CA2016847A1 - Error detection and correction systems for long burst errors - Google Patents
Error detection and correction systems for long burst errorsInfo
- Publication number
- CA2016847A1 CA2016847A1 CA002016847A CA2016847A CA2016847A1 CA 2016847 A1 CA2016847 A1 CA 2016847A1 CA 002016847 A CA002016847 A CA 002016847A CA 2016847 A CA2016847 A CA 2016847A CA 2016847 A1 CA2016847 A1 CA 2016847A1
- Authority
- CA
- Canada
- Prior art keywords
- error
- matrix
- error correction
- symbols
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
- 238000012937 correction Methods 0.000 title claims abstract description 142
- 238000001514 detection method Methods 0.000 title claims abstract description 20
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 21
- 239000011159 matrix material Substances 0.000 claims description 121
- 238000000034 method Methods 0.000 claims description 49
- 239000013598 vector Substances 0.000 claims description 26
- 241000269627 Amphiuma means Species 0.000 claims 2
- 238000002372 labelling Methods 0.000 claims 2
- 102100033962 GTP-binding protein RAD Human genes 0.000 description 12
- 101001132495 Homo sapiens GTP-binding protein RAD Proteins 0.000 description 12
- 102100027362 GTP-binding protein REM 2 Human genes 0.000 description 10
- 101000581787 Homo sapiens GTP-binding protein REM 2 Proteins 0.000 description 10
- 238000012546 transfer Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- OQCFWECOQNPQCG-UHFFFAOYSA-N 1,3,4,8-tetrahydropyrimido[4,5-c]oxazin-7-one Chemical compound C1CONC2=C1C=NC(=O)N2 OQCFWECOQNPQCG-UHFFFAOYSA-N 0.000 description 1
- AYFVYJQAPQTCCC-GBXIJSLDSA-N L-threonine Chemical compound C[C@@H](O)[C@H](N)C(O)=O AYFVYJQAPQTCCC-GBXIJSLDSA-N 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
- H03M13/13—Linear codes
- H03M13/15—Cyclic codes, i.e. cyclic shifts of codewords produce other codewords, e.g. codes defined by a generator polynomial, Bose-Chaudhuri-Hocquenghem [BCH] codes
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11B—INFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
- G11B20/00—Signal processing not specific to the method of recording or reproducing; Circuits therefor
- G11B20/10—Digital recording or reproducing
- G11B20/18—Error detection or correction; Testing, e.g. of drop-outs
- G11B20/1833—Error detection or correction; Testing, e.g. of drop-outs by adding special lists or symbols to the coded information
Abstract
ABSTRACT
The invention is an error detection and correction system which encodes data twice, once for error detection by using a cyclic redundancy check (CRC) code with a generator polynomial, g (x) [in octal form]:
g(x) = 2413607036565172433223 and a second time for error correction by using a Reed-Solomon error correction code. The system then uses the CRC code to check the data for errors. If errors are found the system uses the error location information supplied by the CRC code and the Reed-Solomon code to correct the errors.
The invention is an error detection and correction system which encodes data twice, once for error detection by using a cyclic redundancy check (CRC) code with a generator polynomial, g (x) [in octal form]:
g(x) = 2413607036565172433223 and a second time for error correction by using a Reed-Solomon error correction code. The system then uses the CRC code to check the data for errors. If errors are found the system uses the error location information supplied by the CRC code and the Reed-Solomon code to correct the errors.
Description
ERROR DETECTION AND CORRECTION S~STEM FOR
LONG BURST ERRORS
FIELD OF INVENTION
This invention relates to error detection and correction of stored computer data in general and more particularly to the correction of long burst errors.
BACKGROUND OF THE INVENTION
The importance of error correction coding of data in digital computer systems has increased greatly as the density of the data recorded on mass storage media, more particularly magnetic tape, has increased. With higher recording densities, a tiny imperfection in the tape can corrupt a large amount of data. In order to avoid losing that data, error correction codes ("ECC's") are employed, as the name implies, to correct the erroneous data.
Before a string of data symbols is recorded on a tape, it is mathematically encoded to form redundancy symbols. The redundancy symbols are then appended to the data string to form code words - data symbols plus redundancy symbols. The :`
.
,, code words are stored on the tape. When the stored data is to be accessed~from the tape, the code words containing the data symbols are retrieved from the tape and mathematically decoded. During decoding any errors in the data are detected and, if possible, corrected through manipulation of the redundancy symbols [For a detailed description of decoding see ' Peterson and Weldon, Error-Correcting Codes, 2d Edition, ~IT
Press, 1972]. Stored digital data can contain multiple independent errors. One of the most effective types of error correction codes used for the correction of these multiple errors is a Reed-Solomon code [For a detailed description of Reed-Solomon codes, see Peterson and Weldon, Error-Correcting Codes]. To correct multiple errors in strings of data symbols, Reed-Solomon codes efficiently and effectively utilize the various mathematical properties of sets of symbols known as Galois Fields, represented "GF(P~", where "P" is a prime number and "q" can be thought of as the number of digits, base P, in each element or symbol in the field. "P"
usually has the value 2 in digital computer applications and, therefore, "q" is the number of bits in each symbol.
Data is typically stored on a tape in a long sequence of symbols. Errors in data stored on a tape often occur in long bursts, that is, many erroneous symbols in a row. Techniques designed for detecting and/or correcting single or multiple independent errors, which do not ordinarily occur in bursts, ~re not well suited for the detection or correction o~ these long burst errors. Thus special error detection and/or correction techniques are typically employed to handle these long burst errors.
Techniques for the detection of long burst errors currently in use detect single or double bursts with each burst containing up to a predetermined maximum number of erroneous symbols. A code which detects bursts containing a larger number of erroneous symbols is considered more powerful than a code which detects bursts containing fewer erroneous symbols. For example, a well known Fire code detects double bursts of up to 22 bits each in 2 kilobytes of data and a well known, more powerful Burton code detects double bursts of up to 26 bits in 2 kilobytes of data. As the density of the data stored on tapes increases, the number of data symbols involved in the long burst errors increases, also. Such long burst errors may occur soon in data stored on disks due to increases in the density of the stored data. Thus a mechanism is desirable for detecting longer error bursts.
Once a long burst error is detected the erroneous symbols involved are corrected, if possible. The faster the errors can be corrected, the faster the data can be made available to a user. Thus the effective data transfer rate increases as the speed of error correction increases. Accordingly, a mechanism is desirable for quickly correcting the detected long burst errors.
SUMMARY OF INVENTION
The invention in its broad form re~ides in a system for encoding data and recording it on a magnetic tape while detecting and correcting errors, including:
A. a first encoder for encoding a stream of d data symbols to generate error detection redundancy symbols, said encoder using a cyclic redundancy code having as a generator 0 polynomial g(x) ~in octal form):
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513)*
or g~x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms, said generated redundancy symbols beiny appended to the data symbols to form a series of CRC code words;
B. a second encoder for encoding using an error correction code a predetermined number, h, of data symbols from h CRC
code words to generate error correction redundancy symbols, .-:
the h data symbols and the associated error correction redundancy symbols forming error correction code words; and !
C. recording means for recording on the tape the h CRC code words and the associated error correction redundancy symbols.
As described herein, error correcting system encodes data which is to be recorded on a magnetic tape twice, once horizontally for error detection and once vertically for error correction. First the system encodes a stream of d multi-bit data symbols to form c redundancy symbols using a powerful cyclic redundancy check (CRC) code. The CRC code is characterized by the following generator polynomial g(x): [in octal vector form]
g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms of the polynomial. The CRC code is capable of detecting a single long burst error containing up to 64 bits and double long burst errors containing up to 28 bits each.
The error correction system concatenates the d data symbols and the c CRC redundancy symbols to form a CRC code word. The encoder then forms a group of CRC code words such .
. . ,~
that each of the corresponding symbols in the CRC code words in a group are vertically aligned. The aligned symbols form a series of one-symbol wide columns.
The system further encodes, using a Reed-Solomon error correction code, each column containing data symbols. It does not encode the columns containing the CRC redundancy symbols.
The encoding generates, for each column, a predetermined number, n, of ECC redundancy symbols. The encoder then includes these ECC redundancy symbols in the group such that the ECC redundancy symbols are aligned with the data symbols to which they correspond. Thus the encoder forms a series of columns each of which has b symbols, b-n data symbols and n ECC redundancy symbols.
The ECC redundancy symbols corresponding to the data symbols in a group are themselves encoded using the CRC code.
The various ECC redundancy symbols and the corresponding CRC
redundancy symbols form n ECC code words. Thus the nth ECC
code word contains (i) the nth ECC symbol corresponding to each column of data symbols, and (ii~ the related c CRC symbols. A
complete group of code words contains b-n data code words and n ECC code words. The various code words are recorded on the tape as a group, i.e., ln multiple tracks, or bit-serially, i.e., in a single track. Accordingly, the group may require re-assembly when the data is retrieved.
,: .
" .
, ~
~ hen data is retrieved from the tape it is first assembled in groups, if necessary, and then each group is decoded, code word- by- code word, using for each code word the CRC redundancy bits contained therein. The CRC code detects with certainty a single error of 64 bits or less and/or double long burst errors, each of which may be up to 28 bits long. The code also detects up to seven random errors in each CRC code word. If errors are detected in the data CRC
code words the system uses the Reed-Solomon code ECC
redundancy symbols and one of several decoding algorithms to correct them column by column. Error correction is in most cases simplified because the CRC decoding points to the vertical locations of the errors.
The particular Reed-Solomon code used to encode the data encodes it such that single errors with known locations can be quickly and easily corrected. If the CRC code detects errors in only one code word, code word t, there is at most one erroneous data symbol per column, i.e., the symbol in column location t. To calculate the error value associated with the symbol in location t in a particular column the system adds together (XOR's) the symbols in the column. The sum is the error value, and thus, the system corrects the error by adding together the symbol and the associated error value. If the .
J
r symbol in location t is not erroneous the calculated error value is ZERO.
When the CRC decoding indicates errors in a particular code word, the system treàts as erroneous every data symbol in that code word. Thus the system calculates an error value for each column.
If the CRC decoding indicates that multiple CRC code words contain errors and one or more of the CRC code words is a data code word, the system first determines which of the columns contain errors. The system then engages in error correction for these particular columns.
To determine whether a particular column contains any errors, the system encodes the column data symbols using the Reed-Solomon code and thus generates new ECC redundancy symbols. It then adds (XOR's) these symbols to the corresponding ECC redundancy symbols retrieved from the tape to formulate remainder symbols REMo, REMl,..., REM~. If the remainder symbols are all ZEROS the two sets of ECC symbols are the same, and thus, the data symbols in the column are error-free. If one or more of the remainder symbols are non~zero the two sets of ECC symbols are not the same and the column contains one or more erroneous symbols. The system then performs error correction for each column containing errors.
If the CRC points out as erroneous between one and n CRC
code words the system simultaneously solves a set of n 5 equations based on the remainder symbols, REMo, REMl, REM~, and the specific vertical locations of the erroneous code words. The system uses matrices manipulated by Galois Field operations to solve these equations. The solutions are the error values corresponding to the column symbols in the locations corresponding to the erroneous code words.
If the CRC code indicates that n+1 or more code word contain errors, the system corrects the errors by assuming there are n or fewer actual errors per column and using one of - several correction procedures. It then uses the CRC code to determine whether all the errors in the data code words have been corrected. If the CRC detects further errors either the system repeats the entire error correction procedure until the CRC code indicates that all the errors are corrected, or the system labels the data uncorrectable.
One procedure for correcting the n+1 errors is to assume that there is only one erronaous symbol, with an unknown location, in any particular column. The errors can then be corrected using a mechanism for correcting single errors with ~` g unknown locations. The system does not know in which of the many possible error locations the one erroneous symbol resides, and thus, the location of the error is considered unknown.
First, the system examines the remainder symbols RE~or REMl...... .......,REM~ for a particular column. If all of the remainder symbols are ZEROS, it indicates that the column data is error-free and no error correction is required. If only one of the remainder symbols, for example, REMi, is non-zero, it indicates that the symbol in column location i is erroneous. The error value corresponding to this symbol is REMi and thus the error is corrected by adding REMi to that symbol.
If more than two remainder symbols are non-zero the system uses the remainder symbols to initialize n shift registers of a cyclic shifter. The shifter cyclically shifts the symbols in a column backwards by dividing the remainder . ~ .
symbols by the reciprocal of the Reed-Solomon generator polynomial used to encode the data. Due to the cyclic property of the Reed-Solomon code the error values required to correct the errors in the column are not affected.
The shifting mechanism shifts the column symbols until the contents of registers 1, 2, ..., (n-1) are all ZEROS and the resulting symbol in the nth register is non-zero. If the .
.
shifter shiftq the ~ymbolq a total of L times the symbol in column location n+L is erroneous and the nth register contains the corresponding error value. If after the shifter is shifted b times, that is, one shift for each symbol location in a column, an error locàtion is not found error correction for that particular column fails. The system then goes on to correct the errors in the next column.
Once error correction has been attempted for each of the columns the system again examines the group of code words for errors using the CRC code. If there are further errors the system corrects them, if possible, by using an error correction procedure designed for the number of remaining errors.
BRIEF DESCRIPTION OF DRAWINGS
A more complete understanding of the invention may be had by referring to the following description of preferred embodiments, given by way of example only, to be understood in conjunction with the accompanying drawings, in which:
.
Figure 1 is a view of a magnetic tape recorded in - accordance with a preferred embodiment;
Figure 2 depicts a Reed-Solomon code encoder;
Figure 3 illustrates a matrix M used in error correction Figure 4 depicts a cyclic shifter; and Figure 5 is a view of various error configurations.
DETAILED DESCRIPTION
The mathematical operations involving Reed-Solomon code ECC redundancy symbols are performed using Galois Field addition, multiplication and/or division. Galois Field addition is performed by exclusive Or'ing (XOR) the code symbols bit-by-bit, Galois Field multiplication, represented by the symbol "*", is modulo p(x), where p(x) is a primitive polynomial which characterizes the Galois Field.
Figure 1 illustrates a group 6 of b code words 4, where b = 20 in the preferred embodiment. The code words 4, oriented horizontally, contain d data or information symbols and c associated redundancy bits. The generation of the c redundancy bits is discussed below. The code words 4 are vertically aligned such that the corresponding symbols in each of the code words 4 form a series of ona-symbol wide columns.
- . , ., ~ ,, ~ .
' The c bits in each code word 4 are generated by encoding the code word data or information symbols using a cyclic redundancy check (CRC) code which is well suited to detecting long burst errors. The c CRC redundancy bits are appended to the associated data or information symbols to form a CRC code word. The code words numbered 5 - 20 in Figure 1 contain data symbols and c CRC redundancy bits. The code word numbered 1 -4 contain d information symbols and c CRC redundancy bits. In the preferred embodiment d = 2 kilobytes and c = 64 bits. The CRC code encoding is discussed in more detail below.
The d information symbols in each of the code words 1 - 4 are Reed-Solomon error correction code (ECC) redundancy symbols. The ECC redundancy symbols are generated by encoding each of the columns containing data symbols; the columns containing CRC bits are not encoded. The ECC redundancy symbols recorded in a particular column, that is, the symbols in code words 0 - 4 of a column containing data symbols, correspond to the data symbols in that column. Thus each of these columns forms a code word of the Reed-Solomon code.
The d ECC redundancy symbols in each of the code words 1 - 4 are then encoded using the CRC code. The c CRC redundancy bits generated by this encoding are appended to the ECC
redundancy symbols to form ECC code words, that is, code words in the CRC code. Accordingly, errors in the stored ECC
~' redundancy symbols can be detected. Each column containing CRC bits forms a Reed-Solomon code word, also, because both the Reed Solomon code and the CRC code are linear codes.
To generate the CRC bits, an encoder (not shown) encodes the appropriate data or ECC redundancy symbols using a cyclic redundancy check code characterized by the following degree sixty-four binary generator polynomial g(x) presented as an octal vector:
g(x) = (3)*(3)*(100003~*(102043)*(110013)*(1560513) or g(x) - 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms. The CRC
encoder, which is of conventional design, encodes in a conventional manner the data or information symbols using this CRC generator polynomial. The data may also be encoded using the reciprocal of g(x), g*(x) (in octal form):
g*(x) - (3)*(3)*(140001)*(11420~ 0011)*(1512073) or g*(x) = 3113305362565703417205.
The length of the CRC code is shortened fro~ its natural length of over 8 billion bits to slightly over 17,000 bits.
The shortened code is a powerful code which is capable of .
detecting with certainty in the 17,000 bits of data a single burst error of up to 64 bits in length. The code further detects with certainty double burst errors of up to 28 bits each. The Hamming distance of the code is at least eight, thus the code detects up to seven independent errors in any location, i.e., seven random erron~ous bits as opposed to burst errors.
The shortened code actually detects 28-bit double burst errors in 17,907 bits of data. Thus a block may contain 2 kilobytes of actual data and up to 1523 bits of appended control information.
As set forth above, the columns of data in a group are encoded for error correction purposes using a Reed-Solomon code. The code encodes the data over a Galois Field GF(28) which is characterized by a primitive polynomial p(x) represented by the octal vector p(x) = 455, which in binary form is:
Xa + X5 + X3 + X2 + 1 The particular Reed-Solomon code used to encode the data columns is specified by the generator polynomial G(x):
G(x) = (x ~ ~) * (x + ~l) * (x + ~Z) * (x + ~3) = X4 + a75x3 + oC249X2 + ~X78X + ~X,6 where a is the primitive element of the Galois Field. The data is encoded over Galois Field GF (28) in the preferred . ~ -.
embodiment because the elements of the field are 8-bit symbols, or bytes, and most computer hardware is set up to handle data organized in bytes.
Figure 2 depicts an encoder 10 for encoding the column data symbols. The encoder 10 includes a set of 8-bit registers 16, 18, 20 and 22 that are cleared prior to the encoding operation. The contents of the registers are applied to exclusive OR circuits (XOR's) 11, 12, 13 and 14 respectively. The input to the encoder is applied to the XOR
14 in byte serial, bit parallel fashion over eight input lines.
During the input operation a switch 32 is in position so that the output of the XOR 14 is applied to Galois Field multipliers 24, 26, 28 and 30. The outputs of multipliers 26 28 and 30 are applied to XOR's 11, 12 and 13, respectively, and the output of multiplier 24 is applied to the register 16.
The Galois Field mul-tipliers 24-30 multiply their inputs by the factors a6, a78, a249 and a75 respecti Accordingly, as each input byte is applied to the encoder 10, it is summed bit-by-bit (modulo 2) with the contents of the register 22. The sum is then simultaneously multiplied (modulo p(x)) by Galois Field multipliers 24-30 which contain the coefficients of all but the highest power of x in G(x).
, The respective products of the byte and the coefficients are then added (modulo 2) to the contents of the registers 16-22.
After all of the data bytes in a particular column have been applied to the encoder 10, the encoder switch 32 is placed in position 2 and the contents of registers 16-22 are unloaded as the ECC redundancy symbols, R3, R2, Rl, Ro~
respectively, by shifting their contents to the left. It will be noted that during the shifting operation with the switch 32 in position 2, the vertical inputs of the XOR's pass the contents of the registers 22-26 without modifying them.
Encoders of this type are well known in the art.
The encoding system formulates a complete group by aligning beneath the appropriate column data symbols, in code words 1-4, the ECC redundancy symbols. Thus code word 4 contains the R3 redundancy symbols, code word 3 contains R2 redundancy symbols, et cetera. A complete column, that is, the corresponding data symbols and the associated ECC
redundancy symbols in code words 1-20, form a Reed-Solomon code word.
A processor (not shown) writes the code words to the tape. It will be understood by those skilled in the art that the processor may write the code words in any format, for : .
~ , :
" ~
.. . .
example, it may write them as a string of bits with one code word following another. Thereafter, when the code words are retrieved from the tape, the processor re-assembles the group 6.
When a user requests data from the tape the processor retrieves from the tape the code words 4 (Figure 1) containing the requested data symbols and any other code words require to re-assemble the group 6. An error correction processor (not shown) then determines if there are errors in any of the code words 4 in the group 6 by decoding each of them using the component CRC bits. If the CRC code does not detect any errors the error correction processor forwards the require data to the requesting user. If the CRC code detects errors it points out the erroneous code words, and thus, the locations of the errors. Accordingly, the error correction processor performs error correction as appropriate, to correct the symbols in these code words.
If all the errors are located in code words 1-4, that is, the ECC code words, the data is assumed error-free and the error correction processor immediately forwards the data to the requesting user. However, if errors are detected in the code words containing data symbols, the error correction processor corrects the errors (if possible) using the ECC
redundancy symbols. To correct the errors the error PD~9~0117 correction processor selects an appropriate error correction routine based on the number and location of the detected errors.
If the CRC bits indicate that only one CRC code word is S erroneous, for e~ample, code word bi, the processor corrects the erroneous data symbols in that code word using a procedure that is specifically designed for correcting single errors with known locations. For a given column containing data symbols the column either is error-free or it contains one erroneous symbol, the symbol in the column location corresponding to code word bi.
The error correction processor does not, at this point, determine if a particular column contains an erroneous symbol.
Instead, it attempts to correct the symbol in location bi in every column. Thus, for every column, the error correction processor generates an error value corresponding to the symbol in location bi. It then corrects the symbol by combining it with, i.e., adding it to, the generated error value. The generated error value for a symbol which is not in error is ZERO. Thus adding this error value to the symbol does not change it. Alternatively, the error correction processor could calculate the error values only for columns containing data symbols and ignore the columns containing CRC symbols.
; .
:. . ~, ' `': ~
~: `' : ' To generate an error value :Eor a potentially erroneous symbol in location bi the processor adds together, using Galois Field addition, all the symbols ln a column. This sum is the error value associated with the data symbol in location bi.
Thus the error correction processor adds the sum to the erroneous symbol to correct it. The same result is achieved by adding together all the symbols in the column except the symbol in location bi. This sum is the corrected data symbol.
Thus this sum replaces the symbol in location bi in the data transmitted to the user.
This single-error correction procedure is possible only with Reed-Solomon codes which have (X ~ ~) as a factor of their generator polynomials. Having this term as a factor means that the generator polynomial, which is a degree four polynomial, may have more than two distinct terms with non-~ero coefficients. Thus an associated encoder 10 will most likely include more than two distinct Galois Field multipliers. This may result in an encoder which is not the simplest one for encoding data for single-error correction.
However, the ease of correcting errors contained in a single CRC code word using such a Reed-Solomon code outweighs the increased complexity of the encoder.
If the CRC bits indicate that more that one code word contains erroneous symbols, for example, two code words, the processor corrects the errors using an error correction procedure designed to correct between one and four errors with known locations. First, the error :Location processor determines which columns contain errors by calculating a remainder for each column. The remainder for a particular column is calculated by~ encoding the retrieved column data symbols using encoder 10 to generate new ECC redundancy symbols corresponding to the retrieved data symbols; and (ii) comparing these new ECC redundancy symbols with the retrieved ECC redundancy symbols.
The two sets of ECC symbols are compared by XOR'ing bit-by-bit the corresponding symbols. The resulting symbols are the remainder symbols REMo, REMl, REM2, and REM3. If the remainder symbols are all ZEROS the two sets of ECC symbols are the same and the data column is assumed error-free. If one or more of the remainder symbols are non-zero the column contains errors.
To calculate the error values for the columns containing errors, that is, the columns with non-zero remainders, the error correction processor solves for each of these columns set of four simultaneous equations. Each of these equations involves a different column remainder symbol.
In order to formulate the equations the processor generates a matrix M with rows corresponding to X~ modulo G(x), where e is the symbol location associated with a given matrix row and G(x) is the Reed-Solomon code generator polynomial.
The rows of matrix M can be thought of as vectors with components corresponding to coefficients of (X3, XZ, Xl, X).
Thus, for example, the matrix M row corresponding to X3 iS (1~
0, 0, 0) while the matrix row corresponding to X4 iS (X7s, X249, x7a~ x6) .
The error correction processor may generate the rows of the matrix M using encoder 10 as follows: First, the processor initializes encoder registers 16, 18, 20 and 22 to the values 0,0,0,1, respectively. The registers then contain the elements of the matrix row corresponding to X. The processor next cycles, i.e., clocks, the encoder 10 once with switch 32 in position 1 to generate in the registers the elements of the second row (i.e. x1) of the matrix M. The processor clocks the encoder a total of 19 times with switch 32 in position 1 and uses the contents of the registers 16, 18, 20 and 22 generated by each encoder cycle as the rows X
through X19 of the matrix M. This matrix is then stored for use in the future. Figure 3 depicts the complete matrix M in both oc.tal and binary forms. It will be understood by those skilled in the art that the matrix is generated only once and used thereafter for all error correction.
The error value equations for a particular column with detected errors, for example, in blocks 4 and 6, are:
REM3 = Y (e4) * 17 + Y (e6) * 134 REM2 = Y(e4) * 66 + Y (e6) * 7 REMl = Y (e4) * 170 + Y (e6) * 277 REMo = Y (e4) * 100 + Y (e6) * 345 where Y(e4) and Y(e6) are the error values for the column symbols in column locations 4 and 6, respectively, and the coefficients of the Y(e4) and Y(e6) terms are the elements of matrix M rows X4 and X6, respectively, such that the equation for REM3 has as coefficients the fourth column elements of matrix M rows X4 and X6, the equation for REM2 has as coefficients the third column elements of the rows, et cetera.
The error correction processor could now solve any two of these equations to determine the error values. However, the processor uses all four equations and the procedure set forth below to determine the error values for 2, 3 or 4 detected errors.
Next to solve the equations the error correction processor formulates for each column containing erxors a matrix M'. The matrix M' has rows corresponding to each of the code words in which the CRC detects errors and a bottom row containing as elements the column remainder symbols REMo REMl, REM2 and REM3. In the example there are two code words : ~, ::~
PD~9-0117 containing errors, namely, code words 4 and 6, and thus rows 1 and 2 of matrix M' are the same as rows X~ and x6 of matrix M.
The third row of matrix M' is the column remainder REMo, REMl, REM2 and REM3. In this particular example the column remainder is an octal vector 175, 47, 211, 123. Thus matrix M' in octal form is:
row X4 ¦ 17 66 170 100 ¦
row x6 ¦ 134 7 277 345 ¦
remainder ¦ 123 211 47 175 ¦
The error correction processor next manipulates the matrix M' using column operations to reduce it to a matrix M"
which has as elements of the first two columns of the first two rows a 2x2~identity matrix and has as the remaining elements of rows 1 and 2 only ZEROS. In this example, matrix M" is:
~ ' ~ J
If the CRC indicates 3 errors, the corresponding matrix M' has 4 rows and it reduces to a matrix M" containing a 3x3 identity matrix in the first three columns of the first thre rows and has only ZEROS as the remaining elements of rows 1-3.
The two column operations performed in reducing matrix M' are: (1) a*[column i] + b*[column j] to replace column i, where "a" is non-zero element of GF(28) and "b" is any element of GF(28), and (2) interchanging two columns.
When the matrix M" is in this reduced form the last row of the matrix contains the error values associated with the data or ECC redundancy symbols in the erroneous code words.
Thus the matrix element in the first column, 355, is the error value associated with the ECC redundancy symbol in code word 4 and the element in the second column, 135, is the error value associated with the data symbol in code word 6. The elements in the two matrix columns which do not correspond to errox locations, columns 3 and 4, should be ZEROS. Otherwise, an uncorrectable error condition exists and the calculated error values are rejected.
If only one of the detected errors is in a data code word and each of the other detected errors are in the ECC code words the matrix M' corresponding to the detected errors may be used directly to calculate the error value of the erroneous PD~9-0117 data symbol. Thus the matrix need not be reduced. If, for example, the errors are detected in ECC code words 0 and 3 an data code word 5, and the remainder symbols for a given column are 321, 335, 360 and 375, the matrix M' is:
1 0 ~ I I
The matrix M' has in its first two rows, that is, the rows corresponding to the ECC code words, only ONES and ZEROS.
Thus the matrix is already in a reduced form, though not in the reduced form discussed above. The error value corresponding to the code word 5 data symbol is calculated using this matrix M' by dividing the element in the last row of either matrix column 1 or column 2, that is, either of the columns having ZEROS in rows 1 and 2, by the element in row 3 of the same matrix column. Thus the error value for the code word 5 data symbol, in octal form, is:
360/127 = 335/322 = 115 An alternative error correction procedure for correcting double data-symbol errors with known locations utilizes the ' ~Y-U117 cyclic property of Reed-Solomon codes to "shift" one of the errors into an ECC symbol location. Thus the symbols are cyclically shifted backwards by some number of locations without altering the error values associated with the symbols.
After the shift the error correction processor calculates the error value associated with the symbol which is still in a data symbol location using the procedure set forth above for a single data-symbol error. The processor then calculates the error value corresponding to the remaining erroneous symbol using the single error with known location procedure set forth above, that is, by adding together all the symbols in the column including the corrected data symbol.
Referring now to Figure 4, in order to cyclically shift the symbols in a column backwards the error correction processor uses a cyclic shifter 40, which is a linear shift register set up to multiply the symbols in registers 44, 46, 48, 50 by the reciprocal of G(x). Thus Galois Field multipliers 52-58 contain the coefficients of the terms of the reciprocal of G(x). For example, multiplier 52 multiplies terms by a78/a6, which is the coefficient of the X~ term of G(x) divided by the coefficient of the X term of G(x). The multiplicands of multipliers 54-58 are the X2, X~, and X4 coefficients of G(x) divided by the Xcoefficient (the . :,:
' " ' ' ~ . -, ' ' , PD~9-0117 coefficient of the X4 term is 1, which in GF(28) ls the same as o~255 ) To shift a particular column the error correction processor loads the column remainder symbols REMo, REMl, REM2 and REM3 into registers 44, 46, 48, 50, respectively. It then shifts the cyclic shifter 40 L times until the location of only one of the errors corresponds to a data symbol and the location of the other error corresponds to an ECC symbol. Thus the error locations are shifted from locations e1 and e2 to locations e1-L and e2-L.
The error correction processor then uses the symbols in the registers 44, 46, 48, 50 as the column remainder symbols in formulating matrix M'. The first two rows of matrix M' correspond to the column locations e1-L and e2-L and the third row corresponds to the shifted remainder symbols. The processor then calculates the error value for the data symbol by dividing the appropriate elements of rows 2 and 3 of matrix M'. The processor corrects the symbol by adding to it the calculated error value.
To calculate the error value for the remaining error, that is, the symbol in location e2, the processor adds together the column symbols, including the corrected symbol in location el. The processor then corrects the symbol in location e2 by adding to it the calculated error value.
If more than four errors are detected the error correction processor corrects them by first correcting a predetermined number of errors, f, and repeating the erro correction procedure for any remaining errors. Figure 5 depicts several configurations of five or more erroneous code words. The "X's" denote erroneous symbols. Configuration 1, for example, depicts code words with errors occurring such that each column contains at most one error. Thus the errors can be corrected using single-error correction techniques. Similarly configuration 2 has at most two errors per column and the errors can be corrected using double-error correction techniques, and so forth. The system does not have any mechanism which is capable of determining exactly how the errors in the five or more erroneous code words are distributed, however. Therefore special error correct procedures must be used the correct the errors.
One such error correction procedure calculates the error values associated with a particular number of errors, v, with known locations and a single error with an unknown location.
Thus the error correction processor selects v detected error locations and uses the matrix M rows associated wïth these locations, along with the column remainder symbols, to solve for the single unknown error location. ~fter finding the PD~ ~-011 7 unknown location the processor corrects the v + 1 errors using the procedure for v + 1 errors with known locations. The value of v may be either ONE or TWO. The location of the single error is unknown because the system does not know the S exact location of the erronèous symbol in a particular column, it knows only that the error is ln one of the remaining locations indicated by the CRC code.
The remainder associated with a column is a linear combination of all the symbols in a column, including the erroneous symbols. Thus equations, for example, for two errors with known locations and a single error with an unknown location are:
REM3 = Y (EV1) *X3Vl + Y (EV2) *X3V2 + Y (Eu) *X3U
REM2 = Y (EV1) *X2Vl + Y (EV2) *X2V2 + Y (Eu) *X2U
REMl = Y (EV1) *XlVl + Y (EV2) *X1V2 + Y (Eu) *XlU
REMo = Y (EV1) *XOV1 + Y (EV2) *X0v2 + Y (Eu) *xOU
where Y(El) is the error value, a scalar, associated with the error in location i and Xji is the jth element of row i of Matrix M (Figure 2). These equations can be re-written:
X3U = ~l/Y (Eu) ] *REM3 + [Y (EV1) /Y (Eu) ] *X3Vl + [Y (EV2) /Y (Eu) ] *X3V2 X2U = [l/Y (Eu) ] *REM2 + [Y (EV1) /Y (Eu) ] *X2Vl + [Y (EV2) /Y (Eu) ] *X2V2 Xl = [l/Y (Eu) ] *REMl + [Y (3V1) /Y (Eu) ] *Xl 1 + [Y (EV2) /Y (Eu) ] *X1V2 X0 = [l/Y (Eu) ] *REMo + [Y (EV1) /Y (Eu) ] *XOV1 + [Y (EV2) /Y (Eu) ] *X0v2 ~- :
: :
SY-Ull /
or more generally:
XU = AR*REMC ~ Ai*XVl + Aj*XV2 where the Ab's are scalar coefficients. Thus the vec-tor XU
corresponding to the unknown error location is a llnear combination of (1) the column remainder symbol vector REMC, and (2) the two rows or vectors of matrix M corresponding to -the two known error locations, namely, XVl and XV2.
To solve for the unknown error location, that is, determine which row or vector of matrix M can be expressed as a linear combination of the remainder symbols and the error location vectors, the error correction processor formulates a matrix M' following the procedures set forth above. As an example, let v=2. The error correction processor formulates a 3x4 matrix M' corresponding to the two selected error locations and the column remainder symbols. The error correction processor then reduces matrix ~' using only row operations to a matrix M" which includes within it a (v~l~x(v~1) identity matrix. By using only row operations the error correction processor preserves the relationships between matrix columns and thus preserves the relationships between the various vectors.
In this example the reduced matrix M" includes a 3x3 identity matrix. It may take any of the following forms:
` ' .:
PD89~0117 la3 0 0 1 ¦ ¦o a2 1¦ lo o a~ o o 1 aO¦
¦b3 0 1 0 ¦orlO b21 O¦or¦O 1 bl O¦orlo 1 o bo¦
1~3 1 0 I 11 C2 ol 11 o c, ol11 o o col ~ ~ ~ ` ~ ~ _ ~ J
The two row operations performed in reducing matrix M' to matrix M" are: (1) a*[row i] + ~*[row j] to replace row i, where "a" is any non-zero element of GF(28) and "b" is any element of GF(28), and (2) interchanging two rows.
Next, the error correction processor determines if a row Xi of the large matrix M (F.igure 2) can be expressed as a linear combination of the rows of the matrix M" by first creating a new matrix N which is the matrix M" with the vector Xj appended as a new last row. Thus the last row of matrix N
has elements S j3, S j2, S jl and sjo. The error correction processor then rotates (v=1) or permutates (v=2) the columns of matrix N until the included identity matrix is located in the upper-right corner. Thus matrix N is of the form:
~ 1 ¦a3'0 0 1 ¦
Ib3'0 1 0 1 IC3~1 0 0 I S j3rS j2r S jlr S j ~ 32 . .
,. : .: .
..
:. , : ~, , .
The error correction processor then tests if the row Xj is a linear combination of the matrix M" rows by substituting the appropriate elements of the permutated matrix N into the following equation:
sj3' = a3'*sjo' + b3'*sjl' + C3~ *S j2' If the equation is satisfied Xi is the unknown error location. Otherwise, the error location processor repeats the procedure for the matrix M rows Xj+l, Xi~2... Xi~l9 in which errors are detected until the location of the error is found or all the possible error locations have been tried.
If the error correction processor is attempting to find the unknown error location using only one known error location, i.e., v=l, the equation into which the elements of the rotated matrix N are substituted is one of the following:
sj2' = a2~*sj0~ + b2'*sjl' sj3' = a2~*sj0~ + b3'*sjl' In this case there is a redundant equation and the calculated error location can be checked by substituting the appropriate matrix N elements into the second equation.
After error correction has been tried for all of the group 6 (Figure 1) columns containing data symbols the error correction processor again examines blocks 5 - 20 for any remaining errors using the CRC code. If all the errors have been corrected the processor transfers the error-free data to the requesting user. If all the errors have not been corrected the processor corrects them, if possible, using an error correction procedure which is appropriate for the number and location of the remaining errors.
Error correction for a particular column fails if none of the rows of matrix M can be expressed as a linear combination of the rows of the associated matrix M". The system may then ignore this column and go on to correct the errors in the other columns. Thereafter, when all the columns which are correctable have been corrected and the error correction processor checks the code words for remaining errors using the CRC bits, error correction continues only if the number of code words now containing errors is smaller than the number containing errors before error correction. Otherwise, the processor labels the data uncorrectable.
An alternate method of correcting five or more errors has the error correction processor first correcting the columns containing single errors using a procedure for correcting .
.~ . . i .
single errors with unknown locations and then correcting any remaining errors using a procedure designed for the number and the locations of the errors. Presumably, by correcting the columns with single errors, the number of code words containing the remaining errors is reduced to four or below.
In this way, one of the procedures for correc-ting errors with known locations can be used to complete the error correction.
To correct a single error in a particular column the system first examines the column remainder symbols. If all the remainder symbols are ZEROS there are no errors in the column and thus no error correction is required. If there is only one non- zero remainder symbol, for example, REM3, the error is in column location 3 and the remainder symbol is the associated error value. If only two of the remainder symbols ~EMl and REM3 are non-zero it means that the particular column contains two errors instead of one. The -two errors are in the locations corresponding to the remainder symbols, namely, column locations 1 and 3, and the remainder symbols are the corresponding error values.
If more than two remainder symbols are non-zero the processor loads the remainder symbols in the registers 44, 46, 48, and 50 of cyclic shifter 40 (Figure 4) and clocks the shifter until the symbol in register 50 is non-zero and the symbols in registers 44, 46, 48 are all ZEROS. If the shifter ,~
PD~9-0117 is clocked K times to produce this result the column symbol in location K+3 is~in error and the symbol in register 50 is the error value associated with that symbol. If, after b cycles, an error location not found, it indicates that the column contains more than one error. The processor then ignores this column and goes on to the next one. The processor performs this single error correction procedure for each column and then tests the group 6 (Figure 1) for any remaining errors using the CRC code. If errors remain in the code words containing data symbols the system uses another procedure, as appropriate, to correct them.
The cyclic shifter 40 may be used to correct double errors with unknown locations, also. However, before the shifter is loaded the column remainder symbols are examined as set forth above to determine if only one or two remainder symbols are non- zero. If this is the case, the errors are located in the locations corresponding to the non-zero remainder symbols and the remainder symbols are the error values corresponding to the symbols in these locations.
If more than two remainder symbols are non-zero, the error correction processor loads the remainder symbols into registers 44, 46, 48, 50 of the cyclic shifter 40. The error correction processor then examines the contents of the registers to determine if the symbols in registers 46, 48, and .
50 are proportional to the last three elements of a row Xj of the large matrix M (Figure 2), namely, elements s j3, s j2, s jl.
Thus the error correction processor determines if:
REM3 = Uj* s j3 REM2 = Uj* s j2 REMl = Uj* s If these conditions are satisfied for a particular row X
the errors are at location 0 and location O+j, and the corresponding error values are Uj*s jo + REMo and Uj, respectively. If these conditions are not satisfied for any of the matrix M rows the error correction processor clocks the cyclic shifter 40 to produce in the registers 44, 46, 48, 50 new remainder symbols REMo', REMl', REM2' and REM3'. The processor then determines if these new remainder symbols satisfy the conditions set forth above. The procedure is repeated 19 times, once for each potential error location, if necessary.
It will be understood by those skilled in the art that it is unlikely for a group 6 (Figure 1) to contain more than 4 erroneous blocks in a well designed system. Thus error correction procedures for 5 or more code words are rarely used. Accordingly, a system which includes procedures for correcting up to four code words with errors and labels groups containing 5 or more code words with errors as uncorrectable does not degrade significantly the system performance.
The error correction codes and the error correction procedures set forth above reduce the number of undetectable and/or uncorrectable errors in data recorded on magnetic tape to a level which is comparable to the level achieved with data recorded on magnetic disks. The powerful CRC code used to encode the data for error detection detects longer burst errors than prior art CRC codes. Thus these errors, which are one of the most common types of errors for tapes, are more accurately detected.
Combining the powerful detection capabilities of the CRC
code with a Reed-Solomon code which quickly corrects errors with known locations results in a fast error correction scheme. Thus the transfer rate of the data is high. The particular Reed- Solomon code is particularly well suited for correcting single (column) errors with known locations. The single error correction is so fast that the error correction processor attempts to correct each column rather than determine first which columns contain errors. Accordingly, the length of an error, that is, the burst length, does not affect the time it takes the processor to correct the errors, and long burst errors are corrected as quickly as shorter errors.
.
The foregoing description has been limited to various embodiments of this invention. It will be apparent, however, that further variations and modifications may be made to the invention, with the attainment of some or all of the advantages of the invention. Therefore, it is the object of the appended claims to cover all such variations and modifications as come within the scope of the invention.
.
LONG BURST ERRORS
FIELD OF INVENTION
This invention relates to error detection and correction of stored computer data in general and more particularly to the correction of long burst errors.
BACKGROUND OF THE INVENTION
The importance of error correction coding of data in digital computer systems has increased greatly as the density of the data recorded on mass storage media, more particularly magnetic tape, has increased. With higher recording densities, a tiny imperfection in the tape can corrupt a large amount of data. In order to avoid losing that data, error correction codes ("ECC's") are employed, as the name implies, to correct the erroneous data.
Before a string of data symbols is recorded on a tape, it is mathematically encoded to form redundancy symbols. The redundancy symbols are then appended to the data string to form code words - data symbols plus redundancy symbols. The :`
.
,, code words are stored on the tape. When the stored data is to be accessed~from the tape, the code words containing the data symbols are retrieved from the tape and mathematically decoded. During decoding any errors in the data are detected and, if possible, corrected through manipulation of the redundancy symbols [For a detailed description of decoding see ' Peterson and Weldon, Error-Correcting Codes, 2d Edition, ~IT
Press, 1972]. Stored digital data can contain multiple independent errors. One of the most effective types of error correction codes used for the correction of these multiple errors is a Reed-Solomon code [For a detailed description of Reed-Solomon codes, see Peterson and Weldon, Error-Correcting Codes]. To correct multiple errors in strings of data symbols, Reed-Solomon codes efficiently and effectively utilize the various mathematical properties of sets of symbols known as Galois Fields, represented "GF(P~", where "P" is a prime number and "q" can be thought of as the number of digits, base P, in each element or symbol in the field. "P"
usually has the value 2 in digital computer applications and, therefore, "q" is the number of bits in each symbol.
Data is typically stored on a tape in a long sequence of symbols. Errors in data stored on a tape often occur in long bursts, that is, many erroneous symbols in a row. Techniques designed for detecting and/or correcting single or multiple independent errors, which do not ordinarily occur in bursts, ~re not well suited for the detection or correction o~ these long burst errors. Thus special error detection and/or correction techniques are typically employed to handle these long burst errors.
Techniques for the detection of long burst errors currently in use detect single or double bursts with each burst containing up to a predetermined maximum number of erroneous symbols. A code which detects bursts containing a larger number of erroneous symbols is considered more powerful than a code which detects bursts containing fewer erroneous symbols. For example, a well known Fire code detects double bursts of up to 22 bits each in 2 kilobytes of data and a well known, more powerful Burton code detects double bursts of up to 26 bits in 2 kilobytes of data. As the density of the data stored on tapes increases, the number of data symbols involved in the long burst errors increases, also. Such long burst errors may occur soon in data stored on disks due to increases in the density of the stored data. Thus a mechanism is desirable for detecting longer error bursts.
Once a long burst error is detected the erroneous symbols involved are corrected, if possible. The faster the errors can be corrected, the faster the data can be made available to a user. Thus the effective data transfer rate increases as the speed of error correction increases. Accordingly, a mechanism is desirable for quickly correcting the detected long burst errors.
SUMMARY OF INVENTION
The invention in its broad form re~ides in a system for encoding data and recording it on a magnetic tape while detecting and correcting errors, including:
A. a first encoder for encoding a stream of d data symbols to generate error detection redundancy symbols, said encoder using a cyclic redundancy code having as a generator 0 polynomial g(x) ~in octal form):
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513)*
or g~x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms, said generated redundancy symbols beiny appended to the data symbols to form a series of CRC code words;
B. a second encoder for encoding using an error correction code a predetermined number, h, of data symbols from h CRC
code words to generate error correction redundancy symbols, .-:
the h data symbols and the associated error correction redundancy symbols forming error correction code words; and !
C. recording means for recording on the tape the h CRC code words and the associated error correction redundancy symbols.
As described herein, error correcting system encodes data which is to be recorded on a magnetic tape twice, once horizontally for error detection and once vertically for error correction. First the system encodes a stream of d multi-bit data symbols to form c redundancy symbols using a powerful cyclic redundancy check (CRC) code. The CRC code is characterized by the following generator polynomial g(x): [in octal vector form]
g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms of the polynomial. The CRC code is capable of detecting a single long burst error containing up to 64 bits and double long burst errors containing up to 28 bits each.
The error correction system concatenates the d data symbols and the c CRC redundancy symbols to form a CRC code word. The encoder then forms a group of CRC code words such .
. . ,~
that each of the corresponding symbols in the CRC code words in a group are vertically aligned. The aligned symbols form a series of one-symbol wide columns.
The system further encodes, using a Reed-Solomon error correction code, each column containing data symbols. It does not encode the columns containing the CRC redundancy symbols.
The encoding generates, for each column, a predetermined number, n, of ECC redundancy symbols. The encoder then includes these ECC redundancy symbols in the group such that the ECC redundancy symbols are aligned with the data symbols to which they correspond. Thus the encoder forms a series of columns each of which has b symbols, b-n data symbols and n ECC redundancy symbols.
The ECC redundancy symbols corresponding to the data symbols in a group are themselves encoded using the CRC code.
The various ECC redundancy symbols and the corresponding CRC
redundancy symbols form n ECC code words. Thus the nth ECC
code word contains (i) the nth ECC symbol corresponding to each column of data symbols, and (ii~ the related c CRC symbols. A
complete group of code words contains b-n data code words and n ECC code words. The various code words are recorded on the tape as a group, i.e., ln multiple tracks, or bit-serially, i.e., in a single track. Accordingly, the group may require re-assembly when the data is retrieved.
,: .
" .
, ~
~ hen data is retrieved from the tape it is first assembled in groups, if necessary, and then each group is decoded, code word- by- code word, using for each code word the CRC redundancy bits contained therein. The CRC code detects with certainty a single error of 64 bits or less and/or double long burst errors, each of which may be up to 28 bits long. The code also detects up to seven random errors in each CRC code word. If errors are detected in the data CRC
code words the system uses the Reed-Solomon code ECC
redundancy symbols and one of several decoding algorithms to correct them column by column. Error correction is in most cases simplified because the CRC decoding points to the vertical locations of the errors.
The particular Reed-Solomon code used to encode the data encodes it such that single errors with known locations can be quickly and easily corrected. If the CRC code detects errors in only one code word, code word t, there is at most one erroneous data symbol per column, i.e., the symbol in column location t. To calculate the error value associated with the symbol in location t in a particular column the system adds together (XOR's) the symbols in the column. The sum is the error value, and thus, the system corrects the error by adding together the symbol and the associated error value. If the .
J
r symbol in location t is not erroneous the calculated error value is ZERO.
When the CRC decoding indicates errors in a particular code word, the system treàts as erroneous every data symbol in that code word. Thus the system calculates an error value for each column.
If the CRC decoding indicates that multiple CRC code words contain errors and one or more of the CRC code words is a data code word, the system first determines which of the columns contain errors. The system then engages in error correction for these particular columns.
To determine whether a particular column contains any errors, the system encodes the column data symbols using the Reed-Solomon code and thus generates new ECC redundancy symbols. It then adds (XOR's) these symbols to the corresponding ECC redundancy symbols retrieved from the tape to formulate remainder symbols REMo, REMl,..., REM~. If the remainder symbols are all ZEROS the two sets of ECC symbols are the same, and thus, the data symbols in the column are error-free. If one or more of the remainder symbols are non~zero the two sets of ECC symbols are not the same and the column contains one or more erroneous symbols. The system then performs error correction for each column containing errors.
If the CRC points out as erroneous between one and n CRC
code words the system simultaneously solves a set of n 5 equations based on the remainder symbols, REMo, REMl, REM~, and the specific vertical locations of the erroneous code words. The system uses matrices manipulated by Galois Field operations to solve these equations. The solutions are the error values corresponding to the column symbols in the locations corresponding to the erroneous code words.
If the CRC code indicates that n+1 or more code word contain errors, the system corrects the errors by assuming there are n or fewer actual errors per column and using one of - several correction procedures. It then uses the CRC code to determine whether all the errors in the data code words have been corrected. If the CRC detects further errors either the system repeats the entire error correction procedure until the CRC code indicates that all the errors are corrected, or the system labels the data uncorrectable.
One procedure for correcting the n+1 errors is to assume that there is only one erronaous symbol, with an unknown location, in any particular column. The errors can then be corrected using a mechanism for correcting single errors with ~` g unknown locations. The system does not know in which of the many possible error locations the one erroneous symbol resides, and thus, the location of the error is considered unknown.
First, the system examines the remainder symbols RE~or REMl...... .......,REM~ for a particular column. If all of the remainder symbols are ZEROS, it indicates that the column data is error-free and no error correction is required. If only one of the remainder symbols, for example, REMi, is non-zero, it indicates that the symbol in column location i is erroneous. The error value corresponding to this symbol is REMi and thus the error is corrected by adding REMi to that symbol.
If more than two remainder symbols are non-zero the system uses the remainder symbols to initialize n shift registers of a cyclic shifter. The shifter cyclically shifts the symbols in a column backwards by dividing the remainder . ~ .
symbols by the reciprocal of the Reed-Solomon generator polynomial used to encode the data. Due to the cyclic property of the Reed-Solomon code the error values required to correct the errors in the column are not affected.
The shifting mechanism shifts the column symbols until the contents of registers 1, 2, ..., (n-1) are all ZEROS and the resulting symbol in the nth register is non-zero. If the .
.
shifter shiftq the ~ymbolq a total of L times the symbol in column location n+L is erroneous and the nth register contains the corresponding error value. If after the shifter is shifted b times, that is, one shift for each symbol location in a column, an error locàtion is not found error correction for that particular column fails. The system then goes on to correct the errors in the next column.
Once error correction has been attempted for each of the columns the system again examines the group of code words for errors using the CRC code. If there are further errors the system corrects them, if possible, by using an error correction procedure designed for the number of remaining errors.
BRIEF DESCRIPTION OF DRAWINGS
A more complete understanding of the invention may be had by referring to the following description of preferred embodiments, given by way of example only, to be understood in conjunction with the accompanying drawings, in which:
.
Figure 1 is a view of a magnetic tape recorded in - accordance with a preferred embodiment;
Figure 2 depicts a Reed-Solomon code encoder;
Figure 3 illustrates a matrix M used in error correction Figure 4 depicts a cyclic shifter; and Figure 5 is a view of various error configurations.
DETAILED DESCRIPTION
The mathematical operations involving Reed-Solomon code ECC redundancy symbols are performed using Galois Field addition, multiplication and/or division. Galois Field addition is performed by exclusive Or'ing (XOR) the code symbols bit-by-bit, Galois Field multiplication, represented by the symbol "*", is modulo p(x), where p(x) is a primitive polynomial which characterizes the Galois Field.
Figure 1 illustrates a group 6 of b code words 4, where b = 20 in the preferred embodiment. The code words 4, oriented horizontally, contain d data or information symbols and c associated redundancy bits. The generation of the c redundancy bits is discussed below. The code words 4 are vertically aligned such that the corresponding symbols in each of the code words 4 form a series of ona-symbol wide columns.
- . , ., ~ ,, ~ .
' The c bits in each code word 4 are generated by encoding the code word data or information symbols using a cyclic redundancy check (CRC) code which is well suited to detecting long burst errors. The c CRC redundancy bits are appended to the associated data or information symbols to form a CRC code word. The code words numbered 5 - 20 in Figure 1 contain data symbols and c CRC redundancy bits. The code word numbered 1 -4 contain d information symbols and c CRC redundancy bits. In the preferred embodiment d = 2 kilobytes and c = 64 bits. The CRC code encoding is discussed in more detail below.
The d information symbols in each of the code words 1 - 4 are Reed-Solomon error correction code (ECC) redundancy symbols. The ECC redundancy symbols are generated by encoding each of the columns containing data symbols; the columns containing CRC bits are not encoded. The ECC redundancy symbols recorded in a particular column, that is, the symbols in code words 0 - 4 of a column containing data symbols, correspond to the data symbols in that column. Thus each of these columns forms a code word of the Reed-Solomon code.
The d ECC redundancy symbols in each of the code words 1 - 4 are then encoded using the CRC code. The c CRC redundancy bits generated by this encoding are appended to the ECC
redundancy symbols to form ECC code words, that is, code words in the CRC code. Accordingly, errors in the stored ECC
~' redundancy symbols can be detected. Each column containing CRC bits forms a Reed-Solomon code word, also, because both the Reed Solomon code and the CRC code are linear codes.
To generate the CRC bits, an encoder (not shown) encodes the appropriate data or ECC redundancy symbols using a cyclic redundancy check code characterized by the following degree sixty-four binary generator polynomial g(x) presented as an octal vector:
g(x) = (3)*(3)*(100003~*(102043)*(110013)*(1560513) or g(x) - 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms. The CRC
encoder, which is of conventional design, encodes in a conventional manner the data or information symbols using this CRC generator polynomial. The data may also be encoded using the reciprocal of g(x), g*(x) (in octal form):
g*(x) - (3)*(3)*(140001)*(11420~ 0011)*(1512073) or g*(x) = 3113305362565703417205.
The length of the CRC code is shortened fro~ its natural length of over 8 billion bits to slightly over 17,000 bits.
The shortened code is a powerful code which is capable of .
detecting with certainty in the 17,000 bits of data a single burst error of up to 64 bits in length. The code further detects with certainty double burst errors of up to 28 bits each. The Hamming distance of the code is at least eight, thus the code detects up to seven independent errors in any location, i.e., seven random erron~ous bits as opposed to burst errors.
The shortened code actually detects 28-bit double burst errors in 17,907 bits of data. Thus a block may contain 2 kilobytes of actual data and up to 1523 bits of appended control information.
As set forth above, the columns of data in a group are encoded for error correction purposes using a Reed-Solomon code. The code encodes the data over a Galois Field GF(28) which is characterized by a primitive polynomial p(x) represented by the octal vector p(x) = 455, which in binary form is:
Xa + X5 + X3 + X2 + 1 The particular Reed-Solomon code used to encode the data columns is specified by the generator polynomial G(x):
G(x) = (x ~ ~) * (x + ~l) * (x + ~Z) * (x + ~3) = X4 + a75x3 + oC249X2 + ~X78X + ~X,6 where a is the primitive element of the Galois Field. The data is encoded over Galois Field GF (28) in the preferred . ~ -.
embodiment because the elements of the field are 8-bit symbols, or bytes, and most computer hardware is set up to handle data organized in bytes.
Figure 2 depicts an encoder 10 for encoding the column data symbols. The encoder 10 includes a set of 8-bit registers 16, 18, 20 and 22 that are cleared prior to the encoding operation. The contents of the registers are applied to exclusive OR circuits (XOR's) 11, 12, 13 and 14 respectively. The input to the encoder is applied to the XOR
14 in byte serial, bit parallel fashion over eight input lines.
During the input operation a switch 32 is in position so that the output of the XOR 14 is applied to Galois Field multipliers 24, 26, 28 and 30. The outputs of multipliers 26 28 and 30 are applied to XOR's 11, 12 and 13, respectively, and the output of multiplier 24 is applied to the register 16.
The Galois Field mul-tipliers 24-30 multiply their inputs by the factors a6, a78, a249 and a75 respecti Accordingly, as each input byte is applied to the encoder 10, it is summed bit-by-bit (modulo 2) with the contents of the register 22. The sum is then simultaneously multiplied (modulo p(x)) by Galois Field multipliers 24-30 which contain the coefficients of all but the highest power of x in G(x).
, The respective products of the byte and the coefficients are then added (modulo 2) to the contents of the registers 16-22.
After all of the data bytes in a particular column have been applied to the encoder 10, the encoder switch 32 is placed in position 2 and the contents of registers 16-22 are unloaded as the ECC redundancy symbols, R3, R2, Rl, Ro~
respectively, by shifting their contents to the left. It will be noted that during the shifting operation with the switch 32 in position 2, the vertical inputs of the XOR's pass the contents of the registers 22-26 without modifying them.
Encoders of this type are well known in the art.
The encoding system formulates a complete group by aligning beneath the appropriate column data symbols, in code words 1-4, the ECC redundancy symbols. Thus code word 4 contains the R3 redundancy symbols, code word 3 contains R2 redundancy symbols, et cetera. A complete column, that is, the corresponding data symbols and the associated ECC
redundancy symbols in code words 1-20, form a Reed-Solomon code word.
A processor (not shown) writes the code words to the tape. It will be understood by those skilled in the art that the processor may write the code words in any format, for : .
~ , :
" ~
.. . .
example, it may write them as a string of bits with one code word following another. Thereafter, when the code words are retrieved from the tape, the processor re-assembles the group 6.
When a user requests data from the tape the processor retrieves from the tape the code words 4 (Figure 1) containing the requested data symbols and any other code words require to re-assemble the group 6. An error correction processor (not shown) then determines if there are errors in any of the code words 4 in the group 6 by decoding each of them using the component CRC bits. If the CRC code does not detect any errors the error correction processor forwards the require data to the requesting user. If the CRC code detects errors it points out the erroneous code words, and thus, the locations of the errors. Accordingly, the error correction processor performs error correction as appropriate, to correct the symbols in these code words.
If all the errors are located in code words 1-4, that is, the ECC code words, the data is assumed error-free and the error correction processor immediately forwards the data to the requesting user. However, if errors are detected in the code words containing data symbols, the error correction processor corrects the errors (if possible) using the ECC
redundancy symbols. To correct the errors the error PD~9~0117 correction processor selects an appropriate error correction routine based on the number and location of the detected errors.
If the CRC bits indicate that only one CRC code word is S erroneous, for e~ample, code word bi, the processor corrects the erroneous data symbols in that code word using a procedure that is specifically designed for correcting single errors with known locations. For a given column containing data symbols the column either is error-free or it contains one erroneous symbol, the symbol in the column location corresponding to code word bi.
The error correction processor does not, at this point, determine if a particular column contains an erroneous symbol.
Instead, it attempts to correct the symbol in location bi in every column. Thus, for every column, the error correction processor generates an error value corresponding to the symbol in location bi. It then corrects the symbol by combining it with, i.e., adding it to, the generated error value. The generated error value for a symbol which is not in error is ZERO. Thus adding this error value to the symbol does not change it. Alternatively, the error correction processor could calculate the error values only for columns containing data symbols and ignore the columns containing CRC symbols.
; .
:. . ~, ' `': ~
~: `' : ' To generate an error value :Eor a potentially erroneous symbol in location bi the processor adds together, using Galois Field addition, all the symbols ln a column. This sum is the error value associated with the data symbol in location bi.
Thus the error correction processor adds the sum to the erroneous symbol to correct it. The same result is achieved by adding together all the symbols in the column except the symbol in location bi. This sum is the corrected data symbol.
Thus this sum replaces the symbol in location bi in the data transmitted to the user.
This single-error correction procedure is possible only with Reed-Solomon codes which have (X ~ ~) as a factor of their generator polynomials. Having this term as a factor means that the generator polynomial, which is a degree four polynomial, may have more than two distinct terms with non-~ero coefficients. Thus an associated encoder 10 will most likely include more than two distinct Galois Field multipliers. This may result in an encoder which is not the simplest one for encoding data for single-error correction.
However, the ease of correcting errors contained in a single CRC code word using such a Reed-Solomon code outweighs the increased complexity of the encoder.
If the CRC bits indicate that more that one code word contains erroneous symbols, for example, two code words, the processor corrects the errors using an error correction procedure designed to correct between one and four errors with known locations. First, the error :Location processor determines which columns contain errors by calculating a remainder for each column. The remainder for a particular column is calculated by~ encoding the retrieved column data symbols using encoder 10 to generate new ECC redundancy symbols corresponding to the retrieved data symbols; and (ii) comparing these new ECC redundancy symbols with the retrieved ECC redundancy symbols.
The two sets of ECC symbols are compared by XOR'ing bit-by-bit the corresponding symbols. The resulting symbols are the remainder symbols REMo, REMl, REM2, and REM3. If the remainder symbols are all ZEROS the two sets of ECC symbols are the same and the data column is assumed error-free. If one or more of the remainder symbols are non-zero the column contains errors.
To calculate the error values for the columns containing errors, that is, the columns with non-zero remainders, the error correction processor solves for each of these columns set of four simultaneous equations. Each of these equations involves a different column remainder symbol.
In order to formulate the equations the processor generates a matrix M with rows corresponding to X~ modulo G(x), where e is the symbol location associated with a given matrix row and G(x) is the Reed-Solomon code generator polynomial.
The rows of matrix M can be thought of as vectors with components corresponding to coefficients of (X3, XZ, Xl, X).
Thus, for example, the matrix M row corresponding to X3 iS (1~
0, 0, 0) while the matrix row corresponding to X4 iS (X7s, X249, x7a~ x6) .
The error correction processor may generate the rows of the matrix M using encoder 10 as follows: First, the processor initializes encoder registers 16, 18, 20 and 22 to the values 0,0,0,1, respectively. The registers then contain the elements of the matrix row corresponding to X. The processor next cycles, i.e., clocks, the encoder 10 once with switch 32 in position 1 to generate in the registers the elements of the second row (i.e. x1) of the matrix M. The processor clocks the encoder a total of 19 times with switch 32 in position 1 and uses the contents of the registers 16, 18, 20 and 22 generated by each encoder cycle as the rows X
through X19 of the matrix M. This matrix is then stored for use in the future. Figure 3 depicts the complete matrix M in both oc.tal and binary forms. It will be understood by those skilled in the art that the matrix is generated only once and used thereafter for all error correction.
The error value equations for a particular column with detected errors, for example, in blocks 4 and 6, are:
REM3 = Y (e4) * 17 + Y (e6) * 134 REM2 = Y(e4) * 66 + Y (e6) * 7 REMl = Y (e4) * 170 + Y (e6) * 277 REMo = Y (e4) * 100 + Y (e6) * 345 where Y(e4) and Y(e6) are the error values for the column symbols in column locations 4 and 6, respectively, and the coefficients of the Y(e4) and Y(e6) terms are the elements of matrix M rows X4 and X6, respectively, such that the equation for REM3 has as coefficients the fourth column elements of matrix M rows X4 and X6, the equation for REM2 has as coefficients the third column elements of the rows, et cetera.
The error correction processor could now solve any two of these equations to determine the error values. However, the processor uses all four equations and the procedure set forth below to determine the error values for 2, 3 or 4 detected errors.
Next to solve the equations the error correction processor formulates for each column containing erxors a matrix M'. The matrix M' has rows corresponding to each of the code words in which the CRC detects errors and a bottom row containing as elements the column remainder symbols REMo REMl, REM2 and REM3. In the example there are two code words : ~, ::~
PD~9-0117 containing errors, namely, code words 4 and 6, and thus rows 1 and 2 of matrix M' are the same as rows X~ and x6 of matrix M.
The third row of matrix M' is the column remainder REMo, REMl, REM2 and REM3. In this particular example the column remainder is an octal vector 175, 47, 211, 123. Thus matrix M' in octal form is:
row X4 ¦ 17 66 170 100 ¦
row x6 ¦ 134 7 277 345 ¦
remainder ¦ 123 211 47 175 ¦
The error correction processor next manipulates the matrix M' using column operations to reduce it to a matrix M"
which has as elements of the first two columns of the first two rows a 2x2~identity matrix and has as the remaining elements of rows 1 and 2 only ZEROS. In this example, matrix M" is:
~ ' ~ J
If the CRC indicates 3 errors, the corresponding matrix M' has 4 rows and it reduces to a matrix M" containing a 3x3 identity matrix in the first three columns of the first thre rows and has only ZEROS as the remaining elements of rows 1-3.
The two column operations performed in reducing matrix M' are: (1) a*[column i] + b*[column j] to replace column i, where "a" is non-zero element of GF(28) and "b" is any element of GF(28), and (2) interchanging two columns.
When the matrix M" is in this reduced form the last row of the matrix contains the error values associated with the data or ECC redundancy symbols in the erroneous code words.
Thus the matrix element in the first column, 355, is the error value associated with the ECC redundancy symbol in code word 4 and the element in the second column, 135, is the error value associated with the data symbol in code word 6. The elements in the two matrix columns which do not correspond to errox locations, columns 3 and 4, should be ZEROS. Otherwise, an uncorrectable error condition exists and the calculated error values are rejected.
If only one of the detected errors is in a data code word and each of the other detected errors are in the ECC code words the matrix M' corresponding to the detected errors may be used directly to calculate the error value of the erroneous PD~9-0117 data symbol. Thus the matrix need not be reduced. If, for example, the errors are detected in ECC code words 0 and 3 an data code word 5, and the remainder symbols for a given column are 321, 335, 360 and 375, the matrix M' is:
1 0 ~ I I
The matrix M' has in its first two rows, that is, the rows corresponding to the ECC code words, only ONES and ZEROS.
Thus the matrix is already in a reduced form, though not in the reduced form discussed above. The error value corresponding to the code word 5 data symbol is calculated using this matrix M' by dividing the element in the last row of either matrix column 1 or column 2, that is, either of the columns having ZEROS in rows 1 and 2, by the element in row 3 of the same matrix column. Thus the error value for the code word 5 data symbol, in octal form, is:
360/127 = 335/322 = 115 An alternative error correction procedure for correcting double data-symbol errors with known locations utilizes the ' ~Y-U117 cyclic property of Reed-Solomon codes to "shift" one of the errors into an ECC symbol location. Thus the symbols are cyclically shifted backwards by some number of locations without altering the error values associated with the symbols.
After the shift the error correction processor calculates the error value associated with the symbol which is still in a data symbol location using the procedure set forth above for a single data-symbol error. The processor then calculates the error value corresponding to the remaining erroneous symbol using the single error with known location procedure set forth above, that is, by adding together all the symbols in the column including the corrected data symbol.
Referring now to Figure 4, in order to cyclically shift the symbols in a column backwards the error correction processor uses a cyclic shifter 40, which is a linear shift register set up to multiply the symbols in registers 44, 46, 48, 50 by the reciprocal of G(x). Thus Galois Field multipliers 52-58 contain the coefficients of the terms of the reciprocal of G(x). For example, multiplier 52 multiplies terms by a78/a6, which is the coefficient of the X~ term of G(x) divided by the coefficient of the X term of G(x). The multiplicands of multipliers 54-58 are the X2, X~, and X4 coefficients of G(x) divided by the Xcoefficient (the . :,:
' " ' ' ~ . -, ' ' , PD~9-0117 coefficient of the X4 term is 1, which in GF(28) ls the same as o~255 ) To shift a particular column the error correction processor loads the column remainder symbols REMo, REMl, REM2 and REM3 into registers 44, 46, 48, 50, respectively. It then shifts the cyclic shifter 40 L times until the location of only one of the errors corresponds to a data symbol and the location of the other error corresponds to an ECC symbol. Thus the error locations are shifted from locations e1 and e2 to locations e1-L and e2-L.
The error correction processor then uses the symbols in the registers 44, 46, 48, 50 as the column remainder symbols in formulating matrix M'. The first two rows of matrix M' correspond to the column locations e1-L and e2-L and the third row corresponds to the shifted remainder symbols. The processor then calculates the error value for the data symbol by dividing the appropriate elements of rows 2 and 3 of matrix M'. The processor corrects the symbol by adding to it the calculated error value.
To calculate the error value for the remaining error, that is, the symbol in location e2, the processor adds together the column symbols, including the corrected symbol in location el. The processor then corrects the symbol in location e2 by adding to it the calculated error value.
If more than four errors are detected the error correction processor corrects them by first correcting a predetermined number of errors, f, and repeating the erro correction procedure for any remaining errors. Figure 5 depicts several configurations of five or more erroneous code words. The "X's" denote erroneous symbols. Configuration 1, for example, depicts code words with errors occurring such that each column contains at most one error. Thus the errors can be corrected using single-error correction techniques. Similarly configuration 2 has at most two errors per column and the errors can be corrected using double-error correction techniques, and so forth. The system does not have any mechanism which is capable of determining exactly how the errors in the five or more erroneous code words are distributed, however. Therefore special error correct procedures must be used the correct the errors.
One such error correction procedure calculates the error values associated with a particular number of errors, v, with known locations and a single error with an unknown location.
Thus the error correction processor selects v detected error locations and uses the matrix M rows associated wïth these locations, along with the column remainder symbols, to solve for the single unknown error location. ~fter finding the PD~ ~-011 7 unknown location the processor corrects the v + 1 errors using the procedure for v + 1 errors with known locations. The value of v may be either ONE or TWO. The location of the single error is unknown because the system does not know the S exact location of the erronèous symbol in a particular column, it knows only that the error is ln one of the remaining locations indicated by the CRC code.
The remainder associated with a column is a linear combination of all the symbols in a column, including the erroneous symbols. Thus equations, for example, for two errors with known locations and a single error with an unknown location are:
REM3 = Y (EV1) *X3Vl + Y (EV2) *X3V2 + Y (Eu) *X3U
REM2 = Y (EV1) *X2Vl + Y (EV2) *X2V2 + Y (Eu) *X2U
REMl = Y (EV1) *XlVl + Y (EV2) *X1V2 + Y (Eu) *XlU
REMo = Y (EV1) *XOV1 + Y (EV2) *X0v2 + Y (Eu) *xOU
where Y(El) is the error value, a scalar, associated with the error in location i and Xji is the jth element of row i of Matrix M (Figure 2). These equations can be re-written:
X3U = ~l/Y (Eu) ] *REM3 + [Y (EV1) /Y (Eu) ] *X3Vl + [Y (EV2) /Y (Eu) ] *X3V2 X2U = [l/Y (Eu) ] *REM2 + [Y (EV1) /Y (Eu) ] *X2Vl + [Y (EV2) /Y (Eu) ] *X2V2 Xl = [l/Y (Eu) ] *REMl + [Y (3V1) /Y (Eu) ] *Xl 1 + [Y (EV2) /Y (Eu) ] *X1V2 X0 = [l/Y (Eu) ] *REMo + [Y (EV1) /Y (Eu) ] *XOV1 + [Y (EV2) /Y (Eu) ] *X0v2 ~- :
: :
SY-Ull /
or more generally:
XU = AR*REMC ~ Ai*XVl + Aj*XV2 where the Ab's are scalar coefficients. Thus the vec-tor XU
corresponding to the unknown error location is a llnear combination of (1) the column remainder symbol vector REMC, and (2) the two rows or vectors of matrix M corresponding to -the two known error locations, namely, XVl and XV2.
To solve for the unknown error location, that is, determine which row or vector of matrix M can be expressed as a linear combination of the remainder symbols and the error location vectors, the error correction processor formulates a matrix M' following the procedures set forth above. As an example, let v=2. The error correction processor formulates a 3x4 matrix M' corresponding to the two selected error locations and the column remainder symbols. The error correction processor then reduces matrix ~' using only row operations to a matrix M" which includes within it a (v~l~x(v~1) identity matrix. By using only row operations the error correction processor preserves the relationships between matrix columns and thus preserves the relationships between the various vectors.
In this example the reduced matrix M" includes a 3x3 identity matrix. It may take any of the following forms:
` ' .:
PD89~0117 la3 0 0 1 ¦ ¦o a2 1¦ lo o a~ o o 1 aO¦
¦b3 0 1 0 ¦orlO b21 O¦or¦O 1 bl O¦orlo 1 o bo¦
1~3 1 0 I 11 C2 ol 11 o c, ol11 o o col ~ ~ ~ ` ~ ~ _ ~ J
The two row operations performed in reducing matrix M' to matrix M" are: (1) a*[row i] + ~*[row j] to replace row i, where "a" is any non-zero element of GF(28) and "b" is any element of GF(28), and (2) interchanging two rows.
Next, the error correction processor determines if a row Xi of the large matrix M (F.igure 2) can be expressed as a linear combination of the rows of the matrix M" by first creating a new matrix N which is the matrix M" with the vector Xj appended as a new last row. Thus the last row of matrix N
has elements S j3, S j2, S jl and sjo. The error correction processor then rotates (v=1) or permutates (v=2) the columns of matrix N until the included identity matrix is located in the upper-right corner. Thus matrix N is of the form:
~ 1 ¦a3'0 0 1 ¦
Ib3'0 1 0 1 IC3~1 0 0 I S j3rS j2r S jlr S j ~ 32 . .
,. : .: .
..
:. , : ~, , .
The error correction processor then tests if the row Xj is a linear combination of the matrix M" rows by substituting the appropriate elements of the permutated matrix N into the following equation:
sj3' = a3'*sjo' + b3'*sjl' + C3~ *S j2' If the equation is satisfied Xi is the unknown error location. Otherwise, the error location processor repeats the procedure for the matrix M rows Xj+l, Xi~2... Xi~l9 in which errors are detected until the location of the error is found or all the possible error locations have been tried.
If the error correction processor is attempting to find the unknown error location using only one known error location, i.e., v=l, the equation into which the elements of the rotated matrix N are substituted is one of the following:
sj2' = a2~*sj0~ + b2'*sjl' sj3' = a2~*sj0~ + b3'*sjl' In this case there is a redundant equation and the calculated error location can be checked by substituting the appropriate matrix N elements into the second equation.
After error correction has been tried for all of the group 6 (Figure 1) columns containing data symbols the error correction processor again examines blocks 5 - 20 for any remaining errors using the CRC code. If all the errors have been corrected the processor transfers the error-free data to the requesting user. If all the errors have not been corrected the processor corrects them, if possible, using an error correction procedure which is appropriate for the number and location of the remaining errors.
Error correction for a particular column fails if none of the rows of matrix M can be expressed as a linear combination of the rows of the associated matrix M". The system may then ignore this column and go on to correct the errors in the other columns. Thereafter, when all the columns which are correctable have been corrected and the error correction processor checks the code words for remaining errors using the CRC bits, error correction continues only if the number of code words now containing errors is smaller than the number containing errors before error correction. Otherwise, the processor labels the data uncorrectable.
An alternate method of correcting five or more errors has the error correction processor first correcting the columns containing single errors using a procedure for correcting .
.~ . . i .
single errors with unknown locations and then correcting any remaining errors using a procedure designed for the number and the locations of the errors. Presumably, by correcting the columns with single errors, the number of code words containing the remaining errors is reduced to four or below.
In this way, one of the procedures for correc-ting errors with known locations can be used to complete the error correction.
To correct a single error in a particular column the system first examines the column remainder symbols. If all the remainder symbols are ZEROS there are no errors in the column and thus no error correction is required. If there is only one non- zero remainder symbol, for example, REM3, the error is in column location 3 and the remainder symbol is the associated error value. If only two of the remainder symbols ~EMl and REM3 are non-zero it means that the particular column contains two errors instead of one. The -two errors are in the locations corresponding to the remainder symbols, namely, column locations 1 and 3, and the remainder symbols are the corresponding error values.
If more than two remainder symbols are non-zero the processor loads the remainder symbols in the registers 44, 46, 48, and 50 of cyclic shifter 40 (Figure 4) and clocks the shifter until the symbol in register 50 is non-zero and the symbols in registers 44, 46, 48 are all ZEROS. If the shifter ,~
PD~9-0117 is clocked K times to produce this result the column symbol in location K+3 is~in error and the symbol in register 50 is the error value associated with that symbol. If, after b cycles, an error location not found, it indicates that the column contains more than one error. The processor then ignores this column and goes on to the next one. The processor performs this single error correction procedure for each column and then tests the group 6 (Figure 1) for any remaining errors using the CRC code. If errors remain in the code words containing data symbols the system uses another procedure, as appropriate, to correct them.
The cyclic shifter 40 may be used to correct double errors with unknown locations, also. However, before the shifter is loaded the column remainder symbols are examined as set forth above to determine if only one or two remainder symbols are non- zero. If this is the case, the errors are located in the locations corresponding to the non-zero remainder symbols and the remainder symbols are the error values corresponding to the symbols in these locations.
If more than two remainder symbols are non-zero, the error correction processor loads the remainder symbols into registers 44, 46, 48, 50 of the cyclic shifter 40. The error correction processor then examines the contents of the registers to determine if the symbols in registers 46, 48, and .
50 are proportional to the last three elements of a row Xj of the large matrix M (Figure 2), namely, elements s j3, s j2, s jl.
Thus the error correction processor determines if:
REM3 = Uj* s j3 REM2 = Uj* s j2 REMl = Uj* s If these conditions are satisfied for a particular row X
the errors are at location 0 and location O+j, and the corresponding error values are Uj*s jo + REMo and Uj, respectively. If these conditions are not satisfied for any of the matrix M rows the error correction processor clocks the cyclic shifter 40 to produce in the registers 44, 46, 48, 50 new remainder symbols REMo', REMl', REM2' and REM3'. The processor then determines if these new remainder symbols satisfy the conditions set forth above. The procedure is repeated 19 times, once for each potential error location, if necessary.
It will be understood by those skilled in the art that it is unlikely for a group 6 (Figure 1) to contain more than 4 erroneous blocks in a well designed system. Thus error correction procedures for 5 or more code words are rarely used. Accordingly, a system which includes procedures for correcting up to four code words with errors and labels groups containing 5 or more code words with errors as uncorrectable does not degrade significantly the system performance.
The error correction codes and the error correction procedures set forth above reduce the number of undetectable and/or uncorrectable errors in data recorded on magnetic tape to a level which is comparable to the level achieved with data recorded on magnetic disks. The powerful CRC code used to encode the data for error detection detects longer burst errors than prior art CRC codes. Thus these errors, which are one of the most common types of errors for tapes, are more accurately detected.
Combining the powerful detection capabilities of the CRC
code with a Reed-Solomon code which quickly corrects errors with known locations results in a fast error correction scheme. Thus the transfer rate of the data is high. The particular Reed- Solomon code is particularly well suited for correcting single (column) errors with known locations. The single error correction is so fast that the error correction processor attempts to correct each column rather than determine first which columns contain errors. Accordingly, the length of an error, that is, the burst length, does not affect the time it takes the processor to correct the errors, and long burst errors are corrected as quickly as shorter errors.
.
The foregoing description has been limited to various embodiments of this invention. It will be apparent, however, that further variations and modifications may be made to the invention, with the attainment of some or all of the advantages of the invention. Therefore, it is the object of the appended claims to cover all such variations and modifications as come within the scope of the invention.
.
Claims (19)
1. A system for encoding data and recording it on a magnetic tape while detecting and correcting errors, characterized by:
A. a first encoder for encoding a stream of d data symbols to generate error detection redundancy symbols, said encoder using a cyclic redundancy code having as a generator polynomial g(x) (in octal form):
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513) or g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms, said generated redundancy symbols being appended to the data symbols to form a series of CRC code words;
B. a second encoder for encoding using an error correction code a predetermined number, h, of data symbols from h CRC code words to generate error correction redundancy symbols, the h data symbols and the associated error correction redundancy symbols forming error correction code words; and C. recording means for recording on the tape the h CRC
code words and the associated error correction redundancy symbols.
A. a first encoder for encoding a stream of d data symbols to generate error detection redundancy symbols, said encoder using a cyclic redundancy code having as a generator polynomial g(x) (in octal form):
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513) or g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms, said generated redundancy symbols being appended to the data symbols to form a series of CRC code words;
B. a second encoder for encoding using an error correction code a predetermined number, h, of data symbols from h CRC code words to generate error correction redundancy symbols, the h data symbols and the associated error correction redundancy symbols forming error correction code words; and C. recording means for recording on the tape the h CRC
code words and the associated error correction redundancy symbols.
2. The encoding and recording system of claim 1, wherein:
A. said first encoder further encodes the corresponding error correction redundancy symbols in a predetermined number of error correction code words to generate CRC
redundancy symbols and form CRC code words; and B. said recording means further records said generated CRC code words.
A. said first encoder further encodes the corresponding error correction redundancy symbols in a predetermined number of error correction code words to generate CRC
redundancy symbols and form CRC code words; and B. said recording means further records said generated CRC code words.
3. The encoding and recording system of claim 2, wherein said second encoder encodes the data symbols using a Reed-Solomon code having as a generator polynomial G(x):
G(x) = (x + .alpha.°) * (x + .alpha.1) * (x + .alpha.2) * (x + .alpha.3) = x4 + .alpha.75x3 +.alpha.249x2 +.alpha.78x + .alpha.6 where .alpha. is the primitive element of the Galois Field, "*"
represents Galois Field multiplication and "+" represents Galois Field addition.
G(x) = (x + .alpha.°) * (x + .alpha.1) * (x + .alpha.2) * (x + .alpha.3) = x4 + .alpha.75x3 +.alpha.249x2 +.alpha.78x + .alpha.6 where .alpha. is the primitive element of the Galois Field, "*"
represents Galois Field multiplication and "+" represents Galois Field addition.
4. An error correction apparatus for correcting errors in data, the data being encoded for detection using a Cyclic Redundancy Check (CRC) code to form CRC redundancy bits, the redundancy bits and the data forming a CRC code word, and encoded for error correction using an error correction code to form error correction redundancy symbols, the error correction redundancy symbols and the data forming an error correction code word, said error correction apparatus characterized by:
A. a decoder for decoding the data using the CRC code to detect the locations of errors in the data;
B. an error corrector responsive to the locations of the detected errors for calculating error values for v detected errors, where v is less than a predetermined maximum value, f, said error corrector including:
1. means for calculating a remainder for each error correction code word, 2. means for simultaneously solving the equations, .SIGMA.XikY(ei) = REMk where Xik is the kth term of the vector xi modulo G(x), G(x) being the generator polynomial of the error correction code used to encode the data, and Y(ei) is the error value associated with the symbol in location i, said means including:
i. formulating means for formulating a matrix M' for each error correction code word with a non-zero remainder, the matrix M' having:
a. v rows corresponding to the detected error the v rows having elements corresponding to a vector with components representing X? modulo G(x) where e is the location of a detected error and G(x) is the generator polynomial of the error correction code used to encode the data, and b. a last row corresponding to the code word remainder, ii. means for manipulating the matrix M' columns to form a matrix M" having a v x v identity matrix in the first v matrix columns of matrix rows 1-v and having ZEROS as the remaining elements of the matrix rows 1-v, and, iii. means for combining, by using the matrix M" last row elements as the error values associated with the symbols in the detected error locations, the error values with the corresponding symbols.
A. a decoder for decoding the data using the CRC code to detect the locations of errors in the data;
B. an error corrector responsive to the locations of the detected errors for calculating error values for v detected errors, where v is less than a predetermined maximum value, f, said error corrector including:
1. means for calculating a remainder for each error correction code word, 2. means for simultaneously solving the equations, .SIGMA.XikY(ei) = REMk where Xik is the kth term of the vector xi modulo G(x), G(x) being the generator polynomial of the error correction code used to encode the data, and Y(ei) is the error value associated with the symbol in location i, said means including:
i. formulating means for formulating a matrix M' for each error correction code word with a non-zero remainder, the matrix M' having:
a. v rows corresponding to the detected error the v rows having elements corresponding to a vector with components representing X? modulo G(x) where e is the location of a detected error and G(x) is the generator polynomial of the error correction code used to encode the data, and b. a last row corresponding to the code word remainder, ii. means for manipulating the matrix M' columns to form a matrix M" having a v x v identity matrix in the first v matrix columns of matrix rows 1-v and having ZEROS as the remaining elements of the matrix rows 1-v, and, iii. means for combining, by using the matrix M" last row elements as the error values associated with the symbols in the detected error locations, the error values with the corresponding symbols.
5. The error correction apparatus of claim 4, wherein said for calculating a remainder includes:
A. means for generating new error correction code redundancy symbols corresponding to the data; and B. means for adding, by using Galois Field addition, the new error correction code redundancy symbols to the error correction code redundancy symbols in the data.
A. means for generating new error correction code redundancy symbols corresponding to the data; and B. means for adding, by using Galois Field addition, the new error correction code redundancy symbols to the error correction code redundancy symbols in the data.
6. The error correction apparatus of claim 5, wherein said error corrector further includes means for calculating the error value for the only error in a location corresponding to a data symbol by dividing an element of matrix M' row v by an element of matrix M' row (v-1), the elements being in a matrix M' column containing only ZEROS in rows 1-(v-2).
7. The error correction apparatus of claim 6, wherein said error corrector further includes means for labeling code words with remainders equal to ZERO as error-free and refraining from performing further error correction on the symbols in these code words.
8. The error correction apparatus of claim 7, wherein said apparatus further includes means for determining the location of an error when there are more than f detected errors, said means including:
A. means for selecting v error locations from the detected locations and supplying the selected locations to the formulating means, said formulating means formulating a matrix M' corresponding to the selected rows;
B. means for manipulating the rows of matrix M' to form a matrix M" having a v x v identity matrix in the first v rows;
C. means for appending a vector Xe, where e is a vertical location of a detected error other than a selected location, to the matrix M" as a last row to form a matrix N;
D. means for manipulating the columns of matrix N until the identity matrix contained therein is located in the upper-right-hand corner of the matrix; and E. means for determining if the appended row Xe is a linear combination of the rows of matrix M', said means including:
i. means for determining if the elements of the manipulated matrix N satisfy the conditions:
sj2' = a2'*sjo' + b2'*sj1' sj3' = a3'*sjo' + b3'*sj1' if the number of selected error locations is one, or the condition sj3' = a3'*sjo' + b3'*sj1' + c3'*sj2' if the number of selected error locations is two.
A. means for selecting v error locations from the detected locations and supplying the selected locations to the formulating means, said formulating means formulating a matrix M' corresponding to the selected rows;
B. means for manipulating the rows of matrix M' to form a matrix M" having a v x v identity matrix in the first v rows;
C. means for appending a vector Xe, where e is a vertical location of a detected error other than a selected location, to the matrix M" as a last row to form a matrix N;
D. means for manipulating the columns of matrix N until the identity matrix contained therein is located in the upper-right-hand corner of the matrix; and E. means for determining if the appended row Xe is a linear combination of the rows of matrix M', said means including:
i. means for determining if the elements of the manipulated matrix N satisfy the conditions:
sj2' = a2'*sjo' + b2'*sj1' sj3' = a3'*sjo' + b3'*sj1' if the number of selected error locations is one, or the condition sj3' = a3'*sjo' + b3'*sj1' + c3'*sj2' if the number of selected error locations is two.
9. A method for encoding a stream of data symbols to allow detection of errors, said method using a cyclic redundancy code having as a generator polynomial g(x) (in octal form):
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513) or g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms.
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513) or g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms.
10. A method for encoding data and recording it on a magnetic tape including the steps of:
A. encoding a stream of d data symbols to generate error detection redundancy symbols using a cyclic redundancy code having as a generator polynomial g(x) (in octal form):
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513) or g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms, said generated redundancy symbols being appended to the data symbols to form a series of CRC code words;
B. encoding using an error correction code a predetermined number, h, of data symbols from h CRC code words to generate error correction redundancy symbols, the data symbols and the corresponding error correction redundancy symbols forming an error correction code word;
and C. recording on the tape the h CRC code words and the associated error correction redundancy symbols.
A. encoding a stream of d data symbols to generate error detection redundancy symbols using a cyclic redundancy code having as a generator polynomial g(x) (in octal form):
g(x) = (3)*(3)*(100003)*(102043)*(110013)*(1560513) or g(x) = 2413607036565172433223 where the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms, said generated redundancy symbols being appended to the data symbols to form a series of CRC code words;
B. encoding using an error correction code a predetermined number, h, of data symbols from h CRC code words to generate error correction redundancy symbols, the data symbols and the corresponding error correction redundancy symbols forming an error correction code word;
and C. recording on the tape the h CRC code words and the associated error correction redundancy symbols.
11. The encoding and recording method of claim 10, wherein said error detection encoding step and said recording step further include the steps of:
A. encoding the corresponding error correction redundancy symbols in a predetermined number of error correction code words to generate CRC redundancy symbols and form CRC code words; and B. recording said generated CRC code words.
A. encoding the corresponding error correction redundancy symbols in a predetermined number of error correction code words to generate CRC redundancy symbols and form CRC code words; and B. recording said generated CRC code words.
12. The encoding and recording method of claim 11, wherein said step of encoding the data for error correction further includes encoding the data symbols using a Reed-Solomon code having as a generator polynomial G(x):
G(x) = (x + .alpha.°) * (X + .alpha.1) * (x + .alpha.2) * (x + .alpha.3) = X4 + .alpha.75X3 +.alpha.249X2 +.alpha.78X + .alpha.6 where .alpha. is the primitive element of the Galois Field, "*"
represents Galois Field multiplication and "+" represents Galois Field addition.
G(x) = (x + .alpha.°) * (X + .alpha.1) * (x + .alpha.2) * (x + .alpha.3) = X4 + .alpha.75X3 +.alpha.249X2 +.alpha.78X + .alpha.6 where .alpha. is the primitive element of the Galois Field, "*"
represents Galois Field multiplication and "+" represents Galois Field addition.
13. A method of correcting errors in data, the data being encoded for detection using a CRC code to form CRC redundancy bits, the redundancy bits and the data forming a CRC code word, and encoded for error correction using an error correction code to form error correction redundancy symbols, the error correction redundancy symbols and the data forming an error correction code word, said error method including the steps of:
A. decoding the data using the CRC code to detect the locations of errors in the data;
B. correcting the errors in the detected locations provided that only v errors are detected, where v is less than a predetermined maximum value, f, said error correcting step including the steps of:
1. calculating a remainder for each error correction code word, 2. simultaneously solving the equations, .SIGMA.XikY(ei) =
REMk where Xik is the kth term of the vector Xi modulo G(x), G(x) being the generato polynomial of the error correction code used to encode the data, and Y(ei) is the error value associated with the symbol in location i, said solving step including the steps of:
i. formulating a matrix M' for each error correction code word with a non-zero remainder, the matrix M' having:
a. v rows corresponding to the detected errors, the v rows having elements corresponding to a vector with components representing Xe modulo G(x) where e is the location of a detected error and G(x) is the generator polynomial of the error correction code used to encode the data, and b. a last row corresponding to the code word remainder, ii. manipulating the matrix M' columns to form a matrix M" having a v x v identity matrix in the first v matrix columns of matrix rows 1-v and having ZEROS as the remaining elements of the matrix rows 1-v, and, iii. combining, by using the matrix M" last elements as the error values associated with the symbols in the detected error locations, the error values with the corresponding symbols.
A. decoding the data using the CRC code to detect the locations of errors in the data;
B. correcting the errors in the detected locations provided that only v errors are detected, where v is less than a predetermined maximum value, f, said error correcting step including the steps of:
1. calculating a remainder for each error correction code word, 2. simultaneously solving the equations, .SIGMA.XikY(ei) =
REMk where Xik is the kth term of the vector Xi modulo G(x), G(x) being the generato polynomial of the error correction code used to encode the data, and Y(ei) is the error value associated with the symbol in location i, said solving step including the steps of:
i. formulating a matrix M' for each error correction code word with a non-zero remainder, the matrix M' having:
a. v rows corresponding to the detected errors, the v rows having elements corresponding to a vector with components representing Xe modulo G(x) where e is the location of a detected error and G(x) is the generator polynomial of the error correction code used to encode the data, and b. a last row corresponding to the code word remainder, ii. manipulating the matrix M' columns to form a matrix M" having a v x v identity matrix in the first v matrix columns of matrix rows 1-v and having ZEROS as the remaining elements of the matrix rows 1-v, and, iii. combining, by using the matrix M" last elements as the error values associated with the symbols in the detected error locations, the error values with the corresponding symbols.
14. The method of correcting errors of claim 13, wherein said step of calculating a remainder includes the steps of:
A. generating new error correction code redundancy symbols corresponding to the data; and B. adding, by using Galois Field addition, the new error correction code redundancy symbols to the error correction code redundancy symbols in the data.
A. generating new error correction code redundancy symbols corresponding to the data; and B. adding, by using Galois Field addition, the new error correction code redundancy symbols to the error correction code redundancy symbols in the data.
15. The method of correcting errors of claim 14, wherein said method includes the step of calculating the error value for the only error in a location corresponding to a data symbol by dividing an element of matrix M' row v by an element of matrix M' row (v-1), the elements being in a matrix M' column containing only ZEROS in rows 1-(v-2).
16. The method of correcting errors of claim 15, wherein said method further includes the step of labeling code words with remainders equal to ZERO as error-free and refraining from performing further error correction on the symbols in these code words.
17. The method of correcting errors of claim 16, wherein said method further includes, for determining the location of an error when there are more than f detected errors, the steps of:
A. selecting v error locations from the detected locations and supplying the selected locations to the formulating means, said formulating means formulating a matrix M' corresponding to the selected rows;
B. manipulating the rows of matrix M' to form a matrix M" having a v x v identity matrix in the first v rows;
C. appending a vector Xe, where e is a location of a detected error other than a selected location, to the matrix M" as a last row to form a matrix N;
D. manipulating the columns of matrix N until the identity matrix contained therein is located in the upper-right- hand corner of the matrix; and E. determining if the appended row Xe is a linear combination of the rows of matrix M', by determining if the elements of the manipulated matrix N satisfy the conditions:
sj2' = a2'*sj0' + b2'*sj1' sj3' = a3'*sj0' + b3'*sj1' if the number of selected error locations is one, or the condition:
sj3' = a3'*sj0' + b3'*sj1' + c3' *sj2' if the number of selected error locations is two.
A. selecting v error locations from the detected locations and supplying the selected locations to the formulating means, said formulating means formulating a matrix M' corresponding to the selected rows;
B. manipulating the rows of matrix M' to form a matrix M" having a v x v identity matrix in the first v rows;
C. appending a vector Xe, where e is a location of a detected error other than a selected location, to the matrix M" as a last row to form a matrix N;
D. manipulating the columns of matrix N until the identity matrix contained therein is located in the upper-right- hand corner of the matrix; and E. determining if the appended row Xe is a linear combination of the rows of matrix M', by determining if the elements of the manipulated matrix N satisfy the conditions:
sj2' = a2'*sj0' + b2'*sj1' sj3' = a3'*sj0' + b3'*sj1' if the number of selected error locations is one, or the condition:
sj3' = a3'*sj0' + b3'*sj1' + c3' *sj2' if the number of selected error locations is two.
18. A data encoding system for encoding a stream of data symbols to allow detection of errors, said system including an encoder which encodes data using a cyclic redundancy code having as a generator polynomial g*(x) (in octal form):
g*(x) = (3)*(3)*(140001)*(1142041)*(150011)*(1512073) or g*(x) = 3113305362565703417205 and the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms.
g*(x) = (3)*(3)*(140001)*(1142041)*(150011)*(1512073) or g*(x) = 3113305362565703417205 and the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms.
19. A method for encoding a stream of data symbols to allow detection of errors, said method using a cyclic redundancy code having as a generator polynomial g*(x), where g*(x) is (in octal form):
g*(x) = (3)*(3)*(140001)*(1142041)*(150011)*(1512073) or g*(x) = 3113305362565703417205 and the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms.
g*(x) = (3)*(3)*(140001)*(1142041)*(150011)*(1512073) or g*(x) = 3113305362565703417205 and the vector components are the binary coefficients of the terms of the polynomial representing from left-to-right the coefficients of the highest-to-lowest degree terms.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US373,323 | 1989-06-28 | ||
US07/373,323 US5136592A (en) | 1989-06-28 | 1989-06-28 | Error detection and correction system for long burst errors |
Publications (1)
Publication Number | Publication Date |
---|---|
CA2016847A1 true CA2016847A1 (en) | 1990-12-28 |
Family
ID=23471910
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CA002016847A Abandoned CA2016847A1 (en) | 1989-06-28 | 1990-05-15 | Error detection and correction systems for long burst errors |
Country Status (4)
Country | Link |
---|---|
US (1) | US5136592A (en) |
EP (1) | EP0407101A3 (en) |
JP (1) | JPH03136524A (en) |
CA (1) | CA2016847A1 (en) |
Families Citing this family (108)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2591242B2 (en) * | 1990-04-02 | 1997-03-19 | 松下電器産業株式会社 | Error detection method |
NL9101376A (en) * | 1990-08-16 | 1992-03-16 | Digital Equipment Corp | AN IMPROVED ERROR DETECTION CODING SYSTEM. |
US5280488A (en) * | 1990-11-08 | 1994-01-18 | Neal Glover | Reed-Solomon code system employing k-bit serial techniques for encoding and burst error trapping |
US5210760A (en) * | 1991-02-15 | 1993-05-11 | International Business Machines Corporation | Optimized pointer control system |
JP2554219B2 (en) * | 1991-11-26 | 1996-11-13 | 日本電信電話株式会社 | Digital signal superposition transmission method |
US5375127A (en) * | 1992-03-25 | 1994-12-20 | Ncr Corporation | Method and apparatus for generating Reed-Soloman error correcting code across multiple word boundaries |
ES2099789T3 (en) * | 1992-07-14 | 1997-06-01 | Alcatel Bell Nv | DIVIDING DEVICE TO DIVIDE A FIRST POLYNOMY BY A SECOND POLYNOMY. |
DE69327212T2 (en) * | 1992-09-14 | 2000-07-27 | Nippon Telegraph & Telephone | System for transmission of information bit sequences |
US5446745A (en) * | 1992-10-05 | 1995-08-29 | Mitsubishi Semiconductor America, Inc. | Apparatus for correcting errors in optical disks |
JP2821324B2 (en) * | 1992-11-04 | 1998-11-05 | 三菱電機株式会社 | Error correction circuit |
FI940322A (en) * | 1993-01-25 | 1994-07-26 | Hughes Aircraft Co | Improved error correction decoder and method for receivers in digital cellular communication systems |
US5463642A (en) * | 1993-06-29 | 1995-10-31 | Mitsubishi Semiconductor America, Inc. | Method and apparatus for determining error location |
US5602857A (en) * | 1993-09-21 | 1997-02-11 | Cirrus Logic, Inc. | Error correction method and apparatus |
US5465260A (en) * | 1993-11-04 | 1995-11-07 | Cirrus Logic, Inc. | Dual purpose cyclic redundancy check |
US5629949A (en) * | 1993-09-21 | 1997-05-13 | Cirrus Logic, Inc. | Error correction verification method and apparatus using CRC check remainders |
US6125469A (en) * | 1994-10-18 | 2000-09-26 | Cirrus Logic, Inc. | Error correction method and apparatus |
US5864440A (en) * | 1994-12-27 | 1999-01-26 | International Business Machines Corporation | Data processing method and data storage system |
FR2740280B1 (en) * | 1995-10-20 | 1997-11-21 | Cga Hbs | SYSTEM FOR CORRECT DECODING, IN THE FORM OF A BINARY MESSAGE, OF A MAGNETIC CODE RECORDED ON A MAGNETIC TRACK |
US5828513A (en) * | 1995-12-07 | 1998-10-27 | International Business Machines Corporation | Servo address apparatus and positioning methods for read, write and seek operations in a direct access storage device |
US5721744A (en) * | 1996-02-20 | 1998-02-24 | Sharp Microelectronics Technology, Inc. | System and method for correcting burst errors in digital information |
US5787094A (en) * | 1996-06-06 | 1998-07-28 | International Business Machines Corporation | Test and diagnostics for a self-timed parallel interface |
EP0850473B1 (en) * | 1996-06-13 | 2002-11-27 | Koninklijke Philips Electronics N.V. | A method and apparatus for trackwise burst error correction in a multitrack storage format |
JP3249405B2 (en) * | 1996-09-30 | 2002-01-21 | 株式会社東芝 | Error correction circuit and error correction method |
GB2324445B (en) * | 1997-04-17 | 2002-08-21 | United Microelectronics Corp | Error correction and detection system for mass storage controller |
US6081918A (en) * | 1997-11-06 | 2000-06-27 | Spielman; Daniel A. | Loss resilient code with cascading series of redundant layers |
US6092231A (en) * | 1998-06-12 | 2000-07-18 | Qlogic Corporation | Circuit and method for rapid checking of error correction codes using cyclic redundancy check |
US6307487B1 (en) | 1998-09-23 | 2001-10-23 | Digital Fountain, Inc. | Information additive code generator and decoder for communication systems |
US7068729B2 (en) | 2001-12-21 | 2006-06-27 | Digital Fountain, Inc. | Multi-stage code generator and decoder for communication systems |
JP4071879B2 (en) * | 1998-12-09 | 2008-04-02 | 富士通株式会社 | Error detector, communication system including the error detector, and error detection method |
US6681203B1 (en) * | 1999-02-26 | 2004-01-20 | Lucent Technologies Inc. | Coupled error code protection for multi-mode vocoders |
US6331978B1 (en) * | 1999-03-09 | 2001-12-18 | Nokia Telecommunications, Oy | Generic label encapsulation protocol for carrying label switched packets over serial links |
US7277388B1 (en) | 2000-03-31 | 2007-10-02 | Nokia Corporation | Method and apparatus for random packet marking for differentiated services |
US6701336B1 (en) * | 1999-11-12 | 2004-03-02 | Maxtor Corporation | Shared galois field multiplier |
EP1146650A1 (en) * | 2000-04-10 | 2001-10-17 | Hewlett-Packard Company, A Delaware Corporation | Error detection for data storage and transmission |
EP1146651A1 (en) * | 2000-04-10 | 2001-10-17 | Hewlett-Packard Company | Error detection for data storage and transmission |
US6477777B1 (en) | 2000-06-29 | 2002-11-12 | Joseph A. Oliveri | Gravity knife and backbone |
US6772390B2 (en) | 2000-11-30 | 2004-08-03 | Quantum Corporation | Erasure correction for ECC entities |
US6836869B1 (en) | 2001-02-02 | 2004-12-28 | Cradle Technologies, Inc. | Combined cyclic redundancy check (CRC) and Reed-Solomon (RS) error checking unit |
US6868517B1 (en) * | 2001-05-15 | 2005-03-15 | Marvell International Ltd. | Method and apparatus for checking read errors with two cyclic redundancy check stages |
US20020199153A1 (en) * | 2001-06-22 | 2002-12-26 | Fall Thomas G. | Sampling method for use with bursty communication channels |
US6986092B2 (en) * | 2001-12-04 | 2006-01-10 | Qualcomm Inc. | Erasure-and-single-error correction decoder for linear block codes |
US7111228B1 (en) | 2002-05-07 | 2006-09-19 | Marvell International Ltd. | System and method for performing parity checks in disk storage system |
US9240810B2 (en) | 2002-06-11 | 2016-01-19 | Digital Fountain, Inc. | Systems and processes for decoding chain reaction codes through inactivation |
CN100539439C (en) | 2002-10-05 | 2009-09-09 | 数字方敦股份有限公司 | The system coding of chain reaction sign indicating number and decode system and method |
US7287102B1 (en) | 2003-01-31 | 2007-10-23 | Marvell International Ltd. | System and method for concatenating data |
US7007114B1 (en) * | 2003-01-31 | 2006-02-28 | Qlogic Corporation | System and method for padding data blocks and/or removing padding from data blocks in storage controllers |
US7181677B1 (en) * | 2003-02-10 | 2007-02-20 | Maxtor Corporation | System and method for producing data and ECC code words using a high rate restricted-symbol code |
US7064915B1 (en) | 2003-03-10 | 2006-06-20 | Marvell International Ltd. | Method and system for collecting servo field data from programmable devices in embedded disk controllers |
US7039771B1 (en) | 2003-03-10 | 2006-05-02 | Marvell International Ltd. | Method and system for supporting multiple external serial port devices using a serial port controller in embedded disk controllers |
US7870346B2 (en) | 2003-03-10 | 2011-01-11 | Marvell International Ltd. | Servo controller interface module for embedded disk controllers |
US7219182B2 (en) | 2003-03-10 | 2007-05-15 | Marvell International Ltd. | Method and system for using an external bus controller in embedded disk controllers |
US7492545B1 (en) | 2003-03-10 | 2009-02-17 | Marvell International Ltd. | Method and system for automatic time base adjustment for disk drive servo controllers |
US7076723B2 (en) | 2003-03-14 | 2006-07-11 | Quantum Corporation | Error correction codes |
US7290197B2 (en) * | 2003-06-03 | 2007-10-30 | Quantum Corporation | Correcting data using redundancy blocks |
CN100361222C (en) * | 2003-09-05 | 2008-01-09 | 三洋电机株式会社 | Circuit and method for generating error correction code |
KR101170629B1 (en) | 2003-10-06 | 2012-08-02 | 디지털 파운튼, 인크. | Error-correcting multi-stage code generator and decoder for communication systems having single transmitters or multiple transmitters |
US7228467B2 (en) * | 2003-10-10 | 2007-06-05 | Quantum Corporation | Correcting data having more data blocks with errors than redundancy blocks |
US7526691B1 (en) | 2003-10-15 | 2009-04-28 | Marvell International Ltd. | System and method for using TAP controllers |
US7139150B2 (en) | 2004-02-10 | 2006-11-21 | Marvell International Ltd. | Method and system for head position control in embedded disk drive controllers |
US7228490B2 (en) * | 2004-02-19 | 2007-06-05 | Quantum Corporation | Error correction decoder using cells with partial syndrome generation |
KR101205758B1 (en) | 2004-05-07 | 2012-12-03 | 디지털 파운튼, 인크. | File download and streaming system |
US7120084B2 (en) | 2004-06-14 | 2006-10-10 | Marvell International Ltd. | Integrated memory controller |
US8166217B2 (en) | 2004-06-28 | 2012-04-24 | Marvell International Ltd. | System and method for reading and writing data using storage controllers |
US8032674B2 (en) * | 2004-07-19 | 2011-10-04 | Marvell International Ltd. | System and method for controlling buffer memory overflow and underflow conditions in storage controllers |
US9201599B2 (en) | 2004-07-19 | 2015-12-01 | Marvell International Ltd. | System and method for transmitting data in storage controllers |
US7757009B2 (en) | 2004-07-19 | 2010-07-13 | Marvell International Ltd. | Storage controllers with dynamic WWN storage modules and methods for managing data and connections between a host and a storage device |
US7386661B2 (en) | 2004-10-13 | 2008-06-10 | Marvell International Ltd. | Power save module for storage controllers |
US7240267B2 (en) | 2004-11-08 | 2007-07-03 | Marvell International Ltd. | System and method for conducting BIST operations |
US7802026B2 (en) * | 2004-11-15 | 2010-09-21 | Marvell International Ltd. | Method and system for processing frames in storage controllers |
US7609468B2 (en) | 2005-04-06 | 2009-10-27 | Marvell International Ltd. | Method and system for read gate timing control for storage controllers |
US7676735B2 (en) | 2005-06-10 | 2010-03-09 | Digital Fountain Inc. | Forward error-correcting (FEC) coding and streaming |
US7590920B2 (en) * | 2005-08-05 | 2009-09-15 | Hitachi Global Storage Technologies Netherlands, B.V. | Reduced complexity error correction encoding techniques |
US7694208B2 (en) * | 2005-12-20 | 2010-04-06 | Quantum Corporation | Error correction algorithm using interleaved parity check and Reed-Solomon code |
KR100734307B1 (en) | 2006-01-23 | 2007-07-02 | 삼성전자주식회사 | Post-viterbi error correction method and apparatus for the same |
CN101686107B (en) | 2006-02-13 | 2014-08-13 | 数字方敦股份有限公司 | Streaming and buffering using variable FEC overhead and protection periods |
US9270414B2 (en) | 2006-02-21 | 2016-02-23 | Digital Fountain, Inc. | Multiple-field based code generator and decoder for communications systems |
US7610519B1 (en) * | 2006-03-03 | 2009-10-27 | Xilinx, Inc. | Vector generation for codes through symmetry |
US7971129B2 (en) | 2006-05-10 | 2011-06-28 | Digital Fountain, Inc. | Code generator and decoder for communications systems operating using hybrid codes to allow for multiple efficient users of the communications systems |
US9386064B2 (en) | 2006-06-09 | 2016-07-05 | Qualcomm Incorporated | Enhanced block-request streaming using URL templates and construction rules |
US9380096B2 (en) | 2006-06-09 | 2016-06-28 | Qualcomm Incorporated | Enhanced block-request streaming system for handling low-latency streaming |
US9432433B2 (en) | 2006-06-09 | 2016-08-30 | Qualcomm Incorporated | Enhanced block-request streaming system using signaling or block creation |
US9419749B2 (en) | 2009-08-19 | 2016-08-16 | Qualcomm Incorporated | Methods and apparatus employing FEC codes with permanent inactivation of symbols for encoding and decoding processes |
US9178535B2 (en) | 2006-06-09 | 2015-11-03 | Digital Fountain, Inc. | Dynamic stream interleaving and sub-stream based delivery |
US9209934B2 (en) | 2006-06-09 | 2015-12-08 | Qualcomm Incorporated | Enhanced block-request streaming using cooperative parallel HTTP and forward error correction |
WO2008002174A1 (en) * | 2006-06-28 | 2008-01-03 | Intel Corporation | Modification to meggitt decoder for burst error correction codes |
US7937643B1 (en) | 2006-09-18 | 2011-05-03 | Mediatek Inc. | Mobile communication device and data reception method |
KR101221913B1 (en) * | 2006-12-20 | 2013-01-15 | 엘지전자 주식회사 | Digital broadcasting system and data processing method |
US7836230B1 (en) * | 2007-02-14 | 2010-11-16 | Marvell International Ltd. | Managing multiple host requests in queued commands that corresponds to receipt of stored acknowledgement commands from the host |
US20080205229A1 (en) * | 2007-02-26 | 2008-08-28 | Yung-Chih Li | Method of identifying optical disc |
TWI332611B (en) * | 2007-05-24 | 2010-11-01 | Realtek Semiconductor Corp | Method for writing data in flash memory and error correction coding/decoding method thereof |
KR101129260B1 (en) | 2007-09-12 | 2012-03-27 | 디지털 파운튼, 인크. | Generating and communicating source identification information to enable reliable communications |
US9281847B2 (en) | 2009-02-27 | 2016-03-08 | Qualcomm Incorporated | Mobile reception of digital video broadcasting—terrestrial services |
US9288010B2 (en) | 2009-08-19 | 2016-03-15 | Qualcomm Incorporated | Universal file delivery methods for providing unequal error protection and bundled file delivery services |
US9917874B2 (en) | 2009-09-22 | 2018-03-13 | Qualcomm Incorporated | Enhanced block-request streaming using block partitioning or request controls for improved client-side handling |
US9626243B2 (en) * | 2009-12-11 | 2017-04-18 | Advanced Micro Devices, Inc. | Data error correction device and methods thereof |
US8339726B2 (en) * | 2009-12-22 | 2012-12-25 | Quantum Corporation | Unique media identifier |
US20110280311A1 (en) | 2010-05-13 | 2011-11-17 | Qualcomm Incorporated | One-stream coding for asymmetric stereo video |
US9596447B2 (en) | 2010-07-21 | 2017-03-14 | Qualcomm Incorporated | Providing frame packing type information for video coding |
US8806050B2 (en) | 2010-08-10 | 2014-08-12 | Qualcomm Incorporated | Manifest file updates for network streaming of coded multimedia data |
US9270299B2 (en) | 2011-02-11 | 2016-02-23 | Qualcomm Incorporated | Encoding and decoding using elastic codes with flexible source block mapping |
US8958375B2 (en) | 2011-02-11 | 2015-02-17 | Qualcomm Incorporated | Framing for an improved radio link protocol including FEC |
US9253233B2 (en) | 2011-08-31 | 2016-02-02 | Qualcomm Incorporated | Switch signaling methods providing improved switching between representations for adaptive HTTP streaming |
US9843844B2 (en) | 2011-10-05 | 2017-12-12 | Qualcomm Incorporated | Network streaming of media data |
US9294226B2 (en) | 2012-03-26 | 2016-03-22 | Qualcomm Incorporated | Universal object delivery and template-based file delivery |
WO2017007483A1 (en) * | 2015-07-09 | 2017-01-12 | Hewlett-Packard Development Company, L.P. | Multi-dimensional cyclic symbols |
TWI718710B (en) * | 2019-06-25 | 2021-02-11 | 慧榮科技股份有限公司 | Data storage device and non-volatile memory control method |
CN112130749B (en) | 2019-06-25 | 2023-12-22 | 慧荣科技股份有限公司 | Data storage device and non-volatile memory control method |
CN112130750B (en) | 2019-06-25 | 2023-11-07 | 慧荣科技股份有限公司 | Data storage device and non-volatile memory control method |
Family Cites Families (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4142174A (en) * | 1977-08-15 | 1979-02-27 | International Business Machines Corporation | High speed decoding of Reed-Solomon codes |
JPS5857781B2 (en) * | 1978-01-17 | 1983-12-21 | 三菱電機株式会社 | Encoding/decoding method |
US4359772A (en) * | 1980-11-14 | 1982-11-16 | International Business Machines Corporation | Dual function error correcting system |
US4413339A (en) * | 1981-06-24 | 1983-11-01 | Digital Equipment Corporation | Multiple error detecting and correcting system employing Reed-Solomon codes |
JPS6050669A (en) * | 1983-08-29 | 1985-03-20 | Hitachi Ltd | Demodulating system of data |
US4653051A (en) * | 1983-09-14 | 1987-03-24 | Matsushita Electric Industrial Co., Ltd. | Apparatus for detecting and correcting errors on product codes |
JPS60111531A (en) * | 1983-11-21 | 1985-06-18 | Matsushita Electric Ind Co Ltd | Error correcting method |
WO1985002958A1 (en) * | 1983-12-20 | 1985-07-04 | Sony Corporation | Method and apparatus for decoding error correction code |
JPS6276825A (en) * | 1985-09-30 | 1987-04-08 | Hitachi Ltd | Code error correcting method |
CA1264091A (en) * | 1986-01-10 | 1989-12-27 | Yoichiro Sako | Generator for error correcting code and decoder for the code |
US4763332A (en) * | 1987-03-02 | 1988-08-09 | Data Systems Technology Corp. | Shared circuitry for the encoding and syndrome generation functions of a Reed-Solomon code |
JPS63267018A (en) * | 1987-04-24 | 1988-11-04 | Ricoh Co Ltd | System and device for error correction of long distance code |
US4856003A (en) * | 1987-05-07 | 1989-08-08 | Digital Equipment Corporation | Error correction code encoder |
US4866716A (en) * | 1987-05-15 | 1989-09-12 | Digital Equipment Corporation | Real-time BCH error correction code decoding mechanism |
US4975867A (en) * | 1987-06-26 | 1990-12-04 | Digital Equipment Corporation | Apparatus for dividing elements of a Galois Field GF (2QM) |
-
1989
- 1989-06-28 US US07/373,323 patent/US5136592A/en not_active Expired - Lifetime
-
1990
- 1990-05-15 CA CA002016847A patent/CA2016847A1/en not_active Abandoned
- 1990-06-28 EP EP19900307118 patent/EP0407101A3/en not_active Withdrawn
- 1990-06-28 JP JP2171304A patent/JPH03136524A/en active Pending
Also Published As
Publication number | Publication date |
---|---|
JPH03136524A (en) | 1991-06-11 |
US5136592A (en) | 1992-08-04 |
EP0407101A3 (en) | 1992-04-08 |
EP0407101A2 (en) | 1991-01-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CA2016847A1 (en) | Error detection and correction systems for long burst errors | |
US5724368A (en) | Cyclical redundancy check method and apparatus | |
US4525838A (en) | Multibyte error correcting system involving a two-level code structure | |
US6615387B1 (en) | Method and apparatus for error detection | |
US5491701A (en) | Burst error corrector | |
KR930008683B1 (en) | Reed solomon error correction code encoder | |
US7278085B1 (en) | Simple error-correction codes for data buffers | |
CA1159959A (en) | Dual function error correcting system | |
US7404134B2 (en) | Encoding/decoding device using a reed-solomon encoder/decoder | |
CA1199410A (en) | On-the-fly multibyte error correcting system | |
US5872799A (en) | Global parity symbol for interleaved reed-solomon coded data | |
US5946328A (en) | Method and means for efficient error detection and correction in long byte strings using integrated interleaved Reed-Solomon codewords | |
EP0218413A2 (en) | System and method for correcting errors in coded data recorded in blocks and sub-blocks | |
US5856987A (en) | Encoder and decoder for an SEC-DED-S4ED rotational code | |
US5901158A (en) | Error correction encoder/decoder | |
KR19990028201A (en) | Dedicated ALU Architecture for 10-Bit Reed-Solomon Error Correction Module | |
EP0438907A2 (en) | Improved error trapping decoding method and apparatus | |
EP0909484A2 (en) | Modified reed solomon code selection and encoding system | |
Wolf | An introduction to tensor product codes and applications to digital storage systems | |
EP1213841A2 (en) | Erasure correction for ECC entities | |
US5889794A (en) | Two-level error correction encoder | |
EP0341851A2 (en) | Method and apparatus for interleaved encoding | |
Al Jabri et al. | Array Codes with Error Detection For E cient Storage Systems | |
Kamali et al. | A hybrid decoding method for short-blocklength Reed Solomon codes | |
Bateman | An Introduction to Error Correction Codes |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
EEER | Examination request | ||
FZDE | Discontinued |