Circuitry and method for correcting 3-bit errors containing adjacent 2-bit error

Abstract

A circuitry is proposed for the correction of errors in a possibly erroneous binary word v′=v′1, . . . , v′n relative to a codeword v=v1, . . . , vn, in particular 3-bit errors containing an adjacent 2-bit error (burst error). The circuitry comprises a syndrome generator and a decoder. A modified BCH is used wherein n′ column vectors of a first BCH code submatrix are paired as column vector pairs so that a componentwise XOR combination of the two column vectors of each column vector pair produces an identical column vector K that is different from all column vectors of the first BCH submatrix. A second BCH submatrix comprises corresponding column vectors as the third power, according to Galois field arithmetic, of the column vector in the first BCH submatrix. The syndrome generated by the syndrome generator can be checked against the columns of the first and second submatrices.

Description

BACKGROUND

It is known, in binary sequences or binary words of a certain length N, to correct 1-bit errors and any 2-bit errors using BCH codes by combinational error correction circuits as it is for example described in Okano, H. and Imai, H., “A construction method of high speed decoders using ROM's for Bose-Chadhuri-Hocquenghem and Reed Solomon Codes”, IEEE Trans. Comp. C36 (10) 1165-1175, 1987.

If BCH codes are used over a Galois field GF(2^m), then N≦2^m−1 and the error syndrome s may consist of 2m components, wherein the first m components form the sub-syndrome s₁ and the second m components form the sub-syndrome s₃ as it is common when using BCH codes. If the overall parity is considered, the error syndrome comprises a further binary component which is to be designated by s_P.

It is further known to correct any 3-bit errors by combinational error correction circuits using BCH codes, as it is for example also described in the document by Okano, H. and Imai, H., “A construction method of high speed decoders using ROM's for Bose-Chadhuri-Hocquenghem and Reed Solomon Codes”, IEEE Trans. Comp. C36 (10) 1165-1175, 1987. When correcting any 3-bit errors, in addition to the sub-syndromes s₁ and s₃ a further sub-syndrome s₅ may be used which like the sub-syndromes s₁ and s₃ that generally also comprises a word width of m bits, so that the error syndrome s=s₁, s₃, s₅ generally comprises a word width of 3·m and taking into account the overall parity comprises a word width of 3m+1.

The correction of any 3-bit errors by combinational error correction circuits is here connected with a relatively high hardware effort and a relatively great signal runtime for the determination of the corresponding error correction signals, which may be disadvantageous. In particular, a relatively long signal runtime for the correction signals may be limiting with respect to the clock rate.

For certain circuitries it may be more likely that, when 3 bits in a binary word are erroneous, two of these erroneous bits occur in specific bit positions than that all three erroneous bits are randomly distributed.

An example for one such case may be a data storage whose memory cells may take on more than two states, in general one multivalued state each. If a data storage for example takes on 4 different states, then a memory state of a cell stores the information of two particular bits. These bits which are stored in the same memory cell are here designated as neighboring bits. Generally these bits will also be spatially adjacent in the data word to be stored. It is of course also possible to store the information of two bits which are not directly adjacent in the data word in one memory cell, like the value of the first and the seventh bits, the second and the thirteenth bits, etc. In this way of speaking, the first and the seventh bit and the second and the thirteenth bit are adjacent. In order to make the description as simple as possible, it is in the following always assumed, that neighboring bits which are for example stored in a memory cell are also spatially neighboring in the considered data word. If this is not the case, it may be acquired by a simple exchange of the bits of the data word that neighboring bits are also spatially neighboring. Thus, it is not necessary in the following to differentiate between neighboring bits which are neighboring as they are for example stored in the same memory cell and neighboring bits which are spatially neighboring or adjacent in the considered binary word.

If now an error of a memory state occurs which may e.g. take on 4 states, then two neighboring bits corresponding to this state may both be erroneous at the same time. If the assignment of the multivalued state of a memory cell to binary values is done via a Gray code, as it is common practice and for example proposed in Rupprecht, W., Steinbuch, K., “Nachrichtentechnik”, pp 339-341, Springer Verlag 1967, then errors in state values of a memory cell which only slightly change a memory state into a physically neighboring state value, lead to a 1-bit error in one of the binary values assigned with the memory state.

Errors in a memory state of a memory cell which corrupt the correct memory state into a non-neighboring memory state lead to a 2-bit error in the neighboring binary values of the binary data assigned with the memory state.

It is possible that the bits of a word which are to be corrected are auxiliary binary read values of a ternary memory or a multiple-valued memory as it is described in the U.S. patent application Ser. No. 13/664,495, filed Oct. 31, 2012 and entitled “Circuit and Method for Multi-Bit Correction” which is included here in the description by reference in its entirety.

It would be beneficial to provide an error correction of a 2-bit error in two neighboring bits (or otherwise related to each other) with relatively low hardware effort and/or relatively short signal runtime. It would also be beneficial to provide an error correction of a 2-bit error in two neighboring bits and an additional 1-bit error in an arbitrary bit position.

SUMMARY

Embodiments of the present invention provide a circuitry for the correction of errors in a possibly erroneous binary word v′=, v′₁, . . . v′_n relative to a codeword v=v₁, . . . , v_n. The circuitry comprises a syndrome generator for determining an error syndrome s=(s₁, s₃) according to a modified BCH code with a H-matrix H^mod comprising a first BCH submatrix H₁^mod and a second BCH submatrix H₃^mod and with a code distance d≧5, wherein n′ column vectors of the submatrix H₁^mod are paired as column vector pairs so that a component-wise XOR combination of the two column vectors of each column vector pair produces an identical column vector K that is different from all column vectors of the first BCH submatrix H₁^mod and where n′ is even and 4≦n′≦n applies. The second BCH submatrix H₃^mod comprises a corresponding column vector for each column vector in the first BCH submatrix H₁^mod so that the corresponding column vector is the third power, according to Galois field arithmetic, of the column vector in the first BCH submatrix H₁^mod. The syndrome generator is configured to determine the error syndrome s by multiplying the H-matrix H^mod with the possibly erroneous binary word v′ so that a first error syndrome portion is given by s₁=H₁^mod·v′ and a second error syndrome portion is given by s₃=H₃^mod·v′. The circuitry further comprises a decoder for generating a correction vector e=(e₁, . . . , e_n) with correction values e_j=e_j+1=e_l=1 and e_t=0 for t≠j, j+1, l, if the first error syndrome portion s₁ equals the componentwise XOR combination of the identical column vector K and a column vector at a column position l of the first BCH submatrix H₁^mod, and if the second error syndrome portion s₃ equals the componentwise XOR combination of column vectors at column positions j, j+1, and l of the second BCH submatrix H₃^mod.

Further embodiments of the present invention provide a method for correcting errors in a possibly erroneous binary word v′=v′₁, . . . , v′_n relative to a codeword v=v₁, . . . , v_n. The method comprises determining an error syndrome s=(s₁, s₃) of a modified BCH code with a H-matrix H^mod comprising a first BCH submatrix H₁^mod and a second BCH submatrix H₃^mod, and with a code distance d≧5, wherein n′ column vectors of the submatrix H₁^mod are paired as column vector pairs so that a componentwise XOR combination of the two column vectors of each column vector pair produces an identical column vector K that is different from all column vectors of the BCH submatrix H₁^mod, and where n′ is even and 4≦n′≦n applies. The second BCH submatrix H₃^mod comprises a corresponding column vector for each column vector in the first BCH submatrix H₁^mod so that the corresponding column vector is the third power, according to Galois field arithmetic, of the column vector in the first BCH submatrix H₁^mod. The error syndrome s is determined by multiplying the H-matrix H^mod with the possibly erroneous binary word v′ so that a first error syndrome portion is given by s₁=H₁^mod·v′ and a second error syndrome portion is given by s₃=H₃^mod·v′. The method further comprises a step of generating the correction vector e=(e₁, . . . , e_n) with correction values e_j=e_j+1=e_l=1 and e_t=0 for t≠j+1, l, if s₁ equals the componentwise XOR combination of the identical column vector K and a column vector at a column position 1 of the first BCH submatrix H₁^mod, and s₃ equals the component-wise XOR combination of column vectors at column positions j, j+1, and 1 of the second BCH submatrix H₃^mod.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the present invention will be described using the accompanying figures, in which:

FIG. 1 a shows a schematic block diagram of a circuitry for correcting errors in a binary word according to embodiments;

FIG. 1 b shows a schematic block diagram of a circuitry for correcting errors according to further embodiments in which the overall parity of the binary word is evaluated and used;

FIG. 1 c shows a schematic block diagram of a circuitry for correcting errors according to further embodiments in which the overall parity can be determined from the sub-syndrome s₁, because the submatrix H₁^mod only has columns which comprise an odd number of ones;

FIG. 1 d shows a schematic block diagram of a circuitry for correcting errors comprising an error indication circuit;

FIG. 1 e shows a schematic block diagram of a circuitry for correcting errors that uses a further submatrix H₅^mod in order to differentiate between 3-bit errors with neighboring 2-bit errors and other (non-correctable) 3-bit errors;

FIG. 2 a shows a schematic block diagram of a decoder according to some embodiments;

FIG. 2 b shows a schematic block diagram of a subcircuit for forming auxiliary signals;

FIG. 2 c shows a schematic block diagram of another subcircuit for forming auxiliary signal;

FIG. 3 a shows a possible implementation of the subcircuit for forming auxiliary signals in FIG. 2a;

FIG. 3 b shows a possible implementation of the subcircuit for forming auxiliary signals in FIG. 2b;

FIG. 3 c shows a possible implementation of the subcircuit for forming auxiliary signals in FIG. 2a having further outputs for outputting signals “1-bit error”, “2-bit error”, and “3-bit error”;

FIG. 4 shows a schematic logic gate diagram of a possible implementation of a decoder subcircuit for a single correction value e_j;

FIG. 5 shows a schematic implementation of a subcircuit for realizing the function z=s₁²K⊕s₁K²;

FIG. 6 shows a schematic implementation of a subcircuit for realizing the function u=s₁·α⁸+s₁²·α⁴; and

FIG. 7 shows a schematic flow diagram of a method for correcting errors in a binary word according to embodiments.

DETAILED DESCRIPTION

A theoretical background of the embodiments of the invention will be described first.

For the error correction of randomly distributed multi-bit errors, BCH codes may be used, as it is known to a person skilled in the art and for example described in Lin, S., Costello, D. “Error Control Coding” Prentice Hall, 1983, wherein reference is in particular made to pages 143-160. Likewise reference is to be made to the document by Okano, H. and Imai, H., “A construction method of high speed decoders using ROM's for Bose-Chadhuri-Hocquenghem and Reed Solomon Codes”, IEEE Trans. Comp. C36 (10) 1165-1175, 1987, in which combinational circuits for error correction of BCH codes are represented.

A BCH code is a special linear code which may be described by a parity check matrix H and a generator matrix G which may be derived from the parity check matrix. If a code comprises the length N and if it comprises k information bits, then H is an M,N matrix comprising M lines and N columns, wherein M=N−k. The generator matrix G is then a k,N matrix comprising k lines and N columns, and the code comprises M checkbits.

An unshortened 2-bit error correcting BCH code may be described by an H matrix

$H=\left(\begin{array}{c}{H}_1\\ H_3\end{array}\right)=\left(h_1,\dots ,h_N\right)$ wherein the H matrix is represented in a separate form. Conventionally, H₁ and H₃ are selected to beH₁=(α⁰,α¹, . . . ,αⁱ, . . . ,α^N-1)=(h₁¹, . . . ,h_i¹, . . . ,h_N¹)andH₃=(α⁰,α³, . . . ,α³ⁱ, . . . ,α^3(N-1))=(h₁³, . . . ,h_i³, . . . ,h_N³)when the code is not shortened. Here, α is an element of the Galois field GF(2^m) and may be selected as a primitive element of a finite field GF(2^m), also called Galois field. Here N=2^m−1 applies. The exponents of α^j and of α^3j are here determined to be modulo 2^m−1. H₁ and H₃ are each (m,N) matrices with m lines and N=2^m−1 columns. The elements αⁱ of the Galois field GF(2^m) in their vector representation are m digit binary column vectors.

If L columns of the H matrix of the unshortened BCH code are eliminated, an H matrix of a shortened BCH code of the length n=N−L is obtained. For a shortened code the following applies n=N−L<2^m−1.

It is possible to supplement the H matrix H by a line of only ones. The additional integration of the overall parity corresponds to an additional line of only ones in the H matrix.

An H matrix considering the overall parity may correspond to that of the H matrix

$H=\left(\begin{array}{c}{H}_1\\ H_3\\ P\end{array}\right)$ wherein P is a line of ones only.

Now a shortened BCH code of the length n is considered, wherein n=N−L<2^m−1 applies. A codeword of this code v=v₁, . . . , v_n which is also referred to as a code vector consists of n components v₁, v₂, . . . , v_n. Here, a code vector may be described as a line vector or as a column vector. If a matrix is multiplied with a vector from the right, then the vector is to be interpreted as a column vector and the result is a column vector. In such a case, it is not required to explicitly highlight the corresponding vectors as being column vectors, as this is clear from the context. If it is to be especially highlighted that a vector w is represented as a column vector, it is written as w^T. If a codeword v=v₁, . . . , v_n is corrupted into a word v′=v′₁, . . . , v′_n, then the difference between v and v′ may be described by an error vector e withe=(e₁, . . . ,e_n)=(v₁⊕v′₁, . . . ,v_n⊕v′_n)=v⊕v′. A component e_i of the error vector e is equal to 1, when v_i and v′_i are different and v_i=v′_i⊕1= v_i′ applies. A component e_j of the error vector e is equal to 0, when v_j and v′_j are the same and when v_j=v′_j applies.

If an error may be corrected by an error correction circuit, then the correction values output by the error correction circuit are equal to the components of the error vector and the correction circuit outputs the correction value e_i at the i-th output of its n outputs. The correction values e_i may also be combined into a correction vector. The correction vector is equal to the error vector if the error may be corrected by the code.

The error syndrome s=(s₁, s₃, s_P) of a word v′ is determined bys=H·v′  (1)whereins₁=H₁·v′  (2)s₃=H₃·v′  (3)ands_P=P·v′=(1, . . . ,1)·v′=v′₁⊕v′₂⊕ . . . ⊕v′_n  (4)applies.

The error syndrome of a codeword v is equal to 0, so that for a codeword v the following appliess=H·v=0  (5)and for a non-codeword v′=v⊕es=H·v′=H·(v⊕e)=H·v⊕H·e=H·e,  (6)and the error syndrome s is determined by the error vector e.

In order to determine the assigned correct codeword from the erroneous, interfered with non-codeword v′, those components v′_j are to be inverted for which e_j=1 applies, so that e_j is the corresponding correction value determined by the error correction circuit.

For the considered shortened BCH code, e is to be determined from the error syndrome s=s₁, s₃, s_P with the H matrix

$H=\left(\begin{array}{c}{H}_1\\ H_3\\ P\end{array}\right)$

Here, the following applies

$\begin{array}{cc}{H}_1=\left(h_1^1,h_2^1,\dots ,h_n^1\right)=\left({\alpha }^{i_1},{\alpha }^{i_2},\dots ,{\alpha }^{i_n}\right)& \left(7\right)\\ H_3=\left(h_1^3,h_2^3,\dots ,h_n^3\right)=\left({\alpha }^{3\left(i_1\right)},{\alpha }^{3\left(i_2\right)},\dots ,{\alpha }^{3\left(i_n\right)}\right)\mathrm{and}& \left(8\right)\\ P=\underset{\underset{n}{︸}}{\left(1,1,\dots ,1\right)},& \left(9\right)\end{array}$ wherein the exponents of α, as already indicated, are determined modulo 2^m−1, and the exponents i₁, i₂, . . . , i_n are all different pair-wise. It is not necessary here that i_j=j applies for j=1, . . . , n.

If a 1-bit error exists in the j-th bit, the following appliess₁=α^ij,  (10)s₃=α^3(ij),  (11)s_P=1  (12)and s₁³=s₃. If a 2-bit error exists in the bit positions j and l, the following appliess₁=α^ij⊕α^il,  (13)s₃=α^3(ij)⊕α^3(il),  (14)s_P=0  (15)and s₁³≠s₃.

For a BCH code the error positions of a 1-bit error or of any 2-bit error are determined from the syndrome components s₁ and s₃ of a BCH code. In this respect, a special process for determining the error position is described.

From (13) the following resultsα^3(il)=s₁³⊕s₁α^2(ij)⊕s₁²α^ij⊕α^3(ij) and with (14)s₁³⊕s₃⊕s₁α^2(ij)⊕s₁²α^ij=0  (16)and completely analog to thats₁³⊕s₃⊕s₁α^2(il)⊕s₁²α^il=0  (17)so that a 2-bit error in the positions j and l is determined by the two zeros or roots of the quadratic equations₁³⊕s₃⊕s₁x²⊕s₁²x=0  (18)in the Galois field GF(2^m), wherein s₁=α^ij⊕α^il and s₃=α^3ij⊕α^3il.

If now a 1-bit error is present in the j-th bit position, then s₁³⊕s₃=0 and the two solutions of equation (18) in this case are x=0 and x=s₁=α^ij, so that the position j of the 1-bit error is also determined to be a solution of equation (18) which is unequal 0.

Assuming that only 1-bit, 2-bit and 3-bit errors occur, the following applies:

A 1-bit error exists when s₁³=s₃, s₁≠0 and s_P=1.

A 2-bit error exists when s₁³≠s₃ and s_P=0.

A 3-bit error exists when s₁³≠s₃ and s_P=1.

If s₁=s₃=0, no error exists.

In summary, in case of a 1-bit error or a 2-bit error the bit positions of the errors are determined by the solutions of equation (18) which are unequal 0 and thus from the sub-syndromes s₁ and s₃.

Up to today, no method and no circuitry are known which enable to determine the error positions of a 3-bit error and thus to correct the same from knowing the sub-syndromes s₁, s₃ and s_P of a BCH code.

Surprisingly, it is possible according to the invention to determine the error positions of a 3-bit error from the sub-syndromes s₁, s₃ and s_P when the 3-bit error contains a 2-bit error in neighboring bits, wherein the 2-bit errors occur in certain bit pairs. Such 2-bit errors are for example of interest when bits of a binary sequence are stored in memory cells which may take on more than two states.

To be able to correct the described 3-bit errors with a neighboring 2-bit error, from the original H matrix H of the shortened BCH code a modified H matrix H^mod with the submatrices H₁^mod and H₃^mod is formed by rearranging the columns h′₁, . . . , h′_n of the original H matrix H=(h′₁, . . . , h′_n).

The columns h₁, h₂, . . . , h_n of the modified H matrix H^mod are determined so that for the columns h₁¹, h₂¹, . . . , h_n¹ of the submatrix H₁^mod for n being even the following appliesh₁¹⊕h₂¹=h₃¹⊕h₄¹=h₅¹⊕h₆¹= . . . =h_n-1¹⊕h_n¹=K  (19)and for n being odd the following appliesh₁¹⊕h₂¹=h₃¹⊕h₄¹=h₅¹⊕h₆¹= . . . =h_n-2¹⊕h_n-1¹=K  (20)

Here, K is selected so that K is no column of the submatrix H₁.

With h_j¹=α^ij, equations (19) and (20) correspond to the equationsαⁱ¹⊕αⁱ²=αⁱ³⊕αⁱ⁴=αⁱ⁵⊕αⁱ⁶= . . . =α^in-1⊕αⁱⁿ=K  (21)andαⁱ¹⊕αⁱ²=αⁱ³⊕αⁱ⁴=αⁱ⁵⊕αⁱ⁶= . . . =α^in-2⊕α^in-1=K.  (22)

For a simplest possible description, the different pairs of neighboring bits were here designated as pairs [1, 2], [3, 4], [5, 6], . . . . If the pairs of neighboring bits are for example designated as pairs [1, 7], [2, 11], [3, 5], . . . , thenh₁¹⊕h₂¹=h₃¹⊕h₄¹=h₅¹⊕h₆¹= . . . =K in equations (19) or (20) is simply replaced by the conditionh₁¹⊕h₇¹=h₂¹⊕h₁₁¹=h₃¹⊕h₅¹= . . . =K

If a 3-bit error exists in positions j, j+1 and 1, wherein jε{1, 3, 5, . . . ,}, then the following applies for the sub-syndrome s₁ s₁=α^ij⊕α^ij+1⊕α^il  (23)and withα^ij⊕α^ij+1=K  (24)α^il=s₁⊕K,  (25)from whichα^3(il)=s₁³⊕s₁K²⊕s₁²K⊕K³  (26)results.

For the sub-syndrome s₃ the following applies

$\begin{array}{cc}{s}_3={\alpha }^{3\left(i_j\right)}\oplus {\alpha }^{3\left(i_{j+1}\right)}\oplus {\alpha }^{3\left(i_l\right)}& \left(27\right)\\ ={\alpha }^{3\left(i_j\right)}\oplus {\alpha }^{3\left(i_{j+1}\right)}\oplus s_1^3\oplus s_1{K}^2\oplus s_1^{2}K\oplus K^3.& \left(28\right)\end{array}$

From equation (24) the following resultsα^3(ij+1)=(K⊕α^ij)³=K³⊕Kα^2(ij)⊕K²α^ij⊕α^3(ij)  (29)and from equation (28) for α^ij the following resultss₁³⊕s₃⊕s₁K²⊕s₁²K⊕Kα^2(ij)⊕K²α^ij=0.  (30)Completely analog to that, for α^ij+1 the following resultss₁³⊕s₃⊕s₁K²⊕s₁²K⊕Kα^2(ij+1)⊕K²α^ij+1=0.  (31)so that α^ij and α^ij+1 are solutions of the quadratic equations₁³⊕s₃⊕s₁K²⊕s₁²K⊕Kx²⊕K²x=0.  (32)in the Galois field GF(2^m).

It is thus possible to determine a 3-bit error containing a neighboring 2-bit error merely from knowing the syndrome components as s₁ and s₃, wherein for a 1-bit error the following appliess₁³=s₃ and P=1  (33)For a 2-bit error the following appliess₁³≠s₃ and P=0  (34)and for a 3-bit error the following appliess₁³≠s₃ and P=1.  (35)

The correction of a 3-bit error as a 3-bit error containing a neighboring 2-bit error may in particular be useful when 3-bit errors with neighboring 2-bit errors occur more frequently than 3-bit errors randomly distributed across the n bits, as it may be the case when the bits of binarily encoded data are stored in memory cells which store more than 1 bit per memory cell, for example 2 bits or 4 values per memory cell.

If only a subset of neighboring two-bit errors is to be corrected it is sufficient that a modified H-matrix H^mod is determined in such a way that only for a subset of n′ columns, where 4≦n′≦n applies, the corresponding columns h_i¹, 1≦i≦n′ of the submatrix H^mod are determined such thath₁¹⊕h₂¹=h₃¹⊕h₄¹= . . . =h_n′−1¹⊕h_n′¹=K holds. Thereby K is no column of the submatrix H₁^mod.

It is not necessary that for the remaining n−n′ columns of H₁^mod this condition is valid.

Here the corresponding columns of the submatrix H₁^mod are denoted by h₁¹, h₂¹, h₃¹, . . . , h_n′¹. For any other n′ columns a similar condition can be formulated.

If, for instance for n′=6 only the neighboring two-bit errors in the positions (5, 6), (9, 10) and (12, 13) together with a 1-Bit error in an arbitrary position has to be corrected, the submatrix H₁^mod has to be determined such that we haveh₅¹⊕h₆¹=h₉¹⊕h₁₀¹= . . . =h₁₂¹⊕h₁₃¹=K where K is no column of the submatrix H₁^mod.

If, for example, only errors in the data bits have to be corrected, then the columns h₁¹, h₂¹, h₃¹, . . . , h_n′¹ are columns of the submatrix H₁^mod which correspond to data bits of the corrected word.

It may also be advantageous to correct 3-bit errors containing a neighboring 2-bit error as fast as possible and with a smallest possible hardware effort and to detect 3-bit errors which do not contain a neighboring 2-bit error as non-corrected or as not adequately corrected 3-bit errors. The H matrix H^mod with

$H^{\mathrm{mod}}=\left(\begin{array}{c}{H}_1^{\mathrm{mod}}\\ H_3^{\mathrm{mod}}\\ P\end{array}\right)$ and with H₁(h₁¹, h₂¹, . . . , h_n¹)=αⁱ¹, αⁱ², . . . , αⁱⁿ may be supplemented by a further submatrix H₅^mod=(α⁵⁽ⁱ¹⁾, α⁵⁽ⁱ²⁾, . . . , α⁵⁽ⁱⁿ⁾) into

$\begin{array}{cc}{H}^{\mathrm{mod}}=\left(\begin{array}{c}{H}_1^{\mathrm{mod}}\\ H_3^{\mathrm{mod}}\\ H_5^{\mathrm{mod}}\\ P\end{array}\right)& \left(36\right)\end{array}$

The number of the checkbits may then be no more than 3m+1. It is possible that lines of the submatrix H₅^mod are linearly dependent so that only the linearly independent lines of the matrix H₅^mod are to be considered in the calculation of s₅. By the submatrix H₅^mod then a further syndrome component s₅ is determined bys₅=H₅^modv′  (37)which may include no more than m components.

If now the error correction of a 3-bit error is executed assuming a 3-bit error with a neighboring 2-bit error, then it may now be checked using the submatrix H₅^mod whether the correction was executed properly. If the positions j, j+1 and 1 for a 3-bit error with a neighboring 2-bit error were determined merely on the basis of the syndrome components s₁ and s₃, then after the already executed error correction of the data word v′ into the corrected word v^cor it only remains to be checked whether the following applies for the corrected word v^cor H₅^mod·v^cor=H₅^mod·(v′⊕e)=0  (38)

If this relation is fulfilled, the 3-bit error was corrected properly. If this relation is not fulfilled, the 3-bit error is indicated as a non-correctly corrected 3-bit error.

Apart from that it is possible to detect 4-bit errors (and to differentiate same from a 2-bit error), as for a 4-bit errors₁³s₃⊕s₁·s₅⊕s₃²⊕s₁⁶≠0,and P=0  (39)and for a 2-bit errors₁³s₃⊕s₁·s₅⊕s₃²⊕s₁⁶=0,P=0 and s₁³⊕s₃≠0  (40)applies.

One further possibility to determine whether a corrected 3-bit error is a 3-bit error with a neighboring 2-bit error is to be described now.

For a 3-bit error in the bit positions j, l, r the following appliess₁=α^ij⊕α^il⊕α^ir,  (41)s₃=α^3ij⊕α^3il⊕α^3ir,  (42)s₅=α^5ij⊕α^5il⊕α^5ir,  (43)

If a neighboring 2-bit error exists in the positions j and l, then α^ij⊕α^il=K and thus α^ir=s₁⊕K, so that the following appliess′₁=α^ij⊕α^il=s₁⊕α^ir=K,  (44)s′₃=α^3ij⊕α^3il=s₃⊕α^3ir=s₃⊕s₁³⊕s₁K²⊕Ks₁²⊕K³,  (45)s′₅=α^5ij⊕α^5il=s₅⊕α^5ir=s₅⊕s₁⁵⊕s₁K⁴⊕s₁⁴K⊕K⁵,  (46)and the components s′₁, s′₃, s′₅ of the error syndrome describe a 2-bit error, so thats′₁³s′₃⊕s′₁s′₅⊕s′₃²⊕s′₁⁶=0  (47)has to apply, from which using equations (44), (45) and (46) the following results for a 3-bit error with a neighboring 2-bit errors₁⁶⊕s₃²⊕s₁⁵K⊕Ks₅⊕K³s₃⊕K³s₁³=0.  (48)

If this is originally a 3-bit error comprising no neighboring 2-bit error, the following appliess₁⁶⊕s₃²⊕s₁⁵K⊕Ks₅⊕K³s₃⊕K³s₁³≠0.  (48)

The possibilities of error detection now ought to be summarized again. First of all, the case is considered in which the overall parity P has not been determined.

1. For a 1-bit error the following applies:s₁³=s₃.2. For a 2-bit error the following applies:s₁³≠s₃,s₁³s₃⊕s₁s₅⊕s₃²⊕s₁⁶=0.3. For a random 3-bit error the following applies:s₁³≠s₃,s₁³s₃⊕s₁s₅⊕s₃²⊕s₁⁶≠0.4. For a 3-bit error having a neighboring 2-bit error the following applies:s₁⁶s₃²⊕s₁⁵K⊕Ks₅⊕K³s₃⊕K³s₁³=0.

Now the case is considered that the overall parity P has been determined.

1. For a 1-bit error the following applies:s₁³=s₃,P=1.2. For a 2-bit error the following applies:s₁³≠s₃,P=0.3. For a random 3-bit error the following applies:s₁³≠s₃,P=1.4. For a 3-bit error with a neighboring 2-bit error additionally the following applies:s₁⁶⊕s₃²⊕s₁⁵K⊕Ks₅⊕K³s₁³⊕K³s₁³=0.

A 4-bit error may be differentiated from a 2-bit error in that for a 4-bit error the following appliess₁³⊕s₁·s₅⊕s₃²⊕s₁⁶≠0,and P=0and for a 2-bit errors₁³⊕s₁·s₅⊕s₃²⊕s₁⁶=0,and P=0applies.

According to a possible aspect of the invention, the correction of 1-bit, 2-bit and 3-bit errors in binary data words is facilitated and in particular the error correction circuits required here can be improved so that a least possible hardware effort is required and the signal runtime for determining the correction values is as low as possible.

Apart from that or in addition, in a partial aspect of the invention, the error detection of non-correctable errors is to be enabled, in particular of non-correctable 3-bit errors.

According to another aspect of the invention, the error correction of any 1-bit errors, of any 2-bit errors, and of 3-bit errors containing a neighboring 2-bit error can be enabled in a simple way.

The invention is now to be explained with respect to embodiments.

FIG. 1 a shows an inventive circuitry comprising a syndrome generator Synd 12 and a downstream decoder Dec 14. The syndrome generator Synd 12 comprises a n-bit wide input 11 for inputting an n digit binary word v′=v′₁, . . . , v′_n and m′ binary outputs 13 for outputting an m′ digit error syndrome s having two m digit sub-syndromes s₁ and s₃. The m′ binary outputs 13 of the syndrome generator Synd 12 are connected to m′ binary inputs of the decoder Dec 14 outputting an n digit correction vector e=e₁, . . . , e_n at its n digit output 15. Here, m′≧2m and n≦2^m−1 apply. The syndrome generator Synd 12 is here configured so that it outputs an m′ component error syndrome at m′ binary outputs when inputting the binary word v′. Here, s is determined according to the relations=H^mod·v′.  (50)

The matrix H^mod=h′₁, . . . , h′_n is an (m′, n) matrix of n columns h′₁, . . . , h′_n each comprising m′ components. 2m first components of these columns are designated by h₁, . . . , h_n. For these 2m-component columns h_j for j=1, . . . , n the following applies

$\begin{array}{cc}{h}_j=\{\begin{array}{c}{\alpha }^{i_j}\\ {\alpha }^{3\left(i_j\right)}.\end{array}& \left(51\right)\end{array}$

Further, for n′ being even for j=1, 3, 5, . . . , n′−1 and for n′≦n for pairs of bits [j, j+1] which are neighboring in the data word v′=v′₁, . . . , v′_n the H matrix H is determined so that the following appliesα^ij⊕α^ij+1=K.  (52)

Sub-syndromes s₁, s₃ are determined bys₁=H₁^mod·v′  (53)s₃=H₃^mod·v′,  (54)wherein the (m, n) submatrices H₁^mod and H₃^mod are determined byH₁^mod=(αⁱ¹, . . . ,αⁱⁿ)H₃^mod=(α³⁽ⁱ¹⁾, . . . ,α³⁽ⁱⁿ⁾).

Here, α is an element of the Galois field GF(2^m), which is an m-component binary vector in the vector representation, and the exponents of α are to be interpreted modulo 2^m−1. K is an m-digit binary vector equal to no column of the submatrix H₁^mod, i.e. for which K≠α^ij for j=1, . . . , n applies. Preferably, α is a primitive element of the Galois field GF(2^m). If n is even n′=n may apply and for n being odd n′=n−1 may apply. For the simplicity of representation we describe the cases n′=n for n even and n′=n−1 for n odd.

To make the representation as easily understandable as possible, neighboring bits are here input at neighboring inputs each of the syndrome generator Synd 12, so that pairs of neighboring bits are described by the pairs [1, 2], [3, 4], [5, 6], . . . . It is of course also possible to exchange the bits v′₁, v′₂, . . . , v′_n of v′, so that then a binary word v^vert′=v_1(vert), . . . , v_n(vert) results, whereini(vert)=π(i)applies and π(i) may describe a permutation π of the indices 1, . . . , n.

From the condition α^ij⊕α^ij+1=K after a permutation of the indices then simply the condition α^π(ij)⊕α^π(ij+1)=K results.

If, for example, the first bit is not exchanged and the second bit is exchanged by the seventh bit, then instead of αⁱ¹⊕αⁱ²=K now simply αⁱ¹⊕αⁱ⁷=K.

Further: if the data word v′=v′₁, . . . , v′_n is for example read out of a memory whose memory cells may store more than two states, for example three or four states, then those bits may be designated as neighboring bits which correspond to a state of a memory cell in the stored binary data word. In order not to make the description more complicated than necessary, pairs of neighboring bits are here described as pairs [1, 2], [3, 4], [5, 6], . . . . The data word v′=v′₁, . . . , v′_n may have resulted from errors of a code word v of a code with the H matrix H^mod. If no error exists, then v′=v and for the error syndrome s the following applies s=H^mod·v=0.

If an error exists, then v′≠v or v′=v⊕e, wherein e=e₁, . . . , e_n, as already described, which may be designated as an error vector. If v′ and v are different by r bits, this is an r bit error. In this case, the corresponding r components of the error vector e are equal 1. All other components of the error vector are then equal 0.

The decoder Dec 14 is implemented so that

• • when an error syndrome s=0 is applied to its inputs, it outputs a correction vector e=0 at its output,
  • when an error syndrome s with sub-syndromes s1, s3 is applied to its input withs₁=α^ij,s₃=α^3(ij),
  • (and an odd number of bits is corrupted), it outputs e_j=1 and e_t=0 for t≠j,
  • when an error syndrome s with sub-syndromes s1, s3 is applied to its input withs₁=α^ij⊕α^ir,s₃=α^3(ij)⊕α^3(ir)
  • (and an even number of bits is corrupted), it outputs e_j=1 and e_r=1 and e_t=0 for t≠j, r,
  • when an error syndrome s with sub-syndromes s₁, s₃ is applied to its input withs₁=α^ij⊕α^ij+1⊕α^il,s₃=α^3ij⊕α^3ij+1⊕α^3il
  • and with α^ij⊕α^ij+1 wherein K is no column of the matrix H₁^mod and an odd number of bits is corrupted, and for n being even jε{1, 3, . . . , n−1} and for n being odd jε{1, 3, . . . , n−2} apply, it outputs the bits e_j=1, e_j+1=1 and e_l=1 and e_t=0 for t≠j, j+1, l of the correction vector.

In order to determine whether an even or an odd number of bits has been corrupted, the overall parity s_P=v′₁⊕ . . . ⊕v′_n of the data word v′ may be used.

This may e.g. be done by the H matrix H^mod now comprising m′=2m+1 lines and the columns h′₁, . . . , h′_n of the H matrix H^mod being determined as

$h_j^{\prime }=\{\begin{array}{c}{\alpha }^{i_j}\\ {\alpha }^{3\left(i_j\right)}\\ 1\end{array}$ and the H matrix H^mod having the form

$H^{\mathrm{mod}}=\left(\begin{array}{c}{H}_1^{\mathrm{mod}}\\ H_3^{\mathrm{mod}}\\ P\end{array}\right)$ $\mathrm{with}$ $P=\underset{\underset{n}{︸}}{\left(1,\dots ,1\right)}.$

A corresponding implementation of an inventive circuitry is illustrated in FIG. 1b. In FIG. 1b m′=2m+1 applies. The syndrome generator 12a comprises n binary inputs for inputting the data word v′ and (2m+1) binary outputs fed to (2m+1) binary inputs of the decoder 14a and which carry the components s₁, s₃ and s_P. Here, s_P=0 when an even number of components of the data word are corrupted. s_P=1 applies when an odd number of components of the data word v′ are corrupted. The data word v′ is combined in an XOR circuit 16a component after component (i.e., component-wise) with the components of the correction vector e=e₁, . . . , e_n into v^cor, wherein v^cor or is the corrected data word.

A further possibility to determine whether an even or an odd number of components or of bits has been corrupted in the data word v′ may be to only select such columns h_j¹ for j=1, . . . , n as columns of the submatrix H₁^mod which comprise an odd number of ones. In this case, the sub-syndrome s₁=s₁₁, . . . , s_1m=H₁^mod·v′ comprises an even number of ones when an even number of bits has been corrupted in v′ and comprises an odd number of ones when an odd number of bits has been corrupted in v′, so that then for the parity s_P the following appliess_P=s₁₁⊕ . . . ⊕s_1m and the parity is determined from the sub-syndrome s₁ of the error syndrome.

In the circuitry of FIG. 1c m′=2m. The H matrix H^mod is formed from the submatrices H₁^mod and H₃^mod, wherein the submatrix H₁^mod only comprises columns with an odd number of ones. The syndrome generator Synd 12b outputs the sub-syndromes s₁ and s₃ which are m bits wide each on its m′=2m bit wide output wherein the sub-syndromes are applied as input values to the 2m bit wide input of the decoder Dec 14b, which determines from the sub-syndrome s₁ by an XOR combination of the components of s₁ whether an odd or an even number of components are corrupted or not corrupted in v′.

In order to determine whether an even number or an odd number of components of the data word is corrupted, it is also possible to only select such columns as columns of the submatrix H₃^mod which comprise an odd number of ones. In this case, the sub-syndrome s₃=s_3l, . . . , s_3m=H₃^mod·v′ comprises an even number of ones when an even number of bits has been corrupted in v′, and comprises an odd number of ones when an odd number of bits has been corrupted in v′, so that the following then applies for the parity s_P s_P=s_3l⊕ . . . ⊕s_3m.

In order to determine whether an even number or an odd number of components of the data word is corrupted, it is further possible to select the columns of the H matrix H^mod so that a certain subset of components of the columns which may both belong to H₁^mod and also to H₃^mod comprise an odd number of ones. From the corresponding components of the sub-syndromes s₁ and s₃ then again the parity s_P may be determined by an XOR combination.

When the syndrome generator, apart from the syndrome components s₁ and s₃, outputs an additional syndrome component s_P which forms the parity s_P=v′₁⊕ . . . ⊕v′_n, then the parity s_P is provided as an input value at the input of the decoder Dec 14a. If the parity s_P may be determined from components of the syndrome s=s₁, s₃, for example from the components of s₁ by XOR-ing, then the parity s_P may be determined internally by the decoder, for example by a simple XOR combination of components of the error syndrome.

The decoder Dec 14, 14a, 14b may be implemented as follows. The decoder Dec 14a may comprise m′=2m+1 inputs which are downstream from the 2m+1 outputs of the syndrome generator Syn 12a—when the parity s_P is directly provided by the syndrome generator—which carry the components of the error syndrome s=s₁, s₃, s_P, or the decoder Dec 14b may comprise m′=2m inputs connected to the 2m outputs of the syndrome generator Synd 12. n binary outputs of the decoder Dec 14, 14a, 14b are provided for outputting n binary correction values e₁, . . . , e_n for correcting the corresponding bits v′₁, . . . , v′_n of the binary word v′, wherein the correction values e₁, . . . , e_n are determined from the values of the error syndrome s and the binary vector K. The decoder Dec is configured so that in case of a 1-bit error or a 2-bit error in the data word v′ it outputs a correction value e_j=1 at its j-th output for j=1, . . . , n whens₁³⊕s₃⊕s₁²α^ij⊕s₁α^2ij=0and outputs a correction value e_j=0 at its j-th output whens₁³⊕s₃⊕s₁²α^ij⊕s₁α^2ij≠0.

In case of a 3-bit error the decoder outputs a correction value e_j=1 at its j-th output for j=1, . . . , n, whens₁³⊕s₃⊕s₁²K⊕s₁K²⊕K²α^ij⊕Kα^2ij=0ors₁⊕K=α^ij and in case of a 3-bit error it outputs a correction value e_j=0 at its j-th output for j=1, . . . , n when boths₁³⊕s₃⊕s₁²K⊕s₁K²⊕K²α^ij⊕Kα^2ij≠0and alsos₁⊕K≠α^ij apply.

Here, α^ij forms the j-th column in the matrix H₁^mod.

FIG. 2 a shows one possible implementation of an inventive decoder Dec 2. In FIG. 2a it is assumed that the syndrome generator like the syndrome generator Synd 12a in FIG. 1b outputs at its (2m+1) bit wide output an error syndrome s=s₁, s₃, s_P which is applied to the corresponding input lines of the decoder 2. The decoder in FIG. 2a is set up from a subcircuit Dec 1 (21) and n further subcircuits Dec2₁ (221) . . . , Dec2_j (22j) . . . , Dec2_n (22n). The subcircuit Dec 1 (21) functions as an auxiliary subcircuit that is configured to form at least one auxiliary signal on the basis of the first syndrome portion s₁, the second syndrome portion s₃, and the identical column vector K. The n further subcircuits Dec2₁ (221) to Dec2_n (22n) form a plurality of decoder subcircuits. Each decoder subcircuit Dec2₁ to Dec2_n is associated to a specific bit of the binary word v′. Each decoder subcircuit Dec2₁ to Dec2_n is configured for receiving the at least one auxiliary signal and for determining a corresponding correction value of the correction vector e. To this end, each decoder subcircuit Dec2_j is configured to analyze the auxiliary signals “3-bit error”, (s₁³⊕s₃)′, s₁³⊕s₃⊕s₁K²⊕s₁²K, and the first syndrome component s₁ with respect to the particular column vector h_j=α^ij of the parity matrix H^mod, wherein the column vector h_j corresponds to the specific associated bit of the binary word v′. Typically, the plurality of decoder subcircuits comprises n decoder subcircuits, i.e., one decoder subcircuits for each bit of the binary word v′. According to some embodiments it is nevertheless possible that less than n decoder subcircuits are present. The plurality of decoder subcircuits may also be called bit-related or bit-associated subcircuits. The plurality of decoder subcircuits Dec2₁ to Dec2_n provide a parallel processing regarding the correction values e₁ to e_n, which is capable of providing for a fast and efficient determination of the error correction vector e. For instance, after circuit optimization by commercial synthesis tools, circuits can be jointly optimized.

The inputs of the decoder carrying the values of the sub-syndromes s₁, s₃, s_P are directly connected to the 2m+1=m′ inputs of the subcircuit Dec1 21 for forming the signals

• • “3-bit error”,
  • (s₁³⊕s₃)′
  • s₁³⊕s₃⊕s₁K²⊕s₁²Kwhich the subcircuit Dec1 21 provides at its outputs connected to corresponding three inputs each of the subcircuits Dec2_j for j=1, . . . , n. On a fourth input of each of the subcircuits Dec2_j the sub-syndrome s₁ is applied.

The 1-bit wide signal “3-bit error” indicates whether a 3-bit error exists. The m bit wide signal (s₁³⊕s₃)′ for (s₁, s₃)≠0 is equal s₁³⊕s₃ and for

$\left(s_1,s_3\right)=0\mathrm{equal}\underset{\underset{m}{︸}}{1,0,\dots ,0}.$ The m bit wide value s₁³⊕s₃⊕s₁K²⊕s₁²K is determined by sub-syndromes s₁, s₃ and the constant vector K. The subcircuits Dec2_j 22j output the correction value e_j at their each 1 bit wide output.

In FIG. 2b a subcircuit Dec1′ 21a for forming the signals

• • 3-bit error,
  • (s₁³⊕s₃)′
  • s₁³⊕s₃⊕s₁K²⊕s₁²Kis illustrated which outputs these signals at its outputs when the sub-syndromes s₁, s₃ are applied to its m′=2m bit wide input. It is assumed here that the parity s_P is internally derived from the syndrome components s₁, s₃ by the subcircuit Dec1′. Otherwise, the subcircuit Dec1′ fulfills the same function as the subcircuit Dec1 of FIG. 2a. The (auxiliary) subcircuit Dec1, Dec1′ determines one or more auxiliary signals that are needed by several of the decoder subcircuits Dec2₁ to Dec2_n. Typically, the auxiliary signals provided by the subcircuit Dec1, Dec1′ are used by all of the decoder subcircuits Dec2₁ to Dec2_n. In this manner, the proposed circuitry for the correction of errors is highly efficient because signals that are needed multiple times are generated only once and made available to the decoder subcircuits Dec2₁ to Dec2 for further use.

FIG. 3 a shows a special implementation of the subcircuit Dec1 of FIG. 2a, wherein the parity s_P, as illustrated in FIG. 2a, is provided by the syndrome generator Synd 12a in FIG. 1b. The subcircuit Dec1 21 of the decoder Dec 2 in FIG. 2a in turn consists of, or comprises, a subcircuit 31 for determining the value s₁³⊕s₃, a subcircuit 32 for determining the value s₁²K⊕s₂K², an OR circuit 33 with m inputs and one output, an AND gate 34 with two binary inputs and one output, a NOR circuit 35 with 2m inputs and one output, an XOR circuit 36 with two inputs which are each m bit wide and one m bit wide output and an XOR gate 37 with two 1 bit wide inputs and one 1 bit wide output.

The subcircuit 31 comprises two m bit wide inputs to which the sub-syndromes s₁ and s₃ are applied. It determines the value s₁³⊕s₃ and outputs the same at its m bit wide output. If s₁³=s₃, then s₁³⊕s₃=0. A realization of this circuit as a combinational circuit with a given function is of no difficulty for a person skilled in the art. For example, this subcircuit may be realized as a combinational circuit. It is for example possible to synthesize a combinational circuit with m inputs and m outputs for forming s₁³ via a table of values of this function. The m outputs of this circuit may then be XOR-ed component after component with s₃.

The 2m bit wide input of the NOR circuit 35 is connected to the lines carrying the sub-syndromes s₁ and s₃. This circuit outputs the value 1 exactly when s₁=s₃=0 and no error exists. The output of the NOR circuit 35 is fed to a first input of the XOR gate 37, whose second input is for example connected to the least significant bit of the m-digit output of the subcircuit 31 and whose output carries the least significant bit of the value (s₁³⊕s₃)′. The m bit wide values s₁³⊕s₃ and (s₁³⊕s₃)′ differ by their least significant bit when s₁=s₃=0. The value of the least significant bit of (s₁³⊕s₃)′ in this case is equal 1, while the value of the least significant bit of s₁³⊕s₃ is equal to 0. If s₁, s₃≠0, then s₁³⊕s₃ and (s₁³⊕s₃)′ do not differ.

This is a special implementation which serves for (s₁³⊕s₃)′≠0 when s₁, s₃=0, which may be achieved in different ways. For example, the XOR gate may be replaced by an OR gate. Likewise, the output signal of the NOR gate 35 may be combined with a more significant bit of the output of the subcircuit 31, or the output signal of the NOR gate 35 may be combined with several bits of the output of the subcircuit 31.

The m bit wide output of the subcircuit 31 is fed to the m bit wide input of the OR circuit 33, whose 1-bit wide output is connected to a first input of the AND gate 34 and whose second input is connected to the input carrying the parity signal s_P, and whose output carries the signal “3-bit error”. This signal “3-bit error” is equal to 1, when both s₁³⊕s₃≠0 and s_P=1.

The input carrying the sub-syndrome s₁ is further connected to the m bit wide input of the subcircuit 32 for determining the value s₁²K⊕s₁K², whose m bit wide output is fed to a 2m bit wide first input of the XOR circuit 36, whose second m bit wide input is connected to the output of the subcircuit 31 and which outputs the signal s₁³⊕s₃⊕s₁²K⊕s₁K² at its output.

In FIG. 3b an embodiment for realizing the subcircuit Dec1′ 21a of the FIG. 2b is illustrated in which in contrast to the subcircuit Dec1 21 the parity signal s_P is not provided at the input of the subcircuit but derived from the sub-syndrome s₁. It is assumed here that the columns of the submatrix H₁ each comprise an odd number of ones.

The circuit parts of FIG. 3b, which are not different from the corresponding circuit parts of FIG. 3a, are designated by the same reference numerals and are not to be described again here. In FIG. 3b, the input of the subcircuit 21a carrying the sub-syndrome s₁ is connected to the m bit wide input of an XOR circuit 38 with m inputs and one binary output which outputs the value s_P=s_1l⊕ . . . s_1m at its output.

In FIG. 3c one possible realization of the subcircuit Dec1″ 21b is represented which in addition to the outputs of the subcircuit of FIG. 3a comprises further outputs for outputting signals “1-bit error”, “2-bit error”, and “3-bit error”. The circuit parts of FIG. 3b, which are not different from the corresponding circuit parts of FIG. 3a, are designated by the same reference numeral and are not to be described again here. The input carrying the parity signal s_P is additionally connected to a first input of an AND gate 310 having two inputs and one output and invertedly (e.g., via an inverter) connected with a first input of a further AND gate 39 having two inputs and one output. The output of the OR circuit 33 is additionally connected to the second input of the AND gate 39 and invertedly connected with the second input of the AND gate 310 carrying the signal “1-bit error” at its output. The AND gate 39 outputs the signal “2-bit error” at its output.

FIG. 4 shows one possible realization of a subcircuit Dec2_j 22j of FIG. 2a.

To the 1-bit wide input 410 connected to a control input of a multiplexer 44, the binary signal “3-bit error” is applied, which takes on the value 1 when a 3-bit error exists and takes on the value of 0 when no 3-bit error exists.

To the m bit wide input 411 connected to a first m bit wide input of an XOR circuit 42, the signal (s₁³⊕s₃)′ is applied.

To the m bit wide input 412, connected to a first m bit wide input of an XOR circuit 45, the signal s₁³⊕s₃⊕s₁K²⊕s₁²K is applied. To the second m bit wide input of the XOR circuit 45, the constant value Kα^2ij⊕K²α^ij is applied. This constant value Kα^2ij⊕K²α^ij is determined by the (first or upper) portion h_j¹=α^ij of the column vector h_j of the matrix H^mod so that it is different for each individual decoder subcircuit Dec2_j. In other words, the portion h_j¹=α^ij of the column vector h_j acts as a parameter, or parameterizes, the decoder subcircuit Dec2_j. The output of the XOR circuit 45 is fed to the m bit wide input of an NOR circuit 47 having m binary inputs and one binary output connected to a first input of an OR gate 47 and whose 1-bit wide output is guided into the 1-input of the multiplexer 44.

To the m bit wide input 413 which is at the same time connected to the m bit wide input of a subcircuit 41 for realizing the function s₁α²ⁱ⊕s₁²αⁱ and an m bit wide first input of an XOR circuit 48, the m bit wide value of the sub-syndrome s₁ is applied. The function s₁α²ⁱ⊕s₁²αⁱ is parameterized by the portion h_j¹=α^ij of the column vector h_j that corresponds to the j-th bit of the binary word v′.

The m bit wide output of the subcircuit 41 is connected to the second m bit wide input of the XOR circuit 42 whose m bit wide output is guided into the m bit wide input of NOR circuit 43, whose 1-bit wide output is connected to the 0 input of the multiplexer 44 which outputs the correction value e_j at its output 41.

To the second m bit wide input of the XOR circuit 48 the constant value K⊕α^ij (parameterized by h_j¹=α^ij) is applied. The m bit wide output of the XOR circuit 48 is connected to an m bit wide input of the NOR circuit 49 whose 1-bit wide output is connected to the second input of the OR gate 46.

It is obvious that a person skilled in the art may optimize the subcircuit Dec2_j of FIG. 4. As examples for possibilities of optimizing shall be stated:

Using the XOR circuit 45 the m digit value s₁³⊕s₃⊕s₁K²⊕s₁²K applied to the input 412 is XOR-ed component after component with the constant m digit binary vector Kα^2ij⊕K²α^ij. Exactly for those components of the binary vector Kα^2ij⊕K²α^ij which are equal to 1, the corresponding components of s₁³⊕s₃⊕s₁K²⊕s₁²K are inverted while those components for which the components of the binary vector Kα^2ij⊕K²α^ij are equal to 0 are not inverted. An optimum realization may simply be for the input 412 being connected via corresponding inverters to the m-component input of the NOR circuit 46 for which the function of the XOR circuit 45 may be realized.

By means of the XOR circuit 48 the constant value K⊕α^ij is XOR-ed component after component with the m-component sub-syndrome s₁ applied to the input 413. Functionally equivalent here is a connection of the m input lines of the input 413 to the m binary inputs of the NOR circuit 49 via inverters, while inverters are upstream of inputs of the NOR circuit 49 which are inverted by the constant vector K⊕α^ij, i.e. for which the corresponding components of this vector are equal to 1.

As already discussed, it is possible, in addition to the syndrome components s₁ and s₃ determined by s₁=H₁^mod·v′ and s₃=H₃^mod·v′ to use a further syndrome component s₅=H₅^mod·v′=(α⁵ⁱ¹, . . . , α⁵ⁱⁿ) to differentiate between 3-bit errors with neighboring 2-bit errors and other 3-bit errors which do not comprise neighboring 2-bit errors.

If the overall parity is considered, the used code comprises 3m+1 check bits.

If the used code comprises 3m+1 check bits it is now to be explained with reference to FIG. 1d how a difference may be made between 3-bit errors with neighboring 2-bit errors which are properly corrected according to the invention and 3-bit errors which do not comprise neighboring 2-bit errors and which are not properly corrected.

For an error correction of 1-bit, 2-bit and 3-bit errors with neighboring 2-bit errors, in FIG. 1d only the all in all (2m+1) bit wide syndrome components s₁, s₃ and s_P are used which correspond to the (2m+1) first check bits, although the code comprises more than 2m+1 and a maximum of 3m+1 check bits, wherein further check bits are not used for the correction and only serve to determine whether the correction was executed correctly.

This may be advantageous to realize the correction in a most simple way, which may manifest itself in a relatively limited area requirement of the correction circuit and a relatively fast correction.

Accordingly, the syndrome generator Synd 12c only outputs the (2m+1) bit wide syndrome components s₁, s₃, s_P at its output, which is connected to the input of the decoder Dec 14c and the syndrome generator Synd 12c has a (2m+1) digit output connected to the (2m+1) bit wide input of the decoder Dec 14c. From the syndrome components s₁, s₃, s_P the decoder Dec 14c forms the n-component correction vector e=e₁, . . . , e_n which is combined component-wise with the data word v′ into v^cor in the XOR circuit 16c and is output at the output of the XOR circuit 16c. The output of the XOR circuit 16c is apart from that guided into the m bit wide input of an error indication circuit 17c which outputs an error signal q=q₂ when the error was corrected properly and which outputs an error signal q=q₁≠q₂ when the error was not properly corrected. The error indication circuit may for example form an error signal q withq=s′_5lv . . . vs′_5m  (55)from the components s′_5l, . . . , s′_5m of the sub-syndrome s′₅ withs′₅=H₅^mod·v^cor=(α^5il, . . . ,α⁵ⁱⁿ)·v^cor  (56)and output an error signal q₁=1 when s′₅≠0, . . . , 0. If q=q₁=1, a 3-bit error exists which contains no neighboring 2-bit error and which has not been corrected properly. q=q₂=0 is output when s′₅=0, . . . , 0. A 3-bit error exists then which contains a neighboring 2-bit error and which has been corrected properly. The error signal E(3-bit error) is a component-wise XOR combination of the components s′_5l, . . . , s′_5m of the sub-syndrome s₅. Of course, an error signal q=q₂=0 is also output when no error exists and s=0, or a 1-bit error or a 2-bit error has been corrected properly.

It is also possible to combine only a subset of the components of the sub-syndrome s′₅ by OR operations into an error signal q. This is particularly useful when lines of the matrix H₅^mod are linearly dependent or equal to 0. It is also possible to form an error signal q=F(s′_5l, . . . , s′_5m) wherein F may be a linear or non-linear Boolean function with F(0, . . . , 0)=0.

A neighboring 2-bit error here is always a neighboring 2-bit error occurring at the described positions [1, 2], [3, 4], [5, 6], . . . .

A further possibility to differentiate 3-bit errors containing no neighboring 2-bit errors and which may not be corrected properly from 3-bit errors containing neighboring 2-bit errors which may be corrected properly using a syndrome component s₅ is to be disclosed now.

The used code for example again comprises (3m+1) check bits when the overall parity is considered and it comprises (3m) check bits when the overall parity is not considered.

FIG. 1 e illustrates an inventive circuit comprising a syndrome generator Synd 12d forming an error syndrome s=s₁, s₃, s₅, s_P according to a matrix H^mod

$H^{\mathrm{mod}}=\left(\begin{array}{c}{H}_1^{\mathrm{mod}}\\ H_3^{\mathrm{mod}}\\ H_5^{\mathrm{mod}}\\ P\end{array}\right)$ and outputs same at the output of the syndrome generator Synd 12d connected to the input of the decoder Dec 14d.

The decoder Dec 14d comprises signals for error indication. The signals for error indication may include a signal which indicates whether a 3-bit error exists and they may include a signal which indicates whether a 3-bit error with a neighboring 2-bit error exists which may be corrected properly or may not be corrected properly.

FIG. 2 c shows a subcircuit Dec1″ 21b with a first m bit wide input for inputting the sub-syndrome s₁, a second m bit wide input for inputting the sub-syndrome s₃, a third m bit wide input for inputting the sub-syndrome s₅ and a fourth 1-bit wide input for inputting the sub-syndrome s_P and a first 1-bit wide output for outputting the signal “3-bit error”, a second m bit wide output for outputting the value (s₁³⊕s₃)′, a further m bit wide output for outputting the signal s₁³⊕s₃⊕s₁K²⊕s₁²K and a further 1-bit wide output for outputting a signal “non-correctable 3-bit error”.

The signal “non-correctable 3-bit error” is equal to 1 whens₁⁶⊕s₃²⊕s₁⁵K⊕s₁³K³⊕Ks₅⊕K³s₃≠0  (57)applies.

A direct implementation of the function described in equation (56) may indicate whether a non-corrected 3-bit error exists.

Some modifications are still to be noted.

It is thus possible, instead of the syndrome component s′₅=s′_5l, . . . , s′_5m=H₅^modv^cor to implement only a subset of the syndrome components s′_5l, . . . , s′_5m if it sufficient to detect with a certain probability that a correctable or non-correctable 3-bit error exist.

It is also possible to for example (linearly or non-linearly) compact the m components s′_5l, . . . , s′_5m into m*<m components.

A special possibility of reducing the effort of the decoder is still to be noted.

If a 3-bit error exists containing a neighboring 2-bit error then in case of a correctable 3-bit error the correction always takes place in two subsequent bits i_j and i_j+1, so that the subsequent correction values e_ij and e_ij+1 are always equal in case of a correctable 3-bit error.

Then the termss₁³⊕s₃⊕s₁K²⊕s₁²K⊕Kα^2ij⊕K²α^ij ands₁³⊕s₃⊕s₁K²⊕s₁²K⊕Kα^2ij+1⊕K²α^ij+1 are always both equal to 1 or equal to 0.

This facilitates saving an m-input NOR circuit 47 in FIG. 4 for every second correction value, as the output of this NOR circuit may likewise be used as the output of the corresponding NOR circuit both in the subcircuit 22i_j and also in the subcircuit 22i_j+1.

It is further clear that the subcircuits 22i, i=1, . . . , n can be optimized together by one synthesis tool.

In the following, with reference to examples, the determination of the H matrix is to be shown.

As an example m=4 and the Galois field GF(2⁴) are selected. Common representations of the elements of the Galois field GF(2⁴) in the exponent representation, the polynomial representation and the vector representation with the modular polynomial m(x)=1+x+x⁴ are given in table 1.

Table 1 shows the known representations of the elements of the GF(2⁴) in the exponent representation α⁰, α¹, . . . , α¹⁴ and 0, in the polynomial representation as 16 polynomials p(x)=a₀+a₁x+a₂x²+a₃x³ of a degree less or equal 3 with the polynomial variables x and as 4-digit binary vectors whose components are the coefficients of the corresponding polynomials. The modular polynomial of the Galois field is m(x)=1+x+x⁴.

It may thus be seen from table 1 that the element p(x)=1 (in the polynomial representation) and the binary vector 1, 0, 0, 0 (in the vector representation) corresponds to the element α⁰ (in the exponential representation), as it is described in the first line. It may be gathered from the third line that the representations p(x)=x and 0, 1, 0, 0 correspond to α¹.

As it is known, the addition of two elements in the vector representation is executed as a component-wise addition modulo 2 of the components.

TABLE 1
Elements of the GF(24) generated by the primitive modular polynomial
m(x) = 1 +x + x4 in different forms of representation:
	Exponential	Polynomial	Vector
	Representation	Representation	Representation
	0	0	(0000)
	α0	1	(1000)
	α1	x	(0100)
	α2	x2	(0010)
	α3	x3	(0001)
	α4	1 + x	(1100)
	α5	x + x2	(0110)
	α6	x2 + x3	(0011)
	α7	1 + x + x3	(1101)
	α8	1 + x2	(1010)
	α9	x + x3	(0101)
	α10	1 + x + x2	(1110)
	α11	x + x2 + x3	(0111)
	α12	1 + x + x2 + x3	(1111)
	α13	1 + x2 + x3	(1011)
	α14	1 + x3	(1001)

The addition in the polynomial representation is done by XOR-ing the corresponding coefficients of the powers or potencies of the polynomial variables.

The multiplication of two elements αⁱ and α^j results in the element α^k with k=i+j modulo (2^m−1)=i+j modulo 15. The multiplication of two polynomials p₁(x) and p₂(x) in the polynomial representation results in the polynomial p₃(x) with p₃(x)=p₁(x)·p₂(x) modulo m(x) wherein m(x) is the modular polynomial of the Galois field.

In the Galois field GF(2⁴) there are 15 different columns

$h_1,\dots ,h_{15}=\begin{array}{ccccc}{\alpha }^0& {\alpha }^1& {\alpha }^2& \dots & {\alpha }^{14}\\ {\alpha }^0,& {\alpha }^3,& {\alpha }^6,& \dots ,& {\alpha }^{12}\end{array}$

If the length of the code is to be n=14, then from these columns for example the column [α¹², α⁶]^T with

${\alpha }^{12}=\left(\begin{array}{c}1\\ 1\\ 1\\ 1\end{array}\right)={\left[1,1,1,1\right]}^T=K$ may be deleted. An inventive H matrix

$H^{\mathrm{mod}}=\left(h_1,\dots ,h_{14}\right)=\left(\begin{array}{ccccc}{\alpha }^{i_1},& {\alpha }^{i_2},& {\alpha }^{i_3},& \dots ,& {\alpha }^{i_{14}}\\ {\alpha }^{3i_1},& {\alpha }^{3i_2},& {\alpha }^{3i_3},& \dots ,& {\alpha }^{3i_{14}}\end{array}\right)$ $H_1^{\mathrm{mod}}=\left({\alpha }^{i_1},{\alpha }^{i_2},{\alpha }^{i_3},\dots ,{\alpha }^{i_{14}}\right)$ $\mathrm{and}$ $H_3^{\mathrm{mod}}=\left({\alpha }^{3i_1},{\alpha }^{3i_2},{\alpha }^{3i_3},\dots ,{\alpha }^{3i_{14}}\right)$ may now be formed so that the component-wise XOR sum of the first four components of the pairs of the neighboring columns [1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12], [13, 14] is each equal to K=[1, 1, 1, 1]^T. The first four components of a column of the matrix H^mod here correspond to a column of H₁^mod. Thus, a matrix

$\begin{array}{cc}\begin{array}{c}{H}^{\mathrm{mod}}=\left(\begin{array}{cccccccccccccc}{\alpha }^0& {\alpha }^{11}& {\alpha }^1& {\alpha }^{13}& {\alpha }^2& {\alpha }^7& {\alpha }^3& {\alpha }^{10}& {\alpha }^4& {\alpha }^6& {\alpha }^5& {\alpha }^{14}& {\alpha }^8& {\alpha }^9\\ {\alpha }^0& {\alpha }^3& {\alpha }^3& {\alpha }^9& {\alpha }^6& {\alpha }^6& {\alpha }^9& {\alpha }^0& {\alpha }^{12}& {\alpha }^3& {\alpha }^0& {\alpha }^{12}& {\alpha }^9& {\alpha }^{12}\end{array}\right)\\ =\left(\begin{array}{cccccccccccccc}1& 0& 0& 1& 0& 1& 0& 1& 1& 0& 0& 1& 1& 0\\ 0& 1& 1& 0& 0& 1& 0& 1& 1& 0& 1& 0& 0& 1\\ 0& 1& 0& 1& 1& 0& 0& 1& 0& 1& 1& 0& 1& 0\\ 0& 1& 0& 1& 0& 1& 1& 0& 0& 1& 0& 1& 0& 1\\ -& -& -& -& -& -& -& -& -& -& -& -& -& -\\ 1& 0& 0& 0& 0& 0& 0& 1& 1& 0& 1& 1& 0& 1\\ 0& 0& 0& 1& 0& 0& 1& 0& 1& 0& 0& 1& 1& 1\\ 0& 0& 0& 0& 1& 1& 0& 0& 1& 0& 0& 1& 0& 1\\ 0& 1& 1& 1& 1& 1& 1& 0& 1& 1& 0& 1& 1& 1\end{array}\right)\end{array}& \left(58\right)\end{array}$ can be determined.

The syndrome generator Synd implements the relationss=s₁,s₃=s₁₁,s₁₂,s₁₃,s₁₄,s₃₁,s₃₂,s₃₃,s₃₄=H^mod·(v′₁, . . . ,v′₁₄)or in a multiplied forms₁₁=v′₁⊕v′₄⊕v′₆⊕v′₈⊕v′₉⊕v′₁₂⊕v′₁₃,s₁₂=v′₂⊕v′₃⊕v′₆⊕v′₈⊕v′₉⊕v′₁₁⊕v′₁₄,s₁₃=v′₂⊕v′₄⊕v′₅⊕v′₈⊕v′₁₀⊕v′₁₁⊕v′₁₃,s₁₄=v′₂⊕v′₄⊕v′₆⊕v′₇⊕v′₁₁⊕v′₁₂⊕v′₁₄,s₃₁=v′₁⊕v′₈⊕v′₉⊕v′₁₁⊕v′₁₂⊕v′₁₄,s₃₂=v′₄⊕v′₇⊕v′₉⊕v′₁₂⊕v′₁₃,s₃₃=v′₅⊕v′₆⊕v′₉⊕v′₁₁⊕v′₁₄,s₃₄=v′₂⊕v′₃⊕v′₄⊕v′₅⊕v′₅⊕v′₇⊕v′₁₀⊕v′₁₂⊕v′₁₃⊕v′₁₄,whose implementation presents no difficulty for a person skilled in the art for example using XOR gates.

If the overall parity s_P is considered, then further s_P=v′₁⊕ . . . ⊕v′₁₄ is to be implemented.

Further, the matrix H^mod may be supplemented by a submatrix H₅^mod,H₅^mod=(α⁵ⁱ¹,α⁵ⁱ²,α³ⁱ³, . . . ,α⁵ⁱ¹⁴).

The concrete form of the matrix H₅^mod here is

$\begin{array}{c}{H}_5^{\mathrm{mod}}=\left({\alpha }^1{\alpha }^8{\alpha }^{11}{\alpha }^{11}{\alpha }^8{\alpha }^{11}{\alpha }^1{\alpha }^{11}{\alpha }^{11}{\alpha }^1{\alpha }^8{\alpha }^6{\alpha }^8{\alpha }^1\right)\\ =\left(\begin{array}{cccccccccccccc}1& 1& 0& 0& 1& 0& 1& 0& 0& 1& 1& 1& 1& 1\\ 0& 1& 1& 1& 1& 1& 0& 1& 1& 0& 1& 1& 1& 0\\ 0& 1& 1& 1& 1& 1& 0& 1& 1& 0& 1& 1& 1& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\end{array}\right).\end{array}$

The third line is equal to the second line and the fourth line is identical to 0, so that as additional syndrome components of s₅=s₅₁, s₅₂, s₅₃, s₅₄, in the implementation of the syndrome generator only the two syndrome componentss₅₁=v′₁⊕v′₂⊕v′₅⊕v′₇⊕v′₁₀⊕v′₁₁⊕v′₁₂⊕v′₁₃⊕v′₁₄,s₅₂=v′₂⊕v′₃⊕v′₄⊕v′₅⊕v′₈⊕v′₉⊕v′₁₁⊕v′₁₂⊕v′₁₃,are to be implemented as an XOR circuit, as here s₅₃=s₅₂ and s₅₄=0 apply.

In the following it is to be indicated how the subcircuit 31 may be realized for realizing the function s₁³⊕s₃.

It is possible to realize the function y=s₁³ as a combinational circuit with m=4 binary inputs s₁₁, s₁₂, s₁₃, s₁₄ and m=4 binary outputs y₁, y₂, y₃, y₄. The outputs of this circuit are then XOR-ed component-wise with the syndrome components s₃₁, s₃₂, s₃₃, s₃₄.

A value table of the function y=y₁, y₂, y₃, y₄=s₁³ is represented in table 2.

TABLE 2
value table y = s13 in the GF(24):
	s11	s12	s13	s14	y1	y2	y3	y4
	0	0	0	0	0	0	0	0
	1	0	0	0	1	0	0	0
	0	1	0	0	0	0	0	1
	0	0	1	0	0	0	1	1
	0	0	0	1	0	1	0	1
	1	1	0	0	1	1	1	1
	0	1	1	0	1	0	0	0
	0	0	1	1	0	0	0	1
	1	1	0	1	0	0	1	1
	1	0	1	0	0	1	0	1
	0	1	0	1	1	1	1	1
	1	1	1	0	1	0	0	0
	0	1	1	1	0	0	0	1
	1	1	1	1	0	0	1	1
	1	0	1	1	0	1	0	1
	1	0	0	1	1	1	1	1

This table results from reading the corresponding exponential representation α^j for each of the 16 4-digit binary vectors s₁₁, s₁₂, s₁₃, s₁₄ which represent the vector representation of an element of the Galois field GF(2⁴), determining k=3j modulo 15 and reading the corresponding vector representation in the table for α^k which represents the binary value for s₁³=y₁, y₂, y₃, y₄.

An implementation of a value table as a combinational circuit is no difficulty for a person skilled in the art.

One possible implementation of the subcircuit 32 for realizing the function z=s₁²K⊕s₁K² is to be described now. Here, K=α¹² was selected as an example. In the polynomial representation α¹² according to table 1 corresponds to the polynomial K(x)=1+x+x²+x³. The modular polynomial of the considered Galois field is m(x)=1+x+x⁴. In the polynomial representation s₁ is represented by the polynomial s₁(x)=s₁₁+s₁₂x+s₁₃x²+s₁₄x³ and for z(x) in the polynomial representation the following appliesz(x)=(s₁(x)²·K(x)⊕s₁(x)·K²(x))mod(1+x+x⁴),from which by a direct calculation for z(x)=z₁+z₂x+z₃x²+z₄x³ z(x)=(s₁₁+s₁₄)+(s₁₂+s₁₄)x+(s₁₁+s₁₂+s₁₃+s₁₄)x²+(s₁₂+s₁₃)x³ results.The following appliesz₁=s₁₁+s₁₄ z₂=s₁₂+s₁₄ z₃=s₁₁+s₁₂+s₁₃+s₁₄ z₄=s₁₂+s₁₃,wherein the additions are modulo 2 which logically correspond to an XOR combination. An implementation of the corresponding subcircuit for this embodiment is illustrated in FIG. 5.

The circuit of FIG. 5 comprises four binary inputs 55, 56, 57 and 58 carrying the values s₁₁, s₁₂, s₁₃, s₁₄ and four binary outputs 59, 510, 511 and 512 outputting the values z₁, z₂, z₃, and z₄ and four XOR gates 51, 52, 53 and 54 comprising two inputs and one output each. A first input of the XOR gate 51 is connected to the circuit input 55, while the second input of this gate is connected to the input 58. The output of the XOR gate 51 is connected both to the circuit output 59 and also to a first input of the XOR gate 54. The second input of the XOR gate 54 is connected to the output of the XOR gate 53 which is at the same time also connected to the output 512. The output of the XOR gate 54 is connected to the output 511.

A first input of the XOR gate 52 is connected to the input 56 which is at the same time also connected to a first input of the XOR gate 53. The second input of the XOR gate 52 is connected to the input 58. The second input of the XOR gate 53 is connected to the input 57. A first input of the XOR gate 52 is connected to the input 56 which is simultaneously also connected to a first input of the XOR gate 53. The second input of the XOR gate 52 is connected to the input 58. The output of the gate 52 is connected to the output 510. The second input of the XOR gate 53 is connected to the input 57.

Further, an example of a realization of the subcircuit 41 for realizing the function u=s₁α^2ij⊕s₁²α^ij is to be described.

In the polynomial representation s(x) is again represented as the polynomial s₁(x)=s₁₁+s₁₂x+s₁₃x²+s₁₄x³. As an example, i_j=4 is selected, i.e. for αⁱ, α⁴ is selected. In the matrix H₁^mod according to equation (58), α⁴ corresponds to the ninth column. It may be seen from table 1 that in the polynomial representation α⁴ corresponds to the polynomial α⁴(x)=1+x. For u(x) thusu(x)=s₁(x)·(1+x)²+s₁(x)²·(1+x)mod(1+x+x⁴),results, from which by a direct calculation the following resultsu(x)=s₁₄+(s₁₁+s₁₂+s₁₃)x+(s₁₁+s₁₂)x²+s₁₄x³ and thusu₁=s₁₄ u₂=s₁₁+s₁₂+s₁₃ u₃=s₁₁+s₁₂ u₄=s₁₄

Here, the operation + in the addition of the polynomial coefficients is again to be interpreted modulo 2 and it corresponds to an XOR combination. A corresponding subcircuit for realizing the functionu=s₁·α⁸+s₁²·α⁴ is illustrated in FIG. 6.

The circuit of FIG. 6 comprises four inputs 63, 64, 65 and 66 at which the values s₁₁, s₁₂, s₁₃, and s₁₄ are input and four outputs 67, 68, 69 and 610 at which the values u₁, u₂, u₃, and u₄ are output. It comprises two XOR gates 61 and 62 with two inputs and one output each.

The input 63 is connected to a first input of the XOR gate 61 to whose second input the input 64 is connected. The output of the XOR gate 61 is connected to a first input of the XOR gate 62 and at the same time to the output 69. The input 65 is connected to the second input of the XOR gate 62 whose output is connected to the output 68. The input 66 is at the same time connected to the output 67 and to the output 610.

For the expression K·α^2ij⊕K²·α^ij for K=α¹² and i_j=4α¹²·α⁸⊕α²⁴·α⁴=α²⁰⊕α²⁸=α⁵⊕α¹³=α⁷ is obtained or in the polynomial representation according to table 1 the polynomial 1+x+x³ or the vector representation (1101) is obtained.

In FIG. 4, (s₁³⊕s₃)′⊕s₁K²⊕s₁²K for i=4 is functionally XOR-ed component-wise with the constant value Kα⁸⊕K²α⁴ in the XOR circuit 45.

Practically, this combination may simply be done by inverting the first, the second and the fourth input of the NOR circuit 46 and not inverting the third input. Accordingly, it applies for K=α¹² and α^ij=α⁴ thatα¹²⊕α⁴=α⁶ or in the polynomial representation being equal x²+x³. In the vector representation α⁶=(0, 1, 1, 0) applies. Thus, the component-wise XOR combination of s₁ and K⊕α^ij for K=α¹² and α^ij=α⁴ in the XOR circuit 48 may be implemented by inverting the third and the fourth inputs of the NOR circuit 49 and the first and fourth inputs remaining non-negated.

In the following it is to be explained for one example how an H matrix H^mod may be determined whose m first components of its columns always contain an odd number of ones. As an example here the Galois field GF(2⁵) with 32 elements is considered.

The different representations of the elements of the Galois field GF(2⁵) with the modular polynomial m(x)=1+x²+x⁵ are represented in table 3.

A binary representation of non-modified submatrices H₁, H₃ and H₅ of an unshortened BCH code of the length 31 withH₁=(α⁰α¹ . . . α³⁰),H₃=(α⁰α³ . . . α^3-30),H₅=(α⁰α⁵ . . . α^5-30),may be read in their concrete form directly from table 3.

TABLE 3
Elements of GF(25) generated by the primitive
modular polynomial m(x) = 1 + x2 +
x5 in different forms of representation:
	Exponential	Polynomial	Vector
	Representation	Representation	Representation
	0	0	(00000)
	α0	1	(10000)
	α1	x	(01000)
	α2	x2	(00100)
	α3	x3	(00010)
	α4	x4	(00001)
	α5	1 + x2	(10100)
	α6	x + x3	(01010)
	α7	x2 + x4	(00101)
	α8	1 + x2 + x3	(10110)
	α9	x + x3 + x4	(01011)
	α10	1 + x4	(10001)
	α11	1 + x + x2	(11100)
	α12	x + x2 + x3	(01110)
	α13	x2 + x3 + x4	(00111)
	α14	1 + x2 + x3 + x4	(10111)
	α15	1 + x + x2 + x3 + x4	(11111)
	α16	1 + x + x3 + x4	(11011)
	α17	1 + x + x4	(11001)
	α18	1 + x	(11000)
	α19	x + x2	(01100)
	α20	x2 + x3	(00110)
	α21	x3 + x4	(00011)
	α22	1 + x2 + x4	(10101)
	α23	1 + x + x2 + x3	(11110)
	α24	x + x2 + x3 + x4	(01111)
	α25	1 + x3 + x4	(10011)
	α26	1 + x + x2 + x4	(11101)
	α27	1 + x + x3	(11010)
	α28	x + x2 + x4	(01101)
	α29	1 + x3	(10010)
	α30	x + x4	(01001)

Here, the exponents of α are to be interpreted modulo 31. The following applies

$H_1-\left(\begin{array}{ccccccccccccccccccccccccccccccc}1& 0& 0& 0& 0& 1& 0& 0& 1& 0& 1& 1& 0& 0& 1& 1& 1& 1& 1& 0& 0& 0& 1& 1& 0& 1& 1& 1& 0& 1& 0\\ 0& 1& 0& 0& 0& 0& 1& 0& 0& 1& 0& 1& 1& 0& 0& 1& 1& 1& 1& 1& 0& 0& 0& 1& 1& 0& 1& 1& 1& 0& 1\\ 0& 0& 1& 0& 0& 1& 0& 1& 1& 0& 0& 1& 1& 1& 1& 1& 0& 0& 0& 1& 1& 0& 1& 1& 1& 0& 1& 0& 1& 0& 0\\ 0& 0& 0& 1& 0& 0& 1& 0& 1& 1& 0& 0& 1& 1& 1& 1& 1& 0& 0& 0& 1& 1& 0& 1& 1& 1& 0& 1& 0& 1& 0\\ 0& 0& 0& 0& 1& 0& 0& 1& 0& 1& 1& 0& 0& 1& 1& 1& 1& 1& 0& 0& 0& 1& 1& 0& 1& 1& 1& 0& 1& 0& 1\end{array}\right)$ $H_3=\left(\begin{array}{ccccccccccccccccccccccccccccccc}1& 0& 0& 0& 0& 1& 1& 0& 0& 1& 0& 0& 1& 1& 1& 1& 1& 0& 1& 1& 1& 0& 0& 0& 1& 0& 1& 0& 1& 1& 0\\ 0& 0& 1& 1& 1& 1& 1& 0& 1& 1& 1& 0& 0& 0& 1& 0& 1& 0& 1& 1& 0& 1& 0& 0& 0& 0& 1& 1& 0& 0& 1\\ 0& 0& 0& 0& 1& 1& 0& 0& 1& 0& 0& 1& 1& 1& 1& 1& 0& 1& 1& 1& 0& 0& 0& 1& 0& 1& 0& 1& 1& 0& 1\\ 0& 1& 1& 1& 1& 1& 0& 1& 1& 1& 0& 0& 0& 1& 0& 1& 0& 1& 1& 0& 1& 0& 0& 0& 0& 1& 1& 0& 0& 1& 0\\ 0& 0& 0& 1& 0& 1& 0& 1& 1& 0& 1& 0& 0& 0& 0& 1& 1& 0& 0& 1& 0& 0& 1& 1& 1& 1& 1& 0& 1& 1& 1\end{array}\right)$ $H_5=\left(\begin{array}{ccccccccccccccccccccccccccccccc}1& 1& 1& 1& 0& 1& 0& 0& 0& 1& 0& 0& 1& 0& 1& 0& 1& 1& 0& 0& 0& 0& 1& 1& 1& 0& 0& 1& 1& 0& 1\\ 0& 0& 0& 1& 0& 0& 1& 0& 1& 0& 1& 1& 0& 0& 0& 0& 1& 1& 1& 0& 0& 1& 1& 0& 1& 1& 1& 1& 1& 0& 1\\ 0& 1& 0& 1& 1& 0& 0& 0& 0& 1& 1& 1& 0& 0& 1& 1& 0& 1& 1& 1& 1& 1& 0& 1& 0& 0& 0& 1& 0& 0& 1\\ 0& 0& 0& 1& 1& 1& 0& 0& 1& 1& 0& 1& 1& 1& 1& 1& 0& 1& 0& 0& 0& 1& 0& 0& 1& 0& 1& 0& 1& 1& 0\\ 0& 0& 1& 1& 0& 1& 1& 1& 1& 1& 0& 1& 0& 0& 0& 1& 0& 0& 1& 0& 1& 0& 1& 1& 0& 0& 0& 0& 1& 1& 1\end{array}\right)$

As an example for K, here K=α²⁴ is selected. It may be gathered from table 3 that in the polynomial representationK(x)=x+x²+x³+x⁴ applies and in the vector representation (01111) applies.

To determine H^mod now all columns of the H matrix are deleted which comprise an even number of ones in the components of H₁ and arrange the same so that the component-wise XOR combination of the pairs of columns of the submatrix H₁ in the bit positions [1, 2], [3, 4], [5, 6], . . . each results in [0, 1, 1, 1, 1]^T.

The submatrices H₁^mod, H₃^mod and H₅^mod of the matrix H^mod obtained this way now are

$H_1^{\mathrm{mod}}=\left(\begin{array}{cccccccccccccccc}1& 1& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 1& 1& 1& 1\\ 0& 1& 1& 0& 0& 1& 0& 1& 0& 1& 0& 1& 1& 0& 0& 1\\ 0& 1& 0& 1& 1& 0& 0& 1& 0& 1& 1& 0& 1& 0& 1& 0\\ 0& 1& 0& 1& 0& 1& 1& 0& 0& 1& 1& 0& 0& 1& 0& 1\\ 0& 1& 0& 1& 0& 1& 0& 1& 1& 0& 0& 1& 0& 1& 1& 0\end{array}\right),H_3^{\mathrm{mod}}=\left(\begin{array}{cccccccccccccccc}1& 1& 0& 1& 0& 1& 0& 1& 0& 1& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 1& 1& 1& 0& 1& 0& 1& 0& 0& 0& 0& 1\\ 0& 1& 0& 1& 0& 0& 0& 1& 1& 1& 1& 1& 1& 1& 0& 1\\ 0& 1& 1& 1& 1& 1& 1& 0& 1& 0& 1& 1& 0& 1& 0& 0\\ 0& 1& 0& 0& 0& 0& 1& 1& 0& 0& 1& 0& 0& 1& 1& 0\end{array}\right),H_5^{\mathrm{mod}}=\left(\begin{array}{cccccccccccccccc}1& 0& 1& 0& 1& 1& 1& 1& 0& 1& 0& 1& 0& 0& 1& 1\\ 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 1& 1& 1& 1& 1& 1\\ 0& 1& 1& 0& 0& 1& 1& 0& 1& 0& 0& 1& 1& 0& 0& 1\\ 0& 1& 0& 1& 0& 1& 1& 1& 1& 1& 1& 1& 1& 0& 0& 0\\ 0& 1& 0& 0& 1& 1& 1& 1& 0& 0& 1& 0& 1& 0& 1& 0\end{array}\right)$

Here, all 16 columns of the submatrix H₁ were used which comprise an odd number of ones, so that the number of the columns of H^mod is equal to 16.

It is now possible, as illustrated in FIG. 3b, to derive the overall parity as an XOR combination by the XOR circuit 38 from the sub-syndrome s₁ of the error syndrome s.

The functioning of the invention is now to be demonstrated first of all for the correction of a 2-bit error and subsequently for the correction of a 3-bit error with a neighboring 2-bit error. As an H matrix the H matrix

$H^{\mathrm{mod}}=\left(\begin{array}{c}{H}_1^{\mathrm{mod}}\\ H_3^{\mathrm{mod}}\\ P\end{array}\right)$ is used which comprises 14 columns, wherein the columns of H^mod and of H₃^mod are elements of GF(2⁴) with the modular polynomial m(x)=1+x+x⁴ and K=α¹²=[1, 1, 1]^T, as described in equation (58).

The binary word v′ which is to be corrected is the word v′=(11010011100101). For the syndrome components s₁, s₃ and s_P then the following applies

$s_1=H_1^{\mathrm{mod}}·v^{\prime }=\left(\begin{array}{c}1\\ 0\\ 1\\ 1\end{array}\right)={\alpha }^{13},s_3=H_3^{\mathrm{mod}}·v^{\prime }=\left(\begin{array}{c}1\\ 1\\ 1\\ 0\end{array}\right)={\alpha }^{10},s_P=P·v^{\prime }=\left(0\right),$ as it may be directly recalculated using the H matrix according to equation (58).

As s_P=0 and s≠0, obviously a 2-bit error exists. The zeroes of the quadratic equation s₁³⊕s₃⊕s₁·x²⊕s₁²·x=0 are α¹¹ and α⁴.

This may be checked easiest by putting in all possible values αⁱ for i=0, 1, . . . , 14.

In FIG. 2a, for each component e_j a subcircuit Dec2_ij, j=1, . . . , n, exists which then outputs a correction value e_j=1 when α^ij is the solution of equation s₁³⊕s₃⊕s₁·x²⊕s₁²·x=0.

Solutions of this equation are α⁴ and α¹¹.

FIG. 4 shows one possible implementation of a subcircuit Dec2_j. In the H matrix determined by equation (58) αⁱ²=α¹¹ and aⁱ⁹=α⁴, so that α¹¹ corresponds to the second and α⁴ the ninth column of the H matrix H^mod. In the data word v′=(11010011100101) thus the second and ninth bit positions are to be corrected so that with v′=(11010011100101) the error vector e=(01000000100000) has to be XOR-ed component-wise. The corrected word v^cor is then v^cor=(10010011000101). The components e₂ and e₉ are equal to 1, while all remaining components of the correction vector are equal to 0.

In order to explain the correction of a 3-bit error with a neighboring 2-bit error, as an example the correction of the data word v′=(10100011001101) is explained.

For the syndrome components now the following results

$s_1=H_1^{\mathrm{mod}}·v^{\prime }=\left(\begin{array}{c}1\\ 0\\ 0\\ 1\end{array}\right)={\alpha }^{14}$ $s_3=H_3^{\mathrm{mod}}·v^{\prime }=\left(\begin{array}{c}1\\ 1\\ 0\\ 0\end{array}\right)={\alpha }^4$ $s_P=P·v^{\prime }=\left(1\right)$

As S_P=1 and s₁³=(α¹⁴)³=α⁴²=α¹²≠α⁴=s₃ obviously a 3-bit error exist.

The following applies

$s_1+K={\alpha }^{14}\oplus {\alpha }^{12}=\left(\begin{array}{c}1\\ 0\\ 0\\ 1\end{array}\right)\oplus \left(\begin{array}{c}1\\ 1\\ 1\\ 1\end{array}\right)=\left(\begin{array}{c}0\\ 1\\ 1\\ 0\end{array}\right)={\alpha }^5$

The zeros of the quadratic equations₁³⊕s₃⊕s₁·K²⊕s₁²·K⊕K·x²⊕K²·x are α¹ and α¹³, as it may be checked by directly recalculating.

In the H matrix determined by equation (58) αⁱ¹¹=α⁵, αⁱ³=α¹ and αⁱ⁴=α¹³, so that α⁵ correspond to the eleventh column, α¹ to the third column and α¹³ to the fourth column the H matrix determined by equation (58).

The data word v′=(10100011001101) is to be corrected in the third, the fourth and the eleventh bit position by combining the correction vector e=(00110000001000) component-wise with v′ into v′⊕e=(10100011001101)⊕(00110000001000)=(10010011000101)=v^cor.

FIG. 7 shows a schematic flow diagram of a method for correcting errors in a possibly erroneous binary word v′=v′₁, . . . , v′_n relative to a codeword v=v₁, . . . , v_n. The method comprises a step 902 for determining an error syndrome s=(s₁, s₃) of a modified BCH code with a H-matrix H^mod comprising a first BCH submatrix H₁^mod and a second BCH submatrix H₃^mod. The modified BCH code has a code distance d≧5. According to the modified BCH code, n′ column vectors of the submatrix H₁^mod are paired as column vector pairs so that a componentwise XOR combination of the two column vectors of each column vector pair produces an identical column vector K that is different from all column vectors of the BCH submatrix H^mod. The number n′ is even and 4≦n′≦n applies. The second BCH submatrix H₃^mod comprises a corresponding column vector for each column vector in the first BCH submatrix H₁^mod so that the corresponding column vector is the third power, according to Galois field arithmetic, of the column vector in the first BCH submatrix H₁^mod. The error syndrome s is determined by multiplying the H-matrix H^mod with the possibly erroneous binary word v′ so that a first error syndrome portion is given by s₁=H₁^mod·v′ and a second error syndrome portion is given by s₃=H₃^mod·v′.

The method further comprises a step 904 of generating the correction vector e=(e₁, . . . , e_n) with correction values e_j=e_j+1=e_l=1 and e_t=0 for t≠j, j+1, l, if the following conditions are fulfilled:

• • s₁ equals the componentwise XOR combination of the identical column vector K and a column vector at a column position 1 of the first BCH submatrix H₁^mod, i.e., s₁=K⊕h_l¹; and
  • s₃ equals the componentwise XOR combination of column vectors at column positions j, j+1, and 1 of the second BCH submatrix H₃^mod, i.e., s₃=h_j¹⊕h_j+1¹⊕h_l¹.

This means that the error syndrome s=(s₁, s₃) fulfills the following condition:s₁=α^ij⊕α^ij+1⊕α^il,s₃=α^3ij⊕α^3ij+1⊕α^3il

In this manner, an adjacent 2-bit error in combination with a further 1-bit error (i.e., a 3-bit error in total) can be corrected. Moreover, the method for correcting errors typically also supports the correction of 1-bit errors and 2-bit errors (adjacent or not adjacent).

In case none of the conditions for a 1-bit error, a 2-bit error, or a 3-bit error containing an adjacent 2-bit error are fulfilled, all correction values e_j, j=1 . . . n, are typically chosen to be zero, so that no correction of the binary word v′ is performed. This typically means that no correctable error was detected: Either the binary word v′ was error-free or it contains too many errors in order to be detected and/or corrected in an unambiguous manner.

It is possible that the bits of a word which are to be corrected are auxiliary binary read values of a ternary memory or a multiple-valued memory as it is described in the U.S. patent application Ser. No. 13/664,495, filed Oct. 31, 2012 and entitled “Circuit and Method for Multi-Bit Correction” which is included here in the description by reference in its entirety. The auxiliary binary read values are provided by a subcircuit LH in U.S. Ser. No. 13/664,495 on the basis of ternary (or multi-valued) state values of memory cells which are adapted to take on, at a particular time, one of at least three states. The subcircuit LH is called an “auxiliary read value generator” in the present disclosure. Hence, the technology of ternary or multi-valued memories may benefit from the powerful error correcting capabilities of BCH codes. Ternary memories or multi-valued memories are relatively prone to suffer from adjacent bit errors as a single memory cell influences two or more of the auxiliary binary read values.

Although some aspects have been described in the context of an apparatus, it is clear that these aspects also represent a description of the corresponding method, where a block or device corresponds to a method step or a feature of a method step. Analogously, aspects described in the context of a method step also represent a description of a corresponding unit or item or feature of a corresponding apparatus.

The inventive decomposed signal can be stored on a digital storage medium or can be transmitted on a transmission medium such as a wireless transmission medium or a wired transmission medium such as the Internet.

Depending on certain implementation requirements, embodiments of embodiments can be implemented in hardware or in software. The implementation can be performed using a digital storage medium, for example a floppy disk, a DVD, a CD, a ROM, a PROM, an EPROM, an EEPROM or a FLASH memory, having electronically readable control signals stored thereon, which cooperate (or are capable of cooperating) with a programmable computer system such that the respective method is performed.

Some embodiments according to embodiments comprise a non-transitory data carrier having electronically readable control signals, which are capable of cooperating with a programmable computer system, such that one of the methods described herein is performed.

Generally, embodiments of the present invention can be implemented as a computer program product with a program code, the program code being operative for performing one of the methods when the computer program product runs on a computer. The program code may for example be stored on a machine readable carrier.

Other embodiments comprise the computer program for performing one of the methods described herein, stored on a machine readable carrier.

In other words, an embodiment of the inventive method is, therefore, a computer program having a program code for performing one of the methods described herein, when the computer program runs on a computer.

A further embodiment of the inventive methods is, therefore, a data carrier (or a digital storage medium, or a computer-readable medium) comprising, recorded thereon, the computer program for performing one of the methods described herein.

A further embodiment of the inventive method is, therefore, a data stream or a sequence of signals representing the computer program for performing one of the methods described herein. The data stream or the sequence of signals may for example be configured to be transferred via a data communication connection, for example via the Internet.

A further embodiment comprises a processing means, for example a computer, or a programmable logic device, configured to or adapted to perform one of the methods described herein.

A further embodiment comprises a computer having installed thereon the computer program for performing one of the methods described herein.

In some embodiments, a programmable logic device (for example a field programmable gate array) may be used to perform some or all of the functionalities of the methods described herein. In some embodiments, a field programmable gate array may cooperate with a microprocessor in order to perform one of the methods described herein. Generally, the methods are performed by any hardware apparatus.

While this invention has been described in terms of several advantageous embodiments, there are alterations, permutations, and equivalents which fall within the scope of this invention. It should also be noted that there are many alternative ways of implementing the methods and compositions of the present invention. It is therefore intended that the following appended claims be interpreted as including all such alterations, permutations, and equivalents as fall within the true spirit and scope of the present invention.

The above described embodiments are merely illustrative for the principles of the present invention. It is understood that modifications and variations of the arrangements and the details described herein will be apparent to others skilled in the art. It is the intent, therefore, to be limited only by the scope of the impending patent claims and not by the specific details presented by way of description and explanation of the embodiments herein.

Although each claim only refers back to one single claim, the disclosure also covers any conceivable combination of claims.

Claims

1. A circuitry for a correction of errors in a possibly erroneous binary word v′=v′1, . . . , v′n relative to a codeword v=v1, . . . , vn, the circuitry comprising: a syndrome generator for determining an error syndrome s=(s1, s3) according to a modified BCH code with a H-matrix Hmod comprising a first BCH submatrix H1mod and a second BCH submatrix H3mod, and with a code distance d≧5, wherein n′ column vectors of the BCH submatrix H1mod are paired as column vector pairs so that a componentwise XOR combination of the two column vectors of each column vector pair produces an identical column vector K that is different from all column vectors of the first BCH submatrix H1mod and where n′ is even and 4≦n′≦n applies,wherein the second BCH submatrix H3mod comprises a corresponding column vector for each column vector in the first BCH submatrix H1mod so that the corresponding column vector is a third power, according to Galois field arithmetic, of the column vector in the first BCH submatrix H1mod,wherein the syndrome generator is configured to determine the error syndrome s by multiplying the H-matrix Hmod with the possibly erroneous binary word v′ so that a first error syndrome portion is given by s1=H1mod·v′ and a second error syndrome portion is given by s3=H3mod·v′; anda decoder for generating a correction vector e=(e1, . . . , en) with correction values ej=ej+1=el=1 and et=0 for t≠j, j+1, l, if the first error syndrome portion s1 equals the componentwise XOR combination of the identical column vector K and a column vector at a column position l of the first BCH submatrix H1mod, and if the second error syndrome portion s3 equals the componentwise XOR combination of column vectors at column positions j, j+1, and l of the second BCH submatrix H3mod.

2. The circuitry according to claim 1, wherein the column vectors of the first BCH submatrix H1mod are designated as hj1 with j=1, . . . , n and are equal to a vector representation of elements αij of a Galois field GF(2m) with a being an element of the Galois field GF(2m), and the column vectors of the second BCH submatrix H3mod are designated as hj3 and are equal to a vector representation of elements α3ij of the Galois field GF(2m), and wherein m is a length of the column vectors of hj1 and hj3.

3. The circuitry according to claim 2, wherein α is a primitive element of the Galois field GF(2m).

4. The circuitry according to claim 1, wherein the decoder is configured so that it forms n binary correction values e1, . . . , en for correcting corresponding bits v′1, . . . , v′n of the binary word v′ from the components s1, s3, sP of the error syndrome s, wherein sP is an overall parity of the codeword v, and wherein the correction values e1, . . . , en are determined from the values of the syndrome components s1, s3, sP and the binary vector K so that the decoder Dec in case of a 1-bit error or a 2-bit error outputs a correction value ej=1 at its j-th output for j=1, . . . , n, when s13⊕s3⊕s12αij⊕s1α2ij=0and outputs a correction value ej=0 at its j-th output when s13⊕s3⊕s12αij⊕s1α2ij≠0and in case of a 3-bit error outputs a correction value ej=1 at its j-th output for j=1, . . . , n when s13⊕s3⊕s12K⊕s1K2⊕K2αij⊕Kα2ij=0ors1⊕J=αij and in case of a 3-bit error outputs a correction value e1=0 at its j-th output for j=1, . . . , n, when both s13⊕s3⊕s12K⊕s1K2⊕K2αIj⊕Kα2ij≠0ands1⊕K≠αij apply.

5. The circuitry according to claim 1, wherein the decoder comprises: an auxiliary subcircuit for forming at least one auxiliary signal on a basis of the first error syndrome portion s1, the second error syndrome portion s3, and the identical column vector K; anda plurality of decoder subcircuits, each decoder subcircuit being associated to a specific bit of the binary word v′, being configured for receiving the at least one auxiliary signal and for determining a corresponding correction value of the correction vector e.

6. The circuitry according to claim 5, wherein the auxiliary subcircuit is configured to form the following auxiliary signals: “3-bit error”, which is equal 1, when both s13⊕s3≠0 and sP=1, and which is equal 0 otherwise;an m bit wide signal (s13⊕s3)′ which is equal s13⊕s3 for (s1, s3)≠0, and equal

7. The circuitry according to claim 6, wherein a j-th decoder subcircuit for a corresponding bit v′j of the binary is configured to output a correction value ej=1, when s13⊕s3⊕s12αij⊕s1α2ij=0 ORs13⊕s3⊕s12K⊕s1K2⊕K2αij⊕Kα2ij=0 ORs1⊕K=αij,and to output a correction value ej=0, when s13⊕s3⊕s12αij⊕s1α2ij≠0 OR(s13⊕s3⊕s12K⊕s1K2⊕K2αij⊕Kα2ij≠0 AND s1⊕K≠αij).

8. The circuitry according to claim 1, wherein the possibly erroneous binary word v′ is an auxiliary read value provided by an auxiliary read value generator of a circuit for reading a memory with memory cells with more than two states.

9. The circuitry according to claim 1, wherein for n even at least one pair of the pairs of bits [v′1, v′2], [v′3, v′4] . . . , [v′n-1, v′n] and for n odd at least one pair of the pairs of bits [v′1, v′2],[v′3,v′4] . . . , [v′n-2, v′n-1] is stored in a memory cell with more than two states.

10. A method for correcting errors in a possibly erroneous binary word v′=v′1, . . . , v′n relative to a codeword v=v1, . . . , vn, the method comprising: determining an error syndrome s=(s1, s3) of a modified BCH code with a H-matrix Hmod comprising a first BCH submatrix H1mod and a second BCH submatrix H3mod, and with a code distance d≧5, wherein n′ column vectors of the submatrix H1mod are paired as column vector pairs so that a componentwise XOR combination of the two column vectors of each column vector pair produces an identical column vector K that is different from all column vectors of the BCH submatrix H1mod, and where n′ is even and 4≦n′≦n applies,wherein the second BCH submatrix H3mod comprises a corresponding column vector for each column vector in the first BCH submatrix H1mod so that the corresponding column vector is a third power, according to Galois field arithmetic, of the column vector in the first BCH submatrix H1mod,wherein the error syndrome s is determined by multiplying the H-matrix Hmod with the possibly erroneous binary word v′ so that a first error syndrome portion is given by s1=H1mod·v′ and a second error syndrome portion is given by s3=H3mod·v′; andgenerating a correction vector e=(e1, . . . , en) with correction values ej=ej+1=el=1 and et=0 for t≠j, j+1, l, if s1 equals the componentwise XOR combination of the identical column vector K and a column vector at a column position 1 of the first BCH submatrix H1mod, and s3 equals the componentwise XOR combination of column vectors at column positions j, j+1, and 1 of the second BCH submatrix H3mod.

11. A non-transitory storage medium having stored thereon a computer program having a program code for performing, when running on a computer, a method for correcting errors in a possibly erroneous binary word v′=v′1, . . . , v′n relative to a codeword v=v1, . . . , vn, the method comprising: determining an error syndrome s=(s1, s3) of a modified BCH code with a H-matrix Hmod comprising a first BCH submatrix H1mod and a second BCH submatrix H3mod, and with a code distance d≧5, wherein n′ column vectors of the submatrix H1mod are paired as column vector pairs so that a componentwise XOR combination of the two column vectors of each column vector pair produces an identical column vector K that is different from all column vectors of the BCH submatrix H1mod, and where n′ is even and 4≦n′≦n applies,wherein the second BCH submatrix H3mod comprises a corresponding column vector for each column vector in the first BCH submatrix H1mod so that the corresponding column vector is a third power, according to Galois field arithmetic, of the column vector in the first BCH submatrix H1mod,wherein the error syndrome s is determined by multiplying the H-matrix Hmod with the possibly erroneous binary word v′ so that a first error syndrome portion is given by s1=H1mod·v′ and a second error syndrome portion is given by s3=H3mod·v′; andgenerating a correction vector e=(e1, . . . , en) with correction values ej=ej+1=el=1 and et=0 for t≠j, j+1, l, if s1 equals the componentwise XOR combination of the identical column vector K and a column vector at a column position 1 of the first BCH submatrix H1mod, and s3 equals the componentwise XOR combination of column vectors at column positions j, j+1, and 1 of the second BCH submatrix H3mod.

12. A circuitry for a correction of errors in a possibly erroneous binary word v′=v′1, . . . , v′n, which, if errors occurred, resulted due to these errors from a codeword v=v1, . . . , vn of a linear code with a H-matrix Hmod of m′≧2m lines and n columns and with a code distance d≧5, and if no errors occurred, is equal to the codeword v=v1, . . . , vn of this code, wherein 1-bit errors, 2-bit errors and 3-bit errors containing a neighboring 2-bit error may be corrected, wherein n<2m and m≧3 applies, wherein the circuitry comprises a syndrome generator Synd for determining at least a (2m)-digit error syndrome s from the possibly erroneous binary word v′, and wherein the error correction is executed depending on the error syndrome determined by the syndrome generator by a decoder Dec downstream from the syndrome generator, wherein the syndrome generator Synd comprises n binary inputs and at least 2m binary outputs and is implemented so that when a binary word v′=v′1, . . . , v′n is input at its inputs, it outputs an error syndrome s=(s1, s3) assigned to the input binary word v′ at 2m binary outputs, wherein s1 and s3 are each m-digit binary vectors and the error syndrome s is determined by s=Hmod·v′and Hmod is a [m′,n] matrix of n column vectors h1, . . . , hn having m′ components each and for j=1, . . . , n the column vector hj consists of m first components forming the vector hj1, and of a second vector hj3 of the following m components of the column vector hj, so that the following applies

13. The circuitry according to claim 12, wherein the syndrome generator Synd comprises a (2m+1)-th output for outputting an additional bit sP of the error syndrome, wherein sP is determined as a parity bit sP=v′1⊕v′2⊕ . . . ⊕v′n, and wherein an additional (m+1)-th input of the decoder Dec exists for inputting the syndrome component sP.

14. The circuitry according to claim 13, wherein the decoder forms a parity syndrome sP by an XOR combination from those components of the error syndrome which correspond to the components of the columns of the H matrix Hmod which comprise an odd number of ones.

15. The circuitry according to claim 13, wherein the decoder Dec is configured so that it forms n binary correction values e1, . . . , en for correcting the corresponding bits v′1, . . . v′n of the binary word v′ from the components s1, s3, sP of the error syndrome s, wherein the correction values e1, . . . , en are determined from the values of the syndrome components s1, s3, sP and the binary vector K so that the decoder Dec in case of a 1-bit error or a 2-bit error outputs a correction value ej=1 at its j-th output for j=1, . . . , n, when s13⊕s3⊕s12αij⊕s1α2ij=0and outputs a correction value ej=0 at its j-th output when s13⊕s3⊕s12αij⊕s1α2ij≠0and in case of a 3-bit error outputs a correction value ej=1 at its j-th output for j=1, . . . , n when s13⊕s3⊕s12K⊕s1K2⊕K2αij⊕Kα2ij=0ors1⊕K=αij and in case of a 3-bit error outputs a correction value ej=0 at its j-th output for j=1, . . . , n, when both s13⊕s3⊕s12K⊕s1K2⊕K2αij⊕Kα2ij≠0ands1⊕K≠αij apply.

16. The circuitry according to claim 13, further comprising a combinational circuit for combining the correction values e1, . . . , en output by the decoder with the components v′1, . . . , v′n of a data word into corrected data bits v1cor, . . . , vncor, wherein the combinational circuit is configured to realize for j=1, . . . , n a respectively uniquely invertible combination of the correction value e1 with the corresponding data bit v′j.

17. The circuitry according to claim 12, wherein the columns h1, . . . , hn of the H matrix Hmod are determined so that the number of ones in a subset of components of these columns is odd.

18. The circuitry according to claim 12, wherein for n even at least one pair of the pairs of bits [v′1, v′2], [v′3, v′4] . . . , [v′n-1, v′n] and for n odd at least one pair of the pairs of bits [v′1, v′2], [v′3, v′4] . . . , [v′n-2, v′n-1] is stored in a memory cell with more than two states.

19. The circuitry according to claim 12, wherein Hmod is an [m′,n] matrix of n column vectors h1, . . . , hn each having m′ components and for j=1, . . . , n the column vector hj consists of m first components which form the vector hj1, of m following components which form the second vector hj3, and of m following components of the column vector hj which form a third vector hj5, so that

20. The circuitry according to claim 19, wherein the error indication circuit is implemented so that it realizes the function q=[s51]′[s52]′ . . . [s5m]′wherein([s51]′, . . . ,[s5m]′)=H5mod·vcor applies.

21. The circuitry according to claim 19, wherein the error indication circuit is implemented so that it realizes the function q=[s5i1]′ . . . [s5iM]′wherein([s51]′, . . . ,[s5m]′)=H5mod·vcor applies and {[s5i1]′, . . . ,[s5iM]′} is a subset of {[s51]′, [s52]′, . . . ,[s5m]′} and M<m applies.

22. The circuitry according to claim 12, wherein Hmod is a [m′,n] matrix of n column vectors h1, . . . , hn with m′ components each and for j=1, . . . , n the column vector hj consists of m first components forming the vector hj1 and a second vector hj3 consists of the following m components of the column vector hj and of a third vector hj5 of further m components, so that

23. The circuitry according to claim 12, wherein α is a primitive element of the Galois field GF(2m).

24. The circuitry according to claim 12, wherein subcircuits of the circuitry for correcting of errors in possibly erroneous binary word are partially jointly optimized.