Finite field based short error propagation modulation codes专利检索-..脉冲调制记录或重现的（脉冲编码调制记录的入G11B20/10）专利检索查询-专利查询网

IPRDB

API 数据接口

专利申请

使用指引 chat嘟嘟

会员体验

联系我们

交流群

现在联系顾问~

Finite field based short error propagation modulation codes
申请号	US11016283	申请日	2004-12-17	公开(公告)号	US07486456B2	公开(公告)日	2009-02-03
申请人	William G. Bliss; Razmik Karabed;			发明人	William G. Bliss; Razmik Karabed;
摘要	The invention relates to a data modulation method applicable to make data streams tend to have desired properties, useful for clock recovery, making signals more distinguishable, or enforcing run-length conditions. A stream of input data and a corresponding stream of output data are grouped into elements of a finite field. Input elements of said input data are modified by a transform generating output elements of the output data, such that a current output element is a linear combination of a current input element and at least one previous output element. A multiplier applied to at least one previous output element is a non-zero and non-unity element of the finite field. A set of initial conditions inherent to the transform, is selected such that the output elements resulting from the transform tend to have the desired property.
权利要求	The invention claimed is:1. A data modulation method comprising the following steps:grouping a stream of input data and a corresponding stream of output data into elements of a finite field;applying to input elements of said input data a transform generating output elements of said output data, such that a current output element is a linear combination of a current input element and at least one previous output element, wherein a multiplier applied to the at least one previous output element is a non-zero and non-unity element of the finite field; andselecting a set of initial conditions inherent to the transform, such that the output elements resulting from the transform tend to have a desired property.2. The data modulation method of claim 1, further comprising the steps of:calculating intermediate elements by applying said transform to the input elements with a set of initial conditions of value zero;calculating the effect of the selected set of initial conditions by applying said transform to input elements having value zero and the selected set of initial conditions; andadding said effect to the intermediate elements to obtain the output elements.3. The data modulation method of claim 1, wherein the step of selecting the initial conditions comprises the steps of:defining distinct sets of initial conditions, each set having a single non-zero element at a distinct position; andselecting each non-zero element of the sets of initial conditions such that the output elements tend to have a respective property.4. The data modulation method of claim 1, wherein the transform is such that a current output element is the sum of a current input element and μ times the preceding output element, where μ is a non-zero and non-unity element of the finite field.5. The data modulation method of claim 1, wherein the initial conditions are selected such that each of the output elements is non-zero.6. The data modulation method of claim 1, wherein the transform operates on groups of 14 input elements to produce groups of 15 output elements, and the finite field has 16 elements.7. The data modulation method of claim 1, wherein the initial conditions are selected such that said output data tends to have as many 1s as 0s in the bits constituting the output data.8. The data modulation method of claim 1, wherein the transform operates on groups of six input elements to produce groups of seven output elements, and the finite field has four elements.9. The data modulation method of claim 1, applied in a disk drive for writing data on a disk.10. An inverse data modulation method comprising the following steps:grouping a stream of input data and a corresponding stream of output data into elements of a finite field; andapplying to the input elements of said input data a transform generating output elements of said output data, such that a current output element is a linear combination of a current input element and at least one previous input element, wherein a multiplier applied to a previous output element is a non-zero and non-unity element of the finite field.11. The inverse data modulation method of claim 10, wherein the transform is such that the current output element is the sum of the current input element and μ times a preceding output element, where μ is a non-zero and non-unity element of the finite field.12. The inverse data modulation method of claim 10, applied in a disk drive for reading data from a disk.13. A disk drive comprising:a disk drive assembly including a read/write head for reading and writing data on a disk, actuators for displacing the read/write head, and a motor for rotating the disk;a servo control circuit for controlling the position of the read/write head with respect to the disk;a system processor programmed to control the position of the read/write head through the servo control circuit in response to read/write commands received from a host; anda read/write subsystem responsive to the system processor, operable to:group a stream of input data and a corresponding stream of output data into elements of a finite field;apply to input elements of said input data a transform generating output elements of said output data, such that a current output element is a linear combination of a current input element and at least one previous output element, wherein a multiplier applied to the at least one previous output element is a non-zero and non-unity element of the finite field; andselect a set of initial conditions inherent to the transform, such that the output elements resulting from the transform tend to have a desired property.
说明书全文	BACKGROUND OF THE INVENTION 1. Field of Invention The present invention relates to modulation of data transfer signals, for instance in reading from, and writing to a magnetic medium, such as a hard disk drive. The invention more specifically relates to modulation intended to make the signal properties meet specific criteria, for instance enforcing run-length limited conditions, making signals more distinguishable (increasing “distance”), and providing clock recovery information. 2. Relevant Background FIG. 1 is a schematic representation of a data transfer chain 8, such as used in a hard disk drive. A sequence of user data to be written on a hard-disk is input to an Error Correcting Coding (ECC) circuit 10. An encoder 12, implementing the desired modulation, receives k-bit blocks of output b from the ECC 10 and produces (k+r)-bit blocks c. The (k+r)-bit blocks c are referred to as codewords. The encoder 12 outputs the codeword c to a 1/(1+D²) filter or precoder 14. The term “D” designates a one bit delay and “+” designates the bitwise exclusive OR operator. Thus, the i-th bit x_iin a codeword output c by precoder 14 is expressed as x_i=c_i+x_i-2. Such an operation needs two initial conditions x₋₁and x₀to be set for x, for instance (0, 0). Codeword x passes in a channel 16 through one or more channel filters 18. The channel, which is where the data is written to the hard-disk and read back from the hard-disk, is typically corrupted by additive noise n, such that the received sequence r is defined by r=z+n, where z is the output of filters 18. Based on the received sequence r, a Viterbi detector 20, for example, generates a detected sequence x^, which is a reproduction of the input x to the channel filters 18. Next, bits x^ are filtered by a filter 22 which performs the function (1+D²) that is an inverse of the function performed by precoder 14, and generates g. The output g of the filter 22 is decoded by a decoder 24 to produce a decoded sequence d, which is a reproduction of the ECC output sequence b. An ECC decoder 26 receives the output sequence d and reestablishes the user input to ECC coder 10. As mentioned above, the codewords used in the system have k+r bits, whereas the corresponding user data blocks have a lower number of bits k, whereby there is an efficiency loss. The efficiency of the encoder is called “rate” and it is defined as k/(k+r). Known modulation techniques implemented by the encoder 12 strive to increase the rate and impose desired properties on the codeword. Often this makes each bit of a codeword depend on every bit of the incoming data block. Such techniques have the drawback of increasing “error propagation”—often one corrupted bit in a codeword would cause the loss of most of the bits in the resulting data block. US published application 20040059980, incorporated herein by reference, discloses a modulation method for use in encoder 12, which has short error propagation while imposing desirable properties on the codewords. The modulation has the following generic transform: $\frac{1}{f_{0} + f_{1} D + f_{2} D^{2} + f_{3} D^{3} + \dots + f_{r} D^{r}}$ where D is a one-bit delay and f=(f₀, f₁, f₂. . . f_r) is a set of constant binary values, with f₀=f_r=1, characterizing the modulation scheme. In other words, given an i-th bit b_iof a user data block, the i-th bit a_iof the resulting codeword a is defined as: a_i=b_i+f₁a_{i -1}+f₂a_{i -2}+ . . . +f_ra_i-r, where i varies from 1 to k. This operation requires r initial conditions, one for each of bits a_1-rto a₀. Since each initial condition is one bit, there are 2^rpossible choices for a set of initial conditions. In a first part of the modulation scheme, an intermediate codeword a is calculated as above from b with a set of initial conditions set to zero. Thus: $a_{1 - r} = 0, a_{2 - r} = 0, \dots$ $a_{0} = 0, a_{1} = b_{1}, a_{2} = b_{2} + f_{1} a_{1}, a_{3} = b_{3} + f_{1} a_{2} + f_{2} a_{1}, \dots$ $a_{r} = b_{r} + f_{1} a_{r - 1} + f_{2} a_{r - 2} + \dots + f_{r - 1} a_{1}, a_{r + 1} = b_{r + 1} + f_{1} a_{r} + f_{2} a_{r - 1} + \dots + f_{r} a_{1}$ $\dots$ $a_{k} = b_{k} + f_{1} a_{k - 1} + f_{2} a_{k - 2} + \dots + f_{r} a_{k - r},$ In a second part of the modulation scheme, a set of initial conditions is selected for each codeword c to be generated, depending on a predefined map relating the initial conditions to a predefined set of values for the intermediate codeword a. For instance, if a is all 1s, all 01s or all 10s, use “initial conditions No. 1”, otherwise use “initial conditions No. 2”. Once the set of initial conditions is selected, rather than recalculating the final codeword c by applying the above transform with the selected initial condition set, the effect t (t_1-r, t_2-r, . . . t₀, t₁, t₂, . . . t_k) of the initial condition set is simply added to the intermediate codeword a, i.e. c=a+t. The effect t is calculated by inserting the selected initial condition set in the above transform, and applying the transform to all variables b set to zero. Of course, the zero initial conditions may also be selected, in which case the intermediate codeword a becomes the final codeword c. An interesting property of this modulation technique is that these initial conditions may thus be changed from one codeword to the next without requiring the decoder to be reconfigured. This allows real-time setting of the initial conditions for each codeword so that each codeword may be made to have desired properties. As an example with r=1 and f₁=1, there is one initial condition having two possible values: 0 or 1. For a same data block, switching the initial condition between 0 and 1 switches the resulting codeword to its complement. Therefore, it is certain that one choice of the initial condition will yield a majority of 1s in the resulting codeword. If this is a desired property, the map is such that if the 0 initial condition yields more 0s than 1s in the codeword, the 1 initial condition is selected, otherwise the 0 initial condition is selected. Producing a large number of 1s is often a desired property, because each 1 causes a transition in the signal when it passes through the precoder 14, which transition helps in recovering clock information at the other end of the channel. Since each bit c_iof a codeword c is calculated from r previous bits, corruption of one bit will corrupt r further bits, i.e. the error propagation length is r+1. Therefore, in practical applications, r will be chosen small, often equal to 1 or 2. Choosing r small also increases the rate of the encoder, equal to k/(k+r). The above disclosed modulation technique provides satisfactory results for enhancing signal properties obeying linear laws, which is the case in the specifications for hard-disks with “longitudinal recording”, i.e. having magnetic polarization that changes along the tracks of the disk. Currently, some hard-disks tend to be of the “perpendicular recording” type, i.e. having magnetic polarization changes perpendicular to the disk. The signal specifications for such disks require the “charge” to tend to zero, and this preferably over small sequences of consecutive bits. The charge is defined as the sum of 1s and 0s written on the disk, where each 1 is summed as +1 and each 0 is summed as −1. In other words, the data recorded on the disk should tend to have as many 1s as 0s. The zero charge requirement becomes an additional parameter to be taken into account in the modulation scheme. The known modulation schemes do not offer enough flexibility to address this problem. What is needed, therefore, is a signal modulation scheme with enhanced flexibility, that can in particular make the charge tend to zero while satisfying other requirements in the properties of the signal. SUMMARY OF THE INVENTION According to the invention, this need is satisfied by a data modulation method comprising the steps of: grouping a stream of input data and a corresponding stream of output data into elements of a finite field; applying to input elements of the input data a transform generating output elements of the output data, such that a current output element is a linear combination of a current input element and at least one previous output element, wherein a multiplier applied to at least one previous output element is a non-zero and non-unity element of the finite field; and selecting a set of initial conditions inherent to the transform, such that the output elements resulting from the transform tend to have a desired property. According to an embodiment of the invention, the method comprises the further the steps of: calculating intermediate elements by applying the transform to the input elements with a set of initial conditions of value zero; calculating the effect of the selected set of initial conditions by applying said transform to input elements having value zero and the selected set of initial conditions; and adding the effect to the intermediate elements to obtain the output elements. According to an embodiment of the invention, the step of selecting the initial conditions comprises the steps of: defining distinct sets of initial conditions, each set having a single non-zero element at a distinct position; and selecting each non-zero element of the sets of initial conditions such that the output elements tend to have a respective property. The invention also provides for a decoder or inverse data modulation method comprising the steps of: grouping a stream of input data and a corresponding stream of output data into elements of a finite field; and applying to the input elements of the input data a transform generating output elements of the output data, such that a current output element is a linear combination of a current input element and at least one previous input element, wherein a multiplier applied to a previous output element is a non-zero and non-unity element of the finite field. BRIEF DESCRIPTION OF THE DRAWING The invention is illustrated in the accompanying drawing, wherein: FIG. 1 illustrates a signal processing chain in which the present invention may be implemented. FIG. 2 schematically shows a disk drive system in which the signal processing chain of FIG. 1 may be included. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In an embodiment of the invention, each user data block B=(b₁, b₂, . . . b_k) fed to the encoder 12 of the processing chain of FIG. 1 is subdivided into p=k/m m-tuples B₁, B₂, . . . B_p. Each m-tuple B_iis considered as an element of a finite field GF(2^m). The modulation transform is expressed as: $\frac{1}{α_{0} + α_{1} D_{m} + α_{2} D_{m}^{2} + α_{3} D_{m}^{3} + \dots + α_{r} D_{m}^{r}}$ where D_mis a one m-tuple delay and (α₀, α₁, α₂, . . . α_r, are constant elements of GF(2^m), at least one of which is non-zero and non-unity. In other words, given an i-th m-tuple B_iof a user data block B, the i-th m-tuple A_iof the resulting codeword A is defined as: A_i=α₀B_i+α₁A_i-1+α₂A_i-2+ . . . +α_rA_i-r, where i varies from 1 to p. Of course, all arithmetic is performed over finite field GF(2^m). Addition in a finite field is a bitwise exclusive OR operation. Multiplication is more complex: the two operands are multiplied under their polynomial representation, and the resulting polynomial, modulo the “generator polynomial” of the finite field, is the final result. Such multiplication introduces pseudo-random properties in the results, which contributes to enhanced flexibility of the modulation scheme according to embodiments of the invention. The above transform requires r initial conditions in finite field GF(2^m) for A_1-rto A₀. Since each initial condition is an element of finite field GF(2^m), there are 2^mrpossible choices for a set of initial conditions. In a generic use of the modulation scheme according to embodiments of the invention, a set of constants α=(α₀, α₁, α₂, . . . α_r) is predefined. The corresponding decoder also uses this same set of constants. In operation, a set of initial conditions is found for each codeword being generated in order to make that codeword best match a set of required properties. The choice of the initial conditions does not affect the operation of the decoder, which is one of the interesting features of this type of modulation scheme. The error propagation length of this modulation scheme is rm+1 bits, since one corrupted bit will affect r m-tuples. In a preferred embodiment, the modulation scheme is used in two parts. In a first part, an intermediate codeword A is calculated with the above transform from B with initial conditions set to zero. Thus: $A_{1 - r} = 0, A_{2 - r} = 0, \dots$ $A_{0} = 0, A_{1} = α_{0} B_{1}, A_{2} = α_{0} B_{2} + α_{1} A_{1}, A_{3} = α_{0} B_{3} + α_{1} A_{2} + α_{2} A_{1}, \dots$ $A_{r} = α_{0} B_{r} + α_{1} A_{r - 1} + α_{2} A_{r - 2} + \dots + α_{r - 1} A_{1}, A_{r + 1} = α_{0} B_{r + 1} + α_{1} A_{r} + α_{2} A_{r - 1} + \dots + α_{r} A_{1}$ $\dots$ $A_{p} = α_{0} B_{p} + α_{1} A_{p - 1} + α_{2} A_{p - 2} + \dots + α_{r} A_{p - r} .$ In a second part of the modulation scheme, a map M is defined that relates specific initial condition sets to specific criteria satisfied by the intermediate codeword A. Once the set of initial conditions is selected, rather than recalculating the final codeword C by applying the above transform with the selected initial condition set, the effect T (T_1-r, T_2-r, . . . T₀, T₁, T₂, . . . T_p) of the initial condition set is simply added to the intermediate codeword A, i.e. C=A+T. This is possible, because the transform is linear. The effect T is calculated by inserting the selected initial condition set in the above transform, and applying the transform with input B set to zero. The modulation scheme may be noted ENC(α, M)( X), where α represents the set of constants (α₀, α₁, α₂, . . . α_r) used in the transform, X is the set of p m-tuples to which the transform is applied, and M designates the map that defines initial conditions used in calculating the current codeword. The result of ENC(α, M)( X) is a set of k+rm bits, or r+p m-tuples or elements of GF(2^m). In one embodiment of the invention, instead of exploring all possible initial conditions, only r+1 predefined initial condition sets are used. Map M is thus characterized by r+1 submaps M0, M1, . . . Mr, each associated to a respective one of the r+1 predefined initial condition sets. M0 is associated to initial conditions set to zero, and each of maps Mi, i>0, is associated to an initial condition set where all elements are zero, except the i-th, which is equal to unity, i.e. (0, 0, . . . 0, 1, 0, . . . 0), where 1 is at the i-th position. Thus, the transform, applied to a user data block B, will be expressed as: $\begin{matrix} ENC (\underline{α}, M) (\underline{B}) = ENC (\underline{α}, M 0) (\underline{B}) + \\ β_{1} ENC (\underline{α}, M 1) (\underline{0}) + \\ β_{2} ENC (\underline{α}, M 2) (\underline{0}) + \\ \dots \\ β_{r} ENC (\underline{α}, Mr) (\underline{0}) \end{matrix}$ ENC(α, M0)( B) designates the intermediate codeword A, and all of the other terms represent the effect T of the initial conditions, wherein β=(β₁, β₂, . . . β_r) designates a set of scaling factors in GF(2^m) that will generally change for each codeword A. In fact, map M is such that β=M(X), whatever the value of X. The scaling factors β could form part of their respective submaps, but the above notation allows to better visualize which parameters are adjustable and causes submaps M0, M1, . . . Mr to be constant. The inverse transform, i.e. the decoding operation performed by decoder 24, can be designated DEC(α)(Y). As previously mentioned, the map M does not intervene in the decoding operation. The decoder is such that: B_i=α₀⁻¹(C_i+α₁C_i-1+α₂C_i-2+ . . . α_rC_i-r), where B_iis an m-tuple output by the decoder and C_iis an m-tuple currently input to the decoder. The modulation scheme will be better understood through various examples illustrated below. EXAMPLE 1 1. p=14 2. r=1 3. m=4 4. (α₀, α₁)=(1,μ), where μ is a non-zero and non-unity element of GF(2⁴) 5. Map M1 specifies the use of the unity over GF(2⁴) as initial condition 6. β₁is chosen such that it does not belong to S={0, A₁μ⁻¹, A₂μ⁻², . . . A₁₄μ⁻¹⁴} (reason explained later). This is always possible, since β₁has 16 possible distinct values, whereas S only has 15 elements. In this example, intermediate codeword A=ENC(α, M0)( B) is expressed as: $A_{0} = 0, A_{1} = B_{1}, A_{2} = B_{2} + μ B_{1}, A_{3} = B_{3} + μ B_{2} + μ^{2} B_{1}, \dots$ $A_{14} = B_{14} + μ B_{13} + \dots μ^{13} B_{1}$ The additive effect T of the initial conditions is β₁ENC(α, M1)(0)=β₁(1, μ, μ², μ³, . . . μ¹⁴). β₁is chosen such that T+A has all 4-tuples non-zero, i.e. β₁≠0, μβ₁≠A₁, μ²β₁≠A₂, . . . μ¹⁴β₁≠A₁₄. Hence the choice defined above in item 6. With this choice, each 4-tuple of the final codeword C contains at least one bit at 1, which ensures that there is at least one transition in the signal every 4 bits at the output of precoder 14. This property promotes clock recovery. The search for the desired value of PI requires at most 14 trials out of the 15 non-zero possible values. Each trial requires a comparison with each of the 14 last values of set S. If the 14^thtrial is unsuccessfuil, it is certain that the value searched for is the 15^thnon-zero value. The decoder in this example is such that: B_i=C_i+μC_i-1, where B_iis an m-tuple output by the decoder and C_iis an m-tuple currently input to the decoder. EXAMPLE 2 1. p=6 2. r=1 3. m=2 4. (α₀, α₁)=(1, μ), where μ is a non-zero and non-unity element of GF(2²) 5. Map M1 specifies the use of unity as initial condition. Let Q=ENC(α, M1)(0)=(1, μ, μ², . . . μ⁶) 6. β₁is chosen such that C=A+T=A+β₁Q has the least charge. The charge of C is defined as 2[(−1)^c^-1+(−1)^c⁰+(−1)^c+ . . . (−1)^c¹²], where c₋₁, c₀, c₁, . . . c₁₂are the successive bits of codeword C. (This amounts to adding +1 for each bit at 1 and −1 for each bit at 0, and multiplying the final result by 2.) The search for the required value of β₁is particularly simple in this example, since there are only four values to try. This exemplary modulation does not require a precoder 14 (nor the inverse precoder 22), since the codewords are short (12 bits) and the modulation inherently inserts transitions. Indeed, transitions are necessary to make the charge tend to zero. In using this example in a simulation on random input data, the variance of the charge is about 1.74 over a significant number of consecutive codewords. This result is satisfactory for dealing with perpendicular recording hard-disks. The decoder in this example is also such that: B_i=C_i+μC_i-1, where B_iis an m-tuple output by the decoder and C_iis an m-tuple currently input to the decoder The efficiency of the modulation in reducing charge may be increased by increasing m and k, whereby there will be more values to try for β₁. If several values of β₁happen to reduce the charge, then preferably the one causing most transitions in codeword C is selected, whereby clock-recovery is also promoted. Alternatively, if a precoder 14 is present, the value causing C to have most 1s is selected instead. If multiple properties are to be satisfied by the codewords, r may be chosen equal to the number of properties, whereby there will be as many factors β to search for as desired properties. Factors β will not be independent and it may be necessary to optimize them through several iterations, i.e. if an optimal first factor is found for a first property, a subsequently found optimal second factor for a second property may affect the optimality of the first factor, whereby the first factor is searched for again, which may in turn affect the optimality of the second factor. This may continue until a compromise is found for both factors. FIG. 2 illustrates in simplified form a disk drive system 100 in which the present invention may be embodied. Disk drive system 100 includes a system processor 113 that processes requests and commands from a host computer 101 that direct the drive system to perform specific behavior involving disk drive assembly 107. Examples include reading and writing data to disk drive assembly 107 through a read/write subsystem 105, providing state information such as defect tables, error status and the like. Disk controller unit 103 includes data processing capacity as well as memory in the form of ROM or RAM 112 and buffer memory 104 to generate responses to received commands and requests. The generated responses return data, state information and/or error codes depending on the particular operation being performed. Disk drive assembly 107, e.g., an HDD system, implements physical mass storage typically on a plurality of magnetic disks and read/write head electronics for transferring data with the disks. Disk drive assembly 107 typically includes read channel hardware for preprocessing and amplifying data read from the magnetic media as well as a spin motor for spinning the disks and voice coil motor (VCM) for positioning the read/write head electronics at specific locations with respect to the disk surface(s). A servo control 108 generates drive signals that control the VCM and/or spin motors. These drive signals are in the form of precision voltage or current signals that drive the motors directly. Host 101 typically comprises a data processing device such as a personal computer, server, workstation or the like that requires access to bulk data storage capabilities of disk drive assembly 107. Host 101 sends write commands and data via controller 103 to write data onto the disks as well as read commands to retrieve previously written data from disks within disk drive assembly 107. On both read and write operations the data transmitted from the host 101 to the disk controller 103 includes an indication of a specific location or set of locations on the disk drive assembly that contains the data that is to be accessed. The data that is exchanged through disk controller 103 is typically buffered in buffer memory 104 that is accessible via memory controller 109 and subsequently transmitted to disk assembly 107 or host 101. Buffer memory 104 is used to overcome differences between the speed at which host 101 operates as compared to the speed at which disk assembly 107 operates. In place of or in addition to buffer memory 104, a cache memory may be implemented by appropriate changes (e.g., tag management, hit/miss detection and the like) to memory controller 109. The present invention may be implemented in hardware within the read/write subsystem 105, in software executed within the system processor 113, or in a combined hardware and software mode in processor 113 and subsystem 105. Although the invention has been described and illustrated with a certain degree of particularity, it is understood that the present disclosure has been made only by way of example, and that numerous changes in the combination and arrangement of parts can be resorted to by those skilled in the art without departing from the spirit and scope of the invention, as hereinafter claimed.