Adaptive Window Centering

The present invention deals with reading data from a magnetic disk. More particularly, the present invention deals with adjusting window center settings in a read channel of a magnetic disk drive

A magnetic disk, such as one used in a computer disk drive, is a flat circular platter with a magnetic surface. Data is stored on the magnetic surface by selective polarization of portions of the magnetic surface. The presence or absence of polarity transitions between the polarized portions represents particular binary values. Typically, the magnetically polarized portions are arranged in a plurality of radially concentric tracks on the magnetic surface to aid in location and read back of the data.

In order to read back the data recorded on the magnetic surface, a magnetic transducer moves relative to the magnetic disk along a given track. The magnetic transducer generates an electrical signal (the "read signal"), which represents the states of polarization encountered along the track. Pulses in the read signal correspond to the magnetically polarized portions of the magnetic disk.

Ideally, the read signal would not be influenced by any factors other than the magnetically polarized portions. However, other factors do influence the read signal and must be taken into account in processing the read signal.

During processing of the read signal, the processing logic (i.e. the read chain logic) is configured to recover data from the read signal during a certain period in time known as a read window. The read window represents the time during which the pulses in the read signal, which represent digital data stored on the magnetic disk, are valid. However, the actual window during which the pulses are valid can shift in time due to aging of the circuitry in the magnetic disk drive. Also, as the circuitry heats up during operation, the temperature differential can cause the read window to shift. Further, the optimum read window can shift due to process changes in the magnetic disk drive or due to part-to-part timing variation in the read chain circuitry of the magnetic disk drive.

The present invention is a technique of adjusting a window center setting in a read chain of a magnetic disk drive. A number of read operations are performed by the magnetic disk drive to read data from a magnetic disk. A number of read errors are detected during the read operations and error types are determined. Adjustment of the window center settings in the read chain is controlled based on the number of errors detected and the types of errors determined.

FIG. 1A is a plan view of a conventional disk drive.

FIG. 1B is a partial cross sectional view of the conventional disk drive shown in FIG. 1A.

FIG. 2 is a wave form representing a magnetic transducer-generated read signal.

FIG. 3 is a block diagram of a read chain.

FIG. 4 is a wave form representing a magnetic transducer-generated read signal which is shifted in time with respect to the wave form shown in FIG. 2.

FIG. 5 is a graph showing a number of read errors plotted against percent deviation of window center setting from optimum center setting.

FIG. 6 is a graph showing a probability of the occurrence of Shift Right or Shift Left errors plotted against time.

FIG. 7 shows a portion of a transducer-generated read signal.

FIGS. 8A, 8B and 8C show a flow diagram representing one embodiment of the window centering technique of the present invention.

FIG. 9 is a graph showing a number of read errors plotted against window center settings.

FIG. 1A illustrates a portion of a magnetic disk drive 10. Disk drive 10 includes magnetic disk 12, slider 14, actuator arm 16, rotary actuator 18, and control circuit 19. Although disk drive 10 is shown with only a single magnetic disk 12, it will be understood by those skilled in the art that the present invention is applicable to disk drives including a plurality of magnetic disks.

Magnetic disk 12 is mounted for rotation on a spindle centered on axis 20. Data blocks are arranged on magnetic disk 12 in a plurality of radially concentric tracks 22A-22N. Data tracks 22A and 22N are the outermost and innermost data tracks, respectively, on magnetic disk 12.

Slider 14 carries a magnetic transducer (transducer 26 shown in FIG. 1B). As magnetic disk 12 rotates, slider 14 flies just over the surface of magnetic disk 12 Slider 14 is attached to actuator arm 16, which is, in turn, attached to rotary actuator 18. Actuator arm 16 is mounted for rotation about axis 24 of rotary actuator 18. As slider 14 flies above the surface of magnetic disk 12, control circuit 19 controls rotary actuator 20 to position slider 14 over a selected data track, for example, track 221. If a plurality of magnetic disks were used, control circuit 19 would be programmed to select the appropriate disk to be read from and written to.

FIG. 1B shows a portion of a sectional view of disk drive 10. Magnetic transducer 26 is mounted on a face of slider 14 which is perpendicular to the surface of magnetic disk 12. As transducer 26 flies over magnetic disk 12, transducer 26 senses magnetic flux lines 28 emitted from polarized portions 30 of magnetic disk 12. Magnetic transducer 26 generates a read signal which is characterized by positive going or negative going transitions in pulses associated with the flux reversals of flux lines 28 emitted by polarized portions 30. Read signal 32 is demodulated and data is recovered from read signal 32 by control circuit 19.

FIG. 2 shows a graph of a read signal 32 generated by transducer 26. The positive and negative going transitions of read signal 32 correspond to the flux reversals of flux lines 28 emitted by polarized portions 30 of magnetic disk 12.

FIG. 3 shows a block diagram of a read chain 34 in disk drive 10. Read chain 34 includes transducer 26, amplifier 36, peak detector/synchronizer 38, encoder 40, controller 42 and register 44. Transducer 26 provides read signal 32 to amplifier 36. Typically, amplifier 36 is an automatic gain control amplifier. Amplifier 36 amplifies read signal 32 and provides the amplified signal to peak detector/synchronizer 38. Peak detector/synchronizer 38 detects the peaks in read signal 32 and synchronizes data pulses, representative of those peaks, with a system reference clock.

In detecting the peaks of read signal 32, peak detector/synchronizer 38 "looks" for the peaks during a time period known as the read window. Data is recovered from read signal 32 by determining whether read signal 32 passes above or below a qualification threshold during the read window. Therefore, the read window would be ideally centered around the peaks of read signal 32. If it is not, peak detector/synchronizer 38 may not detect a proper peak and a read error will occur.

Peak detector/synchronizer 38 provides output data, recovered from read signal 32, to encoder 40. Encoder 40 encodes the data into synchronized digital data and provides it to controller 42.

During a read operation, controller 42 receives the synchronized digital data provided by encoder 40 and processes the data by either providing it to, for example, a host computer (not shown) or by performing certain operational steps based on the digital data.

In this preferred embodiment, controller 42 also detects read errors which occur during the read operation. Read errors can be determined in several ways. In this preferred embodiment, a prestored location on magnetic disk 12, such as track 22I or a portion of track 22I, is prewritten with data. During operation of the magnetic disk drive 10, the prestored location is periodically read and read errors are recorded.

Read errors can take many forms including time shift errors (e.g. Shift Left or Shift Right read errors). Such read errors can affect the time position of the optimum read window. Based on these read errors, controller 42 provides a window center setting to register 44. Register 44, in turn, provides the window center setting to peak detector/synchronizer 38. The window center setting provided by register 44 causes peak detector/synchronizer 38 to reconfigure itself so the read windows such as those shown as cross hatched areas in FIGS. 2 and 4, are repositioned to be properly centered around the peaks in the read signal.

The type of read error which causes controller 42 to adjust the window center settings in register 44 vary. For example, FIG. 4 shows a read signal 32' which corresponds to read signal 32 shown in FIG. 2 in all respects except that it is shifted to the right in time with respect to read signal 32. Many things can cause such a shift as that shown in FIG. 4. For example, where magnetic disk drive 10 includes a plurality of magnetic disks, each having an associated slider and transducer, the delay times can vary from disk to disk and transducer to transducer. Also, as magnetic disk drive 10 operates, the electronic components heat up. This temperature variation can cause time shifts in read chain 34. In addition, voltage variations and component-to-component variations in the read circuitry can cause time shifting of the read signal. Further, process or part changes in magnetic disk drive 10 can cause time shifting.

When such a time shift occurs in read signal 32', time shift read errors (such as Shift Left or Shift Right read errors) occur. For example, where read window settings are set for peak detector/synchronizer 38 to those shown as cross hatched windows in FIG. 2, and which are ideally centered around the peaks of read signal 32, they are not properly set when a time shift results in read signal 32' shown in FIG. 4. If such a shift occurs, and the read window is not correspondingly reset, peak detector/synchronizer 38 will be incapable of recovering data from read signal 32' properly. Thus, adjustment of the window center settings in register 44 is required.

FIG. 5 is a graph showing the number of read errors plotted against the percent deviation of the window center setting for peak detector/synchronizer 38 in register 44 from the window center which would be ideal for read signal 32. FIG. 5 shows that, as the window center setting deviates to the right of the optimum window center (which is represented by 0% deviation), the number of read errors detected stays at a minimal level for a certain amount of time and then begins to rise. Thus, it is apparent that, as the window center setting deviates from the optimal window center, but as long as the setting is within the optimal read window, the error rate does not increase greatly.

However, the further the window center setting deviates from the optimal setting, the greater the number of read errors.

The curve shown in FIG. 5 can be determined by setting the window center setting for peak detector/synchronizer 38 to a nominal level and then performing read operations by reading data from magnetic disk 12. Each time a read error is detected, the error is recorded. After a given number of read operations are performed, the window center setting in register 44 is set to shift the window center setting to the right by a given percent. Then, more read operations are performed and the number of read errors detected is recorded. This is repeated until a threshold number of errors are detected for any given window center setting. Then, the window center setting in register 44 is set back to the nominal value and moved to the left in a step wise fashion. By accumulating the number of read errors detected at each window center setting, the curve in FIG. 5 is determined. Thus, the read window for read chain 34 can be roughly determined.

However, it is also desirable to know, as closely as possible, the exact center point within the read window. If the exact center point is known, the window center setting in register 44 can be set to the appropriate value for optimum operation.

A technique of fine tuning the window center settings is used to set the window center setting to the optimum center position within the read window. When data is read from the track, controller 42 accumulates a number of Shift Left read errors and a number of Shift Right read errors detected during the read operations.

In a 2,7 encoded bit stream, a Shift Left error is the result of an asserted bit in the 2,7 encoded bit stream being deasserted and the next bit to the immediate left being asserted. A Shift Left error represents a read errors that occurs as a result of the read window center setting being shifted to the left of the optimal setting.

In addition to Shift Left errors, Shift Left Many errors are also accumulated in controller 42. Shift Left Many errors also result from the read window settings being set earlier than the optimal read window. Shift Left Many errors are the result of an asserted bit in the 2,7 bit stream having been deasserted and a previous bit (more than 1 bit to the left) being asserted For example, in a sample 2,7 encoded bit stream:

Shift Left and Shift Left Many errors are shown in Table 1.

Shift Right errors have similar characteristics to Shift Left errors except that they result from the read window settings being set later in time than an optimal read window. A Shift Right Once error is the result of an asserted bit in a 2,7 encoded bit stream having been deasserted while the adjacent bit to the right has been asserted. Similarly, Shift Right Many errors are the result of a read window center setting being set further to the right than the optimum setting. Shift Right Many errors occur when an asserted bit in the 2,7 encoded bit stream is deasserted while a bit greater than one bit to the right is asserted. For example, in a sample 2,7 encoded bit stream:

Shift Right and Shift Right Many errors are shown in Table 2.

FIG. 6 shows a curve 46 representing the probability that a Shift Left error will occur during a read operation, and a curve 48 representing the probability that a Shift Right error will occur during a read operation. Curves 46 and 48 are shown in relation to the optimum read window 50 and the optimum window center setting 52. FIG. 6 shows that, as the window center setting moves to the right of the optimum window center setting 52, the probability that a Shift Right error will occur during a read operation increases along Gaussian distribution 48. Similarly, as the window center setting moves to the left with respect to the optimum window center setting 52, the probability that a Shift Right error will occur decreases as the probability that a Shift Left error increases along Gaussian distribution 46.

FIG. 6 also shows that, where curves 46 and 48 intersect (where the probability that a Shift Left error will occur equals the probability that a Shift Right error will occur) corresponds to the precise center of optimum read window 50 and optimum window center setting 52. Hence, if the number of read operations performed, the number of Shift Right errors that occur and the number of Shift Left errors that occur, are accumulated, then the probability that a Shift Left error will occur and the probability that a Shift Right error will occur can be determined. Also, the optimum window center setting 52 can be achieved by varying the window center setting until the probability that a Shift Left error will occur equals the probability that a Shift Right error will occur.

FIG. 7 shows the relationship between optimum window center setting 52, as determined from curves 46 and 48, as well as optimum read window 50, and one peak of read signal 32.

In this preferred embodiment, before the window center setting is changed, controller 42 is required to reach a certain confidence level that a change is required. In other words, the window center setting will only be adjusted when controller 42 reaches a certain confidence level that the probability of one type of error occurring is above or below 0.5. Then, controller 42 will adjust the window center setting in the direction of the lower of the two probabilities to equalize the probability of both error types.

The confidence level determined by controller 42 is directly related to the number of samples read from magnetic disk 12. To obtain a threshold confidence level, a certain number of samples must be taken. In this preferred embodiment, a minimum threshold number of errors must occur before controller 42 will do any further processing to determine adjustment to the window center setting. Using the binomial distribution for probability of an event happening in a random sequence, the probability of a certain number of Shift Right errors happening is determined as follows: ##EQU1## where P(r) the probability of a certain number of Shift Right errors happening;

P(1)=the probability of a certain number of Shift Left errors happening;

r=total number of Shift Right errors recorded;

l=total number of Shift Left errors recorded; and

n=total number of errors (r+l) recorded.

Also, ##EQU2##

Assuming, nominally, that P(r)=P(l)=0.5, then: ##EQU3##

If, during the read operations, only one error type is detected (e.g. a Shift Right error), then, using Eq. 3, the following table is generated.

As soon as controller 42 reaches, for example, a 90% confidence level that P(r) is greater than 50%, then it adjusts the window center setting in register 44 to the left, thereby tending to equalize P(r) and P(1). In Table 3 above, that would require four successive Shift Right errors to be detected.

In another example, where one error of a different type occurs in this sequence, the following table can be generated.

In this case, in order to reach a confidence level of 90%, six Shift Right errors must occur while only one Shift Left error occurs.

Practically, it is necessary to set a threshold confidence level in order to keep controller 42 from "hunting" for the optimum level. In one preferred embodiment, a confidence level of 85% is set and another threshold is set whereby at least 3 to 4 errors must be detected before the confidence level is checked. If all errors are of the same type (either Shift Left errors or Shift Right errors), then the confidence level can be immediately extracted. If not, some calculations need to be done. For example, the confidence level that the probability of a Shift Right error occurring is greater than 50% is represented by: ##EQU4##

Similarly, the confidence level that the probability that a Shift Left error will occur is greater than 50% is represented by: ##EQU5##

If the confidence level of either of the above two equations is greater than the threshold confidence level (in this preferred embodiment 85%) then the window center settings are adjusted. If not, no adjustment is made until the threshold confidence level is reached. Alternatively, the maximum number of a certain error type (either Shift Left or Shift Right) can be determined and a confidence level check can be performed on that error type. Then, the appropriate adjustment to the window center setting can be made.

FIGS. 8A, 8B and 8C show a flow diagram illustrating the operation of controller 42 in performing the coarse and fine adjustments to the window center settings in read chain 34. For ease of reference, FIGS. 8A, 8B and 8C should be placed in vertical relationship with one another with FIG. 8A on top and FIG. 8C on the bottom. FIGS. 8A, 8B and 8C, in this preferred embodiment, refer to strobe values. These strobe values correspond to values used with the National Semiconductor 8459 integrated circuit chip. The strobe values vary, in units of percent, between -27% and +27% with 0% corresponding to a nominal strobe value. Therefore, where controller 42 determines that the window center setting must be shifted to the left, the strobe value for the National Semiconductor 8459 integrated circuit chip is moved from, for example, 0% to -1%. Similarly, where controller 42 determines that the window center setting must be shifted to the right, the strobe value on the National Semiconductor 8459 integrated circuit chip is moved, for example, from 0% to 1%, or from 1% to 2%, etc. . . . It will be understood by those skilled in the art that a variety of circuits could be used to perform this function, and the National Semiconductor 8459 is only one preferred embodiment.

In addition, FIGS. 8A, 8B and 8C use the following terms.

CENTER is an array of window center strobe setting values for individual heads (or zones per head). CENTER is loaded with default values upon power up to read various commands for disk drive configuration. Then, CENTER is loaded with a nominal value if coarse tuning is to be performed. If not, CENTER is loaded with stored values of window centers which were stored during previous execution.

RDCNT is a variable representing the maximum number of reads to be performed on a test track.

MAXE is a variable representing the maximum number of errors to allow on a test track.

TESTTRACK is a prestored location on the magnetic disk where the read only test for window center should be performed.

CONFID is a variable representing the confidence level threshold for moving strobe settings (i.e. 85% in this preferred embodiment).

PASSCNT is a variable indicating the number of adjustments made to window center settings-for statistical printout.

STRB is a strobe setting.

NBOUND is the negative boundary of strobe settings during coarse tuning.

PBOUND is the positive boundary of strobe settings during coarse tuning.

ERCNT is the number of errors logged for read operations.

STOPHEAD is a variable indicating coarse tuning has been performed for the last head.

READTK is a variable indicating the number of reads per test track which have been executed.

COUNT is a variable indicating number of errors.

MAX(r,l) is a variable indicating the maximum number of errors of the majority type which have been logged.

SIGN is a variable indicating the direction of correction to be made in adjusting strobe settings.

N is total number Shift Right and Shift Left errors logged.

WIDEBAND is a term used to designate whether the coarse tuning operation should be performed to accelerate window center seeking. It should be noted that the optimum window center setting can be determined using only the fine tuning technique. However, by using the coarse tuning operation, the window center setting can be set more quickly. 0 The coarse tuning routine is initialized and the array CENTER is loaded with nominal values. It should be noted that CENTER could also be loaded with values which were stored during a previous coarse tuning routine. These operations are indicated in blocks 60 and 62. Next, controller 42 determines whether the coarse tuning operation is to be performed. This is indicated in block 64. If not, controller 42 skips to the fine tuning routine. However, if WIDEBAND equals 1 (the coarse tuning technique is to be performed), then controller 42 causes actuator 1B to perform a seek operation for the first head in magnetic disk 10, thereby moving the first head (e.g. head 0) to the test track. This is indicated by block 66.

Next, controller 42 determines whether the particular cylinder over which the selected transducer is flying is a read/writable cylinder. If not, controller 42 determines whether there are any test tracks left to test. If so, the next test track is determined and controller 42 causes actuator 18 to seek to that test track. This is indicated by blocks 68, 70 and 72. If the test track over which the selected transducer is flying is in a read/writable cylinder, then controller 42 sets the strobe value to -15% and sets the negative boundary to -16% and the positive boundary to positive 16%. This is indicated by block 74.

Next, the read errors in reading from the selected test track are read and logged. This is indicated by block 76. Then, the total error count is compared to the maximum number of errors allowed per test track. If the total error count is less than the maximum number of errors allowed per test track, then the number of read operations which have been performed for that test track is compared to the maximum number of reads allowed per test track. This is indicated in block 80. If more read operations are required, then controller 42 repeats the steps taken in blocks 76 and 78. This is indicated in block 80.

Next, controller 42 determines whether the negative boundary is greater than -16%. If not, the negative boundary is set to the current strobe value and the strobe value is incremented. This is indicated in blocks 82, 84 and 86. If the negative boundary is greater than -16%, then the strobe value is simply incremented. This in indicated in block 88.

The strobe value is then examined to see whether it is greater than 15. If not, controller 42 returns to operation block 76. If it is, then the CENTER array is set to the center point between the current positive and negative boundaries. This is indicated in blocks 90 and 92. Then, controller 42 determines whether coarse tuning must be performed for any other heads or test tracks in disk drive 10. If so, controller 42 repeats the appropriate steps. If not, controller 42 updates the CENTER array appropriately and moves on to the fine tuning operation. This is indicated in blocks 94, 96, 98, 100 and 102.

In, in block 78, the total number of errors counted during the read operations is greater than the maximum number of errors allowed on the given test track, then the absolute value of the negative boundary is compared with the absolute value of the positive boundary. This is indicated in block 104. If, however, the answer determined in block 78 is NO, then the strobe is incremented and controller 42 determines whether the present value of the strobe is greater than 15. If not, controller 42 returns to block 76. If so, controller 42 goes to block 94.

If in block 104, the absolute value of the negative boundary is less than the absolute value of the positive boundary, then the positive boundary is set to the current strobe value and the CENTER array is set to the midpoint between the positive and negative boundaries. This is indicated in blocks 110 and 112. Then, controller 42 determines whether any other heads or test tracks must be coarsely tuned. If not, controller 42 moves on to the fine tuning operation.

In the fine tuning technique, the strobe value for the National Semiconductor 8459 integrated circuit chip will only be moved in the smallest increments allowed by the integrated circuit chip. Therefore, all that must be determined is whether the confidence level is above the confidence level threshold, and which direction the window center setting must be adjusted.

Once the fine tuning technique is initialized, controller 42 can either wait for a read command, as indicated in block 114, or can exercise the read operation when the drive is idle. Controller 42 exercises the read command by continuously reading test tracks and updating the window center settings. Where power requirements are not critical, the latter technique may be used. However, in this preferred embodiment, block 116 indicates that controller 42 waits for a read command before reading and logging errors.

Once errors are logged, controller 42 determines which type of (i.e. either Shift Left or Shift Right) errors are logged. This is determined, as discussed earlier, by analyzing the read error characteristics as they are read in the 2,7 encoded bit stream. This is indicated in block 118. Next, controller 42 determines the maximum number of Shift Right or Shift Left errors detected and determines the confidence level using the number of errors of the majority type determined. This is indicated in blocks 120 and 122. If the confidence level is less than the confidence level threshold, then controller 42 returns to block 116 to log more error samples. This is indicated in block 124.

However, if the confidence level is greater than or equal to the confidence level threshold, then a window center setting adjustment is required and controller 42 determines the direction in which the adjustment must be made. This is indicated in block 126. Then, the strobe value stored in the CENTER array is updated by moving the appropriate strobe value one unit in the direction determined in block 126. This is indicated in block 128. Next, the error counters are cleared and the CENTER array is written to the magnetic disk so the strobe values stored in the CENTER array can be retrieved as default settings during the next read operation. This is shown in blocks -30, 132 and 134.

FIG. 9 is a graph showing the number of read errors plotted against window center strobe settings. In some cases, it may happen that the integrated circuit chip used to set the window center settings, such as the National Semiconductor 8459 integrated circuit chip, may have center settings which can only be adjusted in increments. Those increments may be too large to hit the exact optimum window center setting, such as 150 shown in FIG. 9. Instead, the two closest strobe settings may be strobe setting S1 and strobe setting S2 which shift the window center around point 150 but never actually reach exactly the optimum window center setting 150. In such a case, instability may exist in disk drive 10 in that controller 42 would continually oscillate the strobe setting in register 44 between points S1 and S2.

Therefore, in this preferred embodiment, controller 42 stores the strobe setting determined during a previous fine tuning operation and also stores the error rate associated with that strobe setting. Then, when controller 42 determines that an adjustment is required, controller 42 determines whether the adjustment will be back to the previous strobe setting. If so, controller 42 determines whether the number of read errors associated with that previous strobe setting are greater or less than the number of read errors associated with the present strobe setting. If they are less, controller 42 makes the desired adjustment If not, controller 42 disregards the adjustments and leaves the strobe settings at the present value.

The coarse and fine tuning operations can be performed for each transducer in magnetic disk drive 10. Further, such routines can be performed for a plurality of zones on each magnetic disk 12 in disk drive 10. Therefore, very accurate window center settings can be adjusted for each transducer, magnetic disk and zone in disk drive 10. Furthermore, by using the types of errors detected to indicate the direction of needed correction, a closed loop feedback system is used to adaptively adjust the window center settings.

By continuously updating the window center settings for each head or disk, or zone in disk drive 10, controller 42 can adjust for signal time shifting in read signal 32 caused by temperature variation, aging, voltage variation and timing variations in electronic circuits in disk drive 10. This improves the error rate during read operations and also increases manufacturing yield of disk drives.

In addition, if new printed circuit boards or components must be added to disk drive 12, any timing variations caused by the new components or printed circuit boards is almost immediately adjusted for by using the technique of the present invention.

Although the present invention has been described with reference to preferred embodiments, workers skilled in the art will recognize that changes may be made in form and detail without departing from the spirit and scope of the invention.