REPRODUCTION SIGNAL EVALUATION METHOD, REPRODUCTION SIGNAL
EVALUATION UNIT, AND OPTICAL DISK DEVICE ADOPTING THE SAME
Technical Field
[0001] The present invention relates to a reproduction signal evaluation method and
reproduction signal evaluation unit using a PRML signal processing system, and an optical disk
device adopting the same.
Background Art
[0002] Recently the shortest mark length of recording marks have reached the limit for optical
resolution, and an increase in inter-symbol interference and deterioration of SNR (Signal Noise
Ratio) are becoming obvious as the density of optical disk media increases, therefore the use of a
PRML (Partial Response Maximum Likelihood) system as a signal processing method is
becoming common.
[0003] The PRML system is a technology combining partial response (PR) and maximum
likelihood (ML) decoding, and is a known system for selecting a most likely signal sequence
from a reproduced waveform, assuming the occurrence of inter-symbol interference. As a result,
it is known that decoding performance improves compared with a conventional level decision
system (e.g. see Non-Patent Document 1).
[0004] On the other hand, a shift in signal processing systems from a level decision to PRML
has resulted in generating some problems in reproduction signal evaluation methods. Jitter, that
is a reproduction signal evaluation index, which has been used conventionally, is based on the
assumption that signals are processed using a level decision system. This means that in some
cases jitter has no correlation with the decoding performance of the PRML system, of which
signal processing algorithms are different from the level decision system. Therefore a new index
having correlation with the decoding performance of the PRML system has been proposed (e.g.
see Patent Document 1 and Patent Document 2).
[0005] A new index to position shift (edge shift) between a mark and a space, which is critical
for recording quality of an optical disk, has also been proposed (e.g. see Patent Document 3). If
a PRML system is used, this index must also be correlated with the performance of the PRML,

and must quantitatively express the shift direction and quantity of the edge for each pattern,
according to the concept of the PRML system.
[0006] If the PRML system is used, this index must also be correlated with the decoding
performance of the PRML system, and must quantitatively express the shift direction and
quantity of the edge for each pattern according to the concept of the PRML system.
[0007] As the density of magnetic disk media increases further, the problem of inter-symbol
interference and SNR deterioration becomes more serious. In this case, the system margin can
be maintained by using a higher level PRML system (e.g. see Non-Patent Document 1). In the
case of an optical disk medium of which diameter is 12 cm and recording capacity per recording
layer is 25 GB. The system margin can be maintained by using a PR1221 ML system, but in the
case of a 33.3 GB recording capacity per recording layer, a PR 12221 ML system must be used.
In this way, it is expected that the tendency to use a higher level PRML system would continue
in proportion to the increase in densities of optical disk media.
[0008] Patent Document 1 and Patent Document 2 disclose using "a differential metric, which is
a difference of the reproduction signals between the most likely first state transition sequence
and the second most likely second state transition sequence" as the index value.
[0009] If there are a plurality of patterns of "a most likely first state transition sequence and
second most likely second state transition sequence" which have the possibility of causing an
error, these patterns must be statistically processed systematically. This processing method is
not disclosed in Patent Document 1 and Patent Document 2. Patent Document 5 discloses a
method for detecting a plurality of patterns of "a differential metric of reproduction signals
between the most likely first state transition sequence and the second most likely second state
transition sequence" detected in the same manner as in Patent Document 1 and Patent Document
2, and the processing of a pattern group. In PR 12221 ML signal processing, which is disclosed
in Patent Document 5, there are three types of patterns which easily cause an error (pattern group
of merging paths of which Euclidian distance is relatively short). The pattern generation
probability and the number of errors, when the pattern generates errors occur in a pattern, differ
depending on the pattern, so according to Patent Document 5, a standard deviation a is
determined from the distribution of the index values, which are acquired for each pattern, and the
errors to be generated are predicted based on the generation probability of the pattern (generation
frequency with respect to all parameters) and the number of errors to be generated when the
pattern has an error. In Patent Document 5, a method for assuming the distribution of the

acquired index values as a normal distribution and predicting a probability for the index value
becoming "0" or less based on the standard deviation σ thereof and variance average value \i,
that is, a probability of generation of a bit error, is used as an error prediction method. This,
however, is a general method for predicting error generation probability. The method for
calculating the predicted error rate according to Patent Document 5 is characterized in that
generation probability is determined for each pattern, the predicted error rate is calculated, and
this predicted error rate is used as a guideline of signal quality.
[0010] However, with the method according to Patent Document 5, the error rate cannot
be predicted accurately if recording distortion occurs to recording signals. This problem
becomes particularly conspicuous when data is recorded by thermal recording, such as the case
of an optical disk, since recording distortion tends to be generated by thermal interference. As
the density of optical disk increases, space between recording pits decreases even more, and an
increase in thermal interference is expected, therefore this problem will be unavoidable in the
future. The problem of the predicted error rate calculation method according to Patent
Document 5, which cannot appropriately evaluate the signal quality of signals having recording
distortion, will now be described.
[0011] Fig. 15 shows an example of frequency distribution of a differential metric of a specific
pattern, which is used as a signal index in Patent Document 1 and Patent Document 5. Generally
speaking, the spread of the distribution of the differential metric is caused by the noise generated
in an optical disk. The reproduction noise generated by an optical disk is random, so this
distribution usually is a normal distribution. And this differential metric is defined as a
"differential metric of the most likely first state transition sequence and second most likely
second state transition sequence", and is a distribution of which center is a square of the
Euclidean distance between the most likely first state transition sequence and the second most
likely second state transition sequence of an ideal signal (hereafter defined as the signal
processing threshold). The standard deviation of which center is this signal processing threshold
is the index value defined in Patent Document 1, Patent Document 2 and Patent Document 5.
The probability of this differential metric becoming 0 or less corresponds to the predicted error
rate based on the index value. This predicted error rate can be determined using the inverse
function of the cumulative distribution function of this normal distribution.
[0012] Fig. 15A is a distribution diagram when no substantial distortion occurred during
recording, and Fig. 15B and Fig. 15C show distribution diagrams in a state where recording
edges in the recording pits shifted due to thermal interference during recording, and recording

distortion occurred. If distortion occurs due to thermal interference, the frequency distribution of
the differential metric of a specific pattern becomes a normal distribution of which center value
is shifted. This shift of the center position corresponds to the distortion generated by thermal
interference. Fig. 15B and Fig. 15C are cases when a predetermined amount of shift occurred in
the plus or minus direction from the center of the distribution, and an index value to be
determined is the same value for both Fig. I5B and Fig. 15C, and the index value increases since
the center of the distribution has shifted. An increase in the index value should mean an increase
in the probability of error generation, but errors decrease in the case of Fig. 15C. This is because
in the case of Fig. 15B, where the center of the distribution is shifted to the side closer to "0",
error generation probability (probability of differential metric becoming 0 or less) increases, but
in the case of Fig. 15C, where the center of the distribution is shifted to the plus side, error
generation probability decreases. This reversal phenomena is because an error is generated only
when the index value based on the differential metric approaches 0, which is the major
difference from the jitter of the time axis, that is the index value conventionally used for optical
disks. In the case of a conventional jitter of the time axis, errors increase regardless the side,
plus or minus, to which the center position of the distribution shifts, therefore the above
mentioned problem does not occur.
[0013] A problem similar to the above also occurs in the case shown in Fig. 15D. Fig. 15D is a
case when the determined distribution of the differential metric is not normal distribution. This
occurs when the thermal interference during recording is high, and thermal interference is also
received from the recording marks before and after "the most likely first state transition sequence
and second most likely second state transition sequence". The thermal interference amount is
different depending on the length of the recording marks before and after, and the shift of
recording mark positions generates a differential metric distribution where two normal
distributions (distribution 1 and distribution 2) overlap. In distribution 2, where there is a shift to
the plus side from the signal processing threshold, error generation probability drops, but the
index value, which is a standard deviation from the signal processing threshold as the center,
increases because of the influence of distribution 2. Just like the case of Fig. 15C, error rate also
decreases when the index value increases. In this way, if the prior art reported in Patent
Document 1 and Patent Document 5 is applied to a high recording density optical disk of which
thermal interference is high, the correlation of the index value and error rate worsens.
[0014] An idea for solving this problem is disclosed in Patent Document 4. This is a method of
counting a number of differential metrics with which the differential metric, acquired from a

predetermined pattern group, which becomes smaller than a predetermined threshold (e.g. half of
signal processing threshold). A method for determining a predicted error rate based on this
count value is also disclosed. In the case of this method, a side closer to 0 of the differential
metric distribution, that is the side which has a possibility of generating an error, is used for the
evaluation target, so the above mentioned problems in Patent Document 1 and Patent Document
5 do not occur. But a new problem, mentioned herein below, occurs, since a predetermined
threshold is used and a number of differential metrics exceeding this threshold is measured. This
problem will be described with reference to Fig. 15E.
[0015] Fig. 15E shows an example of counting the differential metrics of the distribution which
exceeds the threshold, which is half of the signal processing threshold. The differential metrics
less than this threshold are counted, and the ratio of the parameter of pattern generation and the
count value is used as the signal index. If it is assumed that the distribution of the differential
metric is a normal distribution based on this count value, the probability when the differential
metric becomes smaller than 0 can be determined, and the predicted error rate can be calculated.
Fig. 15F shows an example when the signal quality is good (signal quality with about an 8%
jitter). In such a case, the spread of the distribution of the differential metric becomes narrow,
and the number of differential metrics which exceeds the threshold decreases dramatically. In
the case of Fig. 15F, only about 0.2%, out of the differential metric distribution, can be measured.
This means that a wide area must be measured in order to increase the accuracy of the
measurement, which increases the measurement time and diminishes measurement stability.
Also if there are defects and scratches generated during manufacture of the disks or if there is
dust on a disk surface, the differential metric is generated in an area not greater than the
threshold due to this defect (illustrated in Fig. 15F). In such a case, a number of the differential
metrics, which exceed the threshold generated in the normal distribution, cannot be counted
correctly. An advantage of conventional time axis jitter used for optical disks is that it is not
affected by such defects, since standard deviation of measured time fluctuation is used and all
the measured data is used. The method disclosed in Patent Document 4, on the other hand, does
not have this advantage of the conventional method based on time axis jitter, which is not
affected by the defects, and therefore has a problem when used for the index values of optical
disks, which is a system where such defects as scratches and fingerprints easily occur. In order
to increase the number of differential metrics to be measured using the method according to
Patent Document 4, the threshold could be increased, but if the threshold is increased, another
problem occurs, that is the accuracy of the predicted error rate drops. In an extreme case, if the
threshold is increased to half of the Euclidean distance, a number of differential metrics that

exceed the threshold becomes half of the number of measured samples, therefore it no longer
depends on the spread of distribution, and accurate measurement becomes possible. In this way,
in the case of the method according to Patent Document 4, the value of the threshold must be
adjusted in order to maintain constant measurement accuracy depending on the quality of
measured signals, and such adjustment is possible if the manner of how distribution spreads is
somewhat understood, nonetheless this is a major problem for optical disks, where signal quality
changes significantly.
[0016] Patent Document 4 and Patent Document 5 also disclose a method of using bER
predicted by the differential metric as the index, but if this is used as an index value,
compatibility with the time axis jitter, which has been used as the signal quality evaluation index
of optical disks, is lost, and handling is difficult.
Prior Art Document
Patent Document
[0017] Patent Document 1: Japanese Patent Application Laid-Open No. 2003-141823
Patent Document 2: Japanese Patent Application Laid-Open No. 2004-213862
Patent Document 3: Japanese Patent Application Laid-Open No. 2004-335079
Patent Document 4: Japanese Patent Application Laid-Open No. 2003-51163
Patent Document 5: Japanese Patent Application Laid-Open No. 2003-272304
Non-Patent Document
[0018] Non-Patent Document 1: Blu-ray Disk Books, Ohmsha Ltd.
Non-Patent Document 2: Adaptive Signal Processing Algorithms, Baihukan
Disclosure of the Invention
[0019] It is an object of the present invention to provide a signal processing method and
reproduction signal evaluation unit suitable for a system using a PRML system, and an optical
disk device adopting the same.
[0020] In order to achieve the foregoing object, a reproduction signal processing method
according to an aspect of the present invention is a reproduction signal evaluation method for

evaluating quality of a reproduction signal reproduced from an information recording medium
based on a binary signal generated from the reproduction signal using a PRML signal processing
system, having: a pattern extraction step of extracting, from the binary signal, a specific state
transition pattern which has the possibility of causing a bit error; a differential metric computing
step of computing a differential metric, which is a difference of a first metric between an ideal
signal of a most likely first state transition sequence corresponding to the binary signal and the
reproduction signal, and a second metric between an ideal signal of a second most likely second
state transition sequence corresponding to the binary signal and the reproduction signal, based on
the binary signal extracted in the pattern extraction step; an extraction step of extracting the
differentia] metric which is not greater than a predetermined signal processing threshold; a mean
value computing step of determining an average mean of the differential metrics which is not
greater than a predetermined signal processing threshold and extracted in the extraction step; a
standard deviation computing step of determining a standard deviation which corresponds to an
error rate that is predicted, from the mean value; and an evaluation step of evaluating a quality of
the reproduction signal using the standard deviation.
[0021] According to the foregoing structure, a standard deviation, which corresponds to an error
rate predicted from the mean value of the differential metrics which are not greater than the
extracted signal processing threshold, is determined, and the quality of the reproduction signal is
evaluated using this standard deviation. It is therefore possible to realize a signal evaluation
having very high correlation with the error rate. As a result, a reproduction signal evaluation
method, which is suitable for a system using a PRML signal processing system and which can
evaluate the quality of reproduction signals of information recording media at high accuracy, can
be implemented.
[0022] Other objects, characteristics and advantages of the present invention shall be sufficiently
clarified by the description herein below. The excellent aspects of the present invention shall be
clarified in the following description with reference to the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Fig. 1 is a block diagram depicting a general structure of an optical disk device according
to one embodiment of the present invention;
Fig. 2 is a block diagram depicting a general structure of an optical disk device according
to another embodiment of the present invention;

Fig. 3 is a diagram depicting a slate transition rule which is determined by an RI..L (1,7)
recording code and equalization type PR (1, 2, 2, 2, J) according to one embodiment of the
present invention;
Fig. 4 is a trellis diagram corresponding to the state transition rule shown in Fig. 3;
Fig. 5 is a graph depicting a relationship of a sampling time and a reproduction level
(signal level) on the transition paths in Table 1;
Fig. 6 is a graph depicting a relationship of a sampling time and a reproduction level
(signal level) on the transition paths in Table 2;
Fig. 7 is a graph depicting a relationship of a sampling time and a reproduction level
(signal level) on the transition paths in Table 3;
Fig. 8 is a diagram depicting a distribution of the differential metric of the PR (1, 2, 2, 2,
1) ML according to one embodiment of the present invention;
Fig. 9 is a diagram depicting a distribution of the differentia] metric in a Euclidian
distance pattern of a PR (1, 2, 2, 2, 1) ML according to one embodiment of the present invention;
Fig. 10 is a diagram depicting a distribution of the differential metric in each Euclidian
distance pattern of a PR (1, 2, 2, 2, 1) ML according to one embodiment of the present invention;
Fig. 11 is a diagram depicting a distribution of the differential metric of a PR (1, 2, 2, 2,
1) ML according to one embodiment of the present invention;
Fig. 12 is a graph depicting a relationship of the signal evaluation index value and error
rate according to one embodiment of the present invention;
Fig. 13 is a block diagram depicting a structure of an optical disk device according to still
another embodiment of the present invention;
Fig. 14 is a block diagram depicting a structure of an optical disk device according to still
another embodiment of the present invention;
Fig. 15A is a diagram depicting the distribution of a conventional differential metric;
Fig. 15B is a diagram depicting the distribution of a conventional differential metric;
Fig. 15C is a diagram depicting the distribution of a conventional differential metric;

Fig. 15D is a diagram depicting the distribution of a conventional differential metric;
Fig. 15E is a diagram depicting the distribution of a conventional differential metric; and
Fig. 15F is a diagram depicting the distribution of a conventional differential metric.
Embodiments for Carrying out the Invention
[0024] Embodiments of the present invention will now be described with reference to the
drawings. The following embodiments are examples of carrying out the present invention, and
do not limit the technical scope of the present invention.
[0025] Jn a signal evaluation index detection unit of the present embodiment, a PR 12221 ML
system, which is an example of a PRML system, is used for signal processing of a reproduction
system, and RLL (Run Length Limited) codes, such as an RLL (1, 7) code, are used for the
recording codes. A PRML system is a signal processing that combines waveform equalization
technology for correcting reproduction distortion, which is generated when information is
reproduced, and signal processing technology for selecting a most likely data sequence from the
reproduction signal which includes data errors, by actively utilizing the redundancy of an
equalized waveform.
[0026] First signal processing by a PR12221 ML system will be described in brief, with
reference to Fig. 3 and Fig. 4.
[0027] Fig. 3 is a state transition diagram depicting the state transition rule, which is determined
by the RLL (1, 7) recording codes and the PR12221 ML system. Fig. 3 shows a state transition
diagram which is normally used when PR ML is described. Fig. 4 is a trellis diagram in which
the state transition diagram shown in Fig. 3 is developed with respect to the time axis.
[0028] "0" or "1" inside the parenthesis in Fig. 3 indicates a signal sequence on the time axis,
and indicates the state of possibility of the state transition from the respective state to the next
time.
[0029] In a PR12221 ML system, a number of states of the decoding unit is limited to 10,
because of the combination with the RLL (1,7) code. A number of state transition paths in a
PR12221 ML system is 16, and a number of reproduction levels is 9.
[0030] In order to describe the state transition rule of a PR 12221 ML system, 10 states are
represented, as shown in the state transition diagram in Fig. 3, where the state S(0, 0, 0, 0) at a

certain time is SO, the slate S(0, 0, 0, 1) is S1, the state S(0, 0, 1,1,) is S2, the state S(0, 1, 1, 1) is
S3, the state S(1, 1, 1, 0) is S4, the state S(l, 1, 1, 0) is S5, the state S(l, 1,0, 0) is S6, the state
S(l, 0, 0, 0) is S7, the state S(l, 0, 0, 1) is S8 and the state S(0, 1,1, 0) is S9. In Fig. 3, "0" or
"1" in parenthesis indicates a signal sequence on the time axis, and shows which state a certain
state may possibly become in the state transition the next time.
[0031] In the state transition of the PR12221 ML system shown in Fig. 4, there are an infinite
number of state transition sequence patterns (combination of states) in which two state
transitions can occur when a predetermined state at a certain time transit to a predetermined state
at another time. However, patterns which have a high possibility of causing an error are limited
to specific patterns of which discernment is difficult. By targeting these state transition patterns
in particular which can easily generate an error, the state transition sequence patterns in the
PR12221 ML system can be listed as shown in Table 1, Table 2 and Table 3.

[0035] In each of the tables, Table 1 to Table 3, shows a state transition to indicate a locus of
states which merged from the start state, two possible transition data sequences which underwent
state transition, two possible ideal reproduction waveforms which underwent state transition, and
a square of the Euclidean distance of the two ideal reproduction waveforms.
[0036] The square of the Euclidean distance indicates as sum of the square of the difference of
the two ideal reproduction waveforms. When the error possibility of the two reproduction
waveforms is judged, two reproduction waveforms can be more easily distinguished if the value
of the Euclidean distance is long, therefore a judgment mistake occurs less frequently. If the
value of the Euclidean distance is short, a judgment mistake may more frequently occur, since it
is difficult to distinguish the two waveforms having an error possibility. In other words, state
transition patterns of which Euclidean distance is long are state transition patterns where an error
does not occur very much, and state transition patterns when the Euclidean distance is short are
state transition patterns where an error easily occurs.
[0037] In each table, the first column shows the state transition (Smk-9 → Snk) where two state
transitions, which easily cause an error, branch and merge again. The second column shows the
transition data sequence (bk-1, . . . , bk) which generates this state transition. X in this state data
sequence indicates a bit which has high error generation possibility among this data, and if this
state transition is judged as erred, a number of X (also !X in Table 2 and Table 3) is a number of
errors. In other words, X in a transition data sequence can be either "0" or "1". One of "0" or
"1" corresponds to the most likely first state transition sequence, and the other corresponds to the
second most likely second state transition sequence. In Table 2 and Table 3, !X indicates a bit
inversion of X.
[0038] As described in detail later, each decoding data sequence (binary signal) after a Viterbi
decoding section executes decoding processing is compared with the transition data sequences in
Table 1 to Table 3 (X indicates "don't care"), and a most likely first state transition sequence
having high error possibility and a second most likely second state transition sequence are
extracted. The third column shows the first state transition sequence and second state transition
sequence. The fourth column shows two ideal reproduction waveforms (PR equalization ideal
values) when a respective state transitions completes, and the fifth column shows a square of the
Euclidean distance of these two ideal signals (square of Euclidean distance between paths).
[0039] Table 1 shows state transition patterns that could take two state transitions, and is state
transition patterns in the case when a square of the Euclidean distance is "14". There are 18

types of state transition sequence patterns in the case when a square of the Euclidean distance is
14. The state transition sequence patterns shown in Table 1 correspond to the edge section
(switching of a mark and a space) of the waveforms of an optical disk. In other words, the state
transition sequence pattern shown in Table 1 is a pattern of a 1-bit shift error at the edge.
[0040] Fig. 5 is a graph depicting the relationship of the sampling time and a reproduction level
(signal level) in the transition paths in Table 1. In the graph in Fig. 5, the x-axis indicates a
sampling time (each sampling timing of a recording sequence), and the y-axis indicates a
reproduction level. As mentioned above, in the case of a PR12221 ML system, there are 9 levels
of ideal reproduction signal levels (levels 0 to 8).
[0041 ] As an example, the transition paths when transiting from the state SO (k-5) to the state S6
(k) according to the state transition rule shown in Fig. 3 will be described (see Table 1). In this
case, one transition path is a case when the recording sequence was detected as a transition of "0,
0, 0, 0, 1, 1, 1, 0, 0". If this transition is converted into a recording state, regarding "0" of the
reproduction data as a space portion and "1" as a mark portion, the recording state is 4T or
longer spaces, and 3T marks and 2T or longer spaces. In Fig. 5, the relationship of the sampling
time and the reproduction level (signal level) in this transition path is shown as a path A
waveform.
[0042] The other transition path of the state transition paths from the state SO (k-5) to the state
S6 (k) in the state transition rule in Fig. 5 is a case when the recording sequence is detected as
the transition of "0, 0, 0, 0, 0, 1, 1, 0, 0". If "0" of the reproduction data is regarded as a space
portion and "1" as a mark portion, the recording state corresponds to 5T or longer spaces, and 2T
marks and 2T or longer spaces. In Fig. 5, the PR equivalent ideal waveform of this path is
shown as a path B waveform. The state transition pattern of which square of the Euclidean
distance is 14 in Table 1 always includes one edge information (zero cross point), which is
characteristic thereof.
[0043] Fig. 6 is a graph depicting the relationship of the sampling time and the reproduction
level (signal level) in the transition paths in Table 2. In the graph in Fig. 6, the x-axis indicates a
sampling time (each sampling time of recording sequence), and the y-axis indicates a
reproduction level.
[0044] Table 2 shows state transition patterns which could take two state transitions, just like
Table 1, and shows the state transition patterns in the case when a square of the Euclidean
distance, is 12. There are 18 types of state transition patterns in the case when a square of the

Euclidean distance is 12. The state transition patterns shown in Table 2 are patterns having a 21'
marks or 2T spaces shift error, that is 2-bit shift error patterns.
[0045] Jn this case, one path of which recording sequence transits as "0. 0, 0, 0, 1, 1, 0, 0. 0, 0,
0" is detected, and if "0" of the reproduction data is regarded as a space portion and ''1" as a
mark portion, the recording state corresponds to 4T or longer spaces, and 2T marks and 5T or
longer spaces. In Fig. 6, the PR equivalent ideal waveform of this path is shown as a path A
waveform.
[0046] As an example, the transition paths when transiting from the state SO (k-7) to the state SO
(k) according to the state transition rule shown in Fig. 3 will be described (see Table 2). In this
case, one transition path is a case when the recording sequence was detected as a transition of "0,
0, 0, 0, 1, 1, 0, 0, 0, 0, 0". If this transition path is converted into a recording state, regarding "0"
as the reproduction data as a space portion and "1" as a mark portion, the recording state
corresponds to 4T or longer spaces, and 2T marks and 5T or longer spaces. In Fig. 6, the
relationship of the sampling time and the reproduction level (signal level) in this transition path
is shown as a path A waveform.
[0047] The other transition path, on the other hand, is a case when the recording sequence is
detected as the transition of "0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0". If this transition path is converted into
a recording state, regarding "0" of the reproduction data as a space portion and "1" as a mark
portion, the recording state corresponds to 5T or longer spaces, and 2T marks and 4T or longer
spaces. In Fig. 6, the relationship of the sampling time and the reproduction level (signal level)
in this transition path is shown as a path B waveform. The state transition pattern of which
square of the Euclidean distance is 12 in Table 2 always includes two edge information, the rise
and fall of 2T, which is characteristic thereof.
[0048] Fig. 7 is a graph depicting the relationship of the sampling time and the reproduction
level (signal level) in the transition paths in Table 3. In the graph in Fig. 7, the x-axis indicates a
sampling time (each sampling timing of recording sequence), and the y-axis indicates a
reproduction level.
[0049] Table 3 shows state transition sequence patterns which could take two state transition
sequences, just like Table 1 and Table 2, and shows the state transition sequence patterns in the
case when a square of the Euclidean distance is 12. There are 18 types of state transition
sequence patterns in the case when a square of Euclidean distance is 12. The state transition

sequence patterns shown in Table 3 are areas where 2T marks and 2T spaces continue, that is 3-
bit shift error patterns.
[0050] As an example, the transition paths when transiting from the state S0 (k-9) to the state S6
(k) according to the state transition rule shown in Fig. 3 will be described (see Table 3). In this
case, one transition path is a case when the recording sequence was detected as a transition of "0.
0, 0, 0, 1, 1, 0, 0, 1, 1, 1, 0, 0". If this transition path is converted into a recording state,
regarding "0" of the reproduction data as a space portion and "1" as a mark portion, the
recording state corresponds to 4T or longer spaces, and 2T marks, 2T spaces, 3T marks and 2T
or longer spaces. In Fig. 7, the relationship of the sampling time and the reproduction level
(signal level) in this transition path is shown as a path A waveform.
[0051] The other transition path, on the other hand, is a case when the recording sequence is
detected as the transition of "0, 0, 0, 0, 0, 1, 1, 0, 0, 1,1,0, 0". If this transition path is converted
into a recording state, regarding "0" of the reproduction data as a space portion and "1" as a
mark portion, the recording state corresponds to 5T or longer spaces, and 2T marks, 2T spaces,
2T marks and 2T or longer spaces. In Fig. 7, the relationship of the sampling time and the
reproduction level (signal level) in this transition path is shown as a path B waveform. The state
transition sequence pattern of which square of the Euclidean distance is 12 in Table 3 always
includes at least three edge information, which is characteristic thereof.
[0052] Embodiments of the present invention will now be described.
[0053] (First Embodiment)
An optical disk device having a reproduction signal evaluation unit according to one
embodiment of the present invention will now be described with reference to the drawings. Fig.
1 is a block diagram depicting a structure of the optical disk device 200 according to the first
embodiment.
[0054] An information recording medium 1 is an information recording medium for optically
recording/reproducing information, and is an optical disk medium, for example. The optical disk
device 200 is a reproduction unit which reproduces information from the installed information
recording medium 1.
[0055] The optical disk device 200 has an optical head section 2, a preamplifier section 3, an
AGC (Automatic Gain Controller) section 4, a waveform equalization section 5, an A/D
conversion section 6, a PLL (Phase lacked Loop) section 7, a PR equalization section 8, a

maximum likelihood decoding section 9, a signal evaluation index detection section
(reproduction signal evaluation unit) 100, and an optical disk controller section 15.
[0056] The optical head section 2 converges laser beams transmitted through an objective lens
onto a recording layer of the information recording medium 1, receives the reflected light thereof,
and generates analog reproduction signals which indicate information read from the information
recording medium 1. The preamplifier section 3 amplifies an analog reproduction signal, which
is generated by the optical head section 2, using a predetermined gain, and outputs it to the AGC
section 4. A numerical aperture of the objective lens is 0.7 to 0.9, and is more preferably 0.85.
The wavelength of the laser beam is 410 nm or less, and is more preferably 405 nm.
[0057] The preampl ifier unit 3 amplifies the analog reproduction signal by a predetermined gain,
and outputs it to the AGC section 4.
[0058] The AGC section 4 amplifies or attenuates the analog reproduction signal, and outputs it
to the waveform equalization section 5 based on the output from the A/D conversion section 6,
so that the analog reproduction signal from the preamplifier section 3 has a predetermined
amplitude.
[0059] The waveform equalization section 5 has LPF characteristics to shield a high frequency
area of the reproduction signal, and HPF characteristics to shield a low frequency area of the
reproduction signal, and shapes the reproduction waveform according to desired characteristics,
and outputs it to the A/D conversion section 6.
[0060] The A/D conversion section 6 samples an analog reproduction signal synchronizing with
a reproduction clock, which is output from the PLL section 7, converts the analog reproduction
signal into a digital reproduction signal, and outputs it to the PR equalization section 8, and also
to the AGC section 4 and the PLL section 7.
[0061] The PLL section 7 generates a reproduction clock to synchronize with the reproduction
signal after waveform equalization, based on the output from the A/D conversion section 6, and
outputs it to the A/D conversion section 6.
[0062] The PR equalization section 8 has a function to change the filter characteristics into
characteristics of various PR systems. The PR equalization section 8 performs filtering to be the
frequency characteristics, which is set so that the frequency characteristics of the reproduction
system become assumed characteristics of the maximum likelihood decoding section 9 (e.g. PR
(1, 2, 2, 2, 1) equalization characteristics), performs PR equalization processing on digital

reproduction signals tor suppressing high frequency noises, intentionally adding inter-symbol
interference, and outputs the results to the maximum likelihood decoding section 9. The PR
equalization section 8 may have an FIR (Finite Impulse Response) filter structure, for example,
so as to adaptively control the tap coefficient using LMS (The Least-Mean Square) algorithm
(e.g. see Non-Patent Document 2).
[0063] The maximum likelihood decoding unit 9 is a Viterbi decoder, for example, and uses a
maximum likelihood decoding system, which estimates a most likely sequence based on a code
rule intentionally attached according to the type of partial response. This maximum likelihood
decoding section 9 decodes a reproducing signal which was PR-equalized by the PR equalization
section 8, and outputs binary data. This binary data is output to the optical disk controller 15 in a
subsequent step, as a decoded binary signal, and after execution of predetermined processing,
information recorded on the information recording medium 1 is reproduced.
[0064] Now a structure of the signal evaluation index detection section 100 according to the
present embodiment will be described. The signal evaluation index detection section 100 has a
pattern detection section 101, a differential metric computing section 102, a level decision
section 103, a pattern count section 104, an integration section 105, an error computing section
116 and a standard deviation computing section 120.
[0065] A waveform-shaped digital reproduction signal which is output from the PR equalization
section 8, and a binary signal which is output from the maximum likelihood decoding section 9,
are input to the signal evaluation index detection section 100. In the signal evaluation index
detection section 100, the binary signal is input to the pattern detection section 101, and the
digital reproduction signal is input to the differential metric computing section 102, and
evaluation processing is executed on digital reproduction signals of the information recording
medium 1.
[0066] The pattern detection section 101 has a function to extract specific state transition
patterns, which have the possibility to cause a bit error, from the binary signal. The pattern
detection section 101, according to the present embodiment, extracts specific state transition
patterns (state transition patterns shown in Table 1) of which square of the Euclidean distance
between an ideal signal of a most likely first state transition sequence and an ideal signal of a
second most likely second state transition sequence is 14. In order to implement this, the pattern
detection section 101 stores information of the state transition patterns shown in Table 1. And
the pattern detection section 101 compares the transition data sequences in Table 1, and the

binary signal which :s output from the maximum likelihood decoding section 9. As a result of
this comparison, if the binary signal matches the transition data sequences in Table 1, this binary
signal becomes an extraction target, and the most likely first transition sequence 1 and the
second most likely transition sequence 2, corresponding to this binary signal, are selected based
on the information in Table 1.
[0067] Based on the binary signal extracted by the pattern detection section 101, the differential
metric computing section 102 computes a "differential metric" which is an absolute value of a
difference of "a first metric between an ideal signal of a most likely transition sequence 1
corresponding to the binary signal (PR equalization ideal value: sec Table 1) and the digital
reproduction signal" and "a second metric between an ideal signal of a second most likely
transition sequence 2 corresponding to the binary signal and the digital reproduction signal".
Here the first metric is a square of the Euclidean distance between the ideal signal of the
transition sequence 1 and the digital reproduction signal, and the second metric is a square of the
Euclidean distance between the ideal signal of the transition sequence 2 and the digital
reproduction signal.
[0068] The output of the differential metric computing section 102 is input to the level judgment
section 103, and is compared with a predetermined value (signal processing threshold). The
pattern count section 104 counts a number of differentia] metrics which are less than the signal
processing threshold. Each count value is reflected in a pattern group generation frequency
when an error rate is calculated. The integration section 105 integrates the differential metrics
which are less than the signal processing threshold. A mean value of the differential metrics
which are not greater than the signal processing threshold can be determined by dividing the
integration value determined by the integration section 105 by the pattern generation count. The
error computing section 116 calculates a predicted error rate based on each integration value of
differential metrics which are not greater than the signal processing threshold, and the pattern
generation count. The standard deviation computing section 120 computes the standard
deviation corresponding to this error rate, and determines this standard deviation as the signaJ
index value to evaluate the signal quality. The process by this signal evaluation index detection
section 100 will now be described in detail.
[0069] The reproduction signal reproduced from the information recording medium 1 in the
PRML processing is output from the maximum likelihood decoding section 9 as a binary signal,
as mentioned above, and is input to the signal evaluation index detection section 100. When one
of the transition data sequence patterns in Table 1 is detected from this binary signal, the PR

equalization ideal values of the state transition sequence 1 and the state transition sequence 2 are
determined. For example, if (0, 0, 0, 0, X, 1, 1, 0, 0) is decoded as the binary signal in Table 1,
(S0, S1, S2, S3, S5, S6) is selected as the most likely state transition sequence 1, and (S0, S0, S1,
S2, S9, S6) is selected as the second most likely state transition sequence 2. The PR equalization
ideal value corresponding to the state transition sequence 1 is (1, 3, 5, 6, 5). And the PR
equalization ideal value corresponding to the state sequence 2 is (0, 1, 3, 4, 4).
[0070] Then the differential metric computing unit 102 determines a first metric (Pb14) which is
a square of the Euclidean distance between the reproduction signal sequence (digital
reproduction signal) and the PR equalization ideal value corresponding to the state transition
sequence 1. In the same way, the differential metric computing unit 102 determines a second
metric (Pa14) which is a square of the Euclidean distance between the reproduction signal
sequence and the PR equalization ideal value corresponding to the state transition sequence 2.
The differential metric computing unit 102 determines an absolute value of the difference of the
first metric (Pb14) and the second metric (Pb14), as the differential metric D14 = | Pa14- Pb14 |.
The computing of Pb14 is shown in Expression (1), and the computing of Pb14 is shown in
Expression (2). In these expressions, bk denotes the PR equalization ideal value corresponding
to the state transition sequence 1, ak denotes a PR equalization ideal value corresponding to the
state transition sequence 2, and Xk denotes a reproduction signal sequence.

[0074] In Fig. 9, an area above the signal processing threshold is an area where error does not
occur, and it is therefore unnecessary to predict an error rate. Hence in order to predict an error
rate based on the standard deviation of the differential metric, only an area not greater than the

signal processing threshold becomes a target. A method for calculating this error rate will now
be described.
[0075] The differential metric D14, which is an output from the differential metric computing
section 102, is input to the level decision section 103, and is compared with a predetermined
value (signal processing threshold). In the present embodiment, the signal processing threshold
according to an extraction target specific state transition pattern is set to "14", which is a square
of the Euclidean distance between an ideal signal of the most likely state transition sequence 1
and an ideal signal of the second most likely state transition sequence 2. If the differential metric
D14 is not greater than the signal processing threshold "14", the level decision section 103
outputs the value of this differential metric D14 to the integration section 105, and the pattern
count section 104 counts up the count value. The integration section 105 integrates the
differential metric cumulatively each time the differential metric D14, which is not greater than
the signal processing threshold, is input. Then the error computing section 116 calculates a
predetermined error date using an integration value of the differential metric not greater than the
signal processing threshold and number of generated patterns, counted by the pattern count
section 104. Operation of the error computing section 116 will now be described.
[0076] The mean value of the differential metrics which are not greater than the signal
processing threshold can be determined by dividing the integration value, determined by the
integration section 105, by a number of differential metrics less than the signal processing
threshold (number of generated patterns), which was counted up by the pattern count section 104.
When it is assumed that the mean value of the differential metrics, which are not greater than the
signal processing threshold, is M(x), the mean value of the distribution functions is µ, the
standard deviation is σ14, the probability density function is f, and the distribution function is a
normal distribution, and the absolute mean value m of the differential metrics, which are less
than the signal processing threshold, is given by the following Expression (4).


[0078] Therefore the relationship of the standard deviation σ14 of the differential metrics, which
are not greater than the signal processing threshold and the absolute mean value m of the
differential metrics, which are not greater than the signal processing threshold, is determined by
the following Expression (5).
[0079]

[0080]
As Expression (4) and Expression (5) show, in order to determine the standard deviation σ14 of
the differential metrics which are not greater than the signal processing threshold, the absolute
mean value m of the differential metrics, which are not greater than the signal processing
threshold, is determined, and is then multiplied by about 1.253. Since the signal processing
threshold is fixed, the standard deviation σ14 can be calculated from the absolute mean value m.
The probability of error generation (error rate bER14), which is computed by the error computing
section 116, can be determined by the following Expression (6).
[0081]


[0082] Here d14 in Expression (6) denotes the Euclidean distance between an ideal signal of the
most likely state transition sequence 1 in the extraction target state transition patterns, and an
ideal signal of the second most likely state transition sequence 2. In the case of the present
embodiment, a square of the Euclidean distance d142 = 14 is used. Therefore if the standard
deviation given by Expression (5), which is determined by the integration value and integration
count, is σ14, then the error rate bER14B, predicted based on the computing of the error computing
section 116, is given by the following Expression (7). p14 (= 0.4) is an error generation
probability in the distribution components with respect to all the channel points. Errors
generated in the state transition sequence patterns in Table 1 are 1-bit errors, so the error
generation probability has been multiplied by 1.
[0083] The standard deviation computing section 120 converts this error rate (error generation
probability) bER14 into a signal index value M, to make it to an index which can be handled in a
similar manner as a jitter. By using Expression (7), the standard deviation computing unit 120
converts bER14 into signal index value M using the standard deviation a corresponding to the
predicted error rate.
[0084]

[0085] Here erfc() is an integration value of the complementary error function. If the defining
expression of the signal index M according to the present embodiment is the following
Expression (8), then the index value M using a virtual standard deviation a can be determined by
substituting bER14, calculated by Expression (6), in Expression (7).
[0086]

[0087] In the above description, a virtual standard deviation a and signal index value M are
calculated from a predicted error rate using Expression (6) to Expression (8).
[0088] As described above, according to the present embodiment, the state transition sequence
patterns of merging paths of which Euclidean distance in the PRML signal processing is
relatively small are targeted, and the signal evaluation index M is generated based on the

differential metric information of the state transition sequence patterns. Specifically, a predicted
error rate is calculated from the mean value of the differential metric information which is not
greater than the signal processing threshold, then the virtual standard deviation a is calculated
from the error rate, and the signal evaluation index M including this standard deviation σ of the
normal distribution is generated. As a result, it is possible to realize a signal evaluation method
and evaluation index having very high correlation with the error rate.
[0089] In the case of a conventionally proposed distribution evaluation of a simple differential
metric, it is difficult to calculate a signal index having correlation with an error rate, because of
the recording distortion due to thermal interference generated as a higher density of an optical
disk that will be increasingly demanded in the future. The present embodiment is for solving this
problem, and a key point thereof is that only one side of the distribution components of the
differential metrics, where errors are generated, is targeted to calculate the signal index which
has high correlation with actual errors to be generated, and the standard deviation a of both
virtual sides distribution is determined based on this one sided distribution.
[0090] According to the present embodiment, for the specific state transition pattern which may
cause a bit error, the pattern detection section 101 according to the present embodiment extracts
specific state transition patterns (state transition patterns shown in Table 1) with which a square
of the Euclidean distance between an ideal signal of the most likely first state transition sequence
and an ideal signal of the second most likely second state transition sequence becomes 14, but
the present invention is not limited to this. For example, specific state transition patterns (state
transition patterns shown in Table 2 or Table 3) with which this square of the Euclidean distance
becomes 12 may be extracted.
[0091] The optical disk controller unit 15 functions as an evaluation section, which executes
evaluation processing based on the signal evaluation index M received from the standard
deviation computing section 120. This evaluation result can be displayed on a display section,
which is not illustrated, or stored in a memory as evaluation data.
[0092] In the present embodiment, the optical disk device 200 having the signal evaluation index
detection section 100 was described, but the present invention may be constructed as an optical
disk evaluation unit (reproduction signal evaluation unit) having the optical disk controller
section 15 as an evaluation section. The optical disk evaluation unit can be used for evaluating
the information recording medium before shipment, whether this information recording medium
1 has a quality conforming to a predetermined standard or not.

[0093] The optical disk device 200 having the reproduction signal evaluation unit may be
arranged to perform the following operation. For example, the quality of the reproduction signal
is evaluated for commercial optical disks (blank disks) shipped from a factory, and optical disks
which are judged as not satisfying a predetermined quality are rejected. It is for certain possible
that optical disks recorded by a recorder (recording using a device other than this optical disk
device) can be evaluated by this optical disk device and the optical disk, which are judged as not
satisfying a predetermined quality, and are rejected.
[0094] If the optical disk device 200 can record and reproduce information, then evaluation
based on test recording can be performed before recording information on the optical disk. In
this case, the quality of reproduction signals is evaluated for the test information recorded by the
optical disk device 200, and if NG, recording conditions are adjusted until NG becomes OK, and
the optical disk is rejected if NG continues after a predetermined number of times of adjustment.
[0095] (Second Embodiment)
An optical disk device having a reproduction signal evaluation unit according to another
embodiment of the present invention will now be described with reference to the drawings. A
composing element the same as the first embodiment is denoted with a same element number,
for which description can be omitted. Fig. 2 is a block diagram depicting the structure of the
optical disk device 400 of second embodiment.
[0096] An information recording medium 1 is an information recording medium for optically
recording/reproducing information, and is an optical disk medium, for example. The optical disk
device 400 is a reproduction unit which reproduces information from the installed information
recording medium 1.
[0097] The optical disk device 400 has an optical head section 2, a pre-amplifier section 3, an
AGC (Automatic Gain Controller) section 4, a waveform equalization section 5, an A/D
conversion section 6, a PLL (Phase Locked Loop) section 7, a PR equalization section 8, a
maximum likelihood decoding section 9, a signal evaluation index detection section
(reproduction signal evaluation unit) 300 and an optical disk controller section 15. The
structures and functions of these elements constituting the optical disk device 400 are the same
as the first embodiment, and descriptions thereof are omitted here.
[0098] Now a structure of the signal evaluation index detection section 300 according to the
present embodiment will be described. The signal evaluation index detection section 300, just

like the signal evaluation index detection section 100 of the first embodiment, can be used as an
evaluation unit for judging whether the quality of the information recording medium 1 conforms
to a predetermined standard before shipment. The present signal evaluation index detection
section 300 may also be installed in a drive unit of the information recording medium 1, and
used as an evaluation unit to perform test recording before a user records information on the
information recording medium 1.
[0099] The signal evaluation index detection section 300 has pattern detection sections 101, 106
and 111, differential metric computing sections 102, 107 and 112, level decision sections 103,
108 and 113, pattern count sections 104, 109 and 114, integration sections 105, 110 and 115,
error computing sections 116, 117 and 118, and a standard deviation computing section 120.
[0100] A waveform-shaped digital reproduction signal which is output from the PR equalization
section 8, and a binary signal which is output from the maximum likelihood decoding section 9,
are input to the signal evaluation index detection section 300. The pattern detection sections 101,
106 and 111 compare the transition data sequences in Tables 1, 2 and 3 and the binary data
which is output from the maximum likelihood decoding section 9 respectively. If the binary data
matches the transition data sequences in Tables 1, 2 and 3 as a result of comparison, a most
likely transition sequence 1 and a second most likely transition sequence 2 are selected based on
Table 1, Table 2 and Table 3. And based on the selection results of the pattern detection sections
101, 106 and 111, the differential metric computing sections 102, 107 and 112 compute a metric,
which is a distance between an ideal value of a transition sequence (PR equalization ideal value:
see Table 1, Table 2 and Table 3) and the digital reproduction signal. Then the differential
metric computing sections 102, 107 and 112 compute the difference between the metrics
computed from the two transition sequences respectively, and perform the absolute value
processing on the metric differences having plus and minus values. The outputs from the
differential metric computing sections 102, 107 and 112 are input to the level decision sections
103, 108 and 113 respectively, and are compared with a predetermined value (signal processing
threshold). The pattern count sections 104, 109 and 114 count a number of differential metrics
which are not greater than the signal processing threshold respectively. These count values each
become a pattern generation frequency when an error rate is calculated. The integration sections
105, 110 and 115 integrate the differential metrics which are not greater than the signal
processing threshold respectively. The mean value of the differential metrics which are not
greater than the signal processing threshold can be determined by dividing the integration value
determined by the integration sections 105, 110 or 115 by a number of generated patterns.

[0101] Each integration section integrates differential metrics which are not greater than the
signal processing threshold, and each computing section divides each integration value by a
number of generated patterns, so as to determine a mean value of the differential metrics which
are not greater than the signal processing threshold, but each integration section may integrate
the differential metrics which are less than the signal processing threshold, and each computing
section divides each integration value by a number of generated patterns, so as to determine a
mean value of the differential metrics which are less than the signal processing threshold.
[0102] The error computing sections 116, 117 and 118 calculate a predicted error rate from each
integration value of the differential metrics which are not greater than the signal processing
threshold and the number of generated patterns. The error rates calculated by the error
computing sections 116, 117 and 118 are added by the adding section 119. The standard
deviation corresponding to this error rate is computed by the standard deviation computing
section 120, and this becomes the signal index value to evaluate the signal quality. The process
by this signal evaluation index detection section 300 will now be described in detail.
[0103] The reproduction signal reproduced from the information recording medium 1 in the
PRML processing is output from the maximum likelihood decoding section 9 as a binary signal,
as mentioned above, and is input to the signal evaluation index detection section 300. When one
of the transition data sequence patterns in Table 1 is detected from this binary signal, the PR
equalization ideal values of the state transition sequence 1 and the state transition sequence 2 are
determined. For example, if (0, 0, 0, 0, X, 1, 1, 0, 0,) is decoded as the binary signal in Table 1,
(S0, S1, S2, S3, S5, S6) is selected as the most likely state transition sequence 1, and (S0, S0, S1,
S2, S9, S6) is selected as the second most likely state transition sequence 2. The PR equalization
ideal value corresponding to the state transition sequence 1 is (1, 3. 5, 6, 5). The PR equalization
ideal value corresponding to the state transition sequence 2 is (0, 1, 3, 4, 4).
[0104] Then the differential metric computing section 102 determines a first metric (Pb14) which
is a square of the Euclidean distance between the reproduction signal sequence (digital
reproduction signal) and the PR equalization ideal value corresponding to the state transition
sequence 1. In the same way, the differentia] metric computing section 102 determines a second
metric (Pb14) which is a square of the Euclidean distance between the reproduction signal
sequence and the PR equalization ideal value corresponding to the state transition sequence 2.
The differential metric computing section 102 determines an absolute value of the difference of
the first metric (Pb14) and the second metric (Pb14), as differential metric D14 = | Pa14 - Pb14 |.
The computing of Pb14 is shown in Expression (9), and the computing of Pb14 is shown in

Expression (10). In these expressions, bk denotes a PR equalization ideal value corresponding to
the state transition sequence 1, ak denotes a PR equalization ideal value corresponding to a stale
transition sequence 2, and xk denotes a reproduction signal sequence.
[0105]

[0108] In order to determine a signal evaluation index having a high correlation with the error
rate, an evaluation method considering all the patterns which have a high possibility of error
generation is required in the signal processing based on a PR 12221 ML system.
[0109] Fig. 8 is a diagram depicting the distribution of differential metrics in the signal
processing of the PR 12221 ML system. In Fig. 8, the x-axis indicates a differential metric, and
the y-axis indicates a frequency of a predetermined differential metric value. As the differential
metric (square of Euclidean distance) becomes smaller in the distribution, the possibility of
generating an error is higher in the signal processing based on the PR12221 ML system. As
shown in the graph of Fig. 8, the differential metrics have a distribution group in the sections 12
and 14, and differential metrics higher than this are 30 or more. In other words, in order to
acquire a signal index having a high correlation with the error rate, it is sufficient to target the
differential metrics in the groups 12 and 14. These groups indicate the state transition sequence
patterns shown in Table 1, Table 2 and Table 3. And the pattern detection sections 101, 106 and
111 identify these state transition sequence patterns. This operation of the differential metric
computing unit, which computes the metric difference from these identified state transition
sequence patterns, will be described in more detail.
[0110] Fig. 10(A) shows an output frequency distribution of the differential metric computing
section 102. The processing of the differential metric computing section 107 is shown in

Expression (12) to Expression (14), and the processing of the differential metric computing
section 1 12 is shown in Expression (15) to Expression (17).
[0111]

[0117] The distributions of (A), (B) and (C) in Fig. 10 have a different frequency and center
position respectively. A number of error bits which are generated when each of these patterns
cause an error is also different. The patterns in Table 1, where a square of the Euclidean distance
is 14, are patterns in which a 1-bit error occurs. The patterns in Table 2, where a square of the
Euclidean distance is 12, are patterns in which a 2-bit error occurs, and the patterns in Table 3,
where a square of the Euclidean distance is 12, are patterns in which a 3-bit error occurs. The
error patterns of which the square of the Euclidean distance is 12, in particular, depends on the
number of 2Ts that continue, and in the case of the recording modulation codes which allow a
continuation of 6 units of 2T, a maximum 6-bit error is generated. Table 3 does not cover a 6-bit

error in which 2T continuously errors, but a pattern to evaluate 2T continuous errors can be
defined according to necessity, so as to extrapolate the evaluation target pattern table.
[0118] In the state transition sequence patterns in each table, the error generation probability in
the recording modulation code sequence is also different. For example, the error generation
frequency is: the state transition sequence patterns in Table 1 are about 40% of all the samples,
the patterns in Table 2 are about 15% of all the samples, and the patterns in Table 3 are about 5%
of all the samples. In this way, the distributions shown by (A), (B) and (C) in Fig. 10 have
different weights in terms of the standard deviation a which indicates a dispersion, detection
window (Euclidean distance), error generation frequency and error bit count, so prediction of the
error rate generated by these distributions can also be computed considering these factors. A
predicted error rate calculation method, which is a major characteristic of the present application,
will be described below.
[0119] As described in the above mentioned problem, when a predicted error rate is calculated
for each pattern group, the predicted error rate may not be able to be determined appropriately
depending on the profile of distribution. Therefore in the present embodiment, the standard
deviation a is calculated from the mean value of a portion of the distribution not greater than a
predetermined threshold (signal processing threshold), so as to determine the error rate, whereby
calculation accuracy of the predicted error rate is improved.
[0120] In Fig. 11, the area above the signal processing threshold is an area where an error does
not occur, and it is therefore unnecessary to predict an error rate. Therefore in order to predict an
error rate from the standard deviation of the differential metrics, an area not greater than the
signal processing threshold should be targeted. This error rate calculation method will now be
described. D14, D12A and D12B, which are outputs from the differential metric computing sections
102, 107 and 112, are input to the level decision sections 103, 108 and 113, and are compared
with a predetermined value (signal processing threshold) respectively. In the present
embodiment, the signal processing threshold is set to 14 for D14, and 12 for both D12A and D12B-
If the differential metric is not greater than the signal processing threshold, the level decision
sections 103, 108 and 113 output the value, and increment the count value of the pattern count
sections 104, 109 and 114 corresponding to the respective pattern count. At the same time, the
integration sections 105, 110 and 115 integrate the differential metric that is not greater than the
signal processing threshold. And the error computing sections 116, 117 and 118 calculate the
estimated error rate from the integration value of the differential metrics that are not greater than

this signal processing threshold and the number of generated patterns. Operation of these error
computing sections 116, 117 and 118 will now be described.
[0121] The integration value determined in each integration section 105, 110 and 1 15 is divided
by the number of differential metrics (number of generated patterns) that are not greater than the
signal processing threshold, counted by the pattern count section 104, 109 and 114, then a mean
value of the differential metrics that is not greater than the signal processing threshold is
determined. If it is assumed that the mean value of the differential metrics that are not greater
than the signal processing threshold is M(x), a mean value of the distribution functions is µ, the
standard deviation is σn, the probability density function is f, and the distribution functions have
a normal distribution, then the absolute mean value m of the differential metrics that are not
greater than the signal processing threshold is given by the following Expression (18).
[0122]

[0123] Therefore the relationship of the standard deviation σn of the differential metrics that are
not greater than the signal processing threshold and the absolute mean value m of the differential
metrics that are not greater than the signal processing threshold is given by the following
Expression (19).

[0124]

[0125] Hence it is understood from Expression (18) and Expression (19) that in order to
calculate the standard deviation σn of the differential metrics that are not greater than the signal
processing threshold, the absolute mean value m of the differential metrics that are not greater
than the signal processing threshold is determined, and is then multiplied by about 1.253. Since
the signal processing threshold is fixed, the standard deviation σn can be calculated from the
absolute mean value m. Then the probability of error generation (error rate: bER), calculated by
the error computing sections 116, 117 and 118 respectively, can be determined by the following
Expression (20).
[01261

[0127] Here d in Expression (20) denotes a Euclidean distance between an ideal signal of a most
likely state transition sequence 1 in the extraction target state transition patterns and an ideal
signal of a second most likely state transition sequence 2. In the case of the present embodiment,

a square of the Euclidean distance is d142 = 14, d12A2 =12 and d12B2 =12.
[0128] Therefore if the standard deviations that are determined from the integration values and
integration count by Expression (19) are σ14, σ12A and σ12B, then the predicted error rates bER14,
bER12A and bER12B, which are computed by the error computing sections 116, 117 and 118
respectively, are given by the following expressions.
[0129]


[0130]

[0132]
Here P14, P12A and P12B (0.4, 0.15, 0.05) are error generation probabilities in the
distribution components of all the channel points. An error generated in the state transition
sequence patterns in Table 1 is a 1-bit error, so the error generation probability has been
multiplied by 1, an error generated in the state transition sequence patterns in Table 2 is a 2-bit
error, so the error generation probability has been multiplied by 2, and an error generated in the
state transition sequence patterns in Table 3 is a 3-bit error, so the error generation probability
has been multiplied by 3 respectively. By adding these error rates, the error generation
probability of errors generated in all of the state transition sequence patterns in Table 1, state
transition sequence patterns in Table 2, and state transition sequence patterns in Table 3 can be
determined. If the error generation probability is bERall, bERall can be given by the following
Expression (24).
[0133]

[0134]
The standard deviation computing section 120 converts the bit error rate determined by
Expression (24) into a signal index value, to make it to an index which can be handled in a
similar manner as jitter.
[0135]


[0136]
Here P is a total of P14, P12A and P12B, and erfc() is an integration value of a
complementary error function. If the defining expression of the signal index M according to the
present invention is Expression (26), then the index value M can be determined by substituting
bERall, calculated by Expression (24), in Expression (25).
[0137]

[0138]
In the above description, a virtual standard deviation a is calculated, and the signal index value
M is calculated from a predicted error rate using Expressions (20) to (26). However, the
calculation method for the evaluation index M according to the present embodiment is not
limited to this method, but may be determined by other defining expressions. An example of
other defining expressions will now be described.
[0139] A probability of pattern Pa to be detected as pattern Pb is assumed to be the error
function given by the following Expression (27).
[0140]

[0141]
Here t in Expression (27) denotes a pattern number of Tables 1 to 3. d denotes a Euclidean
distance in each pattern group in Tables 1 to 3. Specifically, in the case of a pattern group in
Table 1, d2 is 14, and in the case of the pattern groups in Table 2 and Table 3, d2 is 12.

[0142] The error generation probability in the pattern group in Table 1, the pattern group
in Table 2, and the pattern group in Table 3 can be calculated by the following Expression (28)
using Expression (27).
[0143]

[0144]
N1, N2 and N3 in Expression (28) are the generation counts of the pattern group defined in Table
1, Table 2 and Table 3 respectively. The difference from Expression (24) is that the error rate of
each pattern group is not calculated based on all channels as a parameter, but is based on the
number of evaluation patterns in Table 1 to Table 3 as a parameter. Expression (24) calculates
an error rate of which parameter is all the channels including the evaluation patterns. Expression
(28), on the other hand, calculates the error rate of which parameter is the evaluation patterns.
When a virtual a is calculated based on the error rates calculated by Expression (24) and
Expression (28), a same value can be calculated by considering which parameter is the target of
a. Expressions (20) to (26) are examples of computations of which parameters are all channels.
The virtual a is calculated based on Expression (28), and the evaluation index M is calculated.
[0145] The virtual standard deviation a can be calculated by the following Expression (29).
[0146]

[0147] Here E-1 denotes an inverse function of Expression (30).
[0148]

[0149] The evaluation index M can be calculated using the following Expression (31), by
normalizing with a detected window.

[0150]

[0151]
In the end, Expression (26) and Expression (31) calculate a virtual a which is generated
in the evaluation patlerns defined in Table 1 to Table 3, so the index value M is calculated as
substantially a same value. The only difference is the evaluation parameter to calculate the error
rate and the notation of the detection window. Either expression may be used to calculate the
signal index value M. The calculation of the signal index value M using Expression (31) can
also be applied to the first embodiment, of which extraction target is only specific state transition
patterns.
[0152] Fig. 12 is an example of a simulation result showing the bit error rate (bER) and the
signal index value % of Expression (18) when reproduction stress, such as tile, defocus and
spherical aberration, is applied. In the graph in Fig. 12, ▲ indicates a defocus stress, • indicates
a spherical aberration stress, ♦ indicates a radical tilt stress, and J indicates a tangential tilt stress.
The solid line in Fig. 12 is a theoretical curve.
[0153] Generally the criteria of the system margin is a bER of about 4.0 E-4, so the signal index
value to implement this bER is about 15 %. As shown in the graph of Fig. 12, the signal index
value M, defined in the present embodiment, is matched with the theoretical curve of the error
rate in the area of the signal index value ≤ 15 %, which is actually used in the system. Therefore
the signal evaluation method and index according to the present embodiment are very effective
in terms of evaluating signals appropriately.
[0154] As described above, according to the present embodiment, state transition sequence
patterns of merging paths, of which Euclidean distance in the PRML signal processing is
relatively short, are targeted, and one signal evaluation index is generated based on the
differential metric information of a plurality of pattern groups, having a different error
generation probability and a different number of errors to be generated. Specifically, predicted
error rates are calculated from the mean values of the differential metric information, which are
not greater than the signal processing threshold of each pattern, the total thereof is calculated,
then a virtual standard deviation (hereafter a) of normal distribution is calculated from the total
of error rates, and the signal evaluation index, including this standard deviation a of the normal

distribution, is generated. As a result, it is possible to realize a signal evaluation method and
evaluation index having very high correlation with the error rate.
[0155] The pre-amplificr section 3, the AGC section 4 and the waveform equalization section 5
in the present embodiment in Fig. 2 may be constructed as one analog integrated circuit (LS1).
The pre-amplifier section 3, the AGC section 4, the waveform equalization section 5, the A/D
conversion section 6, the PLL section 7, the PR equalization section 8, the maximum likelihood
decoding section 9, the signal evaluation index detection section 100, and the optical disk
controller section 15 may be consolidated as one analog-digital-mixed integrated circuit (LSI).
[0156] In each of the above mentioned embodiments, a case of using the reproduction device as
the optical disk device was described. However, needless to say, the optical disk device of the
present invention is not limited to this, but can also be applied to a recording/reproduction device.
In this case, circuits for recording are added, but description thereof is omitted here, since a
known circuit structure can be used.
[0157] (Third Embodiment)
An optical disk device according to still another embodiment of the present invention will
now be described with reference to the drawings.
[0158] Fig. 13 is a block diagram depicting a general structure of the optical disk device of the
present embodiment.
[0159] The optical disk device 600 has: an optical head section 2, a pre-amplifier section 3, an
AGC (Automatic Gain Controller) section 4. a waveform equalization section 5, an A/D
conversion section 6, a PLL (Phase Locked Loop) section 7, a PR equalization section 8, a
maximum likelihood decoding section 9, a signal evaluation index detection section
(reproduction signal evaluation unit) 500 and an optical disk controller section 15. The
structures and functions of these composing elements constituting the optical disk device 600 are
the same as the first embodiment, and descriptions thereof are omitted here.
[0160] The optical disk device 600 according to the present embodiment has a signal evaluation
index detection section 500 as the reproduction signal evaluation unit. The signal evaluation
index detection section 500 has the same structure as the signal evaluation index detection
section 100 of the first embodiment, except for the setting of the signal processing threshold.
Hence composing elements having a similar structure and function as the signal evaluation index

detection section 100 of the first embodiment are denoted with a same symbol, and description
thereof is omitted.
[0161 ] As shown in Fig. 13, the signal evaluation index detection section 500 has a mean value
computing section 121 for computing a mean value of outputs of the differential metric
computing section 102, in addition to the structure of the first embodiment.
[0162] Now the operation of the mean value computing section 121 and how to set the signal
processing threshold will be described. In the first embodiment, a predetermined value, that is, a
code distance of ideal signals (a square of Euclidean distance between an ideal signal of a most
likely first state transition sequence and an ideal signal of a second most likely second state
transition sequence in a specific extraction target state transition pattern) is used as the signal
processing threshold. This is because in optimized recording, the mean value of outputs of the
differentia] metric computing section matches the code distance of the ideal signals. However,
as recording densities of optical disks further improve, recording optimization, to match the
mean value with the code distance of the ideal signals, may not be possible in some cases.
[0163] Therefore the signal evaluation index detection section 500 of the present embodiment
has the mean value computing section 121 for computing a mean value of outputs of the
differential metric computing section 102, and inputs this mean value to the level decision
section 103 as the signal processing threshold.
[0164] According to the foregoing structure, the signal processing threshold can be appropriately
set at the center of distribution, which is output from the differential metric computing section
121. Thereby correlation of the signal index value and the bit error rate, when the recording
density is increased, can be improved compared with the structure of the first embodiment.
[0165] Therefore the structure of the present embodiment, using the mean value of the
differential metric distribution as the signal processing threshold, is particularly useful when a
high density recording medium is adopted as the information recording medium 1.
[0166] (Fourth Embodiment)
An optical disk device according to still another embodiment of the present invention will
now be described with reference to the drawings.
[0167] Fig. 14 is a block diagram depicting the general structure of the optical disk device of the
present invention.

[0168] The optical disk device 800 has an optical head section 2, a preamplifier section 3, an
AGC (Automatic Gain Controller) section 4, a waveform equalization section 5, an A/D
conversion section 6, PLL (Phase Locked Loop) section 7, a PR equalization section 8, a
maximum likelihood decoding section 9, a signal evaluation index detection section
(reproduction signal evaluation unit) 700 and an optical disk controller section 15. The structure
and function of these composing elements constituting the optical disk device 800 are the same
as the second embodiment, so description thereof is omitted here.
[0169] The optical disk device 800 according to the present embodiment has a signal evaluation
index detection section 700 as the reproduction signal evaluation unit 700. The signal evaluation
index detection section 700 has the same structure as the signal evaluation index detection
section 300 of the second embodiment, except for the setting of the signal processing threshold.
Hence a composing element having a similar structure and function as the signal evaluation
index detection section 300 of the second embodiment is denoted with a same symbol, and
description thereof is omitted.
[0170] As shown in Fig. 14, the signal evaluation index detection section 700 has mean value
computing sections 121, 122 and 123 for computing a respective mean value of the outputs of
the differential metric computing sections 102, 107 and 112, in addition to the structure of the
second embodiment.
[0171] Now operation of the mean value computing sections 121, 122 and 123, and how to set
the signal processing threshold will be described. In the third embodiment, a predetermined
value, that is, a code distance of ideal signals (a square of Euclidean distance between an ideal
signal of a most likely first state transition sequence and an ideal signal of a second most likely
second state transition sequence in each extraction target state transition pattern), is used as the
signal processing threshold. This is because in optimized recording, the mean value of outputs
of the differential metric computing section matches the code distance of ideal signals. However,
as the recording densities of optical disks further improve, recording optimization to match the
mean value with the code distance of the ideal signals may not be possible in some cases.
[0172] Therefore the signal evaluation index detection section 700 of the present embodiment
has the mean value computing sections 121, 122 and 123 for computing a respective mean value
of outputs of the differential metric computing sections 102, 107 and 112, and inputs this mean
value to the level decision sections 103, 108 and 113 as the signal processing threshold
respectively.

[0173] According to the foregoing structure, the signal processing threshold can be appropriately
set at the center of distribution, which is output from each of the differential metric computing
sections 121, 122 and 123. Thereby correlation of the signal index value and the bit error rate
when the recording density is increased can be improved compared with the structure of the first
embodiment.
[0174] Therefore the structure of the present embodiment, using the mean value of the
differential metric distribution as the signal processing threshold, is particularly useful when a
high density recording medium is adopted as the information recording medium 1.
[0175] As described above, the reproduction signal evaluation method according to one aspect of
the present invention is a reproduction signal evaluation method for evaluating quality of a
reproduction signal reproduced from an information recording medium based on a binary signal
generated from the reproduction signal using a PRML signal processing system, comprising: a
pattern extraction step of extracting, from the binary signal, a specific state transition pattern
which has the possibility of causing a bit error; a differential metric computing step of computing
a differential metric, which is a difference of a first metric between an ideal signal of a most
likely first state transition sequence corresponding to the binary signal and the reproduction
signal, and a second metric between an ideal signal of a second most likely second state
transition sequence corresponding to the binary signal and the reproduction signal, based on the
binary signal of the state transition pattern extracted in the pattern extraction step; an extraction
step of extracting the differential metric which is not greater than a predetermined signal
processing threshold; a mean value computing step of determining a mean value of the
differential metrics which are not greater than the predetermined signal processing threshold and
extracted in the extraction step; a standard deviation computing step of determining a standard
deviation which corresponds to an error rate predicted from the mean value; and an evaluation
step of evaluating a quality of the reproduction signal using the standard deviation.
[0176] According to the foregoing method, specific state transition patterns which have the
possibility of causing a bit error are extracted from the binary signals generated by reproducing
the information recording medium. Here the state transition pattern which has a possibility of
causing a bit error is a state transition pattern having merging paths which could take a plurality
of state transitions when a predetermined state at a certain time transits to a predetermined state
at another time, and is a state transition pattern of merging paths of which Euclidean distance
between an ideal signal of a most likely first state transition sequence and an ideal signal of a
second most likely second state transition sequence is relatively short. If there are a plurality of

state transition patterns, which have the possibility of causing a bit error, a specific state
transition pattern is selectively extracted.
[0177] Targeting the binary signal of the extracted specific state transition pattern, a differential
metric, which is a difference of "a first metric between an ideal signal of a most likely first state
transition sequence corresponding to this binary signal and the above mentioned reproduction
signal", and "a second metric between an ideal signal of a second most likely second state
transition sequence corresponding to this binary signal and the above mentioned reproduction
signal", is calculated. Here the first metric is a square of the Euclidean distance between the
ideal signal of the first state transition sequence and the reproduction signal, and the second
metric is a square of the Euclidean distance between the ideal signal of the second state transition
sequence and the reproduction signal.
[0178] Of the calculated differential metrics, only those not greater than a predetermined signal
processing threshold are selectively extracted. In other words, an area in which the differential
metrics are great do net contribute to error generation, so an area in which the calculated
differential metrics are greater than the signal processing threshold is eliminated as an
unnecessary area to predict the error rate. Further, by limiting the target to predict the error rate
to differential metrics which are not greater than the signal processing threshold, error rate
prediction accuracy is improved.
[0179] A standard deviation, which corresponds to the error rate to be predicted, is then
determined from the mean value of the above mentioned extracted differential metrics which are
not greater than the signal processing threshold, and the quality of the reproduction signal is
evaluated using this standard deviation, hence signal evaluation having very high correlation
with the error rate becomes possible. As a result, a reproduction signal evaluation method,
which is suitable for a system using a PRML signal processing system and which can evaluate
the quality of the reproduction signal of the information recording medium at high accuracy, can
be implemented.
[0180] It is preferable that the signal processing threshold is a square of a Euclidean distance
between the ideal signal of the most likely first state transition sequence and the ideal signal of
the second most likely second state transition sequence.
[0181] According to the foregoing structure, the signal processing threshold corresponding to the
extraction target specific state transition patterns can be accurately set so as to match with the
Euclidean distance between the ideal signal of the first state transition sequence and the ideal

signal of the second state transition sequence. This is particularly effective to evaluate signals
where a plurality of state transition patterns, which have the possibility of generating an error,
are mixed.
[0182] It is preferable that the foregoing method further comprises a threshold computing step of
setting a computed value, acquired by averaging the differential metrics calculated in the
differential metric computing step, as the signal processing threshold.
[0183] According to this method, a signal processing threshold according to extraction target
predetermined state transmission patterns is determined by averaging the differential metrics
calculated in the differential metric computing step. Therefore a signal processing threshold
suitable for actual reproduction signals of the information recording medium can be acquired
each time. As a result, a more accurate predicted error rate can be determined, and furthermore
an appropriate quality evaluation of reproduction signals of information recording media
becomes possible.
[0184] The reproduction signal evaluation method according to another aspect of the present
invention is a reproduction signal evaluation method for evaluating quality of a reproduction
signal reproduced from an information recording medium based on a binary signal generated
from the reproduction signal using a PRML signal processing system, comprising: a plurality-of-
patterns extraction step of extracting, from the binary signal, a plurality of state transition
patterns, which have the possibility of causing a bit error; a differential metric computing step of
computing a differential metric, which is a difference of a first metric between an ideal signal of
a most likely first state transition sequence corresponding to the binary signal and the
reproduction signal, and a second metric between an ideal signal of a second most likely second
state transition sequence corresponding to the binary signal and the reproduction signal, based on
the binary signal respectively, for each state transition pattern extracted in the plurality-of-
patterns extraction step; an extraction step of extracting, for each state transition pattern, the
differential metrics which are not greater than a predetermined signal processing threshold,
which is set for each of the plurality of state transition patterns; a mean value computing step of
determining, for each state transition pattern, a mean value of the differential metrics which are
not greater than the signal processing threshold and extracted in the extraction step; an error rate
computing step of determining, for each state transition pattern, an error rate predicted based on
the mean value; a standard deviation computing step of determining a standard deviation which
corresponds to a sum of the error rate of each state transition pattern; and an evaluation step of
evaluating a quality of the reproduction signal using the standard deviation.

[0185] According to the foregoing structure, a plurality of state transition patterns which have
the possibility of causing a bit error are extracted from the binary signal generated by
reproducing the information recording medium. And differential metrics are calculated for each
of the extracted state transition patterns, and only the differential metrics which are not greater
than a predetermined signal processing threshold are selectively extracted. In other words, an
area in which the calculated differential metrics are greater than the signal processing threshold
is eliminated as an unnecessary area to predict the error rate, and by limiting the target to predict
the error rate only to differential metrics which are not greater than the signal processing
threshold, error rate prediction accuracy is improved. Then an error rate predicted from the
mean value of differential metrics which are not greater than the signal processing threshold is
determined for each state transition pattern respectively. Then a standard deviation, which
corresponds to the sum of the error rate of each state transition pattern, is determined, and the
quality of a reproduction signal is evaluated using this standard deviation, hence signal
evaluation having very high correlation with the error rate becomes possible. As a result, a
reproduction signal evaluation method, which is suitable for a system using the PRML signal
processing system and which can evaluate the quality of the reproduction signal of the
information recording medium at high accuracy, can be implemented.
[0186] It is preferable that the plurality-of-patterns extraction step extracts state transition
patterns of the binary signal for which a square of a Euclidean distance between the ideal signal
of the most likely first state transition sequence and the ideal signal of the second most likely
second state transition sequence is not greater than 14.
[0187] According to the foregoing structure, a state transition pattern which has the possibility of
causing a bit error is a state transition pattern having merging paths which could take a plurality
of state transitions when a predetermined state at a certain time transits to a predetermined state
at another time, and is a state transition pattern of merging paths of which Euclidean distance
between an ideal signal of a most likely first state transition sequence and an ideal signal of a
second most likely second state transition sequence is relatively short. The state transition
pattern of which square of the Euclidean distance is not greater than 14 is a pattern which has a
very high possibility of causing a bit error, and by targeting only such state transition patterns for
extraction, error rate can be efficiently predicted at high accuracy, and appropriate quality
evaluation for the reproduction signals of the information recording medium can be implemented.
[0188] It is preferable that the PRML signal processing system is a PR12221 system.

[0180] in this way, if the reproduction signal evaluation method is applied to a system using
PR 12221, the quality of the reproduction signal of the information recording medium can be
evaluated at high accuracy.
[0190] The reproduction signal evaluation unit according to another aspect of the present
invention is a reproduction signal evaluation unit for evaluating quality of a reproduction signal
reproduced from an information recording medium based on a binary signal generated from the
reproduction signal using a PRML signal processing system, comprising: a pattern extraction
section for extracting, from the binary signal, a specific state transition pattern which has the
possibility of causing a bit error; a differential metric computing section for computing a
differential metric, which is a difference of a first metric between an ideal signal of a most likely
first state transition sequence corresponding to the binary signal and the reproduction signal, and
a second metric between an ideal signal of a second most likely second state transition sequence
corresponding to the binary signal and the reproduction signal, based on the binary signal of the
state transition pattern extracted by the pattern extraction section; an extraction section for
extracting the differential metric which is not greater than a predetermined signal processing
threshold; a mean value computing section for determining a mean value of the differential
metrics which are not greater than the signal processing threshold and extracted by the extraction
section; and a standard deviation computing section for determining a standard deviation which
corresponds to an error rate predicted based on the mean value.
[0191] According to the foregoing structure, a standard deviation, which corresponds to the error
rate to be predicted based on the mean value of the differential metrics, which are not greater
than the extracted signal processing threshold, is determined, and the quality of the reproduction
signal is evaluated using this standard deviation. Hence signal evaluation having very high
correlation with the error rate becomes possible. As a result, a reproduction signal evaluation
unit, which is suitable for a system using the PRML signal processing system and which can
evaluate the quality of the reproduction signal of the information recording medium at high
accuracy, can be implemented.
[0192] It is preferable that the signal processing threshold is a square of a Euclidean distance
between the ideal signal cf the most likely first state transition sequence and the ideal signal of
the second most likely second state transition sequence.
[0193] According to the foregoing structure, the signal processing threshold corresponding to the
extraction target specific state transition patterns can be accurately set so as to match with the

liuclidean distance between the ideal signal of the first state transition sequence and the ideal
signal of the second state transition sequence. This is particularly effective to evaluate signals
where a plurality of state transition patterns, which have the possibility of generating an error,
are mixed.
[0194] It is preferable that the foregoing structure further comprises a threshold computing
section for setting a computed value, acquired by averaging the differential metrics calculated by
the differential metric computing section, as the signal processing threshold.
[0195] According to the foregoing structure, a signal processing threshold according to the
extraction target specific state transition pattern is determined by averaging the differential
metrics calculated by the differential metric computing section. Therefore a signal processing
threshold suitable for actual reproduction signals of the information recording medium can be
acquired each time. As a result, a more accurate predicted error rate can be determined, and
furthermore, an appropriate quality evaluation of reproductive signals of information recording
media can be implemented.
[0196] The reproduction signal evaluation unit according to another aspect of the present
invention is a reproduction signal evaluation unit for evaluating quality of a reproduction signal
reproduced from an information recording medium based on a binary signal generated from the
reproduction signal using a PRML signal processing system, comprising: a pattern extraction
section for extracting, from the binary signal, a plurality of state transition patterns which have
the possibility of causing a bit error; a differentia] metric computing section for computing a
differential metric, which is a difference of a first metric between an ideal signal of a most likely
first state transition sequence corresponding to the binary signal and the reproduction signal, and
a second metric between an ideal signal of a second most likely second state transition sequence
corresponding to the binary signal and the reproduction signal, based on the binary signal
respectively, for each state transition pattern extracted by the pattern extraction section; an
extraction section for extracting, for each state transition pattern, the differential metrics which
are not greater than a predetermined signal processing threshold, which is set for each of the
plurality of state transition patterns; a mean value computing section for determining, for each
state transition pattern, a mean value of the differential metrics which are not greater than the
predetermined signal processing threshold and extracted by the extraction section; an error rate
computing section for determining, for each state transition pattern, an error rate predicted based
on the mean value; and a standard deviation computing section for determining a standard
deviation which corresponds to a sum of the error rate of each state transition pattern.

[0197] According to the foregoing structure, a standard deviation which corresponds to a sum of
the error rate of each state transition pattern, is determined and the quality of the reproduction
signal is evaluated using this standard deviation. Hence a signal evaluation having very high
correlation with the error rate becomes possible. As a result, a reproduction signal evaluation
device, which is suitable for a system using the PRML signal processing system and which can
evaluate quality of the reproduction signal of the information recording medium at high accuracy,
can be implemented.
[0198] It is preferable that the pattern extraction section extracts patterns of the binary signal for
which a square of a Euclidean distance between the ideal signal of the most likely first state
transition sequence and the ideal signal of the second most likely second state transition
sequence is not greater than 14.
[0199] It is preferable that the PRML signal processing system is a PR12221 system.
[0200] A disk device according to another aspect of the present invention comprises a
reproduction section for generating a binary signal from a reproduction signal reproduced from
an optical disk, that is an information recording medium, using a PRML signal processing
system; and the reproduction signal evaluation unit according to each of the above mentioned
structures.
Industrial Applicability
[0201] The present invention is particularly useful in technical fields in which signal processing
is performed using the maximum likelihood decoding method.

WHAT IS CLAIMED IS:
1. A reproduction signal evaluation method for evaluating quality of a reproduction signal
reproduced from an information recording medium based on a binary signal generated from said
reproduction signal using a PRML signal processing system, comprising:
a pattern extraction step of extracting, from said binary signal, a specific state transition
pattern which has the possibility of causing a bit error;
a differential metric computing step of computing a differentia] metric, which is a
difference of a first metric between an ideal signal of a most likely first state transition sequence
corresponding to said binary signal and said reproduction signal, and a second metric between an
ideal signal of a second most likely second state transition sequence corresponding to said binary
signal and said reproduction signal, based on said binary signal of the state transition pattern
extracted in said pattern extraction step;
an extraction step of extracting said differential metric which is not greater than a
predetermined signal processing threshold;
a mean value computing step of determining a mean value of the differential metrics
which are not greater than the signal processing threshold and extracted in said extraction step;
a standard deviation computing step of determining a standard deviation which
corresponds to an error rate predicted from said mean value; and
an evaluation step of evaluating a quality of said reproduction signal using said standard
deviation.
2. The reproduction signal evaluation method according to Claim 1, wherein said signal
processing threshold is a square of a Euclidean distance between the ideal signal of the most
likely first state transition sequence and the ideal signal of the second most likely second state
transition sequence.
3. The reproduction signal evaluation method according to Claim 1, further comprising a
threshold computing step of setting a computed value, acquired by averaging the differential
metrics calculated in said differential metric computing step, as said signal processing threshold.

4. A reproduction signal evaluation method for evaluating quality of a reproduction signal
reproduced from an information recording medium based on a binary signal generated from said
reproduction signal using a PRML signal processing system, comprising:
a plurality-of-patterns extraction step of extracting, from said binary signal, a plurality of
state transition patterns, which have the possibility of causing a bit error;
a differential metric computing step of computing a differential metric, which is a
difference of a first metric between an ideal signal of a most likely first state transition sequence
corresponding to said binary signal and said reproduction signal, and a second metric between an
ideal signal of a second most likely second state transition sequence corresponding to said binary
signal and said reproduction signal, based on said binary signal respectively, for each state
transition pattern extracted in said plurality-of-patterns extraction step;
an extraction step of extracting, for each state transition pattern, said differential metrics
which are not greater than a predetermined signal processing threshold, which is set for each of
said plurality of state transition patterns;
a mean value computing step of determining, for each state transition pattern, a mean
value of the differential metrics which are not greater than said signal processing threshold and
extracted in said extraction step;
an error rate computing step of determining, for each state transition pattern, an error rate
predicted based on said mean value;
a standard deviation computing step of determining a standard deviation which
corresponds to a sum of the error rate of each state transition pattern; and
an evaluation step of evaluating a quality of said reproduction signal using said standard
deviation.
5. The reproduction signal evaluation method according to Claim 4, wherein said plurality-
of-patterns extraction step extracts state transition patterns of said binary signal for which a
square of a Euclidean distance between the ideal signal of the most likely first state transition

sequence and the ideal signal of the second most likely second state transition sequence is not
greater than 14.
6. The reproduction signal evaluation method according to Claim 4, wherein said PRML
signal processing system is a PR12221 system.
7. A reproduction signal evaluation unit for evaluating quality of a reproduction signal
reproduced from an information recording medium based on a binary signal generated from said
reproduction signal using a PRML signal processing system, comprising:
a pattern extraction section for extracting, from said binary signal, a specific state
transition pattern which has the possibility of causing a bit error;
a differentia] metric computing section for computing a differential metric, which is a
difference of a first metric between an ideal signal of a most likely first state transition sequence
corresponding to said binary signal and said reproduction signal, and a second metric between an
ideal signal of a second most likely second state transition sequence corresponding to said binary
signal and said reproduction signal, based on said binary signal of the state transition pattern
extracted by said pattern extraction section;
an extraction section for extracting said differential metric which is not greater than a
predetermined signal processing threshold;
a mean value computing section for determining a mean value of the differential metrics
which are not greater than the signal processing threshold and extracted by said extraction
section; and
a standard deviation computing section for determining a standard deviation which
corresponds to an error rate predicted based on said mean value.
8. The reproduction signal evaluation unit according to Claim 7, wherein said signal
processing threshold is a square of a Euclidean distance between the ideal signal of the most
likely first state transition sequence and the ideal signal of the second most likely second state
transition sequence.

9. The reproduction signal evaluation unit according to Claim 7. further comprising a
threshold computing section for setting a computed value, acquired by averaging the differential
metrics calculated by said differential metric computing section, as said signal processing
threshold.
10. A reproduction signal evaluation unit for evaluating quality of a reproduction signal
reproduced from an information recording medium based on a binary signal generated from said
reproduction signal using a PRML signal processing system, comprising:
a pattern extraction section for extracting, from said binary signal, a plurality of state
transition patterns which have the possibility of causing a bit error;
a differential metric computing section for computing a differential metric, which is a
difference of a first metric between an ideal signal of a most likely first state transition sequence
corresponding to said binary signal and said reproduction signal, and a second metric between an
ideal signal of a second most likely second state transition sequence corresponding to said binary
signal and said reproduction signal, based on said binary signal respectively, for each state
transition pattern extracted by said pattern extraction section;
an extraction section for extracting, for each state transition pattern, said differentia]
metrics which are not greater than a predetermined signal processing threshold, which is set for
each of said plurality of state transition patterns;
a mean value computing section for determining, for each state transition pattern, a mean
value of the differential metrics which are not greater than said predetermined signal processing
threshold and extracted by said extraction section;
an error rate computing section for determining, for each state transition pattern, an error
rate predicted based on said mean value; and
a standard deviation computing section for determining a standard deviation which
corresponds to a sum of the error rate of each state transition pattern.

11. The reproduction signal evaluation unit according to Claim 10, wherein said pattern
extraction section extracts patterns of said binary signal for which a square of a Euclidean
distance between the ideal signal of the most likely first state transition sequence and the ideal
signal of the second most likely second state transition sequence is not greater than 14.
12. The reproduction signal evaluation unit according to Claim 10, wherein said PRML
signal processing system is a PR 12221 system.
13. An optical disk device, comprising:
a reproduction section for generating a binary signal from a reproduction signal
reproduced from an optical disk, that is an information recording medium, using a PRML signal
processing system; and
the reproduction signal evaluation unit according to any one of Claims 7 to 12. J

A method for evaluating quality of a reproduced
signal in accordance with a binary signal generated
by using a PRML signal processing system from the
signal reproduced from an information recording medium,
comprises a pattern extracting step for extracting a
specific state transition pattern to possibly cause a bit error
from the binary signal, a step for calculating a differential
metric in accordance with the binary signal, an extracting
step for extracting the differential metric equal to
or less than a predetermined signal processing threshold
value, a step for seeking an average value of the differential
metric extracted by the extracting step so as to be
equal to or less than the signal processing threshold value,
a standard deviation calculation step for seeking a
standard deviation corresponding to an error rate predicted
from the average value, and an evaluation step for
evaluating the quality of the reproduced signal by using
the standard deviation.