JP2011181150A - Method and device for correcting asymmetry of data read from recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To correct asymmetry of digital data read from a recording medium by using a histogram. <P>SOLUTION: In a histogram formed by expressing appearance frequency of each output value of digital data as a function of the output value, a value to be a center of symmetry of distribution of corrected output values is calculated as a checkpoint value (S102, S103). One group includes output values of digital data larger than the checkpoint value. The other group includes output values of digital data smaller than the checkpoint value. Based on an output value in the group in which waveform distortion due to asymmetry is relatively smaller, the output value in the other group is corrected (S104). In addition, asymmetry correction using histogram is applied to Viterbi decoding. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to correction of asymmetry (asymmetry) of a digital reproduction signal when information recorded on a recording medium such as a magnetic disk or an optical disk is read and reproduced.

  In recent years, the recording density of magnetic disk devices and optical disk devices has been further improved, and high-density recording can be performed. When performing high-density recording, if data is read under non-optimal playback conditions due to variations in head or disk quality, asymmetry (asymmetry) occurs in the playback signal. In other words, in an ideal state, data can be read using the linear part of the response of the magnetic head or optical head at all times, but the data of the head response is non-linear due to variations in the quality of the disk or head. Happens to have to be read. That is, the output from the magnetic head or the optical head is asymmetrical in the vertical direction when the magnitude of the output signal is viewed in the vertical direction. If this asymmetry can be corrected more powerfully, the yield in manufacturing the head and the disk will increase, and the reliability of the disk apparatus will also improve.

  FIG. 14 schematically shows such asymmetry. The output signal is plotted as a function of time with the magnitude of the output signal on the vertical axis. Originally, the upper side of the output of the head, which should be symmetrical in the vertical direction, is blocked by the nonlinearity of the response of the head, and is asymmetrical in the vertical direction. FIG. 15 shows a histogram in which the number of appearances of the output value (amplitude value) (x axis) obtained as a result of measuring the actual output data with and without this asymmetric distortion is taken on the y axis.

  A plurality of asymmetry correction methods are known. For example, Patent Document 1 discloses a method of performing a branch metric operation on a signal after AD conversion and then inputting the signal to a Viterbi decoder. Patent Document 2 discloses a technique for sampling a analog output signal using an asymmetry detector and calculating a correction coefficient used after AD conversion. Patent Document 3 discloses a technique for improving the accuracy when adaptively controlling the equalization coefficient (tap) of a waveform equalizer (FIR filter) using an asymmetry detector. Patent Document 4 describes a method of performing Viterbi decoding by performing asymmetry correction by nonlinear transformation using the asymmetry amount detected by the offset corrector. Although corresponding corrections are possible by these methods, if more powerful corrections are possible, the manufacturing yield of disks and heads that are becoming more severe as the recording density increases can be further improved.

  In view of the above-described problems of the prior art, the present invention provides a new asymmetry correction method and correction apparatus.

Japanese Patent Laid-Open No. 2001-186027 JP 2008-276882 A Japanese Patent Laid-Open No. 2003-85764 JP 2007-59018 A

  An object of the present invention is to realize more accurate asymmetry correction for digital data read from a recording medium.

  In the present invention, digital data values and the number of appearances of each value are represented in a histogram, and asymmetry correction of the digital data is performed based on the histogram.

  In the present invention, first, a histogram calculation step for expressing the number of appearances of each output value of digital data as a histogram that is a function of the output value, and a value that is to be the center of symmetry of the output value after the asymmetry correction in the histogram. For a checkpoint determination step obtained as a checkpoint value, a set of output values greater than or equal to the checkpoint value, and a set of output values less than the checkpoint value, the distribution of output values in both sets Asymmetry of the data read from the recording medium, including an asymmetry correction step for correcting the output value in one set as a reference value and associating it with the output value in the other set so as to be more symmetrical as the center A correction method is provided. Here, the “center of symmetry” is not an absolute meaning, and the distribution of the output values after the asymmetry correction has regained a certain degree of symmetry, and is almost at the center of the distribution. It means that.

  The present invention also provides a digital data preprocessing step in which A / D conversion is performed on a signal read from a recording medium and then processing by an inverse filter is performed, and the number of appearances of each output value of the digital data is represented by the output value. The histogram calculation step expressed as a histogram that is a function of the above, the sum of the number of appearances of each output value greater than a certain value in the histogram, and the sum of the number of appearances of each output value less than that value are the same or nearly the same A checkpoint determination step for obtaining a value as a checkpoint value, an asymmetry direction determination step for determining whether waveform distortion due to asymmetry is strongly expressed in the larger output value or strongly expressed in the smaller one, and An offset amount determination step for determining an offset amount based on the checkpoint value and the asymmetry directionality. And a set of output values greater than or equal to the checkpoint value and a set of output values less than the checkpoint value, the set of which is determined that the waveform distortion due to asymmetry is small in the asymmetry direction determining step. A lookup table generation step for generating a lookup table that associates the output value of the other set with a certain output value as a reference value, and asymmetry for correcting digital data based on the offset amount and the lookup table An asymmetry correction method for data read from a recording medium, comprising a correction step and an output step for outputting the corrected digital data to a digital equalizer.

  Furthermore, the present invention relates to a preprocessing step for performing processing by a digital equalizer after A / D conversion of an analog signal read from a recording medium, and relating an output value from the digital equalizer to the state of the Viterbi decoder. A histogram calculation step for expressing the number of appearances of each output value as a histogram for each data pattern of the Viterbi decoder, a branch metric calculation step for calculating a branch metric as a lookup table based on the histogram, and a basis for the lookup table Thus, a method for correcting asymmetry of digital data comprising the step of performing Viterbi decoding is provided.

  According to the present invention, histogram calculation means for expressing the number of appearances of each output value of digital data as a histogram that is a function of the output value, and the histogram should be the center of symmetry of the distribution of output values after asymmetry correction. Checkpoint determination means for obtaining a value as a checkpoint value, a set of output values greater than or equal to the checkpoint value, and a set of output values less than the checkpoint value, the distribution of output values in both sets Asymmetry correction means for correcting the output value in one set as a reference value and correcting the output value in the other set by associating it with the output value in the other set so as to be more symmetrical about the value, and outputting the asymmetry correction means An asymmetry correction device for digital data read from a recording medium is provided.

  Furthermore, the present invention represents the number of appearances of each output value of digital data obtained by subjecting a signal read from a recording medium to A / D conversion and inverse filtering as a histogram that is a function of the output value. In the histogram calculation means, a value in which the sum of the number of appearances of each output value equal to or greater than a certain value and the sum of the number of appearances of each output value less than that value are the same or substantially the same is obtained as a checkpoint value. The direction of the asymmetry is determined by judging whether the waveform distortion due to the asymmetry is strongly expressed in the larger or smaller output value from the check point value, and determining the direction of the asymmetry. Offset amount determining means for determining an offset amount based on the checkpoint value and an output value greater than the checkpoint value With respect to a set and a set of output values less than the checkpoint value, an output value in the set determined to have a smaller waveform distortion due to asymmetry is used as a reference value, and the output value of the other set is associated with it. Lookup table generation means for generating a lookup table, offset correction means for performing offset correction based on the offset amount, and correction processing means for performing asymmetry correction based on the lookup table and outputting the result to the digital equalizer An asymmetry correction device for data read from a recording medium comprising:

  Then, the present invention relates to an output value from a digital equalizer obtained by subjecting an analog signal read from a recording medium to A / D conversion and then processing by a digital equalizer, while associating the output value with the state of the Viterbi decoder. On the basis of the lookup table, a histogram calculation unit that represents the number of appearances of each output value as a histogram for each data pattern of the decoder, a branch metric calculation unit that calculates a branch metric as a lookup table based on the histogram, Provided is an asymmetry correction device for digital data comprising a Viterbi decoder for performing Viterbi decoding.

It is a flowchart explaining the method concerning the 1st embodiment of this invention. It is a block diagram which shows the outline | summary of the asymmetry correction | amendment apparatus concerning the 2nd embodiment of this invention. It is a flowchart which shows operation | movement of a parameter calculation unit. A histogram at the start of calculation of parameters A and B at the time when the histogram is generated is shown. It is a flowchart which shows the calculation process of the parameter A. It is a histogram which shows the position of the checkpoint after calculation of parameter A. It is a flowchart which shows the calculation process of the parameter B. It is a flowchart which shows rearrangement of each item of the histogram after parameter B calculation, and change of a name. It is a diagram which shows matching of the output value by the side with the waveform distortion by asymmetry, and the output value by the side without the waveform distortion by asymmetry. It is a flowchart which shows the production | generation process of a lookup table. It is a block diagram which shows the asymmetry correction | amendment apparatus concerning the 3rd embodiment of this invention. It is a histogram of the digital output value calculated about each data pattern, and its appearance frequency. It is a flowchart which shows the process of calculating | requiring an output value and its expression probability from the histogram for every data pattern. It is a graph which shows typically waveform distortion by nonlinear asymmetry. The histogram which shows the frequency | count of appearance of each output value with and without the waveform distortion by asymmetry is shown in the graph.

(First embodiment)
FIG. 1 shows a flowchart of the first embodiment of the present invention. First, a histogram H is created by expressing the number of appearances of each output value of the digital data (S101) read from the recording medium as a function of the output value (S102). The histogram H is as shown in FIG. This histogram H counts the number of appearances of the output value of the digital data. For example, for each output value, the number of appearances of the value increases from the smaller output value to the larger output value (from negative output value to positive output value). ) Can be generated by arranging from bottom to top. Then, for output values greater than or equal to a certain value of the output values and output values less than that value, the sum of the number of appearances is taken, and the sum is the same or nearly the same (the difference between the sums is zero or minimum). The output value is obtained as a checkpoint value (S103). Alternatively, the maximum and minimum output values can be obtained and averaged to obtain a checkpoint value. Steps S102 and S103 constitute a checkpoint determination step. Then, for each set of output values greater than or equal to this checkpoint value and each set of output values less than the checkpoint value, the output value in the set with the smaller waveform distortion due to asymmetry is used as the reference value, and asymmetry The output value in the set with the larger waveform distortion is corrected (S104) (asymmetry correction step). This is because, for example, in the histogram H, the check point value is the boundary, the waveform distortion due to asymmetry is small, or the non-existing side represents the ideal output state, and is used as a reference value. The output value on the opposite side across the value is associated with the reference value. Using the correspondence relationship as a lookup table C, the output value on the side where the waveform distortion due to asymmetry is relatively large is associated with the reference value and converted into a corrected output value. In addition to the method using the look-up table, there are various methods such as finding the correspondence by performing four arithmetic operations based on a certain function. Then, the corrected output value is output (S105).

  According to the embodiment of the present invention as described above, more powerful asymmetry correction can be performed without requiring many hardware resources. Depending on the recording device, there is a tendency that the waveform distortion due to asymmetry is strong or constant in which direction (for example, positive side or negative side starting from zero). Since it may be obvious without measuring the side on which the waveform distortion due to asymmetry is strong, this embodiment does not include the step of measuring on which side of the checkpoint value the asymmetry is greater.

(Second Embodiment)
The second embodiment of the present invention will be described below. FIG. 2 is a block diagram of an apparatus for reproducing data from a recording medium including an asymmetry correction apparatus according to one embodiment of the present invention. A signal read from the recording medium by the magnetic head or the optical pickup is processed by a front end processor (not shown) and then input to the analog / digital converter 1. Thereafter, for example, a digital output having a number of bits of about 6 bits is input to the asymmetry correction apparatus 10. The asymmetry correction apparatus 10 includes an asymmetry correction unit 11 and a parameter calculation unit 12. The asymmetry correction unit 11 includes an inverse filter 13, an offset correction unit 14, and a correction processing unit 15 that corrects asymmetry based on a correction table. The parameter calculation unit 12 includes a histogram calculator 16 and a parameter output unit 17. The histogram calculator 16 represents each value of the digital signal output from the inverse filter 13 and the number of appearances of each value as a histogram. The parameter output unit 17 determines the directionality of asymmetry and the amount of offset correction based on this histogram. The parameter calculation unit 12 generates a lookup table for asymmetry correction based on the histogram, asymmetry directionality, and offset correction amount. The offset correction amount is input to the offset corrector 14 to perform offset correction. Further, the lookup table is input to the correction processing unit 15 in the asymmetry correction unit 11 and further correction processing is performed. The signal output from the correction processing unit 15 is processed by the digital equalizer 3 and the Viterbi decoder 4 and output as binary data.

Here, an FIR filter can be used as the inverse filter 13. This FIR filter may be a FIR filter of about 5 taps, for example. The tap coefficients are five coefficients D0, D1, D2, D3, and D4 which are constants obtained empirically. In the case of this 5-tap, the relationship between the input in (k) and the output X (k) of the inverse filter 13 is as follows.
X (k) = Σ ⁴ _{m = 0} {D _m · in (km)}

Offset adjustment by the offset correction unit 14 is given by the following equation.
Y (k) = X (k) -B
Here, B is the value of the parameter B obtained from the parameter calculation unit 12, more specifically, the parameter output unit 17, and Y (k) is the output.

  The correction processing unit 15 includes a lookup table obtained from the parameter calculation unit 12, more specifically, the parameter output unit 17, and corrects asymmetry based on the lookup table.

  Next, the operation of the parameter calculation unit 12 will be described. In the parameter calculation unit 12, as shown in the flowchart of FIG. 3, first, the histogram calculator 16 obtains the output of the inverse filter 13 (S1) and generates a histogram H (S2). Next, the parameter output unit 17 calculates a parameter A indicating the polarity of asymmetry based on this histogram (S3), and further calculates a parameter B for offset adjustment based on the histogram H (S4). This parameter B is input to the offset corrector 14 and used for the offset correction process. Thereafter, the parameter output unit 17 generates a lookup table C (parameter C) used by the correction processing unit 15 (S5).

  First, a histogram H having a length L is generated by the histogram calculator 16. The value of L generally indicates the range of the output value of the inverse filter 13. For example, if the output is 6 bits, L is 64. However, a portion having no output can be omitted and a value smaller than 64 can be adopted as the value of L. Of course, the output of the inverse filter 13 is not limited to 6 bits. The histogram H counts the number of appearances of each output value of the inverse filter 13 and changes the number of appearances of each output value from the smaller output value to the larger output value (from negative output value to positive output value). F) can be generated by arranging L items from bottom to top. As shown in FIG. 4, the number of appearances of each output value from the inverse filter 13 is tabulated. At this time, the number of appearances of each output value starts from the smaller output value (−32 for 6-bit output) to H (1), and the number of appearances of the maximum output value (31 for 6-bit output) Let H (L). Here, the total number of appearances below the checkpoint in the histogram is Sum1, and the total number of appearances above the checkpoint is Sum2. This checkpoint is at the center of the histogram at the start, and Sum1 and Sum2 are obtained.

  Generally, the values of Sum1 and Sum2 are different, but the parameter A representing the polarity or directionality of the asymmetry can be obtained based on these values. As shown in FIG. 5, first, Sum1 and Sum2 are calculated based on the histogram H (S12). At this time, the check point (check point at the start) is between L / 2 + 1 and L / 2 as described above (this state is called the check point being L / 2). If Sum1 at the time of start is larger than Sum2, parameter A (PA) is set to 1, otherwise parameter A (PA) is set to 0 (S13, S14, S15). If the parameter A is 1, the signal amplitude in the negative direction is small (usually, the nonlinearity appears more strongly in the negative output value). If the parameter A is zero, the signal amplitude in the positive direction is It indicates that it is small (usually a state in which nonlinearity appears strongly at a positive output value). Furthermore, as shown in FIG. 6, the values of Sum1 and Sum2 can be made equal by moving this check point up and down. The parameter B for offset adjustment can be calculated based on the distance between the check point position at the start and the check point position where Sum1 and Sum2 are equal or the difference between them is minimized. The calculation of the parameter B can be performed as shown in FIG. That is, for a given histogram H (S21) with L elements, the checkpoint position M is first set to L / 2 (S22), and Sum1 and Sum2 are obtained (S23). If the previously obtained value of parameter A is zero (S24), it is determined whether Sum1 is greater than Sum2 (S25a). If so, the value of parameter B is changed from the current M value to L. / 2 is subtracted (S26a). If not, the checkpoint position M is increased by 1 (S27a), and the calculation of Sum1 and Sum2 is performed again (S23). If the parameter A is not zero, that is, 1 (S24), it is determined whether Sum2 is greater than or equal to Sum1 (S25b). If so, the value of the parameter B is changed from the current M value to the L value. It is assumed that / 2 is subtracted (S26b). Otherwise, the value of M is decreased by 1 (S27b), and the process returns to the calculation of Sum1 and Sum2 (S23). As a result, the parameter B is determined, and the position of the check point moves from L / 2 to M.

  In this way, when a check point where Sum1 and Sum2 are equal or nearly equal is found and the value of parameter B is determined, the previously obtained histogram is reconstructed. That is, as shown in FIG. 8, the values of H (M + 1) to H (L) above the check point of the histogram are left as they are, and the respective names are referred to as H_up (LM). It is changed to ~ H_up (1) (from (c) to (d) in FIG. 8). This is the same as the original, just by changing the name of each entry. For H (M) to H (1) below the check point, the bottom is such that the value of H (M) at the top of the lower part is at the bottom of the histogram. The histogram is inverted so that the value of H (1) at the top is at the top in the lower part (from (b) to (c) in FIG. 8). Then, H (1) to H (M) are arranged from top to bottom. Here, the name of each entry is changed. That is, H (1) is set as H_dn (M), and sequentially downward, H_dn (M-1) and H_dn (M-2) are set, and H (M) is set as H_dn (1) ((( c) to (d)). Here, up and dn are determined by the polarity of asymmetry, that is, the value of parameter A. That is, the side where the waveform distortion due to asymmetry is strong is set to H_up. H_dn is the one with no or little waveform distortion due to asymmetry, and can be regarded approximately as an ideal output value is obtained. If Sum1 is larger at the start and parameter A is 1, waveform distortion due to asymmetry will be seen in the negative output value direction, H_up will be the upper part of the illustrated histogram where asymmetry is seen, and H_dn will be the lower part Become. If parameter A is zero, the reverse is true.

  Next, as shown in FIG. 9, it is assumed that the histogram H_dn with little or no non-linear distortion represents an ideal output state, and is set as a reference value, and the histogram H_up with larger distortion is set as H_dn. Make it correspond. That is, it corresponds to each term H_up (k) (k = 1,..., LM) and output value i (k) of the histogram H_up. The result is referred to as a look-up table C in association with the output value j (m) (m = 1,..., M) of the term of the histogram H_dn on the smaller distortion side.

  The lookup table C can be created as follows. Since the data on the side with waveform distortion can be considered as “compressed” from normal or ideal data by asymmetry, the correspondence between which output value is compressed to the currently measured value Calculate In this calculation, for example, if H_up (LM) = H_dn (M) + H_dn (M-1) + H_dn (M-2), the output value i (LM) of H_up (LM) is It is determined that the output values j (M), j (M-1), and j (M-2) corresponding to the terms H_dn (M), H_dn (M-1), and H_dn (M-2) are compressed, The corrected value of the output value i (LM) can be determined to be an average value of j (M), j (M-1), and j (M-2). In addition, since it is only necessary to find the correspondence relationship unambiguously, the correspondence relationship is not limited to the above-described calculation method, and various calculation methods can be considered.

  That is, such determination of the lookup table C can be executed as shown in the flowchart of FIG. 10 as an example. As described above, the histogram H and two histograms H_up and H_dn obtained by modifying the histogram H are obtained (S31, S32). Then, variables Pup and Pdn are set as Pup = LM and Pdn = M (S33). Then, Pk that satisfies H_up (Pup) = H_dn (Pdn) +... + H_dn (Pk) is obtained while changing the value of the variable Pk (takes a positive integer value) (S34). When Pk is obtained, the output value i ′ (Pup) corrected as a function of the output values j (Pdn),..., J (Pk) corresponding to the terms H_dn (Pdn),. . That is, i ′ (Pup) = f (j (Pdn),..., J (Pk)). Here, f is an arbitrary function, and is not particularly limited. Typically, f will be an average value of the output values of the corresponding ideal states serving as each variable. Here, Pup = Pup-1 and Pdn = Pk-1 are set, and the process returns to S34. This is repeated until Pup = 1 (S37).

  As a result of obtaining the parameter B and the look-up table C as offset correction values in this way, as shown in FIG. (Offset correction unit 14) associates the actual output value from the correction table with the estimated ideal reference value (correction processing unit 15) to output the asymmetry correction unit 11, and consequently the asymmetry correction device. 10 outputs can be input to the digital equalizer 3. In addition, such a lookup table can be created at the time of manufacture of the disk drive and can be used continuously, or it can be updated at a certain time interval and updated. You can also do it.

  According to the present embodiment, it is possible to perform highly accurate correction for both offset and asymmetry.

(Third embodiment)
Next, an example in which the asymmetry correction technique using the histogram of the present invention is incorporated in a Viterbi decoder will be described as a third embodiment. As shown in FIG. 11, the apparatus of this embodiment includes an A / D converter 21 that receives an output from a data reading device (not shown) from a recording medium, a digital equalizer 25, a Viterbi decoder 26, and the like. An asymmetry correction unit 22 comprising: an output from the digital equalizer 25 and an output from the Viterbi decoder 26 to calculate a histogram; and a parameter D is calculated based on the output of the histogram calculation unit 28 And a parameter calculation unit 23 including a parameter output unit 29. That is, in this embodiment, a histogram is created for each data pattern of the Viterbi decoder 26 for the output value from the digital equalizer 25, and the occurrence probability of each pattern of the input values of the Viterbi decoder is calculated. The Viterbi branch metric is calculated using the appearance probability of the input value of the Viterbi decoder 26 as the parameter D.

First, it is assumed that the Viterbi decoder 26 has 2 ^d states. Here, d is an integer of 1 or more. At that time, the length of the data pattern related to the Viterbi decoder 26 is d + 1. For example, if d = 1, it is a two-state Viterbi decoder, and the length of the data pattern is 2. Specific data patterns are [00], [01], [10], [11], and generally, the number n is n = 2 ^{(d + 1)} . As shown in FIG. 12, a histogram is created for each data pattern. Next, the parameter D is calculated using this histogram, and this parameter D indicates the probability of occurrence of each data pattern for a certain input value.

  A histogram is generated based on the following assumptions. The Viterbi decoding result is V, and at a certain time point k, the Viterbi decoding decoding result is V (k), V (k−1),... V (km) (where V (k), V ( k-1) ... is already known). The input value I (k) at time k is also known. The magnitude of I (k) is one of i (1), i (2)... I (L). When d = 1 (2 state Viterbi), the data pattern is 2 bits. At this time, a data pattern is expressed using V (k) and V (k-1). For example, when V (k-1) = 1 and V (k) = 0, the data pattern indicates [10]. By doing so, after a certain period of time, a histogram for the input value I is generated for each data pattern.

  At time point k, only a calculation for one data pattern can be performed, but after a certain time, a histogram for each input value can be calculated for each data pattern. Using this histogram, the appearance probability of the magnitude i (L) is calculated from the magnitude i (1) of the input values I in each data pattern. Then, decoding is performed by selecting an optimum data pattern by comparing the probabilities calculated in advance with each possible data pattern in the Viterbi decoder.

  As shown in FIG. 12, each histogram represents the number of appearances H (m) of each output value i (m) (m is an integer from 1 to L) from the digital equalizer 25. When d = 1 as described above, since four data patterns are created, histograms H1, H2, H3, and H4 are created.

  Based on the histograms H1, H2, H3, and H4 for each data pattern created in this way, parameters D1, D2, D3, and D4 representing the occurrence probability of each data pattern can be calculated as shown in FIG. it can. First, the histogram H1 is taken as an example. As described above, the number of appearances of each output value i (m) from the digital equalizer 25 is defined as H1 (m), and this is obtained for all output values, and histograms H1 (H1 (1) to H1 (L) are obtained. ) Is created for data pattern 1 (FIG. 13A). The sum of the elements of H1 (m) is SumH1. SumH1 = (H1 (1) + H1 (2) +... + H1 (L). Then, the value of each term in the histogram is divided by SumH1 to obtain new elements (FIG. 13 (b)). The term value H1 (m) / SumH1 remains exactly the same, and its name is the parameter PD1 (m), that is, the parameter PD1 (m) is the output value i (m from the digital equalizer 25 in the data pattern 1. ) Is represented (FIG. 13 (c)).

The Viterbi decoding method of the present embodiment differs from the conventional Viterbi decoding method in the branch metric calculation method, and realizes the asymmetry correction function using the parameter D. First, consider the transition from state S (i) to S (j) at time k. S (i) _is set to _{x k-d x k-d} + 1 ... x k-1, S (j) _is the _{x k-d + 1 x k} -d + 2 ... x k. Then, the conventional Viterbi branch metric BM (r _k ; x _k−d x _{k−d + 1} ... X _k ) is
BM (r _k ; x _k-d x _{k-d + 1} ... x _k )
= {R _k −m (x _k−d , x _{k−d + 1} ,..., X _k )} ²
It becomes. Here, _{r k} is the input value at the time k, m is the pattern in the partial list Pons (PR) system _{(x k-d, x k} -d + 1, ..., x k) is the ideal value of. d is none other than d appearing in paragraph number 0031. Basically, the function of the BM, each time, the input value r _k is the pattern _{(x k-d, x k} -d + 1, ..., x k) is the natural logarithm of the conditional probability density for. If the noise at that time is a Gaussian distribution or if it can be assumed as such, the logarithm of the conditional probability density is (r _k −m (x _k−d , x _{k−d + 1} ,..., X _k )) ² To express. Here, the logarithm is taken for the following reason. First, the Viterbi decoder is a maximum likelihood detector. The operation is done by looking for a decoding that maximizes the product of the next probabilities. Then, as can be seen from the above equation, it is necessary to calculate by taking the product of the probabilities at each time point, but if the logarithm is used, this multiplication can be simplified by adding.

Here, according to this conventional Viterbi detection method, m (x _k−p , x _{k−p + 1} ,..., X _k ) is a pattern (x _k−p x _{k−p + 1} ... X _k ) and a partial response. Since it is calculated by the target value, the influence of nonlinearity cannot be canceled. As a result, the branch metric value representing the difference between the actual value r and the ideal value m cannot be accurately calculated, and the Viterbi detection performance is degraded. Therefore, in this embodiment, an asymmetry correction function is added by calculating a branch metric using the parameter D as follows.
_{BM (r k; x k-} p x k-p + 1 ... x k)
= −log (D (r _k | x _k−p x _{k−p + 1} ... X _k ))
And _{Here, D (r k | x k} -p x k-p + 1 ... x k) is the value of the parameter D corresponding to the input value _{r k,} the pattern of the input value _{r k} point _{k (x k-p} x _{k−p + 1} ... x _k ). This can be used by storing the value of the parameter D obtained as described above with reference to FIG. 12 in the form of a lookup table in a storage device. This look-up table can be set when the recording medium drive is manufactured and used continuously, or can be updated at a certain interval or at a certain timing. The branch metric is calculated using the statistical information D. Since parameter D naturally includes asymmetry nonlinearity, it is considered that more accurate branch metrics can be calculated and compared. Here, when d = 4, the length of the data pattern is 5, and the number of patterns is 2 ⁵ = 32. There are 32 lookup tables. The size of each table is determined by the range of r _k. If the value is 16 capable of r _k, the size of each table is 16. Therefore, the size of the entire lookup table is 32 × 16.

DESCRIPTION OF SYMBOLS 1 A / D converter 3 Digital equalizer 4 Viterbi decoder 10 Asymmetry correction apparatus 11 Asymmetry correction unit 12 Parameter calculation unit 13 Inverse filter 14 Offset correction part 15 Correction process part 16 Histogram calculator 17 Parameter output part

Claims (6)

A histogram calculation step that represents the number of appearances of each output value of the digital data as a histogram that is a function of the output value;
In the histogram, a checkpoint determination step for obtaining, as a checkpoint value, a value that should be the center of symmetry of the output value distribution after asymmetry correction;
For the set of output values greater than or equal to the checkpoint value and the set of output values less than the checkpoint value, the distribution of the output values in both sets is closer to the symmetry around the checkpoint value. An asymmetry correction method for data read from a recording medium, comprising: an asymmetry correction step for correcting an output value in the set by using the output value in the set as a reference value and associating it with the output value of the other set. Prior to the histogram calculation step, a digital data preprocessing step of performing A / D conversion on the signal read from the recording medium and then performing processing by an inverse filter is performed.
In the checkpoint determination step, in the histogram, a value that makes the sum of the number of appearances of each output value equal to or greater than a certain value and the sum of the number of appearances of each output value less than that value equal or substantially equal to the checkpoint value And
The asymmetry correction step includes
An asymmetry direction determining step for determining whether waveform distortion due to asymmetry is strongly expressed in the larger output value or strongly expressed in the smaller output value;
An offset amount determination step for determining an offset amount based on the checkpoint value and the asymmetry direction;
Output values in the set of output values greater than or equal to the checkpoint value and the set of output values less than the checkpoint value that are determined by the asymmetry direction determination step that waveform distortion due to asymmetry is small A reference table, and a lookup table generation step for generating a lookup table that associates the output value of the other set with the reference value;
A correction processing step of correcting digital data based on the offset amount and the lookup table,
2. The method for correcting asymmetry of data read from a recording medium according to claim 1, wherein an output step of outputting the corrected digital data to a digital equalizer is executed after the asymmetry correction step. A preprocessing step of performing processing by a digital equalizer after A / D converting an analog signal read from a recording medium;
A histogram calculation step for representing the number of appearances of each output value as a histogram for each data pattern of the Viterbi decoder, while associating the output value from the digital equalizer with the state of the Viterbi decoder;
A branch metric calculation step of calculating a branch metric as a lookup table based on the histogram;
A method for correcting asymmetry of data read from a recording medium, comprising: performing Viterbi decoding based on the lookup table. Histogram calculation means for expressing the number of appearances of each output value of digital data as a histogram that is a function of the output value;
In the histogram, checkpoint determining means for obtaining a value that should be the center of symmetry of the distribution of output values after asymmetry correction, as a checkpoint value;
For the set of output values greater than or equal to the checkpoint value and the set of output values less than the checkpoint value, the distribution of the output values in both sets is closer to the symmetry around the checkpoint value. An asymmetry correction device for digital data read from a recording medium, comprising: an asymmetry correction means for correcting the output value in the set by using the output value in the set as a reference value and correcting the output value in the other set by associating it. Each output value of the digital data is obtained by subjecting a signal read from a recording medium to A / D conversion and processing by an inverse filter,
The asymmetry correction means is
In the histogram, the direction of the asymmetry is determined by determining whether the waveform distortion due to the asymmetry is strongly expressed in the larger or smaller output value from the check point value and determining the direction of the asymmetry. An offset amount determining means for determining an offset amount based on directionality and the checkpoint value;
With respect to a set of output values greater than or equal to the checkpoint value and a set of output values less than the checkpoint value, an output value in a set determined to have a small waveform distortion due to asymmetry is used as a reference value. Lookup table generation means for generating a lookup table for associating the output values of the other set;
Offset correction means for performing offset correction based on the offset amount;
5. An asymmetry correction device for data read from a recording medium according to claim 4, further comprising: correction processing means for performing asymmetry correction based on the look-up table and outputting the result to a digital equalizer. For each data pattern of the Viterbi decoder, the output value from the digital equalizer obtained by A / D converting the analog signal read from the recording medium and performing processing by the digital equalizer is related to the state of the Viterbi decoder. A histogram calculator that represents the number of appearances of each output value as a histogram,
A branch metric calculator for calculating a branch metric as a lookup table based on the histogram;
An asymmetry correction device for data read from a recording medium, comprising: a Viterbi decoder that performs Viterbi decoding based on the lookup table.

Images