Three-Axis Accelerometer Having Dynamic Bias Compensation

This application is related to the following concurrently filed applications, which are incorporated herein by reference: (1) Shutt, "Three-Axis Accelerometer," U.S. patent application Ser. No. 292,460; (2) Shutt, "Electromagnet Configuration for Three-Axis Accelerometer", U.S. patent application Ser. No. 292,550; (3) Shutt et al., "Three-Axis Accelerometer Having Improved Magnet Configureation," U.S. patent application Ser. No. 292,544; (4) Shutt, "Three-Axis Accelerometer Having Flexure Bearing With Overload Protection", U.S. patent application Ser. No. 292,543; and (5) Shutt, "Suspension for Three-Axis Accelerometer", U.S. patent application Ser. No. 292,541. All of the applications listed above are assigned to Rockwell International Corporation, the assignee herein.

1. Field of the Invention

The present invention relates generally to accelerometers and more particularly pertains to accelerometers in which a proof mass is constrained by a combination of electromagnetic and elastic supports.

2. Description of the Prior Art

Electromagnetic force rebalance accelerometers are well known. In this type of accelerometer, a proof mass is constrained by force generated in a current-carrying conductor or coil located in a magnetic field. Such accelerometers have typically been built as single-degree-of-freedom instruments. Thus, three such accelerometers are required to be included in an inertial navigation system. Such a system obviously tends to be more costly and complex than would be one which performed its function using fewer instruments. Furthermore, in such a single-degree-of-freedom accelerometer, a structural member of substantial size and mass is typically required to center and maintain the alignment of the coil at the full rated acceleration of the instrument in two of the three directions. This structural member, however, tends to augment the null drift or bias uncertainty of the accelerometer, primarily because of dimensional changes which result from ambient temperature variations. In order to reduce this bias uncertainty, accelerometers intended for use in high-performance navigation systems are typically provided with temperature control elements for holding the temperature-sensitive portions of the instrument at a relatively constant temperature. The need for temperature control in accelerometers is another factor tending to increase the cost and complexity of inertial navigation systems.

It is an object of this invention to overcome the above-noted shortcomings of the prior art.

Another object of the invention is to provide an accelerometer in which bias uncertainty due to temperature variations is greatly reduced.

Still another object is to provide an accelerometer which can be used over a wide range of ambient temperature variations without a need for temperature control.

A further object is to provide an accelerometer capable of measuring acceleration simultaneously along the three orthogonal axes of a Cartesian coordinate system.

A still further object of the invention is to provide an accelerometer having high accuracy.

Another object is to provide an accelerometer which is inexpensive to fabricate.

Yet another object is to provide an accelerometer which is suitable for use in a strapdown inertial navigation system.

Still another object is to provide a single accelerometer which provides all of the acceleration information needed for an inertial navigation system.

A further object is to provide an accelerometer wherein residual bias uncertainty is greatly reduced.

According to the present invention, the foregoing and other objects are attained in an accelerometer having a proof mass supported and constrained relative to the accelerometer case or housing by forces generated electromagnetically in each of three orthogonal directions. A magnet is disposed to establish a constant magnetic field across a gap. Coils for carrying electric current are disposed in the gap. When appropriate coils are selectively energized, sufficient force can be applied to the proof mass along each of the three axes of a set of Cartesian coordinates to keep the proof mass substantially centered at its null position. Each of three distinct currents in the coils is associated with and is a measure of the force applied along a corresponding distinct one of the orthogonal axes to maintain the proof mass centered. Thus, each current is also a measure of the acceleration along the corresponding axis.

The magnetic suspension described above provides constraint for the proof mass in three degrees of freedom. To provide constraint for three additional degrees of freedom, the proof mass is also constrained or balanced by an elastic suspension. The magnetic suspension is designed to supply the overwhelming preponderance of the force required to balance or confine the proof mass. Therefore, the elastic suspension is called upon to provide relatively very small forces. In the preferred embodiment, the elastic suspension is an array of very fine filaments preloaded in tension at a force equivalent to an acceleration significantly less than the acceleration required to be balanced by the magnetic suspension. The tendency of this elastic suspension to cause bias uncertainty in the accelerometer due to the dimensional changes resulting from variations in temperature or other causes is thus reduced over that in a prior art single-axis electromagnetic force balance accelerometer having the same range of operation. There is a potential reduction in bias uncertainty by a factor equal to the ratio of the larger load balanced by the magnetic suspension to the smaller load balanced by the elastic suspension.

For example, in one preferred embodiment, the magnetic suspension is capable of balancing the proof mass at linear input accelerations to .+-.15 g. In this case, the filaments need carry only small loads equivalent to 0.5 g. The force on the elastic suspension is a maximum of 400 dynes at 15 g's input acceleration as compared with 30,000 dynes of force required to be balanced by the magnetic suspension. Any tendency of this elastic suspension to cause bias uncertainty in the accelerometer due to variations in temperature or other causes is potentially reduced by a factor of thirty, the ratio of 15 g to 0.5 g, over that in a single-axis electromagnetic force balance accelerometer having the same range of operation.

This reduction in null drift or bias uncertainty in an accelerometer in accord with the invention makes it particularly suitable for use in strapdown navigation systems with gyroscopes having a similarly low temperature sensitivity.

The elastic suspension for the proof mass of an accelerometer in accord with the invention is an array of filaments connecting the proof mass to the accelerometer housing. In the preferred embodiment, the filaments are disposed in a plane perpendicular to the longitudinal axis of the proof mass. Such a filament array is relatively compliant in three degrees of freedom in that relatively small forces, unless balanced, can produce relatively large rotational deflections of the proof mass about its two cross axes and a relatively large translation deflection of the proof mass along its longitudinal axis. In the remaining three degrees of freedom, however, the elastic suspension is relatively stiff. Relatively large forces are required to produce only a small rotational deflection about the longitudinal axis of the proof mass and only small translational deflections of the proof mass along the two cross axes.

Pickoffs, preferably of the capacitive type, are provided in the accelerometer to sense deflection in the three compliant degrees of freedom of the elastic suspension. The pickoffs supply error signals to be fed to amplifiers which, in turn, supply current to the coils in a primary closed-loop force balance system.

An accelerometer in accord with this invention is inexpensive to fabricate because it has a relatively low part count per axis. This makes it simple to assemble. It is a three-axis sensor which is only about as complex as a single-axis unit. In addition, the dimensional tolerances and stabilities required for the individual elements of the structure are relatively relaxed since these factors are not as critical here as in prior-art devices.

In the accelerometer described thus far, the low null drift or bias uncertainty which is achieved results from the novel structure combined with careful selection of materials and studied selection of appropriate dimensions in certain aspects of the structure, as discussed more fully hereinafter. However, there remains a need to reduce the bias uncertainty even further. In an improved version of an accelerometer in accord with the invention, means are provided to greatly reduce the residual bias uncertainty by dynamically compensating or nulling out the bias itself.

The mechanical null point of an accelerometer in accord with the invention is the point at which the proof mass will rest in a zero-g or force-free field with the force-balance control loop deenergized. The electrical null point for the accelerometer is the location of the proof mass in a force-free field with the force-balance control loop energized. In the ideal situation, the mechanical null point and the electrical null point are identical. In the realistic situation, however, there is a a difference between them. In the accelerometer described thus far, the primary force-balance control loop supplies a current component to the forcer coils sufficient to displace the proof mass from the mechanical null point to the electrical null point. This current component is the bias. It is predominantly due to the tension in the filaments of the elastic suspension. It is proportional to the displacement between the mechanical null point and the electrical null point. When sensed, the bias current component cannot be distinguished from forcer coil current due to acceleration.

In an improved version of this accelerometer, an oscillatory signal proportional to the bias is generated by setting up a periodic tension variation in the filaments of the elastic suspension. This signal is detected and amplified in a secondary force-balance control loop. The secondary control loop provides an offset voltage, proportional to the bias, to the primary control loop. This offset voltage substantially cancels and counteracts the bias. The electrical null is shifted toward the mechanical null until they coincide. The bias current is substantially eliminated. The proof mass is thus driven to and maintained substantially at its mechanical null despite dimensional variations due to temperature changes, creep and the like and despite variations in electrical and electronic component parameters due to drift, aging and the like. These variations are the cause of bias uncertainty. The predominant sources of bias uncertainty are changes in the position of the elastic suspension filament mount relative to the pickoffs, changes in the tension of the filaments, and changes in the relative position of the forcer coils with respect to the magnetic field.

Alternatively, and more simply, the signal arising from the periodic variation of filament tension may be used to calibrate or correct the accelerometer output for bias in an open-loop fashion.

In either case, the bias and thus the bias uncertainty can be eliminated or reduced to near nothing without requiring the external application of an acceleration for calibration.

FIG. 1 is a diagrammatic sketch of a proof mass for an accelerometer elastically suspended in accord with the invention.

FIG. 2 is a view in elevation of an assembled accelerometer.

FIG. 3 is a cross-section of the accelerometer of FIG. 2 taken along the line 3--3.

FIG. 4 is a portion of FIG. 3 enlarged to show more detail.

FIG. 5 is a view of electrodes for displacement pickoffs taken along the line 5--5 in FIG. 4 looking in the direction of the arrows.

FIG. 6 is a cross-section taken along the lines 6--6 of FIGS. 2 and 3 showing the disposition of the electromagnetic forcer coils.

FIG. 7 is a cross-section taken along the lines 7--7 of FIGS. 2, 3 and 4 showing filaments for an elastic suspension.

FIG. 8 is a view in elevation of the electromagnetic forcer coils.

FIG. 9 is a sketch showing the longitudinal axis electromagnetic forcer coils interconnected in a closed force-balance control loop.

FIG. 10 is a sketch showing a pair of cross-axis forcer coils interconnected in a closed force-balance control loop.

FIG. 11 is a sketch of an elastic suspension for a proof mass showing a first technique for dynamically nulling bias.

FIG. 12 is a sketch of an elastic suspension for a proof mass showing a second technique for dynamically nulling bias.

Referring now to FIG. 1, there is shown a simplified sketch of a proof mass 10 elastically suspended from a housing or base 12 for an accelerometer in accord with the invention. The elastic suspension connecting the proof mass 10 to the base 12 includes an array of four support filaments 14 disposed in a plane perpendicular to the longitudinal axis, the Z axis, of the proof mass 10. In the example of FIG. 1, the filaments 14 are spaced apart from each other at regular ninety-degree intervals in a plane near an end of the proof mass 10.

The support filaments 14 are preloaded in tension at a force equivalent to an acceleration load less than the rated acceleration load of the accelerometer by a factor of at least about fifty.

In an accelerometer according to the invention, the major proportion of the force required to constrain deflection of the proof mass 10 relative to the base 12 is provided by electromagnetic forcer coils not shown in FIG. 1. The forcer coils are shown in and discussed further in connection with FIGS. 3, 4, 6, 8, 9 and 10. The coils operate in an electromagnetic force balance system. They are capable of being selectively energized to provide a resultant force acting substantially through the center of mass of the proof mass 10 as indicated at 16. A component force can be applied along any one or any combination of the three axes, X, Y and Z, of a Cartesian coordinate system fixed to the proof mass 10. Z is the longitudinal axis of the proof mass 10, as noted above, while X and Y are a pair of cross axes. In FIG. 1, the cross axes X and Y are shown in alignment with corresponding pairs of the filaments 14 for convenience, but this alignment is not a requirement. Since the forcer coils constrain the proof mass 10 in only three degrees of freedom, the elastic suspension of filaments 14 is required to provide restraint in the remaining degrees of freedom and to establish a stable null position.

If the center of mass of the proof mass 10 and the center of action of the electromagnetic forcers are displaced, an additional elastic suspension force will be required to constrain the proof mass. For example, if the center of action and center of mass are displaced by 0.01 cm, an additional elastic suspension force of 200 dynes is required to support the proof mass 10 when there is a 10 g input acceleration normal to the Z axis.

Actually, the elastic suspension of filaments 14 constrains the proof mass 10 in all six degrees-of-freedom. It provides a relatively very stiff translation support along the cross axes X and Y and relatively very stiff rotational support about the longitudinal or Z axis. Thus, relatively large forces can produce only relatively small translational deflections .DELTA.x and .DELTA.y and a relatively small rotational deflection .theta..sub.z. The relatively small translational deflections .DELTA.x and .DELTA.y along the two cross axes are on the order of 3.8.times.10.sup.-8 centimeters per dyne at rated acceleration input. The relatively small rotational deflection .theta..sub.z is on the order of 6.times.10.sup.-7 radians per dyne-centimeter at rated acceleration input. In the remaining three degrees of freedom, the elastic suspension of filaments is relatively compliant. Thus, relatively small forces can produce relatively large rotational deflections .theta..sub.x and .theta..sub.y and a relatively large translational deflection .DELTA.z, along the longitudinal axis of the proof mass 10. The relatively large rotational deflections .theta..sub.x and .theta..sub.y about the two cross axes is on the order of 2.times.10.sup.-3 radians per dyne-centimeter at rated acceleration input. The relatively large translational deflection .DELTA.z along the longitudinal axis is on the order of 2.times.10.sup.-4 centimeters per dyne at rated acceleration input. Thus, at rated acceleration, the translational deflection of the proof mass along the longitudinal axis is greater than the translational deflection along a cross axis by a factor of about 1.9.times.10.sup.4. Similarly, the rotational deflection of the proof mass about a cross axis is greater than the rotational deflection about the longitudinal axis by a factor of about 3.times.10.sup.4. In FIG. 1 the proof mass 10', depicted in phantom, is shown deflected from its null position to illustrate the discussion above. In actual operation of the accelerometer, the electromagnetic force balance system would operate to keep all such deflections negligibly small.

Referring now to FIG. 2, there is shown an assembled accelerometer 20 in accord with the invention. A housing or case 12 for the accelerometer 20 is fabricated from any suitable non-magnetic material such as, for example, aluminum or a ceramic such as beryllia.

The working parts of the accelerometer 20 are shown in FIG. 3 and 4. A magnet ring 24 is mounted on the housing 12. Interior of the magnet 24, coaxial therewith and spaced apart therefrom, is a flux return ring 26 mounted on an inner pole piece mounting member 28. The magnet 25 establishes a constant or time-invariant magnet field across a gap between it and the flux return ring 26. Disposed in the gap is the coil form 30 portion of a proof mass 10. An assembly of electromagnetic forcer coils 34 is disposed on the coil form 30 for applying forces to the proof mass 10 in three orthogonal directions. At an end of the proof mass 10, the left side thereof in FIGS. 3 and 4, the proof mass 10 is flared into a disk portion 36.

The proof mass 10 is suspended from the housing 12 by an array of filaments 14 disposed in a plane substantially perpendicular to the longitudinal axis 38 of the proof mass 10. The outward portion of each filament 14 is clamped between the inner surfaces of a clamping ring 40 and a housing cap 42. The other surface of the clamping ring 40 is fastened to a housing member 43. Housing member 43 is part of the housing 12. Mating grooves in the clamping ring 40 and the housing cap 42 are provided to form a cavity 45 which is directed along a filament 14 and which permits the central portion of each filament 14 to be free of contact with the structure. The inward portion of each filament 14 is clamped between the inner surfaces of a clamping disk 44 and a cover ring 46. The other surface of the clamping disk 44 is fastened to the proof mass 10.

The filaments 14 are preferably formed of a material suitable not only for carrying the required load but a material which is also a good conductor of heat and electricity such as, for example, beryllium copper. Good electrical and thermal conductivity are desired so that the filaments 14 may also serve as conductors for supplying current to the forcer coils in forcer coil assembly 34.

The predominant sources of bias uncertainty in an accelerometer 20 according to the invention are related to a change in the forces exerted by the filaments 14 on the proof mass 10. In order to minimize any change in these forces as temperature variations occur, dimensions of the filaments 14 and the filament mounting parts 40, 42, 43, 44 and 46 and the materials used to form the parts are chosen so that thermal expansion and contraction of the filaments and filament mounting parts over a wide temperature range produces a negligible change, if any, in the tension preload on the filaments 14. The manner in which the filaments 14 are preloaded in tension is discussed below in connection with FIG. 7.

In order to achieve this stability of the preload on the filaments 14 over a wide temperature range, the filament mounting parts 40, 42, 43, 44 and 46 are preferably made of materials that have a high degree of dimensional stability with variations in temperature. Clamping ring 40, housing cap 42 and housing member 43 are preferably made of a ceramic material such as, for example, beryllia. Clamping disk 44 and cover ring 46 are preferably made of a material having only a negligible change in dimensions with changes in temperature such as, for example, Cer-Vit.

In addition, in order to keep the preload on the filaments 14 stable over a wide temperature range, the nominal length of a cavity 45 formed by grooves in the clamping ring 40 and housing cap 42 is selected so that expansion or contraction of the filaments 14 tending to lessen or increase their preload tension is compensated by expansion or contraction of the length of a cavity 45. A cavity 45 expands or contracts due to thermal expansion or contraction in the clamping ring 40 and housing cap 42. As a cavity 45 lengthens or shortens due to changes in temperature, that point at which the outward portion of a filament 14 is clamped between clamping ring 40 and housing cap 42 moves away from or closer to the point at which the inward end of a filament 14 is clamped between clamping disk 44 and cover ring 46. The latter point exhibits negligible change, of course, since clamping disk 44 and cover ring 46 are of Cer-Vit or the like.

An example of thermally caused changes in dimension will be given here for the filament shape shown in FIG. 7. In this example, the nominal length of the unclamped portion of a filament 14 extending through a cavity 45 is 0.40 cm. A line extending from the longitudinal axis 38 to the point at which the outward portion of a filament 14 is clamped between clamping ring 40 and housing cap 42 has a length of 0.73 cm. The angle between said line and said filament is 18 degrees. See the discussion of FIG. 7 for a further explanation of this angle. For a temperature variation from -55.degree. C. to 70.degree. C., the change in the length of the unclamped portion of a filament 14 is 0.0044.times.10.sup.-6 centimeters per degree Celsius. For a filament 14 having a square cross-section of 0.0015 cm on a side, the change in tension is less than four dynes. Since the nominal preload tension is 200 dynes, this change in tension is less than two percent over the temperature range.

When the preload tension in the filaments 14 changes, the accelerometer biases change also. For example, if the filaments 14 are out of plane alignment by 0.001 cm, then the X or Y axis bias uncertainty changes by less than 0.1.times.10.sup.-6 g per degree Celsius and the Z axis bias uncertainty changes by less than 0.5.times.10.sup.-6 g per degree Celsius.

The disk portion 36 of the proof mass 10 serves as a movable plate or electrode for a plurality of capacitor pickoffs for sensing rotational deflection about the cross axes X and Y and for sensing translation deflection along the longitudinal or Z axis of the proof mass 10. The body of the proof mass 10 is excited with an alternating electric potential of, for example, five volts at a frequency of, for example, twenty kilohertz. Pickoff electrodes forming the fixed or immovable plates of sensor capacitors are disposed, preferably by plating, on surfaces of clamping ring 40 and housing member 43 adjacent to the proof mass disk 36. Those are seen more distinctly in FIG. 4 than in FIG. 3. A pair of ring-shaped electrodes 48, one on the clamping ring 40 and one on the housing member 43, form the two fixed electrodes of a capacitor pickoff element for sensing translational deflection along the longitudinal or Z axis of the proof mass 10. Another pair of electrodes 50 is disposed, diametrically opposed to each other, on the clamping ring 40. The electrodes 50 are the two fixed electrodes of a capacitor pickoff element for sensing rotational deflection about a cross axis perpendicular to the plane of the drawing, in this case the Y axis.

FIG. 5 is a plan view of the pickoff electrodes formed on the rear surface of clamping ring 40. One of the two electrode rings 48 for sensing translational deflection of the proof mass 10 is shown. The other ring 48 is on housing member 43 as mentioned above. The pair of sector-shaped electrodes 50 for sensing rotational deflection about the Y cross axis is shown interposed between an additional pair of sector-shaped electrodes 52. The electrodes 52 are for sensing rotational deflection about the X cross axis. Each of the electrode trodes 48, 50 and 52 is spaced apart from the others and is therefore insulated from the others. Each pair of electrodes 48, 50 and 52, respectively, is connected into a bridge circuit, well known to those skilled in the art, for generating a signal proportional to the respective deflection.

It should be noted that, in this accelerometer, the mount for the filaments 14 relative to the pickoff electrodes 48, 50 and 52 has been designed and constructed for maximum stability in the face of temperature variations. This is accomplished by suspending the filaments 14 from the front surface of clamping ring 40 while the majority of the pickoff electrodes are formed on the rear surface of the same single-piece, temperature-stable, ceramic part, i.e., clamping ring 40.

FIG. 7 shows the array of support filaments 14 in more detail. In this embodiment eight support filaments 14 are used. An individual support filament 14 has an outward portion 80, an inward portion 82 and an intermediate portion 84. The outward portion 80 is disposed on the clamping ring 40. The inward portion 82 is disposed on the clamping disk 44. The intermediate portion 84 is the very fine unclamped part of the filament intended to be disposed in the cavity 45 formed by the mating grooves in the clamping ring 40 and housing cap 42.

As is apparent in FIG. 7, the eight support filaments 14 of this embodiment are disposed as four pairs, a pair consisting of each two filaments which are closer to each other than to the remaining filaments. Furthermore, a filament 14 does not extend radially but is directed at an angle relative to a radial line which intersects the filament at the point where the intermediate portion 84 and the outward portion 80 meet, the point at which the filament is clamped between clamping ring 40 and housing cap 42. In this embodiment, the angle between this radial line and each filament 14 is 18 degrees.

The support filaments 14 may be formed from a single sheet of beryllium copper, preferably about 0.0015 cm thick. At an elevated temperature of, for example, 100.degree. C., the outer portion of the beryllium copper sheet is cemented to the beryllia clamping ring 40. The cement is then cured. At a reduced temperature of, for example, 95.42.degree. C., the inner portion of the beryllium copper sheet is then cemented to the Cer-Vit clamping disk 44. After curing the new cement, the beryllium copper sheet is etched to form the pattern of filaments 14 shown in FIG. 7. Each filament intermediate portion 84 may be etched to have, for example, a 0.0015 cm square cross-section. Cer-Vit cover ring 46, as shown in FIGS. 3 and 4, is then cemented over the inner portion of the filaments 4. The new cement is cured as before. This procedure sets up a tension preload in each filament 14. In this example, the tension preload is about 200 dynes. This preload tension will change no more than about 2% as the temperature varies in the range from -55.degree. C. to 70.degree. C.

An electrical conductor 86 is shown connected to an outward portion 80 of a support filament 14 for carrying coil current. Another conductor 88 for carrying coil current is shown connected to an inward portion 82 of a different support filament 14. The conductor 88 extends through the center aperture 90 of clamping disk 44 for connection to a forcer coil.

FIG. 9 shows forcer coils 100, 102, 104 and 106 interconnected in a closed-loop force balance system which provides a resultant force substantially at the center of mass of the proof mass 10, wherein the force is directed along the longitudinal or Z axis. These four coils are connected in series and are disposed circumferentially on the coil form portion 30 of the proof mass 10. As shown in FIG. 7, only one circumferentially-extending leg of each of substantially flat coils 100, 102, 104 and 106 is interposed directly in the gap between the magnet 24 and the flux return ring 26. The remaining circumferentially-extending leg of these coils is displaced from the gap. Thus, the first-mentioned leg of each coil provides the bulk of the force induced when the coil carries current.

Furthermore, the polarity of the radial magnetic field alternates in space from one coil to the next adjacent coil. For example, as indicated in FIG. 9 by the arrows marked B, the magnetic field may be directed positive radially outward at coils 100 and 104. In this case, it is directed positive radially inward at coils 102 and 106. To accommodate this alternating polarity of the field, the coils 100, 102, 104 and 106 are so connected in series that the current is directly oppositely in adjacent coils. Thus, when current is flowing counterclockwise in coils 100 and 104, for example, it is flowing clockwise in coils 102 and 106.

A signal from the pair of fixed pickoff electrode rings 48 for sensing translational deflection is fed to Z axis control amplifier 108. A forcing current to constrain longitudinal deflection of the proof mass 10 is output by the control amplifier 108. The forcing current is carried by a conductor 110 and by an elastic suspension filament 14 and a conductor 111 to forcing coils 100, 102, 104 and 106. The circuit is completed through a conductor 113 and another elastic suspension filament 14, a conductor 112, and an output circuit 114. The signal output 116 of circuit 114 is preferably in digital format.

FIG. 10 shows substantially flat forcer coils 118 and 120 interconnected in a force balance system to provide electromagnetic forcing and constraint along the Y cross axis of the proof mass 10. These two coils are connected in series and are so disposed opposite each other along the X axis on the coil form portion 30 of the proof mass 10 as to provide a resultant force directed along the Y axis acting substantially at the center of mass of the proof mass 10. As shown in FIG. 6, each of the longitudinally-extending legs of each of coils 118 and 120 is interposed directly between the magnet 24 and the flux return ring 26. However, the polarity of the radial magnetic field alternates in space from one longitudinally-extending leg to the other of each of coils 118 and 120. As a result, current flowing in the coils 118 and 120 produces both a force along the Y axis and, since the center of mass of the proof mass 10 is displaced from the point at which it is elastically suspended, a torque about the X axis.

A signal from the pair of fixed, sector-shaped pickoff electrodes 52 for sensing rotational deflection about the X axis is fed to control amplifier 122. A forcing current proportional to the X axis rotational deflection of the proof mass 10 is output by the control amplifier 122. The forcing current is carried by a conductor 124, by an elastic suspension filament 14 and by a conductor 125 to forcing coils 118 and 120. The circuit is completed through conductor 127, another elastic suspension filament 14, a conductor 126, and an output circuit 128. The signal output 130 of circuit 128 is preferably in digital format.

An additional pair of substantially flat forcer coils, displaced about the Z axis by ninety degrees from coils 118 and 120 is disposed on the coil form portion 30 of the proof mass 10. These coils are interconnected in another force balance system similar to the one shown in FIG. 10 to provide electromagnetic forcing along the X cross axis. This additional pair of forcer coils is not shown in FIG. 10.

Returning now to FIG. 6, the full assembly of electromagnetic forcer coils 34 is shown in cross-section disposed on the coil form portion 30 of the proof mass 10. Shown are portions of the Z axis forcer coils 100, 102, 104 and 106, the Y axis forcer coils 118 and 120, and a pair of X axis forcer coils 132 and 134. The coil assembly is disposed in the space between the four-pole magnet ring 24 and the flux return ring 26. As indicated by the polarity markings N on opposing poles 136 and 138 of the magnet ring 24 and by the polarity markings S on opposing poles 140 and 142, magnet ring 24 is so magnetized that the polarity of the radial magnetic field in the gaps between adjacent poles of magnet ring 24 and flux return ring 26 alternates in space. This was mentioned above in the discussion of FIGS. 9 and 10. Each of the cross-axis forcing coils 118, 120, 132 and 134 has one longitudinally-extending leg in the gap between a north pole and the flux return ring 26 and the other longitudinally-extending leg in the gap between a south pole and the flux return ring 26. Each of the Z axis forcing coils 100, 102, 104 and 106 has only one circumferentially-extending leg in the gap between either a north pole or a south pole and the flux return ring 26. This disposition of the coils was mentioned previously in the discussion of FIGS. 9 and 10.

The magnet ring 24 is shunt compensated for temperature variations in a way well known to those skilled in the art. A Carpenter 30 alloy, for example, may be used for this purpose.

FIG. 8 is another view of the electromagnetic forcer coil assembly 34. The Z axis forcer coils 100, 102 and 106 are shown, each with one circumferentially-extending leg directly under a pole of magnet ring 24 and the other circumferentially extending leg displaced from the poles so as to provide a force directed longitudinally with this type of magnet ring 24.

The cross-axis forcer coils 118 and 134 in FIG. 8 each have one longitudinally-extending leg under a north pole and one under a south pole so as to provide a force directed radially with this type of magnet ring 24.

In the discussion of FIG. 7, it was indicated that the support filaments 14 might be fabricated to have a square cross section. Actually, it is preferable that the filaments be made relatively wide and thin. A minimum for the cross-sectional area of a filament is determined by the amount of electric current required to be carried and the permissible electrical loss in the filament. Of course, the filaments 14 must each also have sufficient cross-sectional area and thickness to support the expected tensile loading and to give the suspension a required stiffness. However, for a given cross-sectional area, the smaller the filament thickness is made, the smaller will be the filament spring rates in bending and torsion. Reducing the bending and torsion spring rates is desirable in order to reduce the contributions to bias uncertainty or null drift due to such effects as material creep and the like.

The elastic suspensions for the accelerometer embodiments shown in FIGS. 1-10 each have either four support filaments or four pairs of support filaments 14. This number was chosen by way of example only. When four or more support filaments or support filament pairs are included in an elastic suspension, the suspension provides redundant constraint. The constraint is redundant because the suspension will still provide support for the proof mass 10 if a filament or a filament pair is fractured.

However, it is not necessarily advantageous to provide a redundant-constraint elastic suspension. If the procedure for attaching the support filaments 14 between the proof mass 10 and the base 12 is not sufficiently precise and leaves some slack in a filament or a filament pair, then the null position for the elastic suspension will not have a satisfactory degree of stability. Due to the extreme degree of care required to attach the filaments so that a satisfactory degree of null position stability is achieved in a redundant constraint elastic suspension, it is preferable to design the elastic suspension to use only three support filaments 14. Where no more than three support filaments 14 are used, there can be no slack. Therefore, the null position is relatively more stable. By using only three support filaments 14 in the elastic suspension, the accelerometer may be fabricated with greater ease and with larger tolerances for filament attachment. It may therefore be fabricated at lower cost.

Referring now to FIG. 11, there is shown a simplified sketch of a proof mass 10 elastically suspended from a housing or base 12 for an accelerometer in accord with the invention. This version of the accelerometer has means associated therewith for nulling or eliminating bias. The elastic suspension shown connecting the proof mass 10 to the base 12 is an array of eight support filaments 14 disposed in a plane substantially perpendicular to the longitudinal or Z axis of the proof mass 10.

In FIG. 11, the forcer coil assembly 34 is represented by a cylindrical sheet girdling the proof mass 10.

As mentioned in the discussion of FIGS. 3 and 4, the disk portion 36 of the proof mass 10 serves as a movable plate or electrode for the capacitor pickoffs used to sense rotational deflection about the cross axes X and Y and to sense translation deflection along the longitudinal or Z axis. Portions of the two fixed pickoff electrode rings 48 for sensing translational deflection are shown disposed above and below the movable plate 36. Excitation for the capacitor pickoffs is provided by a generator 200 connected through a conductor 202 and a filament 14 to the proof mass 10. This pickoff excitation was mentioned previously in connection with the discussions of FIGS. 3 and 4.

The primary control loop for maintaining the proof mass 10 of FIG. 11 centered along the Z axis is similar to the control loop shown in FIG. 9. A signal from the pair of fixed pickoff electrode rings 48 is fed, by way of a pair of conductors 204, to a signal amplifier 206. The output of signal amplifier 206 is fed to a synchronous detector 208 which also receives the pickoff excitation signal for reference purposes from the generator 200 by way of a conductor 210. The demodulated output signal from detector 208 is fed to a control amplifier 212.

A forcing current sufficient to maintain the position of the proof mass 10 centered relative to the base 12 under conditions of acceleration is output by control amplifier 212. This forcing current is fed to the Z axis forcer coils in coil assembly 34 by way of a conductor 110, an elastic suspension filament 14 and a conductor 111. The circuit is completed through a conductor 113, another elastic suspension filament 14, a conductor 112, and an output circuit 114.

A torquer coil 214 is mounted on the coil form portion 30 of the proof mass 10. Where the polarity of the magnetic field alternates in space as shown in FIGS. 6, 8 and 9, one longitudinally extending leg of torquer coil 214 is disposed under a north pole and the other longitudinally extending leg is disposed under a south pole of the magnet 24 as shown in FIG. 6. Thus, current in torquer coil 214 torques the proof mass 10 about its Z axis. An alternating torquing current is fed to the torquer coil 214 through a pair of conductors 216. The frequency of the torquing current is within the pass band of the primary control loop but high enough so that there is negligible effect on the operation of the loop. A suitable frequency for this purpose is, for example, 100 Hz.

When the proof mass 10 is torqued about its longitudinal axis by the alternating current in torquer coil 214, a periodic tension variation is set up in the filaments 14 of the elastic suspension. The variation is at twice the frequency of the torquer current because the tension in each filament 14 is increased and decreased twice for each cycle of the oscillatory torquer coil current.

A generator 220 provides a torquing excitation signal at the desired frequency for the periodic tension variation in the filaments 14. This signal is fed to a torquer current amplifier 222 which outputs an alternating current for the torquer coil 214 at half the frequency of the torquing excitation signal.

When the proof mass 10 is at its mechanical null position, the filaments 14 are in a plane substantially normal to its longitudinal axis. With the proof mass 10 at the mechanical null position, and despite the periodic variation in the tension of the filaments 14 as described above, no disturbance occurs in the position of the proof mass 10 in the three compliant degrees of freedom, i.e., about the two cross axes and along the longitudinal axis. To displace the proof mass 10 away from its mechanical null requires a force proportional to the displacement due to the elastic nature of the suspension. Thus, when a source of bias in a primary control loop displaces the proof mass 10 from its mechanical null position, a bias current is set up in the appropriate forcer coils to produce the force. When the tension in the filaments is caused to have a periodic variation as discussed above, the bias current will include a component having a corresponding periodic variation. The greater the bias, i.e., the greater the displacement between the mechanical null position and the electrical null position of the primary control loop, the greater will be the periodic variation in the bias current for a given level of periodic tension variation in the filaments 14. That is to say, both the nominal level of bias current and the amplitude of the periodic variation in the bias current will be proportional to the bias.

A secondary control loop for longitudinal axis bias compensation is shown in FIG. 11. The secondary control loop senses the bias, as represented by the periodic variation in the bias current described above, and operates to null or eliminate its effects. A signal proportional to the periodic variation in the bias current is fed by a conductor 224 through a capacitor 226 and a signal amplifier 228 to a synchronous detector 230. The detector 230 also receives the torquing excitation signal for reference purposes from the generator 220 by way of a conductor 231. The output of synchronous detector 230 appearing on conductor 232 is a unidirectional signal proportional to the bias. It is fed as a control signal to the control amplifier 212 to counteract the effects of bias and therefore to allow the primary and secondary control loops jointly to return the proof mass 10 to its mechanical null position and maintain it there.

Alternatively, when the switch 234 in the conductor line 232 is opened, the signal output by the detector 230 is used to calibrate or correct the accelerometer output signal appearing on conductor 116.

The controls shown in FIG. 11 apply to longitudinal axis operation. However, as will be apparent to those skilled in the art, a secondary control loop for compensating bias may equally well be combined with each of the primary control loops for constraining deflection about the X and Y cross axes.

FIG. 12 shows an alternative approach to achieving the same bias compensation effect as in the embodiment of FIG. 11. In the embodiment of FIG. 12, significant modifications are made to the elastic suspension for the proof mass 10. An array of four fixed support filaments 14 is shown, each attached at its inward end to the proof mass 10 and each attached at its outward end to the accelerometer housing or base 12. The filaments 14 are preloaded in tension. Therefore, they give the elastic suspension the same high stiffness along the X and Y cross axes and about the longitudinal or Z axis as was mentioned above in the discussion of FIG. 1.

Each of the support filaments 14 is disposed substantially parallel to and spaced apart from a corresponding driver filament 236. At its inward end, each driver filament 236 is attached to the proof mass 10 in the same manner as are the support filaments 14. One filament tension spring 238 for each driver filament 236 is suspended as a cantilever from the base 12. Each driver filament 236 is attached at its outward end to a corresponding filament tension spring 238 so that the driver filament 236 is held in tension.

Attached to and inserted between each filament tension spring 238 and the base 12 is a piezo tension changer 240 capable of changing its length when a voltage is applied across it. A piezo tension changer is a device in which a slab of a piezoelectric material is sandwiched between two electrodes. Each piezo tension changer 240 is so disposed as to be capable of displacing its associated filament tension spring 238 and thus varying the tension in the associated driver filament 236. As a result, the proof mass 10 is torqued and tension is varied in the fixed filaments 14 also.

The alternating voltage from the torquing excitation generator 220 is applied in synchronism to all of the piezo tension changers 240. The result is that a periodic variation in tension at the frequency of the voltage output by the torquing excitation generator 220 is set up in the driver filaments 236 and the support filaments 14. A corresponding periodic torque variation is applied to the proof mass 10. When the proof mass 10 is displaced from its mechanical null position by bias, a periodic variation in the primary control loop output current proportional to the bias is sensed and detected in the secondary control loop. As in the embodiment of FIG. 11, the secondary control loop operates to null the bias and return the proof mass 10 to its mechanical null position. The electrical components of the control loops for the embodiment of FIG. 12 are the same as in the embodiment of FIG. 11.

As was mentioned previously in connection with the discussion of FIG. 11, when the switch 234 in the conductor line 232 is opened, the signal output by the detector 230 is used to calibrate or correct the accelerometer output signal appearing on conductor 116.

The controls shown in FIG. 12 apply to longitudinal axis operation, However, as will be apparent to those skilled in the art, a secondary control loop for compensating bias may equally well be combined with each of the primary control loops for constraining deflection about the X and Y cross axes.

Another aspect of the elastic suspension shown in FIG. 12 is that it provides overload protection for the accelerometer in the same manner as the suspension described in Shutt, "Three-Axis Accelerometer Having Flexure Bearing With Overload Protection," U.S. Patent Application Number (to be supplied), filed concurrently herewith and incorporated herein by reference. The support filaments 14 of FIG. 12 of this specification are the same as the support filaments 14 of the incorporated specification. In both cases, the support filaments 14 are disposed for minimum displacement in response to lateral tensile loading. That is to say, loading on the elastic suspension which tends to cause displacement of the proof mass 10 in the plane of the cross axes X and Y produces the maximum effect possible on the tensile loading of the individual support filaments 14. In addition, the driver filaments 236 of FIG. 12 of this specification correspond to the compliant filaments 300 of the above-mentioned incorporated specification. The support filaments 14 are held in tension by the complementary set of spring-loaded driver filaments 236. The four driver filaments 236 are attached to and mounted between the proof mass 10 and cooperating filament tension springs 238. The driver filaments 236 are coplanar with the support filaments 14. The filament tension springs 238 are anchored to the base 12 through the piezo tension changers 240. Each of the filament tension springs 238 is preloaded in tension so as to apply a pull to its corresponding one of the driver filaments 236. The set of driver filaments 236 is disposed to apply a torque to the proof mass 10 about the longitudinal axis thereof. A pull is thereby exerted on each of the support filaments 14. In the absence of external loading, each of the support filaments 14 is pulled into sufficient preload tension to keep the proof mass 10 in equilibrium. The suspension of FIG. 12 is thus designed to be relatively compliant in all six degrees of freedom when the lateral load exceeds a selected design load. Sufficient deflection of the suspension occurs at such an overload to bring the proof mass 10 to a practical mechanical stop for support before excessive stress can fracture a filament. Therefore, the accelerometer of FIG. 12 is capable of suffering a significant amount of overload or shock without sustaining damage.

FIGS. 11 and 12 show elastic suspensions having redundant constraint in the same sense as was discussed above in connection with the suspension of FIGS. 1-10. However, it is preferable to fabricate these suspensions to have only three support filaments as was previously discussed. This allows the accelerometer to be fabricated with greater ease and with larger tolerances for filament attachment. Greater stability of the null position is achieved at lower cost since the adverse effects of a slack support filament are avoided.

While the invention has been described with respect to the preferred embodiment thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as set forth in the following claims.