EP0000261B1

EP0000261B1 - Positioning system employing feedforward and feedback control

Info

Publication number: EP0000261B1
Application number: EP78300061A
Authority: EP
Inventors: Richard Karl Oswald
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 1977-06-29
Filing date: 1978-06-21
Publication date: 1981-10-07
Also published as: CA1100609A; BR7804158A; ES470846A1; IT7823829A0; IT1111180B; AU511484B2; DE2861129D1; US4200827A; JPS5412082A; AU3483578A; EP0000261A1

Description

Technical Field of the Invention

The present invention relates to positioning systems for moving a member between positions in response to a position input command. The invention relates particularly to systems in which each such movement, between a current and a target reference position, is completed before a subsequent command is accepted and comprises a single acceleration phase in which such a member is accelerated to a maximum velocity and a single deceleration phase in which said member is decelerated from said maximum velocity to a state of rest at said target reference position.

Background Art

A typical positioning application to which the present invention relates is the positioning of a data recording head or heads over a selected track of a magnetic disk file. The aspect of this positioning operation which is of interest is the movement of heads between tracks, known as the "track access" or "seek" operation, as opposed to the "track follow" operation which maintains the heads in position over a selected track.
Time optimal motion between tracks implies maximum acceleration and deceleration of the heads within the physical constraints of the system. To control such motion, with typical access times of a few tens of milliseconds, and bring the heads to rest on the target track, with an accuracy better than a hundred microns, feedback control of head velocity has been widely employed. Velocity control offers a higher performance than position control because velocity is the time derivative of position. A typical system for controlling a disk file head access operation is described in an article entitled "Design of a Disk File Head-Positioning Servo" by R. K. Oswald (IBM Journal of Research and Development, Nov 1974, pp 506 to 512). Similar systems directed to the same application are shown, for example in US Patents 2 881 184, 4006394. 4030132 and 4031 443.
In the system described in the Oswald article and in similar systems employed in many commercially available disk files, an access operation is controlled by means of a generated reference velocity trajectory representing the required velocity of the heads for deceleration at the maximum attainable rate to a state of rest over the target track. A velocity transducer (or tachometer) measures the actual velocity of the heads and the measured velocity is compared with the reference velocity trajectory and amplified in an error amplifier to provide a velocity error signal. The velocity error signal is applied to control the head actuator, typically a voice coil motor, to cause the actual velocity to follow the deceleration curve as closely as possible. Initially, actual velocity is low and the heads are accelerated under open loop (saturated) conditions until the actual velocity equals the reference velocity. When actual velocity exceeds the reference velocity, the sign of the velocity error changes and reverse current is applied to the actuator. The reverse current is controlled as a function of the velocity error to cause the head velocity to follow the reference velocity trajectory accurately. Considerable accuracy is possible- since, generally a high bandwidth velocity measurement is available with very little lag. Typically, such a velocity signal has been derived from incremental position signals provided by an external position transducer linked to the head motion or, by a servo head and dedicated servo surface on one of the disks.
Control systems for innumerable other applications than position control are also found in the prior art. A large variety of control schemes also exist, and these have tended to develop according to the particular application and its unique problems. In the process control art, control systems have been developed which have some resemblance to the positioning control system of the present invention.
The problems of the process control art are somewhat different from those of the position control systems. In process control there is often a long delay between the alteration of a manipulated variable and the process response. During such a delay, conditions may change. For example, other variables besides the manipulated one may change and alter the process characteristics. Also further commands may be received before the process has had time to respond to a previously applied command. In such conditions, feedback control alone is not adequate and so-called feed- forward control has also been employed. Feed- forward control involves predicting the change with time of a manipulated variable necessary to produce the desired change in a controlled variable of the process. The feedforward function is based on a model of the process and is applied to the actual system independently of any feedback. Feedforward control is essentially an open-loop technique allowing immediate response to an input command. The accuracy of such control is only as good as that of the process model.
In the process control art, control systems have been proposed which employ both feedback and feedforward control. US Patents 3 657 524 (Bakke) and 3758 762 (Littman) describe such systems. In the system described in the Bakke patent, for example, a process is coarsely controlled by means of a manipulated variable to produce a desired change in a measureable controlled variable of the process. A control and feasible response means is responsive to a command signal, defining a set point of the process, to provide a control signal, in accordance with a predetermined model of the process, for controlling the manipulated variable in such a way as to bring the process to the set point. This is feedforward control. Simultaneously, the control and feasible response means provides a feasible response signal, representing the predicted response of the controlled variable to the change of the manipulated variable in accordance with the feedforward control signal. The feasible response signal is compared with the actual measured response of the controlled variable to provide an error signal. This error signal is summed with the fed forward manipulated variable to provide fine feedback control of the process.
The application of feedforward plus feedback control techniques to positioning systems has been proposed in the prior art. Systems having both types of control are described in an article "Feed forward can improve feedback controls" by R L Dutchie (Control Engineering, May 1959, pp136-140). However, in the systems described in that article, the position input signals, 9i, are arbitrary and may vary continuously. The feedforward action consists of simultaneously generating feedforward signals either directly from or directly in response to variations of the input signals. These feed- forward signals are then applied to various points in the main gain path of the control loop. There is no element of prediction over a period of time and relatively complex tachometers and auxiliary servo systems must be used to enable the feedforward signal to respond in real time to variations of the input.

Disclosure of the Invention

As discussed above, because of the high positioning accuracy and rapid changes of velocity required in many position control applications such as disk file head positioning, the tight control afforded by velocity feedback has been widely used. However, the velocity feedback loop is Type 1 and the reference velocity trajectory is approximately a ramp so that the velocity error can never be completely eliminated. The magnitude of the velocity error is dependent on the overall gain of the feedback loop which also affects the bandwidth of the loop. A compromise is necessary between the reduction of velocity error and the limitations imposed by the mechanical resonances of the system. If the gain is more than unity at a resonance frequency, the system will be unstable. This problem is becoming increasingly acute with the higher frequencies inherent in increased track densities and shortened access ties.
Another problem is posed by the recent trend in disk file head positioning systems to eliminate both external position transducers and dedicated servo surfaces in favour of a sampled servo scheme in which sectors of servo information are interspersed with data on the disk recording surfaces. To maximize the disk area available for data storage, the number and extent of such servo sectors must be kept to the minimum. This in turn reduces the bandwidth of the position and velocity signals which can be derived from the servo information for feedback to a conventional position control system. Using feedback control alone, a high velocity error will result unless access times are severely limited.
The present invention offers a solution to the above problems in providing a positioning system, responsive to a position input command to move a member along a predetermined path of travel between a current and a target reference position defined by said command, each such movement being completed before a subsequent command is accepted and comprising a single acceleration phase in which such a member is accelerated to a maximum velocity and a single deceleration phase in which said member is decelerated from said maximum velocity to a state of rest at said target reference position, said system comprising an electrically controlled actuator for moving said member along said predetermined path in response to electrical drive signals and having an input circuit to which said signals are applied, reference signal generating means responsive to said position input command to generate a reference signal at last a portion of which represents the variation with time of a position related attribute of said member for deceleration of said member in a predetermined manner during said deceleration phase, and a feedback control loop including a transducer for providing a signal indicating the value of said position related attribute of said member, error determining means for producing an error-signal representing any difference between said transducer signal and said reference signal, and feedback means for feeding back said error signal to said actuator input circuit to cause said actuator to move said member in a direction to reduce said error signal characterized in that said system further comprises feedforward control means including predictive drive signal generating menas for generating, concurrently with said reference signal generation, a time varying predictive drive signal at least a portion of which represents the input to the actuator of a nominal system required to decelerate such a member in said predetermined manner during said deceleration phase, and means for feeding forward said predictive drive signal to said actuator input circuit, whereby deceleration of said member is coarsely controlled by said feed- forward control means, and is concurrently finely controlled by said feedback control loop to correct for fine errors resulting from the differences between said nominal and actual systems.
The invention allows time optimal positioning movement while permitting a low bandwidth in the feedback loop by employing in addition to feedback control a type of feedforward control in which the form of the feedforward control signal can be predicted in advance and represents the input to the actuator of a nominal system.
Specifically, when the basic problem is the stability of a system which is subject to mechanical resonances of bandwidth overlapping the bandwidth of said transducer signals, the invention preferably provides that the bandwidth of said feedback control loop as a whole is arranged to be predominantly below the frequency of said resonances, such that the overall loop gain is substantially below unity at the frequency of said rsonances and that the bandwidth of said feedforward control means is higher than that of said feedback control loop and overlaps the frequency of said resonances. In this way the system is desensitized to high frequency disturbances without loss of performance.
As explained above, the present invention also relaxes considerably the constraints on minimum bandwidth required of transducer signals for rapid motion between positions. This further allows normal velocity feedback and the associated velocity transducing circuitry to be dispensed with entirely, if desired, and the more easily derived position information to be employed in the feedback loop. Preferably, the incremental position of the member from equally spaced reference positions along its path of travel can be used. In such a system there is preferably included means for providing a representation of the velocity of said member, said reference signal generating means comprising an integrator for integrating said velocity representation up and down alternately between positive and negative thresholds to generate said reference signal in the form of an incremental position signal.
In the case of a positioning system according to the invention which employs an electromagnetic coil actuator, it is a preferred feature of the invention to provide that said predictive drive signal generating means includes means for reducing the absolute magnitude of said predictive drive signal as a function of velocity during said deceleration phase of said movement.
This feature enables the system to take account of the back of e.m.f. of a coil type actuator.
Feedforward control may advantageously be employed both during acceleration and deceleration of the member and to that end it is a preferred feature of the invention to provide a positioning system further comprising phase indicating means far indicating acceleration and deceleration phases of said motion, and means for providing a representation of the velocity of said member, said feedforward control means being connected to receive said phase indications and said velocity representation and being responsive thereto to generate said predictive drive signal comprising an initial portion consisting of a constant component of one polarity and a component of opposite polarity proportional to said velocity representation during the acceleration phase and having a final portion consisting of a constant component of said opposite polarity and a further component of said opposite polarity proportional to said velocity signal during the deceleration phase of said movement.

Brief Description of the Drawings

FIGURE 1 comprises a diagrammatic illustration of a preferred positioning system according to the present invention;
FIGURE 2 shows a curve generating circuit suitable for use in the system of Figure 1;
FIGURE 3 shows a control signal generating circuit suitable for use in the system of Figure 1:
FIGURE 4 shows waveforms illustrating the operation of the system and circuits of Figures 1 to 3;
FIGURE 5 shows an alternative positioning system according to the present invention;
FIGURE 6 shows a reference periodic position signal generating circuit suitable for use in the system of Figure 5;
FIGURE 7 shows waveforms occuring in the circuit of Figure 5.

Detailed Description

In Figure 1, there is shown a preferred positioning system according to the present invention. Specifically, the positioning system controls the positioning of heads 10 relative to information bearing concentric tracks on disks 11 of a schematically illustrated disk file. The heads are moved by an actuator 122 of the well known voice coil motor type. The mechanical connection between the actuator 12 and heads 10 is schematically indicated by dashed line 13 and includes a carriage (not shown) for supporting the heads. The motor 12, the heads 10 and disks 11 together with other support components including the head carriage constitute a mechanical system 14. This system as a whole has natural resonance frequencies typically of the order of a few thousand Hertz which, as will be explained further below, may affect the stability of the positioning system, if they are excited and amplified.
The input circuit of actuator 12 comprises a power amplifier 15 which amplifies an input drive signal to provide a current to the actuator coil. A feedforward control signal on line 16 and a feedback control signal on line 17 are summed in summing junction 18 and selectively inverted by inverter 19, depending on the direction of motion, to provide the drive signal to the power amplifier 15.
The feedforward control signal is generated by feedforward current generator 20, the details of which will be explained below in connection with Figure 3. The feedforward current trajectory is illustrated as waveform 101 of Figure 4 and represents the actuator current required for a nominal system to cause the actuator to move the heads from one track to another in a minimum time. If the actual system were exactly the same as the nominal system, the heads would be moved to the target track and brought to rest there in a minimum time without further control being necessary. However, since there will be parameter differences between the nominal and actual systems, the actual response of the system is measured and fed back for use in a feedback control arrangement to ensure accurate positioning.
As in the system described in the above referenced Oswald article, the quantity which is measured to determine the response of the system is velocity. The velocity of the heads 10 moving radially across the disk is determined by a velocity transducer circuitry 21 from the integral of the current in the actuator coil and the derivative of a periodic incremental position signal from position transducer circuitry 22. A suitable circuit for deriving a velocity signal from these inputs is described in US Patent 3 820 712 (Oswald). The position transducer circuitry 22 comprises demodulating circuitry for deriving a position error signal from servo signals read by one of the heads 10 from a dedicated servo surface of one of the disks 11. The principles of such circuits are well known and are described in the above referenced Oswald article and also in US Patent 3 691 543 (Mueller). The demodulated position error signal is a cyclic triangular waveform whose zero crossings correspond to track centre.
The measured velocity signal on line 23 is applied to a summing junction 24 to which is also applied a reference velocity signal on line 25. The summing junction forms the difference between the reference velocity signal and the measured velocity signal which is amplified in error amplifier 26 to provide the feedback control signal on line 17.
The reference velocity signal is conventionally produced in response to a position command at input 30 which loads a difference counter 31 with a value equal to the number of tracks between the current track position of the heads and the target position to which they are to be moved. As the heads move towards the target position, zero crossings of the position signal from position transducer circuitry 22 are detected by a zero crossing detector 32. The zero crossing detector output is a series of pulses each of which decrements the difference counter 31 every time a track is crossed. The output of the difference counter 31 is applied on a bus 33 to a digital-to-analog converter 34 which converts the decreasing count to an analog staircase function representative of the instantaneous absolute position error between the heads and the target track. An interpolator 35 receives the track crossing pulses from zero crossing detector 32 and the velocity signal on line 23 and provides a "fill-in" signal which is summed in junction 36 to smooth the output of the digital-to-analog converter 34. The fill-in signal comprises a falling ramp with a slope proportional to velocity which is reset on every track crossing pulse. Circuits for generating such signals are well known and comprise, for example, an integrator for integrating the velocity input signal, which is reset to a predetermined level by the track crossing pulses. Finally, the smoothed absolute position error signal from junction 36 is applied to a function generator 37 whose output on line 25 is the reference velocity signal as shown in curve 103, Figure 4.
As described in the above referenced Oswald article, the function generator 37 modifies the absolute position error signal in shape according to a predetermined function. A simple function which has been used is a square root function as this represents the variation of velocity with position for a constant maximum deceleration. In practice, the relationship of velocity to position may be a more complex function to allow for the effect of the actuator back e.m.f. and to meet servo system stability criteria. A circuit for generating a second order function having both a squared and linear term is described below in connection with Figure 2. The reference velocity signal from function generator 37 represents the required velocity of the heads 10 while decelerating towards target position with the maximum deceleration attainable by a worst case system.
Two other features of the reference velocity generation circuitry are saturation logic 38 and anticipate circuitry 39. The saturation logic 38 is responsive to outputs of the difference counter exceeding a predetermined count to provide a saturation signal to the input of digital-to-analog converter 34 by way of summing junction 40. The presence of the saturation signal causes the DAC output to maintain a constant maximum output. The corresponding velocity reference signal output on line 25 is also constant under these conditions. Thus, the velocity of the heads is limited to a predetermined value, known as the "coast" velocity, to protect them from damage in the event of a control failure.
The anticipate circuit 39 is effective, while the heads are accelerating, to lower slightly the absolute position error signal and thus the reference velocity curve by an amount proportional to velocity. The accelerate phase of the motion is indicated by the output of a flip-flop 44 which is set at input 45 at the start of each new seek. The output of flip-flop 44 is reset by ground level comparator 41, indicating the sign of the velocity error signal from junction 24 and the end of the accelerate phase.
The inverted accelerate signal from flip-flop 44 and the saturation signal from logic 38 are applied to an AND gate 42 to produce a "coast" mode signal which indicates the portion of the motion when the heads are at coast velocity. This signal is used in the feedforward current generator 20.
Another input to the feedforward current generator is a "stop velocity" indication from threshold detector 43. This indicates that the heads have come substantially to rest and that the seek motion is complete.
Part of the reference velocity generating circuitry is shown in greater detail in Figure 2. The digital-to-analog converter 34 receives the output of difference counter 31 on lines 33 and also the output of saturation logic 38 on additional line 50. The digital-to-analog converter output appears on line 51 and is smoothed by the addition of the fill-in signal from interpolator 35 applied at terminal 52. The anticipate circuitry 30 comprises a switching transistor 53 responsive to an inverted accelerate mode indication at terminal 54 to inhibit the anticipate function. The measured velocity from line 23, Figure 1, is applied at terminal 55 and, when transistor 53 is off, acts to lower slightly the digitial-to-analog converter output level on line 51. By lowering the digital-to-analog converter output and thus the reference velocity curve, as shown by waveform 103 in Figure 4, the sign of the velocity error signal from junction 24, Figure 1, is caused to change early so as to allow time for the current in the actuator coil to reverse.
The function generator 37 of Figure 1 is seen in Figure 2, to comprise an amplifier 60 with a resistive feedback connection to provide a linear term of the required function. A two quadrant transconductance multiplier 60 to provide the second order term of the function. The output at terminal 62 represents the reference velocity signal on line 25 of Figure 1.
A preferred implementation of feedforward current generator 20 together with other associated portions of the system of Figure 1 will now be described in greater detail with reference to Figure 3 and the waveforms of Figure 4.
The inputs to the circuit of Figure 3 comprise the accelerate signal from comparator 41 at terminal 70, the coast signal from AND gate 42, or the stop velocity signal from detector 43 at terminal 71, the reference velocity signal from function generator 37 at terminal 72, and measured velocity from velocity transducer circuitry 23 at terminals 73 and 74.
In the circuit of Figure 3, the reference velocity signal at terminal 72 and the measured velocity at terminal 73 are algebraically summed at node 75, corresponding to junction 24 of Figure 1, to produce the velocity error signal. An operational amplifier 76 amplifies the velocity error. The amplified velocity error is provided at output 79. The amplifier output is limited by diodes 77 and 78 to prevent an excessive output signal during accelerate mode when the velocity error is very large. The limiting function also ensures that the feedback control signal cannot exceed more than a small predetermined fraction (around 15%) of the feedforward control signal.
The feedforward current generator comprises a resistive network for providing a current input to an operational amplifier 80, the input being switchable under control of transistors 81 and 82. During accelerate mode, transistor 81 is off and transistor 82 is on. A current 1₀ flows from positive supply through resistor 83. A current I_v proportional to the velocity signal input at terminal 74 is summed with 1₀ so that a combined current I_o + I_v flows through resistor 84 to the inverting input terminal of amplifier 80. Since transistor 82 is on, current flows from positive supply through resistor 85 to ground and there is no net current through resistor 86. A current of magnitude 21_o flows from the input terminal of the amplifier 80 to negative supply through resistor 87. The net input current from this resistive network to the amplifier during accelerate mode is thus - (l_o - I). An inverted representation of the waveform of this input is shown as the dashed line 100 in the left hand half of the upper waveform of Figure 4. As the velocity rises so does I_v and the level of the inverted waveform falls.
In coast mode, transistor 81 is on and transistor 82 is off. The combined current which had been flowing through resistor 84 is diverted to ground and equal and opposite currents of 21a flow in resistors 86 and 87. The net input to amplifier 80 is zero and the fed foward current is zero. The coast condition is not illustrated in Figure 4 but would merely comprise a zero level portion separating the positive and negative current pulses.
In decelerate mode, which is indicated by the absence of both coast and accelerate signals, both transistors 81 and 82 are off. The equal and opposite currents 21a in resistors 86 and 87 cancel out leaving a net current of 1₀ + I_v through resistor 84. An inverted representation of the waveform of this input current is shown by the dashed line 100 in the right hand half of the upper waveform of Figure 4. The feed- forward waveform is reduced to zero when the heads come to rest at the end of a seek by an input to terminal 71 that is provided by the stop velocity detector 43 of Figure 1. Since the input terminal is shared with the coast indication, the switch states of the circuit of Figure 3 are exactly the same as described above in connection with coast mode.
The velocity factor is introduced to represent the effect of back e.m.f. on current in a high performance electromagnetic coil actuator. The back e.m.f. reduces the voltage applied across the coil in the accelerate mode and is added to the voltage applied across the coil in the decelerate mode.
The amplifier 80 is connected in lag-lead filter configuration with a feedback loop comprising resistors 88 and 89 and capacitor 95. The filter modifies the dashed waveform portions 100 of Figure 4 to the shape of continuous line 101. The filtering action represents the effect of motor coil inductance on the transient response of coil current. It will be noted that the feedback control voltage from amplifier 76 is effectively summed with the feedforward function at the input of the amplifier 80 rather than at the output as suggested by summing junction 18 of Figure 1. This difference has no practical effect. The output waveform 101 as drawn in Figure 4 is that which would be produced in the absence of a feedback control signal.
A final element of the circuitry of Figure 3 is selective inversion circuitry responsive to input commands indicating forward or reverse direction at terminals 90 and 91. Amplifier 94 passes the feedforward signal to output 93 without inversion if line 91, indicating the forward direction of motion is active. Amplifier 92 inverts the feedforward signal at output 93 if line 90, indicating the reverse direction of motion, is active. This circuitry corresponds to the selective inversion circuit 19 of Figure 1.
Having described in detail the structure of the feed-forward plus feedback controlled positioning systems of Figures 1, 2 and 3, some additional description of their operation will be given with reference to Figure 4.
The feedforward current waveform 101 of Figure 4 represents a prediction of the actual current which would exist in the coil of an electromagnetic actuator of a nominal system with full forward then full reverse power applied, less a small margin for control. This waveform is fed forward to the power amplifier of the real system and applied as the actuator input current. The velocity of the heads is thus caused to follow the trajectory 102 in Figure 4. At the same time, the reference velocity signal 103 is generated, as described in connection with Figures 1 and 2, which represents the variation of velocity with distance necessary to bring the heads to rest on the target track in the minimum time, i.e., with a worst case system operating at full reverse power. The reference velocity signal is lowered during acceleration from the dashed curve 103' by the action of the anticipate function.
The reference curve 103 is compared with the actual velocity 102 to provide a feedback control to provide fine correction to the feed- forward action. During acceleration, a feedback control signal is produced, but, because of the large velocity error between curves 102 and 103, is always of the maximum amplitude determined by the limiting diodes of the error amplifier. Thus, the maximum amplitude error signal is simply added as a small increment to the positive portions of the feedforward function and full forward power is applied to the actuator in open loop fashion.
When the actual velocity 102 exceeds the reference velocity at point 104, the acceleration is terminated. The change in sign of velocity error causes the polarity of the feedforward waveform to start to reverse and after a short interval substantially full reverse current is applied to the coil. The actual velocity 102 now exceeds the reference velocity by a small error within the linear range of the error amplifier. A negative feedback control signal is thus added to the negative portion of the feedforward waveform and is effective to maintain the velocity error between curves 102 and 103 to a minimum amount determined by the gain of the feedback loop.
In prior art systems employing only feedback control, the gain of the feedback loop had to be large since the feedback control was required to provide the whole corrective effect. A large gain was achieved by means of the error amplifier but at the expense of increased bandwidth. Problems arose when the gain of the system at mechanical resonance frequencies approached unity. At this gain the system would become unstable. Prior systems therefore always had a minimum velocity error which was determined by the mechanical resonance frequencies of their mechanical components. Various filtering techniques have been proposed but all have a deleterious effect on the performance of the system.
With the addition of feedforward control, according to the present invention, the velocity error would disappear entirely, in theory, if the feedforward function were 100% accurate. However, more realistically, if a feedforward function such as waveform 101 is 90% accurate, then the corrective action required from the feedback control loop is only 10% of what would be required without the feed- forward function. Thus, the gain of the error amplifier and, roughly speaking, the bandwidth of the feedback loop can be reduced by a factor of ten while maintaining the velocity error between curves 102 and 103 as small as before. This considerably reduces the limitations on performance imposed by the mechanical resonance frequencies of the system. If mechanical resonances are not a problem, then the gain can be maintained high and the velocity error very much reduced over what was possible with feedback control alone.
Thus, in the system of Figures 1, 2 and 3, the gain of the error amplifier is set sufficiently low to reduce the bandwidth of the feedback loop to a few hundred Hertz, well below the lowest resonance frequency of a few thousand Hertz. The lag-lead filter formed by resistors 88 and 89 and capacitor 90 in Figure 3 does reduce the bandwidth somewhat, but the effect is insignificant compared with that of the gain of amplifier 76.
In Figure 5 there is shown another embodiment of the present invention which makes use of the reduction in feedback loop bandwidth permitted by the addition of feedforward control, to employ a position signal directly as the feedback controlled variable.
The system of Figure 5, like the system of Figure 1, is a system for positioning magnetic heads 210 in relation to tracks on disks 211 of a disk file by means of an electromagnetic voice coil actuator 212. The actuator input circuit comprises a power amplifier 215. The control signal to the power amplifier input comprises a feedforward signal on line 216 and a feedback signal on line 217 which are summed in junction 218 and selectively inverted by inverter 219 in dependence on the direction of motion. The feedforward signal is provided by feed- forward current generator 220 which operates in exactly the same way as the generator 20 of Figure 1, though the inputs to the generator are derived somewhat differently as will be described below.
A periodic position signal is derived by position transducing circuitry 222 from servo signals ready back by one of the heads 210 from a dedicated servo surface on one of the disks 211. The operation of the circuitry and the form of the triangular position signal is exactly the same as for the circuit of Figure 1. However, unlike Figure 1, no velocity transducer circuitry is provided and the periodic position signal is fed back directly to a summing junction 223 for comparison with a reference periodic position on line 224. The difference signal from junction 233 is alternately inverted by inverter 228 in dependence on the slope of the reference periodic position signal as detected by slope detector 227. The alternately inverted difference signal constitutes the position error signpl and is amplified by error amplifier 225 to provide the feedback control signal.
The reference periodic position signal is generated by integrating a reference velocity signal repeatedly up and down between predetermined levels in incremental integrator 226, the operation of which will be described below in connection with Figures 6 and 7. The reference velocity signal comprises both an accelerate and decelerate portion and feedback control is thus available for the complete duration of the motion. The decelerate portion of the reference velocity signal is provided in very similar fashion to Figure 1. A difference counter 230 is loaded at terminal 231 with a value representing the number of tracks to be crossed. The difference counter is decremented by output pulses from zero crossing detector 232 during the motion and its output converted to an analog function by digitial-to-analog converter 233 and smoothed by fill-in signals from interpolator 234. The absolute position error signal thus derived from summing junction 235 is applied to decelerate function generator 236 to produce a reference velocity signal in the manner of Figure 1.
The accelerate portion of the reference velocity signal is produced somewhat similarly. An up counter 240 is set to zero as difference counter 230 is loaded with the difference count. A digitial-to-analog converter 242, summing junction 243 and accelerate function generator 244 produce a rising curve representing the required velocity for time optimal motion at successive positions. The fill-in signal available from interpolator 234 is inverted in inverter 245 before it is applied to summing junction 243. Saturation logic 246 indicates that the up counter has reached a certain predetermined value upon which further counting is inhibited by means of inhibit gate 247.
The acceleration curve and the deceleration curve are passed through a circuit 248 for passing whichever has the lower value. The output of this circuit is the reference velocity curve which is input to the incremental integrator 226. A comparator circuit 249 provides an output signal indicating which of the acceleration and deceleration curves is of greater magnitude. This indication identifies the acceleration phase of the motion and is applied to the feedforward generator 220 as an input.
A second input to the feedforward generator is a "coast" signal, provided by the AND gate 250 from the output of saturation logic 246 and the accelerate signal from comparator 249. A second input to the same line is provided by stop velocity detector 251 which detects when the reference velocity effectively falls to zero, indicating that the seek is complete.
A preferred form of incremental integrator and associated switching circuitry suitable for use in the general system of Figure 5 is shown in Figure 6. Waveforms produced by the circuitry of Figure 6 are shown in Figure 7. The circuitry of Figure 6 is directly applicable to the system of Figure 5 with the modification that two phases of position signal (both measured and reference) are provided. The two phases are of identical form to the sawtooth signals described in the Oswald article, referenced above, but are phase displaced by 90 degrees. One signal is normally referred to as the "normal" (in phase) position signal and the other as the "quadrature" position signal.
As shown in Figure 6, the measured in phase and quadrature position signals are applied at terminals 310 and 311 for comparison with reference quadrature position signals N and Q, Figure 7 in junctions 312 and 313 respectively. The junctions 312 and 313 correspond to the summing junction 233 of Figure 5 and their outputs are altemately selected by logic to be described to remove the effect on the position error signal of the slope changes and peaks of the position signals. A single position error output signal is provided at output 314.
The two phases of reference periodic position signal are produced by applying the reference velocity signal from circuit 248 (Figure 5) to the input 319 of a selective inverter 320. The inverter is controlled by a signal d, Figure 7, from the normal output of a set/reset flip-flop 321. When signal d is up (=1 the reference velocity signal is passed through circuit 320 without inversion. When signal d is down (=0), the reference velocity signal is inverted. An integrator 322 integrates the alternately inverted reference signal to produce a signal a, Figure 7, which is of triangular form and resembles a single phase position signal. The alternation of the flip-flop 321 is controlled by comparators 323 and 324 which compare the magnitude of the integrator output a with predetermined reference levels +V/2 and -V/2. Thus, the integrator output reverses slope every time one of the levels +V/2 is reached.
The normal and inverted outputs of flip-flop 321 are used to clock respective data/clock flip- flops 325 and 326 which produce output signals e and f as shown in Figure 7. These signals are at half the frequency of signal d and are 90 degrees displaced in phase from each other. They are employed to switch selective inverters 327 and 328 in the generation of reference periodic position signals N and Q.
The two signals N and Q are produced by applying waveform a to a level shifting network including amplifiers 329 and 330 to produce two intermediate signals N' and Q', Figure 7, which are centered about +V/2 and -V/2 volts respectively. Application of these intermediate signals N' and Q' to selective inverters 327 and 328 produce the reference periodic position signals N and Q, Figure 7, which are of twice the amplitude and half the frequency of intermediate signals N' and Q'.
The reference position signals N and Q from inverters 327 and 328 are next compared with the measured in phase and quadrature position signals in summing junctions 312 and 313. For the output of the summing junctions to represent the position error, the effects of slope change and inversion of the position signals must be removed. To accomplish this, a switch circuit 331 is employed to select either the "in phase" or the quadrature position error in dependence on the value of a waveform b, also shown in Figure 7. The waveform b is produced by an over- driven comparator 332 in response to the waveform a. The switch 331 operates to select alternately only the position error signal derived from central linear portions of the position signals. This signal will invert according to whether the slope of the position signals is positive or negative when the comparison is made. To remove the slope sign dependence a selective inverter 333 is interposed between the output of switch 331 and output terminal 314. The selective inverter is controlled by a waveform c, shown in Figure 7, derived by data/clock flip-flop 334 from waveform b.
As was the case with Figure 1, the system of Figures 5, 6 and 7 employs feedback control only as a fine correction imposed on the basic feedforward control. As discussed in connection with Figure 1, the use of approximate feed- forward control permits the gain and bandwidth of the minor feedback loop to be significantly lower than where feedback control alone is employed. In the system of Figures 5, 6 and 7, this fact permits the use of the position transducer output directly as a feedback controlled variable. In a pure feedback positioning system, a position feedback loop is not used where high performance is required since the bandwidth available with position signal feedback is low compared to that of a velocity feedback loop.
Although the system has been described in terms of substantially continuous position information from a dedicated servo disk surface, it is also suitable for use with positioning systems where position information is only available at relatively infrequent sampling times. In a disk file context, such a system would be of the type where no dedicated servo surface is provided but, instead, servo information is distributed in sectors on the data surfaces of the disks. In such systems, position information is sampled at servo sector times and interpolated in between sectors. The bandwidth available is much lower than with a continuous source of servo position information. The systems of Figures 5 and 6 are easily modified to accommodate a sampled source of servo information. For example, the position transducer circuitry 222 could have a provision for sampling the information from heads 10 at sector times only and for holding sampled position signals, or interpolating between them, between sectors. Alternatively, the output of the error amplifier 225 could be sampled and then held or interpolated between sectors. The use of feed- forward control in conjunction with such a sampled system would allow a relatively high performance to be achieved.
Finally, it will be apparent to one skilled in the art that the disclosed invention may be applied not only to disk file head positioning, but to other positioning systems in general.

Claims

1. A positioning system, responsive to a position input command to move a member (10; 210) along a predetermined path of travel between a current and a target reference position defined by said command, each such movement being completed before a subsequent command is accepted and comprising a single acceleration phase in which such a member is accelerated to a maximum velocity and a single deceleration phase in which said member is decelerated from said maximum velocity to a state of rest at said target reference position, said system comprising an electrically controlled actuator (12; 212) for moving said member along said predetermined path in response to electrical drive signals and having an input circuit (15, 18, 19; 215, 218, 219) to which said signals are applied, reference signal generating means (31, 34, 37; 230, 233, 236, 226) responsive to said position input command to generate a reference signal (103; N, Q) at least a portion of which represents the variation with time of a position related attribute of said member (10; 210) for deceleration of said member in a predetermined manner during said deceleration phase, and a feedback control loop including a transducer (21; 222) for providing a signal indicating the value of said position related attribute of said member, error determining means (24, 26; 223, 225, 312, 313) for producing an error signal representing any difference between said transducer signal and said reference signal, and feedback means (17; 217) for feeding back said error signal to said actuator input circuit (15, 18, 19; 215, 218, 219) to cause said actuator (12; 212) to move said member in a direction to reduce said error signal characterized in that said system further comprises feedforward control means including predictive drive signal generating means (20; 220) for generating, concurrently with said reference signal generation, a time varying predictive drive signal (101) at least a portion of which represents the input to the actuator of a nominal system required to decelerate such a member in said predetermined manner during said deceleration phase, and means (16, 216) for feeding forward said predictive drive signal to said actuator input circuit, whereby deceleration of said member is coarsely controlled by said feedforward control means, and is concurrently finely controlled by said feedback control loop to correct for fine errors resulting from the differences between said nominal and actual systems.

2. A positioning system as claimed in claim 1 wherein said predictive drive signal generating means (20) includes means (74, 81) for reducing the absolute magnitude of said predictive drive signal as a function of velocity during said deceleration phase of said movement.

3. A positioning system as claimed in claim 1 or claim 2, further comprising phase indicating means (42, 43, 44) for indicating acceleration and deceleration phases of said motion, and means (21) for providing a representation of the velocity of said member, said feedforward control means being connected to receive said phase indications and said velocity representation and being responsive thereto to generate said predictive drive signal comprising an initial portion consisting of a constant component of one polarity and a component of opposite polarity proportional to said velocity representation during the acceleration phase and having a final portion consisting of a constant component of said opposite polarity and a further component of said opposite polarity proportional to said velocity signal during the deceleration phase of said movement.

4. A positioning system as claimed in claim 3 wherein said phase indicating means (42, 43, 44) also indicates a coast phase corresponding to a predetermined constant velocity, said feed- forward control means (20) being responsive to said coast phase indication to reduce said predictive drive signal to zero, and wherein said predictive drive signal generating means comprise an amplifier (80), means (87) for applying a constant current of one sense, 21₀, to an input of said amplifier, first switching means (81) responsive to the absence or presence of said coast phase indication to switch a current, of opposite sense, (1₀ + I), respectively to or away from said amplifier input, where I_v is proportional to said velocity representation, and second switching means (82) responsive to said absence of said accelerate phase indication to switch a current of said second sense, -21₀, respectively to or away from said amplifier input.

5. A positioning system as claimed in any preceding claim wherein said position related attribute of said member (10) is velocity.

6. A positioning system as claimed in any one of claims 1 to 4 wherein said position related attribute of said member (210) is position.

7. A positioning system as claimed in claim 6 wherein said position related attribute of said member (210) is its incremental position, being the incremental deviation of said member from equally spaced reference positions along said path of travel, so that said reference signal and said transducer signal are cyclic incremental position signals.

8. A positioning system as claimed in claim 7, including means for providing a representation of the velocity of said member (210), said reference signal generating means comprising an integrator (226) for integrating said velocity representation up and down alternately between positive and negative thresholds to generate said reference signal in the form of an incremental position signal.

9. A positioning system as claimed in claim 8 wherein said reference signal gnerating means comprises said means for providing a representation of the velocity of said member, which means include : an up counter (240) means (241) for setting said counter to zero in response to said position input command, a difference counter (230), means (231) for loading said difference counter with a count representing the distance between said current and target positions in response to said position input command, means (232) for incrementing said up counter and decrementing said difference counter in step with the cycles of one of said cyclic incremental position signals, means (242, 244) for deriving an increasing velocity signal from said up counter according to a predetermined function, said increasing velocity signal indicating the required change of velocity with time for acceleration of said member by said actuator in a predetermined manner during said acceleration phase, means (233, 236) for deriving a decreasing velocity signal from said difference counter according to a predetermined function, said decreasing velocity signal indicating the required change of velocity with time for deceleration of said member by said actuator in said predetermined manner during said deceleration phase, and means (248) for determining the smaller of said increasing and decreasing velocity signals and providing said smaller signal as said velocity representation to said integrator (226) for generating said reference incremental position signal.

10. A positioning system as claimed in any one of claims 7 to 9 wherein said error determining means comprise: a summing junction (223) to which the incremental position signal from said transducer and said reference incremental position signal are applied to develop a difference signal therebetween, means for inverting said difference signal in alternating fashion in dependence on the slope sign of said incremental position signals, and an error amplifier (225) for amplifying said inverted difference signal to produce said position error signal.

11. A positioning system as claimed in any one of claims 7 to 9 wherein said transducer produces two incremental position signals of first and second phase and wherein said reference signal generating means (322, 327, 328) produces two reference incremental position signals of first and second phase, said error signal generating means comprising: two summing junctions (312, 313) to each of which a respective phase of said transducer incremental position signal and said reference incremental position signal is applied, said summing junctions producing first and second difference signals therefrom, means (331) for alternately selecting said first and second difference signals, means (333) for inverting said selected difference signal in dependence on the slope sign of said incremental position signal of corresponding phase, and an error amplifier for amplifying said inverted and selected difference signal to produce said position error signal.

12. A positioning system as claimed in any preceding claim and which is subject to mechanical resonances of bandwidth overlapping the bandwidth of said transducer signals, wherein the bandwidth of said feedback control loop as a whole is arranged to be predominantly below the frequency of said resonances, such that the overall loop gain is substantially below unity at the frequency of said resonances, and in which the bandwidth of said feedforward control means is higher than that of said feedback control loop and overlaps the frequency of said resonances.

13. A positioning system as claimed in claim 12, in which said error determining means include an error amplifier (76) for amplifying any difference between said transducer signal and said reference signal, the gain of said amplifier being sufficiently low to reduce the overall loop gain below unity at the frequency of said resonances.