JPS6129947A - Code error correction system - Google Patents

Abstract

PURPOSE:To improve the correcting capability by making the processing method of an error flag different depending whether the error location obtained from a syndrome corresponds to any of two words having ''1'' level of error flag or does not correspond to any word. CONSTITUTION:When the number of error flags in a sub block is 2 and both syndromes SPY, SQY are other than ''0'', the two words having ''1'' level of error flags are corrected by the SPY, SQY. When the location of error obtained by the SPY, SQY corresponds to any of the two words having ''1'' level of the error flags, all the error flags follow to the next X direction error detection result or all the error flags are reset, and when the location does not correspond to any word, the error flag of the corrected two words is set and the error flag of the other word follows the next X direction error detection result.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides an error detection code that corrects code errors in a code block obtained by encoding a plurality of data words using an error detection code and two error correction codes. and two error correction codes are added to form a code block as shown in FIG. Patent application 1986-14040
It is disclosed in the specification of No. This patent application 1984-1404
The method disclosed in the specification of No. 0 (hereinafter referred to as the prior invention method) will be explained.

In FIG. 1, one block of data words Wo to W3q consisting of 440 words is arranged in the shape of a rectangular parallelepiped with 10 words in the X direction, 4 words in the Y direction, and 11 words in 2 directions. Each data word sequence in each of the X, Y, and Z directions is encoded as a subblock, with two additional test words P and Q formed by a Reed-Solomon code.

The subscripts x, y, and z of P and Q indicate the expansion direction of the subblock containing the check word, and the numerical subscript corresponds to the number of the first word of the subblock in the expansion direction. The word expressed as Px Py o is
This indicates that it is a test word Px in the X direction and a test word PY in the Y direction, and the numbers of the first PX and PY in each direction are O. Further, a word expressed as QX QYQZ indicates that it is a test word Qx in the X direction, a test word QY in the Y direction, and a test word Qz in the Z direction.

The same applies to other words expressed by combinations of P or Q and subscripts. The reason why one word can serve as a check word in multiple directions at the same time is because the Reed-Solomon code is a linear code.

Here, it is assumed that each word of data and check consists of 8 bits, and that the Reed-Solomon code is a code on the Galois field G[(2). Also, the data word W o-W4! The subscript of q represents the order before three-dimensional arraying, and these W
o~W,! ? are arranged as shown in FIG. 1, and check words are added to them.

2, one word at a time is extracted and recorded or transmitted. The order of the words at this time is Wo, W
n”'W95+ Pxo, Qxo, Woo,
-W22(1, -W33o, -Wau, Px 3
30. Qx 33(1,-Pv o .

-Pysj, PX PYO, QX PYO, QYO, -
QY95. PXQYO, QXQYO, W+, -W2゜...W3. ...W4. ...Ws...
・・W6r・・・W7゜−Wa, −Ws, ・=W+
o, -Woos, PX10, Qx+o,
-0ys+o, -Qx Pz 3311. -Qx QY
In, Pzo, -Pz9s, PxPzo,
QxPzo, -QxPz330° -Qx PY PZ
, -QxQv Pz , -Qz o , ・=Qz9g
, PzQzo, QxQzo, =QxQz33o.

...Qx PYQZ, -Qx Qy Qz.

On the regeneration or reception side, the word sequence is arranged again as shown in Figure 1, and after correcting errors that occurred during recording or transmission, only the data words are extracted in the order of z, x, y, and the original A data word sequence of Wo, W+, . . . W2B5 is obtained.

The reason for changing the order of words during recording or transmission as described above is to disperse erroneous words in the original data word sequence when uncorrectable burst errors occur in regenerated or received data. be.

Check words PI, Qu (U=X, Y or Z) consisting of Reed-Solomon codes in the code block of FIG.
is generated to satisfy the following equation.

ΣW1 dimensionru +Qu =0 -(+), M
'(""" +clFUtenα, = Q ,,,
(2) ml However, the addition is modulo 2, and α is GF (28)
Let it be the primordial light above. Also, V is U=X.

The values of 10, 14, and 11 are assumed for each of Y and Z, and the subscript of W indicates the position of the data word in the subblock, with 1 being the first.

The following equation is obtained from equations (1) and (2).

On the regeneration or receiving side, if the error correction position is known for the subblocks in each direction in the code block in Figure 1, errors in data and check words within 2 words can be corrected, and the error position can be corrected. If not known, errors in one word of data and check word can be corrected. These corrections are made as follows. That is, first, syndromes Spu and Sou defined by the following equations are calculated for data containing errors and check words.

5, o-ΣvJ, 10f, → Q u , −(
5,) However, the subscripts of U, V, and are the same as the subscripts of the equations of Pu and Qu.

Now, if a two-word error occurs at each position k = i, j, and the error patterns are ei and ej, respectively, then
The syndromes Spu and Sou are as shown in the following equations, respectively.

S p u =ei+ej-(7) Sov-a e++a 2-Jej-・---
-(8) From equations (7) and (8), ei and ej are as follows.

It can be seen from equations (9) and (10) that if positions i and j are known, ei and ej can be determined from the above equations, making error correction possible.

Next, if a one-word error occurs at position i and the error pattern is ei, then the syndrome Spu, sou
are as shown in the following equations.

S p u =ei... (11) So u =
a e+-(12) The following formula is obtained from the formulas (11) and (12).

From equation (13), even if position 1 is not known, Spu.

By calculating αi from Sou, 1 is obtained and e
Since i becomes Spu itself, it can be seen that error correction is possible.

In the code block shown in Figure 1, each word in the X direction is continuous on the recording or transmission system, so errors of 3 or more words often occur consecutively in the X direction, and correction is more difficult than in the Y and Z directions. There is a high probability that it will be impossible. Therefore, in the X direction, the Reed-Solomon code had to be used as an error detection code, and if Sp x -8Q Only the value of the error flag indicating the presence or absence of an error can be determined.

In the conventional system, code errors in code blocks as described above are corrected according to the procedure shown in the flowchart of FIG. That is, first, error detection is performed for each subblock in the X direction, and an error flag that is either "1" or "0" is added to all words in the subblock depending on the presence or absence of an error. Next, error correction in the Y direction is performed nine times over the entire code block. Next, error detection in the X direction is performed again, and then error correction in the Z direction is performed over the entire code block. After performing the above correction processing eight times, error detection in the X direction is performed again to complete the correction.

Error correction in the Y direction in the flowchart of FIG. 2 is performed according to the procedure shown in the flowchart of FIG. That is, first, for a subblock whose error flags are all 0'' among subblocks extending in the Y direction, if the syndromes SPY and SQY are both O, it is assumed that no error exists and all error flags are reset. The error flag set to “O” is set to “O” even if an error is detected in the next error detection in the X direction.
, SQY are both 0'', the error flag is “O”.
II, but there are erroneous words, ie, undetected errors. Therefore, when the error position determined by SPY and SQY corresponds to any word in the subblock, that word is determined to be an undetected error word and corrected. After this correction, all error flags are reset to 0'' as if no error exists.S
If the error position determined by PY and SQY does not correspond to any word in the sub-block, there is an undetected error of two or more words, which cannot be corrected by SPY and SQY alone. Therefore, at this time, all error flags are set to 1'', and the set error flags remain 1'' even if no error is detected in the next error detection in the X direction.

Next, for a sub-block where only one error flag is 1'', if both Spy and Soy are 'O', it is determined that the error flag is '1' even though it is a correct word. and reset all error flags to 0''. SPY.

Unless SOY is both 'O'', the error position is determined by SPY and SOY, and the error flag is 1'' at the determined error position.
, the word is determined to be incorrect, corrected, and all error flags are reset. The obtained error position is the error flag II I I
If it does not correspond to the I word, there must be some other undetected error.Correcting with TEARI, SPY, and SQY may result in a false correction, so do not perform any correction and set all error flags. . In addition, Kometoki SpY,
Correction is performed by SQY, and all error flags may be set after correction.

Also, for a sub-block with two error flags '1', if both Spy and Say are not '0', it is assumed that the two words with error flags '11' are incorrect, and the positions of these two words are assumed to be incorrect. SPY, SQ with error position
Make corrections with Y. At this time, the error flag is neither set nor reset, but is set according to the result of the next error detection in the X direction. Both SPY and SQY are 1
10 II, the error flag is It I II 2
It is determined that the error flag is set to 1'' even though both words are correct, and all error flags are reset.

Also, for sub-blocks with three or more error flags 111++, nothing is done if both SPY and SQY are 0'', otherwise SPY and S.

Find the error position using Y. When this error position corresponds to one of the words with an error flag of 1'', it is determined that that word is an error and the error flag is ``1'' even though the other words are correct. Correct one word corresponding to the error position found. If the error position does not correspond to any word whose error flag is 1'', there is an error of two or more words, and even if the error is two words, the position is unclear and correction is impossible. In this case, regardless of the values of SPY and SQY and whether or not they are correctable, the error flag is neither set nor reset, and everything follows the result of the next X-direction error detection. 3 or more II I
In the case of ++, there may naturally be an error of 3 or more words, and in that case, SpY and Sov may both be 110 by chance.
++, or the determined error position may coincidentally correspond to any word whose error flag is 1''.

In the above correction processing, the processing of error flags after correction is changed depending on the number of 1'' error flags before correction and the correction situation, but this is based on the probability that an error is included after correction. Error flags are reset, error flags are set, and the error flags are reset and set for each case.
This is to make processing without setting and resetting the error flag correspond. Also, since the sub-blocks extending in the X direction and the sub-blocks extending in the Y direction intersect with each other, the sub-blocks extending in the X direction are An independently processed error flag is added to each word in the sub-block.

Furthermore, in the X-direction error detection that follows the Y-direction error correction in the flowchart of FIG. 4, the error flag value is set for each word according to the result of Y-direction error flag processing and the X-direction error detection result. I decided to. That is, the error flag reset and the error flag set in the Y-direction error correction become "0" and "1", respectively, regardless of the result of the next X-direction error detection. In addition, error flags that are neither set nor reset during error correction in the Y direction are
If an error is detected at the time of direction error detection, it becomes 1'', and if no error is detected, it becomes "O".

After such error detection in the X direction is performed, there are fewer words with an error flag of 1'' even though they are correct, and words with an error flag of O'' even though they are incorrect, and 7 in the flowchart of FIG. In direction error correction, effective error correction can be performed based on accurate error flags.

Error correction in the Z direction is performed in the same way as error correction in the Y direction.
By performing error detection in the X direction again after error correction in the direction, the next error correction in the Y direction can also be performed based on an accurate error flag.

After error correction is performed 0 times in each of the Y direction and the 7 directions as described above, error detection is performed once again in the X direction in order to detect errors that are ultimately uncorrectable, and the correction is completed. .

As explained above, the system of the prior invention has a high error correction ability and can be said to be suitable for the code block shown in FIG. 1, but cannot correct the error pattern shown in FIG. 4. In FIG. 4, five Y-Z hundred M at the end of the first X-direction error detection for the code block of FIG.
The states of 1 to M5 are shown, 1, ``0'' indicates a word that is incorrect but the error flag is 0'', and 41, II is incorrect and the error flag is also ``1''. ” indicates a word, and an “×” indicates a word whose error flag is 111 II even though it is correct.The Reed-Solomon code with a check word count of 2 has a minimum distance of 3, so , when there is an error of 3 or more words in one subblock, the syndrome may become "O" and the error may not be detected. A 3-word error has occurred that is not detected by the initial error detection in the X direction, and other errors have also occurred. FIGS. 5(A) to 5(D) show one Y-2 plane M2 including M2. That is, FIG. 5(A) shows the Y-2 plane M2 at the end of the first X-direction error detection. -2 side M2 is shown, and (B) of the same figure shows the state of the first Y
The state of the Y-2 plane M2 at the end of the second error detection in the X direction following the error correction in the direction is shown in FIG.
The state of the Y-2 plane M2 at the end of the third X-direction error detection following the first seven-direction error correction is shown in the figure (
D) shows the state of the Y-2 plane M2 at the end of the fourth error detection in the X direction following the second error correction in the Y direction. In FIGS. 5(A) to 5(D), there are two '1'' error flags in sub-block Y1 during the first error correction in the Y direction after the first error detection in the X direction. False corrections will be made to the two 'X'° corresponding to this error flag. False corrections will also be made to the other Y-2 planes M3 and Ma that include '0'1. The word that was falsely corrected is
Two subblocks each contain three words in the X direction, and these are errors that are not detected in the X direction. This is because errors due to false corrections are derived from undetected errors in the X direction, and Reed-Solomon codes are linear codes. In sub-blocks Y2 and Y3, the syndrome in the Y direction is O, but in the other Y-2 plane “...
Since the two words ゛・'' included in the error pattern No. 11 are not corrected, the error flag is set to ``1'' in the second X-direction error detection that is performed after the first Y-direction error correction.
becomes. The first 7 following this second X-direction error detection
During error correction in the direction, false correction is performed on sub-block Z1 in the same way as sub-block Y1, and sub-block Z
2 and Z3, the error flag becomes uncorrectable and the "0" word becomes "." and the other words become "x". At this time, subblock Z2. Z3
If correction is performed as if it is correctable, PA0''' will be corrected, but the error flag will be set.After this, in the second Y-direction error correction following the third X-direction error detection, sub-block Y+, False corrections are made in both Y2 and Y3, resulting in a code block of 27
Undetected word errors will occur. These errors will not be detected or corrected no matter how many times the correction is repeated.

In the prior invention method, in addition to the error patterns shown in FIG. 4, there are error patterns that include undetected errors and whose occurrence probability cannot be ignored, and which cannot be corrected.
These determine the limits of the correction ability of this prior invention method.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a code error correction system that can correct error patterns that could not be corrected using the system of the prior invention, and that has a further improved correction ability.

The code error correction method according to the present invention corrects code errors in a code block consisting of a plurality of data words, an error detection code, and a plurality of check words formed by first and second error correction codes. In addition to the correction processing performed by referring to the error flag during error correction by each of the second error correction code and the data word that has a low probability of being an error at the end of the correction processing, and the error detection code, there is a high probability that the error will be an undetected error. The error flags of the data words are changed to correspond to no errors and errors, respectively, and for error flags for which the above change process is not performed, the next time error detection is performed, the error detection code is used for error detection. A code error correction method in which a value is determined by a result and the value before the correction processing is held when no error detection is performed by the error detection code, wherein the first and second error correction codes are , if the erroneous word is not known, the error in one word in one subblock can be corrected, and if the erroneous word is known, then the error in up to two words in one subblock can be corrected. It is a correctable code, and the number of errors caused by the error flag before correction in the subblock to be corrected is 2.
In this case, as the correction process, the error word indicated by the error flag is corrected by one of the first and second correction codes corresponding to the sub-block to be corrected, and as the change process, the error word is corrected by one of the first and second correction codes, and as the change process, the error word indicated by the error flag is If the number of errors determined by If it indicates an error, the error flags of all words in the subblock corresponding to the one are not changed or are changed to correspond to no error, and if the number of errors determined by the one is not O and the one is The present invention is characterized in that when the error flag of the word at the location of the error determined by the method does not indicate an error, only the error flags of the two corrected words are changed to correspond to the presence of an error.

Part-1-1 Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS. 6 to 11.

The procedure shown in the flowchart of FIG. 2 is also followed when correcting an error in a code block as shown in FIG. 1 using the code error correction method according to the present invention. However, error correction in the Y direction is performed according to the procedure shown in the flowchart of FIG. Further, error correction in the Z direction is also performed according to the same procedure as the error correction in the Y direction.

In the procedure shown in the flowchart of FIG. 6, the processing other than when the number of error flags of l 1' I+ in the subblock is 2 is the same as the procedure shown in the flowchart of FIG. 3. ° If the number of "1" error flags in the sub-block is 2, if SPY and SQY are both "0", all error flags are reset to O", except when SPY and SQY are both "OI+" corrects the two words whose error flag is "1" using SpY and SQY. This is the third
The procedure is similar to that shown in the figure, but the two words are SPY, SQ
When a correction is made using Y, the procedure shown in Figure 3 does not set or reset all error flags, whereas the procedure shown in Figure 6 shows that the position of the error determined from SPY and SQY is determined by the error flag " Correct when the error position determined from Spv and sov does not correspond to any word with an error flag of 1111+ without setting or resetting all error flags when it corresponds to either one of the two words with an error flag of 1". 2 with the error flag ``1''
It is assumed that the error flag of a word is set and the error flags of other words are neither set nor reset.

If the position of the error determined from SPY and SQY corresponds to one of the two words with the error flag "1", the error flag is "1" even though that word is incorrect and the other is not. There is a high probability that the error pattern is an error pattern with At this time, all 1 error flags may be reset.Even if this is done, the final correction result is almost the same as when neither setting nor resetting is done.On the other hand, the correction result obtained from SPY and SQY If the location of the error does not correspond to any word with an error flag of 1'', there is a high possibility that an undetected error exists, and the probability of a false correction cannot be ignored, so correction is not performed. Only the error flag of the word is set, and the error flags of other words are not set or reset. Therefore, errors caused by false correction are detected and corrected in the next stage. Errors that are not detected are detected. However, since other errors in the same sub-block have been detected, there is a high possibility that they will be corrected in the next stage.If no false correction occurs, an error flag is set for the two correctly corrected words. will be set, and the number of words with an error flag of 1'' will increase even though there is no error, but such words can be corrected for each correction by repeating the correction as shown in Figure 2. As the number of erroneous words decreases, the effect on the final error correction ability of the entire correction process is small.Therefore, the method according to the present invention reduces the number of false corrections due to undetected errors compared to the method according to the prior invention. The method has excellent correction ability for error patterns such as those that perform the following, and if the number of words with an error flag of 1'' in the sub-block is 1 or less or 3 or more, the same processing as the prior invention method is performed, so the method is similar. Has high correction ability.

In addition, the last (fifth) error correction in the seven directions in the procedure shown in Fig. 2 will be performed according to the procedure shown in Fig. 3, or the last error detection in the X direction will be performed by erasing all previous error flags. shall be ignored. This is to detect an error only when it is truly uncorrectable after all correction processes are completed.

The fourth error correction method performed by the code error correction method according to the present invention
Figure 7 (A) shows how the error pattern shown in the figure is corrected.
) to (D) of the same figure. FIG. 7(A) shows the state of the Y-2 plane M2 at the end of the first error detection in the X direction, and FIG. 7(B) shows the state of the Y-2 plane M2 at the end of the first error correction in the X direction. (C) shows the state of the Y-2 plane M2 at the end of the error detection in the X direction after the first error correction in the two directions.
The state of the plane M2 is shown, and FIG. 2D shows the state of the Y-2 plane M2 at the end of the fourth error detection in the X direction following the second error correction in the Y direction. During the first Y-direction correction, a false correction is performed in sub-block Y1, a new 2-word error occurs, and the error flag for these 2 words is set to 1''. In sub-blocks Y2 and Y3, the A change similar to the correction by the prior invention method shown in Fig. 5 occurs.After this, a second error detection in the X direction is performed, but no error is detected in the "O" word. In the first two-direction error correction performed following error detection, a false correction is performed in sub-block z1 in the same way as the first Y-direction error correction in sub-block Y1, and a new ?word error occurs. However, the error flag of these two words is set to "°1".

At the same time, in sub-blocks Z2 and Z3, “′・X
The I+ error pattern is corrected. At this time, the values of the error flags of sub-blocks Z2 and Z3 are determined by the result of the third error detection in the X direction that is performed next, and the values of the error flags of ゛... Since the error pattern has also been corrected with the error t1 positive in the first seven directions, the third error detection in the X direction subblock Z2,
The error flags of all words in Z3 become '○''.Incidentally, the error flags in sub-blocks Z2 and Z3 may all be reset after the correction. The error flags of all words in sub-blocks Z2 and Z3 become O''.

In the second Y-direction error correction following the third X-direction error detection, the word °'0'' is corrected in sub-block Y1, the word "・" is corrected in sub-blocks Y2 and Y3, and all The word will be correctly corrected.

The code error correction method according to the present invention is not limited to the error pattern shown in FIG. 4, but can similarly correct other error patterns that include undetected errors and whose occurrence probability cannot be ignored.

Although it has been described above that the correction is performed according to the procedure shown in FIG. 2, various modifications can be made to this procedure. That is, in the correction processing procedure as shown in the flowchart of FIG.
Although error detection in the X direction has been performed, it is possible to partially omit error detection in the other X directions except for the first error detection in the X direction. FIG. 8(A) and FIG. 8(B) are flowcharts showing an example of this. Error detection in the X direction is omitted before correction. If error detection in the X direction is omitted, the value of the error flag must be determined after previous error correction. Therefore, in the Y-direction error correction that follows the procedure shown in the flowchart of FIG. "No error flag set/reset" processing and "Set only the corrected two words,"
Changed the process "No set/reset for all others" to "Keep error flag value before correction" and "No set/reset for all error flags" process when the number of error flags in 111 II is 2. It is necessary to perform the correction process according to the procedure obtained by changing ``all error flags reset'' to ``reset all error flags''. By doing so, all error flag values are determined at the time when error correction in the Y direction is completed. Furthermore, when correcting errors in the Z direction in the procedure shown in the flowchart of FIG. 8(B), it is also necessary to perform the correction process according to a procedure obtained by changing the error flag processing in the procedure shown in the flowchart of FIG. There is. Note that if error detection in the X direction is omitted, the number of words with an error flag of 1'' will increase even though there is no error, so the effect of error correction immediately after will be small, but the number of times correction will be repeated will increase. If there are many, there will be no practical problem.

In the above embodiment, after the first error detection in the X direction, the first error correction is immediately performed for 6 seconds, but the first error detection is performed in two directions and then the first error correction is performed. Good too. For example, error detection in seven directions may be performed after error detection in the X direction, and then error correction in the Y direction may be started. At this time, the value of the error flag is determined after error detection in the Z direction, and the error flag processing is, for example, the syndrome SPY in the flowchart of FIG.

Change SQY to Spz and Soz respectively, remove the correction processing step due to syndrome, and process ``all error flags, set, no reset'' and ``set only 2 corrected words, do not set or reset the others'' An error occurred when the step was changed to "holding the value of the error flag for error detection in the 1x direction", and the process was shifted to the processing step of one-word correction due to the syndrome when the number of "1" error flags is less than or equal to 1. flag processing
This can be done by a procedure according to a flowchart, such as "setting only the error flag of the error position determined by spz and Soz, and resetting the others." Further, in the above embodiment, all error corrections are started from the Y direction, but error correction may be started from seven directions.

In addition, the error detection in the X direction before the end of correction in the flowcharts of FIGS. 2 and 8 is only for detecting residual errors after error detection, and may be used in cases where detection is not necessary or detection accuracy is not required. may be omitted.

Further, various modifications can be made to the flowchart shown in FIG. 6. For example, in the flowchart of Figure 6, 1″
When the number of error flags in is 1, Sp v =SQ
Y = O, that is, if the number of errors obtained from SPY and SQY is 0, all error flags are reset; otherwise, the position of the error obtained from SPY and SQY corresponds to the word whose error flag is "1". When the word is corrected and all error flags are reset, if it does not correspond, all error flags are set without correction, but
The process when the error position determined from SPY or SOY does not correspond to the word with the error flag "1" may be replaced with the process (1) or (II) shown below. (1) SPY , SQY, if the error position corresponds to a word in the subblock, it is corrected, and after correction, all error flags are set.

(II) If the error position determined from SPY and SQY corresponds to a word in the subblock, correct it;
After correction, all error flags are neither set nor reset.

Further, in the above embodiment, a Reed-Solomon code with a check word count of 2 is used as the error detection code and the error correction code, but any other code may be used as the error detection code, and the error correction code is 1 word if the error location is known, and 2 words if the error location is known.
Any code that can correct word errors may be used.

The above has explained the correction of code errors in data words applied to a rectangular parallelepiped code block as shown in FIG.
The present invention can be applied to code error correction of data words in any code block as long as it is formed by being encoded with an error detection code and the above two error correction codes. For example, as shown in Figure 9, a code block of infinite length (in reality, it is finite) with mixed codes, or a code block of finite length is extracted from a code block as shown in Figure 9 and both ends are removed. It is clear that the present invention can also be applied to code blocks as shown in FIG. 10 in which the codes are connected to each other and the codes are circulated.

Incidentally, a code block as shown in FIG. 9 is obtained by an encoding circuit as shown in FIG. 11. In Figure 11, Z
The a data words forming subblocks extending in the direction are provided to a two-way error correction encoder 30. The two-way error correction encoder 30 is configured to output the supplied a data words as they are, and at the same time generate and output check words of a seven-way error correction code. The a data words and the seven direction check words outputted from the Z-direction error correction encoder 30 are provided in delay circuits D11 to D+ a having different delay times, respectively.

By passing through D+(a++) to [)lb, a subblock extending in the Y direction is formed and supplied to the Y direction error correction encoder 31. The Y-direction error correction encoder 31 is configured to output the outputs of the delay circuits Du to D1b as they are, and at the same time generate and output a check word of the Y-direction error correction code. This Y direction error correction encoder 31
The outputs of the delay circuits D21 to D having different delay times are
2a, D2 (a ++ ) ~ D2b, D2
(1) ++ ) to D2C to form a subblock extending in the X direction and supplied to the X direction error detection encoder 32. The X-direction error detection encoder 32 is configured to output the outputs of the delay circuits D21 to D2C as they are, and at the same time generate and output a check word of the X-direction error detection code. This X-direction error detection encoder 32
A code block as shown in FIG. 9 is formed by the output.

As detailed above, in the code error correction method according to the present invention, when the number of "1" error flags in a sub-block is 2 and 2-word correction is performed, the error position determined from the syndrome When the error flag corresponds to either one of the two words with 1'', all the error flags are made to follow the result of the next error detection in the X direction, or all the error flags are reset, and the result is calculated from the syndrome. The location of the error is the error flag is °″1″
When the corrected two words do not correspond to any of the words, the error flag is set, and the error flags of the other words follow the result of the next X-direction error detection, so that false corrections are not made due to undetected errors. The correction ability for such error patterns has been improved, and it has become possible to correct error patterns that were uncorrectable in the system of the prior invention, and as a result, the correction ability has been further improved. Therefore, the code error correction method according to the present invention is suitable for code blocks obtained by encoding with an error detection code and two error correction codes.

[Brief explanation of drawings]

FIG. 1 is a diagram showing a code block obtained by encoding a plurality of data words with an error detection and error correction code, FIG. The figure is a flowchart showing in detail a part of the error correction processing procedure in the first-invention method, FIG. 4 is a diagram showing error patterns that cannot be corrected in the first-invention method, and FIG. FIG. 6 is a diagram showing an embodiment of the present invention; FIG. 7 is a diagram showing the process of correcting the error pattern of FIG. 4 according to the method of the present invention; FIG. 8 is a flowchart showing an outline of the error correction processing procedure in another embodiment of the present invention, FIGS. 9 and 10 are diagrams showing an example of code blocks to which the present invention can be applied, and FIG. teeth,
10 is a block diagram showing an encoding circuit for forming the code block shown in FIG. 9. FIG.

Claims (4)

[Claims] (1) One of a plurality of subblocks obtained by dividing a code block consisting of a plurality of data words and a plurality of check words using each of the first, second, and third division methods.
Any two words of the plurality of data words and check words included in the first, second and third
an error detection code formed for each subblock obtained by the first partitioning method when the error detection code does not exist simultaneously in any of the subblocks obtained by the other two partitioning methods except for one of the partitioning methods; When correcting code errors in the code block when the plurality of check words are formed by first and second error correction codes formed for each sub-block obtained by the second and third division methods, respectively. An error flag indicating the presence or absence of a code error is added to each of the plurality of data words and the plurality of check words, and the error flag is referred to when error correction is performed by each of the first and second error correction codes. Along with the correction process, at the end of the correction process, the error flags of words with a low probability of error and words with a high probability of being an error that will be undetected in the error detection code are changed so that they correspond to no errors and errors, respectively. For error flags that are not subjected to the change processing, the value is determined by the result of error detection by the error detection code when the next error detection is performed by the error detection code, and no error detection is performed by the error detection code. The code error correction method is configured to hold the value before the correction processing is performed when the error word is not corrected, and the first and second error correction codes are configured to hold the value before the correction processing is performed, and the first and second error correction codes are fixed to one subblock when the erroneous word is not known. If the error in one word in a subblock can be corrected and the erroneous word is known, it is a code in which errors in up to two words in one subblock can be corrected, and the code before correction in the subblock to be corrected. When the number of errors caused by the error flag is 2, the error word indicated by the error flag is corrected by one of the first and second correction codes corresponding to the sub-block to be corrected. If the number of errors determined by the one method is 0, the error flags of all words in the sub-block corresponding to the one method are changed to correspond to no errors, and the number of errors determined by the one method is changed. If the error flag of the word at the location of the error indicates an error, the error flags of all words in the sub-block corresponding to the one above are either unchanged or changed so that they correspond to no error, and the error flags are determined by the one above. When the number of errors is not 0 and the error flag of the word at the position of the error determined by the above method does not indicate an error, only the error flags of the two corrected words are changed to correspond to the presence of an error. A code error correction method that uses (2) If the number of errors caused by the error flag before correction in the sub-block to be corrected is other than 2, the number of errors caused by the error flag before correction in the sub-block to be corrected is determined as the correction process. When the value is 0, if the error position determined by the one method indicates any word in the sub-block to be corrected, the word at the error position determined by the one method is corrected, and the word in the sub-block to be corrected is corrected. If the number of errors according to the error flag before correction is 1 or 3 or more, if the error flag of the word at the error position determined by the one method indicates an error, correct the word at the error location determined by the one method, and as the change process, when the number of errors due to the error flag before correction in the sub-block corresponding to the one is 1 or less, the number of errors determined by the one is 0, or the sub-block corresponding to the one If a word in the sub-block is corrected, the error flags of all words in the sub-block corresponding to the one are changed to correspond to no errors, and the number of errors determined by the one is not 0, and the error flag corresponds to the one. Claim 1, characterized in that when the word in the sub-block that has been corrected is not corrected, the error flags of all the words in the sub-block corresponding to said one are changed to correspond to the presence of an error. code error correction method. (3) If the number of errors caused by the error flag before correction in the sub-block to be corrected is other than 2, the number of errors caused by the error flag before correction in the sub-block to be corrected is determined as the correction process. 1 or less, if the error position determined by the one method indicates any word in the sub-block, the word at the error location determined by the one method is corrected, and the correction is made in the sub-block to be corrected. When the number of errors caused by the previous error flag is 3 or more, if the error flag of the word at the error position determined by the one method indicates an error, correct the word at the error location determined by the one method, and perform the changing process. If the number of errors due to the error flag before correction in the sub-block corresponding to the one is 0, then the number of errors determined by the one is 0, or the word in the sub-block corresponding to the one is corrected. If the error flags of all words in the sub-block corresponding to one of the above are changed to correspond to no errors, and the number of errors determined by the one is not 0, the error flags of all words in the sub-block corresponding to the one are changed to If the word is not corrected, the error flags of all words in the sub-block corresponding to one of the words are changed to indicate that there is an error, and the error flags of all words in the sub-block corresponding to the one are changed to correspond to the error flags before correction. When the number is 1, the number of errors determined by the one method is 0, but if the error flag of the word at the location of the error determined by the one method indicates an error, all of the errors in the subblock corresponding to the one method are If the error flag of a word is changed so that it corresponds to no error, and the error position determined by the above one is not the word in which the error flag indicates an error, the error of all words in the sub-block corresponding to the one above is detected. The code error correction method according to claim 1, characterized in that the flag is changed to correspond to the presence of an error. (4) If the number of errors caused by the error flag before correction in the sub-block to be corrected is other than 2, the number of errors caused by the error flag before correction in the sub-block to be corrected is determined as the correction process. 1 or less, if the error position determined by the one method indicates any word in the subblock to be corrected, correct the word at the error location determined by the one method, and correct the subblock to be corrected. If the number of errors caused by the error flag before correction in the error flag is 3 or more, if the error flag of the word at the error position determined by the one method indicates an error, correct the word at the error location determined by the one method, and As the change process, if the number of errors due to the error flag before correction in the sub-block corresponding to the one is 0, the number of errors obtained by the one is 0, or the number of errors in the sub-block corresponding to the one is 0. When a word is corrected, the error flags of all words in the sub-block corresponding to said one are changed so that they correspond to no errors, and the number of errors determined by said one is not 0, and the sub-block corresponding to said one is corrected. If the words in the block are not corrected, the error flags of all words in the sub-block corresponding to the one are changed to correspond to errors, and the error flags before correction in the sub-block corresponding to the one are changed. The number of errors due to
If the number of errors determined by one of the above methods is 0, or if the error flag of the word at the location of the error determined by one of the methods indicates an error, then all words in the subblock corresponding to the one method are errors. error flags of all words in the sub-block corresponding to said one when the flag is changed to correspond without error and the error position determined by said one does not indicate any word in the sub-block corresponding to said one; 2. The code error correction method according to claim 1, wherein the code error correction method is changed to correspond to the presence of an error.