JP2008262611A - Decoding method, decoding device and information reproduction device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that in conventional PRML signal processing applied to a reproduction signal with a large waveform distortion, errors occur frequently, so that a signal having a satisfactory reproduction quality cannot be obtained. <P>SOLUTION: In comparison operation of a path metric in Viterbi decoding, the comparison is calculated by adding a predetermined offset value to a predetermined path metric for a path having no shortest Euclidean distance out of two paths which branch from a predetermined state and join again in the same state, thereby selecting a survival path. For the reproduction signal of large waveform distortion, the survival path can be selected in consideration of the influence of the distortion, so that stable decoding is achieved without an influence of a recording quality. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  An object of the present invention is to perform decoding capable of performing stable decoding without being affected by recording quality even when a reproduced signal having a large waveform distortion is reproduced using PRML (Partial Response Maximum Likelihood) signal processing. The present invention relates to a method and a decoding device.

  In recent years, in an optical disc recording / reproducing apparatus whose density is increasing, a PRML signal processing method is often adopted as a reproduction signal processing method. In PRML signal processing, a playback signal having a large intersymbol interference is obtained by performing waveform equalization on the playback signal using the PR method according to the characteristics of the recording / playback system and performing decoding processing using maximum likelihood decoding (ML) such as a Viterbi decoder. It is also possible to obtain data with a low error rate.

  However, on the other hand, when reproducing a DVD in which marks having an incomplete shape are recorded depending on recording performance or the like, binarization determination is performed in the PRML signal processing using a slice level considering the balance of recording codes. The reproduction quality may be deteriorated as compared with the level discrimination binarization process to be performed.

To deal with this problem, Patent Document 1 measures the mark distortion rate for a mark pattern having a long recording width on a DVD, and switches between PRML signal processing and level discrimination binarization processing using the mark distortion rate as a criterion. Thus, it is possible to obtain reproduction quality that is not affected by mark distortion. In Patent Document 2, it is possible to reproduce information from an optical disc recorded with high density by selecting two types of waveform equalization circuits having different equalization characteristics in accordance with PR equalization of a reproduction signal. It is said.
JP-A-2005-93033 JP 2002-230904 A

  However, when reproducing a mark recorded on a BD (Blu-ray Disc) of higher density than DVD, the shortest recording mark is very small and easily affected by intersymbol interference during reproduction. Reproduction is difficult in the level discrimination binarization process used. In the BD, good reproduction quality can be obtained by using the above PRML signal processing for a reproduction signal having a small waveform distortion, but the conventional PRML signal processing is performed for a reproduction signal having a large waveform distortion. If is used as it is, errors frequently occur, and a signal with good reproduction quality cannot be obtained.

  An object of the present invention relates to a decoding method and a decoding apparatus capable of obtaining stable reproduction quality without being affected by recording quality even when a reproduction signal having a large waveform distortion is reproduced using PRML signal processing. Is.

  According to a first aspect of the present invention, a reproduction signal reproduced from a recording medium is sampled at a predetermined sampling period, and a plurality of branch metrics are obtained using a difference between the sampled reproduction signal and a predetermined expected value. And calculating a plurality of path metrics from the plurality of branch metrics, and comparing the path metrics of the two paths that merge into the same state among the plurality of paths having the plurality of path metrics. In the Viterbi decoding method in which one of the two paths is selected as a surviving path based on the magnitude relationship, the Euclidean distance between the two paths branching from a predetermined state and joining again to the same state is minimized. Among the two paths and other paths, the two paths that do not have the shortest Euclidean distance merge into the same state At this time, a comparison operation is performed by adding a predetermined offset value to a path metric of a predetermined path among the two paths whose Euclidean distance is not the shortest, and one of the two paths is calculated based on the magnitude relationship. Is selected as a surviving path.

  The present invention according to a second aspect is the decoding method according to the first aspect, wherein the predetermined path is a path in which waveform distortion has occurred in a reproduction signal reproduced from the recording medium. .

  The present invention according to a third aspect is the decoding method according to the first aspect, wherein the predetermined path is a path corresponding to a pattern having a long recording width recorded on the recording medium. .

  The present invention according to a fourth aspect is characterized in that the Euclidean distance of two paths that branch from the predetermined state and merge again to the same state is one of two paths having 36. This is a decoding method according to the first aspect.

  According to a fifth aspect of the present invention, there is provided a sampling means for sampling a reproduction signal reproduced from a recording medium at a predetermined sampling period, and a plurality of differences using a difference between the sampled reproduction signal and a predetermined expected value. A branch metric calculation unit that calculates a branch metric, a path metric calculation unit that calculates a plurality of path metrics from the plurality of branch metrics, and a plurality of paths having the plurality of path metrics merge into the same state. In a comparison operation means for performing a comparison operation of the path metrics possessed by two paths, and a Viterbi decoder that selects one of the two paths as a surviving path based on the magnitude relationship thereof, a branch from a predetermined state is performed. Two paths that merge again into the same state, two paths that have the shortest Euclidean distance, and more When two paths that do not have the shortest Euclidean distance merge into the same state, the path metric of a predetermined path out of the two paths that do not have the shortest Euclidean distance is predetermined. The decoding apparatus includes a comparison / selection unit that performs comparison operation by adding an offset value, and selects one of the two paths as a surviving path based on the magnitude relationship.

  The present invention according to a sixth aspect is the decoding apparatus according to the fifth aspect, wherein the predetermined path is a path in which waveform distortion has occurred in a reproduction signal reproduced from the recording medium. .

  The present invention according to a seventh aspect is the decoding apparatus according to the fifth aspect, wherein the predetermined path is a path corresponding to a pattern having a long recording width recorded on the recording medium. .

  According to an eighth aspect of the present invention, the predetermined path is one of two paths having a Euclidean distance of 36 between two paths that branch from the predetermined state and merge into the same state again. The decoding device according to the fifth aspect, characterized in that

  According to the present invention, in a path metric comparison operation in Viterbi decoding, a predetermined path is determined with respect to a path whose Euclidean distance between two paths branched from a predetermined state and joined again to the same state is not the shortest. By performing a comparison operation by adding a predetermined offset value to the metric and selecting the surviving path, it is possible to select the surviving path in consideration of the effect of the distortion for the reproduced signal with a large waveform distortion. Stable decoding can be performed without being affected by the recording quality.

(Embodiment 1)
A decoding method and a decoding apparatus according to a first embodiment of the present invention will be described with reference to the drawings. First, a method for comparing path metrics in Viterbi decoding will be described.

Here, the modulation code satisfies a so-called (d, m) restriction (d and m are integers satisfying d and m ≧ 0, and the shortest and longest run lengths are represented by d + 1 and m + 1, respectively). A run-length limited code that satisfies the condition that the minimum polarity inversion distance is 2 (d = 1) is used. The recording code modulates the modulation code with NRZI (Non Return To Zero Inverted). As the PRML equalization method, the PR (a, b, b, a) method is used. In the PR (a, b, b, a) system, a signal (a + b × D + b × D ² + a × D ³ ) obtained by adding the sampling data of four different times at a ratio of a: b: b: a. It has the feature of generating. In the following, PR (1, 2, 2, 1) will be taken for simplicity. When a recording code having a minimum polarity reversal distance of 2 and a PR (1, 2, 2, 1) equalization method are combined, the recording code b _k (k represents time) and the amplitude value y _{k of the} PR equalization output. Is represented by Formula 1.

b _k-3 represents a recording code 3 bits before, b _k-2 represents a recording code 2 bits before, b _k-1 represents a recording code 1 bit before, and b _k represents a current recording code. Assuming that the current state is S (b _k−2 , b _k−1 , b _k ), and the state one hour before is S (b _k−3 , b _k−2 , b _k−1 ), as shown in Table 1. Can be represented by a simple state transition table.

  Here, the state S (0, 0, 0) at each time is S0, the state S (0, 0, 1) is S1, the state S (0, 1, 1) is S2, and the state S (1, 1, 1) ) Is defined as state S3, state S (1, 1, 0) is defined as state S4, and state S (1, 0, 0) is defined as state S5. The state transition diagram at that time is shown in FIG.

  In FIG. 1, lines connecting the states are called branches and represent state transitions. When the state transition diagram of FIG. 1 is developed in the time axis direction, a trellis diagram as shown in FIG. 2 is obtained.

  In the trellis diagram, to consider all combinations of branches (referred to as paths) generated from an arbitrary state through an arbitrary state is equivalent to considering all possible bit strings. Therefore, if the ideal waveform expected for all paths is compared with the reproduced waveform actually reproduced from the optical disc, the most probable path can be found by searching for the path having the closest waveform, that is, the shortest Euclidean distance. Can be determined as the correct answer path.

  At an arbitrary time, two paths merge in states S0, S1, S3, and S4, and one path merges in S2 and S5. For the states S0, S1, S3, and S4 where the two paths merge, if the process of leaving the shorter Euclidean distance between the ideal waveform and the reproduced signal waveform of each path as a surviving path is performed, at any time, six A total of six paths remain, one for each path.

The Euclidean distance between the ideal waveform and the reproduced signal waveform of each path is called a path metric, and the branch metric obtained as the square of the difference between the expected value x _{k of the} branch and the amplitude value y _k of the PR equalized output is expressed as the path metric. It is obtained by accumulatively adding all constituent branches.

The branch metric and path metric formulas are shown respectively. The branch metrics to each state S0, S1, S2, S3, S4, S5 at time k are BM _k ^S0 , BM _k ^S1 , BM _k ^S2 , BM _k ^S3 , BM _k ^S4 , BM _k ^S5 , BM _k ^S6 , surviving. The path metrics of the paths are L _k ^S0 , L _k ^S1 , L _k ^S2 , L _k ^S3 , L _k ^S4 , and L _k ^{S5, respectively,} and the PR equalized output at time k is the amplitude value y _k . The expected amplitude value x _nk (n is an integer) has seven levels (for example, x _0k = 0, x _1k = 1, x _2k = 2) in the case of ideal PR (1, _2, 2, 1) equalization. , X _3k = 3, x _4k = 4, x _5k = 5, x _6k = 6). As described above, the six branch metrics are calculated as shown in Equation 2, and the path metrics are calculated as shown in Equation 3.

  However, the operator min [A, B] is an operator that selects the smaller one of the path metrics of A and B, and this process corresponds to the determination of the surviving path.

For example, when normalization is performed so that the maximum amplitude becomes ± 3 for convenience, the expected values are x _0k = −3, x _1k = −2, x _2k = −1, x _3k = 0, x _4k = + 1, respectively. Seven values of x _5k = + 2 and x _6k = + 3 are obtained, and the path metric formula is as shown in formula 4.

  If the procedure for determining the surviving path is repeated each time the sample value of the reproduction signal waveform is input in this way, the path having a large path metric is deceived, so that the path gradually converges to one. By making this a correct path, the original data bit string is correctly reproduced.

  Here, considering the conditions under which Viterbi decoding is correctly performed, the path metric of the correct path must be smaller than the path metric of the other path (error path) in the path metric comparison expression of Expression 3. At this time, the quality of the reproduction signal can be measured by using the difference between the path metrics of the two paths. That is, when the reproduction signal quality is good, the difference between the two path metrics is large, so the probability of selecting the correct path is high. When the reproduction signal quality is bad, the difference between the two path metrics is small. It becomes impossible to select the correct path or the incorrect path with confidence, and as a result, the probability of selecting the correct path becomes low.

  This will be described in more detail with reference to FIGS.

In the trellis diagram of FIG. 3, attention is paid to a path where two paths branch from a certain state S (state S0 at time k-4) and merge again in a certain state S (state S4 at time k). Assuming that one of the possible paths is a path A, the path A transits the states S0 _k-4 , S0 _k-3 , S1 _k-2 , S2 _k-1 , S4 _k and represents a path metric indicating the probability of the transition. Becomes Pa. Assuming that the other path is path B, path B transits the states S0 _k-4 , S1 _k-3 , S2 _k-2 , S3 _k-1 , S4 _k, and the path metric indicating the probability of the transition is Pb. It should be noted that by using the metric difference from the path branching to the merging, all the accumulated values of the metrics before the path branching point can be ignored.

  Further, Pa-Pb is used as an index of reliability of the decoding result. If Pa << Pb, the path A is selected with confidence, and if Pa >> Pb, the path B is selected with confidence. On the other hand, if Pa = Pb, it does not matter which path A or path B is selected, and it can be said that whether the decoding result is correct is 5 minutes 5 minutes. Therefore, the value of Pa-Pb can be used to determine the reliability of the decoding result. That is, the larger the absolute value of Pa-Pb, the higher the reliability of the decoding result, and the closer the absolute value of Pa-Pb is to 0, the lower the reliability of the decoding result.

  For example, FIG. 4 shows a schematic diagram of the distribution of Pa-Pb when Pa-Pb is obtained for a predetermined time or a predetermined number of times based on the decoding result. FIG. 4 shows the distribution of Pa-Pb when noise is superimposed on the reproduction signal. For example, when the path A takes the ideal value, Pa = 0, so the frequency becomes maximum as Pa−Pb = −Pb, and when the path B takes the ideal value, Pb = 0. The frequency becomes maximum as Pa−Pb = Pa. Further, when the metrics in both paths have the same value, Pa−Pb = 0, and which path is selected is in a state of low reliability of 50/50. A distance d between the two distributions in FIG. 4 corresponds to the Euclidean distance between the path A and the path B.

As described above, when the recording code having the minimum polarity reversal distance of 2 and the PR (1, 2, 2, 1) equalization method are combined, the shortest Euclidean distance between the two paths is d _min = 10. If it is considered that white noise is dominant among noises included in the reproduction signal, this pattern is a pattern in which a 1-bit shift error occurs, so that an error is most likely to occur. In addition, the Euclidean distance between the two paths is innumerable as 12, 14, 16, 18, 20, 36... Depending on the length of the path, but is included in the reproduction signal as described above. If white noise is dominant among noises, the error frequency decreases in proportion to the Euclidean distance. FIG. 5 shows the relationship. As the Euclidean distance increases from 10 to 2, the error rate increases by an order of magnitude or more, and at d = 36, the error rate becomes almost zero.

  However, there is a problem that this relationship does not hold for a reproduced signal having a large waveform distortion.

  FIG. 6A shows the distribution of the reproduction signal and Pa-Pb when a mark with a long recording width is formed in an ideal state. Similarly, the mark with a long recording width is M-shaped. FIG. 6B is a diagram showing a distribution state of a reproduction signal and Pa-Pb having a large waveform distortion when formed in such an unbalanced state.

  When a mark having a long recording width is formed in an ideal state as shown in FIG. 6A, the reproduction signal has a shape without distortion as in path B. The path A is most likely to be mistaken with respect to the path B. In this case, Pa-Pb is also distributed around the position separated by ± d, and there is almost no probability of mistakes between the path A and the path B. No.

  On the other hand, as shown in FIG. 6B, when a mark with a long recording width is formed in an unbalanced state such as an M shape, the reproduction signal of path B has a lower amplitude at the center than at both ends. Become a shape. For such a reproduced signal, the path B that is the correct answer path approaches the shape of the path A that is the error path, and the Euclidean distance between the paths is shortened. Looking at the distribution of Pa-Pb, the distribution of Pb in which path B is the correct path approaches the Pa side, and the state (Pa-Pb = 0) that does not know which path to select increases, so Viterbi decoding Will increase the probability of mistakes.

In the present invention, two path metric comparison operations are performed again for a reproduction signal having a large waveform distortion in consideration of the waveform distortion. In the first embodiment of the present invention, as shown in FIG. 7, when the center of the distribution of one path approaches the other path due to the influence of waveform distortion, the Euclidean distance between the two paths is the shortest. Otherwise, the offset value α is added to the path metric comparison calculation with the other path having less influence of waveform distortion as a reference. The offset value α is set at a position away from the center of the path distribution with little influence of waveform distortion by the shortest Euclidean distance (d _min ). If the distance is the shortest Euclidean distance, a desired error rate can be obtained. However, if the Euclidean distance between paths is set shorter than the shortest Euclidean distance by the offset value α, the error rate becomes very poor. End up. From this, a plausible path including the influence of the waveform distortion is selected by adding the offset value α in the direction of reducing the influence of the waveform distortion with respect to the path where the distance between the two paths is not the shortest Euclidean distance. Therefore, the probability that Viterbi decoding is erroneous can be reduced.

  Hereinafter, an optical disk reproducing method and apparatus according to a first embodiment of the present invention will be described in detail with reference to the drawings. FIG. 8 is a block diagram showing a configuration example of the optical disc device according to the first embodiment of the present invention. The RF amplifier circuit 801 that amplifies the reproduction signal from the optical pickup 800 and the output from the RF amplifier circuit 801 are shown. A / D converter 802 for sampling, PLL circuit 803 for supplying a clock, digital equalizer 804 for waveform shaping of sampling data, and Viterbi for maximum likelihood decoding based on output data of digital equalizer 804 A decoder 805, a pattern detector 806 that detects a pattern of a reproduction signal, and an offset value generator 807 that generates an offset value used for a Viterbi decoding path metric comparison operation according to the detected pattern.

  In FIG. 8, a reproduction signal reproduced from an optical disk 1 as a recording medium via an optical pickup 800 and an RF amplifier circuit 801 is input to an A / D converter 802 and converted into a digital signal. The digital equalizer 804 performs waveform shaping on the input digital signal so that the frequency characteristic of the reproduction system becomes a predetermined PR equalization method, and the shaped digital signal is obtained by the Viterbi decoder 805 by using the most probable data. Is decrypted as The PLL circuit 803 generates a reproduction clock synchronized with the reproduction signal, and the circuits after the A / D converter 802 operate according to the supplied clock. On the other hand, the pattern detector 806 uses a signal before and after the A / D converter 802, a signal obtained by binarizing the signal of the digital equalizer, a signal in the decoding process of the Viterbi decoding 805, and a decoding result to obtain a predetermined value. Detect patterns. Alternatively, another Viterbi decoder 805 may be provided to extract a pattern therefrom. The offset value generator 807 calculates the Euclidean distance between the detected pattern and a pattern that is likely to be erroneous, and generates an offset value corresponding to the Euclidean distance.

  Here, an example is shown in which a mark having a long recording width of 6T (T is a channel clock) or more is detected as a pattern affected by waveform distortion. When 1 bit is detected continuously from time k-5 to k-4, k-3, k-2, k-1, and k (bit string: 111111), the most likely pattern is the bit string: 110011. The Euclidean distance is d = 36. 7T, 8T, and 9T can be detected by the same method, and all of these Euclidean distances are d = 36. In this way, by simply detecting the Euclidean distance of d = 36, it is possible to adjust the offset value for a mark having a long recording width of 6T or more that is particularly susceptible to waveform distortion. The influence of waveform distortion can be reduced. A specific offset value is obtained according to the following Equation 5.

  This offset value α is used for a path metric comparison operation for a mark with a long recording width of a Viterbi decoder described later.

  Next, the configuration of the Viterbi decoder 805 shown in FIG. 9 will be described. FIG. 9 is a diagram illustrating an internal block of the Viterbi decoder, which includes a branch metric calculation unit 901 that calculates a branch metric at each time, an addition comparison selection unit 902 that calculates the probability of a path that transitions to each state, and each state The path memory unit 903 updates and stores the surviving path based on the probability of the path to be transferred to.

The PR equalization output amplitude value y _k output from the digital equalizer 211 is input to the branch metric calculation unit 901. FIG. 10 is a diagram illustrating a configuration example of the branch metric calculation unit 901. The branch metric calculation unit 901 includes seven subtraction units 1001 and a square calculator 1002. The subtraction unit 1001 and the square calculator 1002 calculate the square of the difference between the amplitude value y _{k of} the input PR equalization output and each expected value x _nk, and calculate each branch metric. The branch metrics are output to the addition comparison selecting unit 902 as BM _k ^S0 , BM _k ^S1 , BM _k ^S2 , BM _k ^S3 , BM _k ^S4 , BM _k ^S5 , and BM _k ^S6 .

Here, the square calculator 1002 may be changed to an absolute value calculator that calculates the absolute value of the difference between the amplitude value y _k of the PR equalization output and the expected value x _k .

FIG. 11 is a diagram illustrating a configuration example of the addition comparison selection unit 902. The addition comparison / selection unit 902 includes ten adders 1101, four comparators 1102, a selector 1103, and six path metric memory units 1104 so as to express the form of Equation 3. The addition comparison / selection unit 902 receives the branch metrics BM _k ^S0 , BM _k ^S1 , BM _k ^S2 , BM _k ^S3 , BM _k ^S4 , BM _k ^S5 , BM _k ^S6, and the time k−1. The path metrics at time _{k from the} path metrics L _k-1 ^S0 , L _k-1 ^S1 , L _k-1 ^S2 , L _k-1 ^S3 , L _k-1 ^S4 , L _k-1 ^S5 according to the calculation of Equation 3 L _k ^S0 , L _k ^S1 , L _k ^S2 , L _k ^S3 , L _k ^S4 , and L _k ^S5 are _obtained . The branch metric at time k and the path metric at time k−1 are added by the adder 1101 and input to the comparator 1102 and the selector 1103. The comparator 1102 compares the magnitudes of the input metrics, and the selector 1103 selects the smaller metric according to the comparison result of the comparator 1102 and outputs it to the path metric memory 1104. The path metric memory 1104 stores the metric in a register, and outputs each path metric to each adder 1101 at the next time k + 1. Note that L _k ^S2 and L _k ^S5 are not compared and selected because there is only one path leading to these states.

  Here, the offset value α generated by the above-described offset value generator 908 is added to a predetermined adder at the same timing. The predetermined adder is the right term of the right expression of Expression 3 (4), and the path metric expression to which the offset value is added is expressed by Expression 6.

Here, from the path metric L k-1 S3 at time k-1, the state transition path metric L k S3 at time k, the bit string: 111 bit string: 111 is a system to transition to, a long recording width This is a system in which the reproduction signal of the mark transitions. As described above, the reproduction signal of a mark with a long recording width is easily affected by waveform distortion. Therefore, by adding an offset to the path metric, the desired error rate is ensured and the influence of waveform distortion is included. A plausible path can be selected.

As a specific example, for example, if L _k−1 ^S3 + BM _k ^S6 in Expression 6 is a path metric A of d = 36, and L _k−1 ^S2 + BM _k ^S5 is a path metric B of d = 36, α = − (D−d _min ) = − (36−10) = − 26, and the path metric comparison operation can be considered as min [path A−26, path B].

  Here, if path A is the correct answer path and an ideal reproduction signal is obtained, path A = 0 and path B = 36, so min [−26, 36], and path A is surely Selected. Here, assuming that waveform distortion has occurred in the reproduction signal of path A, for example, when path A = 18 and path B = 18, min [−8, 10] is obtained, path A = 36, path B When = 0, min [10, 0] is obtained, and the path A is selected in any case.

  On the other hand, if path B is the correct path and an ideal reproduction signal is obtained, path A = 36 and path B = 0, so min [10, 0], and the Euclidean distance between the two paths is The shortest Euclidean distance is set and the path B is selected.

  The comparison results of the comparator 1102 are output to the path memory unit 903 as SEL0, SEL1, SEL2, and SEL3, respectively.

  FIG. 12 is a diagram illustrating a configuration example of the path memory unit 903. A path memory unit 903 includes six flip-flop circuits 1201 and four selectors 1202 as one set (path memory), and holds a predetermined number of path memories.

  Starting from data representing the selected path set in the first flip-flop (0 or 1 corresponding to each state), selection based on SEL0, SEL1, SEL2, and SEL3 output from the addition comparison selection unit 902 and In the process of repeating the shift process, the data set in each flip-flop is rewritten to data corresponding to the surviving path. Then, the data set in each flip-flop at the final stage is output as decoded data.

  When there is a sufficient number of path memories, the final decoded data has the same state in any of the six path memories, and therefore the data in an arbitrary path memory is selected as the decoded data. Of course, the path metrics held by each path memory in the final stage may be compared, and the data of the smallest path memory may be adopted. It may be adopted.

  With the above processing, in the present embodiment, even if the reproduction signal having a large waveform distortion is reproduced using PRML signal processing while minimizing the increase in circuit scale, the recording quality is not affected. Stable reproduction quality can be obtained.

(Embodiment 2)
Next, an optical disk reproducing method and apparatus according to the second embodiment of the present invention will be described. The basic configuration of the second embodiment is shown in FIG. 8 which is the same as that of the first embodiment, but the offset value passed from the offset value generator 807 to the Viterbi decoder 805 is different. Since the other points are the same, the same parts are denoted by the same reference numerals and detailed description thereof is omitted, and only different points will be described in detail below.

  In the first embodiment of the present invention, the offset value is adjusted by paying particular attention to marks with a long recording width of 6T or more that are particularly susceptible to waveform distortion, whereas the second embodiment of the present invention. The embodiment is characterized in that an offset value is adjusted for an arbitrary pattern.

  The pattern detector 806 performs detection for each Euclidean distance between paths. For example, when it is desired to detect a pattern of d = 12, 18, the pattern shown in Table 2 below may be detected.

  In Table 2, x means that either 0 bit or 1 bit is entered, and each Euclidean distance can be obtained only by detecting other bit strings. From the obtained Euclidean distance, an offset value is calculated according to the above-described Expression 5. As shown in FIG. 13, the offset value is added to each metric used in the path metric comparison calculation.

  Here, to which adder the offset value is added can be determined based on 4-bit information from time k to time k-3. For example, when the bit string at times k-3, k-2, k-1, and k is 0110, the offset value α8 is added to the adder (adder A1300) corresponding to the right term on the right side of Equation 3 (5), When the bit string at times k-3, k-2, k-1, and k is 1110, the offset value α7 is added to the adder (adder B1301) corresponding to the right term on the right side of Equation 3 (5). At this time, if 0 is given to other offset values, there is no possibility of adversely affecting other than the target pattern.

  Here, the method of adding an offset value to all the adders is used. However, since the difference information between the two paths is sufficient for the path metric comparison, an offset is added to one of the adders to be compared. You can just add values. Further, it is possible to pay attention to only an arbitrary pattern and add an offset value only to that portion. Further, although the case where the Euclidean distance is d = 12, 18 is shown as an example, the distance is not limited to this as long as the shortest Euclidean distance is not d = 10.

  The offset value is not a fixed value as in the present embodiment, and the offset value may be automatically adjusted. For example, the DC component of the Viterbi decoding result may be acquired and adjusted so that the component becomes zero. Alternatively, test recording may be performed with a pattern known in advance, and adjustment may be made based on the performance at that time. The performance may be anything as long as it represents the reproduction performance, such as the error rate or the likelihood difference between the two paths.

  With the above processing, in the present embodiment, it is not necessary to determine a pattern with a large waveform distortion in advance, and the offset value can be adjusted for any pattern, so Viterbi decoding corresponding to the waveform distortion can be performed more flexibly. It can be performed.

  According to the present invention, in a path metric comparison operation in Viterbi decoding, a predetermined path is determined with respect to a path whose Euclidean distance between two paths branched from a predetermined state and joined again to the same state is not the shortest. By performing a comparison operation by adding a predetermined offset value to the metric and selecting the surviving path, it is possible to select the surviving path in consideration of the effect of the distortion for the reproduced signal with a large waveform distortion. Since stable decoding can be performed without being affected by the recording quality, the present invention is useful as an optical disk reproducing apparatus or the like that performs distortion reduction processing for reducing waveform distortion of a reproduction signal obtained by reproducing information recorded on an optical disk.

State transition diagram determined from the fact that the minimum polarity inversion interval used in the embodiment of the present invention is 2 and the restriction of PR (1, 2, 2, 1) equalization A trellis diagram in which a state transition diagram determined from the restriction of PR (1,2,2,1) equalization being 2 and the minimum polarity inversion interval used in the embodiment of the present invention is developed in the time axis direction A trellis diagram determined from the fact that the minimum polarity inversion interval used in the embodiment of the present invention is 2 and the restrictions of PR (1, 2, 2, 1) equalization The figure which shows the two state transition sequences which can be taken between state S0 _k-4 and state S4 _k-5 in the trellis diagram used by embodiment of this invention. Diagram showing the relationship between Euclidean distance and error rate The figure showing the state of distribution of Pa-Pb which shows the reliability of a decoding result in the reproduction signal when formed in an ideal state and the reproduction signal with a large waveform distortion The figure which shows the setting method of the offset value used for the comparison calculation of the path metric used by embodiment of this invention 1 is a block diagram showing a configuration example of an optical disc apparatus used in the first embodiment of the present invention. 1 is a block diagram showing an example of the overall configuration of a Viterbi decoder used in the first embodiment of the present invention. The figure which shows the structural example of the branch metric calculation part used in the 1st Embodiment of this invention. The figure which shows the structural example of the addition comparison selection part used in the 1st Embodiment of this invention. The figure which shows the structural example of the path memory part used in the 1st Embodiment of this invention. The figure which shows the structural example of the addition comparison selection part used in the 2nd Embodiment of this invention.

Explanation of symbols

800 Optical pickup 801 RF amplifier circuit 802 A / D converter 803 PLL circuit 804 Digital equalizer 805 Viterbi decoder 806 Pattern detector 807 Offset value generator 901 Branch metric calculation unit 902 Addition comparison selection unit 903 Path memory unit 1001 Subtraction Unit 1002 Square operator 1101 Adder 1102 Comparator 1103 Selector 1104 Path metric memory unit 1201 Flip-flop circuit 1202 Selector 1300 Adder A
1301 Adder B

Claims (10)

A reproduction signal reproduced from a recording medium is sampled at a predetermined sampling period, a plurality of branch metrics are calculated using a difference between the sampled reproduction signal and a predetermined expected value, and the plurality of branch metrics are calculated. A plurality of path metrics are calculated, and the path metrics of the two paths that merge into the same state among the plurality of paths having the plurality of path metrics are compared. In the Viterbi decoding method of selecting one of the two paths as a surviving path,
Two paths that branch from a given state and merge again into the same state are two paths that have the shortest Euclidean distance, and two paths that do not have the shortest Euclidean distance are in the same state At the time of merging, among the two paths whose Euclidean distance is not the shortest, a predetermined offset value is added to a path metric of a predetermined path, and a comparison operation is performed. A decoding method, wherein one is selected as a surviving path. The predetermined path is:
The decoding method according to claim 1, wherein the path is a path in which waveform distortion occurs in a reproduction signal reproduced from the recording medium. The predetermined path is:
The decoding method according to claim 1, wherein the path corresponds to a pattern having a long recording width recorded on the recording medium. The predetermined path is:
2. The decoding method according to claim 1, wherein the Euclidean distance between two paths that branch from the predetermined state and merge into the same state is one of two paths having 36. 5. Sampling means for sampling a reproduced signal reproduced from a recording medium at a predetermined sampling period, and branch metric calculating means for calculating a plurality of branch metrics using a difference between the sampled reproduced signal and a predetermined expected value And path metric calculation means for calculating a plurality of path metrics from the plurality of branch metrics, and the path metrics possessed by two of the paths that merge into the same state among the plurality of paths having the plurality of path metrics. In a Viterbi decoder that selects one of the two paths as a surviving path based on the magnitude relationship, and a comparison calculation means that performs the comparison calculation of
Two paths that branch from a given state and merge again into the same state are two paths that have the shortest Euclidean distance, and two paths that do not have the shortest Euclidean distance are in the same state At the time of merging, among the two paths whose Euclidean distance is not the shortest, a predetermined offset value is added to a path metric of a predetermined path, and a comparison operation is performed. A decoding apparatus comprising comparison / selection means for selecting one as a surviving path. The predetermined path is:
6. The decoding apparatus according to claim 5, wherein the decoding signal is a path in which waveform distortion occurs in a reproduction signal reproduced from the recording medium. The predetermined path is:
6. The decoding apparatus according to claim 5, wherein the path corresponds to a pattern having a long recording width recorded on the recording medium. The predetermined path is:
The decoding apparatus according to claim 1, wherein the Euclidean distance between two paths branched from the predetermined state and rejoining the same state is one of two paths having 36. An optical head that irradiates the recording medium with laser light, receives the reflected light, and generates a reproduction signal;
An information reproducing apparatus comprising: the decoding device according to claim 5. The program which expressed the decoding method in any one of Claim 1 to 4 with the expression format which can be read by computer.

Images