Method and apparatus for correcting for systematic errors in timing pattern generation Number:7,136,243 from the United States Patent and Trademark Office (PTO) owispatent Improvements in placement of timing patterns in self-servowriting include correcting for systematic errors due to geometric effects. A correction is made for varying systematic errors, such as when the recording head has spatially separate read and write elements. Further, servopattern rotation due to residual or unmeasured systematic errors is reduced by using a once per revolution clock index derived from the motor drive current waveform or any other sensor. In one aspect of correcting for systematic errors in the writing of timing patterns on a storage medium of a storage device, a time interval between a trigger pattern written at a first radial position of the storage medium and a rotational index is measured. The rotational index is related to the rotational orientation of the storage medium with respect to a fixed frame of the storage device. The location of another trigger pattern to be written is shifted, using the measured time interval to determine the shift in location for the another trigger pattern.<H generation, systematic, Method, and, appa, 7136243.html, Patent, pto, 7136243

Title: Method and apparatus for correcting for systematic errors in timing pattern generation

Abstract: Improvements in placement of timing patterns in self-servowriting include correcting for systematic errors due to geometric effects. A correction is made for varying systematic errors, such as when the recording head has spatially separate read and write elements. Further, servopattern rotation due to residual or unmeasured systematic errors is reduced by using a once per revolution clock index derived from the motor drive current waveform or any other sensor. In one aspect of correcting for systematic errors in the writing of timing patterns on a storage medium of a storage device, a time interval between a trigger pattern written at a first radial position of the storage medium and a rotational index is measured. The rotational index is related to the rotational orientation of the storage medium with respect to a fixed frame of the storage device. The location of another trigger pattern to be written is shifted, using the measured time interval to determine the shift in location for the another trigger pattern.

Patent Number: 7,136,243 Issued on 11/14/2006 to Chainer, et al.

Inventors: Chainer; Timothy J. (Mahopac, NY), Praino; Anthony P. (Poughquag, NY), Schultz; Mark D. (Elmsford, NY), Webb; Bucknell C. (Ossining, NY), Yarmchuk; Edward J. (Mahopac, NY)
Assignee: Hitachi Global Storage Technologies Netherlands B.V. (Amsterdam, NL)

Appl. No.: 11/015,130

Filed: December 17, 2004

Related U.S. Patent Documents


Application Number	Filing Date	Patent Number	Issue Date
09723362	Nov., 2000	6898035
09550643	Oct., 2001	6307697
08882396	Jun., 2001	6252732
08628910	May., 1999	5901003
08348773	Dec., 1994

Current U.S. Class: 360/51 ; 360/75

Current International Class: G11B 5/09 (20060101); G11B 21/02 (20060101)

References Cited [Referenced By]

U.S. Patent Documents


4414589	November 1983	Oliver et al.
4839877	June 1989	Kulakowski et al.
4879608	November 1989	Sano
4912576	March 1990	Janz
5229901	July 1993	Mallary
5416652	May 1995	Lewis
5448429	September 1995	Cribbs et al.
5485322	January 1996	Chainer et al.
5581420	December 1996	Chainer et al.
5600501	February 1997	Yamakoshi et al.
5612833	March 1997	Yarmchuk et al.
5615058	March 1997	Chainer et al.
5682274	October 1997	Brown et al.
5694265	December 1997	Kosugi et al.

Foreign Patent Documents


A-63-000872	Sep., 1988	JP
A-04 023276	Jul., 1992	JP
WO 94/11864	May., 1994	WO

Other References

"Regenerative Clock Technique For Servo Track Writers", IBM Technical Disclosure Bulletin, vol. 33, No. 5, pp. 310-311, Oct. 1990. cited by oth- er.

Primary Examiner: Sniezek; Andrew L.
Attorney, Agent or Firm: Duncan, Esq.; Patrick W. Schiller, Esq.; Blanche E. Heslin Rothenberg Farley & Mesiti P.C.

Parent Case Text

This application is a continuation application of application Ser. No. 09/723,362, filed Nov. 27. 2000, now U.S. Pat. No. 6,898,035, which is a divisional application of application Ser. No. 09/550,643, filed Apr. 14, 2000, now U.S. Pat. No. 6,307,697, issued Oct. 23, 2001, which is a divisional application of application Ser. No. 08/882,396, filed Jul. 10, 1997, now U.S. Pat. No. 6,252,732, issued Jun. 26, 2001, which is a continuation-in-part application of application Ser. No. 08/628,910, filed Apr. 08, 1996, now U.S. Pat. No. 5,901,003, issued May 04, 1999, which is a divisional application of application Ser. No. 08/348,773, filed Dec. 01, 1994, now abandoned. Application Ser. No. 09/723,362 and U.S. Pat. Nos. 6,307,697, 6,252,732 and 5,901,003 are hereby incorporated herein by reference in their entirety.

Claims

What is claimed is:

1. A method for correcting for systematic errors, said method comprising: detecting a circumferential systematic timing error at a plurality of differing circumferential positions, wherein a circumferential systematic timing error is an along-track systematic timing error that varies with circumferential position; and correcting for one or more circumferential systematic timing errors in the writing of one or more timing patterns on a storage medium by a head of a recording device.

2. The method of claim 1, wherein a circumferential position is relative to an index position.

3. A method for correcting for systematic errors in the writing of timing patterns on a storage medium by a head of a recording device, said method comprising: detecting a circumferential systematic timing error at a plurality of differing circumferential positions, wherein a circumferential systematic timing error is an along-track systematic timing error that varies with circumferential position, and wherein said detecting a circumferential systematic timing error comprises computing an integral correction value for a time interval, wherein a non-zero integral indicates a circumferential error; and correcting for one or more circumferential systematic timing errors.

4. The method of claim 3, wherein said computing comprises adding a random error correction for said time interval to said integral correction value to obtain said integral correction value.

5. The method of claim 4, wherein said correcting comprises calculating a target interval for said time interval, said calculating using said integral correction value.

6. An apparatus for correcting for systematic errors, said apparatus comprising: a controlling unit adapted to detect a circumferential systematic timing error at a plurality of differing circumferential positions, wherein a circumferential systematic timing error is an along-track systematic timing error that varies with circumferential position; and said controlling unit being further adapted to correct for one or more circumferential systematic errors in the writing of one or more timing patterns on a storage medium by a head of a recording device.

7. The apparatus of claim 6, further comprising a computing unit adapted to compute an integral correction value for a time interval, wherein a non-zero integral indicates a circumferential error.

8. The apparatus of claim 7, wherein said computing unit is further adapted to add a random error correction for said time interval to said integral correction value to obtain said integral correction value.

9. The apparatus of claim 8, wherein said controlling unit is further adapted to calculate a target interval for said time interval, using said integral correction value to correct for said one or more circumferential errors.

10. The system of claim 6, wherein a circumferential position is relative to an index position.

Description

CROSS REFERENCE TO RELATED APPLICATIONS

This application is related to U.S. Pat. No. 5,485,322 of T. Chainer et al. issued on Jan. 16, 1996 entitled, "Method and System for Writing a Clock Track on a Storage Medium", which is a divisional application of application Ser. No. 08/028,044, filed Mar. 8, 1993, now abandoned. U.S. Pat. No. 5,485,322 is commonly assigned herewith, and hereby incorporated herein by reference in its entirety. It is also related to U.S. Pat. No. 5,612,833 of E. Yarmchuk et. al., issued on Mar. 18, 1997 entitled, "Radial Self-Propagation Pattern Generation for Disk File Servowriting," which is also hereby incorporated herein by reference in its entirety.

TECHNICAL FIELD

The invention relates generally to hard disk drive memory storage devices for computers. More particularly, it relates to disk drive apparatus and to a method for writing servotrack information therein. More specifically it relates to alleviating the need for a complex mechanical and/or optical positioning system to establish servopatterns on the recording surfaces of the recording media.

BACKGROUND ART

Increased levels of storage capacity in floppy and hard disk drives are a direct result of the higher track densities possible with voice-coil and other types of servo positioners, as well as the ability to read and write narrower tracks by using, for example, magnetoresistive (MR) head technology. Previously, low track density disk drives were able to achieve satisfactory head positioning with leadscrew and stepper motor mechanisms. However, when track densities become so great that the mechanical error of a leadscrew stepper motor combination is significant compared to track-to-track spacing, an embedded servo is needed so that the position of the head can be determined from the signals it reads.

Conventional servo-patterns (also referred to as servo-data) typically include short bursts of a constant frequency signal, very precisely located offset from a data track's center line, on either side. The bursts are written in a sector header area, and can be used to find the center line of a track. Staying on center is desired during both reading and writing. Since there can be sixty, or even more, sectors per track, that same number of servo-data areas must be dispersed around a data track. These servo-data areas allow a head to follow a track center line around a disk, even when the track is out of round, as can occur with spindle wobble, disk slip and/or thermal expansion. As technology advances to provide smaller disk drives, and increased track densities, the placement of servo-data must also be proportionately more accurate.

One example of servo-data is shown in FIG. 1, which includes a sector header 2 followed by a pattern to provide radial position information. The sector header includes a Servo ID (SID) Field 4 and a Grey Code Field 6, which require precise alignment track to track. Misalignment in these patterns results in destructive interference of the magnetic pattern and reduces the amplitude of the signal which leads to errors. Specifications on the alignment in modern disk drives is approximately 25 nanosec (3 sigma) track to track for a disk rotation period of roughly 11 milliseconds or 2.3 ppm. This narrow time window therefore requires precise measurement of the disk angular position over many revolutions of the disk.

As disk drives become smaller and track densities increase, there is a desire to reduce the size of the servo-data areas, such that they take up less space on the disk. In order to reduce the size, however, the servo-data are written at higher and higher frequencies. These higher frequencies require tighter timing tolerances from track to track.

In one example, timing is provided by writing trigger patterns at various locations of the disk. It is understood that in writing a trigger pattern a specified time after a trigger, the presence of electronic delays in the trigger and write circuitry is taken into consideration. This is described in IBM Technical Disclosure Bulletin, Vol. 33, No. 5 (October 1990), where the delay between A and B clock areas is measured and stored. This delay value is used to advance the write timing of all subsequent servo-tracks and clock areas.

Although, the IBM Technical Disclosure Bulletin, Vol. 33, No. 5 (October 1990) discussed the presence of electronic delays, it did not discuss how to achieve optimum track to track trigger pattern alignment in the presence of systematic errors (e.g., constant for every sector), which vary as a function of radial position and/or circumferentially in the propagation process.

Therefore, a need still exists for a capability to reduce systematic errors in the writing of timing patterns. In particular, a need exists for an improved capability to: 1) Eliminate a varying systematic error when the recording head has spatially separate read and write elements, such as is the case for magnetoresistive heads. This results in a read to write time delay which is radially dependent. 2) Remove the varying systematic error due to a read element and write element, which are non-parallel, resulting in an error in the measurement and subsequent writing of trigger patterns. 3) Eliminate servo pattern rotation due to residual or unmeasured systematic errors by using a once per revolution clock index derived from the motor drive current waveform or any other sensor. 4) Reduce systematic errors which vary circumferentially. 5) Minimize the error due to a recording head mounted non-parallel to the actuator motion.

SUMMARY OF THE INVENTION

The shortcomings of the prior art are overcome and additional advantages are provided through the provision of a method and apparatus for correcting for systematic errors in the writing of timing patterns on a storage medium of a storage device.

In one example, a time interval is measured between a trigger pattern written at a first radial position of the storage medium and a rotational index. The rotational index is related to the rotational orientation of the storage medium with respect to a fixed frame of the storage device. Further, the location of another trigger pattern to be written is shifted. The measured time interval is used in determining the shift in location of the another trigger pattern.

In another embodiment of the invention, a method and apparatus for correcting for systematic errors in the writing of timing patterns on a storage medium of a storage device are provided. The method includes, for instance, determining a systematic delay for use in writing at least one trigger pattern at one radial position of the storage medium, and updating the systematic delay for use in writing at least one trigger pattern at another radial position of the storage medium. The updated systematic delay is different from the systematic delay for use in writing at the one radial location.

In another aspect of the present invention, a method and apparatus for correcting for systematic errors during self-servowriting of a storage device are provided. In one example, a systematic error is measured at a plurality of predetermined radial positions, wherein a measured systematic error for each of the plurality of predetermined radial positions having a mean whose absolute value is greater than zero is determined. Each of the measured systematic errors is corrected in order to reduce the measured systematic errors.

In yet a further embodiment of the present invention, a method and apparatus for self-servowriting a storage medium in a storage device including a head positioned for interaction with the storage medium are provided. In one example, the method includes writing a first set of trigger patterns on a first track of the storage medium; determining a read to write time delay; determining an index correction value; determining at least one random error correction value; computing a set of delay values which are a function of the read to write time delay, the index correction value and the at least one random error correction value; and triggering from the first set of trigger patterns and writing a second set of trigger patterns on a subsequent track using the set of delay values.

In another embodiment of the present invention, a method and apparatus for correcting for systematic errors in the writing of timing patterns on a storage medium by a head of a storage device are provided. The method includes, for instance, detecting one or more circumferential systematic errors and correcting for the one or more circumferential systematic errors.

The present invention advantageously corrects for a varying systematic error, such as when the recording head has spatially separate read and write elements, such as is the case for magnetoresistive heads. Additionally, the varying systematic error due to a read element and a write element, which are non-parallel, resulting in an error in the measurement and subsequent writing of trigger patterns, is reduced. Further, it reduces servo-pattern rotation due to residual or unmeasured systematic errors by using a once per revolution clock index derived from the motor drive current waveform or any other sensor.

Additional features and advantages are realized through the techniques of the present invention. Other embodiments and aspects of the invention are described in detail herein and are considered a part of the claimed invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The subject matter which is regarded as the invention is particularly pointed out and distinctly claimed in the claims at the conclusion of the specification. The foregoing and other objects, features, and advantages of the invention will be apparent from the following detailed description taken in conjunction with the accompanying drawings in which:

FIG. 1 illustrates a disk and a typical disk servo pattern;

FIG. 2a depicts one embodiment for writing timing marks on a disk surface;

FIG. 2b shows how independent casual paths of pattern propagation are generated and can be eliminated by measuring and by using an additional revolution of the disk;

FIGS. 3a 3c depict one example of a self-servowrite apparatus incorporating and using the capabilities of the present invention;

FIGS. 4a 4b show one example of how servo patterns are written with a dual revolution process;

FIGS. 5a 5c show how the physical separation of read and write elements results in a time delay;

FIGS. 6a 6b show how a non-parallel read/write head results in an apparent shift of the trigger pattern when the head is moved off track center;

FIG. 7a shows non-parallel read and write elements resulting in pattern rotation;

FIG. 7b shows a misaligned head mounting resulting in pattern rotation;

FIG. 8 shows the effect of systematic errors on pattern propagation;

FIG. 9 shows one example of a process to reduce systematic error and control pattern rotation, in accordance with the principles of the present invention;

FIG. 10 depicts one example of a block diagram of the processes used to reduce systematic and random errors in a pattern propagation process, in accordance with the principles of the present invention;

FIG. 11 is one embodiment of the start-up procedure shown in FIG. 10, used to initialize various variables to be used during the systematic and random error reduction process of the present invention;

FIG. 12 depicts one embodiment of the logic used to measure intervals at TP.sub.i, in accordance with the principles of the present invention;

FIG. 13 depicts one embodiment of the read procedure depicted in FIG. 10, in accordance with the principles of the present invention;

FIG. 14 depicts one embodiment of the logic used to measure and update an index estimate, in accordance with the principles of the present invention;

FIG. 15 depicts one embodiment of the logic used to correct measured intervals, in accordance with the principles of the present invention;

FIG. 16 depicts one embodiment of the logic used to compute algorithm corrections and target intervals, in accordance with the principles of the present invention;

FIG. 17 depicts one embodiment of the logic used in the step procedure depicted in FIG. 10, in accordance with the principles of the present invention;

FIG. 18 depicts one embodiment of the logic used during the write procedure of FIG. 10, in accordance with the principles of the present invention;

FIG. 19 shows one example of a technique to remove the effects of spindle velocity jitter, in accordance with the principles of the present invention;

FIGS. 20a 20b show examples of a technique of clock propagation with reduced interval jitter, in accordance with the principles of the present invention;

FIG. 21 shows one example of a technique of clock propagation without incurring an additional revolution of the disk, in accordance with the principles of the present invention;

FIG. 22 shows one example of how servo patterns are written with a single revolution clock propagation process, in accordance with the principles of the present invention; and

FIG. 23 depicts one example of a hardware technique to detect trigger patterns written on a disk, in accordance with the principles of the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION

Self-Servowriter Overview

One embodiment of a self-servowrite apparatus incorporating and using the capabilities of the present invention is shown in FIGS. 3a 3c, and described in detail below.

Referring to FIG. 3a, in one example, a self-servowrite apparatus 300 interfaces to, for instance, a storage device 302 (such as, for instance, a disk storage device). In one embodiment, the storage device includes a recording head or transducer 304, which writes information to and/or reads information from one or more storage medium 306 (referred to herein as storage media). The storage medium rotates via a motor 312 located within storage device 302.

Transducer 304 is coupled to an actuator 308, used in moving the transducer radially along the storage media. Transducer 304 receives its signals from a voice coil motor 310 coupled thereto.

The storage device is coupled to self-servowrite apparatus 300, as described in further detail herein. In one example, self-servowrite apparatus 300 includes a PC controller 314, a radial propagation controller 316, a timing propagation controller 318, a pattern generator 320, a motor controller 322 and a read/write interface card 324, each of which is described below.

PC controller 314 is a standard personal computer (such as, for instance, an Aptiva Model 2176 offered by International Business Machines Corporation) coupled to other components of the system via a system bus 315. In one example, the system bus conforms to ISA standards and enables communication among the PC controller and the other system components, as well as between the components themselves. The bus allows addressing and control of the other components by the PC controller. In particular, the controller is controlling the components on a sector by sector basis during the technique of the present invention. In another example, the bus can be eliminated and each of the components can communicate directly.

Coupled to bus 315 is radial propagation controller 316, which is used to control the writing of radial propagation bursts (described below) on storage medium 306. In particular, the radial propagation controller reads the propagation bursts along a track on the storage media to position the transducer to write subsequent propagation bursts on a next track.

One example of a radial propagation controller is described with reference to FIG. 3b. In one example, radial propagation controller 316 includes an amplitude demodulator 330, which receives as input readback signals from read/write interface card 324. The read/write interface card receives the signals from recording transducer 304. The amplitude demodulator converts the readback signals to a position signal, which is input to an analog/digital (A/D) converter 332. The converter converts the signal to digital and passes the signal to a processor 334 (e.g., a digital processor, such as a TMS 320 offered by Texas Instruments).

The processor calculates the error in the present head position from the desired position and outputs a correction value to a digital to analog converter (DAC) 336. DAC 336 converts the correction value into an analog signal, which is input to a voice coil motor (VCM) driver 338. The VCM driver converts the analog voltage signal into a current, which drives actuator 308 to move transducer 304.

Returning to FIG. 3a, timing propagation controller 318 is also coupled to system bus 315. The timing propagation controller controls the writing of trigger patterns (described below) on a given track. In particular, it reads trigger patterns on one track and writes subsequent trigger patterns on a next track.

One example of a timing propagation controller is described with reference to FIG. 3c. In one example, timing propagation controller 318 includes a trigger pattern detector 340, a time interval analyzer 342, a processor 344, one or more memory units 345 coupled to processor 344, a trigger pattern window 346 and a programmable delay 348, each of which is described below.

Input to trigger pattern detector 340 are readback trigger patterns output from read/write interface card 324. Detector 340 converts the readback trigger patterns into a digital signal, which is input to time interval analyzer 342.

Time interval analyzer 342 precisely measures the time between trigger patterns. In one example, time interval analyzer 342 is a HP 5372A. The measured time is output from analyzer 342 and input into processor 344.

In one example, processor 344 is a digital processor (e.g., a TMS 320 offered by Texas Instruments), which stores the information received from analyzer 342, as well as other components of controller 318. Processor 344 includes the logic used to perform many of the calculations and determinations used in the present invention, as described below with reference to the flow diagrams. For example, the processor is used to calculate a delay value used to shift the location of a subsequent pattern generator output to write a trigger pattern.

In one example, the processor controls the opening of trigger pattern window 346 at a desired time, which enables the trigger pattern detector at the desired times, such that the incidences of false triggers are reduced.

In addition to the above, controller 318 includes programmable delay 348, which receives its input from processor 344. Processor 344 uses the time measurements provided by time analyzer 342 to calculate a shift value, which is input to programmable delay 348. The shift value is used to shift the writing of the product servo-patterns. The programmable delay provides the shifted signals (e.g., in time) to pattern generator 320.

An input to processor 344 is motor index 323. In particular, the motor index is input to a counter 349 within the processor. The counter measures the time between the index and a first trigger pattern. When the motor index comes in, the counter starts counting until the next trigger pattern is detected.

Referring back to FIG. 3a, the system also includes pattern generator 320, used to write patterns and to store them in its memory. It is capable of writing various patterns, such as, for instance, radial propagation burst patterns, trigger patterns and/or product servo-patterns, which are described below.

The radial propagation burst patterns provide radial information to move the actuator inwards while servoing on the edge of the tracks. The trigger patterns are used to derive precise timing or circumferential information for use in writing product servo-patterns. Using the radial and trigger patterns, the product servo-patterns are written. Particularly, the radial and trigger patterns are used to precisely line up the product servo-patterns track to track.

In one example, in order to write a track of servo-patterns, the radial propagation controller instructs the pattern generator to write a sequence of radial bursts on a track of the storage media via read/write interface card 324. The read/write interface card provides a signal to the transducer to write patterns on the storage media. Further, the timing propagation controller controls the pattern generator to write product servo-patterns at specified locations.

Although the pattern generator is depicted as a separate component within the self-servowrite apparatus, it can be a part of other components, such as, for instance, the radial propagation controller and/or the timing propagation controller.

The output of pattern generator 320 (e.g., a stored pattern) is, for example, input to read/write interface card 324, which generates one or more signals representative of the stored pattern. The signals are input to recording head 304, which writes the signals on storage media 306.

In addition to the above, in one example, self-servowrite apparatus 300 includes motor controller 322 used to rotate storage media 306 and to derive an index pulse 323 from motor 312 of the storage device.

Described above is one example of a self-servowrite apparatus coupled to a storage device. The self-servowrite apparatus and/or the storage device may include additional, less and/or different components than that described above without departing from the spirit of the present invention. For instance, one processor may be used to perform the operations of PC controller 314, processor 334 and/or processor 344. Alternatively, a plurality of processors may be employed to perform the functions of one or more of the above-described processors. For example, processor 344 may include one or more processors (referred to as a processing unit).

Further, one or more timing propagation controllers may be used to perform the functions provided by timing propagation controller 318. These one or more controllers form a controlling unit. Similarly, one or more radial propagation controllers may be used to perform the functions provided by radial propagation controller 316. Again, the one or more controllers form a controlling unit. It is also possible that the functions of the radial and timing propagation controllers can be combined into one or more controllers, which also can be referred to as a controlling unit.

Typically, a commonly used, servo architecture known as "embedded servo" in which servo pattern information is present on all disk surfaces is used in disk drives. A dual revolution clock propagation process described herein and in U.S. Pat. No. 5,485,322 is used to gate pattern generator 320 to write servo patterns onto all surfaces. The ability to accomplish this with a minimum of revolutions is important in the case when the servowriter is external to the disk drive which is contemplated to be the preferred commercial implementation of this invention. To write the embedded servo requires a clock to position the embedded servo circumferentially.

Any process used to generate a clock for the pattern generator by self-propagation using only the disk drive data heads typically requires several thousand steps during the servowriting of the disk surface. Since a single head cannot write and read simultaneously, self-propagation requires dividing a track on the disk into alternating sectors, such that timing synchronization is obtained in one sector just prior to writing the next sector as shown in FIGS. 2a 2b. After moving the head to an initial starting track, track #0, STEP 200, trigger patterns (TP) are written on the odd numbered sectors, STEP 202. The head is displaced a fraction of a track to servo track #1, STEP 204, and the odd numbered sectors 1, 3, 5, . . . 59 are used for synchronization in writing the subsequent TP's on the even numbered sectors 0, 2, 4, . . . 60, STEP 206. The head is again displaced a fraction of the track to servo track #2, STEP 208, and the roles are reversed, with the even numbered sectors used for synchronization in writing the subsequent odd numbered sectors, STEP 210. This process continues until the head is displaced across the recording surface.

Each time a sector is written, a small but unavoidable timing error called the base error occurs due to both disk rotational velocity changes and electrical noise in the readback signal used for synchronization. These errors are replicated on the next step as the written sectors change over to synchronization sectors. A new and independent set of random errors is added here as well. Thus, the self-propagation process involves a replication and summation of the random errors in each step. Left uncorrected, statistically such errors grow as the root mean square of the number of steps. Since the process of synchronization and writing is causal, i.e. synchronization occurs before writing on each step, the sequence of random errors that determines the location of a given sector timing pattern can be traced back in a spiral path which is shown by the arrows labeled Path1, Path2 and Path3 of FIG. 2b. The paths run in parallel and for adjacent radial locations at each sector are totally independent so the track to track misalignment ends up as the difference between two independent random walks, giving an additional factor of the square root of 2. Therefore, after 10,000 steps, the RMS track to track error would be 141 times the base error. It should be noted that this process can work with other than alternating patterns limited to two. For instance, two or three patterns may be used for clock propagation.

As described in U.S. Pat. No. 5,485,322, errors in the pattern propagation process can be detected and corrected to produce precise magnetic pattern alignment from track to track. Errors written at each step are measured during a subsequent revolution of the disk and are accounted for and corrected. For example, in FIG. 2b prior to writing TP 26, the interval between TP 24 and TP 42 is measured on an extra revolution of the disk and stored as a time interval T45. The time interval T45 is used when the head is positioned on servo track #3 and triggers on TP 24 to write TP 26 aligned to TP 42. The extra revolution of the disk breaks the causal chains denoted by Path 2 and Path 3, effectively locking together the normally independent random walk paths and preventing the growth of track to track misalignment. This process, as described in U.S. Pat. No. 5,485,322, is referred to herein as the "dual revolution clock propagation process."

In FIGS. 4a 4b, there is shown one embodiment of a process by which trigger patterns are used to write servo patterns on a first surface using a first recording head and other surfaces by switching to other recording heads on other recording surfaces. The recording head is set to read mode and triggers on TP#0, which begins gating pattern generator 320 to write servo pattern SP#0, then the next recording head in the stack of a multiplatter disk drive is selected and a second servo pattern SP#1 is written on surface #1 and repeated to N surfaces. The number N is limited by the time required to write each surface and switch the heads. At the end of the Nth surface, the head is switched to the first surface #0 in time to read the next sector TP#1. This process makes clear the advantage of trigger patterns as their short time allows other surfaces to be written with servo patterns within the same revolution of the disk, thus reducing servo write time. In particular, servo-patterns can be written while writing trigger patterns.

Geometrical Effects Resulting in Systematic Errors

It is desirable for clock propagation processes to have systematic errors reduced during the propagation process. Such errors contribute to track misalignment and/or rotation of the pattern with respect to a fixed frame. The most obvious systematic error, as described in IBM Technical Disclosure Bulletin, Vol. 33, No. 5 (October 1990), is the electronic delay in the read to write circuitry. However, the prior art does not teach that radial varying systematic errors (which change with radial position) due to geometric effects which are particular to a head and actuator design are present. If these errors are not compensated, they can reduce the track to track accuracy of the propagation process and can accumulate to result in rotation of the pattern with respect to an index (fixed reference). This rotation is shown graphically in FIG. 8, where each TP 800 on a subsequent servo track is displaced from the previous servo track due to a systematic error causing both pattern rotation and a fixed track to track misalignment error.

In the case where the pattern rotation matches the trajectory of the actuator during a seek, no shift in timing relative to the index related to the rotational orientation of the disk relative to the base plate (which does not change with radial position) will be present. This is the desired pattern rotation since it simplifies the disk file servo during long seeks. Thus, patterns which are written in such a manner as to have no timing shift relative to an index related to the rotational orientation of the disk will necessarily be of the desired type. Several examples of systematic errors are described below, but it should be understood that other similar effects can result from particular head and disk drive design. In general, for a given systematic error, there are three techniques which can be used to remove the systematic error: 1) Measurement of the error at one or more known radii which are used to compute the correction term at each propagation step; 2) In-process correction of the error by internal measurements during the propagation process; and 3) Measurements of the error in the disk file, by external means, which are used to correct all disk drives of similar design. Radial Dependent Time Delay for Independent Read/Write Sensors

The dual revolution timing pattern process described in U.S. Pat. No. 5,485,322 is modified when the recording transducer has write and read elements which are separate as shown in FIG. 5a. In this case, a time delay or advance (relative to the case where the write and read elements are not separate) is required to write a second trigger pattern a specified physical distance after a readback trigger pattern. This time delay is a function of the disk linear velocity given by:

##EQU00001## where W is the separation between read element 46 and write element 48, and V is the disk linear velocity.

Further, in the case where the head is skewed relative to the track, as for example, would occur on a rotary actuator, the "apparent" separation W of the read and write transducers, as projected along the recorded track, is modified by the cosine of the skew angle of the recording head and will also vary as a function of radius (see FIG. 5b). The time delay due to the separation effect is given by:

.function..THETA..times..OMEGA..times..times. ##EQU00002## where .THETA. is the skew angle of the recording head, .OMEGA. is the disk rotation speed, and R is the radius of the track.

The read to write separation W can be measured, for example, using the product data head by writing a TP 50 (FIG. 5c), triggering on that trigger pattern on a subsequent revolution and writing a second trigger pattern TP 52, a time T54 later. Subsequent measurement of the actual time between the two trigger patterns is equal to the time T54 plus any delay in the electronics (which is fixed) and the delay due to the physical read to write separation W. Repeating this measurement at two known radii (R.sub.1 and R.sub.2), possibly the innermost accessible data track or inner radius (ID) and the outermost accessible data track or outer radius (OD), with a known skew angle can be used to compute the separation between the read and write elements. The measured total delay is given by:

.times..times..function..THETA..times..OMEGA..times..times. ##EQU00003## .times..times..function..THETA..times..OMEGA..times..times. ##EQU00003.2## where this total delay referred to as the read to write delay, includes a delay due to the read to write separation and an additional delay due to the electronics path in either or both the read and write chain of the recording head.

The above relationships can be solved for W as given by:

.OMEGA..times..times..times..times..times..times..function..THETA..times..- function..THETA..times. ##EQU00004##

The delay for each propagation step can then be computed and updated as the timing pattern is propagated across the disk surface (i.e., technique #1 of systematic error removal).

Alternatively, as noted in technique #2, the read to write delay can be measured during the propagation process and used to correct for the variable delay. For the case of the dual revolution timing pattern process, a measurement of the delay and correction values (e.g., HFM, HFS, HFC and SD where gp, gd and gi (each of which is explained later below) are set to 0) is available on each readback measurement revolution, which is described in the section "Process with Systematic Error Removal" below.

In some cases, it may be desirable to use measured delays to find the absolute radial position of the head. For a given disk file, the relationship between radial position of the head and the skew angle may be ascertained.

Once the separation between the read and write elements is known, the electronic component of the delay may be determined by solving the above equation for the read to write delay at a given known radius. The same equation may now be solved to determine absolute radial position by using the read to write delay measured with the head at any other location.

Read to Write Non-Parallelism

An additional systematic error occurs if the read element of a recording head is non-parallel to the writing element. This is shown in FIG. 6a where the written trigger patterns, TP 64, TP 66, and TP 68 are non-parallel to read element 62. TP 68 is written approximately one half off-track from TP 64 and TP 66. When read element 62 on track center 70 reads the trigger patterns, the time interval measured between TP 64 and TP 66 is given by D1. The time interval measured between TP 66 and TP 68 is equal to D2. The time interval D2 will be longer than D1 due to the fact that TP 68 is off track and its angle results in a shift in the apparent center of the trigger pattern. To further clarify this effect, when the read and write elements are parallel as shown by trigger patterns TP 72, TP 74 and TP 76 (FIG. 6b) the measured intervals per the above description result in the two measured intervals D1' and D2', which are equal. It should be noted that any non-parallelism between the read element and the written transition will cause this effect and it need not be limited to a constant angle between the read and write elements. Specifically, a bending or warping of the written transitions arising from edge effects in the write process will produce this effect.

This error can be removed during the propagation process by measuring the systematic error of intervals comprised of both on-track trigger patterns versus those intervals with one on-track and one off-track signal and storing the value.

Alternatively, as noted in technique #2, this error can be measured during the propagation process, as a separate periodic measurement step. In particular in the case where the read and write elements have a more complex geometric misalignment rather than simply non-parallel, then the delay may depend non-linearly on radial location. Correction of the effect would require periodic measurement during the propagation process.

This geometric effect can create a pattern spiral which is shown in FIG. 7A. Trigger patterns TP 82, TP 84 and TP 86 are aligned to minimize the track to track misalignment of the pattern, but the result is a rotation of the pattern from Index 80 in the time domain. Pattern rotation can be reduced by using a once per revolution index (i.e., an index fixed to an absolute frame, e.g., a motor index), as described in the "Process With Systematic Error Removal" below.

Misaligned Head Mounting

An additional geometric effect can occur if the recording head is non-parallel to the direction of actuator motion. When the trigger patterns are aligned to minimize the track to track error, a pattern rotation will result. A simple example is shown in FIG. 7B, in which a head is non-parallel to the motion of a linear actuator. When the trigger patterns TP 90, TP 92, and TP 94 are aligned, a pattern rotation occurs. Removing the pattern rotation shown by TP 90, TP 96 and TP 98 results in a track to track misalignment. A similar effect occurs when a head is misaligned on a rotary actuator.

Process with Systematic Error Removal

In the propagation process shown in FIG. 2a, the odd numbered sectors are used for synchronization in writing the subsequent even numbered sectors and on the next step (after moving the head a short distance radially), the roles are reversed, with the even numbered sectors used for synchronization in writing the subsequent odd numbered sectors. Although the growth of random errors is removed by the measurement feedback process, a systematic error can result in a fixed track to track error and in a rotation of the entire servo pattern with respect to a fixed reference frame. This is shown in FIG. 8 where subsequent written trigger patterns are displaced a fixed distance from the previous servo track due to a systematic error.

One embodiment of a flow chart in which systematic errors are reduced and the growth of random errors are controlled is shown in FIG. 9 and described below. STEP 100) The process is initiated by writing a sequence of trigger patterns and measuring the delay between read and write, as previously described, and storing the delay as a variable D.sub.RW0. The TP shift due to read/write parallelism is checked and measured and stored as D.sub.S0. STEP 102) The read to write delay determined from STEP 100 is used to write, for instance, the even numbered TPs on the disk on, for instance, servo track number 0. The location of the first TP with respect to the index is recorded and the index correction is set to zero. STEP 103) The head is moved a fraction (e.g., 1/3) of the track width. STEP 104) The alternate numbered TPs are written. This is initially, for instance, the odd TPs, but will alternate between even and odd on successive propagation steps. STEPS 105) Every interval corresponding to the time between trigger patterns is measured. At STEP 105A, the intervals are divided into two groups, Interval Full to Half (I.sub.FH) and Interval Half to Full (I.sub.HF). The I.sub.FH corresponds to a time interval beginning with a full on-track TP and ending with an off-track TP. The I.sub.HF corresponds to a time interval beginning with an off-track TP and ending with an on-track TP. Combined these account for all the intervals on the disk. Each interval is measured using, for instance, a time interval analyzer. At STEP 105B, the location of the first TP in the group written with respect to the rotational index is recorded using, for example, a time interval analyzer, during the same revolution of the disk. STEPS 106) The measured intervals are used to calculate the time delay corrections. At STEP 106A, the I.sub.HF are used to determine the read to write delay to be used in the next TP write. At STEP 106B, the I.sub.FH are input to the clock algorithm and corrections are computed to eliminate the random growth of errors in the next TP write. Elimination of the random growth of errors is described in detail in commonly assigned U.S. Pat. No. 5,485,322, issued Jan. 16, 1996, entitled "Method And System For Writing A Clock Track On A Storage Medium," which is hereby incorporated herein by reference in its entirety. At STEP 106C, the location of the first TP to the index is used to compute a new index correction to prevent the rotation of the pattern with respect to index. It should be noted this feature may be disabled for example to minimize track to track error by allowing the pattern to rotate as in FIGS. 7a 7b. The index can be obtained from the current waveform in the DC brushless motor typically used in disk drives or from any other available once-per-revolution signal derived from the orientation of the disk relative to the base plate of the drive. In one example, the index correction is equal to a predetermined desired time interval (e.g., 50 microseconds) minus the recorded time interval (between the index and the first TP). STEP 107) The outputs of each of the above described processes in STEPS 106A, 106B and 106C are used to compute the desired delay to write the alternate trigger patterns at the next radial location. This is described in further detail below. STEP 108) The servo-track count is checked to determine the end of the process (e.g., 18,000 servo-tracks).

If the process has not ended, the process loops back to STEP 103.

In the above process, in practice, one may average the measurements of the systematic errors on two or more subsequent servo-tracks prior to making corrections to the computed delay values for each interval.

Another example of reducing systematic and random errors in a pattern propagation process is described in detail below with reference to FIGS. 10 18. In one embodiment, the logic within those figures is performed by one or more controllers (e.g., a controlling unit), such as timing propagation controller 318 (FIG. 3a). For example, many of the calculations and comparisons are performed by one or more processors (e.g., a processing unit) of the controller. Further, the time measurements are performed by, for example, either the time interval analyzer or the motor index in combination with the counter within the processor, as described herein. Additionally, the programmable delay is used, as one example, to shift location of patterns.

In another embodiment, many of the calculations and comparisons are performed by one or more computing units, which may or may not be a part of controller 318 or any controlling unit. In one instance, a computing unit is a processor. In further embodiments, other components may be used to perform the below-described functions and these components may or may not be a part of controller 318 or any controlling unit. For instance, a measuring unit can be used to measure time intervals, and the measuring unit may or may not be a part of controller 318 or any controlling unit. As a further example, a processing unit (e.g., one or more processors) that is not considered a part of the controller or controlling unit may be used to perform at least a portion of the logic within controller 318.

Referring to FIG. 10, initially a start-up procedure is implemented, which is used to initialize various variables to be used during the systematic and random error elimination process, STEP 1000. One example of a start-up procedure is described in detail with reference to FIG. 11. An initial value of the read/write delay (systematic delay (SD0)) is measured in a manner described above with reference to FIG. 5c, STEP 1100. Additionally, a value for a target index time (IT), which represents the target interval between the index (e.g., the motor) and the first trigger pattern, is either measured or chosen, STEP 1102. In one embodiment, IT is set to 50 microseconds.

Further, a value for a valid interval window (VI) is set equal to a valid interval window nominal value (VI.sub.nominal), such as, for instance, 35 nanoseconds, STEP 1104. The valid interval window is used as an error check; in particular, it defines how far away from the anticipated trigger location an actual measured trigger pattern can be to conclude that it is valid. Also, an estimate of a current index location (IS) is set to the target index time, which is updated as the technique proceeds along to the target, STEP 1106.

In addition to the above, a number of other variables are initialized, STEP 1108. In particular, a step value (N) is set equal to zero. In one example, there are 18,000 steps. Further, an index correction value is set equal to zero, as well as a half to full estimate (HFS) and a half to full correction (HFC). Additionally, a local average interval error (LAIE) is set to zero, as well as an integral algorithm correction (IAC) for all i. A trigger point valid flag is set to false for all of i and a value representing the number of equally spaced trigger patterns (Nint) is set equal to 90, as one example. As used herein, all index i are assumed to be modulo and can range from zero to Nint-1 (i.e., eighty-nine).

After the above values are initialized, an initial set of equally spaced trigger patterns (Nint) are written at intervals TI.sub.i with the first trigger pattern near the target index time with respect to the index, STEP 11110. All TI.sub.i are equal to the target average interval (TAI) to start with, which is the nominal revolution time divided by the number of intervals in a revolution.

Subsequently, a measured interval (MI.sub.0) is measured at trigger pattern zero using, for instance, a time interval analyzer, STEP 1112, and a variable referred to as TS (time slip) is set to zero, STEP 1114. This value represents an error handling variable that keeps track of the magnitude of the error in an erroneous trigger. This logic assumes the existence of hardware, which generates a fake trigger pattern if a valid or erroneous trigger pattern is not received within a specified time; i.e., trigger patterns valid, erroneous or fake are generated near every anticipated trigger pattern location. In the beginning, it is assumed there are no errors and therefore, TS is set to zero. Thereafter, an interval (MI) is measured at trigger pattern 1, STEP 1116, as described in detail below.

One embodiment of measuring intervals at trigger pattern one (TP.sub.(1)) is described in detail with reference to FIG. 12. Initially, the interval at TP.sub.i where i is equal to 1 is measured using, for example, the time interval analyzer, STEP 1200. (That is, the interval between TP.sub.(0) and TP.sub.(1) is measured.) Thereafter, the measured interval is corrected for previous invalid trigger patterns, STEP 1202. This is accomplished by setting MI.sub.i equal to MI.sub.i+TS (i.e., MI.sub.i=MI.sub.(i)+TS). Since the first trigger pattern is assumed to be valid, TS is set to zero and no correction is made.

Next, a determination is made as to whether the immediately corrected value of MI.sub.i minus the value of the target interval (TI.sub.(i)) is greater than the valid interval window (VI), INQUIRY 1204. If |MI.sub.i-TI.sub.i| is greater than the valid interval window, then MI.sub.(i) is considered to be invalid. Thus, the value of time slip (TS) is set equal to MI.sub.i-TI.sub.i, and the trigger pattern valid flag (TPV.sub.(i)) is set to FALSE, STEP 1206. Thereafter, another inquiry is made as to whether the valid interval window is greater than a maximum value that has been set (VI.sub.limit), INQUIRY 1208. If the valid interval window is greater than the maximum (e.g., 700 nanoseconds), then the process for measuring the interval at the first trigger pattern is complete, STEP 1218. However, if the valid interval window is less than or equal to the maximum, then the valid interval window is increased by a valid interval increment (VI.sub.inc), STEP 1210 and the measurement of MI.sub.i is complete. One example of a value for the valid interval increment is 35 nanoseconds.

Returning to INQUIRY 1204, if the measured interval is valid, then the time slip is set equal to zero, the trigger pattern valid flag is set equal to TRUE and the valid interval window is set equal to the nominal value (VI.sub.nominal), such as, for instance, 35 nanoseconds, STEP 1212.

Subsequently, the trigger pattern valid flag for interval i-2 is checked to see if the trigger pattern at that location was valid, INQUIRY 1214. If it was not valid, then again the measured interval process is complete and the local average interval error (LAIE) is left unchanged, STEP 1218. However, if the trigger pattern at that location is valid, then the local average interval error (LAIE) (i.e., the velocity error) is updated, STEP 1216.

In one example, the local average interval error is computed as follows: LAIE=LAIE+q.times.{([MI.sub.(i)+MI.sub.(i-1)-TI.sub.(i)-TI.sub.(i-1)]/2)-- LAIE}, where q represents a value, which determines how fast the LAIE is going to track the difference between actual interval measurements and expected nominal velocity interval measurements, and is equal to a fraction such as 0.1.

Subsequent to updating the local average interval error, the process for measuring the interval at trigger pattern one is complete, as well as the start up procedure of FIG. 10. Therefore, processing proceeds with a read procedure, STEP 1002 (FIG. 10). One embodiment of the read procedure is depicted in FIG. 13 and described in detail herein.

Initially, index i is set equal to 2, since the read is going to read back the interval starting with i=2 and go around past zero to one, STEP 1300. Thereafter, the interval at trigger pattern two is measured, STEP 1302, in the manner described above for measuring the interval at trigger pattern one (see FIG. 12). Subsequent to measuring the interval, the index is incremented by one, STEP 1304, and an inquiry is made to determine whether the process is complete, i.e., whether i has reached 2, again, INQUIRY 1306. If i is not equal to 2, then a further determination is made as to whether i is equal to a value of Nint, INQUIRY 1308. If i is not equal to Nint, then processing continues at STEP 1302 "Measure Interval At TP.sub.i." If however, i is equal to Nint, then i is recalculated by subtracting Nint from i, (i.e., i=i-Nint), STEP 1310. Thereafter, processing continues with STEP 1202 "Measure Interval At TP.sub.i."

In addition to the above, sometime after STEP 1310, an index estimate (IS) is measured and updated. One embodiment for measuring and updating the index estimate is described in detail with reference to FIG. 14. Initially, the time interval between the index and the first available non-fake (i.e., valid or erroneous) trigger pattern after the index, which is usually either at TP.sub.0 or TP.sub.1, is measured using, for example, a time interval analyzer, STEP 1400. Further, a trigger pattern indice (TPind) is set equal to zero, STEP 1402. Subsequently, a determination is made as to whether the measurement (IM) of the interval between the index and the first available non-fake trigger pattern is greater than a target (IT) plus a va