JP2003141823A - Method for evaluating quality of reproduced signal, and information reproducing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for evaluating a signal quality on the basis of an index, by which the error rate of the binarized result obtained by maximum likelihood decoding can appropriately be estimated. SOLUTION: In a maximum likelihood decoding system for estimating a most certain state transition stream out of (n) ((n) is n integer of >=2) ways of state transition streams while having state transition rules capable of taking a plurality of states in a time (k) ((k) is an arbitrary integer) and taking (n) ways of state transition streams from the state in a time k-j ((j) is an integer of >=2) to the state in the time (k), when the certainty of state transition from the state in the time k-j of the most certain state transition stream among (n) ways of state transition streams to the state in the time (k) is defined as PA, the certainty of the state transition from the state in the time k-j of the second certain state transition stream to the state in the time (k) is defined as PB and the reliability of the decoded result from the time k-j to the time (k) is defined as |PA-PB|, the value of |PA-PB| is found for a prescribed time or prescribed times and by finding the diffusion thereof, the index presenting the signal quality correlative with the error rate of the binarized result of maximum likelihood decoding can be provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention can perform a method for evaluating the quality of a decoded signal and a method for evaluating the quality of the decoded signal when the digital information recorded on a recording medium is decoded by the maximum likelihood decoding method. The present invention relates to an information reproducing device.

[0002]

2. Description of the Related Art Recently, in AV equipment, personal computers, etc., HDD (hard disk drive),
2. Description of the Related Art Devices such as an optical disk drive or a magneto-optical disk drive that can reproduce digital information recorded on a recording medium are widely used.

FIG. 1 shows a conventional optical disk drive 900.
The partial structure of is shown. The reflected light from the optical disc 1 is
The reproduction signal is converted by the optical head 2. The reproduced signal is waveform-shaped by the waveform equalizer 3, and then the comparator 4
Is binarized at. The threshold value of the comparator 4 is
Usually, feedback control is performed so that the integration result of the binarized signal output from the comparator 4 becomes zero.

In the optical disk drive 900, the PLL
A (phase looked loop) circuit generates a clock signal (reproduced clock signal) synchronized with the reproduced signal. In order to generate the regenerated clock signal, the phase comparator 5 detects a phase error between the binarized signal output from the comparator 4 and the clock signal output from the VCO (voltage controlled oscillator) 7. The detected phase error is averaged by an LPF (low-pass filter) 6, and this LPF 6 is averaged.
The control voltage of the VCO 7 is set based on the output from the. In this way, the control voltage of the VCO 7 (oscillation frequency of the VCO 7) can be feedback-controlled so that the phase error output from the phase comparator 5 is always zero. As a result, the VCO 7 can output a clock signal synchronized with the reproduction signal. PLL like this
By using the circuit, a clock signal synchronized with the reproduction signal can be stably extracted even when the disc has an eccentricity, for example.

The reproduction clock signal is used to judge whether the recording code (digital information) is 1 or 0. More specifically, the detection pulse of the comparator 4 (that is, the signal portion exceeding the threshold value in the binarized signal output from the comparator 4) exists within the window width (window width) defined by the reproduction clock signal. The digital information can be reproduced by detecting whether or not to do so.

However, the detection pulse output from the comparator 4 deviates from the window width of the reproduction clock due to the inter-code interference of the reproduction signal, the distortion of the recording mark, the circuit noise, the control residual of the PLL circuit, etc. Errors may occur. Such a time difference between the detection pulse of the comparator 4 and the reproduction clock is called "jitter".

When the digital information is reproduced as described above, the reproduced signal quality (error rate) can be detected by obtaining the jitter distribution. The distribution of this jitter is
It can be assumed that the average value has a normal distribution of 0. In this case, the error rate Pj (σ / Tw) is calculated by the following equations (1) and (2) using the standard deviation σ of the jitter distribution. expressed.

[0008]

[Equation 1]

[0009]

[Equation 2]

Here, σ is the standard deviation of the jitter distribution assumed to be a normal distribution, and Tw is the window width.

From the graph shown in FIG. 2, as the jitter standard deviation increases, the error rate (bit error rate BE
It can be seen that R) increases. The jitter of the reproduced signal is T
It can be actually measured using an IA (Time Interval Analyzer). Therefore, even if an error does not actually occur, the signal quality can be evaluated by the standard deviation σ / Tw of the jitter, and thus the easiness of error occurrence can be predicted. If the standard deviation of the jitter is measured in this way, for example, the performance of the drive,
It is possible to confirm and inspect the performance of the recording medium, the performance of the optical head and the like. Further, by adjusting the parameters of the equalizer so that the standard deviation of the jitter decreases, more stable reproducing operation can be performed.

[0012]

On the other hand, unlike the method of directly reproducing the digital information from the binarized signal output from the comparator 4 as described above, the method of reproducing the digital information by the maximum likelihood decoding method. It has been known. As the maximum likelihood decoding method, for example, PRML (Part
ial Response Maximum Likelihood) method is known. In the PRML system, data recording and reproduction are performed in consideration of the occurrence of intersymbol interference when the recording density is high. More specifically, the signal reproduced from the recording medium is subjected to partial response equalization so as to have a predetermined frequency characteristic using a waveform equalizer or a digital filter, and then the maximum likelihood using Viterbi decoding or the like. It is decoded into binary data (most likely). PRM
In the L system, a reproduction signal with low S / N (signal to noise),
It is possible to decode data with a low error rate even from a reproduced signal that is relatively affected by intersymbol interference.

In such a maximum likelihood decoding system, decoding is performed by selecting the most probable state transition sequence based on the reproduced signal. State, generally up to time k
The quantity that represents the likelihood of state transition up to Sn (n is the number of states) is defined by equation (3).

[0014]

[Equation 3]

Here, y _i is the value of the reproduced signal (digital sample data) at time i, and level _v is the expected value of the ideal reproduced signal.

In the maximum likelihood decoding circuit, the state transition sequence is selected so that the quantity representing the certainty obtained by the above equation (3) is minimized. When the maximum likelihood decoding method is used, unlike the above-described method of discriminating between "1" and "0" depending on whether or not the detection pulse is within the window width at each time k, sampling is performed at the reproduction clock at each time k. Euclidean distance (y _k -level _v ) ² is obtained using the _obtained data y _k , and decoding is performed based on this Euclidean distance. Therefore, the decoding result in the maximum likelihood decoding method is also influenced by the sample value y _k of the past sampled reproduced signal.

When such a maximum likelihood decoding method is used, there are cases where an error occurs and cases where no error occurs even if the reproduced signal has the same standard deviation σ of jitter. For this reason,
It is difficult to predict the error rate of the binarized result obtained by maximum likelihood decoding using the standard deviation σ of the jitter of the reproduced signal. Therefore, it is necessary to use an error rate prediction method (signal quality evaluation method) more suitable for the maximum likelihood decoding method.

A method for evaluating the quality of a signal reproduced by the maximum likelihood decoding method is described in, for example, Japanese Patent Laid-Open No. 10-21651. In the device described in this publication, the difference in likelihood between two paths (state transition sequences) that minimizes the Euclidean distance is obtained, and the signal quality is evaluated by statistically processing this difference.

More specifically, in order to obtain the difference in likelihood between two paths having the same state at time k, time k
-1 in two different states (time k in each path
The accumulated value of the branch metrics of the path (surviving path) that has already been determined to be the maximum likelihood in each of the (-1 state) is used. However, when the cumulative value of the branch metric at the time k−1 is used, it may be desired when a path different from the path whose actual likelihood is to be checked is erroneously selected as the path before the time k−1. There was a possibility of using the cumulative value of the branch metric that is not. In the above publication, the minimum Euclidean distance is 2
Although it is described that one path is selected and the difference between these likelihoods is calculated, a method for more reliably calculating the likelihood to be actually calculated for the two paths is specifically described. Not not.

The present invention has been made in view of the above problems, and an object thereof is to evaluate reproduced signal quality using an index having a correlation with an error rate of a binarization result by maximum likelihood decoding. A method and an evaluation device are provided.

[0021]

According to the reproduced signal quality evaluation method of the present invention, from the first state S _{kj at} the time k-j (k is an integer of 3 or more and j is an integer of 2 or more) to the second state at the time k. In the case where the reproduction signal is decoded by the maximum likelihood decoding method that selects the most probable state transition sequence from the n (n is an integer of 2 or more) state transition sequences that transit to the state S _k , the decoding is performed. And a first state S that defines the n number of state transition sequences in a predetermined period j from the time k-j to the time k.
a step of detecting a predetermined combination of _kj and the second state S _k, and a most probable first state transition sequence of the n types of state transition sequences defined by the detected predetermined combination. When Pa is an index indicating the likelihood of the state transition in the predetermined period j, and Pb is an index indicating the likelihood of the state transition of the second most likely second state transition sequence in the predetermined period j. , | P
a-Pb | is used to judge the reliability of the decoding result from the time kj to the time k.

In a preferred embodiment, the Pa
Is defined based on a difference between an expected value and an actual sample value indicated by the first state transition sequence in the predetermined period j, and Pb is the second state transition sequence in the predetermined period j. Is defined based on the difference between the expected value and the actual sample value.

In a preferred embodiment, the Pa
Is an expected value l _kj from time k−j to time k indicated by the first state transition sequence in the predetermined period j.
.., l _k-1 , l _k and the actual sample value y _kj ,.
., Y _k−1 , y _k corresponding to the cumulative value of the square of the difference,
b is the expected value m _kj , ..., _{M k-1} , m _k from time k−j to time k indicated by the second state transition sequence and the actual sample value y _kj ,. This corresponds to the cumulative value of the square of the difference between y _k−1 and y _k .

In a preferred embodiment, n = 2.

In a preferred embodiment, the first
The Euclidean distance between the second state transition sequence and the second state transition sequence has a minimum value.

In a preferred embodiment, the | P
The method further includes determining a variation in reliability of the decoding result by measuring a-Pb | a plurality of times.

In a preferred embodiment, the variation in reliability is indicated by using the standard deviation of the distribution of | Pa-Pb |.

In a preferred embodiment, the variation in reliability is shown by using the standard deviation of | Pa-Pb | and the average value of the distribution of | Pa-Pb |.

In a preferred embodiment, the | P
The variation in reliability of the decoding result is determined by detecting the frequency at which a-Pb | exceeds a predetermined value.

In a preferred embodiment, the minimum polarity inversion interval of the recording code is 2 and PR (C0, C
(1, C0) equalized reproduction signal is decoded.

In a preferred embodiment, the minimum polarity reversal interval of the recording code is 2, and PR (C0, C
(1, C1, C0) equalized reproduction signal is decoded.

In a preferred embodiment, the minimum polarity inversion interval of the recording code is 2 and PR (C0, C
1, C2, C1, C0) equalized reproduction signal is decoded.

In a preferred embodiment, the | P
When calculating a-Pb |
The feature is that the calculation of the power is not performed.

An information reproducing apparatus according to the present invention comprises a gain controller for adjusting an amplitude value of a reproduced signal, a waveform equalizer for shaping the reproduced signal so as to have a predetermined equalizing characteristic, and a synchronous signal for synchronizing the reproduced signal. A regenerated clock generating circuit for generating the regenerated clock, an A / D converter for generating sampling data by sampling the regenerated signal with the regenerated clock, and outputting the sampling data; and the sampling data. The maximum likelihood detector that decodes the most probable digital information from the maximum likelihood detector and the index indicating the likelihood of the state transition in the predetermined period of the first state transition sequence determined to be the most probable by the maximum likelihood detector are Pa. The index indicating the probability of the state transition of the second most likely second state transition sequence in the predetermined period is Pb. To time, | and a difference metric calculator for calculating a | Pa-Pb.

In a preferred embodiment, an additional waveform equalizer for performing waveform shaping so as to have a predetermined equalization characteristic different from that of the waveform equalizer is further provided, and the reproduction clock is the additional waveform or the like. It is generated from the reproduced signal whose waveform is shaped by the digitizer.

[0036]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of a reproduced signal quality evaluation method and an information reproducing apparatus according to the present invention will be described below.

The method for evaluating the quality of the reproduced signal according to the embodiment of the present invention will be described below. In the embodiment described below, as the recording code, a code having a minimum polarity reversal interval of 2 defined according to the (1,7) RLL modulation method or the like is used. That is, two or more 0s or 1s always occur in the recording code. Further, decoding is performed using the PRML method in which the frequency characteristics of the recording system and the frequency characteristics of the reproducing system are set so as to show the PR (1,2,2,1) equalization characteristics as a whole. Hereinafter, a more specific decoding procedure will be described.

The recording code (either 0 or 1) is expressed as follows.

Current record code: b _k Record code before 1 time: b _k-1 Record code before 2 time: b _k-2 Record code before 3 time: b _k-3

Let Level _v be the ideal value of the reproduced signal in the case of PR (1, 2, 2, 1) equalization.
evel _v is expressed by the following equation (4).

Level _v = b _k-3 + 2b _k-2 + 2b _k-1 + b _k (4)

Here, k is an integer representing time, and v is an integer from 0 to 6. In case of PR (1, 2, 2, 1) equalization,
The ideal sample value (expected value) has seven values from 0 to 6 according to the combination of recording codes (Level _v )
Can be taken.

Next, the state transition of the recording code will be described. Let S (b _k-2 , b _k-1 , b _k ) be the state at time k, and
The state at 1 is S (b _k-3 , b _k-2 , b _k-1 ). Time k-1
Considering the combination of the state at and the state at time k,
A state transition table as shown in Table 1 below is obtained. As described above, since the modulation method in which the minimum inversion interval between 0 and 1 is 2 is adopted, the state transitions that the recording code can take are limited to the following 10 types.

[0044]

[Table 1]

For simplicity, the state S (0,0,0) _k at time _k is
S0 _k , state S (0,0,1) _k is S1 _k , state S (0,1,1) _k is S2 _k , state S
(1,1,1) _k is S3 _k , state S (1,1,0) _k is S4 _k , state S (1,0,0) _k is
Notation to and so on S5 _k. State transitions that can occur during the period from time k−1 to time k (the time corresponding to one cycle T of the recovered clock) are shown in the state transition diagram shown in FIG. The trellis diagram shown in is obtained.

[0046] Here, attention is paid to the state S0 _k-5 in the state S0 _k and time k-5 at time k. FIG. 5 shows the states S0 _k and S0 _k-5.
It shows two possible state transition sequences between and. If one of the possible state transition sequences is path A, path A
If S0 _k-5 , S0 _k-4 , S0 _k-3 , S0 _k-2 , S0 _k-1 , and S0 _k are transited, and the other state transition sequence is path B, then path B is in state S0 _k-5 , S
It transits 1 _k-4 , S2 _k-3 , S4 _k-2 , S5 _k-1 , and S0 _k . In addition, in FIG. 4 and FIG. 5, (recording code / Lev
elv) is shown, Levelv is shown as taking seven values from -3 to 3, and each of -3 to 3 corresponds to each of Levels 0 to 6 of the above.

Thus, the state at time k-5 is S0._k-5
And the state at time k is S0_kIf
Transition to either path A or path B described above.
It is estimated that That is, from time k-7 to time k.
Decrypted data at (C_k-7, C_k-6, C _k-5, C_k-4, C_k-3, C_k-2,
 C_k-1, C_k) = (0,0,0, x, x, 0,0,0), where x is 0 or
Or a value of 1) is obtained, path A or path B
It is assumed that the state transition of is most likely.

In this way, the state S0 _k at time k and the time k
If the state S0 _k-5 at -5 is detected (ie (0,0,
0, x, x, 0,0,0) when the decoding result is obtained), pass A
It is determined which of the two is more likely, that is, the pass B or the pass B.
This judgment is based on the magnitude of the deviation between the ideal sample value (expected value) indicated by the path A and the actual sample value, and the deviation between the ideal sample value (expected value) indicated by the path B and the actual sample value. Can be done by comparing with the size of. More specifically, the expected value (Level) from time k-4 to time k indicated by path A and path B, respectively.
_v ) and the respective actual values of the reproduced signals y _k-4 to y _k based on the cumulative result of the square of the difference.
Alternatively, it can be determined which state transition sequence of the path B is more likely.

[0049] Here, the expected value l _k-4 from time k-4 showing the path A to the time _{k, l k-3, l} k- 2, l k-1, l k ( i.e., 0,0, 0,0,0) and the reproduction signal y
_Let Pa be the cumulative value of the square of the difference from the value from _k-4 to y _k ,
Expected values m _k-4 , m from time k-4 to time k of path B
_k-3 , m _k-2 , m _k-1 , m _k (that is, ₁ , ₃ , ₄ , ₃ ,
When the cumulative value of the square of the difference between 1) and the values of the reproduced signals y _k-4 to y _k is Pb, the cumulative value Pa is represented by the following equation (5), and the cumulative value Pb is the following equation. It is represented by (6).

Pa = (y _k-4 -0) ² + (y _k-3 -0) ² + (y _k-2 -0) ² + (y _k-1 -0) ² + (y _k -0 ) ² … (5)

Pb = (y _k-4 -1) ² + (y _k-3 -3) ² + (y _k-2 -4) ² + (y _k-1 -3) ² + (y _k -1 ) ² … (6)

The cumulative value Pa thus obtained is an index showing the certainty of the transition of the path A in the predetermined period from time k-5 to time k, and the smaller the value of Pa, the more reliable the path A is. It will be unique. Also, the cumulative value Pb
Is in a predetermined period from time k-5 to time k,
It is an index indicating the probability of transition of the path B, and the smaller the value of Pb, the more likely the path B is. When the value of Pa or Pb is 0, the probability of path A or path B is maximum.

Next, the meaning of the difference Pa-Pb between Pa and Pb will be described. If Pa << Pb, the maximum likelihood decoding circuit will select the path A with confidence, and if Pa >> Pb, it will select the path B with confidence. However,
If Pa = Pb, it does not matter whether path A or path B is selected, and it can be said that it is 5 minutes and 5 minutes whether the decoding result is correct. Therefore, the value of Pa-Pb can be used to judge the reliability of the decoding result. That is, P
The larger the absolute value of a-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.

Index Pa-P indicating the reliability of this decryption result
b is used to evaluate the quality of the reproduced signal. For this purpose, for example, Pa-Pb is obtained by obtaining Pa-Pb for a predetermined time or a predetermined number of times based on the decoding result.
Get the distribution of. A schematic diagram of the distribution of Pa-Pb is shown in FIG. FIG. 6A shows the distribution of Pa-Pb when noise is superimposed on the reproduced signal. This distribution has two peaks, one having a maximum frequency when Pa = 0 and the other having a maximum frequency when Pb = 0. The value of Pa-Pb when Pa = 0 is -Pst
The value of Pa-Pb when d and Pb = 0 is represented as Pstd. Take the absolute value of Pa-Pb and calculate | Pa-
When Pb | -Pstd is obtained, a distribution as shown in FIG. 6 (b) is obtained.

Assuming that this distribution is a normal distribution, the standard deviation σ of the distribution and the average value Pave are obtained. The standard deviation σ of this distribution and the average value Pave can be used to predict the bit error rate. For example, when the distribution curve estimated as showing the distribution of | Pa-Pb | is gentle and the distribution curve is defined by a function such that the value of | Pa-Pb | can be 0 or less ( That is,
(When the frequency at which | Pa−Pb | takes 0 is not 0),
When it is considered that a decoding error occurs at a frequency according to the probability that the value of | Pa−Pb | becomes 0 or less, the error probability P (σ, Pave) is calculated using the standard deviation σ and the average value Pave.
Can be defined by the following equation (7).

[0056] P (σ, Pave) = erfc (Pstd + Pave / σ)… (7)

As described above, by using the average value Pave and the standard deviation σ obtained from the distribution of Pa-Pb, the error rate of the binarization result by the maximum likelihood decoding method can be predicted.
That is, it is possible to use the average value Pave and the standard deviation σ as an index of reproduced signal quality. In the above example, it is assumed that the distribution of | Pa−Pb | is a normal distribution. However, if it is difficult to consider that the distribution of | Pa−Pb | is a normal distribution, the above is described. Average value P such as
ave and standard deviation σ, instead of | Pa−Pb
The number of times the value of | becomes equal to or less than the predetermined reference value may be counted. The number of counts thus obtained is | P
It can be an index showing the degree of variation of a-Pb |.

As described above, according to this embodiment, a predetermined first state (for example, S
0) to a predetermined second state (e.g., S0), the absolute value of the difference in certainty | Pa-Pb between the two possible paths in the above-described predetermined period.
The reliability of decoding can be determined by calculating |. Furthermore, by measuring | Pa-Pb | multiple times to obtain the degree of dispersion (distribution) of the decoding reliability | Pa-Pb |, it is possible to evaluate the quality of the reproduced signal (predict the bit error rate). it can.

When the signal quality is evaluated by such a method, a combination of state transitions that can take two paths in which an error is most likely to occur (that is, one having a minimum Euclidean distance between the two paths) is used. Selected,
Absolute value of the difference between the probabilities of such two paths | Pa
The signal quality may be evaluated using -Pb |. Hereinafter, this point will be described in detail.

As described above, when the minimum polarity reversal interval is 2, and when a reproduced signal according to the state transition rule using PR (1,2,2,1) equalization is decoded, two paths can be taken. In the range from time k-5 to time k, there are 15 transitions in addition to the above-described transition from S0 _{k -5} to S0 _k . Table 2 below shows the state transition (combination of the state at time k-5 and the state at time k) and the value (Pstd) that can be taken by Pa-Pb in each state transition.

[0061]

[Table 2]

Reliability of the above 16 decoding results Pa-
Pb can be represented by the following formula (8).

(C _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,0,0, x, x, 0,0,0) Pa-Pb = (A _k-4 -B _k-4 ) + (A _k-3 -D _k-3 ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (A _k -B _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _{k- 2} , c _k-1 ,, c _k ) = (0,0,0, x, x, 0,0,1) Pa-Pb = (A _k-4 -B _k-4 ) + (A _{k- 3} -D _k-3 ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (B _k -C _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,0,0, x, 1,1,0) Pa-Pb = (A _k-3 -B _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (E _k -F _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,0,0, x, 1,1,1) Pa-Pb = (A _k-3- B _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (F _k -G _k ) (c _k-6 , c _k-5 , c _{k -4} , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,1,1, x, 0,0,0) Pa-Pb = (E _k-3 -F _k-3 ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (A _k -B _k ) (c _k-6 , c _k-5 , c _{k- 4} , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,1,1, x, 0,0,1) Pa-Pb = (E _k-3 -F _{k -3} ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (B _k -C _k ) (c _k-7 , c _k-6 , c _k-5 , c _{k- 4} , c _k-3 , c _k-2 , c _k-1 ,, _{k k} ) = (0,1,1, x, x, 1,1,0) Pa-Pb = (E _k- 4- F _k-4 ) + (D _k-3 -G _k-3 ) + (C _k-2 -G _k-2 ) + (D _k-1 -G _k-1 ) + (E _k -F _k ) ( c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,1,1, x, x, 1,1,1) Pa-Pb = (E _k-4 -F _k-4 ) + (D _k-3 -G _k-3 ) + (C _k-2 -G _k-2 ) + (D _k-1 -G _k-1 ) + (F _k -G _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _{k -1} , c _k ) = (1,0,0, x, x, 0,0,0) Pa-Pb = (B _k-4 -C _k-4 ) + (A _k-3 -D _{k -3} ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (A _k -B _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,0,0, x, x, 0,0,1) Pa-Pb = (B _k-4 -C _k-4 ) + (A _k-3 -D _k-3 ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (B _k- C _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,0,0, x, 1,1 , 0) Pa-Pb = (B _k-3 -C _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (E _k -F _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,0,0, x, 1,1, When 1) Pa-Pb = (B _k-3 -C _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (F _k -G _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1,1, x, 0,0,0) Pa-Pb = (F _k-3 -G _k-3 ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (A _k -B _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1,1, x, 0,0,1) Pa-Pb = (F _k-3 -G _k-3 ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (B _k -C _k ) (c _k-7 , c _k-6 , c _k-5 , c _{k -4} , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1,1, x, x, 1,1,0) Pa-Pb = (F _k-4 -G _k-4 ) + (D _k-3 -G _k-3 ) + (C _k-2 -G _k-2 ) + (D _k-1 -G _k-1 ) + (E _k -F _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1,1, x, x , 1,1,1) Pa-Pb = (F _k-4 -G _k-4 ) + (D _k-3 -G _k-3 ) + (C _k-2 -G _k-2 ) + ( D _k-1 -G _k-1 ) + (F _k -G _k )… (8)

Incidentally, A_k= (y_k-0)², B_k= (y_k-1)², C_k= (y_k-
2)², D_k= (y_k-3)², E_k= (y_k-Four)², F_k= (y_k-Five) ², G_k= (y_k-6)²so
is there.

When the above equation (8) is divided into the case where Pstd is 10 and the case where Pstd is 36, it is represented by the equation (9) when Pstd = 10, and when Pstsd = 36. It is expressed by equation (10).

(C _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,0,0, x, 1,1 , 0) Pa-Pb = (A _k-3 -B _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (E _k -F _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,0,0, x, 1,1, When 1) Pa-Pb = (A _k-3 -B _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (F _k -G _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,1,1, x, 0,0,0 ), Pa-Pb = (E _k-3 -F _k-3 ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (A _k -B _k ). (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,1,1, x, 0,0,1) Then Pa-Pb = (E _k-3 -F _k-3 ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (B _k -C _k ) ( c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,0,0, x, 1,1,0) When Pa-Pb = (B _k-3 -C _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (E _k -F _k ) (c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,0,0, x, 1,1,1) Pa-Pb = (B _k-3 -C _k-3 ) + (B _k-2 -D _k-2 ) + (D _k-1 -F _k-1 ) + (F _k -G _k ) (c _{k -6} , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1,1, x, 0,0,0) Pa-Pb = (F _k-3 -G _k-3 ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (A _k -B _k ) (c _{k -6} , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1,1, x, 0,0,1) Pa -Pb = (F _k-3 -G _k-3 ) + (D _k-2 -F _k-2 ) + (B _k-1 -D _k-1 ) + (B _k -C _k )… (9)

(C _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,0,0, x, x, 0,0,0) Pa-Pb = (A _k-4 -B _k-4 ) + (A _k-3 -D _k-3 ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (A _k -B _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _{k- 2} , c _k-1 ,, c _k ) = (0,0,0, x, x, 0,0,1) Pa-Pb = (A _k-4 -B _k-4 ) + (A _{k- 3} -D _k-3 ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (B _k -C _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,1,1, x, x, 1,1,0) Pa- Pb = (E _k-4 -F _k-4 ) + (D _k-3 -G _k-3 ) + (C _k-2 -G _k-2 ) + (D _k-1 -G _k-1 ) + (E _k -F _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0, 1,1, x, x, 1,1,1) Pa-Pb = (E _k-4 -F _k-4 ) + (D _k-3 -G _k-3 ) + (C _k-2- G _k-2 ) + (D _k-1 -G _k-1 ) + (F _k -G _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,0,0, x, x, 0,0,0) Pa-Pb = (B _k-4 -C _k-4 ) + (A _k-3 -D _k-3 ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (A _k -B _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,0,0, x, x, 0,0,1) Then Pa-Pb = ( B _k-4 -C _k-4 ) + (A _k-3 -D _k-3 ) + (A _k-2 -E _k-2 ) + (A _k-1 -D _k-1 ) + (B _k -C _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1,1 , x, x, 1,1,0) Pa-Pb = (F _k-4 -G _k-4 ) + (D _k-3 -G _k-3 ) + (C _k-2 -G _{k- 2} ) + (D _k-1 -G _k-1 ) + (E _k -F _k ) (c _k-7 , c _k-6 , c _k-5 , c _k-4 , c _k-3 , c _{k -2} , c _k-1 ,, c _k ) = (1,1,1, x, x, 1,1,1) Pa-Pb = (F _k-4 -G _k-4 ) + (D _{k -3} -G _k-3 ) + (C _k-2 -G _k-2 ) + (D _k-1 -G _k-1 ) + (F _k -G _k )… (10)

Here, in each case, the error rate
Think about getting indicators. Pstd is 10
In such state transition, the maximum likelihood decoding result c_kSatisfies expression (9)
, Pa−Pb, and the standard deviation σ from the distribution_TenAnd average
Value Pave_TenAsk for. On the other hand, Pstd is 36
In such a state transition, the maximum likelihood decoding result c_kSatisfies expression (10)
, Pa−Pb, and the standard deviation σ from the distribution₃₆And average
Value Pave₃₆Ask for. Each distribution is a normal distribution
Assuming that_Ten, P₃₆Is it
They are represented by the following equations (11) and (12), respectively. That is,
The error rate can be estimated for each pattern of the likelihood decoding result.
, Standard deviation σ_TenAnd the average value Pave _Ten, Or
Quasi-deviation σ₃₆And the average value Pave₃₆The quality of the reproduced signal
It can be used as an index.

[0069]

[Equation 4]

[0070]

[Equation 5]

In addition, when the range for detecting the state transition pattern is increased by one time and a combination pattern of state transitions that can take two state transition sequences in the range from time k-6 to time k is detected, The eight patterns shown in Table 3 can be further detected.

[0072]

[Table 3]

Similar to the above equations (11) and (12), the probability P ₁₂ of causing an error in the pattern of Table 3 is given by the equation (13).

[0074]

[Equation 6]

What is important here is the reliability | Pa-Pb |
In order to properly use as an index of reproduction signal quality,
This means that it is only necessary to detect the pattern of state transition that has a high possibility of error (error rate). That is, it is possible to obtain an index that correlates with the error rate without detecting all state transition patterns.

Here, the state transition pattern that is highly likely to be erroneous is a state transition pattern that reduces the maximum value of the reliability | Pa-Pb | (that is, Euclidean which is the absolute distance between the path A and the path B). Pattern that minimizes the distance). Here, Pa or Pb shown in Table 2 is used.
When either one of 0 is 0, Pa-Pb = ± 10
The eight patterns that take

Of the noise included in the reproduced signal, white noise
If Iz is dominant, P_Ten> P₁₂>> P₃₆Tona
Expected to occur. P_TenOnly 1 bit shift error
Other patterns indicate shift error of 2 bits or more.
To taste. Analyze error pattern after PRML processing
And most of them are 1-bit shift errors, so P
_TenAppropriately estimate the error rate of the reproduced signal by Eq. (11) using
it can. In this way, a predetermined state transition is taken during a predetermined period.
Pattern is detected, and in this detected state transition
Standard deviation σ of the distribution of | Pa−Pb | −Pstd_Ten,average
Value Pave_TenIs used as an index to evaluate the quality of the reproduced signal.
It is possible to value.

Although the error rate can be predicted by using the standard deviation σ ₁₀ as described above, for example, the PRML error index MLSA (Maximum Likelihood Sequeque) defined by the following equation (14) is used.
nceAmplitude) may be used as an index indicating signal quality (error rate).

[0079]

[Equation 7]

Here, d ² _min is the square of the minimum value of the Euclidean distances of the two possible paths, and is 10 in the combination of the modulation code of this embodiment and the PRML system. The above-mentioned index MLSA is defined on the assumption that the average value Pave ₁₀ in equation (11) becomes zero. This can be considered that the average value Pave ₁₀ typically takes a value close to 0, and even if the average value Pave ₁₀ is not taken into consideration, an index having a correlation with the error rate can be obtained. Because.

The error rate BER (BitErrorRat) that can be calculated from the index MLSA defined by the equation (14) and the equation (11).
FIG. 16 shows the relationship with e). Similar to the relationship between the jitter and the error rate shown in FIG. 2, it can be seen that the error rate increases as the index MLSA increases. That is, it can be seen that the index MLSA can be used to predict the error rate after PRML processing.

In the above, the general (C0, C1,
As an example of (C1, C0) equalization (C0, C1 is an arbitrary positive number), the case where PR (1, 2, 2, 1) equalization is specifically described, but other ( C0, C
1, C1, C0) equalization (C0, C1 is an arbitrary positive number) is applied, an index having a correlation with the error rate can be obtained by the same procedure as above.

Hereinafter, as a mode different from the above mode,
A recording code having a minimum polarity reversal interval of 2 is used, and PR (C0, C1, C0) equalization (for example, PR
A form to which (1, 2, 1) equalization is applied will be described. Note that C0 and C1 are arbitrary positive numbers.

The recording code (either 0 or 1) is expressed as follows.

Recording code at current time: b _k Recording code before 1 time: b _k-1 Recording code before 2 time: b _k-2

Let Lev be the ideal value of the reproduction signal in the case of PR (C0, C1, C0) equalization.
elv is expressed by the following equation (15).

Levelv = C0 × b _k-2 + C1 × b _k-1 + C0 × b _k (15)

Here, k is an integer representing time and v is an integer from 0 to 3. In addition, the state at time k is S (b _k-1 , b _k )
Then, a state transition table as shown in Table 4 below is obtained.

[0089]

[Table 4]

For simplicity, the state S (0,0) _k at time k is S0 _k , the state S (0,1) _k is S1 _k , and the state S (1,1) _k is S.
2 _k and state S (1,0) _k are written as S3 _k . The state transition in this case is shown in the state transition diagram shown in FIG.
Further, when this is expanded with respect to the time axis, the trellis diagram shown in FIG. 18 is obtained.

Here, the minimum polarity inversion interval of the recording code is 2
And under the condition that PR (C0, C1, C0) equalization is used, two state transitions (path A) are performed when transitioning from a predetermined state at one time to a predetermined state at another time. And there are 6 kinds of state transition patterns (combinations of states) that can take path B) as shown in Table 5.

[0092]

[Table 5]

Here, it is determined which of the path A and the path B is more likely. This judgment is based on the magnitude of the deviation between the ideal sample value (expected value) indicated by the path A and the actual sample value, and the deviation between the ideal sample value (expected value) indicated by the path B and the actual sample value. Can be done by comparing with the size of.

For example, if the state transition S0 _k-3 → S2 _k is estimated, the path A (S0 _k-3 , S0 _k-2, S1 _k-1, S2 _k ) and the path B (S0 _k-3 , S1 _k-2, S2 _k-1, S2 _k ), the state S0 _k-3 is taken at the time k _-3 and the time k
Then take S2 _k . In this case, the difference between the expected value from the time k-2 to the time k and the values y _k-2 , y _k-1 , and y _k of the reproduced signal is 2
Whether the state transition of the path A or the path B is more likely is determined by the cumulative value of the powers. Here, the cumulative value of the square of the difference between each expected value from time k-2 to time k indicated by the path A and the value from the reproduced signal y _k-2 to y _k is Pa.
Let P be the cumulative value of the square of the difference between the expected value of path B from time k-2 to time k and the value of the reproduced signal y _k-2 to y _k.
Assuming b, the cumulative value Pa is expressed by the following equation (16), and the cumulative value Pb is expressed by the following equation (17).

Pa = (y _k-2 −0) ² + (y _k-1 −C0) ² + (y _k − (C1 + C2)) ² (16)

Pb = (y _k−2 −C0) ² + (y _k−1 − (C0 + C1)) ² + (y _k − (2 × C0 + C1)) ² (17)

Here, if Pa << Pb, it is estimated that the possibility of the path A is high, and if Pa >> Pb, it is estimated that the possibility of the path B is high. That is, even when the recording code having the minimum polarity inversion interval of 2 and PR (C0, C1, C0) equalization are used, the reliability of the decoding result can be determined using | Pa-Pb |. Also, the quality of the reproduced signal can be evaluated (error rate estimation) based on the distribution of | Pa−Pb |.

Considering the case where white noise is superimposed on the transmission path, it is considered that the Euclidean distance between the path A and the path B is the shortest in the state transition that is most likely to cause an error. As such two paths that minimize the Euclidean distance, there are two state transition patterns shown in Table 6 below.

[0099]

[Table 6]

Here, when the decoding result is c _k (k is an integer) and the reliability Pa-Pb in the state transition shown in Table 6 is summarized, formula (18) is obtained.

When (c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (0,0, x, 1,1), Pa−Pb = (AA _k-2 -BB _k-2 ) + (BB _k-1 -CC _k-1 ) + (CC _k -DD _k ) (c _k-4 , c _k-3 , c _k-2 , c _k-1 , c _k ) = (1,1, x, 0,0) Pa−Pb = (CC _k−2 −DD _k−2 ) + (BB _k−1 −CC _k−1 ) + (AA _k −BB _k ) ... (18)

Here, AA _k , BB _k , CC _k , and DD _k are represented by the following equations.

AA _k = (y _k −0) ² , BB _k = (y _k −C0) ² , CC _k = (y _k − (C0 + C1)) ² , DD _k = (y _k − (2 × C0 + C1)) ² ,

From the decryption result c _k , Expression (18) is satisfied | Pa
−Pb | − (2 × C0 ² + C1 ² ) is calculated, and the standard deviation σ and the average value Pave are calculated from the distribution. Assuming that the distribution is normal, the probability of error is given by equation (19). Therefore, the error rate of the reproduced signal can be estimated from the standard deviation σ and the average value Pave, and can be used as an index of signal quality.

[0105]

[Equation 8]

In this way, the recording code having the minimum polarity inversion interval of 2 is used and PR (C0, C1,
C0) Even when equalization is applied, the difference in probability | Pa of paths that take a predetermined state transition in a predetermined period | Pa
The quality of the reproduced signal can be evaluated based on −Pb |.

Hereinafter, as a mode different from the above mode,
Codes with minimum polarity inversion interval of 2 and PR (C0, C1, C2,
The form to which C1, C0) equalization is applied will be described. Note that C0, C1, and C2 are arbitrary positive numbers.

The recording code is expressed as follows.

Recording code at current time: b _k Recording code before 1 time: b _k-1 Recording code before 2 time: b _k-2 Recording code before 3 time: b _k-3 Recording code before 4 time : B _k-4

When the ideal value of the reproduction signal in the case of PR (C0, C1, C2, C1, C0) equalization is Levelv, Levelv is expressed by the following equation (20).

Levelv = C0 × b _k-4 + C1 × b _k-3 + C2 × b _k-2 + C1 × b _k-1 + C0 × b _k (20)

Here, k is an integer representing time, and v is an integer from 0 to 8. In addition, the state at time k is S (b _k-3 ,
b _k-2 , b _k-1 , b _k ), a state transition table as shown in Table 7 below is obtained.

[0113]

[Table 7]

For simplicity, state S (0,0,0,0) at time k
_k is S0 _k , state S (0,0,0,1) _k is S1 _k , state S (0,0,1,
1) _k is S2 _k , state S (0,1,1,1) _k is S3 _k , state S (1,1,1
1,1) _k is S4 _k , state S (1,1,1,0) _k is S5 _k , state S (1,
1,0,0) _k is S6 _k , state S (1,0,0,0) _k is S7 _k , state S
(1,0,0,1) _k is _represented as S8 _k , and the state S (0,1,1,0) _k is _represented as S9 _k . The state transition in this case is shown in the state transition diagram shown in FIG. 19, and when this is expanded with respect to the time axis, the trellis diagram shown in FIG. 20 is obtained.

Here, the minimum polarity inversion interval of the recording code is 2
And under the condition that PR (C0, C1, C2, C1, C0) equalization is used, there are two states when transitioning from one state at one time to another state at another time. The state transition patterns (combination of states) that can take transitions (path A and path B) are 9 as shown in Tables 8 to 10.
There are 0 ways.

[0116]

[Table 8]

[0117]

[Table 9]

[0118]

[Table 10]

Note that although Tables 8 to 10 are divided into three tables for convenience, they may be combined into one table.

Here, it is determined which of the path A and the path B is more likely. This judgment is based on the magnitude of the deviation between the ideal sample value (expected value) indicated by the path A and the actual sample value, and the deviation between the ideal sample value (expected value) indicated by the path B and the actual sample value. Can be done by comparing with the size of.

For example, the state transition S0_k-5 → S6_kIs estimated
If there is a transition between path A and path B,
Even if there is, state S0 at time k-5_k-5And at time k
Is S6_kFrom time k-4 to time k
Raw signal value, y_k-4, Y_k-3, Y _k-2, Y_k-1, Y_kAnd expect
Either path A or path B, depending on the cumulative value of the square of the difference
It is determined whether these state transitions are more likely. Path A
The cumulative value of the square of the difference between the expected value and the actual value at
And the square of the difference between the expected value and the actual value on path B
If the cumulative value is Pb, the cumulative value Pa is calculated by the following equation (21).
The cumulative value Pb is represented by the following equation (22).

Pa = (y _k-4 −0) ² + (y _k-3 −C0) ² + (y _k-2 − (C0 + C1)) ² + (y _k−1 − (C0 + C1 + C2)) ² ＋ (y _k − (2 × C1 + C2)) ² … (21)

Pb = (y _k-4 −C0) ² + (y _k-3 − (C0 + C1)) ² + (y _k−2 − (C0 + C1 + C2)) ² + (y _k−1 − (C0 + 2 × C1 + C2)) ² + (y _k − (C0 + 2 × C1 + C2)) ² … (22)

Here, if Pa << Pb, it is estimated that the possibility of the path A is high, and if Pa >> Pb, it is estimated that the possibility of the path B is high. That is, the minimum polarity inversion interval is 2
Even when the code and PR (C0, C1, C2, C1, C0) equalization are used, the reliability of the decoding result can be determined using | Pa-Pb |. Also, the quality of the reproduced signal can be evaluated (error rate estimation) based on the distribution of | Pa−Pb |.

Considering the case where white noise is superimposed on the transmission path, the state transition that is most likely to cause an error is path A.
And the Euclidean distance of the path B becomes the minimum, and there are 16 kinds of state transitions shown in Table 11 below.

[0126]

[Table 11]

[0127] The decoding result of the state transition of the 16 types and c _k (k is an integer), summarized the reliability Pa-Pb in the state transition shown in Table 11 Formula (23) is obtained.

[0128] (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,0,0,0, x, 1,1,0, When 0) Pa−Pb = (AA_k-4-BB_k-4) + (BB_k-3−CC_k-3) + (CC_k-2-EE_k-2) + (DD_k-1-FF _k-1 ) + (DD_k-EE_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,0,0,0, x, 1,1,1, When 0) Pa−Pb = (AA_k-4-BB_k-4) + (BB_k-3−CC_k-3) + (CC_k-2-EE_k-2) + (EE_k-1−GG _k-1 ) + (FF_k−GG_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,0,0,0, x, 1,1,1, When 1) Pa−Pb = (AA_k-4-BB_k-4) + (BB_k-3−CC_k-3) + (CC_k-2-EE_k-2) + (EE_k-1−GG _k-1 ) + (GG_k-JJ_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,0,1,1, x, 0,0,0, When 0) Pa−Pb = (DD_k-4-EE_k-4) + (DD_k-3-FF_k-3) + (CC_k-2-EE_k-2) + (BB_k-1−CC _k-1 ) + (AA_k-BB_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,0,1,1, x, 0,0,0, When 1) Pa−Pb = (DD_k-4-EE_k-4) + (DD_k-3-FF_k-3) + (CC_k-2-EE_k-2) + (BB_k-1−CC _k-1 ) + (BB_k−HH_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,0,1,1, x, 0,0,1, When 1) Pa−Pb = (DD_k-4-EE_k-4) + (DD_k-3-FF_k-3) + (CC_k-2-EE_k-2) + (HH_k-1-II _k-1 ) + (CC_k-II_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,1,1,1, x, 0,0,0, When 0) Pa−Pb = (FF_k-4−GG_k-4) + (EE_k-3−GG_k-3) + (CC_k-2-EE_k-2) + (BB_k-1−CC _k-1 ) + (AA_k-BB_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,1,1,1, x, 0,0,0, When 1) Pa−Pb = (FF_k-4−GG_k-4) + (EE_k-3−GG_k-3) + (CC_k-2-EE_k-2) + (BB_k-1−CC _k-1 ) + (BB_k−HH_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,1,1,1, x, 0,0,1, When 1) Pa−Pb = (FF_k-4−GG_k-4) + (EE_k-3−GG_k-3) + (CC_k-2-EE_k-2) + (HH_k-1-II _k-1 ) + (CC_k-II_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,0,0,0, x, 1,1,0, When 0) Pa−Pb = (BB_k-4−HH_k-4) + (BB_k-3−CC_k-3) + (CC_k-2-EE_k-2) + (DD_k-1-FF _k-1 ) + (DD_k-EE_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,0,0,0, x, 1,1,1, When 0) Pa−Pb = (BB_k-4−HH_k-4) + (BB_k-3−CC_k-3) + (CC_k-2-EE_k-2) + (EE_k-1−GG _k-1 ) + (FF_k−GG_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,0,0,0, x, 1,1,1, When 1) Pa−Pb = (BB_k-4−HH_k-4) + (BB_k-3−CC_k-3) + (CC_k-2-EE_k-2) + (EE_k-1−GG _k-1 ) + (GG_k-JJ_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,1,0,0, x, 1,1,0, When 0) Pa−Pb = (CC_k-4-II_k-4) + (HH_k-3-II_k-3) + (CC_k-2-EE_k-2) + (DD_k-1-FF _k-1 ) + (DD_k-EE_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,1,0,0, x, 1,1,1, When 0) Pa−Pb = (CC_k-4-II_k-4) + (HH_k-3-II_k-3) + (CC_k-2-EE_k-2) + (EE_k-1−GG _k-1 ) + (FF_k−GG_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,1,0,0, x, 1,1,1, When 1) Pa−Pb = (CC_k-4-II_k-4) + (HH_k-3-II_k-3) + (CC_k-2-EE_k-2) + (EE_k-1−GG _k-1 ) + (GG_k-JJ_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,1,1,1, x, 0,0,0, When 0) Pa−Pb = (GG_k-4-JJ_k-4) + (EE_k-3−GG_k-3) + (CC_k-2-EE_k-2) + (BB_k-1−CC _k-1 ) + (AA_k-BB_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (1,1,1,1, x, 0,0,0, When 1) Pa−Pb = (GG_k-4-JJ_k-4) + (EE_k-3−GG_k-3) + (CC_k-2-EE_k-2) + (BB_k-1−CC _k-1 ) + (BB_k−HH_k) (c_k-8, C_k-7, C_k-6, C_k-5, C_k-4, C_k-3, C_k-2, C_k-1, C_k) = (0,1,1,1, x, 0,0,1, When 1) Pa−Pb = (GG_k-4-JJ_k-4) + (EE_k-3−GG_k-3) + (CC_k-2-EE_k-2) + (HH_k-1-II _k-1 ) + (CC_k-II_k)                                                               …(twenty three)

Here, AA _k , BB _k , CC _k , DD _k , EE _k , FF _k ,
GG _k , HH _k , II _k , and JJ _k are represented by the following equations.

AA _k = (y _k −0) ² , BB _k = (y _k −C0) ² , CC _k = (y _k − (C0 + C1)) ² , DD _k = (y _k − (C1 + C2)) ² , EE _k = (y _k − (C0 + C1 + C2)) ² , FF _k = (y _k − (2 × C1 + C2)) ² , GG _k = (y _k − (C0 + 2 × C1 + C2)) ² , HH _k = (y _k −2 × C0)) ² , II _k = (y _k − (2 × C0 + C1)) ² , JJ _k = (y _k − (2 × C0 +2 x C1 + C2)) ²

Expression (23) is satisfied from the decryption result c _k | Pa
−Pb | − (2 × C0 ² + 2 × C1 ² + C2 ² ) is obtained, and the standard deviation σ and the average value Pave are obtained from the distribution. Assuming that this distribution is normal, the probability of error is given by equation (24). Therefore, the standard deviation σ and the average value Pave
Therefore, the error rate of the reproduced signal can be estimated and the signal quality can be evaluated.

[0132]

[Equation 9]

In this way, the recording code having the minimum polarity inversion interval of 2 is used and PR (C0, C1,
Even if (C2, C1, C0) equalization is applied, the quality of the reproduction signal is evaluated based on the difference | Pa-Pb | in the certainty of the path that takes the predetermined state transition in the predetermined period. You can

(Second Embodiment) Hereinafter, the PR shown above will be described.
A specific example of a method of calculating the certainty of each state and the decoding reliability Pa-Pb in the case of performing decoding by the PRML method using (1, 2, 2, 1) equalization will be described in detail.

As described above, PR (1,2,2,1) etc.
When using the truncation, a trellis diagram as shown in Fig. 4 is obtained.
Be done. Here, the confirmation at each time S0 to S5 at time k
L_k ^S0~ L_k ^S5Is a predetermined value at time k-1 as shown below.
Probability of the state of L_k-1 ^S0~ L_k-1 ^S5And the actual at time k
Sample value y_kIt is expressed by the following equation (25) using and. Na
The operator min [xxx, zzz] in the formula below is xxx and zzz
Whichever is smaller, shall be selected.

[0136]

In this embodiment, the probability at time k-1
L_k-1A branch metric (eg (y_k
+3)²) Is always halved and y_k ²/ 2 is subtracted
Shall be done. In the PRML system, the above L_k ^S0
~ L_k ^S5Can be selected by comparing
Therefore, the calculation rule above is L_k ^S0~ L_k ^S5
When it is applied to all the expressions that calculate
It will not be affected. As a result, for each state S0 to S5
Probability L at time k_k ^S0~ L_k ^S5Is expressed by the following equation (26).
Be done.

[0138] ^{_{L k S0 = min [L k}} -1 S0 + (y k +3) 2/2-y k 2/2, L k-1 S5 + (y k +2) 2/2-y k 2 ^{_{/ 2] L k S1 = min}} [L k-1 S0 + (y k +2) 2/2-y k 2/2 ^{_{, L k-1 S5 + (}} y k +1) 2/2-y k 2/2] L k S2 = L k-1 S1 + (y k +0) 2/2-y k 2/2 L k _{^{S3 = min [L k-1}} S3 + (y k -3) 2/2-y k 2/2 ^{_{, L k-1 S2 + (}} y k -2) 2/2-y k 2/2] L k S4 = min [L k-1 S3 + (y k -2) 2/2-y k 2/2 ^{_{, L k-1 S2 + (}} y k -1) 2/2-y k 2/2] L k S5 = L k-1 S4 + (y k +0) 2/2-y k 2/2 ... ( 26)

By expanding this equation (26), the following equation (27) is obtained.

L _k ^S0 = min [L _k-1 ^S0 + 3y _k +9/2, L _k-1 ^S5 + 2y _k +2] L _k ^S1 = min [L _k-1 ^S0 + 2y _k +2, L _k-1 ^S5 + y _k +1/2] L _k ^S2 = L _k-1 ^S1 L _k ^S3 = min [L _k-1 ^S3 -3y _k +9/2, L _k-1 ^S2 -2y _k + 2] L _k ^S4 = min [L _k-1 ^S3 -2y _k +2, L _k-1 ^S2 -y _k +1/2] L _k ^S5 = L _k-1 ^S4 … (27)

Here, A _k , B _k , C _k , E _k , F _k , and G _k are defined as follows.

A _k = 3y _k + 9/2 = (y _k -th4) + (y _k -th5) + (y _k -th6) B _k = 2y _k + 2 = (y _k -th4) + (y _{k -th5) C k = y k} + 1/2 = (y k -th4) E k = -y k + 1/2 = (th3-y k) F k = -2y k + 2 = (th3-y _k ) + (th2-y _k ) G _k = -3y _k + 9/2 = (th3-y _k ) + (th2-y _k ) + (th1-y _k )

Incidentally, th1 = 5/2, th2 = 3/2, th3 = 1/2, th4 = -1 /
2, th5 = -3 / 2, th6 = -5 / 2 are satisfied.

Thus, if L _k ^{S0 to} L _k ^S5 are obtained according to the above equation (27), the sample value y at time k
_{If k} is detected, the difference between the ideal value and the sample value is 2
It is possible to obtain the probabilities L _k ^{S0 to} L _k ^S5 by simple multiplication and addition as shown in A _{k to} G _k without calculating the power. Therefore, there is an advantage that the circuit configuration is not complicated.

Further, as described in the first embodiment, the signal quality can be evaluated by obtaining the difference | Pa-Pb | in the probability of two possible state transition sequences (path A and path B). However, the calculation for obtaining Pa-Pb can also be made relatively simple without including square calculation. Hereinafter, a method of obtaining Pa-Pb will be specifically described.

For example, as described in the first embodiment, P
When R (1,2,2,1) equalization is applied, it is desirable to find Pa-Pb for path A and path B that have the smallest Euclidean distance. That is, among the state transitions shown in Table 2, it is desirable to find Pa-Pb when eight types of state transitions in which Pa-Pb takes ± 10 when Pa = 0 or Pb = 0.

For example, among the above eight state transitions, S0
_The case of obtaining Pa-Pb for the transition of _k-4 → S4 _k will be described. In this case, the path A is S0 → S0 → S1
→ S2 → S4 transition, and path B is S0 → S1 → S2 →
Transition from S3 to S4. At this time, the probability P of path A
a is ^{_{(y k-3 +3) 2}} /2 + (y k-2 +2) 2/2 + (y
represented by ^{_{k-1 +0) 2/2}} + (y k -1) 2/2. Also,
Is the probability Pb of path _{B (y k-3 +2)} 2/2 + (y
^{_{k-2 +0) 2/2}} + (y k-1 -2) 2/2 + (y k -2) 2 /
It is represented by 2.

At this time, Pa-Pb can be expressed by using the above-mentioned A _{k to} G _k . Specifically, Pa-Pb
= Represented by _{(A k-3 -B k-} 3) + B k-2 -F k-1 + (E k -F k). Thus, according to the present embodiment, Pa-Pb
Is expressed using A _{k to} G _k obtained by simple addition and subtraction of the sample value y _k and the set values th1 to th6, and therefore, it is not necessary to perform a square operation, and can be obtained relatively easily. You can Therefore, there is an advantage that the circuit configuration is not complicated.

Although the method of obtaining Pa-Pb for the transition S0 _k-4 → S4 _k has been described above, Pa-Pb for other transitions similarly uses the above A _{k to} G _k . Can be expressed as These examples are shown below.

In the case of the state transition S0 _k-4 → S3 _k , Pa-Pb = (A _k-3- B _k-3 ) + B _k-2 -F _k-1 + (F _k
-G _k ) In the case of state transition S2 _k-4 → S0 _k , Pa-Pb = (E _k-3 -F _k-3 ) -F _k-2 + B _k-1 + (A _k
-B _k ) In the case of state transition S2 _k-4 → S1 _k , Pa-Pb = (E _k-3 -F _k-3 ) -F _k-2 + B _k-1 + (B _k
-C _k ) In the case of state transition S5 _k-4 → S4 _k , Pa-Pb = (B _k-3 -C _k-3 ) + B _k-2 -F _k-1 + (E _k
-F _k ) In the case of state transition S5 _k-4 → S3 _k , Pa-Pb = (B _k-3 -C _k-3 ) + B _k-2 -F _k-1 + (F _k
-G _k ) When the state transition S3 _k-4 → S0 _k , Pa-Pb = (F _k-3 -G _k-3 ) -F _k-2 + B _k-1 + (A _k
-B _k ) In the case of state transition S3 _k-4 → S1 _k , Pa-Pb = (F _k-3 -G _k-3 ) -F _k-2 + B _k-1 + (B _k
-C _k)

(Third Embodiment) An optical disc apparatus 100 for decoding a reproduction signal by the PRML system will be described below with reference to FIG.

In the optical disc device 100, the reproduction signal read from the optical disc 8 by the optical head 50 is amplified by the preamplifier 9, AC-coupled, and then input to an AGC (automatic gain controller) 10. In the AGC 10, the gain is adjusted so that the output of the waveform equalizer 11 in the subsequent stage has a predetermined amplitude. AGC10
The reproduced signal output from is waveform-shaped by the waveform equalizer 11. The waveform-shaped reproduced signal is output to the PLL circuit 12 and the A / D converter 13.

The PLL circuit 12 generates a reproduction clock synchronized with the reproduction signal. The PLL circuit 12 is a conventional PLL circuit (phase comparator 5, LPF 6 shown in FIG. 1).
And a circuit composed of VCO 7). In addition, the A / D converter 13 is the PLL circuit 1
The reproduction signal is sampled in synchronism with the reproduction clock output from 2. The sampling data thus obtained is sent from the A / D converter 13 to the digital filter 1
4 is output.

The frequency characteristic of the digital filter 14 is set so that the frequency characteristic of the recording / reproducing system becomes the characteristic assumed by the Viterbi circuit 15 (PR (1,2,2,1) equalization characteristic in this embodiment). With. The data output from the digital filter 14 is input to the Viterbi circuit 15 which performs maximum likelihood decoding. The Viterbi circuit 15 outputs the binarized data by decoding the PR (1,2,2,1) equalized signal by the maximum likelihood decoding method.

Further, the Viterbi circuit 15 outputs the decoded binarized data and the calculation result (branch metric) of the Euclidean distance for each time to the difference metric analyzer 16. The differential metric analyzer 16 is
The state transition is discriminated from the binarized data obtained from the Viterbi circuit 15, and Pa-Pb indicating the reliability of the decoding result is obtained from the discrimination result and the branch metric. Thereby, the error rate of the decoding result can be estimated.

The Viterbi circuit 1 will be described below with reference to FIG.
5 and the differential metric analyzer 16 are explained in detail
To do. FIG. 8 shows the Viterbi circuit 15 and the differential metric solution.
The structure of the analyzer 16 is shown. Output from digital filter 14
Sampled value y_kIs a branch of the Viterbi circuit 15.
It is input to the trick calculation circuit 17. Branch metric
In the calculation circuit 17, the sampled value y_kAnd the expected value levelv
A branch metric corresponding to the distance is calculated. P
Since R (1,2,2,1) equalization is used, the expected value levelv
Has 7 values from 0 to 6. It at time k
Expected value and sample value y_kA branch that represents the distance from
Metric A_k, B_k, C_k, D_k, E_k, F _k, G_kThat of
Each of them is defined by the following equation (28).

A _k = (y _k -0) ² , B _k = (y _k -1) ² , C _k = (y _k -2) ² , D _k = (y _k -3) ² , E _k = (y _k -4) ² ,, F _k = (y _k -5) ² ,, G _k = (y _k -6) ² … (28)

The branch metric calculated in this way is input to the addition / comparison / selection circuit 18. From the input branch metric at time k and the probability (metric value) of each state at time k−1, the probability of each state S0 to S5 (see FIG. 4) at time k is obtained. The probability of each state is expressed by equation (29). Note that mi
n [xxx, zzz] is a function that selects the smaller value of xxx or zzz.

[0159]

The metric values L _k ^{S0 to} L _k ^{S5 at} the time k are stored in the register 19 and are used to calculate the metric value of each state at the next time k + 1. Also, the circuit 18
The state transition that minimizes the metric value is selected according to equation (29), and the control signals Sel0 to Sel3 are sent to the path memory having the circuit configuration shown in FIG. Output to 20.

L _k-1 ^S0 + A _k ≥ L _k-1 ^S5 + B _k : Sel0 = '1' L _k-1 ^S0 + A _k <L _k-1 ^S5 + B _k : Sel0 = '0' L _k-1 ^S0 + B _k ≥ L _k-1 ^S5 + C _k : Sel1 = '1' L _k-1 ^S0 + B _k <L _k-1 ^S5 + C _k : Sel1 = '0' L _k-1 ^S3 + G _k ≥ L _k-1 ^S2 + F _k : Sel2 = '1' L _k-1 ^S3 + G _k <L _k-1 ^S2 + F _k : Sel2 = '0' L _k-1 ^S3 + F _k ≥ L _k-1 ^S2 + E _k : Sel3 = '1' L _k-1 ^S3 + F _k <L _k-1 ^S2 + E _k : Sel3 = '0'… (30)

The path memory 20 can estimate the most probable state transition sequence according to the state transition rule based on the input control signal, and outputs the binarized data c _k corresponding to this estimated state transition sequence. Output.

On the other hand, in order to evaluate the quality of the reproduced signal, the branch metric output from the branch metric calculation circuit 17 is input to the delay circuit 21, and the signal processing time in the add / compare / select circuit 18 and the path memory 20 is calculated. After being delayed by a minute, the difference metric calculator 22
Is output to. In addition, 2 output from the path memory 20
The binarized data c _k is input to the state transition detector 23, where a predetermined pattern of the binarized data is detected. Specifically, the data pattern corresponding to the eight kinds of state transitions shown in the above equation (9) is detected. When the state transition detector 23 detects a predetermined state transition, the differential metric calculator 22 uses the above equation (9) to
Calculate Pa-Pb for the detected state transition.

It should be noted that Pa-Pb can be obtained by a calculation method that does not include the square operation as described in the second embodiment. According to the method of the second embodiment, Pa−
Pb can be obtained without using the branch metric calculated by the branch metric calculation circuit 17. Therefore, in such a case, a circuit configuration may be adopted in which the sample value y _k output from the digital filter 14 is directly input to the difference metric calculator 22 via only the delay circuit 21. Difference metric calculator 22
Then, Pa-Pb can be obtained from the sampled values y _k according to the method described in the second embodiment.

The value of Pa-Pb for the detected predetermined state transition calculated in this way is input to the average value / standard deviation calculator 24. The average value / standard deviation calculator 24 calculates the average value and standard deviation of the input distribution of Pa-Pb, and calculates these two values (that is, the average value Pav
e ₁₀ and standard deviation σ ₁₀ ) are output. The average value Pave ₁₀ and standard deviation σ ₁₀ output here have the minimum Euclidean distances of the two paths (that is,
It is a value for a predetermined state transition (there is a high possibility of making a mistake in the path). Based on the equation (11), the error rate of the reproduced signal can be estimated from the average value Pave ₁₀ and the standard deviation σ ₁₀ . That is, the standard deviation and the average value obtained from the average value / standard deviation calculator 24 are used as an index indicating the reproduced signal quality correlated with the error rate. As described above, the average value is expected to be close to zero, so P
The error rate may be obtained by regarding ave ₁₀ as zero.

The optical disc apparatus 100 having the configuration shown in FIG. 7 has been described above. The optical disc apparatus has a waveform equalizer B28 having an equalization characteristic suitable for clock reproduction in the PLL circuit as shown in FIG. May be provided. Also in this case,
Similar to the optical disc device 100 shown in FIG.
An average value can be obtained, which makes it possible to evaluate the quality of the reproduced signal. Further, by separately providing the waveform equalizer that performs the waveform shaping suitable for the clock reproduction and the waveform equalizer that performs the waveform shaping suitable for the PRML decoding method, it is possible to generate a preferable reproduced clock signal. , PRML method, it is possible to improve the accuracy of decoding. An optical disc device using two or more waveform equalizers as described above is disclosed in US patent application Ser. No. 09/99 filed by the same applicant as the present applicant.
No. 6,843. No. 09 / 996,843 is incorporated herein by reference.

Further, as shown in FIG. 11, the reproduction clock may be generated from the output (digital signal) of the A / D converter 13. Also in this case, the standard deviation and the average value can be obtained similarly to the optical disc device 100 shown in FIG. 7, and the quality of the reproduced signal can be evaluated by this.

As described above, the standard deviation σ of the Pa-Pb distribution output from the differential metric analyzer 16 and the average value Pave can be used as the index to evaluate the quality of the reproduced signal. It is also possible to perform control for improving the quality of the reproduced signal based on (standard deviation σ and average value Pave). For example, as shown in FIG. 12, by using the frequency characteristic control means 29, the waveform equalizer so that the average value output from the differential metric analyzer 16 becomes 0 or the standard deviation becomes minimum. The reproduced signal quality can be improved by changing the frequency characteristic of 11. Further, in an optical disc device capable of recording information, the recording power and the recording compensation amount are set so that the average value output from the differential metric analyzer 16 becomes 0 or the standard deviation becomes minimum. By controlling, the recording parameters can be optimized.

(Fourth Embodiment) Next, an optical disk device according to a fourth embodiment of the present invention will be described with reference to FIG.

In this embodiment, the differential metric analyzer 1
The reference numeral 60 is configured to output the PRML error index MLSA (M = σ / 2 · d _min ² ) defined by the above equation (14). The PRML error index MLSA is
It is obtained by dividing the standard deviation (mean square error) σ between the most probable state transition sequence and the reproduced signal by the Euclidean distance between the state transition sequence and the second most probable state transition sequence. The PRML error index MLSA is the PRM
This is an index that can favorably evaluate the quality of the reproduced signal when L is used.

The error index MLSA output from the differential metric analyzer 160 is supplied to the frequency characteristic control means 290. The frequency characteristic control means 290 optimizes the characteristic of the waveform equalizer (for example, boost amount or boost center frequency) so that this error index MLSA becomes the minimum.
For example, the boost amount is slightly changed, the PRML error index MLSA at the points before and after the change is compared, and the boost amount with which the MLSA becomes smaller is selected.
By repeating such an operation, the characteristics of the waveform equalizer are optimized and the PRML error index MLSA can be converged to the minimum value.

Further, as shown in FIG. 14, the PRML error index M generated by the difference metric analyzer 160.
The LSA may be supplied to the focus offset search means 291. During signal reproduction, focus servo control is performed so that the beam spot emitted by the optical head 50 can always scan the vicinity of the information recording surface of the optical disk 8. This focus servo control is performed by the servo amplifier 91.
The focus error signal detected by
The focus actuator (not shown) of the optical head 50 is feedback-controlled so as to reach the predetermined target value X0 via the. Here, if the focus offset search means 291 outputs to the subtractor 92 a target value X0 that minimizes the PRML error index MLSA as the predetermined target value X0, the PRML error index MLSA will be the minimum. It is possible to perform focus servo control such that (that is, the error rate is minimized). In order to perform such a search for the target value X0, for example, P when the target value X0 is slightly changed.
The change in the RML error index MLSA may be detected and compared.

Although the focus target value is optimized using the PRML error index MLSA in the present embodiment, the present invention can be applied to the optimization of other servo target values. Using the PRML error index MLSA described above, for example, tracking servo, disc tilt control, lens spherical aberration correction control, etc. can be performed.

Further, as shown in FIG. 15, in an optical disk device having two types of optical heads, a signal reproducing optical head 50 and a signal recording optical head 51, the PRML error index ML output from the differential metric analyzer 160 is output.
The recording power may be controlled using SA. The signal to be recorded on the optical disc is supplied to the signal recording optical head 51 by the recording signal generation means 103 via the modulator 102. The modulator 102 multiplies an appropriate recording power P and the recording signal and supplies the result to the optical head 51. At this time, by supplying the PRML error index MLSA generated by the differential metric analyzer 160 to the recording power control unit 292, the recording power control unit 292 sets the recording power P so that the PRML error index MLSA becomes the minimum. You can decide.

In the optical disk device shown in FIG. 15, the recording operation and the reproducing operation are performed by using different heads, but the function of one head is switched between the recording and the reproducing operation. The operation may be executed. Further, although the example in which the recording power is controlled has been described above, the width and the phase of the recording pulse may be controlled based on the PRML error index MLSA.

[0176]

According to the reproduced signal quality evaluation method of the present invention, in the maximum likelihood decoding method for estimating the most probable state transition sequence from the n types of state transition sequences, the state from the state at time kj to the time The probability of the state transition in a predetermined period until reaching the state of k (for example, the cumulative value of the Euclidean distance in the predetermined period) is Pa, and the time from the state at the time kj of the second most probable state transition sequence is changed to the time. When the probability of state transition in a predetermined period until reaching the state of k (for example, the cumulative value of the Euclidean distance in the predetermined period) is Pb, the reliability of the decoding result from time k−j to time k is | Pa. -Pb | Further, by obtaining the variation of | Pa−Pb | measured multiple times, an index indicating the signal quality correlated with the error rate of the binarization result of maximum likelihood decoding can be obtained.

[Brief description of drawings]

FIG. 1 is a block diagram of a conventional optical disc drive.

FIG. 2 is a diagram showing the relationship between jitter and bit error rate.

FIG. 3 is a minimum polarity reversal interval of 2 used in the embodiment of the present invention.
And state transition diagram determined by PR (1, 2, 2, 1) equalization constraint

FIG. 4 is a minimum polarity inversion interval of 2 used in the embodiment of the present invention.
And the trellis diagram determined by the constraints of PR (1, 2, 2, 1) equalization

FIG. 5 is a diagram showing two possible state transition sequences between states S0 _k and S0 _k-5 in the trellis diagram used in the embodiment of the present invention.

FIG. 6 is a schematic diagram of the distribution of Pa-Pb showing the reliability of the decoding result.

FIG. 7 is a configuration diagram of a reproduced signal quality evaluation device according to a third embodiment of the present invention.

FIG. 8 is a detailed configuration diagram of a Viterbi circuit and a differential metric analyzer of the reproduced signal quality evaluation apparatus according to the third embodiment of the present invention.

FIG. 9 is a configuration diagram of a path memory of the reproduction signal quality evaluation apparatus according to the third embodiment of the present invention.

FIG. 10 is a configuration diagram of another reproduction signal quality evaluation apparatus according to the third embodiment of the present invention.

FIG. 11 is a configuration diagram of still another reproduced signal quality evaluation apparatus according to the third embodiment of the present invention.

FIG. 12 is a configuration diagram of still another reproduced signal quality evaluation apparatus according to the third embodiment of the present invention.

FIG. 13 is a configuration diagram of an optical disk device according to a fourth embodiment of the present invention.

FIG. 14 is a configuration diagram of another optical disk device according to the fourth embodiment of the present invention.

FIG. 15 is a configuration diagram of yet another optical disk device according to Embodiment 4 of the present invention.

FIG. 16: Index MLSA and error rate BER (Bit Error Rate)
Graph showing the relationship with

FIG. 17 is a state transition diagram determined by the fact that the minimum polarity reversal interval used in the embodiment of the present invention is 2 and the constraint of PR (C0, C1, C0) equalization.

FIG. 18 is a trellis diagram determined by the fact that the minimum polarity reversal interval used in the embodiment of the present invention is 2 and the constraint of PR (C0, C1, C0) equalization.

FIG. 19 is a state transition diagram determined by the fact that the minimum polarity reversal interval used in the embodiment of the present invention is 2 and PR (C0, C1, C2, C1, C0) equalization constraints.

FIG. 20 is a trellis diagram determined by the fact that the minimum polarity reversal interval used in the embodiment of the present invention is 2 and PR (C0, C1, C2, C1, C0) equalization constraints.

[Explanation of symbols]

1,8 Optical disc 2 Optical head 3, 11 Waveform equalizer 4 comparator 5, phase comparator 6 LPF 7 VCO 9 preamplifier 10, 28 AGC 12 PLL circuit 13 A / D converter 14 Digital filter 15 Viterbi circuit 16 Differential metric analyzer 17 Branch metric arithmetic circuit 18 Addition / comparison / selection circuit 19 registers 20 pass memory 21 Delay circuit 22 Difference metric calculator 23 State transition detector 24 Selector A 25 Selector B 26, 27 Mean / standard deviation calculator 28 Waveform equalizer B 29 Frequency characteristic control means

Front page continuation (51) Int.Cl. ⁷ Identification code FI theme code (reference) G11B 20/10 321 G11B 20/10 321A 341 341B (72) Inventor Shigeru Komiya 1006 Kadoma, Kadoma-shi, Osaka Matsushita Electric Industrial In-house (72) Inventor Hirohashi Ishibashi 1006 Kadoma, Kadoma-shi, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. F-term (reference) 5D044 BC01 BC02 CC04 FG01 FG02 FG05 GK18 GL32

Claims (15)

[Claims] 1. Time k−j (k is an integer of 3 or more, and j is 2)
Top selects the most likely state transition sequence from among an integer greater than one) n (n is an integer of 2 or more) state transition sequence of as a transition to the second state S _k at time k from the first state S _kj in A method for evaluating the quality of a decoded signal when a reproduced signal is decoded by a likelihood decoding method, wherein the n kinds of state transition sequences are defined in a predetermined period j from the time kj to the time k. The first state S _kj
Detecting a predetermined combination of the second state S _k and the second state S _k, and the most probable first of the n number of state transition sequences defined by the detected predetermined combination.
Let Pa be the index indicating the probability of state transition in the predetermined period j of the state transition sequence of, and the index indicating the probability of state transition in the predetermined period j of the second state transition sequence that is second most likely. Let Pb be | Pa-Pb
Using || to judge the reliability of the decoding result from time k−j to time k. 2. The Pa is defined on the basis of a difference between an expected value and an actual sample value indicated by the first state transition sequence in the predetermined period j, and the Pb is in the predetermined period j. The definition is based on a difference between an expected value indicated by the second state transition sequence and the actual sample value.
The reproduction signal quality evaluation method described in. 3. The Pa is from the time k−j to the time k indicated by the first state transition sequence in the predetermined period j.
Corresponding to the cumulative value of the square of the difference between the expected values l _kj , ..., L _k-1 , l _k and the actual sample values y _kj , ..., y _k-1 , y _k , Pb is an expected value m _kj from time k−j to time k indicated by the second state transition sequence, ...
.., m _k-1 , m _k and the actual sample values y _kj ,.
-, reproduced signal quality evaluation method according to claim 2 corresponding to the square of the cumulative value of the difference between y _k-1, y k. 4. The reproduction signal evaluation method according to claim 1, wherein n = 2. 5. The Euclidean distance between the first state transition sequence and the second state transition sequence has a minimum value.
The reproduction signal quality evaluation method described in. 6. The reproduction signal quality evaluation method according to claim 1, further comprising the step of determining a variation in reliability of the decoding result by measuring | Pa-Pb | a plurality of times. 7. The variation in reliability depends on the | Pa−
The reproduction signal quality evaluation method according to claim 6, wherein the reproduction signal quality evaluation method is performed using the standard deviation of the distribution of Pb |. 8. The variation of the reliability depends on the | Pa−
7. The reproduction signal quality evaluation method according to claim 6, which is indicated by using a standard deviation of Pb | and an average value of the distribution of | Pa−Pb |. 9. The reproduction signal quality evaluation method according to claim 6, wherein the variation in reliability of the decoding result is determined by detecting the frequency at which | Pa−Pb | exceeds a predetermined value. 10. The reproduction signal evaluation according to claim 1, wherein the minimum polarity reversal interval of the recording code is 2 and the reproduction signal PR (C0, C1, C0) equalized is decoded. Method. 11. The reproduction according to claim 1, wherein the minimum polarity reversal interval of the recording code is 2 and the reproduction signal which is PR (C0, C1, C1, C0) equalized is decoded. Signal evaluation method. 12. A recording signal having a minimum polarity reversal interval of 2 and decoding a PR (C0, C1, C2, C1, C0) equalized reproduction signal. Method for evaluating playback signal. 13. When calculating | Pa−Pb |
4. The reproduction signal evaluation method according to claim 2, wherein the square of the actual sampled value is not calculated. 14. A gain controller for adjusting the amplitude value of a reproduced signal, a waveform equalizer for waveform-shaping the reproduced signal so as to have a predetermined equalization characteristic, and a reproduced clock synchronized with the reproduced signal. A reproduced clock generating circuit, an A / D converter which generates sampling data by sampling the reproduced signal with the reproduced clock, and outputs the sampling data, and digital information most likely from the sampling data. And a maximum likelihood detector that decodes the maximum likelihood detector, and Pa is an index indicating the probability of state transition in a predetermined period of the first state transition sequence determined to be most likely by the maximum likelihood detector. When the index indicating the probability of state transition in the predetermined period of the second state transition sequence is Pb, | Pa−P | Information reproducing apparatus and a difference metric calculator for calculating a. 15. The apparatus further comprises an additional waveform equalizer that performs waveform shaping so as to have a predetermined equalization characteristic different from that of the waveform equalizer, and the reproduced clock is waveform-shaped by the additional waveform equalizer. The information reproducing apparatus according to claim 14, wherein the information reproducing apparatus is generated from a shaped reproduced signal.