Position sensor for magnetically suspended member.

This invention relates generally to magnetic suspensions for pointing instruments on a levitated member, movable in rotation about an axis or in linear translation. More particularly, the invention pertains to an electronic circuit configuration for determining the position of a suspension member with respect to associated electromagnetic actuators, for use in linearising the applied deflection force.

Magnetic suspensions are particularly well adapted for substantially frictionless suspension of a movable member. Prior suspensions of this type utilising magnetic bearing assemblies, in which it is desired to provide an output force linearly proportional to a commanded input force, require compensation for the non-linear nature of the magnetic circuit, since the force obtained is a non-linear function of coil current and the air gap between the magnetising coils and the movable member. The resulting magnetic force F has been found to be of the form<maths id="math0001" num=""><img file="EP0218318A2_D0001.tif" /></maths>where I is the current applied to an actuator coil, g is the air gap between a magnetically permeable core of the coil and the magnetic suspension member, and K is a factor of proportionality. Since the magnetic actuator typically consists of two essentially identical magnetic circuits applying oppositely-directed forces to a common rotor or armature, it may be shown that the square law current characteristic can be compensated by applying a fixed bias to the variable control current supplied to each coil. By sensing the position of the armature in the magnetic gap, the currents may be further modulated to negate the effects of the reciprocal gap-squared factor.

Thus prior suspension technology has required accurate measurements of both applied current and air gap between the armature and the coil. While precise current measurements are readily obtained, making an air gap measurement with the requisite linearity and resolution requires a precision sensor which is costly to fabricate and may result in reduced system reliability. A further disadvantage of the aforementioned apparatus is that equation (1) does not allow for the effects of magnetic hysteresis, flux leakage across the gap, and other magnetic anomalies. In consequence, magnetic actuators controlled in this way have at best achieved of the order of one or two percent linearity, even under controlled conditions. In Magnetic Suspension: The Next Generation in Precision Pointing, American Astronautical Society, Paper No. AAS82-034, January 30, 1982, the present inventor discussed an apparatus for correcting for such errors. By measuring the actual force applied to the magnetic bearing, as well as the magnitude of the air gap, and feeding the actual force signal in a closed loop to correct the force command signal, there is obtained improved performance with linearity of the order of one-tenth of one percent and compensation for the effects of the magnetic anomalies. However achieving an accurate air gap measurement requires a precision proximeter transducer (proximeter). The present invention obviates the need for a precise measurement of armature position in the gap when used in a force-loop system.

The invention is defined in the appended claims and provides an improved magnetic suspension for applying a desired force to a magnetically permeable armature which is thereby positioned in a magnetic field formed by energising a plurality of coils, whereat a measure of the airgap between the armature and coils is found by the magnitude of the force applied to the armature. A force sensor is coupled to the armature to develop a signal corresponding to the actual force, and the currents applied to the coils are also measure. By combining the force signal and the currents, a new signal is derived which corresponds to the magnitude of the air gap, and which is used to control the coil currents so that the actual force is linearly proportional to a force command signal.

A magnetic bearing system in accordance with the present invention will now be described in greater detail, by way of example, with reference to the accompanying drawings, in which:- <ul id="ul0001" list-style="none"><li>Figure 1 is an elevation view showing the relationship of a fixed and movable body and associated magnetic suspension assembly,</li><li>Figure 2 is a conceptual view in cross section of a magnetic bearing assembly of the prior art,</li><li>Figure 3 is an electrical schematic diagram of the electronic controls for a magnetic suspension including the force sensor and air gap estimating features of the present invention, and</li><li>Figure 4 is an electrical schematic diagram of an analogue divkjer/mulfiplier circuit for use in the gap estimating circuit of the present invention.</li></ul>

Referring first to Figure 1, a pair of magnetic bearing assemblies 10 and 12 is shown as supporting a movable body 14 above a fixed body 16. Each magnetic bearing assembly is comprised of a stator 18 and 20 for supporting opposing coils for producing a magnetic flux when energised by an applied current. Armatures 30 and 32 are formed from a magnetically permeable material such as iron and disposed within an air gap between coils 22, 24 and 26, 28, respectively. When an electrical current is applied to the coils, a magnetic field is established which causes a predetermined force to be applied to the armatures. Since the armatures 30; 32 are fixed to the movable body 14, the body is caused to be suspended with a predetermined air gap between the respective armature and its associated coils. By varying the applied currents, the resulting magnetic fields will cause the movable body 14 to be levitated linearly or inclined angularly in accordance with the applied currents. Also shown in Figure 1 are transducers 34 and 36 which are used to sense the magnetic force applied to the movable body 14, in a manner to be described.

Referring now to Figure 2, a magnetic bearing assembly 40 of the prior art for use with a magnetic suspension is shown in cross section. Each magnetic bearing assembly 40 includes electromagnets 42 and 44, comprised of a magnetic core and associated coil 46 and 48. An armature or rotor plate 50 of magnetic material is disposed between the face of the electromagnets 42 and 44. The electromagnets 42,44 apply opposing forces to the rotor 50. When a signal current I, is applied to the coil 48, the rotor is attracted in the direction of the electromagnet 44. Conversely, when a signal current 1₂ is applied to the coil 46 of the electromagnet 42, the rotor is attracted in the direction of the electromagnet 42. A proximity transducer 52 is included with each magnetic bearing assembly 40 to produce a signal which is proportional to the distance A g from the position of the displaced rotor 50 to a predetermined reference point go in the plane of the rotor neutral position. The proximiter may be a conventional eddy-current pick-off device, the signal output of which is proportional in magnitude and sense to any change in the gap length from the predetermined value, this signal being processed and amplified so as to produce a current proportional to a change in the airgap. These signals are used for compensating for the inherent non-linearity of magnetic flux resulting from an applied current in the magnetic bearings. It will be shown that the proximiter signal may be used to modulate the electromagnetic flux so as to linearise the resulting force applied to the rotor 50.

Referring now to Figure 3, there is shown a block diagram of a circuit suitable for controlling a magnetic bearing assembly 62. The circuit of Figure 3 represents the elements necessary to control a single magnetic bearing assembly. Like circuits will be provided for each magnetic bearing assembly, of which normally a plurality will be provided. The primary components for controlling the force applied to the rotor 50 are a summing junction 64, integrator and gain 66 and control low network 68, a servo amplifier loop 70, a magnetic bearing assembly 62, and a force sensor 72. A force command signal 74 is applied to an input terminal of the summing junction 64 and processed through the conventional integrator and gain 66. The integrated signal is then applied to the control law 68, function of which is to linearise the applied magnetic force field with respect to the force command 74.

As described above with respect to equation - (1), it is known in the art that the use of a magnetic bearing as a linear force actuator requires that the response be modified in accordance with a control law to compensate for the non-linear nature of the magnetic circuit. Since the magnetic actuator consists of two magnetic circuits applying opposite- directed forces to a common rotor, the current- squared term 1² can be corrected by applying a bias current to each circuit to which a control current is added on one side and subtracted on the other side, which will result in a net force proportional to the control current. Similarly, the currents can be modulated by applying positive and nega- five current increments corresponding to the air gap between the rotor and its respective electromagnet. When the rotor moves in the magnetic gap, the currents will be modulated to compensate for the effects of the inverse gap-squared law. The resulting control law is of the form:-<maths id="math0002" num=""><img file="EP0218318A2_D0002.tif" /></maths>where 1₀ is the bias current, A g is the displacement of the rotor from the gap center g_µ, F is the applied force, and F_MAX is the maximum design force for the magnetic bearing assembly. The applied currents I, and 12may be readily measured by conventional current sampling devices, while the gap signal A g has heretofore been provided by the proximeter 52 as shown in Figure 2. The servo loop 70 is comprised of a summing network 75, servo amplifier 76 and feedback path 78. The loop is closed to ensure that the current in the coils is correctly calibrated despite variations in coil parameters with temperature, and also to reduce the lag in response that results from the large inductance of the coil. The output of servo amplifier 76 is used to energise an electromagnet of magnetic bearing assembly 62.

While the control law 68 compensates to a high degree for the non-linear magnetic field response, errors in the measurement of the gap value with respect to the actual gap will result in a force error. Further errors exist within the magnetic circuit itself since an ideal circuit was assumed in deriving the simple equations of (2). The solution that has been found effective is to measure the actual force applied by the magnetic bearing and feed back a proportional signal to the system input, thus closing the force loop 60. This approach eliminates the need for precise calibration, hysteresis corrections. etc. For this purpose, the force sensor 72 is mechanically coupled to bearing assembly 62 and provides an electrical output proportional to the applied force. The output of force sensor 72 is coupled to the summing network 64 in a polarity opposing that of the input force command 74 so that a force error signal 80 representing the difference between the commanded force signal and the sensed force is used to drive the integrator 66.

The force sensor 72 must be highly linear and stable, have low hysteresis, fast response time, and extremely high resolution. It must be capable of operating in severe environments including a wide range of temperatures and pressures and avoid introducing a resonance in the force loop feedback circuit. One suitable device is a quartz resonator transducer in US-A-4,215,570 which is hereby incorporated by reference.

The remaining circuit elements in force loop 60 are conventional.

A disadvantage of the prior art system, as described with respect to Figure 2, which utilises both a force sensor and a proximity sensor, is the reduced reliability and added expense of providing a multiplicity of sensors. Since simulation studies indicated that under closed loop conditions position errors as high as five percent could be tolerated, an alternate approach which obviated the need for a proximity transducer was sought. Desirably, the present invention permits the armature displacement position from a knowledge of the parameters of applied current and resultant force, thus avoiding the need for the proximity transducer of the prior art. The present invention permits the replacement of the proximity transducer by a simple electrical circuit which will provide the required signal indicative of the armature gap to the control law 68.

From the force equation (1), the net force re- suiting from the application of control current to two opposing electromagnets may be shown to be:-<maths id="math0003" num=""><img file="EP0218318A2_D0003.tif" /></maths>which relates the four variables F_NET, I₁, I₂, and A g, wherein F_NET is the resulting applied force, and 1₂ are the currents applied to the electromagnets, A g is the change in the armature gap from the original position go, and K is a factor of proportionality. Since F_NET I₁, I₂ and go are easily measured and therefore precisely known, the fourth variable A g can be computed. Thus, an exact value of A g may be determined without a need for physical measurement of the value.

While equation (3) is theoretically capable of solution, in practice it involves taking the roots of a fourth order polynomial, which is not readily accomplished with analogue electronics. However, feedback loop 82 of Figure 3 provides a means for estimating the value of A g and by successive approximation returning an exact value of A g to control law 68.

Referring now to loop 82 of Figure 3, the measured force signal F_REF which is representative of the force applied to the suspension is applied to a first input of summing junction 84 which may be comprised of a differential operational amplifier. The output signal F₀is a force error signal representative of the difference between the measured force FREE and a computed force value F_COMP, and is applied to integrator 86 which serves to reduce steady state errors in the feedback loop. If a steady error signal F_e persists, the output of the integrator will be continuously increasing with time. This signal will drive control law 88 in a direction so as to reduce the error, and the integrator output will stop increasing only when the error signal is zero, after which the integrator will supply the necessary steady-state signal to maintain the desired magnetic field. Control law 88 may be comprised of a divider/multiplier which is caused to solve the following equation:<maths id="math0004" num=""><img file="EP0218318A2_D0004.tif" /></maths>The computed force output F_COMP is applied to a second input of the summing junction 84. Since the air gaps g, and g_z may be expressed in terms of a reference position g_o and a variable armature gap displacement Δ g as shown in Figure 2 and equation (3), equation (4) may be used indirectly to solve for the value of A g, and hence determine the true gap position.

Since the force error signal F_e is applied to the integrator 86, the output signal on lead 90 is dependent on the force error. By appropriately configuring the integrator 86, the signal on lead 90 may be made a measure of the armature displacement and hence of the true air gap change A g. The computed force magnitude F_COMP may be seen as directly affecting the value of A g on the lead 90 in the sense that increasing the force error will cause the integrator to charge in a direction so as to increase or decrease A g so as, in turn, to reduce the force error F_e to zero. The integrator transfer function may be expressed in terms of the Laplacian operator as 1/S. The integrator 86 poses- ses extremely high gain at low frequencies and hence will generate whatever Δ g signal is necessary to drive the force error to zero. The output A g is then applied to control law 68, where it is used to compute the required magnet control currents corresponding to the applied force command 74.

Referring now to Figure 4, there is shown - schematically a suitable electronic system for implementing the divider/multiplier 88 of Figure 3. The primary components for accomplishing this are non-linear elements 92 and 94, configured in the form of a divider, and a non-linear element 96 configured in the form of a multiplier.

Equation (3) can be rewritten in the form of:<maths id="math0005" num=""><img file="EP0218318A2_D0005.tif" /></maths>This equation is in the form of the difference to two squares as shown in equation (6) below:<maths id="math0006" num=""><img file="EP0218318A2_D0006.tif" /></maths>The product term can be implemented with a generic non-linear device as shown at the multiplier 96, the output of which is an algebraic product of the form:<maths id="math0007" num=""><img file="EP0218318A2_D0007.tif" /></maths>The dividers 92 and 94 perform an algebraic division which is of the form:<maths id="math0008" num=""><img file="EP0218318A2_D0008.tif" /></maths>For the algebraic divider 92, X, is a constant voltage provided by a Zener diode 98 scaled to represent go and X2 is the computed value of A g, thus making the denominator of equation (8) (go -A g). The input I, to terminal Z represents the current applied to one coil of the magnetic bearing assembly. The signal A g is also coupled to one input of an operational amplifier 100 which is wired to perform an inversion function. Four the algebraic divider 94, X, is again obtained from the Zener diode 98, and the value of Z is found from the applied current 1₂, representative of the current applied to an opposing coil of the magnetic bearing assembly. The signal applied to X₂ is the inverted output of signal Δ g. Thus, it will be seen that the output of divider 92 is of the form of equation (9) below and the output of divider 94 is of the form of equation (10) below:-<maths id="math0009" num=""><img file="EP0218318A2_D0009.tif" /></maths><maths id="math0010" num=""><img file="EP0218318A2_D0010.tif" /></maths>The output 0, of the divider 92 is coupled to the inputs X, and Y, of the multiplier 96. The output O₂ of the divider 94 is coupled to the input X2 of the multiplier 96 and also inverted by an inverter 102 to provide an input to the terminal Y₂ of the multiplier 96. Thus, the output 0, of the multiplier 96 at the input 104 to the summing junction 84 is the product of the sum and differences of equations - (9) and (10) resulting in the solution of equation (5).

The dividers 92 and 94 and the multiplier 96 may be obtained by appropriately configuring an integrated circuit such as the AD 532, as manufactured by Analog Devices Inc., Norwood MA 02062, United States of America. The summing junctions 64, 75 and 84 and the inverters 100 and 102 may be formed from appropriately configured operational amplifiers. the integrator 86 may be configured by using a filter circuit and high gain operational amplifiers in a conventional manner.

While the apparatus for computing the armature displacement signal is shown as analogue for illustrative purposes, it may also be implemented in digital form using conventional analogue-to-digital and digital-to-analogue converters or a programmable microprocessor.

In operation, referring again to Figure 3, a force command signal 74 is applied through the summing junction 64 to the integrator 66 where it is modified by the control law 68 to provide a linear magnetic output with respect to applied currents. The signal from the control law 68 is applied through the servo amplifier loop 70 to actuate an electromagnet (not shown) in the magnetic bearing assembly 62. A force sensor 72 coupled to the magnetic bearing assembly 62 provides an output signal F_REF proportional to the magnitude of the resultant force, which is returned to the summing junction 64 to derive a command error signal 80 representing the difference between the commanded force and the resulting applied force. The control law 68 requires an input A g representative of the air gap between an electromagnet and the armature, which is provided by the feedback loop 82 from the lead 90. The signal from the force sensor 72 is also applied to summing junction 84 and coupled to the integrator 86 which provides a ramped output. So long as there is an input signal provided, the output will continue to increase or decrease depending on the algebraic sign of the input When the input is zero, the output will stay fixed at the previously established value. The integrator output on the lead 90 is applied to the control low circuit 88 in combination with the signals I, and 1₂ representing currents applied to opposing electromagnets. The output F_COMP is a computed force magnitude which is subtracted from the sensed force magnitude F_REF in the junction 84 to provide a force error signal F_e to the integrator 86. If the computed value of the force differs from the sensed force magnitude, then an error will be generated which will cause the output of the integrator 86 to change in a direction either increasing or decreasing the signal at the lead 90 to urge the resulting force error to zero. In the steady state, the computed force will equal the sensed force, which is a necessary and sufficient condition for the output signal A g of the integrator 86 to be a measure of the true gap change. The signal A g is then applied to the control law 68 to complete the circuit operation. Since the gap position is implicitly determined from the known parameters, it need not itself be measured. Thus, advantageously, a conventional proximeter need not be provided, which results in reduced costs and increased reliability.