Generation of a runlength limited digital information signal

[0002] The invention is-related to the field of encoding digital information signals.

[0003] The invention relates to an apparatus for generating a runlength limited (RLL) digital information signal, the digital information signal having a minimum runlength of d′ and a maximum runlength of k′, d′ and k′ being integers larger than zero and where k′ is larger than d′ and to a method for generating such runlength limited digital information signal. The invention further relates to an apparatus for and a method of generating a (d,k) sequence.

[0004] Those skilled in the art are directed to the following related documents:

[0005] (D1) U.S. Pat. No. 5,477,222 (PHN, 14.448)

[0006] (D2) K. A. Schouhamer Immink, Coding Techniques for Digital Recorders, Prentice Hall, 1991.

[0007] (D3) P. Funk; Run-Length-Limited Codes with MultipleSpacing, IEEE Transactions on Magnetics, Vol. MAG-18, 1982, pp. 772-775.

[0008] (D4) U.S. Pat. No. 5,696,505 (PHN 14.746).

[0009] An apparatus as defined in the opening paragraph is eg. known from U.S. Pat. No. 5,477,222, document D1 in the list of related documents.

[0010] Runlength limited digital information signals obtained from channel coding an information signal are described in the book coding techniques for digital records by K. A. Schouhamer Immink, Prentice Hall, 1991, chapter 5, document D2. Such runlength limited digital information signals are generally derived from so-called (d,k) constrained sequences. In such sequences, expressed in NRZ-I (non-return-to-zero-inverse) notation, two logical ‘ones’ in the sequence are separated bva run of consecutive ‘zeros’ of length at least d, and any run of consecutive ‘zeros’ is of length at moist k. A (d,k) sequence is normally converted in a precoding step so as to obtain the runlength limited information signal. In such runlength limited information signal obtained from the above (d,k) sequence, minimum runlengths of d+1 of consecutive like symbols occur and maximum runlengths of k+1 consecutive like symbols occur in sequences expressed in NRZ notation. This precoding of a (d,k) sequence, for d=1 and k=7 as an example, results in a runlength limited information signal having a minimum runlength of 2 and a maximum runlength of 8.

[0011] A choice of a specific channel code, properly adopted to the physical channel characteristics, is limited by the fact that the d-constraint is restricted to the natural numbers (d=0, d=1, d=2, . . . ). It has been proposed in the past to remove this limitation, by using rational d-constraints. Reference is made in this respect to the publication by P. Funk, document D3 in the list of related documents.

[0012] The above documents are hereby incorporated herein in whole by reference.

[0013] The invention has for its object to provide a different alternative for an apparatus which is capable of generating channel codes which fill up the gap between d and d+1 for the traditional (d,k)-based codes.

[0014] In accordance with a first aspect of the invention, the channel code generating apparatus is adapted to generate the runlength limited digital information signal, such that a minimum runlength in the digital information signal occurs isolated from other minimum runlengths in the digital information signal.

[0015] In accordance with a second aspect of the invention, the channel code generating apparatus for generating a runlength limited digital information signal, the digital information signal having a minimum runlength of d′ and a maximum runlength of k′, d′ and k′ being integers larger than zero and where k′ is larger than d′. Also, the channel code generating apparatus is adapted to generate the runlength limited digital information signal, such that a minimum runlength occurring in the digital information signal follows only a directly precedingrunlength larger than d′.

[0016] In accordance with a third aspect of the invention, the channel code generating apparatus for generating a runlength limited digital information signal, the digital information signal having a specific minimum runlength and a maximum runlength of k′, the apparatus including first apparatus for generating runlengths in the digital information signal of length d′+1 at minimum, where d′ and k′ are integers larger than zero and where k′ is larger than d′+1. Also, the channel code generating apparatus further includes second apparatus for generating runlengths of d′ in the runlength limited digital information signal, such that such runlengths of d′ occur isolated from other such runlengths of d′ in the digital information signal.

[0017] In accordance with a fourth aspect of the invention, the channel code generating apparatus is adapted for generating a runlength limited digital information signal, the digital information signal having a specific minimum runlength and a maximum runlength of k′, the channel code generating apparatus including first apparatus comprising means for generating runlengths in the digital information signal of length d′+1, where d′ and k′ are integers larger than zero and where k′ is larger than d′+1. Also, the channel code generating the apparatus further includes second apparatus for generating runlengths of d′ in the runlength limited digital information signal, such that such runlengths of d′ occur only following a directly preceding runlength larger than d′.

[0018] The invention generates (d,k)-based RLL channel codes, with the additional allowance of isolated Id runlengths (that is: isolated runlengths of length d), which are normally forbidden in a pure (d,k) code. An isolated Id runlength has to be surrounded by two other runlengths In with n>d. In contrast, for a pure (d−1,k) RLL channel code, the appearance of Id's is unconditional. Therefore, the capacity of the new (d,k) code with isolated Id's (called for short an iso-Id(d,k) code), is between the capacity of the traditional (d,k) and (d−1,k) channel codes. The increase in capacity of the iso-Id(d,k) code compared to the pure (d,k)-code, goes together with an increase in the highest frequency in the bit-sequence, which equals 1/(2d+1) for the iso-(d,k) code, which is between 1/(2d+2) and 1/2d (for the pure (d,k) and (d−1,k) codes, respectively).

[0019] It is to be understood that, where the claims define a RLL digital information signal with a minimum runlength of d′, d′ should be taken equal to the parameter d introduced above, in order to bring the claims in conformity with the above argumentation.

[0020] These and other objects of the invention will be elucidated below in the following figure description, in which

[0021] FIG. 1 shows a state-transition diagram for an iso-I1(d=1,k=4) code,

[0022] FIG. 2 shows the state-transition diagram for an iso-I1(d=1,k=7) code,

[0023] FIG. 3 shows the state-transition diagram for an iso-I2(d=2,k) code, and

[0024] FIG. 4 shows an embodiment of the apparatus in accordance with the invention,

[0025] In the following figure description, practical examples are worked out for a number of channel codes.

[0026] Iso-Id(d,k) code implementations with mapping of n user bits to m channel bits, have been constructed with the parity preserving property, so that DC-control can be applied. Reference is made in this respect to U.S. Pat. No. 5,477,222, document D1 in the list of related documents. Parity preserving implies that user words of n bits long, and channel words of m bits long, have the same parity (which is the number of 1's in a word, modulo 2). This property must also hold for the substitution encoding tables of the code (with 2n-to-2m mapping), which are used when runlength violations appear upon concatenation of channel words by use of the standard encoding tables (with n-to-m mapping).

[0027] Next a Rate R=4/5, iso-I1(d=1,k=4) Code will be described. The Finite-State Runlength Diagram of the code is shown in FIG. 1. The states numbered 1 to 5 in the state diagram of FIG. 1 describe the normal (d=1,k=4) code. By going from the state 1 to the state 2 and back to the state 1, a sequence...01010...is obtained, which is a representation of the d(=1) constraint. That is: at minimum one ‘zero’ occurs between two subsequent ‘ones’ in the sequence, expressed in NRZ-I notation. Further, by going through the states numbered 1, 2, 3, 4, 5 and back to the state 1, a sequence...01000010...is generated, which represents the k constraint: four ‘zeroes’ at most occur between two consecutive ‘ones’ in the sequence. It can be seen that, upon 1T precoding this NRZ-I sequence, a runlength limited sequence is obtained having runlengths of at minimum 2 and at maximum 5 runs of consecutive like symbols (in NRZ notation), or...+--- ++...and...+-----+..., respectively.

[0028] It can further be seen that, apart from the traditional runlength constraints on the number of successive zeroes, there is now also a constraint on the number of successive ones, which should not exceed 2. This is realized by adding the state numbered 1′. By going through one of the states numbered 2, 3, 4 or 5, via the state 1 and the state 1′, and going back to the state 2, a sequence...0110.... has been obtained. Upon 1T precoding, this results in the occurrence of a run of 1, or ...++-++....

[0029] The capacity of the iso-I1(d=1,k=4) code equals C=0.8376. The code implementation has a rate R=0.8, thus yielding a code efficiency of 95.5%. An iso-I1(d=1,k=3) code, which has a more severe constraint on k, leads to a capacity of C=0.7947. Such code can thus not be realised with a 4-to-5 mapping.

[0030] The standard conversion rules for the iso-I1(d=1,k=4) code are shown in table I A and table I B. The table I A is the conversion table for the conversion of the even parity words and the table I B is the conversion table for the odd parity words.

[0031] Further, substitution tables are needed to maintain the various constraints across the boundaries of consecutive converted words. The table II A corrects for the runlength violation otherwise occurring when concatenating some of the even parity input words, more specifically when concatenating the input words numbered 1 and 7, or when concatenating the input words 1 and 8, or when concatenating the input words 6 and 1 or when concatenating the input words 8 and 1, of the table I A. The middle column in the table II A shows that the runlength violations are in the form of three directly successive ‘ones’ or in the form of five directly successive ‘zeroes’, assuming the conversion in accordance with the rules of the table I A would have been carried out. The right column gives the substitution sequence. It should be noted that the second 5-bit word in the substitution table is not included in the converted words of the tables I A and I B. Therefore, those words can be used as an identification, upon decoding, that the two 5-bit converted words should be reconverted in accordance with the table II A.

[0032] Table II B corrects for the runlength violation otherwise occurring when concatenating some of the even parity input words from the table I A with some of the odd parity input words from the table I B. Table II C corrects for the runlength violation otherwise occurring when concatenating some of the odd parity input words from the table I B with some of the even parity input words from the table I A. Table II D corrects for the runlength violation otherwise occurring when concatenating some of the odd parity input words from the table I B.

[0033] Summarizing, the advantages of the R=⅘iso-I1(d=1,k=4) code are:

[0034] standard mapping (4-to-5) and substitutional mapping (8-to-10) compatible with byte-oriented error correction, and error-propagation only occurs when the substitution tables during decoding involve two successive bytes

[0035] low k-value, thus limited intersymbol-interference (ISI), and proper clocking.

[0036] DC-control via parity preserve principle.

[0037] relatively simple encoding tables.

[0038] Next, a Rate R=6/7 iso-I1(d=1,k=7) code will be described. The finite-state runlength diagram of the code is shown in FIG. 2. The states numbered 1 to 8 in the state diagram of FIG. 2 describe the normal (d=1,k=7) code. By going from the state 1 to the state 2 and back to the state 1, a sequence ...01010... is obtained, which is a representation of the d(=1) constraint. That is: at minimum one ‘zero’ occurs between two subsequent ‘ones’ in the sequence, expressed in NRZ-I notation. Further, by going through the states numbered 1, 2, 3, ......., 7, 8 and back to the state 1, a sequence ...01000000010... is generated, which represents the k constraint: seven ‘zeroes’ at most occur between two consecutive ‘ones’ in the sequence. It can be seen that, upon 1T precoding this NRZ-I sequence, a runlength limited sequence is obtained having runlengths of at minimum 2 and at maximum 8 runs of consecutive like symbols (in NRZ notation), or ...+--+... and ...+--------+.. , respectively.

[0039] It can further be seen that, apart from the traditional runlength constraints on the number of successive zeroes, there is now also a constraint on the number of successive ones, which should not exceed 2. This is realized by adding the state numbered 1′. By going through one of the states numbered 2, 3, 4 or 5, via the state 1 and the state 1′, and going back to the state 2, a sequence ....0110.... has been obtained. Upon 1T preceding, this results in the occurrence of a run of 1.

[0040] The capacity of the RLL code equals C=0.8732. The code implementation has a rate R=0.8571, thus yielding a code efficiency of 98.2%.

[0041] The standard encoding rules for the code is shown in the tables III A and III B. The table III A discloses the conversion of even parity input words into converted words and the table III B discloses the conversion of the odd parity input words.

[0042] Further, the runlength violation problems which would arise upon free concatenation by the standard table, are removed by the substitution tables, shown in the table IV A to IV D, where the table IV A solves the problem when concatenating even parity input words from the table III A, the table IV B solves the problem when concatenating even parity words from the table III A with odd parity input words from the table III B, the table IV C solves the problem when concatenating odd parity input words from the table III B with even parity input words from the table III A and the table IV D solves the problem when concatenating odd parity words from the table III B.

[0043] Next, a rate R=3/5, iso-I2(d=2,k=10) Code is described. This code has a rate between R=1/2 of EFMPlus (d=2,k=10), see document D4 in the list of documents at the end of the description, and R=2/3 of the code disclosed in document D1.

[0044] The Finite-State Runlength Diagram of the code is shown in FIG. 3. It can be seen that apart from the traditional runlengths I3, I4, ...I(k+1), now also runlengths I2 are allowed, on condition that they are not preceeded nor succeeded by another I2. This is outlined in the finite state runlength diagram by going from state 2 to state k+2, emitting a 1, after which two zeroes are emitted on passing along state k+3 on the way to state 3. Continuing from state 3 on, the minimum runlength is an I3. The capacity of the RLL code (for k=10) equals C=0.646057. The code implementation has a rate R=0.6, thus yielding a code efficiency of 92.9%.

[0045] The standard encoding rules are shown in the table V.

[0046] Runlength violation problems could arise upon free concatenation of some of the 3-bit input words upon conversion by use of the standard table. Those combinations of 3-bit input words are given in the table VI and the runlength violations are removed by the combinations of 5-bit channel words given in this table VI.

[0047] It can be summarized that the advantages of the R=3/5 iso-I2(d=2,k=10) code are:

[0048] rate between that of EFMPlus (d=2) and (d=1) codes.

[0049] DC-control via parity preserve principle.

[0050] relatively simple encoding tables.

[0051] A point of attention is that

[0052] standard mapping (3-to-5) and substitutional mapping (6-to-10) are not compatible with byte-oriented error correction. In order to avoid extra error propagation, an error correction code may be devised on formats using 9 bit or 12 bit symbols.

[0053] FIG. 4 shows schematically an apparatus for generating one of the (d,k) sequences described above. The apparatus is identified by the encoding unit 40, which has an input 42 for receiving the sequence of input words. The encoding unit 40 converts the sequence of input words into the (d,k) sequence with the properties described above, and supplies the (d,k) sequence to its output 46. The output 46 is coupled to an input 48 of an aT precoder 50, more preferably a 1T precoder, well known in the art. The 1T precoding carried out on the (d,k) sequence, results in a RLL signal of the type (d,k), which is supplied to an output 52. The output 52 is coupled to an input 54 of a writing unit 56 for writing the RLL signal on a record carrier 60, such as a magnetic record carrier eg. in the form of a tape. The writing unit 56 includes at least one write head 58 for realizing this. Instead of a magnetic record carrier, the RLL signal could have been recorded on an optical record carrier, such as an optical disk 60a.

[0054] Whilst the invention has been described with reference to preferred embodiments thereof, it is to be understood that these are not limitative examples. Thus, various modifications may become apparent to those skilled in the art, without departing from the scope of the invention, as defined by the claims. As an example. The invention has been explained above with reference to embodiments where binary signals have been encoded and decoded. It will be clear that the invention is equally well applicable on m-ary valued signals, where m is larger than 2.

[0055] Further, the invention lies in each and every novel feature and combination of features.

[0056] (D1) U.S. Pat. No. 5,477,222 (PHN 14.448)

[0057] (D2) K. A. Schouharner Imnnink, Coding Techniques for Digital Recorders, Prentice Hall, 1991.

[0058] (D3) P. Funk, Run-Length-Lirnited Codes with Multiple Spacing, IEEE Transactions on Magnetics, Vol. MAG-18, 1982, pp. 772-775.

[0059] (D4) U.S. Pat. No. 5,696,505 (PHN 14.746)