2. Background Art Run length limited codes, generically designated as (d, k) codes, have been widely and successfully applied in modern magnetic and optical recording systems. Such codes, and means for implementing said codes, are described by K. A. Schouhamer Immink in the book entitled"Codes for Mass Data Storage Systems" (ISBN 90-74249-23-X, 1999).

Run length limited codes are extensions of earlier non return to zero recording (NRZ) codes, where binarily recorded"zeros"are represented by no (magnetic flux) change in the recording medium, while binary"ones"are represented by transitions from one direction of recorded flux to the opposite direction.

In a (d, k) code, the above recording rules are maintained with the additional constraints that at least d"zeros"are recorded between successive data"ones", and no more than k"zeros"are recorded between successive data"ones". The first constraint arises to obviate intersymbol interference occurring due to pulse

crowding of the reproduced transitions when a series of"ones"are contiguously recorded. The second constraint arises in recovering a clock from the reproduced data by"locking"a phase locked loop to the reproduced transitions. If there is too long an unbroken string of contiguous"zeros"with no interspersed"ones", the clock regenerating phase-locked-loop will fall out of synchronism.

In, for example, a (1,7) code there is at least one zero' between recorded"ones", and there are no more than seven recorded contiguous"zeros"between recorded"ones". The series of encoded bits is converted, via a modulo-2 integration operation, to a corresponding modulated signal formed by bit cells having a high or low signal value, a one'bit being represented in the modulated signal by a change from a high to a low signal value or vice versa.

A zero'bit is represented by the lack of change of the modulated signal.

The minimum inversion period Tmin, which can be expressed by (d+l) T is thus equal to 2T where T is a bit time interval in the recording wave train. The maximum inversion period Tmax, which can be expressed by (k+1) T, is thus equal to 8T.

By the way, in a train of channel bits generated by a (1,7) code the minimum inversion period Tmin is more frequently observed than inversion periods of length 3T, 4T, etc. The fact that a lot of edge information is generated at short intervals such as 2T and 3T is advantageous to the generation of a clock signal in many cases.

As the recording density is increased, however, the minimum inversion period Tmin this time becomes a problem. That is if

minimum runs 2T are generated consecutively the recording wave train is prone to distortion generated therein. This is because a 2T wave output amplitude is smaller than other wave output amplitudes and, hence, easily affected by factors such as a defocus and a tangential tilt.

In addition, at a high line density, recording of consecutive minimum marks (2T) is also easily affected by disturbances such as noise. Thus, an operation to play back the data will also be prone to errors. In this case, a pattern of errors in reproduction of the data is observed as shifts of the front and rear edges of a minimum mark in many cases. As a result, the length of the generated bit error increases.

As described above, when data is transmitted through a transmission line or recorded onto a medium, the data is modulated into a coded sequence matching the transmission line or recording medium prior to the transmission or recording. If the coded sequence resulting from the modulation contains a direct current (DC) component, a variety of error signals such as tracking errors generated in control of a servo of the disk drive become prone to variations or jitter are generated easily.

The first reason for using said dc-free signals is that recording channels are not normally responsive to low-frequency components. The suppression of low-frequency components in the signal is also highly advantageous when the'signal is read from an optical record carrier on which the signal is recorded in the track, because then continuous tracking control undisturbed by the recorded signal is possible.

A good suppression of the low-frequency components leads to improved tracking with less disturbing audible noise. For this reason it is thus desirable to make as many efforts to prevent the modulated sequence from containing a direct current component as possible.

In order to prevent the modulated sequence from containing a direct current component, control of a DSV (Digital Sum Value) to prevent the modulated signal from containing a direct current component has been proposed. The DSV is a total found by adding up the values of a train of bits, wherein the values +1 and-1 are assigned to 1'and 0'in the train respectively, which results after NRZI modulation of a train of channel bits. The DSV is an indicator of a direct current component contained in a train of sequences.

A substantially constant running digital sum value (DSV) means that the frequency spectrum of the signal does not comprise frequency components in the low frequency area. Note that DSV control is normally not applied to a sequence generated by a standard (d, k) code. DSV control for such standard (d, k) codes is accomplished by calculating a DSV of a train of encoded bits after the modulation for a predetermined period of time and inserting a predetermined number of DSV control bits into the train of encoded bits. In order to improve the code efficiency it is desirable to reduce the number of DSV control bits to a smallest possible value.

Preferably, the encoded signal comprises a sequence of q code words, where q is an integer. Between encoded signal portions are inserted synchronization (sync) signals. Preferably, the sync

signal should not occur in a sequence of the encoded signal.

Conventionally, the sync pattern contains a series of s consecutive bits equal to the logical"0", where s is an integer exceeding k, or alternatively the sync pattern consists of two series of k bits having a logical"0"separated by a bit having a logical"1", i. e. two consecutive runs of k"O"s.

A disadvantage of the usage of such sync patterns is that they are relatively long, and therefore reduce the efficiency of the recording. Therefore, preferably, a short sync pattern is used, which may comprise a sequence of two or more consecutive"0"runs.

An example of the use of such signals to record and read an audio signal on an optical or magneto-optical record carrier can be found in United States Patent Specification 4,501,000. The specification describes the Eight-to-Fourteen (EFM) modulation system, which is used for recording information on Compact Disks (CD) or MiniDisk (MD). The EFM-modulated signal is obtained by converting a series of 8-bit information words into a series of 14-bit code words, and where 3-bit merging words are inserted between consecutive code words.

Respective code words of 14 bits satisfy the conditions that at least d=2 and at most k=10"O"s are placed between two"l"s.

In order to satisfy this condition also between code words, 3-bit merging words are used. Four 3-bit merging words of 8 possible 3-bit merging words are permitted to be used, namely"001","010","000", and"100". The remaining possible 3-bit merging words, namely"111", "011","101", and"110"are not used as they violate the prescribed d=2 constraint.

One of the four allowed merging words is selected such that the bit string obtained after cascading alternate code words and merging words satisfies the (d, k)-constraint, and that in the corresponding modulo-2 integrated signal the DSV remains substantially constant. By deciding the merging words according to above rules, low-frequency components of the modulated signal can be reduced.

The choice for the 3-bit merging words is based on the requirement that, on the one hand the channel signal is satisfied.

Decoding of EFM signals is very simple. A sync pattern is multiplexed between 33 alternative 3-bit merging words and 14-bit code words. The 27-bit sync pattern used in the CD format consists of two consecutive runs of then"O"s plus 3-bit'merging word.

The choice of the merging words avoids the occurrence of said sync pattern in the output sequence. The relative occurrence of the sync pattern in this format is, 27 bits in a total of 588 bits, 4.6%. The decoded skips the 3-bit merging words, and the 14-bit code words are translated, using a look-up table or PLA etc., into the information bytes (8 bits).

Information recording has a constant need for increasing the reading and writing speed. The aim of increased reading speed, however, requires higher servo bandwidth of the tracking mechanism, which, in turn, sets more severe restrictions on the suppression of the low-frequency components in the recorded signal.

Improved suppression of the low-frequency components is also advantageous for suppressing audible noise arising from the tracking mechanism. For this reason, it is desirable to make as

many efforts to prevent the signal from containing low-frequency components.

3. Disclosure of Invention It is an object of the present invention to provide a coding system being able to generate for each data word a corresponding sequence, which can suppress dc-components, and does not contain a sync pattern, long string of O's, and long runs of the smallest runlength d, under the rules of the (d, k) code for recording onto a recordable medium.

It is another object of the present invention to provide method and apparatus of converting a series of data words into a modulated signal, which can write more edge information under the rules of a given (d, k) code, thereby stabilizing a channel clock in reproduction.

A method of converting a series of data words into a modulated signal according to the present invention, generates for each data word a number of intermediate sequences by combining mutually different digital words with the data word, scrambles each alternative sequence to form an alternative sequence, translates each alternative sequence into a (d, k) constrained sequence according to a predefined coding rate m/n, sorts out sequences satisfying k runlength limit of"O"s persisting between the sequence and its previous (d, k) constrained sequence in the translated (d, k) constrained sequences, and selects one (d, k) constrained sequence for recording onto a recordable medium among the sorted sequences.

Another method of converting a series of data words into a

modulated signal according to the present invention, generates for each data word a number of intermediate sequences by combining mutually different digital words with the data word, scrambles each alternative sequence to form an alternative sequence, translates each alternative sequence into a (d, k) constrained sequence according to a predefined coding rate m/n, sorts out sequences, each having at most KSET'0"runs between two"1"s where kSET is smaller than k, in the translated (d, k) constrained sequences, and selects one (d, k) constrained sequence for recording onto a recordable medium among the sorted sequences.

The method and apparatus of converting a series of data words into a modulated signal according to the present invention, makes it possible to manufacture simple hardware for a coding system suppressing dc-components of sequences, and excluding a sync pattern, long string of O's, and long runs of the smallest runlength d from a recording sequence, under the rules of the (d, k) code. In addition, this invention can write more edge information under a given (d, k) code system, so that it can improve stabilization of a reproduction clock, which results in reduction of locking fail.

4. Brief Description of Drawings The accompanying drawings, which are included to provide a further understanding of the invention, illustrate the preferred embodiments of the invention, and together with the description, serve to explain the principles of the present invention.

In the drawings: Figure 1 shows a block diagram of an embodiment a coding

system according to the present invention; Figure 2 shows an explanatory diagram of part of an embodiment of a coding scheme used for carrying out augmenting and scrambling of the digital words; Figure 3 shows a block diagram of a selector for use in a coding system according to the present invention; Figure 4 is a diagram that illustrates the general method for judging the alternative sequences ; Figure 5 shows schematically the k-condition judging operation for alternative (d, k) constrained sequences ; Figure 6 shows an embodiment of dividing a data word into two ones which are inputted to a generator sequentially ; and Figure 7 shows a block diagram of a demodulator for demodulating (d, k) constrained sequences having been recorded according to the present invention.

5. Modes for Carrying out the Invention In order that the invention may be fully understood, a preferred embodiment thereof will now be described with reference to the accompanying drawings.

Figure 1 shows a block diagram of an embodiment of an encoding system according to the invention.

Using a generator 20, a selector 22, the encoding system translates user data 19 into a (d, k) constrained sequence 23, wherein a plurality of predefined subsequence are fully absent or occur with small probability. The (d, k) constrained sequence, in turn, is translated, using a precoder 24, into a runlength-limited sequence 25 with suppressed low-frequency components.

As shown in Figure 1 the coding system comprises a generator 20, whose detailed block diagram is displayed in Figure 2. The generator 20 comprises an augmentor 40 that generates for each word a number of intermediate sequences 41 by combining mutually different digital words with the data word 19. The intermediate sequences 41 can be generated augmentor 40 simply by placing the digital words in front, middle, or rear of the data word 19 The generator 20 further comprises a scrambler 42 that scrambles the intermediate sequences 41, one after each other, in order to form a selection set of alternative sequences 21. The inclusion of the mutually different digital words in the intermediate sequences 41 has the effect that the scrambler 42, which is preferably a self-synchronized scrambler, is initialized for each intermediate sequence 41 with a different digital word.

Hence the alternative sequences 21 are relatively good randomizations of the data word 19.

Preferably the augmentor is embodied so as to generate for each data word 19 2r intermediate sequences 41 by combining all possible digital words of length r with the data word 19. In this way a selection set of alternative sequences 21 is obtained which is optimally randomized.

Figure 3 shows a detailed block diagram of the selector 22.

The selector 22 comprises a (d, k) encoder 50, which translates each alternative sequence 21 into a (d, k) constrained sequence 51. To that end, the alternative sequence 21 is partitioned into q m-bit words, where q is an integer. Under the rules of (d, k) encoder 50, the q m-bit words are translated into q n-bit words, wherein n>m.

The (d, k) encoder 50 can be of a standard type with parameter m=2, n=3, d=1, k=7 or alternatively m=1, n=2, d=2, k=7.

Preferably in order to achieve a high coding efficiency the encoder 50 has parameters m=9, n=13, d=1, or m=11, n=16, d=1, or m=13, n=19, d=1. Reference is made in this respect to the not published PCT application No. PCT/KR00/01292 (corresponding US application No. 09/707,947). The encoder 50 may also has parameters m=6, n=11, d=2 or m=11, n=20, d=2, or m=7, n=13, d=2. Reference is made in this respect to the not published PCT application No.

PCT/KR01/00359.

The selector 22 further comprises means 52 that determine for each alternative (d, k) constrained sequence 51 if the sequence 51 contains an undesired subsequence such as the sync pattern, a long string of"O"s, or a long string of alternative Tmin runs. If such an undesired subsequence is observed, then a judgment circuit will compute the penalty to be associated with that undesired subsequence.

The selector 22 further comprises means 52 that judges each alternative (d, k) constrained sequence 51 on the number of occurrences of undesired subsequence such as the sync pattern, or a long string of"O"s, or long string of alternative Tmin runs, and the contribution of the alternative sequence 21 to the low-frequency components.

Under the rules of the penalty algorithm, the judging means 52 gives a lower penalty for desired and a high penalty for undesired sequences. The selector 22 also comprises means 54 that selects the alternative (d, k) constrained sequence 51 with the

lowest penalty.

Figure 4 is a diagram that illustrates the general method used in accordance with the present invention to judge and select the alternative (d, k) constrained sequence 51 with the lowest penalty.

As depicted in Figure 4 the judging means 52 comprises a number of metric calculators, which measure in parallel the"0"runlength 60, the occurrence of a prescribed sync pattern 62, the alternate Tmin runlength 64, and the low-frequency content 66, respectively.

The"0"runlength metric is used as a measure of consecutive "O"s (commonly referred to as a"0"runlength) detected within an alternative (d, k) constrained sequence 51. As discussed briefly above, when a"0"run persists within the sequence for an extended period, the recorded features such as pits and lands can become prohibitively long, which can be deleterious so that mistracking and errors are more likely to occur. Therefore, the longer a"0" runlength is, the higher the penalty to be given is.

The runlength calculator 60 checks whether or not k-requirement (consecutive"O"s should not run more than k) is satisfied between two consecutive sequences. To do this, a previous sequence is stored temporally, and, if a new alternative (d, k) constrained sequence 51 is entered, consecutive"O"s running across the previous and the new sequence, that is, trailing"0" runlength PS_ZR'of the previous and leading"0"runlength CS-ZR' of the new are added. Then, it is checked whether the added value PS_ZR'+'CS-ZR'is larger than k.

If PS-ZR'+'CS-ZR'>k, the runlength calculator 60 flags the new (d, k) constrained sequence 51, otherwise the sequence remains

unflaged. Figure 5 shows schematically this k-requirement checking operation.

The runlength calculator 60 may use ksET for new consecutive zero requirement of an alternative (d, k) constrained sequence, where kSET is smaller than k which is used in the m/n translation of the (d, k) encoder 50. In this case that kSET smaller than k has been set, the runlength calculator 60 scans, for each alternative (d, k) constrained sequence, inner consecutive"O"s as well as border consecutive"O"s to be connected with each neighboring sequence in order to check whether an alternative (d, k) constrained sequence has"0"runlength exceeding kSET-If it has, the runlength calculator 60 flags current sequence and then applies the sequence to selection means 54.

Although a threshold smaller than k of (d, k) modulation is set for consecutive 0'limit, there is at least one (d, k) constrained sequence having all consecutive"0"subsequences whose length is equal or shorter than kSETif the length r is large because 2 : (d, k) constrained sequences 51 are generated. Therefore, edge information can be recorded more frequently than in k requirement.

If a table for m/n translation of the (d, k) encoder 50 has translation data satisfying k=14 requirement, the runlength calculator 60 selects one among 9,10,11,12, and 13 as kSET-If ksar is too small, there may be no alternative (d, k) constrained sequence satisfying kSET requirement among 2r alternative sequences.

Therefore, too small kSET/for example smaller than 8, is excluded in the selectable set.

If a table of the (d, k) encoder 50 has translation data

satisfying k=18 requirement, kSET is chosen among 9 through 17.

The metric calculator 64 measures the number of consecutive Tmin (if d=l, Tmin is"01", if d=2, Tmin is"001") runlength in order to exclude a sequence having much repetitive Tmin's violating MTR (Maximum Transition Run) constraint, for example, a sequence of"01010101..."or"001001001001...". The metric calculator 64 flags such sequences which will be excluded from selectable alternative sequences by selection means 54.

The sync detector 62 detects if prescribed sync pattern occur in an alternative (d, k) constrained sequence 51. If, indeed, such a sync pattern is detected, the sync detector 62 flags that (d, k) constrained sequence, otherwise the sequence remains unflaged.

The metric calculator 66 measures the DSV of the alternative (d, k) constrained sequence 51 after this sequence had been modulated using a precoding device. Preferably the variance of the DSV is measured, as it is often a more adequate measure when sequence are relatively long, exceeding 100 bits.

The various metrics and the sync detector and k-condition flags are inputs of selection means 54. The selection means 54 finally takes a decision based on weights associated with the various input metrics of which of the alternative sequences will be selected and recorded. In this decision, alternative (d, k) constrained sequences whose flag is set by the judging means 52 are excluded.

In the preferred embodiment of the present invention, a sync pattern is used, which consists of at least two"0"runs, which are shorter than k. As a result, coding efficiency will benefit

from such a relatively short sync pattern.

The alternative (d, k) constrained sequence 51 that has been selected by selector 54 is converted into a modulated signal using the NRZI pre-coding procedure. Then, the modulated signal is generated by the selected (d, k) constrained sequence 51 integrated modulo-2 in which a 1'becomes a transition and a 0'becomes an absence of a transition and forwarded to the recording medium.

Meantime, the length r of the digital word is determined such that bit length A of the data word 19 added by r is multiple of m.

For example, in coding rate m/n=9/13, if A is 728, r satisfying the multiple condition is 1,10,19,..., so that one is selected among these values for r. Considering the conventional ECC block scheme of DVD in which one sync frame has 91-byte user data, the condition of A=728 is used.

If 1 is selected for r, the number of digital words combined with an input data word 19 is two (=21), which generates only two intermediate sequences 41. However, it is unlikely that the selection means 54 can obtain a desired alternative (d, k) constrained sequence 51 among the two. Therefore, r=10 is preferable since 1024 (=210) intermediate sequences 41 can provide at least one desired (d, k) constrained sequence after good randomization.

Figure 6 shows another method to select the length r. In this method, a 728-bit input data word is divided into two 364-bit data words 19 which are applied to the augmentor 40 sequentially. For this division, a dividing means is placed in front of the augmentor

40. The augmentor 40 generates for each 364-bit word 32 intermediate sequences 41 by combining mutually different (=2) digital words with the data word 19 since 364+5 (r) is a multiple of 9 (m). The 32 intermediate sequences 41 are outputted in parallel from each register of the augmentor 40.

If coding rate m/n=6/11 with A=728, r satisfying the multiple condition is 4,10,16,.... Among these bit lengths of r, 4 is preferable because 16 (=24) intermediate sequences 41 probably achieves reasonable randomization and requires proper number of registers.

Another method to select the length r for the case of m/n=6/11 is to divide a 728-bit input data word into eight 91-bit data words which are in turn applied to the augmentor 40. In this division embodiment, the best r is 5 because 91+5 (r) is a multiple of 6 (m).

Therefore, 32 (=25) digital words and intermediate sequences 41 are generated, respectively.

In the above embodiments, it is preferable that the bit length r is proportional to length of a data word 19. Thus, if length of a data word 19 is smaller than the above example, r may be chosen from bit lengths below 5.

Recorded data after modulated according to the above-explained method are restored by a demodulator conducting the above-explained method in reserve, that is, decoding, de-scrambling, and eliminating r-bit digital word, sequentially.

Figure 7 is a block diagram of a demodulator conducting such operations. A sync detector 101 detects a sync code contained in a sequence, whose frame size is (A+rJ *m/n + sync bit length,

reproduced from an optical record carrier, and rem sync code. Then, the sequence 23, whose frame size is (A+r) *m/n, satisfying the (d, k) constraint is applied to a decoder 102 which decodes each n-bit word in the sequence into corresponding m-bit word, thus, frame size of the decoded sequence becomes A+r. A de-scrambler 103 restores the decoded sequence into an intermediate sequence 41 which is one before scrambled. Then, an augment eliminator 104 removes r-bit digital word contained in the front, middle, or rear of the original sequence, and QUtpUtS an original data word 19 whose frame size is A.

The invention may be embodied in other specific forms without departing from the sprit or essential characteristics thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.