Device for encoding/decoding n-bit source words into corresponding m-bit channel words, and vice versa

[0002] The invention relates to a device for encoding a stream of databits of a binary source signal into a stream of databits of a binary channel signal, the device comprising

[0003] input means for receiving the binary source signal,

[0004] merging means for merging a q-bit merging word at equidistant positions in the binary source signal in response to a control signal, so as to obtain a composite binary source signal,

[0005] converting means for converting the composite binary source signal into said binary channel signal,

[0006] control signal generator means for generating said control signal,

[0007] output means for supplying said binary channel signal. The invention also relates to a device for decoding a stream of data bits of a binary channel signal obtained by means of the encoding device, so as to obtain a stream of databits of a binary source signal, to a recording device comprising the encoding device, to a record carrier obtained with the recording device and to a method of encoding the source signal.

[0008] An encoding device mentioned in the foregoing is known U.S. Pat. No. 5,477,222 (PHN 14448). The document discloses a device for encoding a stream of databits of a binary source signal into a stream of databits of a binary channel signal, satsifying a (1,8) runlength constraint. That means that, in a serial datastream of the channel signal at mimimum one ‘zero’ and at maximum eight ‘zeroes’ are present between two consecutive ‘ones’ in the channel signal. It should be noted in this respect that normally an additional precoding step, such as a 1T precoding, is applied to the (1,8) constrained sequence, resulting in a runlength limited sequence with minimum runlength 2 and maximum runlength 9.

[0009] The known conversion is parity preserving. ‘Parity preserving’ means that the parity of the n-bit source words to be converted equal the parity (after modulo-2 addition) of the corresponding m-bit channel words in which they are converted. As a result, the n-to-m conversion device as claimed does not influence the polarity of the signal.

[0010] As the conversion is parity preserving, an efficient DC control can be applied, such as by inserting DC control bits in the datastream of the source words.

[0011] The invention has for its object to provide an improved device for encoding n-bit source words into corresponding m-bit channel words.

[0012] The device in accordance with the invention comprises . . . copy claim 1 in its entirely]. The invention is based on the following recognition. Without applying the measure in accordance with the invention, the encoding process is as follows. After insertion of the q-bit merging words in the bitstream of the binary source signal, a bitstream of the composite binary source signal has been generated. This composite binary source signal is divided into a sequence of subsequent n-bit words for conversion into m-bit channel words in the converting means. This division into a sequence of n-bit words is done such that one boundary between two subsequent words in the composite binary source signal may lie between a merging word and the subsequent source word of the binary source signal in said composite binary source signal. Until the next merging word, all boundaries between two subsequent words in the composite binary source signal are aligned with the boundaries between the subsequent source words of the binary source signal. The said next merging word forms, together with the first q bits of the next source word in the binary source signal, the next n-bit word in the composite binary source signal. As a result, until the following merging word, all boundaries between two subsequent words in the composite source signal are not aligned anymore with the boundaries of the n-bit source words of the binary source signal. Or, an n-bit word in the composite binary source signal is spread out across two subsequent n-bit source words of the binary source signal.

[0013] The n-bit words in the composite binary source signal are converted upon encoding into corresponding m-bit channel words. After transmission and subsequent reception in a receiver provided with a corresponding decoder, the m-bit channel words are decoded into corresponding n-bit words, which n-bit words form a replica of the composite binary source signal. Subsequently, the q-bit merging words are deleted from the composite binary source signal so as to obtain a replica of the binary source signal.

[0014] Errors may occur during transmission, resulting in an erroneous decoding in the decoder. An erroneous word in the replica of the composite binary source signal may lead to one erroneous source word in the replica of the binary source signal, namely for those portions of the replica of the binary source signal where the boundaries between two subsequent source words were aligned with the boundaries of two subsequent words of the replica of the composite binary source signal. However, in those portions of the replica of the binary source signal where the boundaries between two subsequent source words are not aligned with the boundaries of two subsequent words of the replica of the composite binary source signal, an erroneous word in the replica of the composite source signal leads to two erroneous words in the replica of the binary source signal.

[0015] The invention overcomes this error propagation, in that by shifting q bits in each second group of y consecutive bits source words in the binary source signal, the boundaries between two subsequent n-bit source words of the binary source signal are aligned with the boundaries between two subsequent n-bit words in the composite binary source signal. As a result, no error propagation will occur.

[0016] Generally, the channel signal obtained is applied to a 1T-precoder. The purpose of the merging means is to add merging words, such as a 1-bit merging word, that is: a ‘0’- or a ‘1’ bit, to the consecutive code words included in the input signal of the converting means, so as to obtain a precoder output signal which is DC free, or includes a tracking pilot signal having a certain frequency. The precoder output signal is recorded on a record carrier.

[0017] In the example of q=1, the adding of a ‘0’-bit in the input signal of the converter results in the polarity of the output signal of the 1T precoder remaining the same. The adding of a ‘1’-bit results in a polarity inversion in the output signal of the 1T precoder. The merging means therefore influences the output signal of the 1T precoder such that the running digital sum value of the output signal of the 1T precoder can be controlled so as to have a desired pattern as a function of time.

[0018] The invention will be further described in the following figure description, in which

[0019] FIG. 1 shows a first and

[0020] FIG. 2 shows a second embodiment of the encoding device,

[0021] FIG. 3 shows various signals occurring in a prior art encoding device,

[0022] FIG. 4 shows various signals occurring in the encoding device of FIG. 2, with one way of shuffling the composite source signal,

[0023] FIG. 5 shows another way of shuffling the composite source signal,

[0024] FIG. 6 a first embodiment of the shuffling unit in the encoding device,

[0025] FIG. 7 shows a second embodiment of the shuffling unit,

[0026] FIG. 8 shows a first embodiment of the conversion unit in the encoding device,

[0027] FIG. 9 shows a second embodiment of the conversion unit,

[0028] FIG. 10 shows a first embodiment of the decoding device,

[0029] FIG. 11 shows a second embodiment of the decoding device,

[0030] FIG. 12 shows an embodiment of the reconverter unit in the decoding device, and

[0031] FIG. 13 shows again another way of shuffling the composite source signal.

[0032] FIG. 1 shows an embodiment of the encoding device in accordance with the invention. The device comprises an input terminal 1 for receiving a binary source signal, which terminal is coupled to an input 2 of a q-bit word merging unit 4. An output 6 of the unit 4 is coupled to an input 7 of a shuffling unit 8. The shuffling unit 8 has an output 9 which is coupled to an input 11 of a converter unit 10, which has an output 12 which is coupled to an input 14 of an aT precoder 16. An output 18 of the precoder 16 is coupled to an output for supplying a binary channel signal. The output 18 is also coupled to an input 22 of a control signal generator 24, which has an output 26 coupled to a control signal input 28 of the merging unit 4.

[0033] A binary source signal, in the form of a sequence of n-bit source words is applied to the input terminal 1 for encoding. The merging unit 4 is adapted to insert a q-bit merging word at equidistant positions in the binary source signal. More specifically, the merging unit 4 is adapted to insert a q-bit merging word each time between subsequent groups of x consecutive n-bit source words in the binary source signal so as to obtain a composite binary source signal at its output 6. x, n and q are integers, for which holds: x>1, n=2q and q>1.

[0034] The composite binary source signal is applied to the input 7 of the shuffling unit 8, which carries out a shuffling step on each second group of x consecutive n-bit source words in the composite binary source signal, so as to obtain a shuffled composite source signal. The shuffled composite source signal is subsequently supplied to the input 11 of the converter unit 10. The encoding unit 10 carries out a conversion step on the shuffled composite source signal, which results into a channel signal. The conversion step includes the conversion of groups of p n-bit words in the shuffled composite source signal into corresponding groups of p m-bit channel words, said channel words forming the channel signal. This channel signal is supplied to the output 12. The binary channel signal is subsequently precoded into a precoded binary channel signal in the aT precoder 16 and the precoded binary channel signal is subsequently supplied to the output terminal 20, eg. for recording or transmission, as well as to the input 22 of the control signal generator 24. The control signal generator 24 generates a control signal at its output 26 in response to said precoded binary channel signal, which control signal is supplied to the control signal input 28 of the merging unit 4. The merging unit 4 chooses a q-bit word in response to said control signal for merging into the binary source signal.

[0035] The functioning of the device will be further explained with reference to FIG. 2, which shows a slightly modified version of the device of FIG. 1. The device of FIG. 2 comprises a merging unit 4′ which is adapted to insert a 1-bit (q=1) merging word each time between subsequent groups of x consecutive n(=2)-bit source words in the binary source signal so as to obtain a composite binary source signal at its output 6. x could have any integer value. In an example to be described later, x will be equal to 24, for the reason that a 1-bit merging word is inserted between subsequent groups of 6 bytes of the binary source signal.

[0036] The shuffling unit 8′ in the embodiment of FIG. 2 will be described later. First, the functioning of the converter unit 10′ will be explained. The converter 10′ is adapted to convert groups of p 2-bit words in the shuffled composite source signal into corresponding groups of p 3-bit channel words of the channel signal. p is an integer and can vary, as will be made clear further below. The channel signal is precoded in a 1T precoder 16′, well known in the art, so as to obtain the precoded channel signal. The output signal of the 1T-precoder 16′ is applied to the control signal generator 24′ which generates the control signal for the merging unit 4′, so as to control whether a ‘0’ or a ‘1’ is inserted in the serial datastream of the source signal.

[0037] By means of the device shown in FIGS. 1 or 2, it is possible to embed a tracking tone of a certain frequency in the serial datastream of the precoded channel signal, or keep the DC content of the said datastream to zero. Further, when the converter unit 10′ is adapted to generate a (d,k) sequence as explained above, it causes the output signal of the device of FIGS. 1 or 2 to be a (d,k) RLL output signal. Embodiments of the merging unit 4 are given in Bell System Technical Journal, Vol 53, No. 6, pp. 1103-1106.

[0038] The output signal of the 1T precoder 16′ is supplied to a write unit 34 for writing the signal in a track on a record carrier 36. The record carrier 36 can be a magnetic record carrier in longitudinal or disk form. The record carrier could also be an optical record carrier, such as an optical disk 36′. The write unit 34 comprises a write head 38, which is a magnetic write head, when recording the signal on a magnetic record carrier or an optical write head, when recording the signal on an optical record carrier.

[0039] The functioning of the device of FIG. 2 will be explained hereafter. FIG. 3 shows in FIG. 3aschematically the binary source signal applied to the input terminal 1 of the device. The source signal comprises a sequence of 2-bit source words, denoted sw₁, sw₂, . . . sw_i, sw_i+1, . . . Each time, four 2-bit source words of the binary source signal form a byte in the binary source signal. Sequence of each time six consecutive bytes B₀, . . . B₅are shown in FIG. 3a. The merging unit 4′ of FIG. 2 is adapted to insert a 1-bit merging word each time between those groups of six consecutive bytes B₀to B₅in the binary source signal, that is: a 1-bit merging word is each time inserted between subsequent groups of 24 2-bit source words of the binary source signal. In the present example, a merging word mw₀is merged between the source words sw₂and sw₃of the binary source signal and a merging word mw₁is merged between the source words sw_i+3and sw_i+4, resulting into the composite source signal shown in FIG. 3b.

[0040] In the prior art device, no shuffling of the composite source signal is realized and the composite source signal is transferred to the converter unit 10′ without modification. The converter unit 10′ converts 2-bit words in the composite source signal into corresponding 3-bit words. This is shown in FIG. 3cand FIG. 3d. FIG. 3cshows the composite source signal divided into a sequence of 2-bit words, the so-called composite source words, denoted csw₁, csw₂, csw₃. . . Those composite source words are converted into corresponding 3-bit channel words, resulting in the channel signal shown schematically in FIG. 3d. The channel words in the channel signal are denoted cw₁. FIG. 3dshows the channel words cw₂, cw₇, cw₁₁and cw_i, obtained from converting the 2-bit composite source words csw₂, csw₇, csw₁₁and csw_i, respectively.

[0041] Upon reconversion in a corresponding decoding device, the 3-bit channel words are reconverted into replicas of the original 2-bit composite source words. This is schematically shown in FIG. 3e, which shows the sequence of the replicas of the composite source words, denoted rcsw₁, rcsw₂, rcsw₃, . . . rcsw_i. Next, the merging words mw₀and mw₁, included in the replicas rcsw₃and rcsw_i+4are deleted from those replicas, resulting in the replica of the original source signal, shown schematically in FIG. 3f. FIG. 3fshows the replica of the source signal, in the form of a sequence of replicas of the original source words, those replicas being denoted in FIG. 3fby rsw₁, rsw₂, rsw₃. . . , rsw_i, rsw_i+1, . . .

[0042] From the FIGS. 3band 3cit is clear that, because of the insertion of the 1-bit merging words between the 2-bit source words in FIG. 3a, the boundaries between the 2-bit source words in the composite source signal of FIG. 3bdo not always coincide with the boundaries between the 2-bit composite source words in FIG. 3c. More specifically, this statement is valid for the source words sw₃to SW_i+3in the signal of FIG. 3b.

[0043] Suppose now, that upon conversion and subsequent reconversion of the composite source word csw₇, an error occurs, resulting in an erroneous replica rcsw₇. The first bit of the word rcsw₇is the seventh bit b₇of the byte B₀and the second bit in the word rcsw₇is the first bit b₀of the byte B₁in the replica of the source signal of FIG. 3f. As the word rcsw₇is erroneous, so are the bytes B₀and B₁. Thus error propagation occurs, as an error in one word of the replica of the composite source signal results in an error in two bytes of the replica of the source signal.

[0044] In accordance with the invention, a shuffling step is carried out on the composite source signal, prior to supplying the composite source signal to the converter unit 10′. FIG. 4 shows one embodiment of this shuffling step and FIG. 5 shows another embodiment. FIG. 4ashows again the serial datastream of the binary source signal, in which a 1-bit merging word is inserted each time between subsequent groups of 24 2-bit source words. FIG. 4bshows again how a merging word mw₀is inserted after the source word sw₂and a merging word mw₁is inserted before the source word sw_i+4. Further, a shuffling step is carried out on the signal obtained after merging the merging words. The result of the merging and shuffling step is shown by the signal of FIG. 4b, which shows the shuffled composite signal in the form of 2-bit words w₁, w₂, . . . w_i, w_i+1, . . . In addition to inserting the merging word mw₀, the sequence of 2-bit source words following the source word sw₂is shifted to the right in FIG. 4bover an additonal bit position, resulting in an empty bit position between the merging word mw₀and the word w₄of the shuffled signal. Thus, the word w₄in the shuffled signal equals the source word sw₃, the word w₅in the shuffled signal equals the source word sw₄, and so on. Thus, the word w_i+3which equals the source word sw_i+2. The last two bits of the byte B₅, the bits b₆and b₇, are included in the source word sw_i+3. The bit b₆directly follows the word w_i+3, as can be seen in FIG. 4b. The bit b₇, however, is repositioned and placed into the empty bit position between the merging word mw₀and the word w₄. This is indicated in FIG. 4bby the arrow 50. The next merging word mw₁is inserted as the second bit of the 2-bit word w_i+4, which is followed by the 2-bit word sw_i+5which equals the source word sw_i+4, and so on. It is clear from FIGS. 4aand 4bthat the boundaries between the words in the shuffled composite signal of FIG. 4bnow coincide with the boundaries of the source words in the source signal of FIG. 4a.

[0045] Converting the 2-bit words of the shuffled composite signal into 3-bit channel words leads to the channel signal of FIG. 4c. Subsequent reconversion leads to a replica of the shuffled composite signal which is shown in FIG. 4d. Next, the merging words are deleted from the replica of the shuffled composite signal. That means: the first bit of the replica of the word rw₃is deleted and the second bit of the word rw_i+4is deleted. Further, the second bit of the word rw₃is retrieved from the serial datastream, so that the word rsw₃that follows the word rsw₂of the replica of the source signal equals the 2-bit word rw₄of the replica of the shuffled composite signal, and so on. Thus, the word rsw_i+2of the replica of the source signal equals the word rw_i+3of the replica of the shuffled composite signal and the word rsw_i+4equals the word rw_i+5. The bit b₆of the byte B₅of the replica of the source signal equals the first bit of the word rw_i+4of the replica of the shuffled composite signal. Deleting the merging word mw₁means an empty bit position, which is filled in by the bit value of the second bit of the word rw₃, giving the bit b₇of the byte B₅. This repositioning of the bit b₇is shown in FIG. 4eby the arrow 52.

[0046] An error occurring during the conversion/reconversion step results in an erroneous word rw_iin the replica of the shuffled composite signal of FIG. 4d. As the boundaries between the words rw_iin the signal of FIG. 4dand the words rsw_iin the signal of FIG. 4eare aligned, such error results in only one 2-bit word, and thus only one byte, to be erroneous, so that no error propagation occurs.

[0047] In the example of FIG. 4, it was the bit b₇in the source word sw_i+3that was repositioned to the position between the merging word mw₀and the source word sw₃. Another possibility would have been to reposition the bit b₆in the source word sw_i+3into the position between the merging word mw₀and the source word sw₃.

[0048] FIG. 5 shows another way of shuffling the composite source signal. FIG. 5ashows again the serial datastream of the binary source signal. FIG. 5bshows again how a merging word mw₀is inserted after the source word sw₂and a merging word mw₁is inserted before the source word sw_i+4. The shuffling step now comprises repositioning the bit b₀of the byte B₀, that is: the first bit of the word sw₃. The result of the merging and shuffling step is shown by the signal of FIG. 5b, which shows the shuffled composite signal in the form of 2-bit words w₁, w₂, . . . w_i, w_i+1, . . . The insertion of the merging word mw₀in fact means a replacement of the bit b₀of the byte B₀by the merging word mw₀, whilst the bit b₀is temporarily stored. Thus the following 2-bit words of the source signal remain in their original position. Or: the word sw₄in the source signal equals the word w₄in the shuffled composite signal, and so on. Thus, the word sw_i+3of the source signal equals the word w_i+3of the shuffled composite signal. Next, the bit b₀, that was temporarily stored, is now added as the first bit of the word w_i+4and is followed by the merging word mw₁as the second bit of the word w_i+4. This is schematically shown by the arrow 54 in FIG. 5b. The following source words sw_i+4, sw_i+5, . . . are shifted over two bit positions. But in all cases, the boundaries between the 2-bit words w in the shuffled composite signal of FIG. 5band the source words sw in the source signal of FIG. 5aare aligned.

[0049] Converting the 2-bit words of the shuffled composite signal into 3-bit channel words leads to the channel signal of FIG. 5c. Subsequent reconversion leads to a replica of the shuffled composite signal which is shown in FIG. 5d. Next, the merging words are deleted from the replica of the shuffled composite signal. That means: the first bit of the replica of the word rw₃is deleted and the second bit of the word rw_i+4is deleted. The open bit position of the merging word mw₀is filled by the bit value of the first bit of the word rw_i+4, resulting in the first bit of the 2-bit word rsw₃, and thus resulting in bit b₀of the byte B₀in the replica of the source signal. Further, the word rsw₄that follows the word rsw₃of the replica of the source signal equals the 2-bit word rw₄of the replica of the shuffled composite signal, and so on. Thus, the word rsw_i+3of the replica of the source signal equals the word rw_i+3of the replica of the shuffled composite signal. The first bit of the word rw_i+4was already repositioned into the first bit position of the word rsw₃. As a result, deleting the merging word mw₁means that the words rw_i+5(and so on) directly follow the word rsw_i+3as the words rsw_i+4(and so on) of the replica of the source signal. The repositioning of the bit b₀is shown in FIG. 5eby the arrow 56.

[0050] An error occurring during the conversion/reconversion step results in an erroneous word rw_iin the replica of the shuffled composite signal of FIG. 5d. As the boundaries between the words rw_iin the signal of FIG. 5dand the words rsw_iin the signal of FIG. 5eare aligned, such error results in only one 2-bit word, and thus only one byte, to be erroneous, so that no error propagation occurs.

[0051] In the example of FIG. 5, it was the bit b₀in the source word sw₃that was repositioned to the position between the source word sw_i+3and the merging word mw₁. Another possibility would have been to reposition the bit b₁in the source word sw₃into the position between the source word sw_i+3and the merging word mw₁.

[0052] Further, it should be noted that the shuffling step has been described as a step that follows the merging step. It should however be noted that the shuffling step could precede the merging step, or could coincide with the merging step.

[0053] FIG. 6 shows an embodiment of the shuffling unit 8′ for carrying out the shuffling step described with reference to the FIGS. 4aand 4b. The shuffling unit of FIG. 6 comprises a memory in the form of a shift register 40 having, in the example given, 47 storage locations. Further first and second controllable switches 42 and 44 are available having terminals a, b and c. Further, a control unit CPU is available for controlling the positions of the switches 42 and 44 via a control line 46 and for controlling the shifting of the information through the shift register 40 via a control line 48. The input 7 of the shuffling unit is coupled to the a-terminal of the switch 42. The b-terminal of the said switch is coupled to an input of the shift register 40. The c-terminals of the switches 42 and 44 are interconnected. An output of the shift register 40 is coupled to the b-terminal of the switch 44 and the a-terminal of the switch 44 is coupled to the output 9 of the shuffling unit.

[0054] The functioning of the shuffling unit of FIG. 6 is as follows. Under the influence of a control signal on the line 46, the switches 42 and 44 are both in their position a-b and under the influence of a control signal on the line 48, bits are stored in the first memory location M₁of the shift register 40 and shifted further to the right in FIG. 6 in the shift register 40. It is assumed that the signal of FIG. 4a, with the merging bits included, is supplied to the input 7 of the shuffling unit. Storage of source words is carried out by storing the source word sw₁, and subsequent source words thereafter. Storing the information is continued until the source word sw₁is stored in the memory locations M₄₇and M₄₆, the source word sw₂is stored in the memory locations M₄₅and M₄₄and the merging word mw₀is stored in the memory location M₄₃. Five subsequent clock signals on the line 48 results in shifting the source words sw₁and sw₂and the merging word mw₀out of the shift register 40, so that they are applied to the output 9. The shift register now has the byte B₀, formed by the source words sw₃to sw₆, stored in the memory locations M₄₇to M₄₀. The byte B₁, formed by the source words sw₇to sw₁₀, is stored in the memory locations M₃₉to M₃₂, and so on. The bits b₀to b₆of the byte B₅are stored in the memory locations M₇to M₁. Further, the bit b₇of the byte B₅now available at the input 7. FIG. 6 shows the situation at this moment.

[0055] Under the influence of the control signal on the line 46, the switches 42 and 44 now switch over to their positions a-c, so that the bit b₇of the byte B₅supplied to the output 9 for further processing in the converter unit 10′. Next, the switches 42 and 44 are again positioned into their position a-b, so that the bit b₀, stored in the memory location M₄₇, and the subsequent bits stored in the shift register 40, can be supplied to the output 9. Further the source word sw_i+4and following source words are stored in the shift register 40. In this way, the bit b₇of the byte B₅actually inserted in the serial datastream of the shuffled composite signal between the merging word mw₀and the bit b₀. This is done for each second group of 24 subsequent source words.

[0056] FIG. 7 shows an embodiment of the shuffling unit 8′ for carrying out the shuffling step described with reference to the FIGS. 5aand 5b. The shuffling unit of FIG. 7 comprises a memory M having one storage location. Further first and second controllable switches 42 and 44 are available having terminals a, b and c. Further, a control unit CPU is available for controlling the positions of the switches 42 and 44 via a control line 46 and for controlling the storage of information in the memory M via a control line 48. The input 7 of the shuffling unit is coupled to the a-terminal of the switch 42. The b-terminal of the said switch is coupled to an input of the memory M. The c-terminals of the switches 42 and 44 are interconnected. An output of the memory M is coupled to the b-terminal of the switch 44 and the a-terminal of the switch 44 is coupled to the output 9 of the shuffling unit.

[0057] The functioning of the shuffling unit of FIG. 7 is as follows. Under the influence of a control signal on the line 46, the switches 42 and 44 are both in their position a-c, so that the serial datastream of the source signal supplied to the input 7 is directly fed to the output 9. In this way, the source words sw₁and sw₂and the merging word mw₀in the composite source signal are supplied to the output 9. Under the influence of the control signal on the line 46, the switch 42 is switched into its position a-b, at a time instant that the bit b₀in the source word sw₃appears at the input 7. Further, upon the influence of the control signal on the line 48, this bit b₀is stored in the memory M. Next, uypon the appearance of the bit b₁of the source word sw₃, the switch 42 is again switched into of the position a-c, so that the bits b₁, b₂, . . . and so on of the byte B₀can be supplied to the output 9. This continues until the source word sw_i3appears at the input 7 and is supplied to the output 9. Next, under the influence of a control signal on the line 46, the switch 44 is switched into its position a-b, so that the bit b₀of the source word sw₀is supplied to the output. Next, under the influence of the control signal on the line 46, the switch 44 now switches over to its position a-c, so that the merging word mw₁and subsequent source words sw_i+5, sw_i+6, . . . and so on, appearing at the input 7, can be supplied to the output 9. In this way, the bit b₀in the source word sw₃in byte B₀is actually inserted in the serial datastream of the shuffled composite signal between the source word sw_i+3and the merging word mw₁. This is done for each second group of 24 subsequent source words.

[0058] FIG. 8 shows a converter 10′ that is capable of converting 2-bit words of the shuffled composite source signal into 3-bit channel words. The device described here is in fact the device also disclosed in U.S. Pat. No. 5,477,222, with further modifications to realize the objectives of a minimization of the repeated minimum transition runlength.

[0059] The converter 10′ has an input 60, for receiving the stream of databits of the shuffled composite binary source signal S. The input 60 is coupled to an input of a shift register 62 having eight cells X₁to X₈, in the present example, so as to receive eight consecutive bits of the shuffled composite source signal S. The shift register 62 functions as a serial-parallel convener. The outputs of the cells are coupled to corresponding inputs i₁to i₈, respectively, of a logic circuit LC, for supplying the logic values (x₁, . . . , x₈) of the bits present in the cells.

[0060] The converter 10′ further includes a second shift register 64 having twelve cells Y₁to Y₁₂. The logic circuit LC has twelve outputs o₁to o₁₂. Those outputs of the logic circuit LC are coupled to corresponding inputs of the twelve cells Y₁to Y₁₂, respectively, of the shift register 64. An output 66 of the shift register 64 is coupled to an output 68. The shift register 64 functions as a parallel-serial converter, so as to obtain the binary channel signal C.

[0061] Further, a detector unit 70 is available for detecting specific sequences in the serial datastream of the source signal S. To that purpose, the outputs of the eight cells X₁to X₈of the shift register 62 are coupled to corresponding inputs, denoted 72, of the detector unit 70. In the present embodiment, the detector unit 70 has three outputs, denoted o₁, o₂and o₃, for generating a first, second and third control signal, respectively. Those outputs are coupled to corresponding control signal inputs c₁, c₂and c₃, respectively, of the logic circuit LC.

[0062] The functioning of the logic circuit LC in response to the control signals applied to its inputs c₁, c₂and c₃, is as follows.

[0063] The logic circuit LC is capable of converting 2-bit source words w_iof the shuffled composite source signal into 3-bit channel words, such that the conversion for each 2-bit source word is parity preserving. That means that the number of ‘ones’ in the 2-bit word to be converted equals the number of ‘ones’ in the corresponding 3-bit channel word, a modulo-2 addition on the ‘ones’ in the channel word being carried out. Or, otherwise said: if the number of ‘ones’ in the 2-bit word is even, the number of ‘ones’ in the 3-bit channel word will be even. And: if the number of ‘ones’ in the 2-bit word is odd, the number of ‘ones’ in the 3-bit channel word will be odd.

[0064] As an example, the converting means LC is adapted to convert the 2-bit words w_iinto 3-bit channel words cw_iin accordance with the following table:

[0065] It should be noted here, that the first bit in the 2-bit word w_iis applied first to the shift register 62 and that the first bit in the channel word is supplied first from the output 66 of the shift register 64.

[0066] It should be further noted here, that the logic circuit LC converts 2-bit words w_istored in the cells X₁,X₂into 3-bit channel words cw_iand stores those channel words in the cells Y₁,Y₂,Y₃, of the shift register 64, in response to an absence of any control signal at the control signal inputs c₁,c₂and c₃. Each conversion in this way, is followed by a shift over two positions to the left in the shift register 62, and a shift over three positions to the left in the shift register 64. The shift over two positions in the shift register 62 is required so as to make the shift register 62, and thus the converter, ready for a subsequent conversion. The shift over three positions in the shift register 64 is required in order to output the 3-bit channel word cw_igenerated.

[0067] The converter 10′ of FIG. 8 can be used to generate a channel signal C in the form of a (d,k) sequence satisfying the d=1 constraint. That means that at least one ‘zero’ is present between two subsequent ‘ones’ in the serial datastream of the channel signal. That is, that a concatenation of two or more ‘ones’ in the channel signal is prohibited.

[0068] It might occur that the unmodified conversion, such as by means of the converter of FIG. 8, of combinations of two subsequent 2-bit words w_imight violate the d=1 constraint. Those combinations are the combinations; ‘00 00’, which by unmodified conversion would lead to the two 3-bit channel words ‘101 101’; ‘00 01’, which by unmodified conversion would lead to the two 3-bit channel words ‘101 100’; ‘10 00’, which by unmodified conversion would lead to the two 3-bit channel words ‘001 101’and ‘10 01’, which by unmodified conversion would lead to the two 3-bit channel words ‘001 100’.

[0069] The occurrence of such combinations should be detected so that a modified encoding of blocks of two 2-bit words w_i,w_i+1into blocks of two 3-bit channel words cw_i,cw_i+1, can take place. Therefore, the converter of FIG. 8 is, in addition to the ‘normal’ encoding of 2-bit words w_iinto 3-bit channel words cw_i, capable of detecting the above identified combinations, and is capable of realizing a modified encoding, such that the d=1 constraint in the channel signal is still satisfied.

[0070] Because of the fact that the outputs of the cells X₁to X₄of the shift register 62 are coupled to corresponding inputs of the detector unit 70, this detector unit 70 is capable of detecting the position in the serial bitstream of the shuffled composite source signal where unmodified encoding of single 2-bit words w_iin the bitstream into corresponding single 3-bit channel words cw_iwould lead to a violation of the d=1 constraint in the channel signal C, and are adapted to supply a control signal at its output o₁in response to such detection.

[0071] More specifically, the detector unit 70 detects whether the cells X₁to X₄comprise one of the 4-bit sequences that are given in table II, and generates a first control signal at its output o₁. As soon as the detector circuit 70 detects a combination of two 2-bit words w_i,w_i+1present in the four cell positions x₁,x₂,x₃,x₄, which combination equals one of the combinations given in the left hand column of table II, the logic circuit LC converts the combination in accordance with the modified coding as given in table II:

[0072] As can be seen from the table, unmodified conversion of the single two 2-bit words leads to a violation of the d=1 constraint, as two ‘ones’ occur at the boundary between the two channel words obtained. The logic circuit LC is therefore adapted to convert in a modified coding mode, the blocks of two 2-bit words given in the left column of the above table into the blocks of two 3-bit channel words as given in the right column in the above table II. As can be seen, no violation of the d=1 constraint occurs anymore. Moreover, the modified encoding in the same way is parity preserving. Further, one of the two 2-bit words, which is in the above table the second one, is encoded into a 3-bit channel word which is unequal to one of the four channel words of table I. The reason for this is that on the receiver side, a detection of this 3-bit channel word not belonging to the set of four 3-bit channel words of the table I is possible, so that a corresponding decoding, which is the inverse of the encoding as defined with reference to table II, can be realized.

[0073] The block of two 3-bit channel words obtained by means of the encoding in conformity with table II, is supplied by the logic circuit LC to its outputs o₁to o₆, which channel words are supplied to the six cells Y₁to Y₆of the shift register 64.

[0074] It will further be clear that, a conversion of two 2-bit words into two 3-bit channel words by the converter unit LC, is followed by a shift over four positions to the left in the shift register 62 and a shift over six positions to the left in the shift register 64. The shift over four positions in the shift register 62 is required so as to make the shift register 62, and thus the converter, ready for a subsequent conversion. The shift over six positions in the shift register 64 is required in order to output the two 3-bit channel words generated.

[0075] The k-constraint in a (d,k) sequence means that a concatenation of at most k ‘zeroes’ between two subsequent ‘ones’ in the channel signal are allowed.

[0076] It might occur that the unmodified conversion of three subsequent 2-bit words might violate the k-constraint.

[0077] As an example: the sequence of 2-bit words ‘11 11 11’ in the shuffled composite source signal would, by unmofidied conversion, lead to the three 3-bit channel words ‘000 000 000’. If a (d,k) sequence should be obtained where k equals 6, 7 or 8, such combination of three 3-bit channel words should not occur.

[0078] Another example is the sequence of 2-bit words ‘11 11 10’ which by unmodified conversion would lead to the three 3-bit channel words ‘000 000 001’. This combination of three 3-bit channel words does not satisfy a k=6 or k=7 constraint. Moreover, this combination of three 3-bit channel words can follow a previous channel word that ends with a ‘0’, so that it might lead to a violation of a k=8 constraint. Further, the combination ends with a ‘1’, so that it might lead to a violation of the d=1 constraint, if the combination is followed by a 3-bit channel word that starts with a ‘1’. An equivalent reasoning is valid for the sequence of 2-bit words ‘01 11 11’.

[0079] A further example is the sequence of 2-bit words ‘01 11 10’ which by unmodified conversion would lead to the three 3-bit channel words ‘100 000 001’. This combination can, in the same way as given above, lead to a violation of the d=1 constraint.

[0080] The occurrence of such combinations should be detected so that a modified encoding can take place. Therefore, the converter of FIG. 8 is, in addition to the ‘normal’ encoding of 2-bit words into 3-bit channel words, as well as the modified encoding in accordance with table II, capable of detecting the above identified combinations, and is capable of realizing a modified encoding, such that the k-constraint in the channel signal is still satisfied.

[0081] Because of the fact that the outputs of the cells X₁to X₆of the shift register 62 are coupled to corresponding inputs of the detector unit 70, this detector unit 70 is capable of detecting the position in the serial bitstream of the shuffled composite source signal where unmodified encoding would lead to a violation of the k-constraint in the channel signal C, and are adapted to supply a control signal at its output o₂in response to such detection.

[0082] More specifically, the detector unit 70 detects whether the cells X₁to X₆comprise one of the 6-bit sequences that are given in table III, and generates a second control signal at its output o₂.

[0083] As soon as the detector circuit 70 detects a combination of three 2-bit words present in the six cell positions x₁,x₂,x₃,x₄,x₅,x₆, which combination equals one of the combinations given in the left hand column of table III, the logic circuit LC converts the combination in accordance with the modified coding as given in table III:

[0084] The logic circuit LC converts in the second modification of the coding mode, the blocks of three 2-bit words given in the left column of the above table III into the blocks of three 3-bit channel words as given in the right column in the above table. By realizing the modified encoding as per table III, a channel signal has been obtained which satisfies the k=8 constraint. Moreover, the modified encoding in the same way is parity preserving. This means in the present situation that, if the number of ‘ones’ in the combination of three 2-bit words is odd (even), the number of ‘ones’ in the combination of the three 3-bit channel words obtained is odd (even). Further, two of the three 2-bit words, which is in the above table the second one and the third one, is encoded into a 3-bit channel word which is unequal to one of the four channel words of table I. The reason for this is that on the receiver side, a detection of these two consecutive 3-bit channel words not belonging to the set of four 3-bit channel words of the table I is possible, so that a corresponding decoding, which is the inverse of the encoding as defined with reference to table III, can be realized.

[0085] The block of three 3-bit channel words obtained by means of the encoding in conformity with table III, is supplied by the logic circuit LC to its outputs o₁to o₉, which channel words are supplied to the nine cells Y₁to Y₉of the shift register 64.

[0086] It will further be clear that, a conversion of three 2-bit words into three 3-bit channel words by the converter unit LC, is followed by a shift over six positions to the left in the shift register 62 and a shift over nine positions to the left in the shift register 64. The shift over six positions in the shift register 62 is required so as to make the shift register 62, and thus the converter, ready for a subsequent conversion. The shift over nine positions in the shift register 64 is required in order to output the three 3-bit channel words generated.

[0087] A further requirement for encoding the shuffled composite source signal is that the repeated minimum transition runlength in the channel signal should be limited. The repeated minimum transition runlength is defined as the length of the sequence of subsequent transitions between ‘0’ and ‘1’, or: a sequence‘. . . 01010101010 . . . ’, in the case where the d-constraint equals 1. As an example, the bit sequence ‘00 01 00 01’ results, after modified conversion using table II, in the bit sequence ‘101 010 101 010’. In the same way, the bit sequence ‘10 01 00 01’ results, after modified conversion using table II, in the bit sequence ‘001 010 101 010’. Such sequences deteriorate the bit detection in a receiver. Restriction of the length of 01-sequences is thus required.

[0088] Because of the fact that the outputs of the cells X₁to X₈of the shift register 62 are coupled to corresponding inputs of the detector unit 70, this detector unit 70 is capable of detecting the position in the serial bitstream of the shuffled composite signal where unmodified encoding would lead to a violation of the requirement that repeated minimum transition runlenght is limited and is adapted to supply a control signal at its output o₃in response to such detection.

[0089] More specifically, the detector unit 70 detects whether the cells X₁to X₈comprise one of the 8-bit sequences that are given in table IV, and generates a third control signal at its output o₃.

[0090] As soon as the detector circuit 70 detects a combination of four 2-bit source words present in the eight cell positions x₁,x₂,x₃,x₄,x₅,x₆,x₇,x₈, which combination equals one of the combinations given in the left hand column of table IV, the logic circuit LC converts the combination in accordance with the modified coding as given in table IV into a 12-bit bitsequence as given in the right column of this table IV.

[0091] The modified conversion as per table IV is again aparity preserving.

[0092] It should be noted that the bitstream of the channel words is in NRZI (non-return to zero-inverse) notation, which means that a ‘one’ results in a transition in the write current for recording the channel signal on a magnetic record carrier.

[0093] In the foregoing, the situations where a modified encoding is required are detected by the detector unit 70 from the 2-bit words in ther shuffled composite source signal. It should however be noted that detection could be carried out on the channel words generated. Reference is made in this respect to FIG. 2bin U.S. Pat. No. 5,477,222.

[0094] FIG. 9 shows another embodiment of the converter in which detection of the situations where a modified encoding is required is done based on the channel words generated by an unmodified coding as per table I.

[0095] The converter of FIG. 9 includes a detector 70′ having 12 inputs for receiving the four subsequent 3-bit channel words obtained by means of an unmodified encoding as per table I in the circuit LC′. The detector 70′ detects whether the two subsequent 3-bit channel words at the outputs o₁to o₆of the circuit LC′, obtained using the unmodified coding, equal one of the four 6-bit sequences given in the middle column under ‘unmodified coding’ of table II. If so, the detector 70′ issues a switching signal at its output 72 and an address signal AD at its output 72′. The switching signal is applied to a switching signal input 65 of the shift register 64′. The address signal AD is applied to an address signal input 76 of a ROM 77. The detector 70′ generates one of four possible address signals AD1 to AD4, in response to the detection of a corresponding one of the four 6-bit sequences in the middle column of table II. As an example, the address signal AD1 is generated when the detector 70′ detects the sequence ‘101101’ and generates the address signal AD4 upon detection of the 6-bit sequence ‘001100’. The ROM 77 has the 6-bit sequences shown in the right column of table II stored. Upon the receipt of the address signal AD1, the ROM supplies the 6-bit sequence ‘100 010’ at its outputs o₁to o₆, and upon the receipt of the address signal AD2, the ROM supplies the 6-bit sequence ‘101 010’ at those outputs. Upon the receipt of the address signal AD3, the ROM supplies the 6-bit sequence ‘000 010’ at those outputs, and upon the receipt of the address signal AD4, the ROM supplies the 6-bit sequence ‘001 010’ at those outputs. Each memory location of the shift register 64′ has now two inputs, one of them being coupled with a corresponding output of the logic circuit LC′, the other being coupled to a corresponding output of the ROM 77. In response to the switching signal applied to the input 75, the shift register accepts the information supplied to its lower inputs and shifts its contents over six positions to the left. As a result, a modified 6-bit sequence is supplied by the shift register 64′ to the output 68.

[0096] The detector 70′ also detects whether the three subsequent 3-bit channel words at the outputs o₁to o₉of the circuit LC′, obtained using the unmodified coding, equal one of the four 9-bit sequences given in the middle column under ‘unmodified coding’ of table III. If so, the detector 70′ issues a switching signal at its output 72 and an address signal AD at its output 72′. The detector 70′ generates one of four possible address signals AD5 to AD8, in response to the detection of a corresponding one of the four 9-bit sequences in the middle column of table III. As an example, the address signal AD5 is generated when the detector 70′ detects the sequence ‘000 000 000’ and generates the address signal AD8 upon detection of the 9-bit sequence ‘100 000 000’. The ROM 77 has the 9-bit sequences shown in the right column of table III stored. Upon the receipt of the address signal AD5, the ROM supplies the 9-bit sequence ‘000 010 010’ at its outputs o₁to o₉, and upon the receipt of the address signal AD6, the ROM supplies the 9-bit sequence ‘001 010 010’ at those outputs. Upon the receipt of the address signal AD7, the ROM supplies the 9-bit sequence ‘101 010 010’ at those outputs, and upon the receipt of the address signal AD8, the ROM supplies the 9-bit sequence ‘100 010 010’ at those outputs.

[0097] In response to the switching signal applied to the input 75, the shift register accepts the information supplied to its lower inputs and shifts its contents over nine positions to the left. As a result, a modified 9-bit sequence is supplied by the shift register 64′ to the output 68.

[0098] The detector 70′ further detects whether the four subsequent 3-bit channel words at the outputs o₁to o₁₂of the circuit LC′, obtained using the unmodified coding, equal one of the following two 12-bit sequences: ‘101 010 101 010’ or ‘001 010 101 010’. If so, the detector 70′ issues a switching signal at its output 72 and an address signal AD at its output 72′. The detector 70′ generates one of two possible address signals AD9 and AD10, respectively, in response to the detection of a corresponding one of the two 12-bit sequences given above. As an example, the address signal AD9 is generated when the detector 70′ detects the sequence ‘101 010 101 010’ and generates the address signal AD10 upon detection of the 12-bit sequence ‘001 010 010 101’. The ROM 77 has the 12-bit sequences shown in the right column of table IV stored. Upon the receipt of the address signal AD9, the ROM supplies the 12-bit sequence ‘100 010 010 010’ at its outputs o₁to o₁₂, and upon the receipt of the address signal AD10, the ROM supplies the 12-bit sequence ‘000 010 010 010’ at those outputs.

[0099] In response to the switching signal applied to the input 75, the shift register accepts the information supplied to its lower inputs and shifts its contents over twelve positions to the left. As a result, a modified 12-bit sequence is supplied by the shift register 64′ to the output 68.

[0100] In the normal situation, when none of the constraints is violated, unmodified conversion is carried out in accordance with table 1, and the switching signal is absent so that the shift register accepts the bits supplied by the logic circuit LC′via the upper inputs of the shift register 64′.

[0101] It has been said previously that other conversion rules for converting single 2-bit words into single 3-bit channel words are possible. Those conversion rules are given in the following three tables.

[0102] It is evident that extensions of those conversion rules for encoding blocks of two, three or four 2-bit words into blocks of two, three or four 3-bit channel words, respectively, can be obtained using the teachings given above.

[0103] FIG. 10 shows an embodiment of a decoding device for decoding the channel signal generated by the encoding device of FIG. 1 or 2 into a replica of the binary source signal. The decoding device has an input terminal 80 for receiving the channel signal, comprising a sequence of m-bit channel words. The input terminal 80 is coupled to an input 82 of a conversion unit 84, which has an output 86 coupled to an input 88 of a reshuffling unit 90. An output 92 of the reshuffling unit 90 is coupled to an input 94 of a deletion unit 96, which has an output 98 which is coupled to the output terminal 100 of the decoding device.

[0104] The reconverter unit 84 is adapted to convert subsequent blocks of p consecutive m-bit channel words in said channel signal into corresponding subsequent blocks of p consecutive n-bit words. The datasteam of n-bit words at the output 86 is a replica of the shuffled composite binary source signal, where q-bit merging words are present at equidistant positions in the shuffled composite binary source signal, and where subsequent q-bit merging words in said shuffled signal are separated by groups of x.n bits in said shuffled composite binary source signal. The reshuffling unit 90 is adapted to carry out a reshuffling operation on each second group of said groups of x.n bits in the said replica of the shuffled composite binary source signal, more specifically, by repositioning q bits of a second group of x.n bits into another position in the said group of x.n bits. This results in a replica of the composite binary source signal at the output 92. The deleting unit 96 deletes the q-bit merging words present in said replica of the composite binary source signal, resulting in a replica of the binary source signal at the output 98, which replica is supplied to the output terminal 100.

[0105] It should be noted that x is an integer larger than 1, n and q are integers for which holds n=2q and q≧1, m and p are integers for which holds m>n>1, p≧1, and where p can vary.

[0106] Further, the reshuffling operation comprises the step of repositioning said p bits such that the n-bit words in the replica of the composite binary source signal are aligned with the n-bit words in the replica of the binary source signal.

[0107] FIG. 11 shows a slightly modified version of the decoding device of FIG. 10. The decoding device of FIG. 11 is for converting a channel signal comprising 3-bit channel words into a replica of the binary source signal comprising 2-bit source words. The channel signal is obtained from reading the channel signal from a track on a record carrier, which is either in the form of a magnetic record carrier 36 in longitudinal or disk form, or in the form of an optical record carrier, such as an optical disk 36′. A read unit 102 is therefore provided which comprises a read head 104, which is a magnetic read head, when recording the signal on a magnetic record carrier or an optical write head, when recording the signal on an optical record carrier.

[0108] The channel signal thus obtained is converted in the reconverter unit 84′, which is adapted to convert subsequent blocks of p consecutive m-bit channel words in said channel signal into corresponding subsequent blocks of p consecutive n-bit words in said shuffled composite binary source signal. The reshuffling unit 90′ is adapted to reshuffle 1-bit in each second group of x.n bits to another position, as per the shuffling shown in either the FIGS. 4dand 4eor the FIGS. 5dand 5e, so as to obtain the replica of the composite binary source signal. The deletion unit 96′ deletes the 1-bit merging words from the replica of the composite binary source signal so as to obtain the replica of the binary source signal.

[0109] The reshuffling unit 90′ for reshuffling the 1-bit b₇into its original position, as shown in the FIGS. 4dand 4e, can be in the form of the shuffling unit shown in FIG. 7.

[0110] Under the influence of a control signal on the line 46, the switches 42 and 44 are both in their position a-c, so that the serial datastream of the replica of the shuffled composite source signal supplied to the input 7 is directly fed to the output 9. In this way, the words rw₁and rw₂and the merging word mw₀are supplied to the output 9. Under the influence of the control signal on the line 46, the switch 42 is switched into its position a-b, at a time instant that the bit b₇in the word rw₃appears at the input 7. Further, upon the influence of the control signal on the line 48, this bit b₇is stored in the memory M. Next, uypon the appearance of the word rw_3,the switch 42 is again switched into of the position a-c, so that the words rw₃, rw₄. . . and so on can be supplied to the output 9. This continues until the word rw_i+4appears at the input 7 and the bit b₆in the said word is supplied to the output 9. Next, under the influence of a control signal on the line 46, the switch 44 is switched into its position a-b, so that the bit b₇stored in the memory M is supplied to the output. Next, under the influence of the control signal on the line 46, the switch 44 now switches over to its position a-c, so that the merging word mw₁and subsequent words rw_i+5, rw_i+6, . . . and so on, appearing at the input 7, can be supplied to the output 9. In this way, the bit b₇in the word rw₃is actually inserted in its original position of the serial datastream of the replica of the composite binary source signal. This is done for each second group of 48 subsequent source bits in the replica of the shuffled composite binary source signal.

[0111] The reshuffling unit 90′ for reshuffling the 1-bit b₀into its original position, as shown in the FIGS. 5dand 5e, can be in the form of the shuffling unit shown in FIG. 6.

[0112] Under the influence of a control signal on the line 46, the switches 42 and 44 are both in their position a-b and under the influence of a control signal on the line 48, words rw_iare stored in the shift register 40 and shifted further to the right in FIG. 6 in the shift register 40. Storage of words is carried out by storing the word rw_i, and subsequent words thereafter. Storing the information is continued until the word rw₁is stored in the memory locations M₄₇and M₄₆, the word rw₂is stored in the memory locations M₄₅and M₄₄and the merging word mw₀is stored in the memory location M₄₃. Five subsequent clock signals on the line 48 results in shifting the words rw₁and rw₂and the merging word mw₀out of the shift register 40, so that they are applied to the output 9. The shift register now has the bit b₁stored in the memory location M₄₇. Further, the bit b₀, in the word rw_i+4is present at the input 7.

[0113] Under the influence of the control signal on the line 46, the switches 42 and 44 now switch over to their positions a-c, so that the bit b₀is supplied to the output 9. Next, the switches 42 and 44 are again positioned into their position a-b, so that the bit b₁, stored in the memory location M₄₇, and the subsequent words rw₄, rw₅, . . . stored in the shift register 40, can be supplied to the output 9. In this way, the bit b₀is actually inserted in its original position in the serial datastream of the replica of the composite signal. This is done for each second group of 48 subsequent source bits in the replica of the shuffled composite binary source signal.

[0114] It will be clear that, where in the example of FIG. 4 the bit b₆in the source word sw₃was repositioned in the shuffling step, a corresponding reshuffling step should be carried out to reposition the bit b₆into its original position. Further, where in the example of FIG. 5, the bit b₁was repositioned in the shuffling step, a corresponding reshuffling step should be carried out to reposition the bit b₁into its original position.

[0115] Further, it should be noted that the reshuffling step has been described as a step that precedes the deletion step. It should however be noted that the reshuffling step could follow the deletion step, or could coincide with the deletion step.

[0116] FIG. 12 shows an embodiment of the reconverter unit 84′ of FIG. 11 for decoding the channel signal so as to obtain said replica of the shuffled composite binary source signal. The converter unit 84′ has an input 150 for receiving the channel signal, which input terminal 150 is coupled to an input 156 of a shift register 151, comprising twelve cells Y₁to Y₁₂. The shift register 151 functions as a serial-parallel convener so that blocks of four 3-bit channel words are applied to inputs i₁to i₁₂of a logic circuit 152. The logic circuit 152 comprises the four tables I, II, III and IV. Outputs o₁to o₈of the logic circuit 152 are coupled to inputs of cells X₁to X₈of a shift register 154, which has an output 157 coupled to an output terminal 155. A detector circuit 153 is present having inputs i₁to i₉, schematically indicated by the reference number 160, coupled to outputs of cells Y₄to Y₁₂respectively of the shift register 151, and outputs o₁, o₂and o₃coupled to control inputs c₁, c₂and c₃, respectively, of the logic circuit 152. The detector circuit 153 is capable of (a) detecting a ‘010 010 010’ bit pattern in the cells Y₄to Y₁₂of the shift register 151, (b) detecting a bit pattern ‘010 010’ in the cells Y₄to Y₉of the shift register 151, whilst the bits in the cells Y₁₀, Y₁₁and Y₁₂are unequal to ‘010’, and (c) detecting a ‘010’ bit pattern in the cells Y₄, Y₅and Y₆, whilst the bits in the bitcells Y₇, Y₈and Y₉are unequal to ‘010’.

[0117] Upon detection of the ‘010 010 010’ bitpattern, the detector circuit 153 generates a control signal on its output o₁, upon detection of the ‘010 010’ bit pattern in the cells Y₄to Y₉, the detector circuit 153 generates a control signal on its output o₂, upon detection of the ‘010’ pattern in the cells Y₄to Y₆, the detector circuit 153 generates a control signal at its output o₃, whilst, when there is no ‘010’ bit pattern in the cells Y₄to Y₁₂, it generates no control signal on its outputs.

[0118] In the absence of the control signals, the logic circuit 152 converts the 3-bit channel word stored in the cells Y₁, Y₂and Y₃into its corresponding 2-bit word, as per the conversion table I, and supplies the 2-bit word to the cells X₁and X₂. In the presence of the control signal at the input C₃, the logic circuit 152 converts the block of two 3-bit channel words stored in the cells Y₁to Y₆into a block of two 2-bit words, as per the conversion table II, and supplies the two 2-bit words to the cells X₁to X₄. In the presence of the control signal at the input c₂, the logic circuit 152 converts the block of three 3-bit channel words stored in the cells Y₁to Y₉into a block of three 2-bit words, as per the conversion table III, and supplies the three 2-bit words to the cells X₁to X₆. In the presence of the control signal at the input c₁, the logic circuit 152 converts the block of four 3-bit channel words stored in the cells Y₁to Y₁₂into a block of four 2-bit words, as per the conversion table IV, and supplies the four 2-bit words to the cells X₁to X₈.

[0119] In this way, the serial datastream of the channel signal is converted into the replica of the shuffled composite binary source signal.

[0120] FIG. 13 shows again another shuffling scheme that can be carried out on the source signal. In this embodiment, the shuffling is not done on the basis of groups of an integer number of the source words. In the present example, the shuffling is carried out on each second group of y bits of the source signal. In the present example, y=45, as can be seen in FIG. 13.

[0121] FIG. 13ashows again the serial datastream of the binary source signal. FIG. 13bshows how a merging word mw₀is inserted before the source word sw₁and a merging word mw₁is inserted before the source word sw_i. The shuffling step now comprises repositioning the bit b₄of the byte B₅, that is: the first bit of the word sw₂₃. The result of the merging and shuffling step is shown by the signal of FIG. 13b, which shows the shuffled composite signal in the form of 2-bit words w₁, w₂, . . . w_i, w_i+1, . . . The insertion of the merging word mw₀in fact means a shift of the source words in the source signal over two bit times, whilst the bit b₄of the byte B₅is stored in the second bit position of the word w₁.

[0122] This is schematically shown by the arrow 154 in FIG. 13b. The following source words Sw₁, Sw₂, . . . are shifted over two bit positions. In the third group of y bits shown in FIG. 13a, again the last bit of the group os shifted to a location in front of the first bit in the group. This is schematically shown by the arrow 154′ in FIG. 13aand 13b.

[0123] In all cases, the boundaries between the 2-bit words w in the shuffled composite signal of FIG. 13band the source words sw in the source signal of FIG. 13aare aligned.

[0124] Converting the 2-bit words of the shuffled composite signal into 3-bit channel words leads to the channel signal only schematically shown in FIG. 13c. Subsequent reconversion leads to a replica of the shuffled composite signal which is shown in FIG. 13d. Next, the merging words are deleted from the replica of the shuffled composite signal. That means: the first bit of the word rw₁is deleted, the first bit of the word rw₂₄and the first bit of the word rw_i+1, is deleted.

[0125] Further, the bit b₄at the second bit position of the word rw₁is repositioned to the first bit position of the word rsw₂₃, see the arrow 156, and the arrow 156′ indicates the repositioning of the bit b₁in the third group of 45 bits to its original position.

[0126] An error occurring during the conversion/reconversion step results in an erroneous word rw_iin the replica of the shuffled composite signal of FIG. 13d. As the boundaries between the words rw_iin the signal of FIG. 13dand the words rsw₁in the signal of FIG. 13eare aligned, such error results in only one 2-bit word, and thus only one byte, to be erroneous, so that no error propagation occurs.

[0127] It should be noted that the shuffling step has been described as a step that follows the merging step. It should however be noted that the shuffling step could precede the merging step, or could coincide with the merging step.

[0128] 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 in the claims. As an example, the converter unit of FIG. 12 could be modified into a converter unit in which the detector 153 detects the various modified decoding situations from the decoded information, instead of from the encoded information, as disclosed in FIG. 12.

[0129] Further, the invention lies in each and every novel feature or combination of features.