1615, and is able to rock back and forth about that axis, and in so doing a pair of teeth 1617 and 1618 distal from the axis interface with the teeth of a ratchet wheel 1610. The verge 1611 can be rotated anticlockwise about the pin 1612 by a first SMA actuator 1614 connected between a pin 1613 which freely rotates in the verge, and the stator 1615. When actuator 1614 is not pulling anticlockwise, the verge is stably held in the position shown in the right-hand view of figure 16, by spring 1616 connected between pin 1619 which freely rotates in the verge, and the stator 1615. This ensures that the verge tooth 1617 engages with a pair of teeth of the ratchet wheel 1610. This blocks rotation of the wheel. Alternatively a second SMA actuator can be used in place of the spring. When instead the SMA actuator 1614 contracts and pulls the verge anticlockwise, the verge moves through the position shown in the left-hand view of figure 16, where its tooth 1618 can be seen to interact with the teeth of the ratchet wheel 1610, as shown in the left-hand view.

The arrangement of the verge teeth and its rocking motion driven by successive cycles of activation and deactivation of SMA actuator 1614 cause the ratchet wheel to rotate clockwise by precisely one tooth width on each cycle, and because at all times at least one of the verge teeth 1617, 1618 are engaged with the ratchet wheel teeth, the rotation of the wheel is completely under the control of the verge and actuator/spring combination.

Although as outlined above a linear inverse escapement version of this type of mechanism may be used to drive the SMA stepper actuator which is the subject of the present invention, the rotary inverse escapement as described above can also be combined with parts of the balanced shuttle mechanism shown in figure 15, to drive a linear SMA stepper actuator. For example, the mechanism as shown in figure 16 and as described above can be mounted such that its ratchet wheel 1610 is coaxial to and/or mounted to the same shaft as one of the cogs (e.g. cog 1510 of the embodiment shown in figure 15), such that a cycle of the escapement turns the ratchet wheel and thus the cog clockwise by one tooth width. This will then cause the shuttle 1503a to advance to the right by one tooth width. No other stepper mechanism would be required to move the shuttle 1503a.

If in addition an anticlockwise (mirror image) of the mechanism of figure 16 were added to the second cog 1511 of Fig.15, then a bidirectional stepper mechanism becomes available, providing that also an additional facility is included to allow one of the escapements to disengage from the linear stepper when the other is engaged and stepping (since both escapements block rotation in both directions when engaged and not stepping). Such a dual inverse escapement driven linear stepper provides a fast mechanism, as a complete step cycle involves only a single SMA element contraction/expansion (heating/cooling) cycle-time.

Typically the verge contains at least two separate drive“teeth” mechanically connected together, and the arrangement of the verge teeth and their directions of motion are such that at least one of the two verge teeth is always engaged with teeth on the shuttle to be moved. Furthermore, when one of the verge teeth is engaged with the shuttle and causes the shuttle to move, then before that tooth disengages from the shuttle, the verge’s other tooth will have already engaged with the“next” tooth-pair along the shuttle from the tooth-pair it was previously engaged with, prior to the start of the verge movement cycle. So as can be seen from the mechanism described above, an inverse escapement mechanism driven in at least one direction of its cyclic movement by an SMA-wire actuator engages with the teeth of the shuttle in such a way as to cause the shuttle to move in the desired direction by just one step while preventing any further movement independently of the nature of the within- specification forces applied externally to the load of the stepper- actuator. The return motion of the verge can be provided by a spring, or by a second SMA-wire actuator.

Other Combinations Of Features

Several distinct ratchet, rack, pawl, symmetric and asymmetric tooth-slope, pawl spring, fast- return-tension spring, linear to rotary conversion, inverse escapement driven, g-force resistant balanced shuttle, pulse driven, and damping/buffering arrangements have been described, and some combinations of these described for clarity of explanation

Those versed in the art will realise once they have read the above that many other combinations of some or all of the features described are possible and useful in certain circumstances. All such combinations should be considered to be included herein as part of the invention, and the invention should not be considered to be limited only to the specific combinations of features described in the embodiment above.

For example, if what is required is a unidirectional linear stepper motor that steps as fast as possible and rapidly returns to zero at the end of its stroke, then the simple two-pawl ratchet scheme on just one side of the shuttle (i.e. only one set of shuttle ratchet teeth, with tooth slope chosen to give the required resistance), a fixed end- stop on the shuttle, and a long tension spring for the return stroke that is always engaged with the shuttle (i.e. dispensing with the travelling pawls on the shuttle, and instead mechanically linking the tension spring to the shuttle) can be used.

Such a device has minimal hardware, is very simple to drive electronically, and can step very rapidly, especially if the SMA wires are embedded in a thermal heat- conducting compound in the manner described herein.

Fundamental SMA-Stepper-Actuator

One major difference between known SMA actuators and the actuator of the present invention is that the SMA element in the present invention is always fully actuated to produce an output movement, and the SMA element then moves the output load only a fraction or 'step' of the actuator's full stroke upon full SMA element actuation. It is the mechanical mechanism of the actuator that integrates these steps into the total output-stroke of the actuator. This is in contrast to known mechanisms, where the SMA element is partially actuated to produce a partial stroke of actuation. In the present invention, typically the steps are all of equal length d, and where there are Ns of them in total, then the stroke of the actuator S is given by S = Ns . d

Crawler Actuator

In this aspect of the invention, the shuttle of the actuator is fixed, and the body of the actuator moves relative to the shuttle.

Any of the actuators described so far may be converted into a crawler actuator, by reversing the physical arrangement of the pawls and shuttle teeth, such that the shuttle is now on the outside of the actuator with the shuttle teeth facing inwards, and the teeth of the pawls press outwards and onto the inside teeth of the shuttle when engaged with it.

If the external shuttle is fixed, it will be seen that the actuator body will move along the line of shuttle teeth as it is actuated, so the actuator moves itself and not an external load. This is a crawler actuator.

In all other respects this can be identical to any of the nominally linear actuators described above. However, in this variation, there is no reason why the line of shuttle teeth needs to be geometrically straight, as long as its radius of curvature is large compared to the length of the engagement path with the one or more pawls of the actuator body. The shuttle itself may optionally no longer be regarded as a proper part of the actuator - rather, it may be regarded as an external body with which the actuator pawl(s) engages.

In a variation of these crawler actuators the meshable teeth on the pawl and shuttle are replaced with a friction grip between the two parts. Given adequate normal contact force, surface roughness, and a high enough coefficient of friction between the two parts, sufficient tangential force between the contact regions may be generated to allow the crawler to move itself along the shuttle (i.e. external body). Precise one-to-one registration and precision-positioning afforded by a toothed ratchet and pawl will be lost in this friction-interfaced/driven version of the actuator. However, this variation still provides useful if less precise movement.

Another possible variation is that the interfacing surface of the shuttle external body of such a crawler actuator is located inside a tube or pipe (either synthetic, or a component of an animal/human body). In this case, placing two or more step-pawls opposite each other (e.g. disposed at roughly 180deg, 120deg, etc around a circle inside the pipe) allows the contact force between the pawls and shuttle to be generated independently of gravity by squeezing the pawls roughly radially outwards into the inside of the pipe at (opposite) points around a circular section of the pipe, thus allowing the pawls to grip the inside of the pipe (by pushing outwards against each other) in order to provide the axial force needed for movement.

Such a device would be capable of insertion into a bodily aperture (such as a nostril) and then allowed to move itself along the passage (e.g. nostril-air-passage - trachea - bronchial etc into the lungs), and given adequate motive actuator force could pull a cable, catheter or delivery- pipe, waveguide or other flexible device along behind it. Signals on the cable could be used to power and control the actuator. The catheter could deliver drugs to a target reachable via the pipe. The delivery-pipe could deliver friction reducing fluid to the crawler head. The waveguide could transmit radiation to such a target. The crawler could be fitted with a miniature camera so that a surgeon could see what was ahead of the crawler and to control treatment processes and targets. Then the delivery-pipe could deliver fluid to clean the camera lens. If more than one set of drive pawls were fitted to the actuator, and/or were the pawls to be individually controllable, then the lateral direction of the crawler could be controlled - i.e. it could be steered or navigated along the tube it was inside - and for example determine which branch to follow where the tube branched.

Rotary SMA Stepping Actuator In a variation, the shuttle can be curved or even circular in a plane through the set of teeth, but of such large radius that on the scale of a tooth size it is nonetheless essentially straight, thus forming a large diameter gear wheel (possibly of contrate gear form as well as pinion form), with number of teeth N. Then an actuator could operate as described on this circular shuttle and upon each step advance the gear wheel by one tooth and thus rotate it through 2pi/N radians. However, it is now able to continue turning the gear indefinitely without reaching any end stop. Such a device can be referred to as a Rotary SMA Stepping Actuator. In practice the limitation on radius above can be removed, by using pawls with either one tooth, or, multiple teeth shaped and positioned along an arc to appropriately intermesh with multiple teeth of even a small diameter gear, whilst simultaneously modifying the motion of the drive pawl when actuated by its SMA drive wire to first begin to engage its teeth with the gear, and then to provide a circumferential force to the gear teeth so as to turn the gear sufficiently to step its position. Operation is otherwise similar to the linear SMA stepper actuators described.

Blocked-Motor

In a scenario where the shuttle is mechanically blocked (e.g. by external load overload) when a step-cycle is executed, then the step-cycle will necessarily fail to complete correctly, and the shuttle may not move at all.

For applications where it is important to know where the shuttle is, then a blocked step-cycle event is problematic. To overcome the attendant loss of shuttle position, the following two options are available: 1) Detect the blocked cycle (and so account accordingly for the current position; 2) include a return-to-reference-position facility (a return-to-zero, or RTZ, although it should be noted that zero does not need to be the actual reference position), so that when required, e.g. periodically or when some sort of fault condition is detected, the shuttle may be returned to some known reference position, independent of its actual current position.

Blocked cycle detection is best done by sensing the motion (or lack of it) of the shuttle, for example by adding a step-sense mechanism that gives a distinctive yes/no signal when a step is successfully or unsuccessfully (respectively) completed, then the actuator control system can take account of the blocking, in its position counting (and can also be used to report error conditions if more blocks occur in a given stepping sequence than is deemed permissible). The teeth on the shuttle provide a convenient mechanism for this to be achieved. When the shuttle successfully completes a step, a whole tooth will pass by any point adjacent to the shuttle. So, for example, if a metal leaf spring of suitable length projects radially into the gaps between adjacent teeth, then during a step-cycle such a spring will be deflected by the motion of a tooth going by, and will then flick back as it rides over the tooth and into the next gap. This will produce a particular characteristic motion of the leaf (slow deflection one way and then very rapid reverse motion) which can be detected for example by a simple capacitive sensor with the leaf spring forming one element of the capacitor, and transduced into an electrical signal, and then signalled to a processor capable of distinguishing slow from fast motion signals. A typically blocked cycle will produce little or no motion and so little or no signal. A half- blocked and then reversed step will typically produce slow motion in both directions and yet a different characteristic signal.

RTZ may be simply implemented by mounting one or more“limit- switches” which are actuated by the motion of the shuttle preferably when it reaches one end or the other (or both) of the allowable shuttle travel. The operation of the switches may then be detected electrically and e.g. sensed by a microprocessor GPIO pin, used to halt further motion in that direction, and to reset the position counter to the now known position of the shuttle. Once one or more RTZ limit-switches (which can be galvanic switches, magnetic or optical sensors, resistive pressure sensors, or any other suitable form of sensor) are fitted to the shuttle then these can be used to reset the position counter whenever they are activated, even if no error condition has been detected.

Another way to achieve this result is to ensure that there are one or more mechanical end-stops which guarantee that the shuttle cannot move past them. In this case a “blind” position-reset may be attempted by repeatedly driving the shuttle at one or other of the end-stops a sufficient number of step-cycles that were it not to be blocked then it would definitely have reached the end stop (with a significant number of over-steps to allow for miss-steps or blocked steps, if required). In this way there is a high probability that the shuttle will be at the end stop.

Optimised Actuator

A specific implementation of the SMA stepper actuator is described below, and shown in figures 51 , 52, and 53. In this description, directional indications (e.g. left, right, up down, etc) should be taken as indicative of directions on the drawings only - that is, relative directions, and not absolute.

This embodiment of the stepper actuator is configured to have a stroke of ~19mm; a step size of ~300micron; and a step number of 64. It has been found that a stepper actuator according to this embodiment has a positional precision of <10micron; a positional stability of <10micron, lasting approximately 24hrs, with or without power; a maximum working load of Igramme; an overload tolerance of 25gramme; a maximum step rate >10 steps/sec, and; an energy consumption per step of <15mJ.

As shown in figures 51 , 52 and 53, the SMA stepper actuator 5100 is formed by stacking a number of punched, stamped, etched or laser/ion-beam (or other suitable high resolution cutting process) cut thin layers of stainless-steel sheet, each between 200 and 300 micron thick, the layers assembled over a printed-circuit-board (PCB) base. Each separate component has guide holes formed through it, that are aligned with a series of pins 5040 mounted in the PCB base. These serve to align all the components in the assembly. The PCB provides mechanical support, rigidity, and placement precision (along with its guide pins 5040) for all of the other components, as well as providing electrical connections to the SMA wires. The top plate 5041 is shown uppermost in figure 51. In one of the lower layers (underneath layer 5041) can be seen a small portion of a fast slewback return spring 5031. Extending from the right-hand end in this particular implementation can be seen the overload protection spring 5029 for the SMA drive wire 5025 (shown on figure 53). The overload protection spring 5029 is shown in the un-pretensioned state and has not yet been assembled to the SMA drive wire 5025 in this illustration.

The first metal layer directly above the PCB is shown in detail in figure 52 and figure 53. The first metal layer directly above the PCB is the primary functional layer, containing the drive-pawl 5012 and lock-pawl 5001 , their associated leaf springs 5011 , 5002, guides for controlling the range of possible movement of the shuttle 5003, 5004, as well as cavity space for the two SMA wires 5020, 5025. The shuttle

5030 itself is also sited in this layer but is a separate component, which in some variants may be added through the edge of the actuator after the other layers have been stacked and clamped together. The next layer above the first metal layer preferably contains the overload-protection spring 5029 for the drive-pawl SMA-wire 5025 (or spring 5029 may alternatively be sited beneath this layer, in another layer, or even externally as seen in figure 51 in its un-pretensioned state). This layer also acts as a spacer between the functional layer and the fast-slew-back return spring

5031 in the next layer above this one. The functional components (pawls, springs and shuttle) are interlinked as required between layers by fold-down or fold-up tabs which extend out of the plane of the layer in which they originate, through holes in any intervening layers, and interlock with matching holes or slots in the other layer they connect to. The final layer 5041 is a cover layer that keeps dirt out and keeps in any lubricating materials and heat-sinking materials.

All components which move within a layer (e.g. pawls, springs, levers, shoes), are surface etched, so that they are slightly thinner than the surrounding“frame” or stator part of that layer. This helps to ensure positive clearance top-bottom when the layers are clamped together. This clearance is typically 20micron.

This embodiment of the actuator allows adjustment of certain mechanical parameters for optimisation of the operation of the SMA stepper actuator 5100. Figure 52 shows the primary functional layer only. The multiple parallel-leaf spring 5002 extends downwards and rightwards from the top portion of stator. Extending downwards (in the drawing) from spring 5002 is the lock-pawl 5001 , which carries a set of downwards facing triangular teeth that mesh with similar teeth on the shuttle 5030 which is shown out of position in this figure, for clarity. The natural position of the leaf spring is as shown (i.e. the position as manufactured). However, insertion of the shuttle 5030 between the shuttle guide faces 5003 and 5004 lifts up the lock-pawl 5001 against the pressure of its leaf spring 5002, providing the required

retaining/locking force Flk for the shuttle, while the action of the parallel leaf spring maintains its row of teeth parallel to those on the upper edge of the shuttle. Because the teeth are triangular (in this case with a tooth slope angle of around 45deg) pushing the shuttle laterally (L-> R or R-> L) with a force slightly greater than Flk in either direction will cause the lock-pawl 5001 to be pushed upwards against this spring force and allow the shuttle 5030 to move. This provides external overload- force protection since out-of-spec load forces on the shuttle simply cause slippage without damage. The pawl 5001 , its leaf spring 5002 and the adjacent section of the stator 5000 are all one piece of material, which could be etched out of a thin sheet of stainless steel (or 3D printed), as a single component. No assembly is required of these separate parts. It will be seen that this method of fabrication enables the pawl and its spring to be made as just one integral part with the surrounding frame/spacer, enabling very low manufacturing and assembly costs.

To the right of the lock-pawl 5001 is an unlock lever system that comprises an unlock lever arm 5005 which is driven by the unlock SMA actuator wire 5020 (shown in figure 53), so as to lift the lever 5005 against the spring force of the long supporting arm or body of the lever 5005 when the wire 5020 is actuated. The unlock lever system is integral with the frame 5000 of this layer, connecting to the frame 5000 via the spring arm body of the lever arm 5005, allowing the left-hand end of the lever arm 5005 to swing up and down. A peg 5006 on the left-hand end of the lever arm 5005 fits loosely in a slot 5007 on the right-hand edge of the drive-pawl 5001. The width of the slot 5007 is such that when the shuttle 5030 is assembled into the actuator the drive pawl 5001 is free to move up and down, driven by the shuttle teeth meshing and un-meshing with the pawl teeth, when the shuttle moves left or right without interfering with 5005. However, when the unlock SMA wire 5020 is actuated this pulls the unlock lever 5005 upwards, sufficiently far that the peg 5006 engages with the upper edge of the slot 5007 and lifts the drive-pawl 5001 completely clear of the shuttle teeth. With the unlock lever 5005 in this actuated position, the shuttle is free to move left or right - i.e. is unlocked.

A drive-pawl 5012 extends upwards from the bottom portion of stator 5010, the drive- pawl mounted upon a multiple parallel-leaf spring 5011. Pawl 5012 carries a set of upwards facing triangular teeth capable of meshing with similar teeth on the lower edge of the shuttle 5030 (when assembled). The natural position (as manufactured) of the leaf spring 5011 is as shown in figure 52. The un-activated SMA drive wire 5025 (not shown in figure 52, shown in figure 53) when assembled is attached to the right edge of the drive-pawl 5012 and wire 5025 pulls rightwards on this pawl 5012 and is held in sufficient tension by the then pre-tensioned leaf spring 5011 to pull the drive-pawl 5012 upwards and rightwards (because of the direction of the spring leafs of 5011) such that the teeth of the pawl 5012 are positioned just below but completely clear of those on the shuttle lower edge (after assembly as shown in Figure 53). The action of the parallel leaf spring 5011 is to keep the row of drive-pawl teeth at all times parallel to the row of shuttle teeth. When the drive wire 5025 is actuated it shortens, and pulls the drive-pawl 5012 further to the right and further upwards, engaging its teeth with those on the shuttle, and because of the carefully arranged relative phase of the two sets of teeth, by the time the sets of teeth are fully meshed, the drive-pawl 5012 has pulled the shuttle 5030 more than half a tooth length to the right, against the external load force, internal friction forces, and the forces imparted on the shuttle by the lock-pawl. Because the shuttle 5030 has moved more than half a tooth length under the action of the SMA drive wire 5025, the lock-pawl 5001 has by now just lifted over the shuttle teeth previously immediately on the left of its teeth, and begun its descent sliding down their LH sides. The drive wire 5025 is de-actuated and as the drive wire 5025 cools and re-expands under the return force of the drive-pawl spring 5011 , the drive pawl 5012 disengages from the shuttle teeth and the lock-pawl 5001 under the pressure of its leaf spring 5002 pushes down onto and into mesh with the teeth of the shuttle, moving it rightwards the rest of the whole tooth-length, and then locks it firmly and precisely into place, exactly one tooth further to the right. The bottom end 5013 of the drive pawl leaf spring 5011 is firmly anchored into the stator 5010, and is in the most preferred form an integral part of this component so as to avoid separate parts and an assembly step. The remainder of the stator 5010 in this layer of the actuator also forms a reference position for the“fixed” end of the drive SMA wire 5025. Both sections 5000, 5010 of the stator in this layer are fitted with guide holes 5015 which upon assembly locate the stator sections over guide pins (not shown) fixed in the base plate of the actuator (not shown), to ensure precise relative positioning of parts.

It will be seen that apart from the SMA actuator wires 5020, 5025, all of the other components in this layer (stator, pawls, springs, levers and shuttle) may be etched or punched etc out of a single sheet of stainless steel in one operation, resulting in just three components - an upper plate, a lower plate, and a shuttle. Assembly consists simply of placing the upper and lower plates over the guide pins in the base (not shown) to precisely locate them relative to each other, and then inserting the shuttle. In some implementations the shuttle 5030 may be inserted after the entire actuator has been assembled.

Figure 53 shows the same layer of the actuator in assembled form. The shuttle 5003 is now located slidingly between the upper 5003 and lower 5004 edges of stator parts 5000 and 5010, and is in full tooth-mesh contact with lock-pawl 5001. The unlock SMA wire 5020 can be seen between its crimp contacts 5022 and 5021 which locate in the unlock-lever 5005 and stator 5000 respectively. Wire 5020 is held taut when de-actuated by the prestress spring force in the deflected unlock lever 5005 which is held by wire 5020 just clear of the shuttle upper teeth.

Because the lower teeth of the shuttle are triangular (again in this case with a tooth slope angle of around 45deg) pushing the shuttle laterally in either direction with a force sufficient to overcome the meshing force provided by the SMA drive wire 5025 will cause the drive-pawl to be pushed downwards against this force and allow the shuttle to move. This provides a measure of external overload-force protection for the drive-pawl mechanism and drive wire 5025 since out of spec load forces on the shuttle cause slippage without damage. However, an additional level of protection for the SMA drive wire 5025 is provided by the use of a pre-tensioned overload spring 5029 (not shown in this figure) which firmly pushes the right-hand terminal 5026 (which is slidingly mounted in slot 5027 of stator 5010) of SMA drive wire 5025 against an end-stop 5028 integrated into stator 5010, with a force Fovl, where Fovl < Fwmax, Fwmax being the maximum tolerable wire tensile force. For wire 5025 tensions < Fovl terminal 5026 remains stationary and pressed against end stop 5028 by spring 5029. Should the SMA drive wire 5025 experience for any reason a tensile force greater than Fovl, then the overload spring 5029 will allow the terminal 5026 to move left with the wire tension now being a function of the overload spring 5029 pre tension force, it’s spring-rate, and the amount of leftwards movement. With suitable values of these parameters, the SMA drive wire 5025 can be maintained at a safe tension < Fwmax under all conditions.

The method of fabrication described above enables the pawl 5012 and its spring 5011 to be made as just one integral part with the surrounding frame/spacer 5010, enabling very low manufacturing and assembly costs. It is also possible where the materials of the actuator are electrically conductive (e.g. stainless steel, phosphor bronze) to integrate at least one of the SMA wire terminals of each wire 5020, 5025, with the relevant part of the actuator assembly, preferably via a crimp terminal also integral with the structure. For example, for the drive wire 5025 it may be most convenient to integrate the left-hand terminal 5031 of wire 5025 with the drive pawl 5012, relying on the metal frame of the actuator to provide a return path for electrical heating current of the wire 5025. For the unlock wire 5020, the moving end 5022 may also be crimped directly to unlock lever 5005 with an integral crimp formed out of the same material as the lever. The other ends 5026, 5021 , may conveniently be crimped directly to electrically isolated crimps mounted on the base (e.g. a PCB) to complete the electrical circuits of these SMA wires without additional wires., or where an overload protection spring is used (as described for SMA drive wire 5025) the material of the overload protection spring 5029 itself used to provide a conduction path from terminal 5026 (then integrated with moving end of spring 5029) to the fixed end of spring 5029 which itself is then conveniently mounted into the PCB for external connection via a conductive PCB track. Such techniques eliminate separate components, assembly steps, and minimise auxiliary wiring and connection problems.

It can be seen that:

• a step of the shuttle may be achieved simply by pulsing on the drive SMA wire for a short time and then turning it off, the final part of the stroke happening automatically under control of the lock-pawl spring. This provides a very fast step cycle.

• Because the drive pawl need move the shuttle only just greater than half a tooth length, the length of SMA wire needed (and therefore the power required to activate it, are smaller than most other implementations, which lowers cost and drive power requirements.

• This embodiment of stepper actuator is resistant to external overload force damage because in un-actuated mode, excess load force simply pushes the shuttle harmlessly beneath the spring-loaded lock-pawl’s teeth.

• It is also easy to protect the SMA drive-wire against excess force from any overload or fault condition, by the use of an overload-force pre-tensioned spring.

• For unidirectional stepping no lock-pawl lift is required. Return of the shuttle may be achieved simply by an external force >Flk pushing on the shuttle, which can slip under the lock pawl without damage.

• The lock pawl and drive pawl may be placed either on opposite sides or the same side of the shuttle as convenient, in the latter case only one set of shuttle teeth then being required.

• A second drive pawl with the direction of its spring and SMA drive wire

assembly reversed L to R may be fitted on either side of the shuttle wherever teeth are provided, which then allows full bidirectional operation of the shuttle (so long as simultaneous drive of both drive pawls is avoided). Again in this case no SMA unlock drive wire is necessary.

• Adding a fast-slewback spring as described elsewhere herein allows rapid shuttle return to zero-position just by activating the SMA unlock wire.

• The method of manufacture of this device as a series of stacked (thin) layers allows very simple manufacture of assembly, whereby most of the otherwise separate components (pawls, springs, levers, interlocks, crimps, wires, end stops, guides) are all made simultaneously in one step (e.g. by etching, laser cutting, ion-beam cutting, punching, stamping) from one substrate and require little additional assembly work, which all helps to reduce manufacturing costs.

• This particular form of the SMA linear stepper actuator lends itself particularly well to conversion into a rotary actuator: if the straight shuttle is replaced by a toothed cog, and the lock-pawl and drive pawl(s) are disposed around the cog so as to not interfere with each other, but to otherwise interact with the cog as previously they did with the toothed shuttle (in the previous configuration described above wherein all the pawls were on just one side of the shuttle), and, where the drive and lock pawl(s) are modified in shape to allow their set of one or more teeth to wrap around the cog and appropriately mesh with multiple teeth, then, the configuration with one lock pawl (no unlock SMA wire) plus one drive pawl is capable of continuously stepping the cog in the same direction, providing a unidirectional rotary SMA stepper motor; if a second drive pawl is added driving in the reverse direction around the cog, then a fully bidirectional rotary SMA stepper motor results; if a lock-pawl unlock assembly with an unlock SMA wire as before is added to disengage the lock-pawl from the cog when the SMA wire is activated, then the cog can be allowed to rotate freely, and for example, should a spiral clock-spring be attached to the shaft of the cog, a very fast rewind of the cog (after a series of forward steps) can be produced.

Optimised bidirectional rotary stepper actuator

Figure 54 shows a form of SMA rotary stepper actuator, similar in operation to the optimised SMA linear actuator just described, except that the linear shuttle 5030 has been replaced by a rotary cog 6000 rotatably mounted in stator 6050 and the remaining mechanism has been modified accordingly, as described below.

A lock-pawl 6002 mounted on a pre-tensioned leaf spring 6006 that is firmly rooted in the actuator base/stator 6050 at 6018, has teeth intermeshed with those of the cog 6000 providing a lock force resisting rotary motion of the cog 6000 up to torque level Tlk (a function primarily of the force provided by the leaf spring, and the radius of the cog). An anticlockwise drive pawl 6001 with teeth capable of meshing with those of cog 6000 is mounted on leaf spring 6005, which is pre-tensioned by connecting wire 6014 and anticlockwise SMA drive wire 6008, is held just completely clear of cog 6000 when SMA wire 6008 is unheated/deactivated. Similarly a clockwise drive pawl 6003 is mounted on leaf spring 6007 and held just clear of cog 6000 by tension in connecting wire 6015 and clockwise SMA drive wire 6009, when 6009 is un activated.

In each case the connecting wires and SMA wires are mechanically linked by slidably mounted crimp blocks 6012 and 6013 which transmit tension freely between the wires and are merely guided by slots fixed to the base 6050. Both drive pawl leaf springs are firmly rooted at their base ends 6016, 6017 to the actuator base/stator 6050. The connector wire and connector block arrangement shown is optional and is here in this embodiment to keep the SMA wires moving substantially in a straight line during a step cycle, via guide holes in stator 6050 (not visible in figure 54) so that active cooling techniques may be applied over the length of each SMA wire to speed up the step cycle. Otherwise the SMA drive wires may be connected directly to the drive pawls. In either case the return electrical connections from the moving ends of drive wires 6008. 6009 can be made by e.g. flexible flying leads or via the pawls and leaf-springs to which they are mechanically and optionally electrically connected (and thus most conveniently connected to the stator 6050 as a common ground connection).

In operation, for an anticlockwise step:

• SMA wire 6015 is un-actuated.

• SMA wire 6008 is heated and actuated.

• This pulls on the anticlockwise pawl 6001 via the block 6012 and

connecting wire 6014 in a direction so as to first engage its teeth with those of the cog 6000 and then as it progresses towards full mesh it pushes the cog teeth it contacts in an anticlockwise direction around the cog shaft, rotating the cog by greater than half a tooth width.

• Thereafter the SMA drive wire 6008 is de-actuated (and allowed to cool) whereupon it is pulled back to full martensite length by the spring return force of spring 6005.

• In so doing the drive pawl 6001 disengages from cog 6000 at which point the lock pawl 6002 under the force of pre-tensioned leaf spring 6006, pushes its teeth down onto and into mesh with the teeth of the cog in the next clockwise position of the cog, which guarantees that the cog rotates anticlockwise by precisely one tooth width.

• Thereafter the cog is held firmly in that position against any load torques smaller than Tlk connected to the cog shaft.

• Rotation in the clockwise direction is similarly produced by activation and deactivation of clockwise SMA drive wire 6009.

• In a manufactured SMA rotary stepper actuator of this type, all of the pawls, springs and guides and possibly at least one crimp end for each SMA wire can be integrated into a single piece of suitably

etched/punched etc metal as described previously for the linear stepper actuator on which this device is based, making for a very low cost and easy to assemble device. If the active cooling device feature isn’t needed then one end of each SMA drive wire 6008, 6009 may be mechanically connected directly to each of the two drive pawls in place of the connecting wires 6014, 6015, and the connecting wires and guide holes dispensed with.

SMA Wire Mechanical Overload

As outlined in the 'background' section above, an issue that can occur with SMA wire-actuators is overload, where the force applied to the SMA-wire due to abnormal conditions at the load (e.g. out of specification load conditions) is sufficiently large to damage or even break the SMA-wire. For a simple SMA-wire actuator, applying significantly more load force than the designed maximum allowable wire stress will tolerate can permanently damage or break the SMA wire. Mechanical load-end-stops alone, positioned to limit the travel of the free-end of the wire to within the designed stroke range, are ineffective in preventing such damage when the SMA-wire is in the austenite phase, since its maximum safe length will then be inside the limits allowed by the end-stops which will therefore not prevent over-stretching.

An analysis of the conditions is detailed in Appendix A.

Specific embodiments of the pre-stressed spring mechanisms intended to assist with overcoming this issue are shown in figures 17 to 22.

In figure 17, an embodiment of the SMA-wire mechanical overload prevention mechanism is shown. A wire mount 71 is mechanically coupled to one end of a pre stressed tension spring 73 via a first pin 72 that is securely coupled to mount 71.

The spring 73 is stretched so that its other end may be coupled to a second pin 74 which rests in slots 78 in the mount 71 , the pin being held in those slots by the pre stressed tension Fm in the spring 73. A wire terminal 75 with crimp-end 76 is mechanically attached to pin 74.

The SMA-wire 77 of the actuator, of which this assembly is a component, is crimped within the crimp-end 76 of the terminal 75. It can be seen that while the tension in wire 77 is smaller than Fm then pin 74 will not move away from mount 71 , so the effect is that SMA-wire 77 is for all intents and purposes rigidly attached to mount 71 , specifically in the direction of tension along the wire. When the tension in wire 77 exceeds Fm then spring 73 will stretch beyond its pre-stressed length allowing pin 74 and the attached terminal 75 and thus SMA-wire 77 end to move away from mount 71 , the spring-rate of spring 73 determining the distance moved for such forces greater than Fm. Thus if Fm is chosen to be smaller than the safe SMA-wire tension limit, then overloading of the SMA-wire 78 will be reduced. Adding an end-stop to the terminal attached to the other end of wire 77 limiting its ability to travel more than a fixed distance away from mount 71 then provides complete overload protection of wire 77 subject to a suitable choice of spring-rate for spring 73 and suitable pre stress force Fm.

In the embodiment shown in figure 17, the wire-overload protection assembly is made up of a number of parts. For large wires this is practical. However, for smaller wires in particular, it is possible to combine all of these functions into one component, as shown in figure 18.

In figure 18, the terminal 85 is extended away from its crimp-section 86, with the SMA-wire 88 attached as described in the previous embodiment. The extended section of 85 is bent to form a serpentine tension spring 82, loops of which are shown. An extended loop nearest to the crimp-end allows for extension of the spring portion of 85 in the direction of the wire. The end of terminal 85 farthest from the crimp-end 86 is the attachment point for the mounting 81. Mounting 81 in this schematic embodiment has been modified to provide an end-stop for the spring portion of terminal 85 which holds the spring in tension against its installed pre stressed force - i.e. the spring portion as shown is already stretched from its natural length with a pre-stress force Fm as before, and is held in tension against this force by two different sections of the rigid mount 81. Pin portion 82 of mount 81 holds the distal end of the spring portion of terminal 85, and another portion of mount 81 is inserted within the extended loop in line with the line of the SMA-wire, to hold that end in place against the spring tension. Thus the entire pre-stressed- spring overload protection device can easily, simply and cheaply be fabricated from the same piece of material (e.g. phosphor bronze or stainless steel) used to crimp the wire end, and this single component can thus provide electrical connectivity, rigid mechanical anchorage for the wire, as well as stress-overload protection.

The embodiment shown in figure 19 is a variation of the embodiment of figure 18, optimised for insertion directly into a printed circuit board (PCB). A single piece of flat metal (in the preferred embodiment a phosphor bronze or stainless steel, although other suitable metals can also be used) is formed into a crimp 96 at one end, a pair of PCB insertion-pins 91 , 92 and a serpentine tension spring, some loops of which are designated 93. In use the tension spring section is stretched into some restraining mounts (to pre-stress the spring) and held in place by the pins 91 , 92. The SMA-wire 97 is crimped into the crimp-end 96 and excess tension in the wire (i.e tension above the pre-stress put into the tension spring), then causes the spring to stretch further thus providing overload protection for the SMA-wire.

Figure 20 shows the one-piece terminal and protection spring of figure 19 inserted into a PCB, with pin 91 inserted under tension into a tight-fitting hole, and preferably soldered to a PCB track (not shown) to provide additional mechanical and excellent electrical connection. Pin 92 is inserted into a slot 500, one end of which provides an end-stop for the pre-stress tension in the spring while the rest of the slot allows clearance for the spring to stretch further in overload situations. The slot 500 should be long enough to provide complete clearance for the pin 92 under the maximum overload anticipated.

The underside of the arrangement of figure 20 is shown in figure 21 , which shows detail of the location of the hole 501 for pin 91 in the PCB and a clearer view of the slot 500, giving clear space for the pin 92 to move freely under overload conditions, constrained only by the force of the spring.

The SMA-wire overload protection invention as applied to a bowstring actuator is shown in figure 22. Here an SMA-wire 600 crimped at both outer ends to mechanical and electrical connections engages with a load-attach-pin 60 at or near the wire’s centre. The load attach pin 60 is mechanically connected to the push-rod 61 of the actuator (the moving part, to which the external load is coupled) via a connecting member 64 and pin 68 which in turn engage with a spring mechanism 63 whose other end is mechanically connected to the push-rod 61. The spring mechanism 63 is pre-stressed with a force Fm to hold pin 68 firmly in the push-rod, until such time as the tension of the wire is great enough to pull the load attach pin 60 with a force greater than Fm (an overload situation). In this case Fm is set at a value guaranteed to keep the SMA-wire at a tension level smaller than its safe maximum working tension at all actuation positions of the wire. In a preferred embodiment of the bowstring actuator overload protection mechanism, the connecting member 64, pin 68, spring mechanism 63 and load-attach-pin 60 are all made from the same contiguous piece of material (preferably metal or polymer) for simplicity, low cost and ease of assembly.

An actuator using a double bowstring SMA-wire is shown in figure 23, with both SMA wires protected from overload by pre-stressed spring terminations as described above. An alternative embodiment of actuator using a double bowstring SMA-wire is shown in figure 24, this embodiment using a common overload spring.

Optimised Very High Speed High-Speed Actuation

A detailed analysis of optimised high-speed actuation with SMA-wire is included in Appendix B. Specific embodiments of the pre-stressed spring mechanisms intended to assist with overcoming this issue are shown in figures 25 to 27.

A fast cycle-time dual SMA actuator is shown in figure 25. The SMA actuator wire 257 pulls load 25150 in opposition to SMA wire 25107. Both SMA wires 257 and 25107 have one of their ends mechanically attached to base 1 via pre-stressed springs 2523 and 25123 respectively, whose pre-stress forces F2 and F1

respectively and spring constants are calculated as described in Appendix A, and which consequently hold the crimp-ends 256 and 25106 of each wire against backstops formed by the edges of holes cut in base 251. The actuator is shown in a state where wire 257 is shorter than wire 25107 and the load 25150 is displaced significantly towards the root end of wire 257.

This can be taken to be the starting position of a high-speed cycle, with both wires in martensite state, and cool (below their Austenite start temperature).

Then wire 25107 is heated very rapidly (for very high speed actuation), wire 25107 becomes rapidly converted to austenite, gets stiff and short, and pulls load 25150 towards the spring 25123, and in so doing pulls the load-end of wire 257 (cool, martensite, and less stiff) with it. If at any time during this transition, the tension in the wires exceeds a certain threshold force F1 , then spring 25123 with pre-stress force F1 will start to expand, allowing the end of wire 25107 to move away from its end stop, relieving excess tension in the wire. The load reaches its destination position and if external load forces are in spec (<F1) then spring 25123 will again hold crimp end 25106 against its backstop. Wire 257 is then very rapidly heated (again as described above), even if wire 25107 is still hot / mainly or wholly austenite and stiff and short. Wire 257 pulls with a force Fw >F1 and in fact sufficiently hard to move the load 25150 (and if necessary pull the SMA-wire-end of spring 25123 with it) until the load reaches it’s final position. If at any time during this transition, the tension in the wires exceeds a certain threshold force F2, then spring 2523 with prestress force F2 will start to expand (in which case spring 25123 will already be expanded as it’s pre stress force F1 <F2), allowing the end of wire 257 to move away from its endstop, again relieving excess tension in the wire. Once the load has reached its final position, and both wires have cooled, the springs 2523 & 25123 will contract and once again hold the SMA wire crimps against the end stops, ready for another bi directional cycle. Note that the cycle-cycle time of this actuator is determined by the sum of the heating times of both wires plus the cooling time of the slowest to cool wire (usually the second-to-be-heated wire). However, the pull-one-way then pull- back-the-other-way cycle-time itself can be much shorter and is equal to just the sum of the heating times of both wires. Typically, for a 25pm SMA wire, the cycle-cycle time in free air may be as long as 100ms or more; however, the cycle-time itself can be as short as 1 ms-5ms.

A variation of this arrangement is shown in figure 26, which shows a double bowstring SMA-wire actuator. SMA wire 267 is the first wire of a first bowstring actuator, with its ends crimped at 2620, 2621 into mechanical and electrical terminals with pre-stressed overload protection spring 263. SMA-wire 26151 is the second, opposed, bowstring actuator and similarly overload protected with its own separate protection spring.

Another variation of this arrangement is shown in figure 27, which shows a double bowstring SMA-wire actuator. SMA-wire 277 is a first SMA wire of a first bowstring actuator, with its ends crimped at 2720, 2721 into mechanical and electrical terminals with shared pre-stressed overload protection spring 273a. The second SMA-wire 27151 of a second, opposed, similarly overload protected bowstring actuator with a protection spring 273b is also shown. The external load (not shown) is connected to the junction of springs 273a, 273b.

Definitions

Where the phrase 'hooking point mechanism' is used in this specification, this should be taken to mean meshing sets of substantially similar teeth on a pair of components.

Where the phrase 'hook mechanism' is used in this specification, this should be taken to mean features on another component capable of hooking into the hooking-point mechanism on the first component, which then do not need to be similar to each other, although symmetrical teeth fit within this broader definition.

Where 'pawl' is used in this specification, this should be taken to mean a mechanical component that moveably engages with a set of teeth on another component that may lock or at least resist relative movement in at least one direction. This definition should be taken to include pawls with single or multiple hook means including teeth, that can lock or partially resist motion in one or both directions, and which may be additionally slidably mounted or free-floating relative to a stator. Where the phrase 'ratchet tooth' is used in this specification, this should be taken to mean a gear tooth having one side radial and the other inclined so that a pawl will catch firmly on the former and slide over the latter. This definition should be taken to also include ratchet-teeth on a straight rack of teeth (instead of teeth around a gear) where 'radial' then can be taken to mean 'normal to the line of teeth', and 'inclined' can be taken to mean an angle of less than 90deg to the line of teeth. 'Ratchet- teeth should also be taken to mean asymmetrical mechanical teeth such that when engaged they grip in one direction more strongly (the grip- direction) and slip in the other direction more easily (the slip-direction).

Where 'shuttle' is used in this specification, this should be taken to mean a component that is regarded as moved by the action of the actuator(s) - for example by engagement with the moving pawl(s). In cases where the body of such an actuator is fixed, then the shuttle moves relative to it. However, the shuttle can also be regarded as fixed, with the actuator body moving when the actuator is actuated.

The embodiments of the invention as described above help to overcome the long term positional stability issues of SMA linear actuators, including the position sensitivity to mechanical load, as well as providing a precise, stepped linear motion. A true zero- power fixed-position capability is also provided so that power is consumed only when the actuator is changing its position. These embodiments also provide an actuator where the power consumption is not related to the total actuator stroke, and which for most applications will require significantly lower power than standard SMA-wire linear actuators with similar performance.

When discussing an actuator, the following definitions are used:

• a baseplate onto which are mounted the other components, made of material with linear coefficient of thermal expansion a _B ;

• an SMA wire of length = 2Lc when at the SMA martensite-finish temperature, Mf,

• the SMA wire has length = 2Lh < 2Lc when at the SMA austenite-finish

temperature, A†,

• a mechanical push-rod mechanically connected to the moving end of the SMA wire (“the wire”) being the effective output port of the actuator, made of material with linear coefficient of thermal expansion a _P ; • the length of the push-rod between the connection of the push-rod to the SMA-wire, and the actuation point (i.e. the load to be moved) is D;

• the actuator is necessarily mounted to some substrate or system frame of reference. Let M be the orthogonal distance along the line through the SMA wire-ends from the fixed SMA wire end to the system frame of reference;

• k is the wire factor, defined by Lh = Lc (1- k) or equivalently k = 1 - (Lh/Lc), and is typically between 3% and 8% for commercially available SMA wires.

In what follows when discussing a bowstring actuator, the following definitions are used:

• the SMA wire has length = 2Lh < 2Lc when at the SMA austenite-finish

temperature, A†,

• distance between the clamped (e.g. crimped) ends of the SMA wire =2Y <

2Lh;

• a mechanical push-rod mechanically connected to the centre of the SMA wire (“the wire”) being the effective output port of the actuator, made of material with linear coefficient of thermal expansion a _P ;

• the push-rod is oriented more or less orthogonal to the line through the crimp- ends of the SMA-wire;

Because Lc >Lh > Y, when held taut by orthogonal thrust from the push-rod, the wire forms a V-shape (hence the name“Bow-string”), with the push-rod connection to it at the point of the V, and the wire-ends at the two tips of the V. Depending on the direction of thrust in the push-rod, the point of the V is either on the same side ( Case - 1) of the line through the wire-ends as the load-end of the push-rod, or the opposite side (Case- 2).

In a multilayer printed circuit board (PCB) a technique called plated-through-hole (PTH) is used to interconnect any set of contiguous layers (i.e. not necessarily from the top copper layer all the way through to the bottom copper layer). The term 'via generally refers to such a PTH that goes all the way through from top to bottom, providing direct electrical connectivity. For DC and low frequencies this works in the same manner as a wire. At RF, if the spacing between vias is «wavelength, then the "wall of vias" acts like a contiguous metal sheet" completely screening one side of the wall from the other at such high frequencies (in a similar manner to how a wire mesh acts like a continuous conductive surface on some antennas). Cell Phone Camera Zoom Lenses

The present invention can be used to overcome issues of miniature-camera zoom- function actuation. Using any of the SMA linear stepper actuators of the present invention - for example, the embodiments described above - it is possible to provide a miniature SMA actuator with the following capabilities:

a) move one or more optical elements a distance independent of the length of the SMA-wire forming the drive element of the SMA actuator;

b) hold the optical element(s) precisely in the commanded position

indefinitely, without the application of any power to the actuator; c) to consume power only during a change of position of the optical

element(s);

d) to provide a multitude of zoom factors (e.g. ~12 steps between 1X and 10X zoom) with just one miniature camera fitted with zoom optics and an SM linear stepper actuator of any of the varieties described herein; e) use only a relatively short piece of SMA wire (length L related to the step- size of movement, and not to the total length of movement of the optical element) to produce actuation, thus incurring a low cost of manufacture. An alternative arrangement can use one of the SMA rotary stepper actuators described herein, to turn a lead-screw to which the optical element(s) of the miniature camera are attached by a nut which moves axially along the lead screw when the screw rotates.

For bi-directional zoom control (the most common form used in cameras), then one of the bidirectional forms of the SMA stepper actuators described herein may be used as the mover of optical element(s). Alternatively, a possibly less convenient option is to use a uni-directional linear stepper-actuator with a“fly-back” mechanism (generally a return spring and a lock-pawl release actuator, which when activated causes the linear stepper to return quickly under spring return-force, to its nominal; zero position) as are also described herein.

It is frequently the case in zoom cameras that two or more optical elements need to be moved simultaneously but with differing motions. As in conventional (known) zoom lens embodiments, this differential motion of the 2 or more lenses may be achieved by coupling the two or more optical elements via a mechanical

linkage/mechanism that compels each of the elements to move relatively as desired when any one of them is caused to move by an outside influence (e.g. an SMA stepper actuator as described above). However, because a suitable SMA stepper actuator for zoom implementation on a miniature camera is potentially very small and lightweight, it may in fact be advantageous to dispense with one or more such mechanical linkages between optical elements, and drive each element (or group of elements) independently each via their own connected SMA stepper actuator. This has the advantage of adding a degree of freedom of placement of the two (or more) separate stepper actuators, and elimination for any mechanical coupling between the optical elements themselves.

So in an embodiment of the present invention, a miniature camera comprises at least one optical element capable of movement substantially along the optical axis of the camera and mechanically directly connected to the moving shuttle or output node of an SMA linear stepper actuator as described in any of the embodiments herein, wherein the body of said actuator is fixed relative to the body of the camera, the direction of motion of the shuttle of the actuator is substantially aligned with the optical axis of the camera, such that actuation of the actuator causes the at least one optical element to move along the optical axis of the camera.

In another aspect of the present invention similar to the previously described aspect except for the following differences, the at least one optical element is indirectly connected to the moving shuttle of the SMA linear stepper actuator via a linkage mechanism, and the orientation of the movement axis of the actuator shuttle need not be aligned with the camera optical axis, the linkage being such that movement of the actuator shuttle along the actuator movement axis causes via the linkage, movement of the at least one optical element along the camera axis.

So in a further aspect of the present invention, a miniature camera comprises at least two optical elements or groups of elements capable of movement substantially along the optical axis of the camera, each element or group thereof mechanically directly connected to the moving shuttle or output node of a different independently controlled SMA linear stepper actuator as described in any of the embodiments described herein (i.e. one actuator per element or group thereof), wherein the body of said actuators are fixed relative to the body of the camera, the directions of motions of the shuttles of the actuators are substantially aligned with the optical axis of the camera, such that independent actuations of the actuators causes the optical element or groups thereof connected to each actuator to move independently of each other along the optical axis of the camera. In a further embodiment of the present invention any one or more of the SMA linear stepper actuators in any of the three previously described aspects of the miniature camera optics inventions, is replaced by an SMA rotary stepper actuator connected to a leadscrew either directly or via a gearing or belt so as to rotate the leadscrew when the stepper actuator rotates, and where the leadscrew is fitted with a nut that moves axially along the leadscrew when the leadscrew rotates, and wherein the at least one optical element or group thereof is connected directly or via a linkage to the nut, instead of (as previously) to the linear stepper actuator shuttle.

Phase or Frequency Tuneable RF Device Exploiting Properties of SMA

Standard prior-art SMA linear-actuators capable of precision positioning to multiple different positions (as opposed to on-off actuators with essentially just two fixed positions) essentially rely on a continuously powered length of SMA material (most conveniently in the form of a Nitinol wire or tape or strip).

In the embodiments described in this specification:

• the SMA wire/tape/strip length is independent of total actuator stroke;

• Zero drive power is required to hold at a fixed position;

• Long-term position stability is independent of SMA material ageing, ambient temperature, external load force (if kept within rated limits);

• Operating power is independent of total actuator stroke;

• Faster operation may be achieved than is usually feasible;

The embodiments of stepper-actuators described and shown in this specification were initially designed for use with Phase or Frequency Tuneable RF Devices, and are particularly suited for this purpose.

Some of the known prior art related to tuneable RF filters are outlined in the

'background' section above. The problems with current low-loss fully tuneable filters are that they are big, heavy, expensive to make and even more costly to actuate (i.e. to control dynamically in service). Conventional tuneable RF filters are primarily machined out of solid metal, usually aluminium, which is a slow and expensive process, even when using CNC machining. Such fabrication methods also restrict the geometric shapes that may be easily made - e.g. holes or cavities blind at both ends must be fabricated in sections and then assembled together with well-fitting precision joints required to be highly electrically conductive. As far as actuation of moving mechanical components is concerned, rotary electric motors may be used but these motors are costly, bulky and require a significant amount of electronic control, and frequently also require a position sensor with related electronics to provide an accurate indication of the position of the moving component. Rotary electric motors also require some form of rotary-to-linear motion converter as well as a gearbox to produce the required component motions at the right speed, all of which add to the cost, size, weight and unreliability of the solution. Finally brushed electric motors are inherently unreliable and produce conductive particulate pollutants, whereas brushless motors are significantly more expensive and require more elaborate electronic control. Where a given tuneable filter has multiple tuning elements (often five or many more are required for high-performance) the cost of separately driving each of these elements is often prohibitive.

Complicated schemes of mechanical linkages to drive multiple tuning elements with one motor might possibly reduce the cost but add to the complexity and unreliability, and are generally less flexible than individual element tuning drives.

For example, in document WO2016/202687, because of RF energy leakage problems around metal tuning elements, the inventor is forced to instead use dielectric tuning elements. This is one of the problems we address below. Also in this same patent there is no teaching relevant to the integration of actuators to allow for programmable tuning (as opposed to manual tuning) and this is one of the primary issues we address below. Lastly, the coupling between cavities is fixed and not tuneable, and in what follows we advantageously address this issue also.

The embodiments below outline apparatus and methods by which these problems of low-loss RF tuneable filters can be addressed, by the use of electronically-controlled Shape Memory Alloy (SMA) actuators to move and or deform the tuning elements and/or the resonators and/or novel filter electromagnetic and mechanical

configurations. A useful form of SMA actuator for these functions is an SMA wire actuator where the contraction of a long thin wire of SMA material is used to provide a pulling force upon heating of the wire above its Austenite start temperature. We refer hereafter to such actuators very broadly as SMA wire actuators.

However, in certain instances it can also be useful to use SMA material in an actuator in forms other than wires, e.g. strips, sheets or even rods and bars. For convenience of description we include all such SMA actuator forms in the term“SMA wire” hereafter. In the present invention conventional tuning screws are replaced by moveable and/or deformable elements actuated by SMA wires. Both capacitive and inductive tuning elements (and combinations thereof) are used, and we tune not only the resonators to shift centre- frequencies, but also tune input-output (I/O) couplings, inter-cavity couplings, as well as the phase angle of predominantly phase-shifting elements.

By using one small SMA actuator (each with one or more wires, and where there is only one SMA wire in an actuator it is to be understood that some sort of mechanical spring force is used to achieve a return stroke, as is well known in the art of SMA actuators) to control each tuneable element separately, we avoid complicated and unreliable mechanical mechanisms, gears and levers otherwise needed to keep multiple tuning elements in synchronisation (in sync). Instead the synchronisation is done entirely in electronics and/or software for maximum flexibility and minimum complexity. There may occasionally be instances where it is convenient and appropriate to move more than one tuning element with a single SMA actuator, and such cases are to be understood to be included herein, though it is our contention that the most advantage will usually be gained by separately controlling each tuneable element with a dedicated actuator ideally independently controlled by software.

However, one drawback of simple SMA actuators / SMA wire actuator is their long term positional stability. Another is that to maintain their set position such actuators need to be powered continuously. Thirdly, their absolute positional precision over the long term can drift, because of SMA material ageing, fatigue, as well as other factors. Fourthly, the greater the stroke required of a simple SMA actuator, the greater the length of SMA wire is required to achieve that stroke (all else being equal) and thus it is difficult to make compact SMA actuators with long stroke. Fifthly, for precise (even short-term) position-control a complex high-precision electronic controller is required (for each SMA wire) capable to accurately estimate the SMA wire resistance (from which the wire length and thus the actuator position is estimated) or, some other form of precision position sensor needs to be added, all of which adds considerable size, power consumption, complexity and cost.

The embodiments of the SMA actuators described above assist with resolving these issues. These SMA actuators have: zero-power position hold; required total SMA wire-length is independent of actuator stroke; near-ideal short- and long-term positional stability, unaffected by material ageing and fatigue; very high positional precision; simple electronic control with no need for precision ADC and DAC components. In the description below it is to be understood that wherever any form of SMA actuator or SMA wire actuator or SMA-wire actuator is mentioned or shown below in what follows as part of a functional description or an embodiment or claim of the present invention, then an SMA-stepper-actuator in any of the forms described above may advantageously be substituted, because of the exemplary features of this particular form of SMA actuator, and all such substitutions are to be considered as included in this invention.

This is particularly relevant to the control of RF devices, since frequency- and phase- stability are generally one of the most critical parameters of such devices. Making such an RF device tuneable in some way (as we here describe using SMA actuators) in no way diminishes that requirement for stability, of which ever parameter(s) has been tuned. Where specific descriptions below refer to the actual mechanics of the SMA wire itself (e.g. when describing using twisted-pairs of SMA wire, or using insulated SMA wire) then they can additionally be read as referring to the SMA wire(s) within an SMA-stepper-actuator where that option is being considered as the specific form of (generic) SMA wire actuator.

Algorithms and/or look-up-tables (LUTs) in a controller determine how each tuneable element and its associated actuator needs to be moved or deformed to achieve the desired filter performance and then further algorithms are used to separately control all of the actuators in parallel to provide optimum performance. By having the same controller monitor the temperature inside the tuneable cavities and modulate the actuator control algorithms accordingly, the control system is also able to

temperature-compensate the whole filter against environmental conditions as well as heat build up within the system.

Within the tuneable filter are one or more components or elements whose properties may be changed by moving or deforming them, or both. These are generically designated as tuneable elements hereafter. A tuneable element may be a component specifically introduced solely to tune another electromagnetic component (such as a resonator, or a coupling device or a tapping point), or may be one of these electromagnetic components themselves in the case that the electromagnetic properties of that component may be changed by an external effect (e.g. mechanical force, which might be used to move or deform the electromagnetic component).

The SMA actuators themselves can be small and can be very low-cost, may require no lubricants, are highly reliable, are frequently silent and generate no significant magnetic fields. Such an SMA actuator may be mounted outside an RF cavity and control its respective tuneable element via a mechanical connection to the tuneable element inside the cavity; alternatively the tuneable element itself may be extended to protrude through the cavity to the outside and become an integral part of the external actuator.

Because the SMA wires are so small - typically only 25microns (25im) in diameter for the roles envisaged herein - the SMA actuators may advantageously be positioned right inside the RF cavities or waveguides if suitable precautions are taken, and the SMA actuator can then be designed to be mechanically integral with the tuneable element to be controlled. To put this in perspective, the total volume of an SMA wire actuator made from even a 20mm length of 25im wire is less than 1/100th of a cubic millimetre, i.e. < 0.01mm3 and yet such an actuator can pull a ~10gm to15gm load a distance or stroke of 0.75mm or more, which stroke can be greatly increased by use of suitable mechanical leverage techniques such as for example those described for the actuators above, which in some forms may reduce the force by the same ratio. Such tiny, relatively high-force, actuators positioned within the RF cavities can be highly beneficial in terms of reducing overall filter size, reducing the RF leakage otherwise encouraged by moving mechanical devices penetrating the walls of the RF cavities, reducing the amount of structure necessary to support and house the actuators (i.e. no additional housing at all is needed if all of the actuators are sited within the RF cavities) as well as reduce the mechanical loads the SMA actuators are required to move, because of the elimination of coupling mechanics and in many cases sliding supports and additional friction as well. No other form of actuator is as versatile or well suited to this tuning task as small SMA actuators, because the high-force small-volume very high reliability characteristics of SMA actuators is unmatched. MEMS and electrostatic actuators fail on one or more of stroke, force and reliability. Other electric motors are simply too big, too expensive, too unreliable or some combination of these to be usefully fitted within the

waveguides or cavities.

Further innovations made possible by this use of small SMA actuators especially when closely integrated with the filter structures themselves are the use of thin glass dielectric wafers and corundum (AI203 crystalline aluminium oxide) dielectric layers grown directly onto aluminium components in novel ways particularly to increase the tuning effect of the tuning elements, thin blade-shaped or ribbon-shaped tuning elements instead of cylindrical tuning screws and plungers, integral RF chokes positioned along the lengths of conductive tuning elements where they do protrude through cavity walls, and the use of shape-changing of RF elements. Where the material of a tuning element within the cavity either extends through the cavity wall or is mechanically connected to the outside of the cavity by some connective structure this will in general produce a leak-path for RF energy to escape from the cavity. By suitably shaping this structure to produce capacitive and inductive sections an integral RF choke can be created that will completely eliminate such RF leakage. In conventional tuneable filters the resonators and couplings are mechanically fixed shaped components with galvanic contacts implemented for electrically connecting the tuning pins, and generally only the positions or rather the lengths of the tuning elements, if anything, changes. The provision of small internal (to the cavities) SMA actuators or actuator wires allows the resonator and tuning element components themselves to be mechanically deformed by the contraction and expansion of SMA wires when suitably arranged and attached. Thus, a resonator, e.g. that may conventionally have been structured as a rigid rod of metal, may now be structured as a thin conductive strip (but still considerably thicker than RF skin-depth and therefore of similarly low loss) which is easily bent by the pull of an SMA wire or SMA actuator mechanically attached at one end to some point on the resonator and to a ground plane (or in some instances to a different point on the same component) at the other. Because in the active region of the SMA wire it contracts with increased heating, if the heating in the wire is induced by an electrical current in the wire and the metal of the resonator is used as the return path for the current, then the resonator made of normal metal (i.e. not SMA) will expand slightly with increasing current, thus increasing the overall bending effect of the resonator/SMA-wire combination.

So in the present invention an RF filter has one or a plurality of stages, and the filter is either: i) tuneable between upper and lower frequency limits and is a low-pass, band-pass, band-stop or high-pass configuration, or ii) produces a tuneable phase shift, or iii) both of these. The RF filter adjacent stages are electromagnetically coupled. The filter is either constructed of two or more spaced preferably parallel conductive ground planes (though non-parallel ground planes can also be used) with solid conductive joining walls connecting between the conductive planes and/or with conductive vias positioned between the conductive planes, and has inside and between the ground planes one or a plurality of separate cavities separated by either solid conductive partitions or by walls formed by a plurality of conductive vias positioned between the conductive planes or by both, and in each of those cavities is zero, one or a plurality of resonators. The resonators may be either conductive or dielectric or some combination of these. Each of any conductive resonators in a cavity may be either connected to ground at one end only or connected to ground at each end or connected to ground in the middle or connected to ground at one or more half wavelength intervals or not be connected to ground at all resulting in a floating conductive resonator.

One or more of the cavities has each either one or more mechanically moveable or deformable resonators, or one or more deformable or movably mounted tuning elements penetrating into or wholly contained within the cavity such that the deformation or movement of the tuning element changes the electromagnetic characteristics of the cavity. The moveable or deformable resonators are caused to move or deform each by an associated SMA-wire actuator. The tuning element or elements are caused to deform or move each by an associated SMA-wire actuator.

In the case that the conductive ground planes are electrically connected by conductive vias then the spacing between the vias is chosen to produce the required degree of coupling including zero coupling between any cavities separated by the vias or between any cavities and the outside of the filter delineated by the vias. Each of the conductive vias between the ground planes may be either connected to a ground plane at each end or connected to one ground plane at one end only resulting in a blind via with one end left open-circuit or be connected to no ground plane at either end resulting in a buried floating via (with both ends left open-circuit).

A tuneable cavity either contains one or more tuneable resonators thereby giving that cavity a degree of tuneable frequency selectivity, or instead, a tuneable cavity contains no resonator in which case it can have very high Q and be very frequency selective. In the case where there is no resonator in a cavity it may advantageously be used as a tuneable phase-shifter by incorporating one or more tuneable reflector elements instead, which may be tuned by movement or deformation thereof.

In a multistage tuneable filter where each stage is implemented by one or more tuneable cavities there is generally one primary path through the filter stages in a certain sequence from the input port to the output port, although in more complex filters there may also be secondary signal paths as well. Stages or cavities which are sequential along that primary signal path are referred to here as path-adjacent stages or path-adjacent cavities, to distinguish them from merely physically-adjacent (here called adjacent) stages or cavities. Path-adjacent cavities are coupled to each other by couplings where the strength of each coupling can be between zero (no coupling) and one (fully coupled). In addition to path-adjacent cavity couplings, there may also be additional couplings between adjacent but not path-adjacent cavities and such couplings are also referred to as couplings. These non-path- adjacent cavity couplings may be used to implement more complex filters with more desirable filter characteristics. Such techniques are well explained in several well established theoretical works, for example:

• George L. Matthaei, L. Young, E.M.T. Jones, "Microwave Filters, Impedance matching Networks and Coupling Structures";

• Richard J. Cameron, Raafat Mansour, Chandra M. Kudsia, "Microwave Filters for Communication Systems : Fundamentals, Design and Applications".

Where two coupled cavities are separated by a solid conductive partition then the coupling is formed by one or more appropriate sized apertures cutting through that conductive partition and the coupling response is related to the cavity separation distance and size and shape of the coupling aperture(s). Where two coupled cavities are separated by a plurality of conductive vias in the region between the cavities to be coupled, then the coupling is formed by strategically locating gaps and the sizes of gaps between these vias and the number, position and dimension of the vias around the gaps define the amount of coupling between adjacent cavities.

Alternatively, the coupling is formed by providing additional non-grounded conductive tracks printed on an insulating layer formed on the inside of one or both of the ground planes or cavity walls sandwiching the cavities to be coupled, and the conductive tracks protrude into both of the adjacent cavities without electrically connection to anything else. Alternatively, the coupling is formed by non-grounded cross-coupling wires protruding into both of the adjacent cavities without electrically connection to anything else. Both of these couplings are capacitive in nature. If instead the tracks or wires are grounded at both ends then such a coupling becomes inductive instead coupling may also be achieved by some combination of these coupling techniques.

The resonators are made of conductive material or are made of low-loss dielectric material or are made of non-conductive material coated or plated with conductive material or are made of some combination of these. The dielectric resonators are preferably made of high permittivity low loss RF ceramic. One or more of the resonators are in the form of strips or T-shaped strips or rings or spirals or crosses or other shapes that resonate at the required frequency. In one variant of the present invention the cross-section of the resonator is such that it is easy to bend or flex elastically in at least one dimension, e.g. a thin strip of metal or metallised plastic or laminate. Where a resonator has a geometry with several eigenmodes (e.g. X- shaped or star-shaped) then concurrent modes in the resonator may be suppressed by shorting to ground the corresponding ends or points of the branches of the resonator structure. A more compact assembly can be achieved with a dual-mode resonator or a triple-mode resonator with a minimum of two or three mutually orthogonal branches with a single common point. Such an orthogonal three branch resonator necessarily has a 3D configuration and in this case at least one of the three branches may protrude through one of the ground planes if their spacing is too close to fully contain the 3D resonator. Alternatively multi-mode waveguide cavities can also be employed where orthogonality of modes is enforced by the boundary conditions on the walls and the degree of symmetry in the cavity.

External signal connections are provided in the form of input and output tapping points to the first (input) and the last (output) cavity; where there is only one cavity, then it becomes both the input (first) and the output (last) cavity, i.e. the first is the same as the last. Where there is a resonator in the cavity being tapped then the tapping point is preferably adjacent to that resonator; where there is no resonator in the cavity being tapped (e.g. if this is a phase- shifter cavity) then the tapping point is into the cavity itself, e.g. a waveguide port.

So a tuneable filter according to the present invention has an input port, an input tapping point, one or more cavities at least one of which is tuneable, and where more than one cavity, then one or more couplings, zero or more resonators, an output tapping point, and an output port, which may be the same port as the input port.

At least one electromagnetic element (input tap, cavity, resonator, coupling, or output tap) of the tuneable filter is tuneable, either by physically changing its shape (by deformation), or by moving it relative to the filter body i.e. the ground planes, or by the provision of an adjacent changeable tuning element. A tuning element operates to tune its associated electromagnetic element by moving relative to it, and it may do this either by whole-body movement of the tuning element or by deformation of the tuning element such that the portion of the tuning element close to the

electromagnetic element moves relative to it or relative to a ground plane, or to both.

Where an electromagnetic element or tuning element operates to tune by virtue of deformation, then that element may advantageously be contained wholly within the cavity that it tunes, thus avoiding the need to provide for moving structures to penetrate the cavity walls.

Where a tuning element operates to tune by virtue of movement relative to an electromagnetic element (input tap, cavity, resonator, coupling, or output tap) of the tuneable filter, then that tuning element is movably mounted and either penetrates into the cavity being tuned or is wholly contained within that cavity.

In both the case of deforming elements and the case of moveably mounted tuning elements, the tuning effect comes about because the movement changes the capacitive loading or inductive loading or both of the associated electromagnetic element.

The effectiveness of moving or deforming a tuning element to tune its associated electromagnetic element may advantageously be enhanced by interspersing a thin dielectric material into the space between the tuning element and the

electromagnetic element to be tuned. A thin glass wafer is a suitable dielectric material in a suitable form for this function. However, such glass wafer thicknesses are limited to about 300um or greater because of manufacturing and handling issues. An alternative and much superior effect can be obtained by making from aluminium, either the conductive part of the element to be tuned (or at least the portion of it adjacent to the tuning element) or instead, the tuning element itself, and growing a thin smooth crystalline layer of corundum (crystalline AI203) on the aluminium surface between the tuning element and element to be tuned. Such layers may have useful thickness in the range 1 to 30um (thicker layers are possible but become increasingly difficult to maintain the quality of the dielectric), have typical dielectric breakdown voltage of >16KV/mm, high dielectric constant of ~9.8 (@1 MHz), and very low RF loss with dissipation factor as low as 0.0002. If the aluminium surface to be oxidised is first polished then the resulting corundum coating grown upon it also has an external surface of more or less the same surface finish (i.e. almost polished) which allows intimate and low friction contact between it and the adjacent element.

Similarly when an electromagnetic element is self-tuned by deforming it the effectiveness of such tuning may be enhanced by the suitable insertion of a dielectric element, such as a glass wafer or corundum coating on one of the adjacent surfaces, between the moving part of the deformed element and an adjacent ground plane.

Depending on the performance requirements of the tuneable filter, any, some, or all of the electromagnetic elements of the filter may be provided with such a deforming or movably mounted tuning element, or alternatively may themselves be constructed so as to be easily deformable or moveable and be tuned by virtue of such

deformations or movement of the electromagnetic element.

Each tuning element may have the shape of a flat strip, or a rod, or a bar, or a tube, or more generally a long prismatic section with flat or curved or corrugated surfaces, but preferably a shape such as a thin flat strip is used to reduce the mass and volume of the tuning element, making it both easier to move with a small SMA wire actuator, and easier to fit inside a compact filter structure with multiple tuning elements.

The tuning elements are made of conductive material or are made of low-loss dielectric material or are made of non-conductive material coated or plated with conductive material or are made of some combination of these. Dielectric tuning elements are preferably made of high permittivity low loss RF ceramic or alternatively are made from a glass wafer.

A tuning element is preferably aligned in the same direction of greatest extension as the resonator it is tuning so that the gap between the tuning element and the resonator is also aligned with the resonator. For example, where a resonator is in the form of a rectangular section prismatic bar with a greater width than thickness, then the associated tuning element would preferably be in the form of a thin strip having a width similar to the width of the resonator, be positioned close to and parallel to the wide face of the resonator, and have its long axis aligned parallel to the long axis of the resonator, so that progressive movement of the tuning element in this axial direction would cause progressively increasing or decreasing overlap of the resonator by the tuning element and thus increasing or decreasing capacitance. This affords a large range of tuneability with nearly linear tuning characteristics.

A resonator may have a longitudinal slot or slots in it into which a tuning element may fit without touching the resonator, to increase the variability of capacitance between the tuning pin and resonator. A resonator may have a longitudinal slot or slots in it into which a tuning element may fit without touching the resonator, to increase the variability of self-inductance per unit of length of the resonator.

Each tuning element may either be entirely contained within the RF cavity, or alternatively may extend through an aperture in the cavity wall and even further to the outside region beyond. In this context we mean by RF choke an additional auxiliary fixed low-pass filter with a stop-band covering at least the entire operating bandwidth over the full tuning range of the main tuneable filter.

The one or more tuneable elements of the tuneable filter of the present invention are caused to move or deform (shape change) by one or more actuators (defined below), with one or more of the tuneable elements sharing an actuator, so that the number of actuators can vary from one, where all of the tuneable elements are moved by the same actuator, up to the number N of tuneable elements where each tuneable element is driven independently of all of the others by its own actuator. Preferably there is a separate actuator provided to independently control each tuneable element, so that there are N Actuators.

An actuator is herein defined to be an SMA actuator, with the SMA material in the form of a thin wire, strip, or sheet. The most preferable forms of SMA actuator are: i) any of the forms of SMA-Stepper-actuator described above; and ii) an SMA-wire actuator where the length of one or more sections of SMA wire are caused controllably to change by controllably changing the SMA wire temperature(s). The temperature of an SMA wire may advantageously be changed by controlling the magnitude of electric current passing through the SMA wire. This electric current in turn is preferably under the control of a programmable device such as a

microprocessor. The length-changing SMA elements (e.g. wire or wires, strips or sheets, or the body and output element of a SMA- stepper-actuator) are then mechanically connected either directly or indirectly between the filter body and the moveable or deformable tuneable elements of the tuneable filter, which causes the tuneable elements to move relative to the filter body or to change their shape (deform). For a deformable tuneable element the length-changing SMA element(s) may instead be mechanically connected between two (or more) separate points on the deformable tuneable element itself, which causes the tuneable element to change its shape (i.e. deform).

The mechanical linkage of a tuneable element to its respective actuator may be direct and immediate, in which case there is no independent mechanical coupling component between the tuneable element and the actuator. In a fully integrated actuator, part of the tuneable element itself may be used as part of the actuator structure and there then will be no discernibly separate actuator and tuneable element, but instead a single component with moveable or deformable parts capable of changing the electromagnetic environment within a cavity or between cavities. So the tuneable element may optionally and preferably be integral with the actuator structure.

Each actuator connected to one or more of its respective tuneable elements may be positioned outside of the RF cavity or cavities of its tuneable element(s), or may instead be positioned within the walls of the RF cavity or cavities, or instead be placed partially or wholly within the RF cavity or cavities.

Where any tuneable element is made of a dielectric material and its associated actuator is not wholly separated from the inside of a respective cavity or cavities by the solid conductive wall of the cavity(s) it may be RF electrically isolated by the positioning suitably close to the tuneable element of one or more conductive vias connecting between the conductive walls of the cavity(s). In particular, straddling the tuning element with conductive vias spaced closer than a half wavelength of the highest operational frequency of the filter in this way, will eliminate RF leakage via the dielectric element.

Where any tuneable element is made of a conductive material and protrudes through a cavity wall then to prevent TEM mode propagation along the tuning element of RF energy from within the cavity to the outside of the cavity and towards its associated actuator, two or more buried vias are located adjacent to and along the longitudinal line of the tuning element and separated by the appropriate interval which is approximately a half-wavelength but corrected for the reactance introduced by the adjacent vias. This is sufficient for the propagation at this wavelength to be blocked by capacitively loading the leaking TEM mode, and will completely stop the leakage.

In the case that more than one actuator in total controls the movement of the totality of tuneable elements, then the synchronisation of the movements of all of the tuneable elements is electrically controlled by the synchronisation of the appropriate control signals to the plurality of actuators, for example by means of a pre-computed look-up table kept in the memory of the controller or by a real-time algorithm generating the actual required positions of all tuneable elements to achieve the required state of the filter.

If the tuning elements are each movably supported by a tuning support structure (Support) which may be partially or fully dielectric or partially or fully conductive, then each tuning element is associated with at least one SMA wire, and is mechanically connected directly or indirectly to that wire such that changes in length of the SMA wire caused by heating and cooling of the SMA wire cause changes in position or shape of the tuning element. In this case each SMA wire may be enclosed within a dedicated void in one of the one or more supports to ensure free movement of the SMA wire relative to the support. Each tuning element is then positioned slidably in a channel through the support to ensure free movement of the tuning element while maintaining a precise gap between and accurate distance from the tuning element to the corresponding resonator, coupling or tapping point for all positions of the tuning element controlled by the actuator. Each such actuator may be fully integrated into the filter, for example by being buried inside the support. A resonator is preferably mechanically fixed relative to the ground planes and is not moveable. In an alternative variant of the present invention a resonator may be movably mounted within a cavity and caused to so move by mechanical connection to an actuator in which case the resonator becomes itself a tuneable element. In a further alternative variant, a resonator may be constructed so as to be easily mechanically deformable and is caused to so deform by mechanical connection to an actuator which then applies stress to the resonator in which case the resonator becomes itself a tuneable element. Viable deformable resonator forms include thin strips, flat-section spirals, flat-section helices, bellows and other shapes which have at least one direction of easy (low force) deformation.

Optionally there may be more than two parallel or near-parallel ground planes and in this case there may be openings in any ground plane that separates at least two other ground planes to form either inductive and/or capacitive cross-couplings between physically adjacent cavities on either side of that separator ground plane. These coupled cavities may contain one or more resonators tuned by one or more tuning elements actuated by actuators, and such cross-couplings themselves may also preferably be tuned by one or more tuning elements each actuated (to move or deform) by an actuator.

A filter so constructed within three or more ground planes is called a 3-D folded filter. In such a folded filter where there is only one cross-coupling (i.e. through a ground plane) then the filter topology is no different to non-folded filters, and topologies are limited to Chebyshev filters (defined by a diagonal coupling matrix). A 3D folded filter with more than one cross-coupling (i.e. through a ground plane) allows more complex filter topologies to be realised including elliptical filters, extracted pole filters and other types, characterised by a generalised coupling matrix with non-diagonal non-zero elements.

In conjunction with the use of multiple vias to separate and connect the ground planes more complicated tuneable filter topologies become possible while still being compatible with SMA actuated tuning elements, as the filter may now be folded in 3D. This in turn allows for tuneable couplings between non-path-adjacent cavities which may now become physically- adjacent in 3D, which enables the realisation of for example fully tuneable elliptical filters.

Such topologies make the manufacture of these structures more compatible with 3D printing and with 2.5D processes such as PCB and microstrip technology and wafer level integration potentially allowing fully printed designs for dramatic cost saving. Such topologies may also be manufactured by CNC machining in solid metal when high power handling and low PIM requirements dictate.

Using multiple stacked ground planes and multiply folded filters with several levels of cross-coupling may advantageously be used to modify the overall volume of a filter with given performance characteristic, and because of the greater flexibility in defining the elements of the coupling matrix can better optimise performance parameters such as group delay equalization.

Where at least the portion of a tuning element outside of a cavity is non-conducting (e.g. dielectric, or insulator not coated or plated with conductive materials) then the at least one SMA wires that change the position of that tuning element may be attached directly to it. For tuning elements with only conductive portions outside of a cavity, then the at least one SMA wires that change the position of that tuning element may be attached to it via an electrically insulating structure, e.g. plastic or ceramic, to electrically isolate the SMA wire heating current from the tuning element.

Where the section of a tuning element outside a cavity (the tail) is conductive it may be RF isolated from the RF energy within the cavity by: an intrinsic RF choke formed along the length of the tuning element wherein the longitudinal extent of the element provides a series inductance and the proximity of the tail to adjacent ground planes provides a parallel capacitance; and/or an at least 2-section RF choke, similarly formed as in a) by shaping the profile of the tuning element in the tail region so that it comprises successive wider and narrower sections electrically in series along the length of the element, the wider sections being predominantly capacitive and low impedance, the narrower sections being predominantly inductive and high

impedance. With this arrangement advantageously, the SMA wire or wires provided to cause motion of the tuning element, may be attached to the low-impedance capacitive sections of the so-shaped tail so as to maximally isolate them from any RF energy transmitted from within the cavity; such series

inductive/capacitive/inductive/capacitive series sections effectively form a multi section RF choke or low pass filter.

Where the tuneable filter is constructed between two or more ground planes, then the conductive ground planes are preferably parallel.

An actuator of the tuneable filter may advantageously be positioned partly or wholly within the cavity containing the tuneable element that the Actuator serves to move or deform. In this case where the actuator is at least partly inside the RF cavity it is necessary to minimise the RF field coupling to the conductive SMA material (wires , strips or sheets) of the actuator. In order to prevent the SMA material from coupling strongly to the RF field in the cavity one or more of the following approaches may be taken: for a straight-wire strip or sheet SMA actuator where the SMA material is very thin, e.g. SMA material thickness « (electric wavelength of cavity), then the SMA material may be located entirely on or within the electric wall of the cavity (or simply parallel to the electric wall for mode TE10), or alternatively, the line or plane of the SMA material should be positioned orthogonal to and symmetrical to the magnetic walls of the cavity; where the SMA material or the SMA material and its conductive electrical connections (e.g. wires to an external SMA-actuator controller) deviate from a straight line within the cavity then they should be constrained to lie in a plane and that plane positioned orthogonal to and symmetrical to the magnetic walls of the cavity; further isolation of the SMA material and connecting wires thereto from the RF field may be achieved by sandwiching the planar arrangement of SMA material and connecting wires between two thin low RF-loss glass wafers held parallel to the electric field of the cavity, which effectively“suck-in” the surrounding RF field greatly reducing its amplitude in the vicinity of the actuator structure. These glass wafers may be arranged not to touch the static or moving parts of the actuator, but also may advantageously be used as support elements for the actuator structure and even as the primary mechanical static portions of the actuator; the SMA element(s) of the actuator may also be electrically screened from the RF fields by partially or wholly surrounding them with conductive surfaces preferably metal or metallised plastic.

A tuneable filter that is to act as a phase-shifter has a nominally constant amplitude and linear phase response across its passband, and when tuned it is predominantly the magnitude of phase-shift at each frequency within the passband that changes, not the centre frequency or edges of the passband. Such a tuneable phase-shifter filter consists of one cavity or a plurality of coupled cavities each containing one or more movable elements.

In its simplest form a tuneable phase-shifter filter consists of a cavity in the form of a section of waveguide with conductive walls, open at one end and preferably partly or fully closed at the other, the open end serving as both the input-port and the output port, i.e. the I/O port. Within the cavity a moveable or deformable element, the tuning element, is constrained to at least in part move along the waveguide in a direction towards and away from the I/O port, with the movement or deformation caused by an actuator, situated within the cavity or external to it, as described above for general tuneable filters of the present invention. The tuning element itself forms an electromagnetic discontinuity in the waveguide which thus reflects some of the RF energy back to the I/O port. Ideally the tuning element reflects all of the incident RF energy and at least a portion of it may preferably take the form of a plane conductive sheet or plate or conductive plated surface of a plane insulator almost filling the cross-section of the waveguide but preferably without touching it, and preferably without making electrical contact with the electrically conductive waveguide walls, movably supported so as to allow it to travel along the direction of the waveguide towards and away from the I/O port.

Alternatively the tuning element may consist of a shaped resonator with a broad resonance across the passband of the phase shifter, made out of metal or a printed conductive pattern on a dielectric substrate. In operation RF energy propagates down the waveguide (e.g. in TE10 eigenmode) from the input port, is reflected back from the moveable element, and then exits the waveguide at the output port, the phase of the output wave relative to the input wave being directly proportional to twice the length of waveguide extending from the I/O port to the current position of the moveable tuning element. Because of the reflective nature of this tunable phase- shifter configuration it has the advantageous property that the amount of phase difference dphi produced by moving the moveable element a distance x is:

dphi = 4pi.x/L radians

where L is the wavelength in the medium of propagation within the waveguide of the wave being phase shifted. The extra factor of 2 achieved by the reflection (rather than pass-through) of the wave thus requires only half of the movement of the moveable element otherwise needed for the same amount of phase shift change, and this can simplify and/or lower the cost of the actuator provided to move it. The RF field behind the moveable tuning element (i.e. on the opposite side of it to the I/O port) can be made small by suitable design of the tuning element (primarily by making it a highly efficient reflector) and this low RF field makes it relatively unproblematic to site the actuator directly behind the moveable tuning element and within the cavity. In this case it is still preferable to follow the rules set out above for minimising interaction between in-cavity actuators and RF field. Alternatively the actuator may be sited within the thickness of or outside of the cavity wall opposite the I/O port (or indeed outside any of the adjacent side walls) when again the low RF field within this portion of the cavity minimises RF leakage issues around any moving parts passing through the cavity walls, which can be reduced further by the introduction of in-line RF chokes again as described above. The phase-shifter as described above advantageously can be modified to reduce the required movement of the moveable element to achieve a given phase shift, by partially filling the cavity with dielectric. For example by placing glass wafers or other high dielectric constant material on the inside walls of the cavity it is possible reduce the effective wavelength of propagation within the cavity, whereupon a given change in phase shift is produced with a reduced movement of the moveable tuning element.

In another aspect of the invention the tunable phase shifter filter comprises not one but two co-moving tuning elements separated by a distance m along the waveguide, where m is approximately half a wavelength of propagation in the cavity at the mid range frequency of the phase shifter; the optimal separation m differs from an exact half wavelength due to the reactive conductance of the tuning elements, as well as the other structures within the cavity, primarily the actuator or actuator coupling mechanics that link an external actuator to the moveable member. The beneficial effect of the second tuning element is an increase in bandwidth of the tuneable phase-shifter and an increase in reflectivity achieved and a reduction of RF losses. The two tuning elements may be mechanically joined by a stiff strut attached between them, which is preferably made of low loss dielectric, e.g. a glass wafer. In a preferred embodiment of this aspect of the invention, the actuator used to move the moveable elements is sited inside the cavity in the gap between them, and as described above the conducting SMA material of and to the actuator are held in the electric wall in the cavity to minimise coupling to the RF field. The two tuning elements may be identical or be made to differ so that they introduce reactances of different sign, thus providing the possibility to control the dispersion of the phase shift, or otherwise the linearity of the introduced time delay over frequency within the passband, or alternatively, achieving more compact design with reduced distance between the tuning elements.

Where a pair of tuning elements is used in any of these phase-shifter variants, then they may advantageously be mounted sandwiched between glass wafers with elements etched on both faces of the copper-plated glass wafer providing stable and well defined electrical distance between them and at the same time achieving economical design.

The tuning elements in each and all of the above aspects of the tuneable phase shifter can take the form of conductive rectangles, conductive squares, conductive rings or conductive crosses, each of which has its own advantages and

shortcomings. Selection of the optimal configuration of the tuning element is linked to the operating / dominant mode of the phase shifter. The element that provides the maximum coupling to the operating mode combined with the lowest achievable insertion loss should be selected. Narrow band phase shifters may employ resonant elements while more wideband devices will benefit from using non- resonant elements providing only capacitive or only inductive response.

A further aspect of the invention is a tuneable phase-shifter constructed as a waveguide cavity as described in all variants of two tuning element phase-shifter above with the difference that now both ends of the waveguide are open (i.e. there is now no closed end). In this aspect one end of the waveguide cavity acts as the input- port and the other end acts as the output port.

The reflections back to the input-port of the two tuning elements are now arranged to cancel each other at the input-port to minimize the in-band return loss. The degree of coupling between the tuning elements will define the width of the passband, and consequently the amount of phase-shift in-band. Moving them simultaneously along the waveguide will not be useful in this configuration. Instead an actuator is used to controllably change the distance between them which controllably changes their coupling, and each moveable element will be independently tuned by an actuator (one per moveable element) to keep the elements tuned to the same central frequency. Thus this configuration will require at least three actuators. In this through- waveguide phase-shifter configuration (as opposed to reflective waveguide phase- shifter configuration) the two tuning elements are optimally separated by a distance m along the waveguide, where m is approximately one quarter of a wavelength of propagation in the cavity at the mid-range frequency of the phase shifter; the optimal separation m differs from an exact quarter wavelength due to the reactive

conductance of the tuning elements, as well as the other structures within the cavity, primarily the actuator or actuator coupling mechanics that link an external actuator to the moveable member.

A phase shifter can be formed by two orthogonal transmission lines (Line V and Line H) supporting waves of orthogonal polarisations propagating in a direction P. The two lines may each for example be formed by pairs of spaced parallel conductors. A sliding plate placed orthogonal to the direction of propagation P and within the space between the four conductors forming the transmission lines contains a resonating structure formed by a metal structure layout on the plate surface resonating at the frequency of operation, to facilitate the reflection of an incoming wave. These metal structures are designed to resonate at the operating frequency; for example - conductive strips forming dipoles may be used, one dipole in each of the two orthogonal directions. The plate thus contains two types of structures - each designed to interact with the wave of corresponding polarization. This configuration of phase shifter provides dual polarized operation with identical phase shift introduced for Vertical and Horizontal polarizations supported by Lines V and H respectively.

A phase shifter for independent control of the two orthogonal polarizations is formed by two orthogonal transmission lines (Line V and Line H) supporting waves of orthogonal polarisations propagating in a direction P. Two plates orthogonal to each other and lying parallel to the direction of propagation P are placed between the two transmission lines. One plate is orthogonal to the walls of line V and the other orthogonal to line H. A resonating structure designed to interact with one of the waves of each polarization is placed on each plate, one for each polarisation. One structure on one plate interacts with the wave supported by Line V, and the other structure on the other plate interacts with the wave supported by Line H. Each plate has independent freedom of movement in the direction of propagation P, and can be moved independently of the other with the help of dedicated slot in one plate that allows the other plate to move within it (in the slot) without mechanical interaction. A slot can also be placed in both pates to achieve the same end. Each plate is attached to a separate actuator capable of moving the plate far enough to achieve the desired range of phase-shift of the waves.

So in one aspect of the present invention a tuneable RF filter comprises one or more resonant or reflective elements positioned in a waveguide with conductive walls and the one or more resonant or reflective elements are caused to move axially along the waveguide each by an actuator. In a particular embodiment the waveguide conductive walls are formed from alternate metal and dielectric layers with adjacent metal layers joined together by rectangular arrays of conductive vias through the dielectric layers the rectangular arrays forming the walls of the waveguide whose axis is orthogonal to the metal and dielectric layers, and the waveguide cavity is formed by the removal of the dielectric and metal layers within and between the waveguide walls. In a different embodiment the waveguide conductive walls are constructed of conductive metal or conductively coated insulating material such as polymer, by, for example, metallising the polymer. In a preferred embodiment the one or more resonant or reflective elements are constructed so as to reflect as perfectly as practically possible all of the RF energy incident at one end of the waveguide back to that same end of the waveguide with a phase directly proportional to the axial position of the moveable elements along the waveguide thus providing a single-port reflective tuneable phase-shifting filter. Preferably in this embodiment the length of the waveguide is at least half the wavelength within the waveguide of the waves of interest so that the reflected wave may be delayed by any phase angle between 0 and 360deg, allowing, for example, the construction of a phased-array antenna with an array of such phase-shifters. In another preferred embodiment, the length of the waveguide is a multiple (greater than 0.5, and quite possibly >5 or >10 or even more)) of the wavelength within the waveguide of the waves of interest, so that the reflected wave may be delayed by a time anywhere between 0 and the time taken to traverse the waveguide in both directions, which may advantageously be many cycle- times of the wave, allowing, for example the construction of a true-time-delay-array antenna, with greater bandwidth than an otherwise similar phased-array antenna. In another preferred embodiment two resonant or reflective elements are constructed so as to reflect as little as practically possible of the RF energy incident at one end of the waveguide back to that same end of the waveguide such that nearly all of the RF energy emerges from the other end of the waveguide with a phase directly proportional to the axial positions of the moveable elements along the waveguide and wherein a second actuator is used to control the axial separation of the two resonant or reflective elements to optimise the input return loss with operating frequency thus providing a dual-port tuneable phase-shifting filter. In a further aspect of this invention two separate sets of one or more resonant elements are positioned in the waveguide, each set independently of the other moveable axially along the waveguide by independently controllable actuators, and each set of resonant elements is responsive to only one of two different polarisations of waves incident on one end of the waveguide, for plane polarisation waves the different polarisations being orthogonal to each other, and for circular polarisation the different polarisations being of opposite sign.

Most generally:

In one aspect of the invention a phase or frequency tuneable device (hereinafter Device A) comprises an RF cavity exploiting the thermo-mechanical properties of SMA material in the shape of wires or ribbons or sheets so arranged to form an actuator applied in such a way as to achieve controllable deformation or controllable movement of the walls of the RF cavity, or controllable movement or controllable deformation of additional electromagnetic structures in the vicinity of or inside the RF cavity, so as to affect the electric and/or magnetic components of one or several of the eigenmodes of electromagnetic field supported by the RF cavity. In a further aspect of the invention an RF tuneable filter device (hereinafter Device B) comprises two or more phase or frequency tuneable devices A as just described wherein these two or more devices A are each electromagnetically coupled to at least one other of the plurality of such devices.

In a further aspect of the invention an RF tuneable filter device B as just described has at least one of the electromagnetic couplings between devices A in the form of an iris penetrating the solid walls or ground planes separating the phase or frequency tuneable devices or by an iris formed by a gap in a wall of conductive vias separating those devices.

In a further aspect of the invention an RF tuneable filter device B has at least one of the electromagnetic couplings between devices A formed by the provision of additional non- grounded conductive tracks formed (e.g. printed) on an insulating layer itself formed on the inside or outside of one or both of the ground planes sandwiching the cavities to be coupled, and wherein the conductive tracks protrude into both of the adjacent cavities of the devices A either without electrical connection to anything else or with both ends grounded.

In a further aspect of the invention an RF tuneable filter device B has at least one of the electromagnetic couplings between devices A formed by non-grounded cross coupling wires protruding into both of the cavities of the adjacent devices A through an iris either without electrical connection to anything else or with both ends grounded.

In a further aspect of the invention an RF tuneable filter device B of any of the variants described above has at least one of the electromagnetic couplings between RF cavities of the devices A tuneable by a tuning device comprising SMA material in the shape of wires or ribbons or sheets applied in such a way as to achieve controllable deformation or controllable movement of a conductive or dielectric tuning element in the vicinity of the electromagnetic coupling.

In a further aspect of the invention an RF tuneable filter device as in any of device A or device B variants described above comprises one or a plurality of stages, the filter either being of the low-pass, band-pass, band-stop, high-pass or phase-shifting configuration and further comprising two or more spaced conductive ground planes with joining walls connecting between the conductive planes and/or conductive vias positioned between the conductive ground planes, having inside between the ground planes one or a plurality of separate RF cavities separated by solid conductive partitions and/or by a plurality of conductive vias positioned between the conductive planes and in each of those cavities is zero, one or a plurality of resonators or electromagnetic reflectors, and where there is a plurality of cavities each cavity is electromagnetically coupled to at least one other cavity by an iris penetrating the solid walls or ground planes or by an iris formed by a gap in a wall of conductive vias between ground planes, and wherein one or more of the RF cavities has each one or more tuning elements penetrating into or wholly contained within the RF cavity and wherein each such tuning element is either wholly moveable or is deformable in such a way that the movement or deformation thereof changes the electromagnetic characteristics of the RF cavity so as to satisfy the tuneability requirement of the filter and wherein the movement or deformation of at least one of the tuning elements is caused by the expansion and contraction of one or more associated SMA structures each under the heating influence of a controlled electric current passing through said SMA structure and where each SMA structure is located outside of the RF cavity or within the walls of the RF cavity or located wholly within the RF cavity.

In all of the SMA actuated RF tuning and phase-shift devices described herein, there is a possibility of direct and unwanted interaction between the RF waves themselves and the SMA elements of the actuator which are by their nature electrical conductors. In receiver applications this is unlikely to cause any problems for the SMA elements, although they may act as unwanted absorbers of RF signals. However in

transmission devices where the power levels can be high and the electric fields intense, it may become important to prevent unwanted heating of the SMA elements by the RF energy within the RF device being controlled, and this especially so when the actuator is advantageously placed within the RF cavity. A convenient and effective way to minimise RF/SMA interaction, when the SMA element is in the form of thin wires, is to firstly, use insulated SMA wire, and secondly to use the insulated SMA wire in the form of tightly twisted-pairs as are familiar to those practised in the electronics signal processing art as an effective way to cancel interactions between fields and wires. A twisted-pair SMA insulated-wire element will contract upon heating and allow expansion upon cooling much as a single SMA wire will, but will provide approximately twice the pulling force, with little if any increase in cooling time constant.

There is the added advantage that the heating current to such an SMA element may be provided entirely from one end - where an SMA element has one fixed and one moving end (very common) then at the fixed end current may be supplied up one strand and down the other, and all that is necessary at the moving end is to connect the two strands to each other, and to nothing else. This eliminates the otherwise sometime difficult problem of creating a reliable current return from a moving component.

So in this aspect of the present invention any actuator used near or within an RF environment, and especially within any of the tuneable RF devices described herein, is comprised of twisted-pair insulated SMA wire elements, with the length between successive twists at most half a wavelength of the RF energy of concern, and preferably much smaller than half such a wavelength.

More generally it may often be useful to use insulated SMA wire for the SMA elements of any SMA actuators used to tune the RF devices described herein. And to avoid other interactions with RF currents flowing through the conductive structures of the RF devices being tuned, it is preferable in all cases to provide a return path for any SMA element heating current that is separate from the conductors of the RF energy.

Figure 35 shows a schematic section of a conventional solid wall cavity, with two opposing sections (the rest is not shown, for clarity) of solid conductive wall 3501 and 3502 parallel to each other containing in between them a resonator 3503 which is galvanically grounded at one end to wall 3502 and unconnected at its other end, which instead has a glass wafer 3504 attached to its surface. Also not shown (again for clarity) are the remaining conductive walls of the cavity that define its shape and volume. A tuning element 3505 of the present invention nominally parallel to wafer 3504 penetrates wall 3501 through an aperture in that wall and is positioned a small precise orthogonal distance from glass wafer 3504, and is slidably mounted such that it can move in and out of the cavity formed between walls 3501 and 3502. Its movement is in turn controlled by actuator 3506 which is shown only schematically, and is mechanically attached to actuator 3506 by link-pin 3507, so that when actuator 3506 moves the tuning element 3505 may be positioned at a range of distances along the end of resonator 3503. This has the effect of modifying the capacitance (primarily) at the end of resonator 3503 which in turn changes its resonant frequency, providing a tuning function. Actuator 3506 is mechanically attached (not shown) to the outside walls of the cavity. In this example the tuning element 3505 is made of conductive material and extends through the cavity wall 3501 to the outside and provides an unwanted potential leakage path for RF energy from inside the cavity. Such leakage is prevented in this example by the integral RF choke built into tuning element 3505 in the form of a series of wide (capacitive) and narrow (inductive) sections of the tuning element, formed by cutting notches one of which is shown at 3508 into either side of the element and spaced along the element in the direction along the tuning element. A further pair of notches similar to those visible at 3508 are cut into the portion of tuning element 3505 where it penetrates the cavity wall 3501 and which are thus not visible in this view. As described previously, the glass wafer 3504 could advantageously be replaced by a corundum coating on either of the two adjacent faces (i.e. resonator or tuning element) to act as a very high quality high dielectric-constant dielectric between them.

Figure 36 shows a schematic section of a cavity constructed between two parallel conductive ground planes 35010 and 35011 , only sections of which are shown for clarity) and which are electrically and mechanically connected to each other by a number of conductive vias 3509 most of which are not shown, for clarity. The size and volume of the cavity are defined not only by the ground planes 35010 and 35011 but also by“walls” made up of suitably spaced conductive vias such as 3509 but these additional cavity“walls” are not shown in this figure, for clarity.

Contained in between them is a resonator 3503 which is galvanically grounded at one end to a pair of vias 3509 and unconnected at its other end, which instead has a glass wafer 3504 attached to its surface. A tuning element 3505 of the present invention nominally parallel to wafer 3504 penetrates into the cavity through an iris formed by an adjacent pair of vias 35096, the positioning and separation of which are chosen to minimise RF energy leakage from the cavity along the direction of the tuning element 3505. The tuning element 3505 is positioned a small precise orthogonal distance from glass wafer 3504, and is slidably mounted such that it can move in and out of the cavity formed between ground planes 35010 and 35011. Its movement is in turn controlled by Actuator 3506 which is shown only schematically, and is mechanically attached to Actuator 3506 by link-pin 3507, so that when Actuator 3506 moves the tuning element 3505 may be positioned at a range of distances along the end of resonator 3503. This has the effect of modifying the capacitance (primarily) at the end of resonator 3503 which in turn changes its resonant frequency, providing a tuning function. Actuator 3506 is mechanically attached (not shown) to the outside of one of the ground planes. In this example the tuning element 3505 is made of conductive material and extends outside the cavity through the iris formed by two of the vias 35096 and could provide an unwanted potential leakage path for RF energy from inside the cavity. Such leakage is prevented in this example by the integral RF choke built into tuning element 3505 as can be seen more clearly in the tuning pin drawing in Figure 35. However, clearly shown in Figure 38 is a conductive plate 350801 grounded in this example by the two vias 35096, which is arranged parallel to and very close to the surface of tuning element 3505 to provide strong capacitive decoupling to the region in and around the notches forming the integral RF choke. Such grounded capacitive elements (i.e. like 350801) are to be understood to be similarly positioned over the notched RF choke sections of all the tuning elements in all of the figures, but are mostly not shown for clarity (i.e. to allow visibility of components and structures that would otherwise be hidden by the presence of these capacitive elements. They form an integral part of the in-line RF chokes.

Figure 38 shows two coupled resonators 3503 and 35031 grounded at one end by vias 35093 and sandwiched between two parallel ground planes 35010 and 35011 as before in figure 36, except that ground plane 35010 has been removed in figure 38 to reveal the structures between them. The two resonators have glass wafers 3504 and 35041 covering part of one face near the end opposite the grounded end, and the resonators are tuned by moveable grounded tuning elements 3505 and 35051 which are in turn moved by Actuators 3506 and 35061 which are shown only schematically. The two resonators are electromagnetically separated from each other by a set of closely spaced vias 3509 lying between them and connecting at each end to the ground planes 35010 and 35011 , and this line of vias effectively forms one“wall” of each of the two cavities in which sit the resonators 3503 and 35031. The remaining walls of these cavities which are similarly constructed of lines of closely spaced vias are not shown, for clarity. However, near the tuned end of the resonators there is a carefully shaped and positioned and sized gap in the cavity-separating wall of vias between the vias 35091 and 35092 and at that location this gap functions as the coupling iris which electromagnetically couples the two resonators 3503 and 35031. The notches some of which are indicated at 35081 forming the integral RF chokes in the tuning element 35051 can be clearly seen in this drawing, although only one notch at 3508 is visible in tuning element 3505 because the figure shows the capacitive cover plate 350801 grounded by the vias 35096 which normally covers all such notches in such integral RF chokes and is an integral part of the RF choke structure. Other cover plates like 350801 have been removed in the drawings to more clearly show the structures beneath but it is to be understood that some conductive structure serving the same capacitive function would be expected to be present in all such integral RF chokes.

Figure 38 is a further development of the structure shown in figure 37 and is now in this drawing complete enough to form a fully functional tuned filter. The additional elements will now be described. The coupling iris between the two resonators 3503 and 35031 previously (in figure 37) formed between two vias 35091 and 35092 has now become a tuned coupling and is formed between via 35092 and conductive tuning element 35052 which is moved towards and away from via 35092 by actuator 35062 shown schematically only. Again there is an integral RF choke along the stem of 35052 formed by a series of notches, and a grounded capacitive cover plate over these (not shown) as described above. So while the two resonators 3503 and 35031 may be independently tuned by actuators 3506 and 35061 , the coupling between them may also be independently tuned as well, by Actuator 35062, the action of which is to open and close the gap in the electromagnetic wall between the resonators, as tuning element 35052 moves. An input port to the filter is formed by a stepped cylinder 350111 the function of which is to convert the impedance at the input end 350121 (e.g. typically 50ohm if this was a coaxial connection) to the impedance of the resonator 35031 to which impedance converter 350111 is electrically bonded (e.g. soldered). The overall function of this subassembly is to connect with low return loss an input signal into the cavity containing resonator 35031. Similarly, an output port is provided at the end 350120 of a stepped cylinder impedance converter 350110 electrically connected to resonator 3503 at the other end. Typically the output port 350120 might also be matched to a coaxial connector, e.g. at 50ohm impedance.

These two input/output ports (I/O ports) are each separately tuned by grounded conductive tuning element 35053 (for the input port at 350121) and grounded conductive tuning element 35054 (for the output port at 350120) and the tuning is accomplished by moving these tuning elements towards and away from their respective stepped cylindrical impedance converters, the movement being produced by actuators 35063 and 35064. These tuning elements 35053 and 35054 operate in magnetic mode rather than electric mode, and are grounded to the inside of the cavity by elements not shown for clarity.

Figure 39 illustrates an alternative arrangement of the components of the two- resonator tuneable filter shown in figure 38. Here one of the resonators, 3503, has been flipped through 180 degrees end to end so that it still lies parallel to the other resonator 35031 but has its grounded end at the opposite side of the ground plane to the other resonator. Its I/O port 350120 and tuning elements 3505 and 35054 (and associated components) have all moved with it, so that functionally this layout is more or less identical to that in figure 38. However, when putting together a number of filter sections to make a more complex design, this ability to flip resonator cavities gives great flexibility, particularly when building triple (or more) ground plane 3D filters.

Figure 40a illustrates a 3D folded tuneable filter with three ground planes 35010, 35011 , 35012, six resonators and with I/O ports at 350121. This is effectively a development of the complete tuneable filter shown in figure 38 and figure 39, now with six resonators (three between ground planes 35010, 35011 and three more between ground planes 35011 , 35012) and additional inter-cavity couplings between each pair of cavities adjacent to each other on opposite sides of the central ground plane 35011. The following parts of Figures 40b - 40e show all this in more detail.

In figure 40b the top ground plane 35010 has been removed to show the top layer internals, where three resonators 3503, 35031 , 35032 can be seen each situated in their own cavities defined by closely spaced walls of conductive vias connecting between ground planes 35010, 35011 not all of which are shown, for clarity. Each resonator 3503, 35031 , 35032 is tuned respectively by tuning elements 3505, 35051 , 35052 themselves made moveable by actuators 3506, 35061 , 35062. The tuning elements are RF isolated from their Actuators by integral RF chokes implemented as a sequence of wide and narrow sections along the length of the elements and these sections are capacitively coupled to ground by grounding sections not all of which are shown for clarity. An I/O port 350121 can be seen to couple into the cavity containing resonator 3503 and this coupling is magnetically tuned by tuning element 35056 moved by actuator 35066. The coupling between the cavities of resonators 3503, 35031 is tuned by tuning element 35057 moved by actuator 35067 and that between the cavities of resonators 35031 , 35032 by tuning element 35058 moved by actuator 35068. The cavity of resonator 35032 is coupled through an aperture or iris in ground plane 35011 (not visible in this view), to the cavity beneath it between ground planes 35011 , 35012 and this coupling is tuned by tuning element 35059 (only the far end of which is visible in this drawing where it is connected to its associated actuator 35069).

This iris coupling represents the sequential path through the six-resonator filter. However there are additional couplings through the central ground plane 35011 between the other two pairs of cavities which lie adjacent to each other on opposite sides of central ground plane 35011 , though these additional couplings are not visible in this drawing. These additional through- ground-plane-couplings are also each independently tuned by a further pair of tuning elements moved each by their own Actuators, again not shown in this drawing for clarity. figure 40c is a close-up view of the through-plane iris 350200 coupling the cavities of resonator 35032 and 35033. Resonator 35032 which lies between planes 35010, 35011 as can be seen in figure 40b has been removed in this drawing to reveal the coupling iris 350200 beneath it, which coupling can now be seen to be tuned by grounded conductive tuning element 35059 (also beneath resonator 35032) moved by actuator 35069. Resonator 35031 which is parallel and adjacent to (removed in the drawing) resonator 35032 is shown in the drawing. The iris-end of tuning element 35059 has a tab or ridge 350280 on it which is narrower than the iris slot 350200 so that when element 35059 partially covers iris 350200 it forms a ridge-loaded waveguide between the cavities on either side of plane 35011 which assists coupling and tuneability. A ridge-loaded waveguide section is therefore formed inside or in the vicinity of the coupling iris 350200 (i.e. a waveguide loaded with a ridge in the middle) and this structure has the lowest cut-off frequency which is strongly dependent on the ridge and the shape of the waveguide, and in particular on the gap between the ridge tip and the opposite wall of the coupling waveguide. Even when the ridge does not fill the entire thickness of the iris 350200 it will strongly affect the coupling value. A possible variant (of the structure shown in figure 40c) has a coupling pin with a thicker tip protruding through the entire thickness of the iris 200. In this view the resonator 35033 in the cavity below plane 35011 is mostly hidden from view by ground plane 35011 although a portion of it closest to iris 350200 is visible through the iris. Just visible to the left of iris 350200 is the capacitive grounding plate 350802 of the tuning element 35052 which is used to tune resonator 35032. Also now clearly visible is tuning element 35058 moved by actuator 35068 which tunes the coupling iris between the cavities of resonators 35031 and 35032 (latter not shown for clarity).

Figure 40d is another close up of this through-ground-plane iris 350200 where the centre ground- plane 35011 has now also been removed to reveal the structures beneath. The iris 350200 is shown in outline directly beneath the tip 350259 of iris coupling tuning element 35059. Now also visible is tuning element 35053 (moved by actuator 35063) which in conjunction with dielectric element 35043 tunes resonator 35033, several conductive vias 3509, as well as actuator 35062 which moves tuning element 35052 which tunes resonator 35032.

Figure 40e shows a variant on the coupling tuning element of figure 40c, where a true ridged waveguide between the two coupled cavities is formed. Tuning element 35059 now has an extra thickened tab 3502591 which can be seen more clearly in the disassembled tuning element 35059A. Element 35059 moves in the direction shown by the arrow 001 over the surface of ground plane 35011 so that the thickened tab 350259 now fills the full thickness of the iris 350200 in ground plane 35011 so that a true ridged waveguide is formed. A copy of 35059 is shown at 35059A to clearly reveal the shape of the modified tab end 350259, 3502591.

In all of the drawings of tuned filters so far (figures 35 - 41) the actuators have been sited either outside of the RF cavities or in some cases merged with them, and in general could be any suitable type of actuator, though as aspects of the present invention, as mentioned above, they are any sort of SMA actuator, suitably configured to give enough stroke with adequate load-force capability. In many cases a double-bowstring actuator will serve this purpose well as its stroke can be tailored within wide limits even in a compact actuator, as the forces required for tuning these tuned filters are small typically <1 gram, and even a 25micron diameter SMA wire can pull nearly 15 gram and so quite high leverage can be applied to increase the stroke above that available directly from the contraction of a straight wire ( typically 3- 4%, but sometimes more, of its length). So it should be understood that where these figures label components as“actuators” (e.g. 3506, 35061, 35062, 35066, 35067, 35068, 35069 in figure 40b) it is intended to mean that a suitably specified SMA actuator should be used for that component, as per the definition of actuator above.

Figure 41x is a development of the aspect of the invention shown in figure 38 with which figure 41 x should be compared. The standalone actuator 3506 of figure 38 used to tune resonator 3503 has now been replaced with a completely integral actuator built in and around resonator 3503 in figure 41x. So in this figure can be seen an SMA wire 350171 running the length of resonator 3503 and in this case housed inside a groove in resonator 3503 which groove is lined with electrically insulating material, preferably grown on the surface of 3350 which if of aluminium may be aluminium oxide AI203 (in which case a further cost saving and simplification can be achieved by simultaneously growing a similar very thin insulating dielectric layer of AI203 on the top surface of the end of the resonator to replace the separate dielectric element 4 which forms part of the tuning structure). SMA wire 350171 is mechanically anchored in an insulating member 350170 positioned on the grounded base end of resonator 3503 and is electrically terminated here to a wire (not shown) connecting it to a controller which supplies heating current to change the length of the SMA wire. The other end of SMA wire 350171 is mechanically attached to grounded tuning element 3505 which is moveably mounted so as to be able to slide over thin dielectric layer 3504 (as before in figure 38) along the direction of the length of the resonator 3503. This end of the SMA wire may also optionally be electrically bonded to grounded tuning element 3505 which then forms a return path without additional wires, for the SMA-wire heating-current. When the SMA wire 350171 is heated by the controlled current and subsequently contracts, tuning element 3505 is pulled in the direction towards the resonator’s grounded base and in so doing more overlaps resonator 3503 increasing the capacitance between these elements and thus tuning the resonator. Also attached to tuning element 3505 is one end of a spring, a leaf spring 350172 in this variant, whose other end is attached to the base plate ground plane 35011 via some form of mechanical mounting, shown in figure 41x as pin 350173. The spring force is arranged to oppose the pulling force of SMA wire 350171 such that when the heating current to the SMA wire is reduced, and the SMA wire cools, spring 350172 stretches the SMA wire back to its original (unheated) length, thus performing the return-stroke function of the actuator so formed. An alternative to return spring 350172 is to add a second SMA wire capable of pulling 3505 in the opposite direction to SMA wire 350171. This may also be conveniently arranged to run along the length of resonator 3503, but in this case one end of the wire is mechanically attached to base 35011 (or 35010) in the region of pin 350173 (but in line with wire 350171) and the other end to an insulating mechanical rod or strut or link extending from tuning element 3505 above the length of resonator 3503 to a point near to or above mount 350170 at which point the second SMA wire is attached to the rod. In this way both SMA-wires can have similar lengths and be nearly co-located so taking up very little extra space, but may operate as a push-pull pair of actuator wires as is known in the art. The actual detailed configuration of the integral SMA actuator can of course take many forms known to those skilled in the art and the configuration shown here is not meant to be in any way limiting but merely indicative of what is possible. In this drawing tuning element 3505 is grounded and covered by grounding plate 350801 although it may also be arranged to be grounded by any other suitable means, to either of the adjacent ground planes of the filter (i.e. 35010, 35011). One of the immediate advantages of this implementation of a tuning actuator may be seen to be that it is entirely enclosed within the RF cavity surrounding resonator 3503 and takes up hardly any additional volume, and also requires no RF chokes or other countermeasures to reduce RF energy leakage along mechanical linking structures. It might also be expected to cost less to implement than a separate actuator structure.

Figure 41a shows a close-up view of the grounded end of resonator 3503 from figure 41x, and illustrates the location of SMA wire 350171 sited in groove 350175 of the resonator 3503 and mechanically terminated on insulated mount 350170, and shows the electrical termination point 176 of the SMA wire. The moveable tuning element 3505 can be seen at the opposite end of the resonator separated from it by dielectric layer 3504 together the latter’s grounding plate 350801. Several conductive vias 3509 are also visible but note that the ground planes 35010, 35011 , 35012 are not shown in this drawing.

Figure 41b shows a close-up view of the non-grounded end of resonator 3503 from figure 41x now looking from the underside (i.e from the direction of ground plane 35011), and again illustrates the location of SMA wire 350171 passing along a groove in the surface of resonator 3503 and mechanically (and in this case electrically too) terminating in a conductive mounting point 350177 connected to tuning element 3505. The return spring 350172 is clearly visible in this view attached at its free end to tuning element 3505 at point 350178 and at its fixed end to ground plane 35010 (not shown) via post 350173. In this view spring 350172 is significantly extended and SMA-wire 350171 significantly contracted (i.e. the SMA wire is hot) which causes tuning element 5 to significantly overlap resonator 3503 increasing its capacitive coupling to ground.

The SMA wire actuator integral with a resonator as illustrated in figures 41x, 41c, and 41b, may of course be used in any place in a tuneable filter of the present invention where a tuneable resonator or other form of tuneable or moveable element is required. One disadvantage of this particular arrangement shown is that a straight- wire actuator configuration is used. While the length of a resonator may be sufficient to house enough length of SMA wire in this way to achieve the required range of tuning (i.e. sufficient actuation distance), there is no leverage between the moving end of the SMA-wire and the moveable element even though the force capability of even a 25micron SMA-wire is usually many times greater than that required to move such moveable tuning elements, and as a consequence, the power consumption (proportional to SMA-wire length) of this configuration of actuator will be higher than otherwise achievable with a suitably long lever (e.g. as may be achieved with a bowstring or double bowstring actuator). So this particular representation of integral SMA-wire actuator is meant only to be illustrative of what can be achieved and in no way limiting to all of the other ways that such leverage may advantageously be incorporated in an integral SMA actuator, whereupon compactness, low cost, simplicity as well as low-power consumption may all be achieved simultaneously.

Figures 42x, 42a, 42b, 42c, 42d and 42e show a tuneable filter where the phase is the primary parameter changed by tuning (a phase-shifter) which is one aspect of the present invention. There are three layers of metal 350510, 350520, 350530 each on top of a layer of dielectric 350511 , 350521 , 350522 and these layers are all sandwiched together as shown in figure 42. The top layer of metal 350510 can be seen to have an aperture in it of rectangular shape with rounded corners, exposing a portion of the dielectric 350511. The other layers of metal are of similar shape. The top metal layer 350510 is joined to the middle metal layer 350520 by a rectangular array of closely spaced conductive vias 3509 piercing the dielectric layer 350511 , two such vias being labelled 3509. The diameter and spacing of the vias are chosen appropriately so as to form a waveguide for the RF frequency of interest, as is known in the art. The second metal layer 350520 is similarly joined to the next metal layer 350530 by a further similar rectangular array of closely spaced conductive vias piercing the dielectric layer 350521 thus extending the waveguide so formed. Each layer of dielectric 350511 , 350521 , 350522 has a rectangular slot 350519 cut through it somewhat smaller than the aperture in the metal layers in such a way as to minimize the impact to propagation conditions of the dominant operating mode of the waveguide. For example, a large slot aperture may eventually move the cutoff frequency upwards to the extent it reduces the operating frequency band. The aperture size is chosen to ensure that this does not happen. In practice this is defined by the dielectric parameters of the manifold substrate and those of the structures supporting the resonators, and may require an additional dielectric filler in front of the resonator to keep the cut-off frequency within desired limits. This dielectric filler would then become another dielectric layer atop element 350512 and become part of the moving element of this filter. There are graphs of cutoff frequency for such structures available in the literature and known to those skilled in the art which may be used to determine the required dimensions. In this slot in the fixed dielectric layers 350511 , 350521 , 350531 , is a moveable element including a resonator 350512 supported by a dielectric constant support 350513 (which may be a low-dielectric constant material), which moves in a direction orthogonal to the plane of the metal layers. In the aspect of the invention shown in figures 42x to 42e the resonator is a conductive flat metal element in the shape of a capital letter T (more clearly seen in figure 42a), but other useful shapes are possible and known in the art. For example, the T shape of elements shown may advantageously be rotated through 90deg in the plane of its thickness and suitably re-scaled to extend to the limits of the support 350513 so that the stem of the T points along the height of the waveguide rather than its width. The moveable element is caused to move by an actuator 350561 a small portion of which is visible beneath the bottom dielectric layer 350531. In operation the top aperture in metal layer 350510 is a waveguide I/O port and RF energy entering the port travels down the waveguide in both air (in the central cavity 350519) and dielectric (between the cavity walls and the walls of conductive vias 3509 surrounding the cavity), is reflected from the structures of the moving element (further described below) and returns to the I/O port with a phase proportional to the distance of resonator 350512 from the I/O port entry defined by the metal layer 350510. The output phase of the phase-shifter is thus controlled by the position of the moveable element which in turn is controlled by Actuator 350561 , which may in turn be controlled by electronic / digital commands from a controller.

Figure 42a is a top view of the phase-shifter of figure 42x which now clearly shows the T shaped resonator 350512 supported by support 350513, and also visible is the small clearance gap between these moveable parts and the surrounding dielectric layers 350511 , 350521 , 350531. The top metal layer 350510 with its round-cornered rectangular waveguide aperture cut through it is also clearly seen, as are the conductive vias 3509 connecting it to the metal layer 350520 below which form the walls of the waveguide.

Figure 42b is a cut away view of the phase-shifter where the dielectric layers 350511 , 350521 , 350531 have been removed to reveal the walls of vias 3509 connecting the metal layers 350510, 350520, 350530 forming the waveguide. The resonator 350512 on the moveable element support 350513 is also just visible. In this version of the phase-shifter there are three layers of buried vias 3509 rather than two and a fourth layer of metal on the bottom of dielectric layer 350531 , which lengthens the waveguide and allows for a greater range of travel of the moveable element while still remaining within the waveguide, accommodating greater values of phase shift.

Figure 42c shows another cutaway view of the phase-shifter looking upwards from beneath, with all but the top layer of metal and vias removed, as well as the three dielectric layers removed, to reveal the moving element comprised of its support 350513 and resonators, the second of which 350514 can now be seen on the opposite side of 350513 to where resonator 350512 is positioned. This pair of resonators 350512, 350514 are spaced apart in the direction of movement by a fixed distance chosen to maximise the reflection of RF energy back to the I/O port, and this distance is equal to half the wavelength of operation of the phase-shifter corrected for the reactance of the resonators. The actuator 350561 (which is supported by the rear of dielectric layer 350531 not shown) is mechanically coupled to the moving element by a link 350515 made of any suitable non-conductive material. Figure 42d shows a cutaway view of a slightly simplified version of the phase-shifter described above in figures 42x, 42a, 42b, and 42c, constructed by replacing the pair of resonators 350512, 350514 on the moveable element by a single conductive metal reflector or resonator, in which case the spacing support 350513 is no longer required, and the actuator 350561 may then be coupled directly to the first and only reflector 350512. In the drawing the resonator/reflector is shown as a self- supporting conductive (e.g. metal) T shape, but this could be replaced by other shapes with the appropriate resonance and/or reflective properties, e.g. a flat rectangular conductive sheet filling the whole aperture through the dielectric in the waveguide except for a small clearance gap around the edges, or a conductive flat ring. An alternative construction would have the conductive metal resonator/reflector etched or plated onto a thin dielectric backing support material.

Figure 42e is a more detailed development of the type of phase-shifter sketched in Figure 42d where we have shown more of the conductive metal layers 350510, 350520, 350530 and via waveguide walls 3509 (with some cut away for clarity). The moveable resonator/reflector 350512 is now directly mechanically connected to the moving member 350568 of a double-bowstring SMA wire actuator comprised of base 350567, SMA wires 350562, 350563 terminated mechanically and electrically to pins 350564 fixed to base 350567 at their ends, and half-looped around pins 350565, 350566 fixed to moving member 350568. The base 350568 is mechanically fixed to the body of the phase-shifter, e.g. to the lowest layer of dielectric, not shown in this drawing for clarity. When one of the SMA wires is heated enough it will shorten and pull 350568 with it, after which heating the other SMA wire causing it to shorten and allowing the first to cool will pull 350568 with it in the opposite direction, axially along the waveguide. In this particular configuration moving member 350568 is guided by a slot in base 350567. The SMA wires 350562, 350563 are positioned in the electric wall of the waveguide and will have minimum coupling to any RF field behind the resonator/reflector 350512. It will be seen that the actuator has now been brought inside the waveguide. The small size and flat profile of such SMA wire actuators allows them to be safely placed inside RF cavities and waveguides so long as they lie in the plane of the electric wall and thus remain decoupled. Thus an alternative configuration of the phase-shifter in this drawing could have one of the SMA actuator wires on either side of the resonator/reflector 350512 so long as these rules are adhered to. This makes for a more compact assembly. Another variant has the base 350567 of the actuator extending (further left and right in the drawing) into and/or beyond the walls of conductive vias to allow longer SMA actuator wires and thus greater stroke to be achieved. It is only necessary to ensure that the SMA wires do not touch the grounded parts of the metalwork (vias or ground planes).

The more detailed description of better integrating an SMA actuator with a moveable element outlined in the description of figure 42e can be applied to most or all of the other actuator configurations shown in the previous figures 35 to 41 to make low- cost, compact and efficient tuneable filters of all of the varieties described.

figure 43 shows a different aspect of the present invention in the form of a phase- shifting filter similar in many ways to that of figures 42x - 42e except that the device shown in figure 43 shows a through-waveguide phase- filter and not a reflective filter, where one end of the waveguide formed by the rectangular arrays of conductive vias 3509 is an input port (e.g. the top in figure 43) and the other end (e.g. the bottom in figure 43) is the output port. As in figure 42e the layers of dielectric in and around the conductive vias has been removed for clarity, and the metal layers 350510, 350520, 350530, 350540, 350550 too have been cut away to show the internals as have some of the arrays of vias. Inside the cavity in the dielectric layers can now be seen two actuators, illustrated here in non-limitative fashion as double bowstring actuators, the top one comprising two SMA wires 350562, 350563 in a bowstring configuration (for example, and not in any limiting way), a moveable element 350569 connected mechanically to a top reflector/resonator element 350512 which it can move axially along the waveguide relative to actuator support structure 350567. A second resonator/reflector element 3505121 is mechanically connected to Actuator support 350567 and this whole assembly of top actuator and both reflector/resonator elements is in turn caused to move axially along the waveguide by a second actuator below it in the figure, fixed to the phase-shifter body structure, for example to one of the dielectric layers of the stack. This second lower actuator again shown for illustration only in this example has a pair of SMA wires 3505621 , 3505631 in a bowstring configuration which move a moveable element 3505681 which

mechanically connects to the first actuator above and causes it to move relative to the waveguide; its base 3505671 is fixed mechanically to the phase-shifter body. The control wires (for heating the SMA wire) of the upper actuator can be either flexible (to allow free movement of this actuator) and constrained to the electric wall and led out of the sides of the waveguide though small apertures in the dielectric and between the walls of vias, or instead maybe led out to the lower actuator with flexible leads (to allow free relative movement) and then carried to the outside alongside the lower actuator’s control wires, again through suitable small apertures in the dielectric walls. The purpose of the first actuator is to maintain the correct spacing between elements 350512, 3505121 for minimum reflection to input port at the operating frequency at any time. So these two actuators are now right inside the waveguide cavity where the RF energy of interest is flowing (in this case from top to bottom) and so as to interfere as little as possible with the RF, the plane of the SMA actuator wires and their external connections (not shown) is aligned with the electric wall in the cavity; to further enhance the decoupling of all of these wires, the actuators are each sandwiched between a pair of dielectric (e.g. glass) wafers 350569, 350569A for the top actuator, and 3505691 , 3505692 for the bottom actuator, some of which have been cut away in the drawing to show the actuator structure within, and these dielectric wafers effectively“suck in” the RF field near the wires and so minimise its interaction with them. Now the two reflector/resonator elements 350512, 3505121 are no longer identical and their shapes and separation are chosen to minimise the reflection of RF energy back to the input port whilst achieving the required phase- shift at the output port at the operating frequency. In operation the upper actuator adjusts the axial separation of the reflector/resonators 350512, 3505121 and the lower actuator moves them both bodily along the waveguide axially. While figure 43 shows T-shaped resonator/reflectors, these may equally well be other shapes such as circular or elliptical rings, or cross-shaped elements; in fact any shape with the correct reflective and phase-shifting properties. A further refinement (not shown in this drawing for clarity) is the provision of additional actuators and tuning elements (which could take the form of moveable capacitive elements, or inductive elements as described elsewhere above) for each of the reflector/resonators 350512, 3505121 to minimise reflections to input port right across the frequency band of interest, each of these being co- moving with their respective resonator/reflector element. In a slightly different embodiment the two resonators 350512, 3505121 are each moved along the waveguide independently by separate SMA actuators, the actuators being mounted rigidly to the waveguide structure, thus allowing complete freedom of the absolute and relative positions of the two resonators provided they always remain on the same side of each other. The very low mass and compactness of these SMA- wire actuators makes such a device quite practical. Thus a very compact, very- low- RF-loss and low-cost through-waveguide tuneable phase-shifting filter may be provided in this way, which is simply not possible with conventional tuning elements and actuators such as electric motors which could not be placed inside the waveguide, and with no moving parts penetrating the waveguide walls and thus providing no leakage paths for RF energy. Such multiply tuned through waveguide phase-shifters have particular application where extremely low loss across a variable frequency waveband is required and some examples where they may be useful include industrial microwave ovens, timber driers, and crude-oil warming equipment, and indeed anywhere where tuning speed is secondary but cost and low RF loss are paramount.

Figure 44 illustrates another aspect of the invention where the tuning elements, previously operable by moving relative to their target tuned element, now instead deform rather than move as a rigid body, the deformation being maximised in the region close to the tuned element. The advantages of deformation instead of movement includes no sliding and so no friction, simplicity, potentially use of shorter SMA wire actuators which result in lower cost and reduced drive power, the elimination of the need for a separate return spring for the actuator (the deformed element itself provides that function), extreme compactness, highest reliability, and the ability to place the entire actuated tuning element right inside the waveguide to eliminate RF leakage to the outside. In figure 44 we see the same resonator 35031 and capacitive tuning element 35051 as was previously illustrated in figure 38 where it was moved by a separate actuator 35061 that had to be isolated from the RF in the cavity by an integral RF choke structure comprising the slots 35081 in figure 38. However, now in figure 44 there is a thin, conductive, elastic flexible strip

mechanically mounted and grounded at one end by for example a pair of vias to a ground plane (not shown), with a free end 350518 which is turned up in figure 44 for several purposes: first it stiffens the thin strip laterally, second it provides a mechanical anchor point for an SMA-wire 350178 in the centre at 350179, third it can provide an electrical termination for the SMA wire in which case 350179 could conveniently be a crimp formed out of the material of element 35051, and fourth, importantly an out-of-plane attachment point for that SMA-wire to provide sufficient leverage to allow the SMA wire 350178 when heated and controllably contracted to pull the end of the strip 35051 out of the plane of its base mounting and so bend up the strip away from the dielectric 35041 between it and the resonator 35031 , in so doing changing the capacitance between 35051 and 35031 significantly and thus tuning the resonator 35031.

The other end of SMA wire 350178 is mechanically terminated in an insulating mount 350177 and controlled heating current is fed into end 350176 by a control wire (not shown), the return current conveniently run through the grounded end of the wire at 350179. Only a short length of SMA wire 350178 is needed to produce significant bending. The strip 35051 can be very thin, just a few multiples of skin-effect-depth (e.g. ~1um in copper at 5GHz) at operating frequency being adequate to produce insignificant losses, and the material of strip 35051 can be chosen for its mechanical properties alone (e.g. phosphor bronze, or a polymer) if it is plated or coated with a suitable conductive surface e.g. silver. So long as the strip is adequately elastic that it returns to a flat shape (and produces adequate return force to re-stretch the SMA- wire) when the SMA-wire is unheated its mechanical properties are not critical, though thermal expansion coefficient might be chosen to cancel thermal expansion coefficient effects of the rest of the tuned cavity. Note that although in this illustration the deformation of RF element 35051 is produced directly by a section of SMA wire 350178, the same effect with somewhat more complexity may be produced by substituting an actuator in place of wire 350178, for example by an SMA-stepper- actuator, to provide improved long-term-position stability, plus other benefits as described above. Similar substitution of SMA-wire by stepper-actuator considerations apply to the deformable structures with integral SMA wires described in figures 45,

46, 47a and 47b, and are included herein.

Figure 45 shows another example of the deformation-tuning technique described above and shown in figure 44. The device shown in figure 45 shows an I/O port tuning element 35053 first seen in figure 38 where it was moved longitudinally towards and away from the I/O port components 350111 , 350121 by an external actuator 35063 and RF isolation was performed by an integral RF choke built into the stem of 35053. Here in figure 45 in this embodiment, the tuneable element 35053 comprises a thin, conductive, elastic flexible strip mechanically mounted and grounded at the end farthest from the I/O port 350111 , 350121 to be tuned, in this case by one or more conductive pillars 350971 , 350972 (which could be vias). An SMA wire 350779 is mechanically and preferably electrically connected to an offset / out-of-plane tab 350578 of element 35053 on one side of 35053. The other end of the SMA wire is mechanically mounted to an insulating mount 350777 attached to ground plane 35011 (not shown) and has an electrical connection point 350776 to which a control wire (not shown) is electrically connected. When the controlled heating current is passed through the SMA wire (using in this example the grounded end as return path) the wire heats and contracts controllably causing the thin conductive element 35053 to bend out of its natural plane and because of the offset position of the SMA wire (i.e. asymmetric to the principal axes of 35053) element 35053 will twist and so also bend so that its top edge (where the SMA wire attaches) will no longer be parallel to the axis of I/O port components 350111 , 350121 which will have a strong tuning effect as this is a magnetic tuning element. Similar mechanical and conductivity considerations as applied to the bending element and SMA wire in 44 apply here too. The net result of this configuration is a very small low cost controllable tuning element that can fit right inside the waveguide allowing no leakage of RF energy.

Figure 46 illustrates another aspect of the present invention and shows a variant of part of the assembly of figure 38. As previously, item 35031 is one of the resonators but now instead of being a solid rigid conductor it is instead a thin folded elastic conductive sheet comprising a long face of similar size to the resonator it replaces in figure 38, which has a grounded end (here by two vias 35093) and at that end it has also a folded side panel on which is mounted the I/O port assembly 350111 , 350121 again as before, which is folded around once more to rejoin the via 35093, the rigid U- section so formed designed to keep the location of the I/O port fixed in space. The other end of the long main face has a twice folded over end inside of which can just be seen an insulating mounting point 3501030 which serves as a mechanical anchor for an SMA wire 3501010 which extends back to the base where the SMA wire it is again anchored on another mounting 3501040. Wires carrying controlled heating current for 3501010 (not shown for clarity) are attached to each end of SMA wire 3501010 the one from the ungrounded end of 31 being run down the inside face of 35031 to the grounded base-end where it is effectively Faraday shielded from the RF field on the outside face. When the controlled current heats SMA wire 3501010 sufficiently it controllably contracts and in so doing pulls diagonally across the “thickness” of the envelope of 35031 causing it to bend significantly which changes its effective electrical length as well as its capacitance to the ground planes and any other grounded surfaces that may conveniently be arranged to lie close to its unbent position. In this way the resonator itself may be controllably tuned with this completely integral SMA-wire actuator, and by suitable arrangement of additional sections of folded pieces of 35031 the RF field can be adequately decoupled from the very thin SMA wire 3501010, There will be no RF leakage to the outside of the cavity containing resonator 35031. Several useful variants of this type of resonator structure by modifying the folded sections of 35031 which control its local rigidity, and by moving the end point mechanical attachments of the SMA wire 3501010. For example if a much shorter SMA wire is mechanically attached to the inside of the long main section of 35031 much closer to the grounded base, and the other SMA wire end moved further away from that face at its base than as shown in figure 46, then because of the greater leverage provided by the length of 35031 a significantly greater movement of the free tip can be achieved for significantly lower power consumption (which is proportional to SMA wire length), so long as the flexibility of the structure of 35031 in the direction of curvature so caused, is kept low enough to enable that degree of bend to be caused by the force capability of the SMA wire (e.g. a maximum of ~10 to 15gram for a 25micron SMA wire. Such an arrangement is also significantly easier to Faraday shield the SMA wire from the RF energy in and around the resonator 35031.

47a illustrates another useful deformable form of tuneable resonator 35031. This is made of thin folded elastic conductive sheet, e.g. thin metal, and consists of two curved surfaces with facing each other connected at the top (non-grounded end) and optionally also at the bottom (grounded end) with a further piece of the continuous thin conductive sheet. An SMA wire runs up the centre of the inside of the structure and its bottom end 3501120 is just visible in its mechanical and electrical insulated mounting 3501140, while its top end (not visible in this drawing) is mechanically and electrically mounted to the centre of the top section of 35031 joining the top ends of the two curved surfaces. When the SMA wire is controllably heated by a current applied to end 3501120 by a control wire (not shown) using return current path to ground via the body of resonator 35031 , the SMA wire within 35031 contracts which causes the flexible long curved walls to curve through a tighter radius and thus for the entire resonator structure envelope to shorten changing its inductance as well as its capacitance to the surrounding grounded components (not shown for clarity). The SMA wire is symmetrically placed between the two curved walls and is effectively in a complete Faraday cage, completely isolated from the RF energy in the cavity as can be seen more clearly in the side view in Figure 47b. The end result is a completely controllably tuneable resonator taking up no additional space (over and above that of a solid metal component), having lower mass, very high Q (e.g. the external portion of 31 could be silver plated to a few skin depths at negligible cost) and no leakage of RF energy to the outside of the cavity, and all at very low cost, and high reliability. Another useful variant of this aspect of the present invention is to instead have the two curved surface curving outwards rather than inwards, when more use can easily be made of their capacitance to nearby ground planes for example. This particular curved wall SMA-wire tuned resonator is presented as an illustration of what is possible using thin flexible curved wall structures with an internal Faraday-cage screened SMA wire to actuate the tuning function, and it will be clear to those skilled in the art that an infinite variety of such curved (and curved and straight) shapes can be used in the same way and all such are included herein. Figure 50 shows a phased array antenna comprised of a 2D array of phase shifters, which may be any of the types of phase-shifter actuated by an actuator, described herein. An RF feeding system 3501401 irradiates the antenna array (in transmission mode, which will be described from hereon, but the same device also works similarly in receive mode). Each array element is formed by the metal layout configuration that resonates at the frequency of operation to facilitate the strongest possible interaction with the incoming wave. In figure 50 the elements are shown realised as dipoles 3501403 with geometrical dimensions chosen to obtain maximum interaction with the incoming wave. The dipole metal structures may conveniently be laid out on the surface of a dielectric panel supporting the whole structure (not shown in figure 50).

Each dipole is connected to an independently variable length transmission line 3501404, designed to channel the energy received by the dipole to the reflector 3501405. The reflected energy then returns to the dipole with a phase determined by the specific length of each transmission line. The dipole than reradiates the energy with phase as determined by the length of its respective transmission line. The length of each transmission line is independently changed by a mechanically connected actuator (not shown in figure 50). The array of phase shifters (for each polarisation independently where dual polarisation structures e.g, as per figure 49, are used), can be individually positioned so as to form a directional beam of radiation of the reflected waves and form an antenna pattern 3501406, for example, as shown in figure 50, for one polarisation only for clarity - the other polarisation may also be formed similarly into a beam which is independently steerable from the first polarisation in the case of dual polarisation phase-shifters. The direction and shape of the antenna pattern is adjustable by suitable positioning of each of the actuators connected to the phase- shifters.

The array as described may be improved by layering dielectric and metal layers behind all the other elements including the reflectors, to act as a shield and minimise back radiation, which metal layers may advantageously be interconnected with other structures in the array with vias through the thickness of the plane of the antenna.

A further variant of this phased array dispenses with the variable length transmission lines altogether and instead makes the individual dipoles independently moveable orthogonal to the plane of the array, the movement of each being controlled by a dedicated Actuator capable of moving the dipole precisely with 4-bit, or preferably 5- bit or more preferably 6-bit precision. In such an arrangement the dielectric panel supporting the whole structure may now be pierced with an array of slots in which the dipoles move and to which the Actuator stators may conveniently be mounted. Again it is advantageous to add a rear dielectric layer and behind that a plane metal layer, to minimise unwanted rear-radiation.

A phase or frequency tuneable device comprising an RF cavity exploiting the thermo- mechanical properties of SMA material applied in such a way as to achieve controllable modification of electric and/or magnetic components of one or several of the eigenmodes of electromagnetic field supported by the RF cavity. Close integration of SMA actuation functions with the rest of the electromagnetic structure of the device is described, resulting in the capability of very high performance, low- RF-loss, high reliability, compact and low-cost fully tuneable RF filters and phase shifter.

APPENDIX A - AN ANALYSIS OF SMA WIRE MECHANICAL OVERLOAD

In the embodiments of actuators described in this appendix, one end of the wire is mechanically connected via a pre-tensioned or pre-stressed spring mechanism, rather than via a rigid connection. There is a pre-stress force Fs in the spring mechanism, so that the spring mechanism does not change in length in the direction of tension in the SMA-wire while the magnitude of that tension is below the spring mechanism's pre-stress force Fs, thus providing a tight mechanical link to the SMA- wire. However, once the tension in the wire rises above the pre-stress level in the spring mechanism, then the spring mechanism begins to change in length according to its spring rate k. With a suitable mechanical design including end-stops to limit the overall change in length of SMA-wire plus spring means, a spring rate k may be chosen such that the maximum tension ever experienced by the SMA-wire under overload conditions is always within the wire’s safe limit. Under these worst-case conditions, the spring mechanism will have extended a distance Xs and will thus exert a total force on the SMA-wire of FS+k.Xs and by design this force is kept lower than the maximum safe force capability of the SMA-wire, Fwmax.

One way to implement such a scheme is to use the spring mechanism to hold with a preset force at least one end of the SMA-wire against wire-end-stop means which may for example comprise part of, or be fixed to, the actuator base structure.

Alternatively the wire-end-stop mechanism may be mechanically connected to the load to be moved by the actuator. The spring force directed along the direction of wire tension has a magnitude and sign sufficient to hold the end of the wire against the wire-end-stop means for wire-tensions smaller than Fs.

The pre-stressed spring mechanism holds constant mechanical length until the force applied to it exceeds the pre-stress force level Fs. Thus the pre-stressed spring mechanism applies no retarding force to the SMA-wire during a normal (non overload) actuator stroke.

That is, an SMA-wire actuator has at least one wire-end mechanically connected to its mounting (load-end or fixed end) via a mechanical pre-stressed spring mechanism with the magnitude of the spring mechanism pre-stress force component in the direction along the line of the wire no greater than the maximum safe working tension of the SMA-wire.

Alternatively, the actuators can use a pre-stressed spring-mounted SMA-wire mechanical terminal mechanism that holds the SMA wire-end(s) against a defined end-stop with a force less than or equal to the maximum safe tension force for the wire, such that when the tension in the SMA-wire rises above this pre-stressed spring-force, the spring changes length allowing the SMA wire-end to move away from the end-stop at least to a distance sufficient to allow the load to be prevented moving further by the load-end-stop, and the spring-constant is designed such as to keep the wire tension within safe limits during this change in length even when the SMA-wire is fully actuated in the fully austenite phase.

An SMA-wire actuator designed in this manner will have the spring mechanism pre stress force magnitude and spring mechanism spring-constant such as to keep the tension in the SMA-wire at or below the maximum safe working tension of the SMA- wire at all positions of the load between the mechanical load end-stops even when the SMA-wire is fully actuated, for example in the case of NiTi SMA, in the fully austenite phase.

When the actuator is fully actuated (e.g. SMA all in austenite state) and the overload force is sufficient to move the actuator output to its load end-stop we have (using the definitions of actuator gain Gs(x) defined in Appendix B):

• x = S; (actuator output at full stroke)

• 2L = 2Lh; (wire length at hot minimum)

• Sw = x/Gs(x) = S/Gs(S); (change in length of SMA-wire, and also of spring means) Fw = Fwmax; (Wire tension no more than max tension loading, by design)

and the spring means has changed length by Sw, so the required spring-constant k of spring means is:

• k = Fwmax / (S/Gs(S)) = Gs(S). Fwmax /S; (by Hooke’s Law) [equn P1]

The spring mechanism may be either a tension or compression spring mechanism, and may be embodied with coil-springs, leaf-springs, flexures, stretchable or compressible materials (e.g. rubber, polyurethane), or other mechanical spring forms, or even magnetic couplings. It is preferred that the spring mechanism is fabricated out of the same continuous piece of metal (e.g. phosphor-bronze, or stainless steel) that is used for the SMA-wire crimp terminal (where the wire is crimped) and is mounted to the base under pre-stressed tension or compression of a magnitude as described above.

An SMA-wire actuator has one or both of the SMA-wire ends crimped into a piece of metal for mechanical and/or electrical connection, and preferably the spring mechanism is fabricated out of the same continuous piece of metal used as the SMA-wire crimp. More preferably, the metal is phosphor bronze or stainless steel.

In the case of a single bowstring-actuator, the load force does not pull directly along the direction of the SMA-wire. An overload mechanism as described above could be applied to either or each end of the bowstring actuator wires and would provide significant wire overload protection. However, because the load is applied to the centre portion of the SMA-wire, unless the mechanical set-up is perfectly balanced it is likely that one of the wire-ends will release upon overload before the other, and this will cause the point of contact of the SMA-wire with the load-attach pin (at or near the wire centre) to change. That is, the wire will be dragged around the load-attach pin, and if such overload conditions occur frequently may cause wear-damage to the SMA-wire middle section.

A variant of this overload protection mechanism can be used as well a or instead of the above-described pre-stressed spring-loaded wire terminal mechanism. In this variation, an SMA-wire bowstring-actuator has the actuator load connected to the centre region of the SMA-wire via a mechanical pre-stressed spring mechanism. A load-attach pin mechanically connects with the SMA-wire of the bowstring actuator and is mechanically connected to the actuator load via a mechanical pre-stressed spring mechanism. The spring mechanism pre-stress force Fs is set close to or above the rated actuator output force. Thus when the actuator is operating within its rated limits the pre-stressed spring mechanism does not change length. When the force exerted on the spring mechanism exceeds the pre-stress force, the spring mechanism changes length according to the value of its spring constant which is designed such that even when the SMA-wire of the actuator is fully actuated, and when the maximum rated overload force is present, the spring mechanism extends sufficiently to maintain the force on the SMA wire at or below its maximum safe operating tension.

In an overload-protected bowstring actuator, it is preferred that the load-attach pin is slidably mounted - sliding in the direction of the load force - into the bowstring push- rod, and is held against push-rod-end-stop means on the push-rod by a pre-stressed spring mechanism, which holds the load-attach pin against the push-rod-end-stop means by a spring force of magnitude approximately equal to or greater than the rated output-load-force of the actuator. In this way, should the load subject the push- rod to a force greater than the rated output-load-force then the spring mechanism will deflect and allow the load-attach pin to move relative to the push-rod (and thus to the load) such that it no longer continues to drag the SMA-wire along with the load movement. This decoupling above a fixed load force prevents excess tension in the SMA-wire. The actuator configuration should be such that when the load has moved the push-rod to the push-rod’s mechanical limit end-stop, then the force applied to the load-attach pin is no more than the SMA-wire tension is safely able to apply even when in the full austenite (actuated) state. In this case the spring mechanism does not change length during a normal (non- overload) actuator cycle, and the spring mechanism provides no opposing force to the movement of the actuator load until overload-tension is reached. An SMA-wire bowstring-actuator has the output load connected to the centre region of the SMA-wire via mechanical pre-stressed spring mechanism. It is preferred that the pre-stress force magnitude in the direction along the line of the actuator load travel is close to and not much greater than the maximum safe output-force of the actuator, that the load mechanical travel is limited by load-travel mechanical end-stops, and that the spring mechanism pre-stress force magnitude and spring mechanism spring-constant are such as to keep the tension in the SMA-wire at or below the maximum safe working tension of the SMA-wire at all positions of the load between the mechanical load end-stops.

Also, it is preferred that the spring mechanism pre-stress force magnitude and spring means spring-constant are such as to keep the tension in the SMA-wire at or below the maximum safe working tension of the SMA-wire whatever the actuation state of the SMA wire, and particularly when the SMA-wire is fully actuated in the fully austenite phase.

When the bowstring-actuator is fully actuated (SMA all in austenite state) and the overload force is sufficient to move the actuator output to its load end-stop, then:

• x = S; (actuator output at full stroke)

• 2L = 2Lh; (wire length at hot minimum)

• dl_S = x; (change in length of spring means at maximum overload)

• Fw = Fwmax; (Wire no more than max tension loading, by design)

• FL = Gf (x). Fwmax = Gf (S). Fwmax ; (load force is no more than Gf (S) times the wire maximum tension, by design, under max load conditions).

Now the spring pre-stress tension is Fp, produced by stretching or compressing the spring by the appropriate amount, and this Fp may be set at or slightly above the rated actuator output load force Fop (a design input specification). At maximum overload, the spring mechanism has been stretched or compressed by an additional distance x because it is now in series with the load, not the wire as previously, and this additional stretch produces an additional spring force of FL - Fp. So the required spring-constant k of spring means is:

k = ( Gf (S). Fwmax - Fp) /x; (by Hooke’s Law)

In a double bowstring actuator, where the two individual bowstring actuators drive in opposition the same load, or share a push-rod, one or both of the bowstring actuators’ load- attach pins may be pre-stressed spring mounted and optionally slidably mounted on and spring loaded against the push-rod, as described above, thus protecting the whole double bowstring actuator against mechanical overload.

So, a double bowstring actuator may be constructed by combining any one or two of the overload-protected single bowstring actuators described above, and in the case of one of the above with a non-protected single bowstring actuator.

However, in the case of using two overload-protected bowstring actuators, there is also the opportunity to simplify the mechanism by combining the two otherwise separate spring mechanisms of the two actuators into one, so that one pre-stressed spring mechanism connects to both SMA-wire central regions. One pre-stressed spring mechanism holds each of the two load-attach pins against their own end-stops on the push-rod, with each end of the single spring connected to one of the two load- attach pins, and mounted under a pre-set (pre-stressed) tension of magnitude to support nearly full or full specified load-force on either pin without either pin lifting off its end stop. A double-bowstring actuator using this arrangement has single a pre stressed mechanical spring mechanism to connect the centre regions of both SMA- wires to the load.

In this case the spring constant e is the same as for the single bowstring actuator described above:

k = ( Gf (S). Fwmax - Fp) /x; (by Hooke’s Law)

For simplicity, reliability and low cost, it is preferable to minimise the total number of components in a mechanism, and in the overload protection device of the present invention this is done most conveniently by making the SMA-wire terminal and crimp out of thin sheet metal such as phosphor-bronze or stainless steel, and also fabricating the overload-protection spring mechanism from the same contiguous piece of metal, and if possible making the external electrical connection terminal also out of this same single piece of metal. This avoids multiple electrical and mechanical joints providing much higher reliability, and makes for a compact, light-weight and low-cost assembly.

In all the embodiments of the present invention a control system can be provided that is capable of measuring the position of the actuator output by the use of suitable position sensor. The sensor can for example use electrical resistance measurement of one or more of the SMA-wires of the actuator embodiments, and from the measurement(s) of resistance the control system estimates the instantaneous length(s) of the SMA-wire(s), and from these length estimates together with knowledge of the specific actuator mechanical configuration and geometry the controller is capable of estimating the output position of the actuator and thus is able to accurately position the actuator output.

APPENDIX B - DETAILED ANALYSIS OF OPTIMISED HIGH-SPEED ACTUATION WITH SMA-WIRE

In this analysis, the same definitions as outlined on Appendix A for actuators and bowstring actuators should be used.

Very high speed actuation with SMA-wire is usually possible only in one direction - i.e. the heating phase of the whole cycle - because while it is easy in principle to apply heat arbitrarily quickly (e.g. by applying a large enough voltage Vto the ends of an SMA-wire and relying on the V^/R internal Joule-heating), it is far more difficult to induce very fast cooling as there is no equivalent internal method of cooling, and all the heat energy must be removed through the surface of the wire, primarily by conduction and convection. However, there are practical limits even to the allowable rate of heating without damaging the SMA-wire of the actuator.

This analysis defines the limits and thus describes the optimal heating method for fastest safe operation. The analysis is applicable to any form of SMA-wire actuator, and is intended to apply generally to all such forms of SMA-wire actuator including the simple straight-wire actuator, and the bowstring-actuator, as specific cases.

For any one-wire of a generic SMA-wire actuator with active SMA-wire mass M _w (i.e. the mass of the SMA-wire that is actively changing shape during actuation, and ignoring end-effects), and total wire-length L _tot = 2L _C when initially at temperature TO (below or equal to the austenite start temperature A _s ). The austenite finish

temperature is A _f > A _s . The wire is coupled to a moveable load mass M _L (including the mass of any push-rod and other co-moving coupling components). Define the stroke S of the actuator as the maximum distance the actuator is able to move the load and let x be the distance moved along the direction of actuator output thrust, at any point of the actuator cycle. Then 0 <= x <= S.

Define a mechanical stroke slope-gain g _s (x) where g _s (x) =dx/ d (2L) = ½ dx/dL, which is the differential gain, when the load is at displacement x, and the SMA wire has total length 2L. Note that this is not in general the same as the mechanical stroke bulk-gain G _s (x) where G _s (x) = x/(L _tot - 2L), when the actuator load has moved a total distance x, and the actuator wire has changed (shortened) from initial length L _tot to 2L, when some fraction of the wire has phase-transformed from martensite to austenite. We also define a mechanical force slope-gain g _f (x), where g _f (x) = dF _L (x)/dF _w (2L), again a differential gain, where F _L (x) is the force acting on the moving load at position x, and F _W (2L) is then the tension in the wire and the wire-length 2L. Similarly, the mechanical force bulk-gain G _f (x) is given by G _f (x) = F _L (X)/F _W (2L). For a simple straight-wire SMA actuator (where the load is moved directly by the free-end of the SMA-wire), then g _s (x) = g _f (x) = G _s (x) = G _f (x) = 1 (within a sign term). However, for the bowstring actuator, and other more complex forms of actuator, none of these quantities is a constant, nor are they the same as each other, and all are functions of x (the actuator output displacement).

A given SMA-wire will have some maximum safe operating load (tension) F _wmax , repeated application above which load the wire properties will change, the actuator cease to operate according to specification, and if greatly exceeded, the wire will fail. The intent is to show how to drive the wire’s temperature-time profile so as to move the load as quickly as possible, whilst at all times keeping the tension in the wire below or at the safe value F _wmax . To operate any SMA-wire actuator load at the greatest possible speed, we need to create the largest possible acceleration in the load at all times during the stroke, which implies that the load force is always as great as possible, which in turn requires the SMA wire to be operating at maximum safe operating tension F _wmax throughout the stroke. This can also be used to show how to control the electrical drive to the wire to achieve this optimum temperature-time profile. In the following analysis we describe all the factors that need to be taken into account to achieve optimum high-speed drive. In practice, F _wmax will not be a constant, but will at least be a function of wire temperature, i.e. F _wmax = F _wmax (T). In the analysis that follows we treat F _wmax as a constant, but the analysis may be modified suitably by introducing F _wmax (T) instead of F _wmax , and such an enhanced extension of the analysis and results thereof are included in the present invention.

The material of the SMA wire has a specific heat H _s (typically H _s ~3200J/kg.K for NiTi SMA), and latent heat of transformation (from martensite to austenite) H _L (typically H _L -24300 J/kg for NiTi SMA). For simplicity we assume that during the active stroke of the actuator, the percentage shortening of the SMA-wire is directly proportional to the percentage of martensite in the initial (cool) wire converted into austenite in the subsequently hot(ter) wire. This is a reasonable approximation and better

approximations may be used instead to improve the result, without significantly changing the scheme of the result, or the nature of this invention.

Let the electrical resistance of the SMA-wire at the austenite start temperature A _s be R ₀ , and again for clarity, let the initial temperature T ₀ = A _s . This simplifies the analysis but leaves out nothing crucial, since below temperature A _s the wire changes length by only very small amounts so there are no significant dynamics involved. Thus for practical purposes the wire may be heated up to temperature A _s arbitrarily quickly and the inertial effects will generally not be damaging.

The length of the SMA wire at temperature A _s and under tension F _wmax will be 2L ₀ , and let the length of the SMA wire at temperature A _f and under tension F _wmax be 21^ ( < 2l_o), and define the whole-wire-stroke S _w therefore to be S _W =2L ₀ - 2Li .

The latent heat of the whole wire is H _WL = M _W .H _L while the wire’s thermal capacity is H _w s = M _W .H _S

The total temperature range T _t of the martensite to austenite transformation is by definition T _t = A _f - A _s . A small temperature change dT (within the range A _s to A _f ) will convert a fraction dT/T _t of the wire into austenite and will produce a shortening of the wire of S _w .dT/T _t , where S _w is the wire-stroke as defined above. This in turn will cause a movement of the actuator load of g _s (x).S _w .dT/T _t . The maximum actuator output force F _max — G _f (x) F _Wmax -

The push-rod (if any) and actuator load necessarily have to slide on some bearing surface (unless the load is hanging freely & vertically in space), and be supported by that surface with a normal force component F _n . Let the coefficient of sliding friction between the push-rod/load and its bearing surface be m, and the frictional force induced under movement be F _f where

F _f = m-F _n. In the limit where the push-rod and sliding load are horizontal and all their weight is supported by the bearing, the maximum value of F _f is given by F _f = F _fmax = Ju.M _L .g (g is acceleration due to gravity, -9.81 m/s/s).

The result of interest is high-speed actuation, and therefore it can be assumed that the heat lost from the wire to ambient (at temperature T _am ) during the actuation period is negligible, but this turns out not to be true, especially if the wire is pre heated to temperature - A _f prior to the high-speed stroke because this is often enough to allow convection to come into play, so we will account for that here as well. The SMA-wire has a heat transfer coefficient h (e.g. measured in [W/K]) which is a function of wire length, diameter and orientation. This may be determined either by measurement for a particular wire configuration, or by calculation, Strictly, h is itself a function of temperature T, h (T), because for very small DT temperature rises above ambient , the dominant heat loss is conduction not convection and h is small, whereas for larger DT convection dominates and h is larger, while for very large DT radiation may dominate. However, ignoring this for simplicity of description and regarding h as a constant simplifies things and gives good accuracy, but the variation of h may be included in the analysis that follows if higher accuracy is needed.

It is assumed that the mass of the wire itself M _w is negligible in comparison to the total moving mass. If this last assumption is not valid then to a good approximation, the wire mass M _w can be incorporated as an addition 6M _L to the moving load mass M _L where 6M _L ~ V ₄ M _w / g _s (x) ² (because the average velocity of the wire is ~ _½ g _s (x) times smaller than that of the load).

Operating the actuator as fast as possible without overloading the wire, requires that the wire is operated at maximum tension F _wmax (but no more) which results in an actuator output force of F _L (x) = G _f (x).F _wmax . Thus the force remaining available to accelerate the load mass M _L is F _L (x) - F _f , and this will in general vary throughout the high-speed stroke (if G _f (x) or F _wmax is not constant) giving rise to complex non- constant-acceleration dynamics. Thus, the acceleration a(x) of the load (& any push- rod) is given by

a(x)— (G _f (x) . F _wmax F _f )/M _L

According to the assumptions, injecting a small amount of heat energy dE into the wire at temperature T, in short time interval dt, will do the following:

a) Raise the wire temperature by dT

b) Convert a fraction dT/T _t of the wire into austenite;

c) Shorten the wire by 2dl_ = S _w dT/T _t ; (L is the wire half-\ength)

d) Move the load mass by a distance dx = g _s (x).2dl_ = g _s (x). S _w dT/T _t ;

e) accelerate the load mass by a(x) = dv/dt from velocity u to velocity v = u+du, f) supply the heat lost to ambient in time dt;

each of a), b), e) and f) will require energy:

the temperature change requires H _ws .dT

the transformation to austenite requires H _WL .dT/Tt

the load mass acceleration requires 0.5.M _L .(v ² -u ² ) = 0.5.M _L .(2.a(x).dx) — M _L .a(x).g _s (x). Sw-dT/T _f — (G _f (x).F _wmax - F _f ).g _s (x). Sw-dT/T _f the heat loss to ambient is fj.(T-T _am ).dt, where h is the heat transfer coefficient for the wire (primarily due to conduction, and perhaps convection if there has been enough time with T » T _amb for convection to develop); The energy balance equation is therefore:

dE = H _ws .dT + H _WL .dT/Tt + (G _f (x).F _wmax - F _f ).g _s (x). S _w .dT/T _t + h.( T - T _amb ).dt

and thus:

dE/dt = Hws dT/dt + H _w JTt .dT/dt + (G _f (x).F _wmax - F _f ).g _s (x). S _w /T _t .dT/dt + h.{ T - T _amb ) Rearranging we get

dT/dt = (dE/dt - h.(T T _amb ))/(H _ws + HwJTt + (G _f (x).F _wmax - F _f ).g _s (x). S _w /T)

i.e. for a given heating power, dE/dt , this equation tells us the rate of temperature rise of the wire, dT/dt.

For electrical drive purposes we also need an accurate estimate of the wire resistance under dynamic load conditions. It is known in the art (see e.g. Ref.2) that the resistance R(T, å), where T is the wire temperature and å is the wire stress, can be expressed as

R(T, å) = R _M .(1-Q(t)) + R _A .Q(t)

where Q(t) is the phase fraction of austenite at time f, while R _M and R _A are calculated as functions of temperature and stress as follows:

RM ⁼ RM(T, å) = R _O M + (T-T _O M) 6RM/5T + å 6RM/5å

R _A = R _A (T, å) = R _OA + (T-T _0A ) 6R _A /6T + å 6R _A /6å

6R _M /5T and 6R _A /6T are just the temperature coefficients of resistance b _M and b _A in the martensite and austenite states respectively, with approximate values:

b _M = 0.001209 /°C and b _A = 0.0001664 /°C.

6R _M /5å and 6R _A /6å are the stress coefficients of resistance with experimental values:

6R _M /5å = 1.616 x10 ⁴ , and 6R _A /6å = 8.091 x10 ⁵ .

T _OM and T _0A are the temperatures at which the fully martensite and fully austenite phase resistances are measured, respectively (i.e. T _0M <= Mf <= As, and T _0A >= Af >= Ms).

However, because in what follows we operate the SMA wire at approximate constant stress (in fact, at near maximum allowable stress), then for the time interval of interest we can safely ignore terms in 6R _M /5å and 6R _A /6å. This leaves somewhat simplified equations:

RM ⁼ RM(T) = ROM (1 ⁺ (T-TOM) bM)

R _A = R _A (T) = R _oa .(1 + (T-T _0A ) b _A )

giving the final result for the wire resistance:

R(T, å) = R(T) = R _om {1 + (T-T _O M) b _M }{1-0(1)} + R _0A {1 + (T-T _0A ) ^Q{t) Equivalently, we can just treat the wire resistivity p =p(T) as being made up as:

p(T) ={1-Q(t)}.p _M (T) +Q(t).p _A (T) , where p _A is the austenite resistivity and p _M is the martensite resistivity, and so:

P(T) = POM{1 ⁺ (T-TOM) _M }{1-Q(t)} + POA{1 ⁺ (T-T _0A ) _A }Q(t)

For a practical solution:

the dynamics are computed as follows, where many quantities are potentially functions of time indicated by the form f(t) \ [the dimensions of the quantities are shown at right;] the net force on the load is F _L (t) = G _F (L(t)). F _wmax - F _f [kg.m.s-2] the load acceleration is a(L(t)) = F _L (t)/M _L

[m.s-2] the load velocity is v(t) = jF _L (t)/M _L dt, integrated from 0 to t [m.s-1] the load position is x(t) = ifF _L (t)/M _L dt ² , integrated from 0 to t [m] the wire length is 2L(t) = L _tot - x(t)/G _s (x(t)) [m] the wire stroke is S _w (t) = L _tot - 2L(t) = x(t)/G _s (x(t)) [m] the fraction of wire full-stroke is

Q(t) = x(t)/{G _s (x(t))./i.L, _ot }, 0£Q(t)£1 . equnGO

[1] and then the thermals are computed as follows:

the austenite phase fraction of the wire is Q(t) [1] the wire temperature is T(t) = As + Q(t).T _t

. equnGI

[K] the energy required to heat the wire mass to T(t) from As is E _h (t) = H _ws .( T(t) - As) [J] the energy required to transform the austenite phase fraction is E _t (t) = H _WL Q(t) [J] the energy required to accelerate the load mass to velocity v(t) is E _k (t)= ½ M _L .v(t) ² [J] the power required at time t to counter heat lost to ambient is P _a (t) = Ai.{T (t) - T _amb }[W] the energy required to counter heat lost to this point in the stroke is

E _a (t) = f h.{T(t) - T _amb } dt, integrated from 0 to t [J] the total energy required to this point in the stroke is E(t)=E _h (t) +E _t (t) + E _k (t)+E _a (t) [J] the total heating power required at time t is P(t) = d E(t)/dt . equnG2 [W] and finally the electrical drive requirements are computed as follows:

the wire resistivity at time t is p(t) ^« p _0M {1 + (T-T _0M ) b _M }{1-0(ί)} + P _OA {1 + (T-T _0A ) b _A 0(ί) the wire resistance at time t is R(t) = 2 L(t) p(t)/A _w , . equnG3 where A _w is the wire cross-sectional area

the electrical heating current is l _h (t) = V{P(t)/R(t)} . equnG4

the electrical drive voltage is V _h (t) = V{P(t).R(t)} . equnG5

These equations are tedious to solve analytically for non-linear actuators (e.g. the bowstring actuator) but trivial to solve numerically. For linear actuators (e.g. the simple straight-wire actuator without any mechanical leverage) a straightforward analytical solution is possible.

The end result is that if one has any sort of SMA wire-actuator, then driving it with a current l _h (t) as defined above (where the SMA is electrically conductive) or more generally heating at a rate defined by equnGI, T(t) = As + Q(t).T _t , will cause it to actuate as fast as is practically possible without subjecting the SMA wire to mechanical stress overload.

So it can be seen that a generic SMA-wire actuator with parameters as defined above is controllably heated at a rate T(t) as defined by equation G1 ( equnGI above). In a similar aspect, power P(t) is controllably delivered to the wire of a generic SMA-wire actuator at a rate defined by equnG2. In another similar aspect, the resistance of the wire R(t) of a generic SMA-wire actuator is controlled to change at a rate defined by equnG3. Preferably the wire of an SMA-wire actuator is heated electrically by an electrical heating current l _h (t) at a controlled rate defined by equnG4, or alternatively by a drive voltage at a controlled rate V _h (t) defined by equnG5. Each of these aspects of the invention are designed to produce a controllably optimal fast actuation time while avoiding over-stressing the SMA-wire.

It is important to realise that the equations above can be used to compute the control values for a completely general transition of the actuator at highest possible safe speed from any starting position to any finishing position x ₂ , where 0 <= C _Ϊ <= x ₂ <= S, in the direction of the actuation or heating half-cycle of the actuator (as opposed to the de-actuation or cooling half cycle).

We next examine how this result relates specifically to the simple straight-wire SMA actuator. The total stroke S = k. L _tot , where L _tot is the length of the whole SMA-wire.

When the whole SMA-wire length is at intermediate length 2L (so Lh < L < Lc) then

x = U _ot - 2L, so dX/dL = -2 ; The straight-wire actuator mechanical stroke slope-gain g _s (x) = dx/d (2L) = ½ dx/dl_

= -7; (the negative sign merely indicates that as the wire shortens, x increases, but one may choose to define this as +7 without loss of usefulness))

The straight-wire actuator mechanical stroke bulk-gain G _s (x) where G _s (x) = x/(L _tot - 2L)

= x/x = 7 ;

The straight-wire mechanical force slope-gain g _f (x), where g _f (x) = dF _L /dF _w = 1 , because the force experienced by the wire is the exact same force experienced by the load.

The straight-wire mechanical force bulk-gain G _f (x) is given by G _f (x) = F _L (X)/F _W (2L) =1 , for the same reason. These gain factors all being constant greatly simplifies the analysis.

Then:

the net force on the load is F _L = F _wmax - F _f

the load acceleration is a(t) = F _L /M _L

the load velocity is v(t) = JF _L /M _L dt = F _L t/M _L

the load position is x(t) = JiF _L /M _L dt ² = F _L t ² /M _L

the wire length is 2L(t) = L _tot - x(t)

the wire stroke is S _w (t) = L _tot - 2 L(t) = x(t) = F _L t ² /M _L

the fraction of wire full-stroke is Q(t) = F _L t ² /(M _L . c. L _tot ) = x(t) /(Ac. L _tot ) [1] and then the thermals are computed as follows:

the austenite fraction of the wire is Q(t)

the wire temperature is T(t) = As + T _t F _L t ² /(M _L . Ac. L _tot ) . equnSI

[K]

the energy required to heat the wire mass to T(t) from As is

E _h (t) = H _ws . T _t F _| _ t ² /(M _L . Ac. L _tot ) [J] the energy required to transform the austenite fraction is E _t (t) = H _WL F _L t ² /(M _L .Ac. L _tot ) the energy required to accelerate the load mass to velocity v(t) is

E _k (t)= ½ M _L .(F _L t/M _L ) ² = ½ F _L ² t ² /M _L [J] the power required at time t to counter heat lost to ambient is

P _a (t) = As + T _t . F _L t ² /(M _L . AC. L _tot ) - T _amb } [W] the energy required to counter heat lost to this point in the stroke is E _a (t) = f h.{As + T _t . FL t ² /(M _L . C. L _tot ) - T _amb } dt,

= h.{( As - T _amb )t+T _t . FL t ³ /(3M _L ./i.L _tot )} [J] the total energy required to this point in the stroke is E(t)=E _h (t) +E _t (t) + E _k (t)+E _a (t)

=(F _L t ² /M _l ){ (H _ws . T _t + H _WL )/(/c.L, _ot )+½ F _L + Ai.T _t . t/(3/c.L _t0 ,)}+ h.(As - T _amb )t the total heating power required at time t is P(t) = d E(t)/dt

equnS2

[W] and finally the electrical drive requirements are computed as follows:

the wire resistivity at time t is p(t) ^« F _L t ² /(M _L ./i.L _tot ).p _A + {1-F _L t ² /(M _L ./i.L _tot )}.p _M

p(t) = POM{1 + (T-TOM) bwKI-F _L t ² / ( M _L . c . L _tot )}

+ POA{1 + (T-TOA) A}F L t ² /(M _L ./r.L _tot )

the wire resistance at time t is

R(t) = (Ltot - FL t ² /M _L ) p(t)/A _w . equnS3 and finally:

the electrical heating current is l _h (t) = V{P(t)/R(t)} . equnS4 the electrical drive voltage is V _h (t) = V{P(t).R(t)} . equnS5 both of these last equations can be made explicit using the above relations.

Of particular interest is equnS3 above for the wire resistance:

(t) = {L,ot - FL t ² /M _L } p(t)/Aw = {Ut - x(t)} p(t)/A _w

= {Uot - x(t)}[ POM{1 + (T-TOM) b _M }{1- x(t) /(Ar. U t)} + POA{1 + (T-T _0A ) b _A } ^c (ί) /(Ac. L _tot )]/Aw which has terms in x(t) ² for R(t).

So in order to produce an optimally fast actuation time while avoiding over-stressing the SMA-wire, a simple straight-wire SMA-wire actuator with parameters as defined above is heated at a rate T(t) as defined by equation S1 ( equnSI above). In a similar aspect, power P(t) is delivered to the wire of a simple straight-wire SMA-wire actuator at a rate defined by equnS2. In another similar aspect, the resistance of the wire R(t) of a simple straight-wire SMA-wire actuator is controlled to change at a rate defined by equnS3. Preferably the wire of a simple straight-wire SMA-wire actuator is heated electrically by an electrical heating current l _h (t) at a rate defined by equnS4, or alternatively by a drive voltage at a rate V _h (t) defined by equnS5. Each of these aspects of the invention are designed to produce an optimally fast actuation time while avoiding over-stressing the SMA-wire.

We now examine how the general result above relates specifically to the bowstring-actuator previously described, which introduces several non-linear terms.

Ref.1 states that for the bowstring actuator, the stroke S = A - B, which using the definitions therein gives:

S = V(LC ² -Y ² ) - V(l_h ² -Y ² ), where all terms are constants.

When the whole SMA-wire length is at intermediate length 2L(t) (so Lh £ L(t) £ Lc) then

x(t) = V(LC ² -Y ² ) - V{L(t) ² -Y ² }

and L(t) = V[{V(l_c ² -Y ² ) - x(t)} ² + Y ² ]

The bowstring actuator mechanical stroke slope-gain g _s (x) = dx/d (2L) = ½ dx/dL

= ½ (-2L(t))/V{L(t) ² -Y ² } = -L(t)/V{L(t) ² -Y ² }

Because for this actuator the actuation direction is orthogonal to the average wire-direction, the sign used for x is arbitrary. So for convenience here we drop this minus sign and use g _s (x) = g _s (x, t) = L(t)/V{L(t) ² -Y ² } = V[{V(l_c ² -Y ² ) - x(t)} ² + Y ² ] / V{L(t) ² -Y ² }

The bowstring actuator mechanical stroke bulk-gain G _s (x) where G _s (x) = x(t)/(L _tot - 2L(t))

=G _s (x, t) = [ V(LC ² -Y ² ) - V{L(t) ² -Y ² }]/(L _tot - 2L(t))

and it will be seen that for this actuator G _s (x) ¹g _s (x), and neither is a linear function of the variables. I.e. the stroke gains vary throughout the stroke.

The mechanical force slope-gain g _f (x) = dF _L (x)/dF _w (2l_). The mechanical force bulk-gain G _f (x) is given by G _f (x) = F _L (X)/F _W (2L). Both these gains are easily derivable from the above equations for the bowstring actuator, and both are functions of x(t).

We can plug these non-constant values for the gains into general equations equnGO through equnG5 to derive bowstring-actuator specific equations equnBO through equnB5 respectively, for Q(t), temperature, power, resistance, current and voltage as functions of time. We have solved these equations numerically for convenience.

So in order to produce an optimally fast actuation time while avoiding over-stressing the SMA-wire, a bowstring SMA-wire actuator with parameters as defined above is heated at a rate T(t) as defined by equnBI described above. In a similar aspect, power P(t) is delivered to the wire of a bowstring SMA-wire actuator at a rate defined by equnB2. In another similar aspect, the resistance of the wire R(t) of a bowstring SMA-wire actuator is controlled to change at a rate defined by equnB3. Preferably the wire of a bowstring SMA-wire actuator is heated electrically by an electrical heating current l _h (t) at a rate defined by equnB4 , or alternatively by a drive voltage at a rate V _h (t) defined by equnB5. Each of these aspects of the invention are designed to produce an optimally fast actuation time while avoiding over stressing the SMA-wire of a bowstring actuator.

Analysis for a Bi-Directional High-Speed Actuator

As noted earlier, it is challenging to design an SMA actuator with a high-speed stroke in both directions because of the relatively slow cooling time of the SMA material. Active cooling can help (see e.g. Ά Novel Differential Shape Memory Alloy Actuator for Position Control', by Gorbet RB & Russell A), but it’s relatively clumsy, complicated and expensive and detracts from the primary advantages of SMA actuators. Where continuous operation at high speed is not required, but where single-shot complete cycles in the shortest possible time are needed, then a solution can be provided as follows.

Here we combine the optimal high-speed drive (as described above) of the SMA-wires in a dual differential or opposed SMA-wire actuator, with the SMA-wire mechanical overload pre stressed spring overload protection mechanisms described above, and make use of the fact that SMA wires can frequently provide more force than is needed to actuate the load.

Consider first a pair of opposed in-line simple straight-wire SMA actuators each with one end fixed to a base and each with the other end attached to an output-member via a pre-stressed spring device as described above for overload protection. The output-member may be a discrete component, or may be the actuator output-load itself. With this arrangement one of the wires WireP can pull the load in the positive x-direction, and the other wire WireN can pull the load in the negative x-direction. It is desired to apply an accelerating force £F _P|US to the output-member in the positive direction, and an accelerating force £F _minus in the negative direction, with a total actuator output stroke of length S. In what follows we ignore friction acting in the x-direction but that is easily taken into account as described below. Thus the pre-stress on the overload protection spring of WireP has magnitude at least F _P|US so that it will not change length during the positive acceleration stroke induced by heating WireP above temperature A _s at a rate no faster than described above for safe-wire operation, while WireN is kept below temperature A _s .

The operation of the high speed stroke is as follows:

The load is positioned at nominal zero displacement position, which is the static position to which WireN when fully actuated would pull the load in the absence of dynamic forces or any force from WireP. Both wires are originally at ambient temperature, or alternatively one or both of them may be pre-heated to temperature less than or not much more than A _s - this latter has two useful effects but uses more power: firstly, there is smaller or no preheat delay required to get the wire or wires to temperature A _s at which temperature the strokes begin; secondly, temperature A _s may be high enough above ambient to start a significant convection flow around the heated wire(s) which will decrease the wire cooling time after the heating phase ends, and thus allow a faster cycle repetition rate.

Next the positive acceleration stroke is induced by heating WireP above temperature A _s at a rate no faster than described above for safe-wire operation, and possibly all the way to temperature A _f , while WireN is kept below or not much higher than A _s . Neither spring means will extend during this positive stroke if the forces produced by WireP remain below the prestress levels.

After WireP has completed its high-speed stroke in the positive x-direction (accelerating the load with force £ F _P|US throughout the stroke) with the heating power preferably controlled as described above for optimal high-speed actuation, to whatever position is desired, and after any desired dwell time which may be zero, heating power may be removed from WireP, or at least reduced, though this is not necessary in what follows.

At this point heating power is applied to WireN at a rate no faster than described above for highest-speed safe-wire operation, causing WireN to begin shortening and pulling the output- member in the negative x-direction with a net accelerating force £F _minus · In order to do this it must overcome the spring force of the overload protection spring of WireP which is initially at level ~ F _pius and which increases at a rate K _I (the spring-rate of WirePs protection spring) as the WireP spring means stretches throughout the return stroke.

When the output-member has returned to the origin, the total stretch of this spring can be as great as the stroke S so the maximum force it can apply to the output-member is:

Fi _max - F _p|us +Ki .S.

Thus, WireN must be capable of safely producing at least this much pulling force, and WireN' s overload protection spring must be pre-stressed to at least F _{1 max} . WireP meanwhile must also be able to pull with at least this same force to maintain its fully austenite-phase position as the return stroke may have completed before WireP' s temperature has fallen below temperature M _s . So the specification for the two actuator wires is stroke ³ S, and output force in the fully actuated state ³F _{1 max} . And the specification for the wire-protection springs is: for that of WireP, prestress ~ ³ F _P|US and spring constant Ki as described below for SMA wire-protection; for that of WireN, prestress ~ ³ F _{1 max} and spring constant K ₂ again as described below for SMA wire-protection. Note that the spring means prestress values are only approximate because if e.g. for WireP, prestress is slightly less than F _P|US then the actuator will still function more or less as designed, because much of the actuator output force is needed to overcome inertial forces and having a slightly softer spring slightly reduces the acceleration achievable and so slightly reduces the highest speed of operation. Note also that in normal operation of a non-overloaded high-speed bidirectional stroke as described, WireN's protection spring will not stretch at all, so it is not strictly necessary - only the spring means on WireP is necessarily required for the mechanism to work as described. However, because the nature of this actuator arrangement is to drive the wires close to their safe limits, the extra protection provided by an overload protection string on WireN is well worth including since it guarantees the safe operation of both wires even in overload conditions.

So in this arrangement an actuator is comprised of two SMA-wire actuators each

mechanically connected to a mechanical load so as to drive the load in opposite directions when each is actuated, wherein the mechanical connection to the load of at least one of them is via pre-stressed spring means. Preferably the magnitude of the prestress force F _PP of the at least one pre-stressed spring means is substantially equal to the maximum designed accelerating force to be applied to the load, or approximately F _P|US . Preferably the spring constant KI of the at least one pre-stressed spring means is as defined by equnPI using F _PP for the value of F _wmax , so

Ki = G _s (S). F _PP /S . equnHI

Preferably the mechanical connection to the load of both of SMA-wire actuators is via pre stressed spring means. Preferably the magnitude of the prestress force F _pm of the second pre-stressed spring means is given by

F _pm ~ Fi _max — F _p|US +Ki .S . equnH2

Preferably the spring constant K ₂ of the second pre-stressed spring means is as defined by equnPI using F _{1 max} for the value of F _wmax , so

K ₂ ^— G _s (S).F _{1 max} /S . equnHS

With this arrangement a very high-speed complete bi-directional actuator cycle is possible while protecting both SMA-wires from overload. The only“cost” is that actuator wires with higher force output than required to accelerate the load alone must be used. Complete actuator cycles (end to end and back again) faster than 2ms are possible in this way. Note however that the cycle repetition rate is still limited by the cooling rate of the SMA wires. Thus the repetition rate can be increased by any of the methods known in the art for actively increasing the cooling rate of SMA wires, including heat-sinks and forced fluid cooling. A similar arrangement to the above for a bidirectional high-speed actuation cycle can be used with a pair of any generic SMA wire actuators, instead of a pair of opposed straight-wire actuators, and in particular with a double-bowstring actuator. In this latter case the protection springs are preferably attached between each of the wire-centre regions and the output- member (as described below for overload protection) with a similar calculation performed to determine the required spring pre-stresses and spring constants. Note that in both cases (straight-wire and bowstring actuators) the additional force requirements for high-speed full- cycle operation may be met by using 2 or more SMA-wires in parallel rather than a single thicker wire, in order not to increase the wire-cooling time significantly, if the highest possible repetition rates are to be achieved.

In any of the above aspects of the present invention a control system may be provided capable of measuring the position of the actuator output by the use of suitable position sensor means. In a preferred embodiment the position sensor means are implemented by electrical resistance measurement of one or more of the SMA-wires of the actuator embodiments, and from the measurement(s) of resistance said control system estimates the instantaneous length(s) of the SMA-wire(s), and from these length estimates together with knowledge of the specific actuator mechanical configuration and geometry the controller is capable of estimating the output position of the actuator and thus is able to accurately position the actuator output.

Figures 28, 29, and 30 show the analytically modelled results for an embodiment of a simple straight- wire SMA actuator optimised for high-speed drive as in one aspect of the present invention, comprised of 17mm of 25micron diameter SMA wire, and pushing a load of mass of 225mg and a frictional force of 0.011 N. In figure 28 the vertical scale shows output stroke in mm (0.0 to ~ 0.7mm) while the horizontal scale is time from initiation of stroke, in milliseconds. It will be seen that full stroke is achieved in around 1 ms, while keeping the 25micron wire within its maximum safe tension limits at all times. In figure 29 the drive current [mA] on the vertical scale against time [ms] on the horizontal scale is shown for maximum safe actuation speed. It will be seen that a non-linear current ramp is required (as defined by the appropriate equations above) to keep the wire within safe operating limits. The form of the non-linearity is a quartic, even for the simple straight-wire actuator. In figure 30 the wire resistance [ohm] is plotted against the actuator stroke [mm] and this too will be seen to be non-linear, having a quadratic form.

Figures 31 , 32, 33, and 34 show the analytically modelled results for an embodiment of a single bowstring SMA-wire actuator optimised for high-speed drive as in one aspect of the present invention, comprised of 17mm of 25micron diameter SMA wire, with a crimp separation of 16mm, and pushing a load of mass of 225mg and a frictional force of 0.011 N. In figure 31 , the vertical scale shows output stroke (or distance moved by the actuator load) in mm (0.0 to ~ 1.2mm) while the horizontal scale is time from initiation of stroke, in

milliseconds. It will be seen that full stroke is achieved in around 2.5ms, while keeping the 25micron wire within its maximum safe tension limits at all times. The greater stroke of the bowstring actuator, than the straight-wire actuator, comes at the cost of smaller available load accelerating force, resulting in a longer stroke time. In figure 32 the drive current [mA] on the vertical scale against time [ms] on the horizontal scale is shown for maximum safe actuation speed. It will be seen that again a non-linear current ramp is required (as defined by the appropriate equations above) to keep the wire within safe operating limits. The form of the non-linearity is approximately quartic. In figure 33, the load acceleration [m/s/s] is plotted against time [ms] and this shows how the available acceleration falls as stroke increases, primarily due to the increasing stroke gain of the bowstring actuator as the wire straightens. Finally in figure 34 we show the drive voltage (for optimum safe high speed drive) in volts against time in milliseconds, and this can be seen to turn over and start to reduce after the first ~2ms due to the falling resistance of the wire and the gain increase with stroke. These plots (figures 31 to 34) clearly demonstrate the efficacy of the high-speed SMA-actuator design technique, achieving very short half-cycle times, and when applied similarly to a bidirectional high-speed actuator as described herein or a double bowstring actuator of design as per the present invention, also very short whole-cycle times too in the low millisecond range.