DE4021800A1 - Controlling positioning unit e.g. for machine tool - by continuous comparison of actual and target positions, deriving control voltage from filtered step generated control curve - Google Patents

Abstract

The method of controlling a positioning unit (2) of an automatic machine, e.g. a machine tool, from a starting position to a target position involves continuous comparison of the instantaneous position (xa) of the positioning unit as the actual value with the target position (xn) and using a demand value curve dependent on their difference to derive the control unit current. The positioning unit is accelerated using maximum current up to a defined limit speed and decelerated at the end of the positioning process. Sampled values of the step generated curve are filtered (52) to round off steps and processed to generate a control voltage. USE/ADVANTAGE - Method enables further reduction in mean positioning time to be achieved with improved elimination of unwanted mechanical vibrations.

Description

The invention relates to a method for controlling a Positioning unit of a machine according to the preamble of Claim 1 and a device for implementation this procedure.

In a number of areas of technology is a variety of Use cases where a positionable unit in a given coordinate system from a current one Actual position as quickly as possible and at the same time precisely in a new one Able to be referred to as the current target position to perform a given function in this target position. Examples of such applications can be found at Machine tools, in particular program-controlled machines, such as drilling and milling machines, in robotic devices such as manipulators, with so-called X / Y slides, with automatic placement machines or even in laser technology, where a laser beam like e.g. B. in laser welding, precise relative to that to be machined Workpiece must be positioned.

Without this list of use cases conclusive However, it is already clear from it that in these Use cases again and again the comparable task is to solve, in a given coordinate system, with a given Grid an adjustable positioning unit if possible quickly and precisely from a known current position to one control defined target position. The coordinate system only linear, but also two- or three-dimensional be.

Detailed explanations cannot be given for all applications are shown, therefore in the following is a special  Application referred to this problem more closely to explain. Another, at the same time also special boundary conditions typical use case namely with magnetic disk storage. There is a magnetic head supporting positioning unit with respect to a magnetic disk stack set to predetermined tracks on the magnetic disks, to carry out write or read operations. The reference on this use case has the further advantage that the positioning here, with a permissibly simplified view, basically done one-dimensionally and thus the principle easier to display.

The average access time is known for magnetic disk storage in addition to the storage capacity one of the essential Performance criteria. The track density and the average access time, this being the pure positioning time and include settling time on a selected track may, the structure of the positioning unit and the positioning process. For example, the group is more efficient 5¼ ″ magnetic disk storage the conventional Positioning procedure in that one depends on the track distance Speed profile, d. that is, a certain speed as a function of the distance covered during the positioning process Distance, as a setpoint curve and the sequence of the positioning process then regulates. An analysis of the positioning process on this basis, for example, in "Magnetic Recording ", Vol. II, edited by C. D. Mee and E. D. Daniel, McGraw Hill Verlag, 1988, described in chapter 2.3.5. With this setpoint specification in the form of a speed profile access times of around 20 ms are still manageable.

If, on the other hand, you want to significantly reduce the positioning times, d. that is, you have to reach an average of 14 ms and less Above all, two options: You can use the positioning unit so build that higher maximum speed and greater Acceleration becomes possible. With conventional positioning processes there are also limits to these possibilities.  One skilled in the art immediately understands that the goal, despite the higher maximum speed of the positioning unit Drive into the target lane at least as precisely as with conventional ones Magnetic disk storage, only with an appropriate one Increasing the "dynamic range" of the speed control is feasible. But there are conventional standard circuits, for use in positioning procedures of magnetic disk storage were no longer suitable.

Another limitation is that there is also an increase the permissible acceleration is not easily achievable is. This would mean, first of all, what in terms of a time-optimal solution of the positioning task itself reasonable appears, the current to accelerate the drive of the positioning unit to invest as soon as possible. One tries but to achieve the goal in this way you will see that sudden changes in the acceleration of the positioning unit directly stimulate mechanical vibrations, d. i.e., the Extend settling time and z. B. also cause operating noise.

For this purpose, from international patent application WO 88/02 913 already known, despite the goal of optimal positioning an acceleration profile instead of a speed profile to specify for the positioning unit such that the acceleration or deceleration is preferably only linear changes.

In this known positioning method, three groups are used of positioning processes depending on the occurring Distinguished track distance. When positioning below the positioning unit reaches a certain track distance not yet the maximum specified acceleration. In this The positioning unit falls in the first quarter of the positioning time accelerated linearly, then the change in acceleration returns their direction around so that at three quarters of the positioning time  the greatest delay occurs, which is linear thereafter drops again until the end of the positioning process. The Acceleration curve over time therefore corresponds to two Time axis symmetrical, directly placed next to each other Triangles.

In a second group of positioning operations over medium Track distances become the maximum acceleration or deceleration not yet, but the maximum speed of the Positioning unit reached. The acceleration profile contains therefore additional sections with constant acceleration or Delay and corresponds to two symmetrical with respect to the time axis mutually trapezoids, which are placed directly next to each other are.

In the third group of positioning operations over large In the meantime, track speeds will also become the limit speed the positioning unit is reached. The acceleration profile then corresponds to two symmetrical to the time axis lying trapezoids that are spaced apart from each other lie.

In all of these cases, the acceleration portion of the overall profile with the opposite sign identical to the delay component, d. that is, the setpoint curve is point symmetrical. This symmetry makes it much easier to determine each one Setpoint profile under real-time conditions as well the requirement that the acceleration is in all positioning processes or the delay with a fixed predetermined Slope changes linearly.

It can be seen that this approach helps reduce the operating noise because extreme values for changes the acceleration can be avoided as required. It can also be seen that settling times with the known method the positioning unit over the target track shortened and therefore average access times despite the speed-maximized compared to  Positioning process no longer optimal Setpoint specification can be achieved the same or even better than these older conventional ones Method. On closer inspection, however, it is still immediate obvious that even with this known method the task of access time while reducing operating noise to reduce, has not yet been optimally resolved. Each of the acceleration profiles described contains more or less Many discontinuities in the time derivative of this acceleration function, which are still the reason for the excitation mechanical vibrations and thus for the generation of operating noises are, though due to the gradual increase and Decrease in acceleration or deceleration an extension the positioning time is accepted.

The present invention is therefore based on the object to create a process of the type mentioned that at further reduction in the mean positioning time causes for the generation of unwanted mechanical vibrations better as previously eliminated and a facility for implementation to specify this procedure with inexpensive construction a high degree of flexibility to adapt to the respective application having.

In a method of the type mentioned at the outset, this task by those described in the characterizing part of claim 1 Features solved.

A major advantage of the solution according to the invention is in that it succeeds in accelerating the drive the positioning unit in the entire course of the setpoint curve "soft", i.e. H. in other words, without discontinuities to be rounded off. But it won't be a linear increase the acceleration or deceleration accepted, rather, the aim is rather to optimize time by a corresponding current specification maximum acceleration or To achieve deceleration values as quickly as possible. This carries  to shorten the mean positioning time as well as the rounding of the acceleration profile at the end of a positioning process, which with optimized transient response of the Positioning unit as accurate a retraction as possible Target position allowed. In other words, the positioning process is based on a setpoint curve that is rounded off is set in a time-optimized manner and yet also with regard to settling to the target position is optimized.

As further developments of the invention show, after which the positioning process regardless of the respective positioning distance is always divided into three time intervals, it is possible all positioning operations without case distinction after one to regulate uniform positioning procedures. It is immediate to see that this means that the distance d. that is, the positioning distance, dependent determination of a certain one To simplify profiles as a setpoint is. This increases the flexibility of the invention Procedure with reduced effort and helps to shorten the time Computing time at what is happening under real-time conditions Operations is of crucial importance.

Another significant advantage of the method according to the invention is based on the fact that this is also able to in the control for the drive of the positioning unit Reserves, d. that is, to make full use of the influence of the back EMF. In other words, this means detaching the setpoint profile of that with conventional positioning methods point symmetry given for reasons of simplification. The detachment the point symmetry practically results in an in shortened deceleration interval with respect to the acceleration interval and thus uses a further influencing variable, which is used for Reduction of the average positioning time contributes.

How a further development of the invention according to claim 7 shows, this method is using a commercial signal processor capable of the control process  to perform under real time conditions without that of the respective positioning distance-dependent profiles of the setpoint profile previously fixed, for example in a setpoint memory, are specified. Therefore, there is a need for a facility to carry out this procedure no specialized, integrated so far not available on the market Circuits. Rather, the method can be used alone realize conventional building blocks, whereby a conventional, with regard to certain addition and multiplication processes optimized signal processor with its properties is fully used. This possibility also contributes to the The inventive method in a variety of applications use with advantage, since the adaptation to the respective Use case can be carried out relatively easily. At These adjustments - based on the principle - play none significant difference whether the positioning process along a linear path, i.e. one-dimensional or multi-dimensional he follows. When positioning in a two-dimensional or spatial coordinate system are the procedural steps only performed for each given coordinate direction.

Exemplary embodiments of the invention are described below the drawing explained in more detail. This is done for reasons of clarity preferably for an application in a magnetic disk storage Referenced because of the positioning in in this case essentially a one-dimensional coordinate system underlying. It shows

Fig. 1 shows schematically in a block diagram of a control unit for a magnetic disk storage with an actual value and a setpoint generator in which target values are generated for performing the control for each positioning operation,

FIGS. 2a to 2c schematically in the form of timing diagrams, the principle of a velocity profile of a corresponding current profile for the drive of the positioning unit, as well as a profile for the terminal voltage to a drive coil of the positioning unit,

Fig. 3 is an equivalent circuit diagram for the drive of the positioning unit,

FIGS. 4a and 4b, 4c and 4d and 4e and 4f respectively, the comparison of a dog world acceleration profile and a corresponding this amplitude spectrum at the basis of a typical mechanical transmission function for explaining the excitation of mechanical vibrations and operational noises

Fig. 5 shows the equivalent circuit diagram for a FIR filter,

Fig. 6 corresponding in form of pulse diagrams a step function and a step function of this filter response of an FIR filter,

Fig. 7 is a block diagram for a target value generator of FIG. 1,

Fig. 8 in the form of a timing diagram of an example of a current or acceleration profile, as for example, initially determined for all positioning,

Fig. 9 shows an embodiment of a setpoint generator in hard-wired circuitry,

FIG. 10a to 10c in different pulse diagrams each embodiments for an acceleration profile, a velocity profile and a path profile, which is derived with the set value transmitter of Fig. 9 after setting an acceleration profile of FIG. 8,

Fig. 11 shows diagrammatically in a block diagram of another embodiment of a device for implementing the positioning method using a signal processor,

Fig. 12 shows diagrammatically in a block diagram of the processes taking place under control of the signal processor processes the adjustment of the in Fig. 8 represented acceleration or current profile in a profile of FIG. 10a and

Fig. 13 is a flowchart for the control or computing operations performed by the signal processor.

In FIG. 1, a device control 1 is shown schematically in a block diagram to explain a positioning method for a magnetic disk storage device, via which the magnetic disk storage device connects to its environment. The device controller controls the data transfer from and to the magnetic disk storage and - which is of more important interest in the present case - also triggers a lane change if a writing or reading process is to be continued in another track of the magnetic disk storage. Such device controls are generally known. There is therefore no need for any further explanation here, because when a lane change is to be carried out, this device control merely provides a distance address ADR which outputs a corresponding jump distance in the form of the number of recording tracks to be overrun and the respective direction and a start signal START which defines the start of the lane change operation.

In Fig. 1, a positioning unit 2 for the magnetic disk memory is also shown schematically, which is to be designed as a rotary positioner. Rotary positioners for magnetic disk storage are particularly common for 5¼ ″ or 3½ ″ disk storage, so that no detailed description appears necessary for this unit either. It should therefore only be pointed out here that conventional rotary positioners have a swivel piece which is rotatably mounted in the disk storage housing. On the one hand, the swivel piece carries, via a plurality of radially protruding support arms, a corresponding plurality of magnetic head systems which are precisely aligned with one another and, when the swivel piece swivels, pivot in together over the surfaces of the storage disks of the magnetic disk storage assigned to them. The swivel piece is driven by a magnetic drive coil, to which a corresponding coil current is fed via a controlled power controller. The magnetic drive coil is surrounded by a system of permanent magnets, so that it moves in dependence on the coil current and thus deflects the swivel piece. Such a drive device 201 of the positioning unit 2 is only indicated schematically in FIG. 1.

Typically, write / read electronics 202 , which in particular contains write or read amplifiers assigned to the magnetic head systems, are also arranged directly in the positioning unit 2 . It is assumed here that this read / write electronics 202 additionally also includes evaluation electronics for processing, rectifying and selectively comparing servo signals. It is thus able to feed these prepared read signals RS to a phase-locked loop 3 , with which, in the case of magnetic disk memories, a dependent internal clock pulse sequence CLK is derived from these read signals RS, which forms the time basis for all internal control processes and, for reasons of simplicity, also for example the control of the control processes when positioning is used.

In addition, this read / write electronics 202 in the form of the evaluation electronics should also have switching devices which generate position error signals PES from the read servo signals. These usually form the control signals for tracking in magnetic disk memories. As indicated schematically, the position error signals here have a sawtooth-shaped course, the zeros of which occur when the positioning unit 2 continuously crosses recording tracks in the course of a track change.

All of the units described above are conventional Magnetic disk storage in a wide variety of embodiments used and are therefore consistently in the professional world well known. Beyond the introduction above Therefore, no further detailed description is required, which is only the clarity could complicate.  

In the following, however, the positioning process in a magnetic disk memory is to be considered in principle with regard to the positioning method to be described. The magnetic disk memory is a memory with random access, ie the positioning system must be adjustable to any track on the magnetic disk each time a track is changed. In addition, every positioning process should always run optimally, since the average positioning time is a key performance criterion for magnetic disk memories. A predetermined supply voltage V _s is available in the drive device 201 of the positioning unit 2 and is to be used optimally in order to supply the positioning device 2 with maximum acceleration current in principle in the acceleration phase of the magnetic drive coil until the predetermined maximum speed of the positioning unit 2 is reached. This speed is maintained until the braking phase begins, in which the positioning unit 2 is braked with the maximum braking current, until it enters the target track as precisely as possible. Naturally, the braking phase starts at short distances, ie short track distances, before the maximum speed is reached.

This theoretical consideration makes it clear that a specific speed profile v _i (t) can be used as a basis for each track distance in order to optimally carry out the positioning process.

A speed profile v _i (t) is shown in FIG. 2a, FIG. 2b shows a corresponding current curve i (t), that is to say a curve which corresponds to the time profile of the current supplied to the magnetic drive coil in the drive device 201 as a control variable, and FIG . 2c shows the corresponding curve of the terminal voltage V (t) to the drive means 201 of the positioning unit. 2

In connection with an equivalent circuit diagram for the drive device 201 shown in FIG. 3, the influences which result from the structure of the drive device in practice are to be explained first. FIG. 3 shows a series connection of a current source 203 and a magnetic drive coil 204 of the drive device 201 arranged between the supply voltage V _s and ground. In practice, the current source 203 is implemented by a power amplifier, to which a control voltage for setting the coil current i (t) is supplied. This power amplifier has a predetermined own requirement, which is effective as a voltage drop. You also need a certain voltage reserve to be able to regulate during the positioning process. These two influencing variables are indicated in Fig. 2c as partial voltage V₁. The equivalent circuit diagram of the magnetic drive coil 204 illustrates the ohmic resistor 205 and the inductive resistor 206 of the magnetic drive coil, which are predetermined by the dimensioning of the coil 204 . Finally, the influence of the back emf is indicated in the form of a voltage source 207 , which builds up as a voltage contribution that is linearly dependent on the speed and the motor constant K of the magnet drive coil 204 .

If the terminal voltage on the magnetic drive coil 204 is to be kept constant despite the increasing speed of the positioning unit 2 , a coil current i (t) must be specified, as is shown in FIG. 2b. The coil current i (t) follows an exponential drop over time according to the relationship (1):

i (t) = I ₀ exp (-K²t / R), (1)

where I ₀ = (V _s -V ₁ ) / R

K means the motor constant, R the ohmic resistance of the magnetic drive coil 204 , V _s the supply voltage and V₁ the partial voltage, which corresponds to the self-consumption of the controlling power amplifier 203 and the control reserve.

In the positioning process described here, this is Influence of back EMF is used to advantage. The back emf acts as a brake when accelerating, but can then in the braking phase  can be used as a support. However, this has to Consequence that the setpoint curve for the coil current i (t) is asymmetrical will, d. that is, the braking phase becomes shorter than the acceleration phase.

FIGS. 2a to 2c also illustrate as an example that to control the positioning process, a specific profile must be specified for each possible track distance in order to carry out the positioning process in an optimally time-efficient manner. In the case of a plurality of conventional magnetic disk memories, a trapezoidal speed profile v _i (t) is used for this. But if, for example, with 5¼ ″ magnetic disk storage media today the aim is to achieve average positioning times of less than 15 ms, then this means that the control technology can only be mastered with considerable effort due to the additional constraint that the target track must be approached as precisely as possible. In the present case, therefore, a path profile x _n (t) is used for controlling the positioning process in connection with a current specification in (t), the latter determining the acceleration of the positioning unit 2 in the initial phase or its deceleration in the final phase of a positioning process.

So far, another essential problem has not been considered in detail, namely the problem of the excitation of mechanical vibrations that result from the oscillating behavior of the positioning unit 2 . Theoretically, it is desirable to apply maximum current to the drive device 201 of the positioning unit 2 as quickly as possible, in order to allow the positioning process to be carried out in a time-optimized manner at high accelerations. This would be possible with a current profile i (t) or a corresponding acceleration profile a (t) with a step-like course, as shown by way of example in FIG. 4a. In a typical mechanical transfer function according to the positioning model considered here, such a current curve causes mechanical vibrations, the amplitude spectrum of which, ie the vibration amplitudes A plotted against the frequency f, qualitatively produces the image shown in FIG. 4b. The amplitude spectrum drops exponentially from a maximum value occurring at low frequencies, but still has relatively high maximum values even in the range of 10 kHz.

FIG. 4c schematically shows a current curve i (t) or the corresponding acceleration curve a (t), which has a linear course that is pieced together. The associated amplitude spectrum A shown in FIG. 4d with the same transfer function shows a significantly more favorable curve in comparison to the illustration in FIG. 4b, but at least also significant amplitude values at frequencies above 5 kHz.

In Fig. 4e a current and acceleration curve i (t) and a (t) is now shown, the flanks of rise or fall relatively steeply, but which is rounded in the transition areas, for example, according to a cosine function. The associated amplitude spectrum A shown in FIG. 4f shows the best values in a comparison of all FIGS. 4b, 4d and 4f. This shows convincingly that it is very important to design the current curve i (t) in such a way that all transitions run as smoothly as possible in order to avoid the excitation of mechanical vibrations and thus malfunctions in the positioning process, such as long settling times and high operating noises. In other words, if only changes in the current or acceleration over time excite the mechanics to vibrate and then it is possible to specify the coil current i (t) with as few harmonics as possible, higher resonances in the amplitude spectrum can be avoided.

In this case, approximately the relationship (2) applies mathematically expressed transfer function for the assumed positioner model:

F (s) = x (s) / i (s) = K (Rs ² + βs + d) ^-1 (2)

Here x is the deflection related to the servo head, R is the moment of inertia of the positioning unit 2 , β is the friction proportional to the speed and d is a restoring force of the positioning unit, which can be attributed, for example, to the flexible cable feeds. Now one can assume that positioning times in a range from 1 ms to a maximum of 50 ms occur. Expressed in the frequency domain, this means fundamental frequencies from 200 Hz to about 1 kHz in the excitation spectrum. For these frequencies, the relationship (2) can be approximately expressed by a relationship (2a):

F (s) = K / Rs² (2a)

This approximate relationship corresponds to the fact that the displacement of the positioning unit 2 is proportional to the double integration of the current in the magnetic drive coil 204 , as long as the coil is not oversaturated.

Now it must be made clear that such a current curve i (t), which differs from linear sections, must be made available during the positioning process. This is a task that is hardly feasible under real-time conditions with such a complex profile. As stated above, the current curve i (t) cannot be rounded. Now analog technology generally offers to round a pulse train by pulse shaping, ie filtering. If this were also possible here, an unrounded pulse sequence could be specified as a "hard" current profile i (t) and rounded off by filtering. The above-mentioned requirements may apply that the path x can be represented as a double integral of the acceleration a and that the acceleration a is directly proportional to the coil current i, ie that there is linear, saturation-free behavior. Furthermore, it is assumed that an analog control variable i _s (t) fed to the drive device 201 of the positioning unit 2 is linearly proportional to the current curve i (t). Under these conditions, it would be irrelevant whether this control variable is based on an already rounded current profile with the corresponding path x or whether one first specifies a hard current profile i (t), as indicated for example in FIG. 2b, with the associated path and then both in sends functions expressed in electrical signals through the same filters, thus rounding them off.

This is opposed to the fact that defined starting or end conditions apply to exact positioning, namely that exactly defined positions are taken by the positioning unit 2 at the beginning or at the end of each positioning process and at the same time there at least the first and second derivatives of the path according to the position Time must be zero. If you think of the filtering under consideration as being implemented by means of a low-pass filter, it can immediately be seen that the end condition cannot be met in any case, since an exponential entry into the end position is systematically predetermined according to the transfer function of such a filter. In theory, this means that the end position would only be taken exactly after an infinitely long time. Common analog filters are therefore not an acceptable solution to this problem.

However, finite impulse response filters are known from digital filter technology, which are commonly referred to as FIR filters. A circuit design of such a filter is. e.g. B. in IEEE Journal of Solid-State Circuits, Vol. 23, No. 2, April 1988, pages 536 to 542, and discussed in detail. Knowledge of this type of filter can therefore be assumed here. Therefore, in FIG. 5 only better understanding is depicted schematically filter FIR because an equivalent circuit diagram for such. Since the rules of the z-transformation are to be applied to digital filters, each block z ^-1 in this equivalent circuit diagram means the shift of the input signal SI by one sampling period Δt. The filter coefficients are denoted by c ₀ to c _n , so that according to relationship (3)

F (z ^-1 ) = c ₀ + c ₁ z ^-1 + c ₂ z ^-2 +. . . + c _n z ^-n (3)

the transfer function F (z ^-1 ) of the FIR filter can be specified for an n-order filter.

Fig. 6 schematically illustrates this filter function for an FIR filter 5th order. A broken line pulse is shown with broken lines as the input signal SI. The filter response, that is to say the output signal SO of the FIR filter, is the step-shaped signal curve which, after a number n of sampling periods corresponding to the order n of the filter, corresponds exactly to the end value of the input signal SI.

This illustrates the decisive property of the FIR filter, according to which the FIR filter has "settled" precisely on the input signal after a finite, predetermined time. Furthermore, it can be seen immediately that, by selecting the filter coefficients c ₀ to c _n accordingly, it is possible to round the filter response, ie the output signal SO, in any desired manner between the start and the end value. Using an FIR filter, it is therefore possible to specify the current profile "hard" and still achieve a "soft" setpoint specification with an adjustable rounding curve while observing the boundary conditions described.

This explains the basics for the positioning procedure here. Now, returning to the illustration of FIG. 1, the implementation of the principles described above can be explained. The block diagram shows an actual value transmitter 4 to which the sawtooth-shaped position error signal PES is fed. This signal has a period that is identical to the track grid and has a stroke that fully utilizes the available supply voltage for a single track distance. The device address 1 supplies this actual value transmitter 4 with the distance address ADR, which defines the jump distance for the upcoming lane change. Triggered by the start signal START, which is also supplied by the device controller 1 , the actual value transmitter 4 evaluates the supplied position error signal PES in such a way that it reduces the distance address ADR by the value "1" for each detected track crossing. The respectively remaining current remainder is output as the actual path value x _a for the displacement of the positioning unit 2 which is still to be covered.

A possible implementation for this actual value transmitter 4 is shown in FIG. 7. It then has a loadable counter 41 designed as a down counter, at whose data inputs DI the distance address ADR is present and is loaded when the start signal START occurs. A zero crossing detector 42 is provided for evaluating the position error signal PES. This derives a rectangular pulse sequence from the position error signal, which is supplied to the counter 41 as a clock signal. After each leading edge of this clock signal, the counter 41 thus outputs a distance address reduced by "1" via data outputs DO. This is fed to an adder 43 , at whose second inputs the position error signal PES, which is converted into a binary value via an analog-digital converter 44, is present. The actual path value x _a output by the adder 43 is therefore composed of the distance still to be covered, expressed in the number of tracks still to be crossed, and a fine adjustment value which corresponds to the current offset from the center of the track of the last crossed track. This already indicates that, in contrast to many conventional designs in the case of magnetic disk memories, the positioning method here does not switch between control systems for tracking and control systems for changing lanes.

In Fig. 1, a setpoint generator 5 is also indicated schematically, which generates nominal values x _n and i _n for the distance still to be covered or for the predetermined coil current. This setpoint generator 5 is - as a possible embodiment - constructed in hard-wired circuit technology and has a function generator 51 which, depending on the value of the distance address ADR supplied to it after the start signal START has been supplied, continuously generates the non-rounded current specification curve i (corresponding to the respective positioning process) t) generated. This unrounded current specification curve is fed to a FIR filter 52 in the form of binary values. This filter should be of a sufficiently high order, for example 20th order, in order to generate a filter response which is rounded in the desired manner, preferably cosine, and which is output in binary form as a nominal current specification i _n .

Under the conditions described above, the nominal current specification i _n linearly corresponds to a nominal acceleration, ie, the second derivative of the remaining nominal path x _{n over} time. Therefore, the nominal current specification i _{n is} fed to an integration stage 53 , which forms the double integral of the current value and outputs it as the nominal displacement specification x _n . The processes in the setpoint generator 5 are controlled by a z. B. generated by the phase locked loop 3 internal clock pulse sequence CLK.

The calculation basis on which the function generator 51 is based is explained below with reference to FIG. 8. In order to minimize the circuitry and to be able to carry out the current specification in real time during the lane change, this basis should be as simple as possible and should be valid for the full range of all possible lane changes over short and long distances. The scheme shown in FIG. 7 for an unrounded, "hard" current specification curve i (t) applies to all lane changes over longer distances with speed limitation of the positioning unit 2 . This current specification curve takes into account the described influence of the back emf. To simplify the calculation, however, the exponentially falling branches of the current curve i (t), as clearly shown in FIG. 2b, are linearized, so that the hard current specification curve is composed of straight sections.

As shown in FIG. 8, three time intervals δt1 to δt3 are distinguished. In order to clearly define the linearly specified current specification curve i (t), it is only necessary to define the basic values for the current at the beginning and at the end of these time intervals δt1 to δt3. Here i₀ is the instantaneous value for the current at the start time t0. the acceleration phase ends at time t1 with a current value i 1 that jumps to zero. At time t2, the braking phase begins with a current value -i₂, and at time t3 the braking phase is completed with a current value -i₃, which in turn jumps to zero. The linear sections of the current specification curve i (t) during the time intervals δt1 and δt3 are described by the current drop di / dt approximating the back emf as slopes m₁ and m₃.

The parameters i₀, i₃, m₁ and m₃ are each independent of the respective track distance, d. that is, they are device constants. In order to It is clear that the individual positioning processes differ only by the length of the time intervals δt1 to δt3.

A special case of this hard current specification curve i (t) shown in FIG. 8 applies to lane changes over short lane distances up to a single lane change. In these cases, the positioning unit 2 does not reach its predetermined limit speed, that is to say that the principle is that the time interval δt2 = 0.

Under these conditions, all variable parameters for all track distances can be calculated in advance, ie specified in a suitable manner. In FIG. 9, an embodiment for the function generator 51 , embodied in a hard-wired circuit, is shown as a block diagram on which this calculation basis is based. The variable parameters, but expediently also the constants, are stored in a programmable read-only memory 510 for all track distances and can be selected by the distance addresses ADR. At the start time for a positioning process, triggered by the start signal START, the current start value i₀ is loaded into a first counter 511 . This counter is designed as a down counter, which is controlled by the internal clock pulse sequence CLK. Its counter reading decreases in time with this control signal and is continuously read out to a multiplexer 512 . This multiplexer, set by the start signal START, switches this variable current value during the time interval t1 to the output of the function generator 51 and thus to the FIR filter 52 connected to it. This outputs the rounded nominal current control voltage value i _n derived therefrom via a delay element 54 , the function of which will be explained below.

In addition, a first digital comparator 513 is provided, the first inputs of the programmable read-only memory 510 of the current value i 1 is supplied and the second inputs are also connected to the outputs of the first counter 511 . This comparator continuously compares the counter reading of the first counter 511 with the final value of the current at the end of the first time interval δt1 of the unrounded current specification curve i (t). The first comparator thus determines the expiry of the first time interval δt1 when the signals supplied to it are identical. At this time, the multiplexer 512 is switched by the output signal of this first comparator 513 and at the same time a second counter 514 is loaded with a value that corresponds to the length of the second time interval δt2. According to the circuit principle described above, the outputs of this second counter 514 are connected to inputs of a second digital comparator 515 . This digital comparator is also supplied with a value "0", preferably from the programmable read-only memory 510 , so that it determines when the second time interval δt2 has expired.

At the end of the second time interval δt2, the second comparator 515 outputs an output signal which switches the multiplexer 512 and at the same time controls the loading of a third counter 516 with the braking current value -i₂ at the beginning of the third time interval δt3. Again, in the same way in the course of the third time interval δt3, the counter status of this third counter 516, which is continuously reduced by the internal clock pulse sequence CLK, is fed to the multiplexer 512 and, at the same time, inputs of a third comparator 517 . This comparator is also offered from the read-only memory 510 of the final value -i₃ for the braking current. At the end of the third time interval δt3, this comparator 517 determines the identity of the input signals and emits a corresponding output signal. This control signal STOP indicates the end of the unrounded current specification for the positioning process and is supplied to the device controller 1 and the integrator 53 shown in FIG. 1.

An embodiment for this integrator 53 is also shown in FIG. 9. Under the conditions described above for a linear, ie, saturation-free, operation, as stated, the output signal output by the FIR filter 52 represents the second derivative of the nominal path specification x _n . Since the nominal current specification i _n is present as a digital value, the integration of this value can be attributed to a continued addition. Therefore, the integrator 53 is composed of two adders 530 and 531 . The first adder is connected to the FIR filter 52 with first signal inputs. Its outputs are both connected to first inputs of the second adder 531 and also fed back to its second inputs. The second adder 531 is connected in a corresponding manner, so that both adders carry out continuous additions of the input signals supplied in each case and jointly carry out a double integration of the output signals of the FIR filter 52 . The integrator 53 thus generates the nominal path specification x _n .

In Fig. 10a to 10c are timing diagrams are shown which illustrate by way of example the effect of the FIR filter assembly 52. In FIG. 10 a, a current or acceleration profile a (t), i (t) is shown with broken lines, which corresponds to the unrounded specification of FIG. 8. The associated filtered default curve a _n (t), i _n (t) is shown with solid lines. In addition to a certain range in m _s delay lying opposite the unrounded presetting curve can be seen in particular Verrundungseffekt of the FIR filter 52nd It is particularly important that the defined final state is reached exactly after a finite delay predetermined by the order of the FIR filter 52 . In Fig. 10b and 10c are similar to the acceleration profile of FIG. 10a speed profile v (t), v _n (t) and Wegprofile x (t), x _n (t) shown unrounded filtered. The three time segments of the setpoint curve are illustrated there analogously by the times t₀, t₁, t₂ and t₃.

This describes in detail how the actual and target values are obtained for controlling the positioning process. Returning to the block diagram of FIG. 1, it is shown there that the actual travel value x _a and the nominal travel specification x _n are fed to a digital comparator 6 , which outputs _a digital signal corresponding to the deviation between the actual travel value x _a and the nominal travel specification x _n . On the basis of the control principle, this control deviation should ideally be zero during the positioning process, but in the practical case it should represent a finite small value that only takes into account faults or influencing factors of the mechanics that are not captured by the calculation of the target values.

The essential reference variable for the positioning process, on the other hand, is the nominal current specification i _n , to which the output signal of the digital comparator 6 is added in a delay stage 54 in an adder 7 . The delay element 54 compensates for transit times in the integrator 53 and in the digital comparator 6 , so that the input variables offered to the adder 7 are phase-synchronized. The output signal of this adder 7 thus represents the actual reference variable for the positioning process, which is converted into the analog control variable i _s (t) in a digital / analog converter 8 . This control variable is supplied to the drive device 201 of the positioning unit 2 as a control signal for the power amplifier 203 forming the switched current source.

A first exemplary embodiment has been described above,  that in hard-wired circuit technology the implementation of the underlying positioning procedure allowed here. In spite of all optimization of the described process steps is one such hard-wired solution always relatively expensive and also with regard to adjustments, for example the change of device constants, not very flexible. There is a way out here a programmable controller. This is what digital signal processors stand for today available for this use case offer because they are optimized for filter calculations, where multiplications and additions predominate. The A person skilled in the art is, for example, from the publications "Digital Signal Processing Applications with the TMS 320 Family ", 1986, and "First Generation TMS 320 User's Guide", 1987, both published from Texas Instruments, a comprehensive range of instruments known how to use the signal processor TMS 320 C10® useful in a wide variety of applications can use. The processor mentioned has only a summary here explains a cycle time of 200 ns at a Word length of 16 bits, with the accumulator being 32 bits wide and by the multiplier with a thickness of 16 × 16 A 32-bit result is returned. Referring to the second of the above publications is used here for illustration the performance of this signal processor is exemplary referred to a command sequence MPY and LTD, which are there is described on pages 4/49 and 4/41. This command sequence causes only two clock cycles in total, i.e. H. so, in 400 ns, 16 × 16 bit multiplication, 32 bit addition to the accumulator, loading a RAM cell into the multiplier and shifting the contents of this RAM cell into the next higher address.

In FIG. 10, as a further embodiment, an intelligent positioning control unit 9 is shown schematically in a block diagram, which is connected to the device controller 1 . In addition to a digital signal processor 91 of the type described above, an 8-bit wide analog / digital converter 92 and a 10-bit wide digital / analog converter 93 are provided in this positioning control unit. The above-mentioned units are connected to one another via an interface 94 which, as in many conventional microprocessor applications, has simple logic circuits which enable the signal processor 91 to exchange data with the connected units.

The analog / digital converter 92 and the digital-analog converter 93 can be implemented by an AD7569 module from Analog Devices®, which includes an 8-bit analog / digital converter and two 8-bit digital / analog converters contains corresponding input / output and selection logic. This module can be used such that its two digital / analog converters together form the digital / analog converter 93 . For this purpose, they are switched so that one only outputs the signal corresponding to the current specification curve i _n during a positioning process and the other outputs the actual control signal for tracking. The gain of the first digital / analog converter is four times higher than that of the second. The result is that the resolution for the nominal current command i _n corresponds to a 10-bit value. Because of the availability of a corresponding fast module, this division is more cost-effective and possibly also more expedient in view of the real-time conditions than a correspondingly broad individual digital / analog converter, as is indicated schematically in FIG. 10.

The design of the positioning control unit 9 is based on the principles which have already been explained in detail in connection with the first exemplary embodiment. An essential difference in the implementation here is, however, that the default values, made possible by the performance of the digital signal processor 91 , are determined in real time during a positioning process, which is particularly advantageous with regard to a possible adaptability in this embodiment for carrying out the positioning method . Since the calculation methods for determining the time intervals t1 to t3 are optimized with regard to the properties of the digital signal processor 91 used , they are explained below.

Here, too, it is assumed that the following basic conditions apply to each positioning process. The final speed v (t₃) at the end of the positioning process must be zero, the positioning distance, ie the distance x (t₃) covered at time t3, corresponds to the nominal path specification x _n , and the following relationship applies to the limit speed during the positioning process (4) :

The validity of the following relationship (5) can also be derived from FIG. 8 also for positioning processes without speed limitation in which the second time interval δt₂ is zero per se:

2a₀δt₁ + m₁δt₁² = -2a₃δt₃ + m₃δt₃² (5)

This relationship (5) says nothing more than that, because of the energy theorem, the areas of the two parts of the current or acceleration specification curve according to FIG. 8 must be the same. Finally, the following relationship applies to the distance over the full track distance (6):

(½) a₀δt₁² + (¹ / ₆) m₁δt₁³- (½) a₃δt² + (¹ / ₆) m₃δt₃³ = x _n (6)

The relationships (5) and (6) form an equation system with two unknowns, namely δt₁ and δt₃ depending on the distance x _n for the full track distance. However, these relationships cannot be explicitly resolved, since no analytical functions δt₁ = f₁ (x _n ) or δt₃ = f₃ (x _n ) can be established.

This problem is supported by an estimate of the length of the first Time interval δt₁ depending on the particular Track distance solved. For this estimate δt₁ 'is only provided that the condition δt₁ '<δt₁ is met. Then  can be an estimate δt₃ 'based on the following relationship (7), which is derived from relationship (5):

With these estimates, the relationship (6) results in an estimated positioning range xs', which must be smaller than the full positioning range xs due to the requirements δt₁ '<δt₁ and δt₃'<δt₃. The difference between the actual positioning distance xs and the estimated positioning distance xs' is compensated for by introducing the time interval δt₂. During this time interval, the positioning unit 2 is constantly moved at the speed v 1 at the end of the time period δt 1 '. With reference to relationship (5), this speed v 1 is calculated according to the following relationship (8):

v₁ = a₀δt₁ ′ + (½) m₁δt₁′² (8)

According to the following relationship (9), the length of the Determine time interval δt₂:

δt₂ = (xs-xs ′) / v₁ (9)

The derivation described above initially has the advantage that the estimate for the time interval δt₁ 'a simple Underlying relationship, as opposed to an exact calculation much less programming effort, computing time and requires storage space. Since the time interval δt₁ ′ is estimated anyway, it can always be considered an integer Value of sampling periods Δt can be selected, which makes correction calculations the default curve in the transition area between the Time intervals δt₁ and δt₂ can be avoided. A very important one Another advantage is that regardless of the track distance a time interval with each positioning process δt₂ is introduced or, in other words, not between positioning operations distinguished over short or long track distances must become.  

Provided that saturation-free operation is required, a speed limit for the positioning unit 2 must of course be specified for longer track distances, but this results automatically by limiting the first time interval δt 1. The maximum length of this time interval results from the following relationship (10):

the result is positive, since m₁ represents a negative quantity. This maximum value for the first time interval is δt₁ an operating constant that occurs once when restarting of the magnetic disk storage is calculated in a startup routine becomes.

The explanation of the first exemplary embodiment has already been made described that one for the determination of the default values for current and path because of the assumed linearity of the Transfer functions only one of the two default curves not rounded and then filtered for rounding, where it doesn't matter whether you look at the current or acceleration values or the values for the way. Exemplary was assumed in the first embodiment of the current specification and gained the path from integration. In the second exemplary embodiment, a path specification is used as a basis, from which, after filtering, the rounded nominal Current specification determined by double differentiation becomes.

As explained, the unrounded specification is to be reshaped by filtering in such a way that it specifies the current for exciting the magnetic drive coil of the positioning unit 2 with as little harmonics as possible. This goal can only be adequately achieved with a higher order FIR filter. This is based on an FIR filter, for example of the 20th order, which is implemented by a filter program of the signal processor 91 . Because of the high order of the FIR filter, a corresponding amount of computing time arises, which is to be minimized as far as possible, since the determination of the default values for current and distance takes place in a time-critical area.

In order to relieve the digital signal processor 91 , an extrapolation routine for the route is therefore initially used for the calculation of the unfiltered route specification, which extracts a current route setpoint x _k for the unrounded route specification from the three route setpoints of the previous sampling periods Δt according to the following relationship (11) :

x _k = 3x _k-1 -3x _k-2 + x _k-3 + m _i Δt³ (11)

i is one of the values 1, 2 or 3, which refer to one of the time intervals δt₁, δt₂ or δt₃. The last term from relationship (11), namely m _i Δt³, is constant within these time intervals. The relationship (11) thus corresponds formally to a filter operation which can be processed by the signal processor 91 in a time- optimized manner.

In this extrapolation routine it should be noted that when moving from a time period, e.g. B. δt₁, on the next period, e.g. B. δt₂, there are no corresponding previous travel setpoints per se. For these transitions in the non-rounded path specification curve, these three previous path setpoints must therefore be simulated, ie calculated before the actual execution of the positioning process and made available in a memory. It must also be clarified that the path information, for example a current setpoint value x _k , is an absolute information which relates to the width of the entire data band on the magnetic storage disks. For reasons of accuracy, a data format of 32 bit positions is therefore expedient for the x variables in a 5¼ ″ magnetic disk memory, this format corresponds to two data words in a 16 bit signal processor. This extrapolation from three previous travel setpoints could also be avoided. Instead of this, one could determine the path setpoint x _{k of} the current sampling period from the last path setpoint and its two derivatives, as will be shown, as can easily be seen from the previous considerations.

After the current path setpoints x _{k have been} continuously calculated, the corresponding rounded quantity is to be determined by filtering with the aid of the FIR filter. It has already been indicated above that this filter should be of the 20th order. Referring to Fig. 5 and 6 has already been explained that the filter order results from the ratio of smoothing time to the sampling period. If the sampling period is 60 ms and a filter of the 20th order is used, the result is a smoothing time of 1.2 ms, which is still acceptable in the present application.

If, however, the path information is in a 32-bit format, the selected high-performance digital signal processor 91 would still result in an excessively long computing time despite the algorithms described. One solution to this problem is not to filter the absolute value for the current travel setpoint x _k , but only to use its travel increment δx _k to the previous travel setpoint x _k-1 . This path increment can be represented with sufficient accuracy in a 16-bit format. In other words, one forms the path increment δx _k before filtering and then filters it. The filtered path increment obtained is added to the previous absolute value of the nominal path specification x _nk-1 and thus the absolute size for the filtered path specification, ie the nominal path specification x _nk for the current sampling period, is obtained.

In the present exemplary embodiment, the rounded current specification i _{n is} derived from the nominal path specification x _n by double differentiation. In the present case, this can be reduced to the formation of the second difference according to the following relationship (12):

i _nk = (x _nk -2x _nk-1 + x _nk-2 ) / Δt² (12)

However, since, as explained above, the path increments δx _k are already used as the basis for filtering the path specification, this differentiation is further simplified, as is expressed in the following relationship (12a):

i _nk = (δx _nk -δx _nk-1 ) / Δt² (12a)

This relationship obviously illustrates how easily the nominal current specification i _n can be obtained. This applies in particular if the sampling period Δt is nominally set to "1". This also eliminates the division according to relationship (12) or (12a), and the nominal current specification i _n results from a simple subtraction of the corresponding path increments.

The above calculation basis is based on the assumption of ideal mechanics. However, the losses that occur in practice with a positioning unit 2 can be taken into account if necessary if the relationship (12a) is expanded to the following relationship (12b):

i _nk = (δx _nk -δx _nk-1 ) + α _₀ + α ₁ · x _nk + α ₂ · δx _nk (12b)

The expansion term α₀ corresponds to a constant acceleration, which compensates for an unbalance of the positioning unit 2, for example. The second extension term α ₁ · x _nk corresponds to a path-proportional acceleration that takes into account a force that acts on the positioning unit 2 in a path-proportional manner and, for example, compensates for a restoring force caused by connecting lines. The third expansion term α ₂ · δx _nk corresponds to an acceleration proportional to the speed and thus takes into account, for example, friction in the positioning unit 2 . This shows that with the help of the relationship (12b) it is also possible to control a positioning unit 2 which deviates to a considerable extent from the quantities on which an ideal mechanics is based.

The above-described calculations to be carried out by the digital signal processor 91 are illustrated in FIG. 11 for illustration and in summary form. Block 110 shows an equivalent circuit diagram for the path extrapolation described. Line 111 shows basic values x _k-1 , x _k-2 and x _k-3 or m _i Δt³, which are loaded into the register of signal processor 91 at the beginning of a new time interval δt₁, δt₂ or δt₃ Initialize path extrapolation appropriately. As a variant, the current path setpoint x _k can already be loaded in a register 112 and the simulated previous path setpoints x _k-1 and x _k-2 in corresponding registers. This saves the calculation of the first extrapolation value, ie the current travel setpoint x _k , and thus compensates for the time required to load the basic values. The three basic parameters for one of the time intervals δt _i result from the basic parameters a₀, v₀ and x₀ of this time interval according to the following relationships (13), (14) and (15):

x _k = x ₀ (13)

x _k-1 = x ₀ -v ₀ Δt + (½) a ₀ Δt²- (¹ / ₆) m _i Δt³ (14)

x _k-2 = x ₀ -2v ₀ Δt + 2a ₀ Δt²- (⁸ / ₆) m _i Δt³ (15)

wherein relationships (14) and (15) are derived from relationship (6). These relationships also make it clear that - as indicated as a variant above - the current travel setpoint x _{k could also be determined} from the previous travel setpoint x _k-1 and its two derivatives.

As illustrated in FIG. 11 with dash-dotted lines, the content of the register 112 for the path setpoint x _{k is} continuously loaded into a register 113 as the calculation proceeds, which register thus contains the old path setpoint x _{k-1 of} the previous sampling period Δt. In addition, with the content of the register 112 with progressive path extrapolation, the contents of the registers for the basic values are successively updated.

As stated above, the first time interval δt₁ is always an integer multiple of the sampling period. So that the transition from the sampling interval δt₁ to the sampling interval δt₂ is always in sync with the scanning grid. This is different in the transition from the time interval δt₂ to the time interval δt₃, because - as stated above - the second time interval δt₂ is not an integer multiple of a sampling period Δt. This must be taken into account when calculating the basic values for the start of the time interval δt₃. If the period of time between the start of the third time interval δt₃ or the next sampling period Δt is designated as Δt _f, the following corrected values result from the relationships (16), (17) for this first sampling period Δt occurring after the beginning of the third time interval or (18):

a _e3 = a ₃ + m ₃ Δt _f (16)

v _e3 = v ₁ + a ₃ Δt _f + (½) m ₃ Δt _f ² (17)

x _e3 = x ₁ + v ₁ (δt ₂ + Δt _f ) + (½) a ₃ Δt _f ² + (¹ / ₆) m ₃ Δt _f ³ (18)

This defines all boundary conditions for path extrapolation, which is illustrated in FIG. 11 by the equivalent circuit diagram 110 .

The equivalent circuit diagram 114 reproduces the described path increment formation, in which the path setpoints x _k and x _{k-1 of} the current or the previous sampling period are subtracted from one another and give the path increment δx _k loaded in a register 115 .

Block 116 represents an equivalent circuit diagram for the FIR filter, in which path increments x _k-1 , evaluated with corresponding filter coefficients c _i , are summed up. Here i means an integer value from 0 to n, where n indicates the order of the FIR filter.

The result of this filtering is a value loaded in a register 117 for the rounded path increment δx _nk , which is stored progressively after each sampling period in a further register 118 as a path increment of the previous sampling period. The block 119 thus designates the equivalent circuit diagram for the described two-fold differentiation of the path increments, so that the current value i _nk for the nominal current specification i _n results from this. The value temporarily stored in the register 117 for the rounded path increment δx _nk is - as indicated by a delay element 121 - added to the content of a register 120 with a delay, which contains the absolute value for the path specification x _{nk-1 of} the previous sampling period. The current absolute value x _nk for the nominal path specification x _n is thus continuously output in synchronism with the current value for the current specification i _n .

FIG. 12 shows a flow chart for the entire positioning routine. After the start of the positioning process, the values for the time intervals δt₁, δt₂ and δt₃ are first calculated in method step 1201 . With these values obtained in this way, the basic parameters for each time interval, as specified in method step 1202 , are determined before the actual positioning process is carried out. By zeroing the values for the time interval δt and the time t in step 1203 , the positioning control unit 9 and thus its signal processor 91 are set to a defined initial state at the beginning of the actual positioning process. Thereafter, the value for the current time interval is incremented by "1" to initiate the first time interval δt 1 according to method step 1204 . After the initiation of this new time interval, the basic parameters assigned to it are first loaded in accordance with method step 1205 .

In the course of the first time interval δt 1, the path increments δx _{k are then} calculated in accordance with method step 1206 and filtered in accordance with method step 1207 . Generally speaking, the current path increment output at the output of the FIR filter is added to the absolute size of the previous path setpoint, and thus the current absolute size x _{nk of} the nominal path specification x _{n is} specified in method step 1208 .

In step 1209 , this is followed by the calculation of the corresponding current variable i _nk for the nominal current specification i _n . These absolute values x _nk thus calculated for the path specification and i _nk for the current specification are output, that is to say they are supplied to the comparator 6 according to FIG. 1 or to the digital-to-analog converter, as indicated in method step 1210 .

The time sequence in method step 1211 is then incremented by the value “1” and, according to branch 1212, the condition is queried as to whether t is already greater than the calculated total time for the positioning process. If this is the case, the positioning process is finished. Otherwise, a further condition is queried in method step 1213 as to whether the current time interval δt₁, δt₂ or δt₃ has expired. If this is not the case, the current setpoint value x _k for the applicable time period is calculated in accordance with method step 1214 and the program sequence continues with the calculation of the next path increment δx _{k in} accordance with method step 1206 . However, if the branching condition in method step 1213 is fulfilled, ie the current time interval has been completed, then the program flow continues according to method step 1204 with the incrementing of the time interval.

Finally, it should be pointed out that for the positioning control unit 9 there is no difference between a positioning process with a setting for a new track and a control process for tracking a selected track. There are therefore no switching processes between a control system for tracking and a control system for positioning on a new track.

During the control processes for tracking a selected track, two variables, namely the actual travel value x _a and the nominal travel specification x _{n, are managed} as absolute values in the memory of the signal processor 91 . During the control process, only the difference between these two variables is evaluated, each of which has a 32-bit format. With these two data words, the high-order data word contains the information about the track number in whole numbers, while the low-order data word contains the deviation from the track center in the form of fractions of a track. With this format, the entire width of the data tape on the magnetic storage disks can be represented linearly with high accuracy.

Any nominal track and additionally a selectable deviation from this track can thus be specified via the nominal path specification x _n . As was explained on the basis of the description of FIG. 1, the actual path value x _{a is} continuously updated via the position error signals PES which are read in, track crossings being automatically taken into account. The next value to be expected is extrapolated from the course of this actual path value in the past and the corresponding edge of the position error signal PES is selected. As is known, in the case of magnetic disk memories, adjacent information tracks are distinguished by the track type which is defined in the servo cells containing the track information on the servo surface of the magnetic disk memory. In connection with this differentiation, the described method can ensure the correct track selection even if, at the limit speed of the positioning unit 2 of, for example, more than 1.5 m / s, up to six tracks are "blindly" flown over in a sampling period Δt. Due to the separate management of the actual travel value x _a and the nominal travel specification x _n , any travel errors are permitted in principle, this applies both to the positioning process and when settling onto a selected track. This principle also allows, if expedient, positioning operations over short track distances to be brought about solely by the sudden change in the nominal path specification x _n , although a corresponding settling time must be expected.

In the above description, there are implementation options for the positioning process and a device for Execution of the method in use with a magnetic disk storage described. It would be beyond the scope of the present Exceed description, this method now also in view to the multitude of other possible applications in to explain each. From the detailed description above The person skilled in the art receives one from exemplary embodiments sufficient teaching that enables him to apply of the principles described embodiments of the described Adapt the procedure to other applications. A summary overview of such use cases was already given in the introduction to the description, so that there is no need for repetition here. From the detailed above Description of embodiments can, however Those skilled in the art will recognize that the method described also applies to such Use cases offers special advantages. Positioning operations proceed using the described method quiet and low excitation, allow quick positioning with high accuracy for reaching and holding the target position. With the electrical or electromotive used Drives occur due to the properties of the described Process low electrical losses. Since one the method is preferably processor-controlled the controller can be flexibly adapted to different applications, and finally it is another characteristic Characteristic that one using the described method performed control of a positioning unit too is long-term stable. A majority of the properties mentioned here is often used in such controls for positioning units required, the procedure described is sufficient for all of these Conditions.

Finally, it should be noted that the detailed description above of embodiments relate to the controller a positioning unit in only one coordinate direction along a linear path. But this means  no restriction, because the specialist prepares it on the basis of above description no difficulty, the described Procedure also on positioning processes in a two-dimensional or three-dimensional coordinate system to extend. At The positioning unit is an X / Y slide, for example when positioning between two points of a surface emotional. The positioning path thus corresponds to a location vector in this two-dimensional coordinate system. This location vector can, as is obvious to the expert, in his components each pointing in one of the coordinate directions disassemble that clearly define him. To adapt the described One has to proceed with this application assume that the positioning unit for movements in the two Coordinate directions each have a drive member. To control this drive element according to the described The method is derived from the components derived from the location vector each assigned an individual setpoint curve and thus in an individual for each coordinate direction Control voltage derived, which drives the positioning unit as a control variable for one of the coordinate directions is fed. In extension of this principle this also applies analogously to control processes in which the positioning unit spatially, d. that is, based on a three-dimensional coordinate system, is shifted like this for example manipulators and others, introductory mentioned use cases is the case.

Claims (15)

1. Method for controlling a positioning unit ( 2 ) of an automatic machine, for example a machine tool, from a starting position to a target position by continuously comparing the current position (xa) of the positioning unit as the actual value with the target position (xn) as the target value using one of the distance between the start and target position-dependent setpoint curve for the current supplied to a drive ( 201 ) of the positioning unit, characterized in that at least one setpoint curve designed as a step function is used, according to which the positioning unit with a maximum nominal current specification (i _n ) at the beginning of the positioning process if necessary, accelerated until a predetermined limit speed (v _1im ) is reached or, at the end of the positioning process, that the sample values corresponding to at least one setpoint curve with a predetermined sampling period (Δt) are continuously provided by an FIR filter arrangement (e.g. . 52 ) are supplied, whose filter coefficients (c ₀ . . . c _n ) are selected such that the kinks in the setpoint curve are rounded and that a control voltage [i _s (t)] is derived from these sampled values (i _nk or x _nk ), which drives the drive ( 201 ) of the positioning unit ( 2 ) is supplied as a control variable. 2. Positioning method according to claim 1, characterized in that a positioning process is divided into three time intervals (δt₁ to δt₃) regardless of the respective positioning distance, that the drive ( 201 ) of the positioning unit ( 2 ) current supplied in the first time interval (δt₁) such is controlled that it takes linear with consideration of the back emf with rounded transitions from an initial maximum to about the end of this time interval and then drops steeply, that this current in the second time interval (δt₂) is essentially zero and that this current at the beginning the third time interval (δt₃) using the back emf of the drive ( 201 ) of the positioning unit ( 2 ) with a steep increase quickly reaches an excessive negative extreme value, from which it initially decreases linearly and then steeply towards the end of the third time interval. 3. Positioning method according to claim 2, characterized in that the length of the first time interval (δt₁) for a positioning process is always chosen such that this length corresponds to an integer multiple of the sampling period (Δt), being shorter than the corresponding length at an absolute The setpoint curve, which is designed to be time-optimal and does not exceed a predetermined maximum value ( _{δt 1max} ), even with large positioning _distances , which corresponds to a limit speed (v _1im ) of the positioning unit ( 2 ) at which it is ensured that the drive ( 201 ) of the positioning unit still works in a saturation-free range . 4. Positioning method according to claim 3, characterized in that for the nominal path and current specification (x _n or i _n ) only one of these variables is explicitly derived from a common setpoint curve and then filtered and that in the case of a nominal current specification derived in this way (i _n ) the nominal route (x _n ) is obtained by double integration. 5. Positioning method according to one of claims 1 to 4, characterized in that the samples (x _k ) in the filter arrangement (z. B. 52 ) are rounded according to a cos function. 6. Positioning method according to one of claims 1 to 5, characterized in that the actual value for the current position of the positioning unit ( 2 ) as a distance already traveled for the respective positioning distance (x _a ) from position error signals (PES) in the form of the number of during of the positioning process already covered distance increments of the positioning distance, that is, units of a predetermined scale, as well as a current storage of the end value of the last determined path increment and compared with the nominal path specification (x _n ) as a setpoint that the deviation of the nominal path specification (x _n ) superimposed on the actual value (x _a ) of the nominal current specification (i _n ) and from this the control signal [i _s (t)] supplied to the positioning unit for setting the current amplitude for the drive ( 201 ) of the positioning unit is derived, and that the nominal Current specification (i _n ) essentially the control process via d The positioning distance and the deviation of the nominal current specification (x _n ) from the actual value (x _a ) determines the position control for holding the predetermined position. 7. Positioning method according to one of claims 4 to 6 using a with a digital processor, optionally a signal processor ( 91 ) equipped positioning control unit ( 9 ), characterized in that first the calculation of the sample values (x _k ) of the setpoint curve, then the filtering thereof Sampling values and finally the determination of the nominal path and current specification (x _n or i _n ) during the positioning process under real-time conditions with the aid of the processor ( 91 ). 8. Positioning method according to claim 7, characterized in that, before the start of each positioning process, the length of the first time interval (δt₁) is derived from the respective existing positioning distance (x _s ), taking into account the opposite in the first time interval and in the third time interval (δt₃) supporting effective influence of the back emf of the drive ( 201 ) of the positioning unit ( 2 ) determines the length of this third time interval and finally the length of the second time interval (δt₂) is determined such that it is the remaining difference between the full positioning distance and the sum of the compensated distances of the positioning unit to be covered in the first and third time interval. 9. Positioning method according to one of claims 7 or 8, characterized in that within each of the time intervals (δt₁, δt₂ or δt₃) for each sampling period (Δt) a new sample (x _k ) of the setpoint curve by extrapolation from samples (z. B. x _k-1 , x _k-2 , x _k-3 ) previous sampling periods is determined and that at the beginning of each one corresponding to the time intervals, geometrically derived from the setpoint curve, initial values are given as parameters. 10. Positioning method according to one of claims 6 to 9, in which for filtering the samples of the setpoint curve according to the order (n) of the selected FIR filter arrangement (n + 1) successive signal values (eg x _kn... X _k ) are each evaluated and added up with a filter coefficient (c ₀ ... c _n ) predefined in accordance with the transfer function of the FIR filter, characterized in that for the filtering of the sampled values (x _k ), instead of their absolute values, only their increments (e.g. x _k -x _k-1 ) to the previous sample value (x _k-1 ) and that the filtered absolute size (x _nk ) from the corresponding filtered increment (δx _nk ) with addition to the absolute size (x _nk-1 ) of the previous sampling period (Δt) is obtained. 11. The positioning method as claimed in claim 10, in which the filtered sample values (x _nk ) derived from the setpoint curve represent the nominal path specification (x _n ), characterized. that to derive the corresponding nominal current specification (i _n ), the difference is formed from the filtered increments (δx _nk , δx _nk-1 ) of the samples of the current or the previous sampling period (Δt). 12. Positioning method according to one of claims 1 to 11, at which the distance between the start position and the destination position as Location vector in a given, two- to three-dimensional Coordinate system is present, characterized in that that this location vector at the beginning of the positioning process  broken down into its components, each one an individual setpoint curve is assigned and that individual control voltages are derived for each coordinate direction be the drive of the positioning unit as Control variables for movements in one of the coordinate directions are fed. 13. Arrangement for performing the positioning method according to one of claims 1 to 12, characterized by an actual value transmitter ( 4 ) for deriving digital actual values (x _a ) from the sequence of analog positioning error signals (PES) output by the positioning unit ( 2 ), by a Setpoint generator (e.g. 5 ) to which an address (ADR) of a target position or a quantity corresponding to the positioning distance is fed for a positioning process and which is constructed in such a way that it continuously generates a digitized value for the nominal path and current specification (x _n or i _n ) is derived by a digital comparator (e.g. 6 ), to which the actual value (x _a ) given by the actual value transmitter and the value of the nominal path specification (x _n ) generated by the setpoint generator are fed as input variables and outputs as control deviation a value corresponding to the difference between these input variables, by means of an adder (for example 7 ) connected to the digital comparator, the also the digitized value for the nominal current specification (i _n ) is supplied and by a digital / analog converter (z. B. 8 ), which is connected to the output of the adder and outputs the control signal [i _s (t)] for the drive ( 201 ) of the positioning unit ( 2 ). 14. Arrangement according to claim 13, characterized by an actual value transmitter ( 4 ) with a zero crossing detector ( 42 ) and an analog / digital converter ( 44 ), to which the position error signals (PES) are supplied in the form of a sawtooth pulse sequence, the pulse period of which is the geometric dimension of a Wegeinrementes corresponds, further with a down counter ( 41 ), which can be loaded at the beginning of a positioning process with a digitized value corresponding to the respective positioning distance (ADR) and is triggered by output signals of the zero-crossing detector, and with an adder ( 43 ), which is connected to both the outputs of the Down counter and that of the analog / digital converter is connected and outputs the digitized actual value (x _a ) as the sum of its input variables. 15. Arrangement according to claim 13 or 14, characterized by a setpoint generator ( 5 ) with a memory ( 51 ) in which, for each positioning distance, a set of the basic values defining the course of the associated setpoint curve composed of linearly running sections (e.g. i₀, δt₂) and can be selected by an address (ADR) that corresponds to the respective positioning distance and is provided with switching networks (e.g. 511, 512, 513 ) for generating consecutive, possibly linearly decremented samples in a predetermined time pattern Dependency on the stored corner values of the setpoint curve and read out into the switching networks, the setpoint generator also containing the FIR filter arrangement ( 52 ) to which these generated sample values (x _k ) are supplied.