High efficiency disk format and synchronization system

FIELD OF INVENTION

0001The invention relates generally to magnetic disk sector formatting and data synchronization.

BACKGROUND OF THE INVENTION

0002Digital data processing systems typically include data storage devices, for example, multi-disk disk drives. Data is recorded on the disks in concentric tracks. The tracks are divided into sectors, and each sector is further divided into a header section and a data section. Data is recorded in (written) or retrieved from (read) the data portions of the sectors while address information, that is, disk surface number, track number and sector number, is stored in the header sections.

0003Data is read from or written to the disks using read/write heads. Each head is associated with a particular disk surface or portion of a disk surface. As the disks rotate under the heads, data is, for example, read from a particular sector when the associated head is over the data portion of that sector.

0004To begin a read or write operation the appropriate head is moved to the track containing the desired sector. Unless the head was used in the preceding read or write operation, or it has been kept in synchronization by appropriate timers, for example during idle times, the head must then be synchronized to the disk so that the sector addresses rotating under it can be read. Once the head is synchronized, it locates the selected sector by reading the sector addresses and then it performs the read or write operation. Head switching, which may involve switching between heads on a common actuator arm or switching between heads on different arms, occurs quite often during the normal operation of the disk drive and thus synchronization is performed often also. The speed with which the heads are accurately synchronized to the disks affects the speed with which data can be transferred to or from the drive.

0005Using prior technology, timing information for synchronizing is recorded either on an extra servo surface or as an additional information zone between sectors, referred to as an embedded-servo zone. If the disk drive is one which uses fixed block architecture and data surface only embedded-servo information, certain information must be contained within the space of each sector and related servo zone to enable the system to operate in real time, that is, to perform certain functions within the time it takes for the sector to rotate under the head. For example, sector timing must be established, and bit, data symbol and data word synchronization must be determined using the recorded servo and synchronization information. Also, if alternate sectors can be used to replace corrupted, or bad, sectors flags indicating which sectors are good and which are bad must be read and interpreted. Additionally, each sector contains data that must be read and validated, and if necessary corrected, using recorded data redundancy symbols. It is desirable to perform all of these functions with minimum disk overhead, that is, with minimum consumption of extra recording space, leaving the disk space free for data.

0006Prior technology embedded-servo disk drives typically use three-zone formats. The three zones are a servo zone, containing servo position information which is permanently written at the factory when the disks are formatted, and two sector zones, a header zone containing address and bad sector information which can be re-written during reformatting to indicate which sectors have degraded from good to bad, and a data zone containing data and data redundancy symbols which can be easily written and re-written by the user.

0007Using three zones requires the disk drive to synchronize three times, once per zone, in order to perform a read or write operation in a sector. Thus each zone must contain synchronization and timing information. This causes increased overhead, resulting in a loss of recording efficiency. Thus it is desirable to reduce the number of zones while retaining all the functionality of the three-zone architecture. Also, as users demand disk drives with more and more data storage capacity, it is more and more desirable to use less disk space for synchronization.

0008Prior technology also requires significant dedicated timing hardware, for example, timing logic or analog electronics is required to interpret the recorded timing information and synchronize the heads. The servo-zone typically contains a long synchronization burst written in a frequency which is unlikely to occur in the data, servo information for track following, and several information bits for track identification. Thus the embedded servo system must use analog filters to locate the synchronization burst and special timing and control logic to find the servo information, as well as to synchronize the head to the servo zone. This additional hardware increases the cost of the system. Thus it is desirable to perform synchronization using much of the same hardware as the normal read and write operations, and in particular using mostly digital logic. It is also desirable to perform synchronization quickly, even if the heads are not in a known position with respect to the servo information that would be used to align the data head on the track.

0009WO83/01858 discloses a magnetic disk storage system in which each sector provides servo data which provides track position data to enable radial positioning of the head. The document does not discuss sector identification.

0010EP-0 166 373, according to which claim 1 has been put in the two-part form, describes a disk drive system in which the disk surface is formatted in sectors with each sector having an address portion and a data portion.

SUMMARY OF THE INVENTION

0011The invention is a disk drive system in accordance with claim 1, and which may provide all the required functions within each sector including: track error positioning information, sector bit, data symbol and data word synchronization, cylinder, track and sector addresses and address validation, good/bad sector flags, and data and data redundancy. All of these functions are provided within two zones to provide high data efficiency, that is, low overhead per sector. The inventive system synchronizes a controller state machine to the real time relation between the head and the disk by using hardware which is also used in read/write operations, thus reducing the required hardware.

0012Basically each sector is formatted with a header and a data section. The header section includes a preamble, a synchronization character and an address field, as well as servo information for track following. The data section of each sector includes a data preamble, a data synchronization character, a bad sector bit map, and the data and data redundancy information. The header section of at least one sector in a track includes a short DC-erase field, that is, a transitionless segment.

0013The inventive system first looks for a DC-erase field, using system hardware that normally is used in the read/write operations. Once it detects a DC-erase field it looks for the header preamble and synchronization character. If each of these are detected within predetermined times, a timer is set which allows the system to time to the address field and then to the servo information contained in the header. If the address field contains a valid sector address, the system assumes it is synchronized and the servo information is then used to center the head on the track. Thereafter, the desired sector is located by reading the addresses of the succeeding sectors. If any of the header sub-sections following the DC-erase field are not detected correctly within the predetermined times, the system looks for another DC-erase field and then continues the synchronization process. In this way the system is able to find the beginning of a sector and to synchronize using only the short DC-erase field and other header information.

0014The bad sector bit map is used to identify all bad sectors within a track or a section of a track. The bit map is redundant and located within each data zone within the track or section of the track to which it pertains. Thus the loss of one or more sectors is not likely to cause loss of knowledge about which sectors are bad.

DESCRIPTION OF THE DRAWINGS

0015The invention will be pointed out with particularity in the appended claims. The above and other advantages of the invention may be better understood by referring to the following description taken in conjunction with the accompanying drawings, in which: <ul id="ul0001" list-style="none" compact="compact"><li>Figure 1 is a functional block diagram of a system built in accordance with the invention;</li><li>Figure 2 illustrates the format of a disk recorded for use with the invention;</li><li>Figure 3A is a flow chart of the operations of a system built in accordance with the invention, including the steps of looking for a DC-erase field, looking for a preamble, looking for a synchronization character and setting the data channel gain;</li><li>Figure 3B is a more detailed flow chart of the steps used in looking for the DC-erase field;</li><li>Figure 3C is another detailed flow chart, showing the steps used in looking for the preamble;</li><li>Figure 3D is a further flow chart, illustrating the steps used in looking for the synchronization character;</li><li>Figure 3E is a flow chart of the steps used for setting the gain of the data channel;</li><li>Figure 4 is a functional block diagram of a controller shown in Figure 1; and</li><li>Figure 5 is a functional block diagram of a data separator shown in Figure 1.</li></ul>

DETAILED DESCRIPTION

0016With reference to Figure 1, a system 8 is shown which may be used for synchronization and reading or writing data to or from magnetic disks (not shown). Data is, for example, read from a disk by read/write heads (not shown) and sent over line 12 to a read channel consisting of an automatic gain control (AGC) 20, a data separator 14 which "converts" peaks in the signal read from the disk to data, and a phase-locked loop (PLL) 16. During a read operation the AGC 20 sets the gain of the data channel such that the system 8 can distinguish data signals from noise. The data signals typically consist of relative positive and negative peaks within a predetermined amplitude range while noise signals typically consist of peaks below that range. Thus the gain of the AGC 20 is set either by feed-back from the signal or a setting from an external source, such as controller 22, to a value which will amplify the data signal enough to "read" it but which will not amplify the noise signal to bring its peaks, which fall below the data signal range, within that range. If the AGC 20 gain is too large, noise peaks may be misinterpreted as data.

0017The PLL 16 sets the timing of signal sampling, that is, the times at which the analog signal from the read/write heads is sampled by the data separator 14. These samples are then turned into digital symbols representing the data written to the disk. The digital symbols are thereafter demodulated to binary data bits in a demodulator (not shown). The drive also contains a write channel (not shown) for writing data to the disk.

0018When data is to be read or written to a disk the system 8, using a controller 22 and a microprocessor 24, selects the head (i.e., the disk surface) associated with the sector involved in the operation and moves the head to the track containing the sector. If the selected head is not the head used in the preceding read or write operation on this disk surface, that is, if a head switch occurs, the timer and the clocks associated with the controller state machine operation must be re-synchronized to the recorded data on the disk. This is often referred to as synchronizing the head.

0019The head is synchronized by first finding the beginning of a sector and using the sector header information to set both the gain of the AGC 20 and the phase of the PLL 16 to values appropriate for the system 8 to read and interpret the header information. The sector to be read or written is then located by reading the addresses recorded in consecutive sectors on the track. Once the appropriate sector is found, the read or write operation is performed in the data section of that sector.

0020With reference to Figure 2, a disk 28 is recorded with a radial contiguous format, that is, with portions of the header section of each sector recorded such that the start of each bit in these portions is radially contiguous with the corresponding bits in the adjacent tracks. For example, the bits in the preamble of header section 29B in track 29A are radially contiguous with the corresponding bits in the preamble of header section 29D in track 29C. In the preferred embodiment the sector addresses are Gray coded and thus there is only a one-bit difference between the addresses of contiguous sectors on adjacent tracks.

0021At least one sector, and preferably several or all sectors on a track, are formatted with a header section that includes a DC-erase field 30 at the beginning of the sector, followed by a preamble 32, a synchronization character 34, an address field 36 and a servo field 38. The DC-erase field 30 is a field without any magnetic transitions, that is, a signal recorded on the disk without magnetic flux reversals. It has a length in the preferred embodiment corresponding to the length of between three and eight bytes of recorded data.

0022The preamble 32 is a pattern of ONES and ZEROS containing a predetermined number of ONES, preferably the maximum number of ONES, that is, flux reversals, for a given number of recorded bits. The data recorded on the disk are typically encoded to aid in the demodulation of the signal when it is later read from the disk. Thus the code may prohibit ONES in consecutive bit locations. In the preferred embodiment the preamble 32 is the pattern 1010...101. The preamble 32 is used to set the gain of AGC 20 and to lock the phase of the PLL 16 to the bit frequency and phase of the header.

0023The synchronization character 34 is a pattern of ONES and ZEROS which allows the controller 22 to determine the boundary of the decode group, that is, the number of bit cells on the disk used for recording a given number of binary data bits, and for locating the start of the byte boundaries. For example, some codes encode two data-bits into three bits for recording, again to aid in the demodulation of the read-back signal, and thus the start of each three-bit decode group must be determined. The synchronization character 34 also allows the controller 22 to determine the boundaries of bytes of information recorded in the sector. At the end of the synchronization character a timer 22A (Figure 4) is started which is used to time to the remaining header fields and to the next sector header.

0024The synchronization character used in the preferred embodiment is, s = <u>1</u>0<u>1</u>0<u>1</u>0<u>0</u>0<u>1</u>0<u>0</u>0<u>1</u>0<u>1</u>0<u>0</u>0<u>1</u>0<u>1</u>0<u>0</u>1<u>0</u>1<u>0</u>0<u>1</u>0<u>0</u>1. The system, using a clock from the PLL 16 which is one-half the bit frequency, as described in US-4908721 entitled A Method of Handling Data From a Storage Device, filed May 12, 1988 which has a common assignee, examines only every other bit. Thus the portion of the synchronization character which is actually used to synchronize is s' = 1110101101100010, the under-lined portions of s. The bits which are interspersed with s' substantially eliminate inter-bit interference caused by, for example, the magnetic flux reversals representing two ONES interacting with each other. Such interference may cause peak shifts, resulting in the PLL 16 taking longer to lock to the correct bit-tracking phase or increased error in data seperating this pattern.

0025Alternatively, the preamble 32 can be used to determine the decode group. If, for example, the preamble is a pattern 100100...1 it can be used to determine the boundary of the three-bit decode group. The synchronization character 34 would then be used to determine the byte boundaries.

0026Following the synchronization character 34 are the address field 36 and the servo field 38. The address field 36 contains at least one copy of the sector address, in the preferred embodiment it contains four copies. The servo field 38 contains information which is used by the system 8 to center the head on the track.

0027The data section contains a data preamble 40 and a data synchronization character 42 which are identical with the preamble 32 and synchronization character 34 in the header. Following the data synchronization character is a bad sector bit-map which contains information about the good and bad sectors in the track, or a portions of the track. The bit-map is redundant in that it is recorded in multiple sectors in the track. The bit-maps can be re-written and thus up-dated, that is, as sectors degrade from good to bad the maps can be revised to reflect the degradation. Recording the map in multiple sectors serves to protect the map from loss due to the degradation of any one sector or group of sectors. The bit-map is followed by the data and data redundancy symbols.

0028The radial contiguous formatting of the disk 28 allows the preamble 32 and the synchronization character 34 to be located, regardless of where the associated head is placed with respect to the tracks. Without radial contiguous formatting, if, for example, the head straddles two tracks, the data signal in one track may be read as canceling the data signal in the other track, that is, a segment in which negative peaks in one track correspond to positive peaks in the other track may be read as a transitionless signal. Thus a DC-erase may be detected in the data section of the sector. If the acquire mode of the AGC 20 is then enabled, that is, during the detection of a false DC-erase, the acquired gain may cause the read channel to saturate and the system would not recover in time to "see" the real data. However, by recording the preambles 32, the synchronization characters 34 and the address fields 36 in a radial contiguous format bit canceling due to an off-center head is eliminated, and thus, each of these fields can be located once a true DC-erase is detected. Gray coding the addresses, also, allows the address to be determined to within two tracks, even if the head is straddling two tracks.

0029With reference to Figure 3A, the system 8 operates in one of two modes during synchronization. If the system has not previously synchronized to the track, the appropriate gain of the read channel for data retrieval from the track is not known. Accordingly, the system first determines the approximate gain (Figure 3E) before it begins the synchronization process at step 100. However, if the system has previously synchronized to the selected track, the approximate gain for that track has already been determined and stored in an appropriate memory (not shown). The system 8 retrieves that value and adjusts the gain of the AGC 20 to this fixed value. It then begins synchronization (step 102).

0030The synchronization process begins with a search for a DC-erase field 30 (step 104). The search operation is discussed in more detail with reference to Figure 3B below.

0031With further reference to Figure 3A, when a transitionless signal segment is detected, it is assumed that it is a DC-erase field (step 106). Thus the start of a sector is assumed found and a "search" for the preamble is started (step 108). It is likely that false DC-erase fields will be frequently found due to an off-center head's "canceling" of data signals. However, if the start of a sector has in fact been detected, the preamble will be detected within a predetermined time thereafter. The preamble will be detected regardless of whether or not the head is centered on the track due to the radially contiguous format of the disk.

0032If the DC-erase detected is not the start of a sector, the preamble will not be detected within the predetermined time and the search for the DC-erase field is begun again (steps 114-115). The preamble search is described in more detail with reference to Figure 3C below. DC-erase fields are continuously searched for during the preamble search. If another DC-erase field is detected at any time during the preamble search, the search is re-started (Step 166, Figure 3C).

0033Again referring to Figure 3A, once the preamble is detected, a search for the synchronization character is begun (step 116). The synchronization character will be found within a predetermined time if the previously detected DC-erase field and preamble denote the start of a sector. Sensing of the synchronization character allows the controller 22 to set a sector timer 22A (Figure 4). The timer 22A is then used to enable controller operations on specific sector segments, for example, the end of the address field and then the start of the servo field, and also to the header section of the next sector. If the synchronization character is not sensed within the specified time, an indication of false detection of the DC-erase field and the preamble, the DC-erase field search is again started (step 119A). The synchronization character search is described in more detail with reference to Figure 3D below. The DC-erase field search also continues throughout the synchronization character search. If a DC-erase is found at any time, the synchronization process is re-started (step 174A, Figure 3D).

0034With further reference to Figure 3A, when the synchronization character is sensed in the specified time interval, the next bits read from the sector are interpreted as the sector address field (step 120). The validity of the address retrieved from this field is then checked (step 122). In the preferred embodiment the address field contains a number of copies of the address, for example, four copies, and if some minimum number of copies match each other, for example two copies, the address is deemed valid.

0035If the address is valid, the controller 22 (Figure 4) uses the sector timer 22A to time and enable the sampling of the servo field 38 and this sampling may then be used to move the head to the center of the track. It then uses the timer 22A to time to the next sector where it expects to find the sector header and enables the acquire modes of both the PLL 16 and AGC 20 in the preamble to synchronize to this next sector. It then reads the sector address field (steps 124-128) and determines the validity of this address by comparing a minimum number of copies to the expected sector address (step 130). Thus if, for example, the first sector found is sector 10 the next sector should be sector 11. If the sector address matches the expected address the head is assumed to be synchronized. If the address is an invalid address or it does not match the expected address, the system times to the next sector and reads that address to determine which sector is then rotating under the head. If the location of the head cannot be determined in a reasonable number of sectors the system re-starts the DC-erase field search (steps 130-133).

0036The DC-erase field, preamble and synchronization character searches are all performed by the controller 22 (Figure 4). The controller 22 receives from the data separator 14 samples of the signal read from the disk. The controller 22 examines these samples as a bit-stream and determines if any segment of the signal match the signal pattern for which the controller is then looking. For example, if the controller is conducting the DC-erase field search it will examine the signal segments for a segment without transitions.

0037The system controller 22 can vary the number of bit-matches required to detect a synchronization character and/or the number of symbol-matches required to signify a valid address, thus allowing for more or less severe checking. For example, during the factory checking of a drive the controller 22 may require perfect matching, while much less than perfect matching may be required during system error recovery procedures.

0038With reference to Figure 3B, the DC-erase field search is begun by loading all ONES into a correlator, or shift register, 23 in the controller 22 (Figure 4). Thereafter signal samples from the data separator 14 are shifted into the correlator 23 and the ONES are correspondingly shifted out. Each time a sample is shifted into the correlator 23 the contents are examined to determine if the signal segment represented by the samples contains fewer than some predetermined number of ONES, indicating, with some allowance for error, a transitionless signal segment (step 144). For example, if a forty-bit signal segment is examined, a DC-erase field may be indicated if four or fewer ONES are contained in the correlator 23.

0039After detecting a DC-erase field, the controller 22 starts the preamble search (step 108 Figure 3A). With reference to Figure 3C, the controller starts the timer 22A, and initializes the correlator 23 to all ZEROS (step 148). Signal samples are then shifted into the correlator from the data separator 14 and the contents of the correlator are repeatedly compared with the preamble pattern, that is, with a "pattern" of a minimum number of ONES in a predetermined number of bits, where ONES represent transitions, although not necessarily in bit alignment (step 150). The preferred embodiment looks for four ONES in a twelve-bit signal segment. If a signal segment has a sufficient number of ONES, allowing for some error, the preamble is presumed to have been detected (step 150-152). If, however, the timer 22A reaches time T₁ before a preamble is detected, the synchronization process is re-started with the DC-erase field search (step 170).

0040While the controller searches for the preamble pattern it also checks the signal segment to determine if it contains another DC-erase field by tracking the number of ONES in the segment (step 150). Another DC-erase field is deemed to have been detected if the segment contains fewer than a predetermined number of ONES, where the predetermined number may or may not be the same as that used in the DC-erase field search (step 166). In the preferred embodiment a stricter test is used. For example, if fewer than four ONES are found in any twelve-sample signal segment a DC-erase field is presumed detected and the synchronization process is re-started.

0041Once the preamble is detected (step 152), acquire modes of the AGC 20 and the PLL 16 are enabled. This allows the AGC 20 to determine the appropriate read channel gain for the header section, by feeding back the signal magnitude to set the gain to a constant value, and the PLL 16 to lock to the header bit-frequency and phase. The time to the end of the preamble is sufficient for the AGC 20 and PLL 16, respectively, to lock the appropriate gain and phase (step 154). Thereafter the gain of AGC 20 is held for the remainder of the header and the PLL 16 tracks the frequency and phase of the header data-bits until the end of the address field 36. At this point the AGC gain may be stored for future use by the microprocessor 24. At the end of the field, PLL 16 is locked to the phase of the reference clock 18 (Figure 1) because the servo field 38, which follows, does not contain information recorded with the same frequency and phase as the data-bits. Timing diagrams for the AGC 20 and the PLL 16 are shown in Figure 2.

0042The operation of the AGC 20 is described in more detail in US-4849711 entitled Automatic Gain Control System, filed September 4, 1987. That application, which has a common assignee with this application is incorporated herein by reference. The operation of the PLL 16 is described in more detail in US-4908841 entitled Data Decoding Circuit including Phase-Locked Loop Timing, filed October 30, 1987. These patents, which also have a common assignee with this application, are also incorporated herein by reference.

0043At the time indicating the end of the preamble, the synchronization character search is started (step 116, Figure 3A). With reference to Figure 3D, this search is begun by initializing the correlator 23 (Figure 4) with the synchronization character and shifting signal samples from the data separator 14 into the correlator 23 (step 172). The signal segments are then examined for a match with the synchronization character pattern. If the pattern is found before time T₁ (Figure 2), a synchronization character is deemed detected (steps 174, 176). At the end of the synchronization character the timer 22A is re-started to time to the remaining header fields and to the header section of the next sector (step 178). The timing diagram for the timer is shown in Figure 2. If, however, the synchronization character is not detected before time T₁, the synchronization process is re-started with a DC-erase field search (step 175).

0044If the approximate AGC 20 gain setting for the data track to which the head is being synchronized is not known at the start of the synchronization process, a gain value is first determined which allows the gain to be pre-set and the DC-erase field to be detected and the same steps are followed thereafter to complete synchronization (step 100, Figure 3A).

0045With reference to Figure 3E, the gain of the AGC 20 is first set to a maximum value (step 182). The timer 22A (Figure 1) is then started and a search is made for the DC-erase field (step 184). The search uses the same steps as the DC-erase field search described above. If a start of sector DC-erase field is not found before time T_A, the gain of the AGC 20 is pre-set to another value, the timer 22A is re-started and a search for the DC-erase field is again started (steps 194-198).

0046It is unlikely that the DC-erase field will be found when the gain of the AGC 20 is set at the maximum value because noise signals associated with the field will be amplified to look like data transitions and thus a transitionless signal will not be found. This searching and gain adjusting process continues until either the DC-erase field is found or system recovery procedures are implemented, for example, if after a predetermined number of tries no DC-erase fields can be found on the track (steps 198-204).

0047At the end of a successful synchronization process the gain of AGC 20 is stored in an appropriate memory (not shown) so that it may be used the next time the head is switched to that track. Pre-setting the gain of the AGC 20 allows the system to simultaneously enable the acquire modes of the AGC 20 and the PLL 16 without having the PLL 16 try to lock to the phase of amplified noise signals, for example, if the gain of the AGC 20 was too high. By presetting the AGC gain to the appropriate value, the normal AGC action does not have to correct very far, and it can thus acquire the appropriate gain within a relatively short time. This allows the PLL 16 to operate on almost all of the preamble 32 to acquire phase-lock and the preamble can therefore be relatively short in length.

0048During the DC-erase field search and portions of the preamble search the system has no knowledge of the bit frequency of the sector header information. Thus the data separator 14 and the PLL 16 may have to sample the signal read from the disk even if the frequency and phase to which the PLL 16 is locked is not the data-bit frequency and phase. With reference to Figure 5, the data separator 14 includes a comparator 300 and flip-flops 302 and 304. The flip-flops 302,304 are clocked by the PLL 16 which is locked to the reference clock 18 (Figure 1). The signal read from the disk is input to the comparator 300. The comparator output is asserted when the signal peak is within a predetermined range, indicating a flux reversal on the disk, which is interpreted as a ONE. The asserted output of the comparator is then input to flip-flop 302 and appears at its output.

0049The output of flip-flop 302 is input to flip-flop 304 when it is clocked by the PLL 16. Flip-flop 302 is then cleared. When flip-flop 304 is next clocked, the contents of the flip-flop are stored in the correlator 23 as a ONE (Figure 4). If the signal peaks do not come within the predetermined amplitude range, the output of flip-flop 302 remains de-asserted and ZEROS are input to the correlator 23.

0050The foregoing description has been limited to a specific embodiment of this invention. It will be apparent, however, that variations and modifications may be made to the invention, with the attainment of some or all of the advantages of the invention. Therefore, it is the object of the appended claims to cover all such variations and modifications as come within the true scope of the invention as defined in the appended claims.