Capacitively Commutated Brushless DC Servomotors

This invention relates to efficient brushless DC servomotors that are capable of precise motion in both one and two dimensions thereby forming linear, rotary, planar and cylindrical actuators.

In the field of electric motors it is well known that brushless designs are advantageous (See the book, D.C. Motors, Speed Controls and Servo Systems, Chapter 6, Electro-Craft Corporation, Hopkins, MN (1980). Such motors typically employ Hall effect magnetic sensors, together with permanent magnets suitably attached to the motor shaft, to provide commutation information for the motor windings. These widely used one dimensional rotary motor designs have also been extended to one dimensional linear applications using either magnetic or optical position sensors, as shown in U.S. Pat. No. 4,509,001 to N. Wakabayashi et al (April 2, 1985) or in U.S. Pat. No. 4,618,808 to J. Ish-Shalom (October 21, 1986). In all such cases the position sensor is an additional element, rigidly attached to the motor itself and accurately positioned with respect to the motor armature. It must be particularly noted in addition that no brushless DC motor in the prior art can move freely through arbitrarily large distances in more that one dimension.

On the other hand there exists a different class of motor, namely stepping motors (See the book by Vincent Del Toro, Electric Machines and Power Systems, p. 433ff, Prentice Hall, Englewood Cliffs, N.J. (1985). Such motors are open-loop devices in which the motor position advances incrementally in step with sequential electrical driving signals. Motors of this type have been extended to provide full two dimensional motion on planar surfaces over arbitrary distances, as shown in U.S. Pat. No. Re. 27,436 to B. A. Sawyer (July 18, 1972). Some efforts have also been made to produce a true two-dimensional brushless DC motor from such stepping motors by attempting to measure the motor position magnetically (but now in two dimensions) and again using this information to commutate the motor windings. However such schemes to date have turned out to be both difficult to implement and complex (U.S. Pat. No. 3,857,078 to B. A. Sawyer (Dec. 24, 1974)) and no commercally viable motor of this type has yet been reported. One of the key difficulties in such attempts is that of determining the two dimensional commutation (position) information rapidly and accurately and in such a way that motion in one direction does not disturb the position information available in the perpendicular direction.

There does, however, exist a different scheme for position determination in two dimensions which turns out to be well suited to this task. This is by measurement of the capacitance between a suitably designed electrode array and an appropriately patterned two dimensional surface. (See my U.S. Pat. No. 4,893,071, issued on January 9, 1990.)

It is an object of the current invention that this capacitive position-sensing scheme be applied to both one and two dimensional actuators.

A further inherent problem that arises in variable reluctance motors using unidirectional driving currents is that of dissipating the stored magnetic energy that is present in each winding at the turn-off instant. This problem is exacerbated if the motor uses a fine magnetic pitch and moves at high speed, thereby necessitating rapid inductor switching action.

While energy-recovery schemes for variable reluctance motors exist, some of the prior approaches require separate multiple windings on the motor armature to achieve the intended result. See, for example, "Variable Reluctance Motors for Electric Vechicles," NASA TECH. BRIEF, Vol. II, No. 10, Item 113, JPL Invention Report NOP-16993ISC-1444, J. H. Lang and N. L. Chalfin (Dec., 1987.) The separate motor windings impair other qualities of the motor.

A subsidiary objective of the current invention to arrange for both a speed up of the turn-off process and a corresponding speed up of the turn-on process in the next motor winding, further to arrange that the collapsing magnetic field in one motor winding uses the back-emf thereby generated to temporarily increase the voltage available to the next motor winding to be switched on, and furthermore, to achieve all the foregoing without extra motor windings and to apply all of the foregoing to reversible three phase motors, unlike those of U.S. Pat. No. 3,486,096 to G. W. Van Cleave (Dec,. 23, 1969) and U.S. Pat. No. 4,533,861 to J. E. Rogers, et al (Aug. 6, 1985), existing in the prior art that are also intended to provide rapid magnetic field switching in stepping motors by use of the back emf effect. These latter two prior schemes are unable to handle three phase motors including direction reversal.

Brushless DC motor action is achieved in all implementations of the invention by determining the relative position of the armature and the stator of a motor by RF capacitance measurements and using this information to control the drive to the motor windings. The capacitance measurements either make use of a separate electrode structure, or else, as in a preferred embodiment of the motor, employ the motor ferromagnetic cores themselves as the capacitive sensing elements. This leads to a self-aligned and compact design requiring no extraneous position sensing components.

The use of four such capacitively commutated linear servo motors mounted along the edges of a planar square air bearing supporting plate then gives rise to a true two-dimensional servo motor. This two dimensional servo motor, which can operate in a plane of, for example, a factory assembly work station, can in turn be modified to provide a two-dimensional cylindrical servo motor providing simultaneous and independent rotation and axial translation in the same actuator.

All of these motor structures lend themselves to high speed motion and this poses a known problem in multiphase motor operation described above, namely the difficulty of rapidly and sequentially switching the magnetic fields in the motor windings. This problem is handled in the present case by a novel passive system that either rapidly transfers the energy in a winding being turned off into the next motor winding being turned on or transfers it back into the power supply. This energy transfer scheme works for motors with arbitrary numbers of phases and furthermore automatically handles motor direction reversals.

This feature more specifically employs a novel diode-coupled autotransformer scheme. This allows a rapid turn-off of each inductive winding, coupled with the efficient return of essentially all the available stored magnetic energy to the driving power supply or other storage device during each turn-off transient. Furthermore this transformer system achieves this end without wasting any of the available motor winding space.

The same techniques used in my above-cited copending patent application for avoiding the effects of spurious capacitances can be advantageously used in the motor itself.

These and further features and advantages of the invention will become apparent from the ensuing drawings and detailed description.

FIG. 1 is a pictorial and schematic showing of a basic scheme of a three phase variable reluctance linear motor which uses the motor armature ferromagnetic cores themselves as capacitive position sensing elements.

FIG. 2 is a cross-sectional diagram indicating the maximum and minimum capacitance values of an electrode above a patterned ground plane.

FIG. 3 shows a transformer-coupled self-shielding method of measuring the capacitance of an electrode to ground.

FIG. 4 shows a method of interconnecting a linear array of ferromagnetic motor cores to provide three phase spatially-averaging capacitive position information.

FIG. 5 is an electronic block diagram showing the derivation of the three phase position outputs and height information from the motor electrode capacitances.

FIG. 6 shows curves of the form of the three phase position output signals .phi..sub.1, .phi..sub.2 and .phi..sub.3 provided by the position sensing system.

FIG. 7 shows an arrangement for derivation of the Sine/Cosine position outputs from the three phase signals, together with the motor winding commutation commands.

FIG. 8 illustrates the method of removing the effect of the motor winding capacitances from the position-sensing signals.

FIG. 9 shows interconnection of multiple motor cores with cancellation of winding-to-core capacitances.

FIG. 10 illustrates the footprint of multiple motor cores to show the transverse motion signal independence.

FIG. 11 shows a form of "Chevron" three phase spatially averaging printed circuit capacitive position sensing element.

FIG. 12 shows interconnection on the back of the three phase printed circuit position sensing electrode.

FIG. 13 shows a one dimensional rotary motor using the motor armature ferromagnetic cores as position sensing elements.

FIG. 14 shows a one dimensional rotary motor utilizing a separate capacitive position sensing element.

FIG. 15 shows a two dimensional planar servomotor.

FIG. 16 shows a two dimensional cylindrical servomotor.

FIG. 17 is a schematic diagram of an energy conserving high speed motor winding turn-off system.

FIG. 18 is a schematic diagram of an energy transferring high speed motor magnetic field switching system.

The basic form of a preferred embodiment of the invention is shown in FIG. 1.

Here three electrically separate but identical ferromagnetic cores 1, 2, 3 are mounted just above a flat electrically grounded ferromagnetic plate 4 which contains a uniform two dimensional pattern of holes. Each core is connected to a winding 5, 6, 7 on a separate 1:1:1:1 multifilar RF transformer 8 whose primary 9 is driven by an RF oscillator 10. The three cores each contain a winding on their center leg (omitted for clarity, but see FIG. 8) and all three cores are rigidly mechanically attached on one another with particular spacings. This arrangement comprises a variable reluctance linear motor in which the array of ferromagnetic cores (electromagnets) are supported on an air bearing (see FIG. 15) just above the ferromagnetic plate 4 or "platen". Turning on the center leg winding of an appropriate core tends to pull it into step with the nearest "line" of platen material that does not contain holes. Sequentially and appropriately turning the electromagnets on and off in turn then causes the motor to move across the platen. This arrangement of using a ferromagnetic platen with an array of "holes" instead of protuberances or "posts" (which are also usable) has fabrication advantages, allows higher net magnetization, is convenient for air bearings, and provides better position sensing signals.

Of course the synchronization of the electromagnet current drives is critical since each must be turned on and off at precisely the correct spatial position with respect to the array of platen holes. The way in which the required position sensing is achieved can be understood with the simple diagram of FIG. 2. Here a small conducting electrode 13 is envisaged as being either directly over a conducting region of the platen 14, or else over region with a hole (the holes may or may not be filled with plastic or other dielectric, non-ferromagnetic material). It is evident that the electrical capacitance of the electrode to ground is different in the two cases, with corresponding intermediate values at intermediate positions.

It is this capacitance effect which is used in FIG. 1 to sense the relative position of the ferromagnetic cores with respect to the platen. The way in which this is done, using a multifilar RF transformer, is identical to that described in my previously mentioned co-pending patent application "Interpolating Incremental Capacitive Encoders" Serial No. 197800, filed May 24, 1988. The basic scheme is indicated in FIG. 3. An RF oscillator 17 drives the primary 18 of a bifilar 1:1 transformer. The secondary 19 drives the capacity sensing electrode of interest 20, which is itself advantageously surrounded by a driven guard or shielding electrode 21 driven by winding 18. In this way only the RF displacement current flowing from electrode 20 to ground 22 is measured at point 23. This measurement is independent of all potentially interfering effects of capacitances C2, C3, C4 and C5, as explained in detail in the previously mentioned co-pending patent application. This basic scheme is the one used in FIG. 1 and of course extends directly to the use of multiple sets of cores as shown in FIG. 4. Here cores 25, 26, 27 are connected together with other identical cores in similar relative positions in such a way as to provide a common three-phase output E1, E2, E3. All the electrodes are shielded by a single common guard electrode 28.

The output signals E1, E2, E3 in turn drive the system shown in FIG. 5. This is again identical to that used in the aforementioned co-pending patent application. The differences of the capacitive signals (E1-E2), (E2-E3) and (E3-E1) drive the synchronous rectifiers 31, 32, 33. These then provide the position outputs 37, 38, 39. An RF oscillator 40 drives amplifier A3 through a linear attenuator 41. The total RF amplitude to the E1, E2, E3 electrodes is thereby servoed in such a way as to provide a constant total displacement current. This action provides spacing-independent position signals 37, 38, 39 together with an independent height output signal 42.

The form of the position signals 48, 49, 50, (labeled .phi..sub.1, .phi..sub.2, and .phi..sub.3) is shown in FIG. 6. These are interleaved at 120.degree. (spatial) degrees with respect to one platen period 51. A particular advantage of this capacitance based position measuring system is that it averages over a large area of the platen. This has the effect of ensuring that the system has good differential linearity, i.e., good cycle to cycle reproducibility as the motor moves. This results in corresponding distances such as AB and BC being very nearly equal, which is important for accurate position interpolation.

The three-phase position signals .phi..sub.1, .phi..sub.2, and .phi..sub.3 are combined as shown in FIG. 7 to provide Sine and Cosine outputs. These two signals are then used to drive up/down counters to determine the motor position. The same signals are also digitized by an 8 bit ADC to provide fine position interpolation within one platen period. Motor commutation signals are derived from the .phi..sub.1, .phi..sub.2, and .phi..sub.3 position signals. These commutation signals are used in conjunction with the current drivers (which are described subsequently) so as to provide controlled force and direction to the motor motion.

In order to obtain the most accurate position sensing signals it is desirable to reduce to a minimum all stray and unwanted capacitances associated with the motor ferromagnetic cores. One of these stray capacitances is that between the winding on the motor center leg and the core itself. The effect of this stray capacitance is removed by the scheme shown in FIG. 8. Here the electromagnet winding 61 is driven through two additional windings 62 and 63 on the RF transformer. By arranging that the small capacitors 66 and 67 hold one end of each of the transformer windings 62 and 63 at RF ground it follows that the whole of the electromagnet winding 61 itself is driven exactly in unison with the remainder of the ferromagnetic core 60. As a consequence essentially no RF current flows through the stray capacitance between 60 and 61, i.e., effect of this unwanted capacitance has been removed. It will be appreciated that the large DC current that drives the electromagnet coil 61 has no effect on the (ferrite cored) RF transformer 62, 63, 64, 65. This follows since the current through 62 and 63 flows in opposite directions through a bifilar winding and therefore links no net flux with the RF ferite core.

In principle it would be possible to arrange to repeat this procedure for each of the three cores of FIG. 1, for example. However, the same effect can be achieved in all cases, but with fewer windings, as shown in FIG. 9. Here six cores 70 through 75 are shown with their six electromagnet windings 85 through 90. It will be noted that only 8 RF windings (76 through 83) are needed, and indeed only that number is needed for an arbitrarily large number of cores.

The footprint of three cores 93, 94, 95 is shown overlaid on the platen array of holes 96 in FIG. 10. It will be noted that since each core has a lateral footprint length, it its motion-independent direction, of exactly an integral number of platen periods, the associated capacitance values remain essentially unchanged for translations in that one (motion independent) direction, while varying essentially sinusoidally (as required) in the other (measuring) direction.

All of the preceding motor description has centered on situations in which the motor armature elements themselves can act as their own capacitive position sensors. In cases where this is not possible a separate capacitive position sensing array can still be used, of the type described in the previously mentioned co-pending patent application. An example is indicated in FIG. 11. This shows a printed circuit array of triples 98, 99, 100 of capacitive sensing elements. These form a slightly meandering "Chevron" pattern that further aids in ensuring that the sensor only reads out for longitudinal motions and not for transverse motions. Each triple of electrodes is interconnected on the back of the printed circuit board (via plated-through holes) to the three common readout lines 103, 104 and 105. Such an electrode structure can be rigidly attached to a suitable set of electromagnets such as cores 70-75 to provide linear motor commutation information. Its advantage over Hall effect or other sensors for this application lies in the fact that it has large area and therefore provides good differential linearity by virtue of providing a high degree of spatial averaging.

Although the previously described systems have all related to one dimensional linear motors, the extension to one dimensional rotary motors is clear. An example in shown in FIG. 13. Here the three armature elements 109, 110, 111 are doubling as their own capacitive position sensing elements in a fashion identical to that of cores 1, 2, 3 of FIG. 1.

Similarly, if the rotary motor armature elements cannot double as their own position sensors it is still possible to provide a separate rotary capacitive sensor. An example is shown in FIG. 14. Here the interleaved triples of sensing electrodes E1, E2, E3 are conveniently formed photolithographically on one side of a piece of printed circuit board 115. The metallization 116 on the other side of the printed circuit board, (shown moved away for clarity), is itself again driven from the RF source as before so that it moves electrically in unison with the electrodes E1, E2, E3, thereby acting as a driven shield. The sensor array is mounted a small distance away from a rotating, grounded, patterned electrode 117 that is rigidly attached to the motor shaft. Although overall this is a flat disc-shaped sensor it will be obvious how to extend it instead to an essentially cylindrical design.

From what has been described regarding one dimensional linear capacitively commutated motors of the type shown in FIG. 1 it will be clear that numbers of such motors can be grouped to operate in unison. An examples is shown in FIG. 15 where four such motors 120, 121, 122, 123 lie along the edges of a flat, square, air-bearing plate 124 with air bearing nozzles 128. This design provides for controlled two-dimensional motion over the entire platen area, which can in principle be of indefinite extent. Sensor signals, motor drive signals, power and compressed air (for the air bearing) are provided by a flexible umbilical cord 125. The system provides four separate independent position readout signals X.sub.1, X.sub.2, Y.sub.1, Y.sub.2. The position (X,Y) of the center of the motor is defined by X=1/2(X.sub.1 +X.sub.2) and Y=1/2(Y.sub.1 +Y.sub.2). The angular rotation .theta. of the motor (limited to .+-.5.degree. in current designs) is obtained, redundantly, from (X.sub.1 -X.sub.2) and (Y.sub.1 -Y.sub.2). Motor position, which is needed for closed-loop servo control, is provided by up-down counters driven from the sine-cosine outputs from each motor as indicated in FIG. 7. It will be appreciated that this motor design provides both full two-dimensional planar motion and also controlled rotation over a small (acute) angular range. The commutation signals remain unambiguous for the design of FIG. 15 up to an angle of rotation of about 5.degree.. This angular control can be useful in the application of these motors to assembly operations. The motor position information is redundant, providing the four numbers X.sub.1, X.sub.2, Y.sub.1 and Y.sub.2 to specify the three position parameters X, Y and .theta.. This provides a consistency cross-check. Furthermore, as will be recalled from FIG. 5, the sensor system also provides the heights h.sub.1, h.sub.2, h.sub.3, h.sub.4 of each sensor above its local ground plane. This provides a measure at each motor of the air-bearing thickness and this information is again redundant. In this way the four motor sensors provide a redundant readout of all six degrees of freedom of the motor motion. It will be noted also that successive groups of motor cores can have a slight offset from the preceding group for the same motion direction to provide the same effect as the chevron pattern of FIG. 11.

As a practical matter the planar motor uses in addition a permanent hold-down magnet 126 to maintain correct air bearing operation even when all four motors are turned off. This magnet is required because the motors themselves are true variable reluctance designs and therefore do not contain any permanent magnets such as those employed in hybrid stepper motors.

Given a two-dimensional planar design of the form of FIG. 15 it is clear that it also, if desired, can instead use four separate capacitive position sensors of the type shown in FIGS. 11 and 12. It is also clear that, independent of which form of sensing is employed, it is possible to convert it to a two-dimensional cylindrical motor of the type shown in FIG. 16. Here the ferromagnetic platen of FIG. 15 has been wrapped around to form a cylinder 129, the surface of which contains, as before, a two-dimensional pattern of either holes or posts. Shown outside and spaced from this cylindrical surface are two separate capacitive sensors 130 and 131, to read out Z and .theta., respectively. Equally well, of course, these two sensors can be self-sensing motors of the type of FIG. 1 and FIG. 13 to provide Z and .theta. control. In this cylindrical case it is noteworthy that there is no readout redundancy since the cylinder is constrained to allow only Z and .theta. motion.

All of the proceeding has related to design and capacitive position sensing of different motors; but nothing has yet been said regarding the actual electrical driving of the motor coils. Typically, for small motors of the kind discussed here, the coil inductances are perhaps .apprxeq.10 mh and the maximum drive currents .apprxeq.2 amperes The corresponding stored energies are therefore .apprxeq.2.times.10.sup.-2 Joules. At high motor speed the switching frequency can be .apprxeq.1KHz, thereby leading to coil flyback powers of .apprxeq.20 watts. In conventional hybrid motors employing bipolar current drives all of this energy is automatically returned to the driving power supply. In true variable reluctance motors with unidirectional drives (of the kind used here) however, the situation is very different and this energy is often intentionally dissipated as heat in a resistor, or else dissipated in a high voltage zener diode. The advantage of a zener diode is that it clamps the coil flyback voltage at a high and constant value (much higher than that of the driving power supply) which causes the coil magnetic field to collapse very rapidly, which is desirable. However, this is wasteful of energy.

A simple scheme that retains the advantage of rapid field collapse, while in addition wasting essentially none of the available flyback energy, is shown in FIG. 17. Here the three motor coils 134, 135, 136 are star-connected and driven by three power transistors 137, 138, 139. Supposing for example that 137 had been on and is suddenly turned off, then the flyback voltage from inductor 134 rises rapidly until diode 140 turns on. Since the autotransformer 143 has an N:1 step down ratio the clamping voltage seen by inductor 134 is essentially N times that of the power supply 145. This is because the voltage clamping action itself is caused by diode 144 turning on as its cathode is driven one diode drop below ground potential. This therefore provides very rapid coil turn-off action, while simultaneously returning the stored magnetic energy to the power supply.

This scheme is a convenient, efficient and simple one, which in addition works for any number of motor phases, driven in any commutation sequence. However, it only provides for a rapid and efficient coil turn-off, doing nothing to speed up the corresponding coil turn-on.

The turn-on process itself can be speeded up in suitable cases by intentionally extending a property of the circuit of FIG. 17. This property is that when diode 144 turns on (during coil flyback) positive current flows back into the positive terminal of the power source 145. As a result the output voltage of that supply necessarily undergoes a small positive excursion, of a magnitude proportional to its output impedance. This excursion can instead be intentionally made very large indeed by using the system shown in FIG. 18. Here, as before, if one envisages that transistor 151 had been on (carrying current I.sub.o) and was suddenly turned off then the voltage across diode 154 suddenly goes positive, i.e., it turns on. Now, however, due to the 1:1 autotransformer 157 and capacitor 160, as diodes 154, 158 and 159 all turn on the circuit is equivalent to that of a simple parallel resonant LC tank with inductor 148 (of inductance L) directly across capacitor 160 (of capacitance C). In one half of a cycle of this resonant tank all the initially stored energy in inductor 148 is transferred to capacitor 160. After that diodes 154, 158 and 159 all turn off and the voltage available for the turn-on of the next motor coil is then (V.sub.B +I.sub.o .sqroot.LC) This can be much larger than V.sub.B itself and therefore leads to a more rapid build up of the current in the next motor inductor to be switched on.

In practice the situation is more complex that that just described, although the circuit used is precisely as shown in FIG. 18. The commutation cycle employed has only one coil turned on at a time, with exact coincidence of the turn-on and turn-off signals to the power transistors 151, 152 and 153. Theoretically the total field switching time (decay of one coil and build up of the next) is given by T=0.7.pi..sqroot.LC. However, this time is not observed in practice since there are inevitably significant eddy-current and other losses in both the platen and the motor cores. This means that all the magnetic field energy is not automatically transferred from one coil to the next and the difference has to be supplied through diode 161, which takes additional time. However, speedups of a factor of 2 or 3 are obtainable in practical situation.

Two further features of FIG. 18 are noteworthy. First, since only one coil is on at a time, the common resistor 162 provides current feedback information for all the coils and is used in a current servo. Second it is clear that the sequence of the powering of coils 148, 149, 150 is immaterial to the operation of the system. However, this corresponds precisely to reversal of the motor, i.e., motor reversals are handled automatically.

As a final point it should be noted that all of the proceeding has concentrated on three phase motors and the corresponding use of triples of capacitive pickup electrodes. However, it will be obvious to those skilled in the art how to modify the methods presented to handle motors of different designs with different numbers of phases. It should also be apparent that one can produce the desired sinusoidal signals in the measurement direction with a number of different periods and phase offsets of the capacitive elements and/or footprints of electromagnetic cores.