Sliding Mask Variable Resolution Velocity Trajectory for Track Following Servo

This invention relates to servo systems in which a velocity command is generated which is a measure of the desired velocity of the load controlled by the servo system.

Many current magnetic recording disk files imploy servo systems for controlling the movement of one or more magnetic heads to different concentric recording tracks on the magnetic recording disk or disks (track seeking) and for maintaining the head or heads accurately positioned over the center of the selected track or tracks during read or write operations (track following). At the start of a track seeking operation, most of such systems generate a velocity command or curve which is a measure of the desired velocity of the heads in moving from one track to the selected new track location. As would be expeted, this velocity command is a function of the distance to be moved by the head or heads, the desired velocity being higher the longer the distance to be moved. Prior art disk files usually generated different velocity curves or trajectories for different seek lengths using a number of different techniques.

However, velocity trajectory implementation has always presented practical difficulties to the servo designer. Some disk files use a multiplier technique, while others use a diode breakpoint network. Both the multiplier and the diode network have many deficiences. The diode breakpoint causes large disturbances or "bumps" in the current waveform. The multiplier is not very flexible in terms of curve shape, suffers dc drift, and it suffers in the transition to the so-called linear region.

The use of a digital trajectory curve solves virtually all of the above problems associated with the diode breakpoint and multiplier techniques, but it has several problems of its own. The digital curve requires an unreasonable volume of points to be stored in a look-up table to get a reasonable resolution of position vs. velocity.

The present invention provides a digital servo system employing a sliding mask technique to solve both the storage volume and the resolution deficiencies of the standard digital curve.

FIGS. 1A, 1B and 1C are diagrams showing different prior art implementations of velocity curve generation;

FIG. 2 shows circuitry for implementing the sliding mask technique of the present invention;

FIG. 3 is a graph of an input/output relationship produced in accordance with the present invention;

FIG. 4 is a flow chart illustrating the operation of the present invention;

FIG. 5 is a graph of an input/output relationship generated in accordance with the present invention with a lower resolution than the curve of FIG. 3;

FIG. 6 is a velocity trajectory for a current magnetic disk file;

FIG. 7 is a modification of the trajectory curve of FIG. 6 using the sliding mask technique of the present invention;

FIG. 8 contains graphs of variations in the current, velocity and PES during a seek operation in a current disk file whose velocity trajectory is shown in FIG. 6; and

FIG. 9 shows the improvement in the current graph of FIG. 8 obtained by use of the present invention.

FIGS. 1A, B, C show block diagrams of the aforementioned prior art techniques.

FIG. 2 shows how the sliding mask 31 of the present invention is inserted into a digital curve implementation. The key concept of the present sliding mask technique is that high resolution in position is not necessary at large values of velocity command. The resolution must be better and better as the servo approaches the target.

As seen in FIG. 2, one of the inputs to the sliding mask is the difference count, which is a count of the difference between the present track location of the head and the location of the track to which the head is to be moved. This difference count is decremented with each track crossing by the head and reaches zero when the head reaches the selected track, as is well known in the art. The other input to the sliding mask 31 is the digitized position error signal (PES) from the digital servo system. This PES signal is generated from signals produced in the disk file servo read head during track crossings. These inputs are operated on to form an output address which is the input to a look-up table in the form of a programmable read only memory (PROM) 32. The output from PROM 32 is a group of digitized values for the velocity curve. If a digital servo is employed, these digitized values can be utilized directly as the digital velocity command. If an analog servo system is used, the digital values can be converted to analog values in a digital-to-analog converter (DAC) 33.

The sliding mask technique involves dividing the length of the desired velocity curve into a number of segments from the beginning of the curve, each segment being one half the size of the preceding segment. For each segment there is stored an equal number of values for the velocity curve for that segment and these velocity values are used to generate the velocity curve during each segment. This has the effect of reducing the resolution of the curve in the early or high velocity segments of the curve and providing much higher resolution in the critical lower portion of the curve as the head approaches the on track position. This is so because in the lower portions of the curve each segment is much smaller that the early segments, but each segment has the same number of velocity curve values available. This greatly reduces the memory size required to store the velocity values while giving totally acceptable or improved system performance.

The operation of the present invention can best be understood by reference to the curve of FIG. 3 and the flow chart of FIG. 4. The curve of FIG. 3 shows the generation of the addresses in PROM 32 by which velocity values stored in PROM 32 can be retrieved. As indicated, the curve is divided into segments, the first such segment being segment 0, with each succeeding segment being one half the length of the preceding segment. Referring to FIG. 4, this shows in flow chart form the operation of the present invention for a mask length M.sub.1 of 5 bits, the number of difference count bits DC.sub.n equal to 8 bits, and the number of digitized PES bits P.sub.n equal to 5.

If the difference count is zero, the sliding mask technique of this invention is not used, since the servo system is in its linear range. If the difference count is not zero, the input address for sliding mask is formed as represented by block 41. The format of the input address is shown to the right of box 41 and includes the eight difference count bits DC.sub.0 -DC.sub.7 and the five digitized PES bits P.sub.0 -P.sub.4. The next step is to initialize the segment number, as represented by block 42. As indicated to the right of box 42, the segment number is encoded in three bits NS.sub.0 -NS.sub.2, and at the start of a seek, it would be initialized to segment 0, as is apparent from FIG. 3.

Next, the 13 bit register R is loaded with the 13 bit input address, as represented in block 43. The value of the input address in register R is checked to determine if the value of the most significant bit therein, R.sub.12, is 1. (block 44) If it is (as it would be at the start of a long seek, for example) the output address is formed (block 46) using the segment number N.sub.s and the five bits R.sub.11 -R.sub.7 to the right of bit R.sub.12 in register R. This output address is then used (block 47) to address locations in PROM 32 where the required velocity values are stored. Using a mask M.sub.1 of 5 bits permits the addressing of 32 PROM locations to retrieve 32 velocity curve values for each segment. This is shown in FIG. 3 with the 32 staircase values for the sliding mask output in segment 0, and the output velocities 32 staircase values in FIG. 7.

The sliding mask aspect of the invention accurs when R.sub.12 in register R no longer has a value of 1, indicating that the servo system has moved the head past the inner boundary of segment 0. When this occurs (block 48), the segment number N.sub.s is incremented by one and register R is shifted one bit to the left so that bit R.sub.11 therein is now the most significant bit in the register. This produces a "yes" output from block 44 so that the output address is now formed from segment number S, and the five bits to the right of bit R.sub.11 in register R. This permits the addressing of 32 locations in PROM 32 to access 32 values for segment 1 of the velocity curve. Since segment 1 is only half as long as segment 0, it can be seen from FIG. 3 and FIG. 7 that this results in an increase in the resolution of the segment 1 portion of the curve relative to that of segment 0.

The above process continues through the eight segments which are encoded in the three segment under bits NS.sub.0 -NS.sub.2, resulting in the retrieval and use of 32 velocity curve values for each segment and producing a velocity curve of increasing resolution as the head approaches the on track position, as is apparent from FIG. 7.

The improved resolution obtainable using a mask length M.sub.1 of 5 bits as compared to that resulting from a mask length of 3 bits (and hence 8 velocity curve values per segment) is apparent from a comparison of FIGS. 3 and 5.

Thus, the output is formed by sliding a mask M.sub.1 bits long to the right as the distance to the target gets smaller. The resolution in position (R.sub.x) then varies with position.

N.sub.s =Segment Number

M.sub.1 =Mask Length in Bits ##EQU1## where, x.sub.max =maximum distance for which curve is defined.

An important point is that in almost any velocity trajectory technique, when the linear region is met (usually X is less that TW/2), the velocity command comes directly from the PES, or input address, and the curve or look up table is bypassed.

The storage size of the look-up table required for the present invention is:

where

N.sub.sm =PROM storage size required (words)

N.sub.dc =number of bits in difference count

M.sub.1 =mask length in bits

For the standard digital curve technique, the required storage size is:

where

N.sub.sc =standard digital curve look up table size

N.sub.pes =number of bits in digitized PES.

For a comparison of storage size requirements,

assume N.sub.dc =8, N.sub.ps =5, M.sub.1 =5

For the present technique,

N.sub.sm =256 words

For the standard digital method,

N.sub.sc =8096 words

So the sliding mask technique of the present invention requires 32 times less storage for the same resolution at low velocities. (With given assumptions)

FIG. 6 shows the diode breakpoint velocity trajectory used in a current disk file. FIG. 7 shows the same curve implemented with the sliding mask technique of this invention. The absence of diode breakpoints can be clearly seen in FIG. 7.

FIG. 8 shows basic seek waveforms for the current disk file whose velocity trajectory is shown in FIG. 6. For exactly the same servo, with a sliding mask digital curve, the same basic seek waveforms appear in FIG. 9. The improved current waveform for the sliding mask can be noted. Also, it has been found that the sliding mask method produces increased speed in total seek time.