JPWO2009041598A1 - Signal quality evaluation apparatus, method, and information recording / reproducing apparatus - Google Patents

Abstract

The level average value calculator generates a reference value for branch metric calculation in the Viterbi detector from the detection data string output from the Viterbi detector and the output of the equalizer. The error vector calculator obtains a difference in reference values between paths of a pair of signal sequences that are prone to error, and generates an error vector having values based on the difference in reference values as components. The effective noise calculator calculates an inner product of the error vector and a vector having the Viterbi error time series data obtained by the Viterbi error calculator as an element, and calculates a mean square of the inner products. The evaluation index calculator calculates a signal quality evaluation value based on the magnitude of the error vector and the mean square of the inner product calculated by the effective noise calculator.

Description

  The present invention relates to a signal quality evaluation apparatus, method, and information recording / reproducing apparatus. More specifically, the present invention relates to a signal quality evaluation apparatus that reproduces data obtained by equalizing a reproduced signal obtained from an information recording medium and performing data discrimination by viterbi detection. The present invention relates to a signal quality evaluation apparatus, method, and information recording / reproducing apparatus for evaluating quality.

  Optical discs are classified into two types: read-only optical discs on which data is recorded in advance and optical discs on which information can be recorded. In a read-only optical disk, data is recorded by forming an emboss (physical unevenness) on the optical disk by using an exposure process usually called a mastering process. In a recordable optical disc, recording is performed by irradiating a laser beam condensed on the optical disc to change some physical characteristics of the recording film. Until now, the quality evaluation of signals recorded on an optical disc has been generally performed by measuring the jitter characteristics of a reproduction signal obtained by irradiating the optical disc with a laser. Signal quality evaluation based on jitter is suitable when the zero-cross point of a reproduction signal can be detected with high accuracy, that is, when binarization of information can be stabilized by level slices.

  As the recording density of the optical disc is increased, the signal amplitude of the reproduction signal decreases and it becomes difficult to detect the zero cross point of the reproduction signal. As an effective technique for reproducing data under such a high density condition, a signal detection method called PRML (Partial-Response Maximum-Likelihood) is known. In this method, the reproduced waveform is equalized (PR equalized) to a waveform having intersymbol interference so as not to increase noise components, and data is identified by a method called Viterbi decoding (ML). PR equalization is defined by the amplitude for each data period (clock). For example, PR (a, b, c) has an amplitude at time 0 of a, an amplitude at time T of b, and an amplitude at time 2T. C, and the amplitude at other times is zero. The total number of components whose amplitude is not 0 is called the constraint length.

In PRML, data is decoded by Viterbi decoding using a value obtained by sampling a reproduced waveform at every clock cycle, rather than detecting edge positions and decoding data. For this reason, it is difficult to estimate the PRML detection performance using only temporal fluctuation information of the edge position such as jitter. As a method for evaluating the quality of a reproduction signal in PRML, there is a method described in Patent Document 1. In Patent Document 1, a target signal is obtained from a predetermined data string and predetermined partial response characteristics, an equalization error that is a difference between the target signal and a reproduction signal for each clock cycle is calculated for each clock cycle, and the equivalent signal is calculated. Evaluate signal quality based on error autocorrelation. The predetermined data string here means a 1/0 (or 1 / -1) data string recorded on the medium, and the predetermined partial response characteristic is, for example, PR (1, 2, 2, 2). , 1) means a PR polynomial.
JP 2004-213862 A

  Since PRML is premised on linear intersymbol interference, there is a problem that it is vulnerable to nonlinear distortion such as asymmetry and nonlinear reproduction waveform distortion observed under high recording density. As a technique for improving resistance to non-linear distortion, a technique called adaptive PRML has recently been used. In the adaptive type PRML, for example, (1) equalization is performed using a target signal calculated from a PR polynomial as in the conventional PRML, and (2) a branch at the time of Viterbi detection according to the level distribution characteristic after equalization. The reference value at the time of metric calculation is adaptively changed, and Viterbi detection is performed based on the reference value determined in (3) and (2) to determine data (1/0 (or -1)). Data reproduction (data discrimination) is performed by the procedure described above.

  In order to obtain an index for estimating the error rate in PRML detection, it is necessary to estimate the error rate in Viterbi detection in which data discrimination is performed. For error rate estimation, it is necessary to calculate the Euclidean distance corresponding to the signal amplitude in the level slice and the effective component of noise including distortion (in the case of colored noise, the effective component considering noise autocorrelation). is there. In normal PRML, a reference value (noise 0) at the time of branch metric calculation in Viterbi detection is determined by a convolution operation between a predetermined data string and a predetermined partial response characteristic. In this case, as described in Patent Document 1, the Euclidean distance can be derived from the PR polynomial. Also, a weighting coefficient for noise autocorrelation for calculating the effective noise component in Viterbi detection can be derived from the PR polynomial. Therefore, the signal quality can be evaluated by calculating the ratio between the Euclidean distance and the effective component of noise.

  However, in the adaptive PRML, since the reference value for calculating the branch metric does not necessarily match the target signal level for equalization, a coefficient for multiplying the Euclidean distance and the noise autocorrelation cannot be obtained from the PR polynomial. For this reason, it has been difficult to formulate an index for performing signal quality evaluation. In addition, since there is no appropriate index for evaluating signal quality in adaptive PRML, it is difficult to optimize recording or reproduction conditions under reproduction conditions in which nonlinear distortion occurs.

In Viterbi detection, a distance (branch metric) between a reference output string defined according to a state transition between the equalized reproduction output string u _i and a predetermined data string (branch metric) is calculated, and a data string giving the minimum branch metric And 1 / −1 binary data string determination is performed. An error occurs in Viterbi detection when an erroneous reference output sequence gives the minimum branch metric, and such erroneous detection is likely to occur when the Euclidean distance between the reference output sequences is small. As for the expression of binary data, there are a case where binary data is expressed as 1 / -1, and a case where it is expressed as 1/0. Here, the expression of 1 / -1 is used. The same applies to the case where the 1/0 expression is used.

Hereinafter, the case of PR (1, 2, 2, 2, 1) will be described as an example. As data strings that are likely to be erroneous, consider data string A “11111-1-1-1-1” and data string B “1111-1-1-1-1-1”. The reference output string v ₁ of the data string A is “1, 3, 5, 7, 8, 6, 2, −2, −6, −7, −5, −3, −1”. The reference data string _{v 2} of the data sequence B is "1,3,5,7,6,2, -2, -6, -8, -7, -5, -3, -1". The data string A and the data string B have different bits at the fifth time. As a result, there is a difference of 2 × (1, 2, 2, 2, ₁₎ between the reference output trains v ₁ and v ₂ after the fifth time. The Euclidean distance between the data string A and the data string B is Σ (v _{1, i} −v _{2, i} ) ² = 4 × 14.

Let u _i be the equalized reproduction output sequence input to the Viterbi detector. In the Viterbi detector, the branch metric value M ₁ = Σ (u _i −v _{1, i} ) ² corresponding to the data string A and the distance M ₂ = Σ (u _i −v _{2, i} ) corresponding to the data string B ² is calculated, and the magnitude relationship between the two is compared. If the data string originally recorded on the disc is the data string A, if M ₁ <M ₂ , the data is correctly reproduced. However, if M ₁ > M ₂ due to the influence of noise, the data string B is selected and a detection error occurs. Therefore, the probability that a detection error occurs is equivalent to the probability that DM = M ₂ −M ₁ <0.

The data string had been recorded on the disc with the data sequence A, the time-series data of noise and _{n i,} a _{u i = v 1, i +} n i, DM is
_{DM = Σ (v 1, i} -v 2, i + n i) 2 -Σn i 2 = Σ (v 1, i -v 2, i) 2 -
2Σn _i × (v _{1, i} −v _2i )
It is expressed. If one item is amplitude S, and two items are noise N, the probability of DM <0 is equivalent to the probability of erroneous detection in normal amplitude detection, and the SNR representing the signal quality is the Euclidean distance ED. SNR = S ² / σ ² = ED ² / σ ² . If the noise is white, σ ² = ED × E [n _i ² ], so SNR = ED / E [n _i ² ], the numerator is the Euclidean distance, the denominator is the dispersion of white noise, Reduced to a known SNR. Here, E [] is an operator representing obtaining an average of the numerical values in [].

However, since noise is generally not white, σ ² = 4E [(Σn _i × (v _{1, i} −v _2i )) ² ] must be calculated including autocorrelation. As the error vector EV, EV = v ₁ −v ₂ , and as the noise vector n, n = [n ₁ , n ₂ , n ₃ ,. . . ]
S = Σ (v _{1, i} −v _{2, i} ) ² , N = −2Σ n _i × (v _{1, i} −v _2i ),
If | EV | ² = Σ (v _{1, i} −v _{2, i} ) ² , σ ² = E [N ² ] is substituted into SNR = S ² / σ ² ,
SNR = | EV | ⁴ / (4 × E [(n · EV) ² ]) (“·” represents an inner product)
It becomes. This is a generalized equation for SNR in Viterbi detection. Therefore, if EV can be defined, the SNR in Viterbi detection can be obtained.

  In normal PRML, the reference output sequence is defined by convolution of data and a known (target) PR polynomial, and therefore can be expressed explicitly as EV = (A−B) × H. Here, A and B are vectors corresponding to the data strings A and B. For bit shift errors, AB = 1, 2T for shift errors, AB = [1, 0, −1], for 2T continuous shift errors, AB = [1, 0 for error. , -1, 0, 1], and for PR (1, 2, 2, 2, 1), the EV is 2 [1, 2, 2, 2, 1], 2 [1, 2, 1 , 0, -1, -2, -1], 2 [1, 2, 1, 0, 0, 0, 1, 2, 1]. The SNR calculated using this EV is the SNR disclosed in Patent Document 1.

  In adaptive PRML, in order to improve robustness against non-linear disturbances such as asymmetry, the reference output is defined by the level average value of the equalized reproduction output sequence rather than the convolution of the data sequence and the partial response waveform. For example, in the combination of the modulation method with the run length limit d = 1 and the ML with the constraint length 5, for each of the allowed 16 types of 5 time bit string combinations ([11111], [11110], [11100],...). A reference output is defined. In general, the reference output is obtained by extracting an output corresponding to each bit string from the equalized reproduction waveform and averaging it. Thus, in the adaptive PRML, since the reference output is not uniformly defined, the EV cannot be uniformly defined, and it is difficult to obtain the SNR.

  The present invention relates to a signal quality evaluation apparatus and method capable of obtaining an evaluation index capable of evaluating the signal quality of a reproduction signal by an adaptive PRML in which a reference value for branch metric calculation is not determined based on a PR polynomial. And an information recording / reproducing apparatus.

The present invention provides a quality evaluation method for a reproduction signal in a data reproduction apparatus according to the first aspect, wherein a reproduction signal read from an information recording medium is equalized and data discrimination is performed from the equalized reproduction signal by viterbi detection. And
A series of reference values used for branch metric calculation in Viterbi detection is calculated based on the output distribution of the reproduction signal,
Finding a difference in the reference value corresponding to the state of each clock period between both paths of a signal pair including a signal sequence that is likely to be erroneous, and generating a first vector having a component based on the difference in the reference value;
A second vector having a Viterbi error, which is a difference between the reference value and the equalized reproduction signal, for each clock period, having time series data of the obtained Viterbi error as an element, and having the same number of components as the first vector Produces
An inner product of the first vector and the second vector is obtained, a signal evaluation value is obtained based on the signal amplitude and the inner product, with the magnitude of the first vector as a signal amplitude, and the signal evaluation value Provided is a reproduction signal quality evaluation method characterized by evaluating the signal quality of a reproduction signal.

The present invention is the reproduction signal quality evaluation method in the data reproduction apparatus according to the second aspect, wherein the reproduction signal read from the information recording medium is equalized, and data discrimination is performed from the equalized reproduction signal by viterbi detection. And
A series of reference values used for branch metric calculation in Viterbi detection is calculated based on the output distribution of the reproduction signal,
A Viterbi error, which is a difference between the equalized reproduction signal and the corresponding one of the reference values, is determined for each clock period;
Provided is a reproduction signal quality evaluation method characterized in that an effective noise (σ ² ) exerted on signal quality by a Viterbi error is calculated by a product-sum operation of autocorrelation of the Viterbi error.

The present invention, in a third aspect, is a signal quality evaluation apparatus for evaluating reproduction signal quality in an information reproduction apparatus having an equalizer and a Viterbi detector,
A data string A of length L having a constraint length of Viterbi detection as L and 1 or 0 as a component is represented as A = [a1, a2,. . . , AL], the average value of the reproduction signal output corresponding to the data string A or the average value of the output of the equalizer corresponding to the data string A is calculated for each possible data string A to detect Viterbi A level average calculator for generating a series of reference values for branch metric calculation in
An error vector calculator that obtains a difference between the reference values between paths through which signal sequence pairs including signal sequences that are likely to be erroneous each other pass, and generates a first vector having a value based on the difference between the reference values as a component;
A Viterbi error calculator for obtaining a Viterbi error that is a difference between the equalizer output and the average value calculated by the level average value calculator;
The time series data of the Viterbi error is used as an element, a vector having the same number of components as the first vector is used as a second vector, an inner product between the first vector and the second vector 2 is obtained, and the inner product is obtained. And an effective noise calculator for calculating the root mean square.

In a fourth aspect, the present invention is an information recording / reproducing apparatus that has an equalizer and a Viterbi detector, and performs recording / reproduction of information on / from an optical disc,
A data string A having a length L having a constraint length of Viterbi detection as L, 1 or 0 as a component is represented by A = [a1, a2,. . . , AL], the average value of the reproduction signal output corresponding to the data string A or the average value of the output of the equalizer corresponding to the data string A is calculated for each possible data string A to detect Viterbi A level average calculator for generating a series of reference values for branch metric calculation in
An error vector calculator that obtains a difference between the reference values between paths through which signal sequence pairs including signal sequences that are likely to be erroneous each other pass, and generates a first vector having a value based on the difference between the reference values as a component;
A Viterbi error calculator for obtaining a Viterbi error that is a difference between the equalizer output and the average value calculated by the level average value calculator;
The time series data of the Viterbi error is used as an element, a vector having the same number of components as the first vector is used as a second vector, an inner product of the first vector and the second vector is obtained, An effective noise calculator for calculating the mean square;
An evaluation index calculator for calculating a signal quality evaluation value based on the magnitude of the first vector generated by the error vector calculator and the mean square of the inner product calculated by the effective noise calculator. An information recording / reproducing apparatus is provided.

  In the signal quality evaluation apparatus, method, and information recording / reproducing apparatus of the present invention, recording is performed on an information recording medium under a high recording density condition or a non-linear reproduction waveform, which is difficult to reproduce a signal by normal PRML The quality of the signal can be evaluated.

  The above and other objects, features, and advantages of the present invention will become apparent from the following description with reference to the drawings.

1 is a block diagram showing a configuration of an information reproducing apparatus including a signal quality evaluation apparatus according to an embodiment of the present invention. 3 is a graph showing a relationship between a signal-to-noise ratio (SNR) and a bit error rate (bER) in the first embodiment. The block diagram which shows the structure of the information reproducing | regenerating apparatus containing the signal quality evaluation apparatus of Example 2 of this invention. 10 is a graph showing a relationship between a signal-to-noise ratio (SNR) and a bit error rate (bER) in Example 2. The block diagram which shows the structure of the information recording / reproducing apparatus containing the signal quality evaluation apparatus of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows an information reproducing apparatus including a signal quality evaluation apparatus according to an embodiment of the present invention. The information reproducing apparatus 10 includes at least an A / D converter (ADC) 11, a PLL circuit 12, an equalizer 13, a Viterbi detector 14, a tap coefficient calculator 15, an FIR filter 16, a level average value calculator 17, an error A vector calculator 18, an evaluation index calculator 19, a Viterbi error calculator 20, and an effective noise calculator 21 are included. In FIG. 1, the output of the A / D converter 11 is input to the PLL circuit 12, but a configuration in which the order of the A / D converter 11 and the PLL circuit 12 is reversed is also possible. A pre-equalizer that corrects the frequency characteristics of the signal and the like may be provided in the signal path leading to the equalizer 13.

  The A / D converter 11 converts a reproduction signal obtained by reproducing the information medium into a digital signal. The PLL circuit 12 receives the output of the A / D converter 11, performs phase control so as to be synchronized with the input channel frequency, and outputs a sampling signal interpolated from a data string having a fixed sampling rate. The equalizer 13 equalizes the waveform of the output of the PLL circuit 12 and outputs it. The Viterbi detector 14 receives the equalized signal output from the equalizer 13 and outputs binary detection data (1 / -1) corresponding to the equalized signal by Viterbi detection.

  The FIR filter 16 performs a convolution operation on the output data string of the Viterbi detector 14 and a predetermined partial response waveform, for example, the illustrated response waveform (1, 2, 2, 2, 1), and outputs an equalization target signal. Generate. The tap coefficient calculator 15 calculates the tap coefficient in the equalizer 13 so that the square error between the equalized signal output from the equalizer 13 and the equalization target signal input from the FIR filter 16 is substantially minimized. Control.

  The level average value calculator 17 receives the equalized signal output from the equalizer 13 and the detection data output from the Viterbi detector 14. The level average value calculator 17 classifies the detected data into data strings for each constraint length (“5” if PR (1, 2, 2, 2, 1)), and equalizes each classified data string Calculate the average value of the signal. The Viterbi detector 14 performs maximum likelihood detection using the average value of the equalized signal for each data string classified by the level average value calculator 17 as a reference value at the time of branch metric calculation.

The reference value when calculating the branch metric in the adaptive PRML with the constraint length “5” can be expressed as follows:
[-1-1-1-1-1] → Z1;
[1-1-1-1-1] → Z2;
[1-1-1-11] → Z3;
[-1-1-1-1] → Z4;
[-1-1-111] → Z5;
[1-1-111] → Z6;
[-1-111-1] → Z7;
[-1-1111] → Z8;
[11-1-1-1] → Z9;
[11-1-11] → Z10;
[−111-1-1] → Z11;
[111-1-1] → Z12;
[-1111-1] → Z13;
[1111-1] → Z14;
[-11111] → Z15;
[11111] → Z16.
In normal PRML, branch metrics (Z1 to Z16) are determined by convolution of a PR polynomial and a data string (for example, [-1-1-1-1]). On the other hand, in the adaptive PRML, Z1 to Z16 are adaptively determined based on the characteristics of the reproduction signal. The level average value calculator 17 calculates Z1 to Z16.

The level average value calculator 17 examines the output data string of the Viterbi detector 14, takes the average value of the equalized reproduction signal output corresponding to the data string for each predetermined data string, and performs the above Z1 to Z16. Ask for. For example, the level average value calculator 17 calculates the level average value Z16 corresponding to the data string [11111] by E [w _i (11111)]. Here, w _i (11111) represents a value corresponding to the data string [11111] in the output w _i of the equalizer 13. E [] is an operator representing obtaining an average of the numerical values in []. Note that the level average value calculator 17 may obtain Z1 to Z16 based on the average value of the reproduction signal before equalization instead of the equalization signal.

  The error vector calculator 18 calculates an error vector EV based on the output of the level average value calculator 17. Hereinafter, a method for obtaining the error vector EV by the adaptive PRML will be described. In the following, the case where the constraint length is “5” and the run length limit is “d = 1” will be described, but the calculation procedure does not change even when the constraint length and the run length limit are different.

An error that is likely to cause an error in Viterbi detection is a bit shift error in which one bit is different between two types of data strings (signal pairs), a 2T shift error in which a 2T bit is shifted, and a 2T bit is successively shifted. It can be almost integrated into 3 types of 2T continuous shift errors. These three types of errors are further classified into two types according to the expansion / contraction of the mark and the expansion / contraction of the space, so that there are a total of 3 × 2 = 6 types of error prone errors. These error patterns can be classified into the data strings shown in Table 1 when the constraint length is 5 and the run length limit is d = 1. In Table 1, the data notation is not 1 / -1, but 1/0.

Among the patterns corresponding to the bit shift errors corresponding to the expansion / contraction of the mark in Table 1, the data string A “11111-1-1-1-1” and the data string B “1111-1-1-1-1” Consider the combination with “−1”. Table 2 shows the relationship between the data string and the branch metric reference output string. In the data string A, 5-bit time-series data corresponding to the constraint length 5 is [11111], [1111-1], [111-1-1], [11-1-1-1] [1-1- 1-1-1], the branch metric reference output strings are Z16, Z14, Z12, Z9, and Z2. On the other hand, in the data string B, the 5-bit time series data is [1111-1] [111-1-1] [11-1-1-1] [1-1-1-1-1] [-1- 1-1-1-1], the branch metric reference output sequences are Z14, Z12, Z9, Z2, and Z1. Accordingly, the error vector between the data string A and the data string B is obtained by (Z16-Z14, Z14-Z12, Z12-Z9, Z9-Z2, Z2-Z1).

For the other 8 types of bit shift error patterns corresponding to the bit shift error corresponding to the expansion / contraction of the mark among the signal pairs whose data strings are shown in Table 1, corresponding error vectors are calculated in the same manner as described above. be able to. Since the nine types of error vectors corresponding to the bit shift error corresponding to the expansion / contraction of the mark are not necessarily the same, the error vector EV (1) for the bit shift error is calculated by averaging the nine types of error vectors. It will be defined. Since the appearance frequency of the data string shown in Table 1 is not so different for each pattern, it is considered reasonable to estimate the detection performance by averaging the corresponding error vectors. Summarizing the results of calculating the average of nine error vectors,
EV (1) = [1/3 (Z8 + Z15 + Z16-Z7-Z13-Z14), 1/3 (Z
13 + 2 * Z14-Z11-2 * Z12), Z12-Z9, 1/3 (Z10 + 2 * Z9-Z3-2 * Z2), 1/3 (Z2 + Z3 + Z6-Z1-Z4-Z5)]] (1)
It becomes.

Error vector EV (2) when space expands and contracts Error vector EV (3) when 2T mark shifts Error vector EV (4) when 2T space shifts, 2T mark, 2T space shifts The error vector EV (5) in the case, the 2T space, and the error vector EV (6) in the case where the 2T mark shifts are also obtained as an average between the nine types of error vectors, as described above. That is, EV (2) to EV (6) are obtained by the following formulas 2 to 6.
EV (2) = [1/3 (Z3 + Z4 + Z10-Z1-Z2-Z9), 1/3 (Z6 + 2 * Z5-Z3-2 * Z4), Z8-Z5, 1/3 (Z13 + 2 * Z15-Z7-2 * Z8), 1/3 (Z12 + Z14 + Z16-Z11-Z13-Z15)] (2)
EV (3) = [1/3 (Z8 + Z15 + Z16-Z7-Z13-Z14), 1/3 (Z
13 + 2 * Z14-Z11-2 * Z12), Z12-Z10, Z10-Z6, Z6-Z8, 1/3 (Z7 + 2 * Z8-Z13-2 * Z15), 1/3 (Z11 + Z13 + Z15-Z12-Z14-Z16) ] (3)
EV (4) = [1/3 (Z3 + Z4 + Z10-Z1-Z2-Z9), 1/3 (Z6 + 2 * Z5-Z3-2 * Z4), Z7-Z5, Z11-Z7, Z9-Z11, 1/3 ( Z3 + 2 * Z2-Z10-2 * Z9), 1/3 (Z1 + Z4 + Z5-Z2-Z3-Z6)]
(4)
EV (5) = [1/3 (Z8 + Z15 + Z16-Z7-Z13-Z14), 1/3 (Z
13 + 2 * Z14-Z11-2 * Z12), Z12-Z10, Z10-Z6, Z6-Z7, Z7-Z11, Z11-Z9, 1/3 (Z10 + 2 * Z9-Z3-2 * Z2), 1/3 ( Z2 + Z3 + Z6-Z1-Z4-Z5)] (5)
EV (6) = [1/3 (Z3 + Z4 + Z10-Z1-Z2-Z9), 1/3 (Z6 + 2 * Z5-Z3-2 * Z4), Z7-Z5, Z11-Z7, Z10-Z11, Z6-Z10, Z8-Z6, 1/3 (Z13 + 2 * Z15-Z7-2 * Z8), 1/3 (Z12 + Z14 + Z16-Z11-Z13-Z15)] (6)
The error vector calculator 18 applies Z1 to Z16 obtained by the level average value calculator 17 to Equations 1 to 6, and calculates six EVs (1) to EV (6). The data string in Table 1 is used to determine coefficients for calculating the error vectors EV (1) to EV (6) based on the branch metrics Z1 to Z16.

The Viterbi error calculator 20 calculates a Viterbi detection error defined by the difference between the output of the equalizer 13 and the output of the level average value calculator 17. The effective noise calculator 21 uses the error vector and the Viterbi detection error to calculate an effective noise component σ ² for each of the six types of error prone to error. The evaluation index calculator 19 calculates the SNR for each of the six types of error prone to error based on the error vector and the effective noise component σ ² . The minimum value among the six calculated SNRs is defined as an evaluation index representing signal quality. The SNR is calculated using SNR = | EV | ⁴ / {4 × E [(n · EV) ² ])}. EV in this equation is an error vector, and n is a Viterbi error defined by the difference between the equalized signal and the reference value at the time of branch metric calculation.

であるので、ＳＮＲ（１）は、下記式７で表される。

式７において、ｎ_ｉは時刻ｉにおけるビタビ誤差であり、Ｍは実効ノイズを算出するためのサンプル数である。Ｍが十分に大きければ、ＳＮＲは、ほぼＭに依存せず収束した値となる。Ｍは、例えば１００００以上の値にする。Hereinafter, six types of SNR calculation will be described. First, the SNR corresponding to the bit shift error will be described. The error vector EV (1) corresponding to the bit shift _{error, EV (1) = [e} 1_1, e 1_2, e 1_3, e 1_4, e 1_5] and. From Equation 1,
e _{1_1} = 1/3 (Z8 + Z15 + Z16-Z7-Z13-Z14),
_{e 1_2 = 1/3 (Z13} + 2 × Z14-Z11-2 × Z12),
_{e1_3} = 1/3 (Z2 + Z3 + Z6-Z1-Z4-Z5),
e ₁ — ₄ = Z12-Z9,
e 1 — ₅ = 1/3 (Z10 + 2 × Z9−Z3−2 × Z2).
When the SNR obtained by substituting this EV (1) into SNR = | EV | ⁴ / (4 × E [(n · EV) ² ]) is SNR (1),

Therefore, SNR (1) is represented by the following formula 7.

In Equation 7, n _i is the Viterbi error at time i, M is the number of samples for calculating the effective noise. If M is sufficiently large, the SNR becomes a converged value almost independent of M. For example, M is set to a value of 10,000 or more.

評価指標算出器１９は、上記式７〜１２により、ＳＮＲ（１）〜ＳＮＲ（６）を算出し、
ＳＮＲ（１）〜ＳＮＲ（６）のうちの最小値を、信号品質指標とする。Similarly, regarding other easily errorable errors, SNR (2) to SNR (6) can be obtained by substituting EV (2) to EV (6) into the SNR calculation formula. SNR (2) to SNR (6) are obtained by replacing each element of EV (2) to EV (6) with EV (i) = [e _{i_1} e _{i_2} e _{i_3} . . . ] Can be represented by the following formulas 8-12.

The evaluation index calculator 19 calculates SNR (1) to SNR (6) by the above formulas 7 to 12,
The minimum value among SNR (1) to SNR (6) is set as a signal quality index.

  According to the study by the present inventors, SNR (1) and SNR (2) are almost close values, and SNR (5) and SNR (6) are often close values. I understood. Therefore, it is not always necessary to calculate the six types of SNRs in the evaluation index calculator 19, and one of SNR (1) and SNR (2), SNR (3), SNR (4), and Any one of SNR (5) and SNR (6) may be calculated. That is, the SNR calculation may be performed for at least four types of errors that are prone to error.

と定義すれば、式７〜式１２の分母はΣβ_ｉＲ_ｉの形で表現できる。従って、実効ノイズ
σ^２は、

で算出できる。実効ノイズ算出器２１は、このような計算を用いて、実効ノイズ成分σ^２を算出してもよい。この場合、評価指標算出器１９におけるＳＮＲの算出式は、

となる。By the way, the denominators of Equation 7 to Equation 12 are
Σ (n _{i e_1_1} + n _{i + 1} e _{i_2} + n _{i + 2} e _{i_3} + ...) ²
It has become. If this equation is expanded and the number of samples is large enough,
Using the fact that E [n _{i i + m} ] = E [n _{i + 1} n _{i + 1 + m} ] = E [n _{i + k} n _{i + k + m} ] (k and m are arbitrary integers greater than or equal to 0), autocorrelation R _m is

In other words, the denominators of Equations 7 to 12 can be expressed in the form of Σβ _i R _i . Therefore, the effective noise σ ² is

It can be calculated by The effective noise calculator 21 may calculate the effective noise component σ ² using such a calculation. In this case, the calculation formula of the SNR in the evaluation index calculator 19 is

It becomes.

  As described above, in the reproduction signal quality evaluation method of the above embodiment, a series of reference values used for branch metric calculation in the Viterbi detection, which is the basic configuration of the first aspect of the method of the present invention, is used. A difference between reference values corresponding to the state of each clock period is calculated between both paths of a signal pair including a signal sequence that is likely to be erroneous, calculated based on the output distribution, and a component having a component based on the difference between the reference values is obtained. 1 vector is generated, a Viterbi error, which is the difference between the reference value and the equalized reproduction signal, is determined for each clock period, the time series data of the determined Viterbi error is used as an element, and the same number of components as the first vector is obtained. A second vector is generated, an inner product of the first vector and the second vector is obtained, and a signal evaluation value is obtained based on the signal amplitude and the inner product, with the magnitude of the first vector as the signal amplitude. By adopting a configuration that evaluates the signal quality of the reproduced signal based on the signal evaluation value, recording is performed on the information recording medium under a high recording density condition or a non-linear reproduction waveform, which is difficult to reproduce the signal by normal PRML. The quality of the received signal can be appropriately evaluated.

Further, in the modified example of the reproduction signal quality evaluation method of the above embodiment, a series of reference values used for branch metric calculation in Viterbi detection is calculated based on the output distribution of the reproduction signal, and one corresponding to the equalized reproduction signal is obtained. It is also possible to employ a configuration in which a Viterbi error that is a difference from the reference value is obtained for each clock period, and an effective noise (σ ² ) that the Viterbi error has on signal quality is calculated by a product-sum operation of the autocorrelation of the Viterbi error. The same effect can be achieved.
In the quality evaluation apparatus of the above embodiment, first, the level average value calculator 17 determines the output sequence output from the Viterbi detector 14. Next, it is determined which reference value of the branch metrics Z1 to Z16 is the equalization target value, and the Viterbi error calculator 20 calculates the output of the equalizer 13 and the output of the equalizer 13 of the equalization target value. The difference is calculated as a Viterbi error (equalization error). Thereafter, the effective noise calculator 21 calculates an evaluation index based on the SNR definition formula. The information reproducing apparatus employs recording conditions and reproducing conditions that maximize the evaluation index.

  Examples will be described below. First, Example 1 will be described. A phase change optical disk is formed on a polycarbonate substrate having a thickness of 0.6 mm, a guide groove having a pitch of 0.4 μm, and a depth of 25 nm, and an optical head having a wavelength of 405 nm and an objective lens NA = 0.65 is used. Recording and playback evaluation was performed. (1-7) modulated data was recorded at a linear velocity of 5 m / s and a clock frequency of 64.8 MHz, detected by PRML, and an error rate (bit error rate: bER) was evaluated. Under this recording condition, the length of the shortest mark formed on the disc is about 100 nm, which is a condition with a very high recording density equivalent to the optical cut-off.

The partial response characteristic for equalization was (1, 2, 2, 2, 1), and Viterbi detection was performed using the level average value of the equalized reproduction signal as a reference value at the time of branch metric calculation. Table 3 below shows reference values for branch metric calculation. In the case of Viterbi detection for normal PR (1, 2, 2, 2, 1), when normalized so that the amplitude becomes ± 1, the reference value is “−1, −0.75, −0.5. , -0.25, 0, 0.25, 0.5, 0.75, 1 ". The averaged level output average value under this condition (reference values Z1 to Z16 at the time of branch metric calculation) is as shown in Table 3, which is different from nine values in normal PRML. .

Table 4 shows error vectors obtained by substituting the reference values Z1 to Z16 of Table 2 into Equations 1 to 6. Further, when the coefficient to be multiplied by the autocorrelation of the Viterbi error, that is, the coefficient β in Expression 13 is calculated based on the error vector in Table 4, the following Table 5 is obtained.

表６から、本実施例条件下での再生信号品質を表すＳＮＲは１７．３であることが分かる。一方、実際のｂＥＲは、４×１０^−５であった。Table 6 shows SNR (1) to SNR (6) calculated by calculating the error vectors in Table 4 using Expressions 7 to 12. Note that the same result can be obtained by substituting β in Table 5 into Equation 13 to obtain SNR (1) to SNR (6).

From Table 6, it can be seen that the SNR representing the reproduction signal quality under the conditions of this example is 17.3. On the other hand, the actual bER was 4 × 10 ⁻⁵ .

  The recording density was changed by changing the recording linear velocity between 4.8 m / s and 5.1 m / s. Under each condition, the SNR and bER were measured in the same manner as described above, and the relationship between the two was examined. . FIG. 2 shows the results. The horizontal axis in FIG. 2 represents the minimum value of SNR (1) to SNR (6) under each condition, and the vertical axis represents bER under each condition. Referring to FIG. 2, the quality evaluation index (minimum value of SNR) calculated by the evaluation index calculator 19 (FIG. 1) and bER correlate very well and are effective as a signal quality evaluation index. Can be confirmed.

  Next, Example 2 will be described. In Example 1, the reference value at the time of branch metric calculation was obtained from the equalized reproduction output according to the configuration of the above embodiment. In this example, according to a modification of the above embodiment, a reference value at the time of branch metric calculation is obtained from a reproduction signal before equalization. The configuration of this modification is shown in FIG. The flow until the evaluation index is calculated using the output of the level average value calculator 17 is the same in FIG. 1 and FIG. The difference between the configuration of FIG. 3 and the configuration shown in FIG. 1 is that the input of the level average value calculator 17 is changed from the output of the equalizer 13 to the output of the PLL circuit 12. In FIG. 1, the FIR filter 16 performs a convolution operation between the output data string of the Viterbi detector 14 and a predetermined partial response waveform, and the resulting signal is used as an equalization target signal. On the other hand, in the configuration of FIG. 3, the output of the level average value calculator 17 is used as an equalization target.

表７の基準値を、式１〜式６に代入してエラーベクトルを算出すると、表８の通りとなる。また、表８のエラーベクトルに基づいてビタビ誤差の自己相関に乗ずる係数、すなわち、式１３における係数βを算出すると、表９の通りとなる。

Hereinafter, the result of evaluating the reproduction signal under the condition of the linear velocity of 5 m / s used in Example 1 with the configuration of FIG. 3 will be described. The branch metric reference values obtained from the output of the PLL circuit 12 are as shown in Table 7.

When the error values are calculated by substituting the reference values in Table 7 into Equations 1 to 6, Table 8 is obtained. Further, when the coefficient to be multiplied by the autocorrelation of the Viterbi error based on the error vector of Table 8, that is, the coefficient β in Equation 13, is calculated as shown in Table 9.

表１０から、本実施例条件下での再生信号品質を表すＳＮＲは２２．８であることが分かる。一方、実際のｂＥＲは、３×１０^−６であった。Table 10 shows SNR (1) to SNR (6) calculated by calculating the error vectors in Table 8 using Equations 7 to 12. Note that the same result can be obtained by substituting β in Table 9 into Equation 13 to obtain SNR (1) to SNR (6).

From Table 10, it can be seen that the SNR representing the reproduction signal quality under the conditions of this example is 22.8. On the other hand, the actual bER was 3 × 10 ⁻⁶ .

The recording density is changed by changing the recording linear velocity between 4.8 m / s and 5.1 m / s,
Under each condition, SNR and bER were measured in the same manner as described above, and the relationship between the two was examined. FIG. 4 shows the results. The horizontal axis in FIG. 4 represents the minimum value of SNR (1) to SNR (6) under each condition, and the vertical axis represents bER under each condition. FIG. 4 also shows the results of Example 1. Referring to FIG. 4, the quality calculated by the evaluation index calculator 19 both when the reference value at the time of branch metric calculation is calculated from the reproduction signal before equalization and when it is calculated from the reproduction signal after equalization. The evaluation index (minimum value of SNR) and bER correlate very well, and it can be confirmed that the evaluation index is effective as a signal quality evaluation index. In FIG. 4, since no error was detected under the best SNR condition, data is shown at 10 ⁻⁷ for convenience.

  The evaluation index calculated by the evaluation index calculator 19 can be used to optimize various conditions for recording or reproduction. FIG. 5 shows a configuration of an information recording / reproducing apparatus including a signal quality evaluation apparatus. A reproduction signal from the optical disk is input to the A / D converter 11 (FIG. 1 or 3). The evaluation index calculated by the evaluation index calculator 19 is input to the controller 31 that adjusts the recording / playback conditions. The controller 31 controls the recording or reproducing conditions in the optical head 32 so that the input evaluation index is maximized when recording on the optical disk or reproducing information from the optical disk.

  For example, when reproducing the data recorded in the ROM, the controller 31 controls the tilt in the optical head 32 so that the evaluation index becomes maximum. By doing so, it is possible to perform reproduction under the condition that the error rate is minimized. Further, when recording on a rewritable or write-once optical disc, the controller 31 checks the relationship between the recording strategy parameter and the evaluation index, and performs recording with the strategy parameter that maximizes the evaluation index. By doing so, it is possible to perform recording and reproduction under conditions that minimize the error rate even under conditions where nonlinear distortion occurs in the reproduction waveform.

  Although the invention has been particularly shown and described with reference to illustrative embodiments, the invention is not limited to these embodiments and variations thereof. It will be apparent to those skilled in the art that various modifications can be made to the present invention without departing from the spirit and scope of the invention as defined in the appended claims.

  This application is based on and claims the priority of Japanese Patent Application No. 2007-249614 filed on Sep. 26, 2007, the entire disclosure of which is incorporated herein by reference. join.

Claims (14)

A reproduction signal quality evaluation method in a data reproduction apparatus for equalizing a reproduction signal read from an information recording medium and performing data discrimination by viterbi detection from the equalized reproduction signal,
A series of reference values used for branch metric calculation in Viterbi detection is calculated based on the output distribution of the reproduction signal,
Finding a difference in the reference value corresponding to the state of each clock period between both paths of a signal pair including a signal sequence that is likely to be erroneous, and generating a first vector having a component based on the difference in the reference value;
A second vector having a Viterbi error, which is a difference between the reference value and the equalized reproduction signal, for each clock period, having time series data of the obtained Viterbi error as an element, and having the same number of components as the first vector Produces
An inner product of the first vector and the second vector is obtained, a signal evaluation value is obtained based on the signal amplitude and the inner product, with the magnitude of the first vector as a signal amplitude, and the signal evaluation value A reproduction signal quality evaluation method characterized by evaluating the signal quality of a reproduction signal. The signal evaluation value is expressed by using the first vector as EV, the second vector as n, and E [] as a function for obtaining an average value.
SNR = | EV | ⁴ / {4 × E [(n · EV) ² ]}
The reproduction signal quality evaluation method according to claim 1, wherein the reproduction signal quality evaluation method is calculated by:   3. The reproduction according to claim 1, wherein a difference between the reference values is obtained for a plurality of signal sequence pairs, and an average value of the obtained reference value differences is used as a component of the first vector. Signal quality evaluation method.   The plurality of signal sequence pairs are divided into a plurality of patterns corresponding to the cause of signal error occurrence, and the first vector is obtained for each of the plurality of divided patterns. The reproduction signal quality evaluation method described.   Using the first vector obtained for each of the plurality of divided patterns, the signal evaluation value is obtained for each of the plurality of patterns, and the reproduction signal is obtained using the minimum value of the obtained signal evaluation values. 5. The reproduction signal quality evaluation method according to claim 4, wherein the signal quality is evaluated.   A data string A of length L having a constraint length of Viterbi detection as L and 1 or 0 as a component is represented as A = [a1, a2,. . . , AL] for each possible data string A, an average value of the reproduction signal output corresponding to the data string A is obtained, and the obtained average value is used as the series of reference values. The reproduction signal quality evaluation method according to any one of 1 to 5.   A data string A of length L having a constraint length of Viterbi detection as L and 1 or 0 as a component is represented as A = [a1, a2,. . . , AL] for each possible data string A, an average value of the equalized reproduction signal corresponding to the data string A is obtained, and the obtained average value is used as the series of reference values. Item 6. The reproduction signal quality evaluation method according to any one of Items 1 to 5. A reproduction signal quality evaluation method in a data reproduction apparatus for equalizing a reproduction signal read from an information recording medium and performing data discrimination by viterbi detection from the equalized reproduction signal,
A series of reference values used for branch metric calculation in Viterbi detection is calculated based on the output distribution of the reproduction signal,
A Viterbi error, which is a difference between the equalized reproduction signal and the corresponding one of the reference values, is determined for each clock period;
An effective noise (σ ² ) effected by Viterbi error on signal quality is calculated by a product-sum operation of autocorrelation of the Viterbi error. The autocorrelation of the Viterbi error is R _i , the multiplication coefficient in the product-sum operation is β _i , and the effective noise σ ² is
σ ² = Σ _{i = 1, m} β _i × R _i (m is a natural number of 2 or more)
The reproduction signal quality evaluation method according to claim 8, wherein the reproduction signal quality evaluation method is calculated by: 9. The multiplication coefficient in the product-sum operation is prepared for each pattern corresponding to at least four types of signal error occurrences, and the effective noise σ ² is calculated for each pattern. 10. The reproduction signal quality evaluation method according to 9. A difference between the reference values is obtained for both paths of a signal sequence pair including signal sequences that are likely to be erroneous, a first vector having a value based on the difference between the reference values as a component is generated, and the magnitude of the first vector The reproduction signal quality evaluation method according to claim 8, wherein a signal evaluation value is obtained based on the effective noise σ ² . A signal quality evaluation apparatus for evaluating reproduction signal quality in an information reproduction apparatus having an equalizer and a Viterbi detector,
A data string A of length L having a constraint length of Viterbi detection as L and 1 or 0 as a component is represented as A = [a1, a2,. . . , AL], the average value of the reproduction signal output corresponding to the data string A or the average value of the output of the equalizer corresponding to the data string A is calculated for each possible data string A to detect Viterbi A level average calculator for generating a series of reference values for branch metric calculation in
An error vector calculator that obtains a difference between the reference values between paths through which signal sequence pairs including signal sequences that are likely to be erroneous each other pass, and generates a first vector having a value based on the difference between the reference values as a component;
A Viterbi error calculator for obtaining a Viterbi error that is a difference between the equalizer output and the average value calculated by the level average value calculator;
The time series data of the Viterbi error is used as an element, a vector having the same number of components as the first vector is used as a second vector, an inner product between the first vector and the second vector 2 is obtained, and the inner product is obtained. A signal quality evaluation apparatus comprising: an effective noise calculator that calculates a root mean square. An information recording / reproducing apparatus having an equalizer and a Viterbi detector, for recording information on an optical disc and reproducing information,
A data string A having a length L having a constraint length of Viterbi detection as L, 1 or 0 as a component is represented by A = [a1, a2,. . . , AL], the average value of the reproduction signal output corresponding to the data string A or the average value of the output of the equalizer corresponding to the data string A is calculated for each possible data string A to detect Viterbi A level average calculator for generating a series of reference values for branch metric calculation in
An error vector calculator that obtains a difference between the reference values between paths through which signal sequence pairs including signal sequences that are likely to be erroneous each other pass, and generates a first vector having a value based on the difference between the reference values as a component;
A Viterbi error calculator for obtaining a Viterbi error that is a difference between the equalizer output and the average value calculated by the level average value calculator;
The time series data of the Viterbi error is used as an element, a vector having the same number of components as the first vector is used as a second vector, an inner product of the first vector and the second vector is obtained, An effective noise calculator for calculating the mean square;
An evaluation index calculator for calculating a signal quality evaluation value based on the magnitude of the first vector generated by the error vector calculator and the mean square of the inner product calculated by the effective noise calculator. An information recording / reproducing apparatus. A recording / reproduction condition adjuster for evaluating a reproduction signal from the information recording medium by the evaluation index calculator and controlling at least one of the reproduction condition and the recording condition of the information recording medium based on the evaluation result; The information recording / reproducing apparatus according to claim 13, further comprising:

Images