CORRECTION OF BIT ERRORS

Abstract

Processing of a bit sequence is proposed, wherein (i) a first partial error syndrome s1 of an error syndrome and a second partial error syndrome s2 of the error syndrome are determined for the bit sequence, (ii) a first comparison value is determined on the basis of a bit position and the first partial error syndrome, (iii) a second comparison value is determined on the basis of the bit position and the second partial error syndrome, and (iv) the bit position is corrected should a comparison of the first comparison value with the second comparison value assume a specified value and otherwise the bit position is not corrected.

Description

TECHNICAL FIELD

The correction and/or detection of bit errors in binary data, for example sequences of a plurality of bits, using error codes is known.

BACKGROUND

An error code can be an error-correcting code, an error-detecting code, or a combination of an error-detecting and error-correcting code. An example of an error code is the Bose—Chaudhuri—Hocquenghem (BCH) code.

In the context of a BCH code, bit error positions can be corrected based on the zeros of a locator polynomial of the BCH code. A specific procedure for 2-bit error correction in a Galois field GF(2^m) can start with an appropriate second order locator polynomial, and this polynomial is transformed such that of the zeros of the transformed polynomial are determinable by a parameter of word width m. This parameter is determined from the syndrome components. Following their determination, the zeros of the transformed polynomial are to be transformed into the zeros of the original locator polynomial. Using a decoder, the bit positions to be corrected can be calculated from the zeros of the original locator polynomial.

A disadvantage here is that it takes a relatively long time until the erroneous bit positions can be corrected because the parameter of the transformed locator polynomial is initially determined from the syndrome components. After the zeros of the transformed polynomial have been determined, it is still necessary to transform the zeros of the transformed polynomial into the zeros of the original locator polynomial.

A further known option for error correction using a BCH code consists of initially determining, from the syndrome components of the error syndrome, the locator polynomial whose zeros then in turn specify the bit positions to be corrected. In this case, the zeros of the locator polynomial can be determined according to Chien's algorithm. In some regards, the locator polynomial is referred to as “error-location polynomial”. A disadvantage in this case is the great time outlay for determining the zeros of the locator polynomial.

SUMMARY

The present disclosure provides techniques for accelerating error correction and hence being able to carry out a faster error correction.

In some embodiments, a method is proposed for processing a bit sequence. The method determines a first partial error syndrome si of an error syndrome and a second partial error syndrome s₂ of the error syndrome for the bit sequence. The method also determines a first comparison value based on a bit position and the first partial error syndrome, and determines a second comparison value based on the bit position and the second partial error syndrome. A bit value at the bit position is changed to a corrected bit value when a comparison of the first comparison value with the second comparison value assume a specified value. Otherwise the bit value is not changed and, thus, is left at the bit value. The bit position can located in the error syndrome (e.g., the first partial errror syndrome and/or the second partial error syndrome), and thus could correspond to a transformed bit sequence, which is based on the bit sequence.

Consequently, an error at the bit position may be corrected (or else not corrected) depending on a result of the comparison. In this case, it is advantageous that the two comparison values are determined locally only in relation to the current bit position and which further bit position is erroneous need not be known at this time, for example even in the case of a 2-bit error correction.

In some embodiments, an error code capable of correcting at least a 2-bit error is used for the correction.

In some embodiments, the error code is a BCH code.

In some embodiments, the bit position is corrected should the first comparison value equal the second comparison value.

In a development, the first partial error syndrome Si and the second partial error syndrome S₂ each have m components of the error syndrome and m>2.

In a development, the bit position is assigned a value β, the first comparison value is determined as (S₁+β)^k and the second comparison value is determined as S₂+β^k, wherein k≠=0, and S₁, S₂ and β are elements of a Galois field.

By way of example, S₁, S₂ and β can be elements of the Galois field GF(2m). The operation “+” then denotes addition in the Galois field and (S₁+β)^k and β^k denote the k-th power of S₁+β and β, respectively, in the Galois field.

The approach described here can be used advantageously for correcting a bit position when a 2-bit error and/or a 1-bit error occur.

In some embodiments, the Galois field is GF(2^m)and m>2.

In some embodiments, r bit positions i₀, i₁, . . . , i_r−1 of the bit sequence are assigned r values αⁱ⁰, αⁱ¹, αⁱ², α^ir−1, wherein first comparison values and second comparison values are determined for bit positions to be corrected, wherein r≥1 and α is a generator of the Galois field.

In some embodiments, k=3 or k=−1.

In some embodiments, the method is used for error correction of a 2-bit error at bit positions i and j, wherein the first partial error syndrome is determined as s₁=αⁱ+^j and wherein the second partial error syndrome is determined as s₂=α^k·i+α^k·j. The comparison supplies a correction signal for correcting the i-th bit. The correction signal is determined such that it assumes a first value if (s₁+αⁱ)^k=s₂α^k·i applies, and assumes a second value if (s₁αⁱ)^k≠s₂α^k·i applies, wherein αⁱ, α^j, α^k·i, α^k·j are elements of a Galois field GF(2^m), m≥3 and k is an integer not equal to 0, and + denotes addition in the Galois field.

In some embodiments, a plurality of correction signals are determined, in each case one for a different bit, wherein the plurality of correction signals are at least partially determined with overlapping timing.

Consequently, the plurality of correction signals can be determined for example in parallel, e.g., simultaneously by parts of an item of hardware in particular. In this case, “parallel” can be understood to mean that the correction signals are not only determined in sequential fashion but are at least partially determined simultaneously. Consequently, there can be at least two correction signals which are determined simultaneously or parts which lead to these correction signals may be determined simultaneously.

In some embodiments, different correction signals are determined by means of different components.

Each of these components may comprise at least one hardware component and/or at least one software component.

In some embodiments, the first comparison value and the second comparison value are at least partially determined simultaneously.

In some embodiments, a third partial error syndrome is determined.

In some embodiments, the third partial error syndrome is a syndrome of a 3-bit error-correcting BCH code.

In some embodiments, the specified value in the correction of a 2-bit error is constant.

In some embodiments, in the correction of the 3-bit error, the specified value is determined on the basis of the respective bit position to be corrected.

In some embodiments, when the bit position to be corrected is an erroneous first bit of a 3-bit error, the specified value equals the value of the product of the second and the third bit position and the sum of the second and the third bit position, where the second bit position represents the position of the second bit error and the third bit position represents the position of the third bit error and where the product and the sum of the bit positions are determined in a Galois field GF(2^m).

In some embodiments, the specified value is determined on the basis of a symmetric function of the erroneous bit positions.

In some embodiments, the symmetric function is determined using partial error syndromes of a BCH code.

In some embodiments, a parity is determined for the bit sequence.

In some embodiments, the specified value is constant if the parity of the bit sequence equals 0 and the partial error syndrome si is not equal to 0.

In some embodiments, the specified value is constant if the parity of the bit sequence equals 1 and s₁³=s₂, where si is the first partial error syndrome and s₂ is the second partial error syndrome.

In some embodiments, a device for processing a bit sequence is also proposed, said device being configured to: determine a first partial error syndrome si of an error syndrome and a second partial error syndrome s₂ of the error syndrome for the bit sequence, determine a first comparison value on the basis of a bit position and the first partial error syndrome, determine a second comparison value on the basis of the bit position and the second partial error syndrome, correct the bit position should a comparison between the first comparison value and the second comparison value assume a specified value, not correct the bit position should the comparison between the first comparison value and the second comparison value not assume the specified value.

In some embodiments, the device comprises a plurality of components, more particularly processing units and/or circuits, on the basis of which the correction or the non-correction of bit positions is at least in part implementable in simultaneous fashion.

In some embodiments, the device is part of a memory or a memory system or is embodied separately from the memory or the memory system.

The device may be part of the memory or embodied separately from the latter. In particular, the device may comprise a processing unit which carries out the steps mentioned here. The processing unit may comprise at least one of the following components: a processor, a microcontroller, a circuit, an ASIC, an FPGA, a logic unit.

The processing unit can be realized in the form of at least one physical unit. In particular, the processing unit may be distributed among a plurality of physical units.

The explanations given above in relation to the methods apply accordingly to the device. The steps described here of the method may be carried out by means of the device.

The above-described properties, features and advantages of this invention and the manner in which they are achieved will be described below in conjunction with a schematic description of exemplary embodiments, which are explained in more detail in conjunction with the drawings. In this case, the same elements or elements with the same effect may be provided with identical reference signs for a better overview.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows an exemplary symbolic block diagram for a possible correction of a multiplicity of bit positions;

FIG. 2 shows a block diagram based on FIG. 1, additionally with a bit error coding signal, on the basis of which the correction can be switched inactive;

FIG. 3 shows a first alternative for realizing the comparison value generator, shown in FIG. 1, for processing the partial error syndrome s₁;

FIG. 4 shows a first alternative for realizing the comparison value generator, shown in FIG. 1, for processing the partial error syndrome s₂;

FIG. 5 shows a second alternative for realizing the comparison value generator, shown in FIG. 1, for processing the partial error syndrome s₁; and

FIG. 6 shows a second alternative for realizing the comparison value generator, shown in FIG. 1, for processing the partial error syndrome s₂.

DETAILED DESCRIPTION

By way of example, the determination of a correction value for a bit to be corrected is proposed. By way of example, the bit is part of a bit sequence. The correction value can be determined using an error code.

Error syndromes, in particular partial error syndromes, can be determined for the error code. By way of example, a first partial error syndrome and a second partial error syndrome may be determined. A first comparison value is ascertained on the basis of the first partial error syndrome and a bit position to be corrected, and a second comparison value is ascertained on the basis of the second partial error syndrome and the bit position to be corrected. If the two comparison values are identical, the correction value can be set equal to a first value (e.g., “1”). If the two comparison values are not equal, the correction value can be set equal to a second value (e.g., “0”).

In this context, the first comparison value and the second comparison value, and hence also the correction value for the bit position, in each case are determined advantageously directly from one of the partial error syndromes and the bit position, without it being necessary to ascertain a locator polynomial and the zeros thereof by way of a search method, for example.

If use is made of a Galois field GF(2^m) with m>2, each bit position to be corrected may be reversibly assigned an m-component value. On the basis of this value and the first syndrome component, it is possible to determine the first comparison value. Accordingly, the second comparison value can be determined using this value and the second syndrome component.

In one option, correction values for a plurality of bit positions, in particular for all bit positions, are determined according to this scheme. Preferably, correction values for at least two bit positions can be determined in parallel. In this case, “determined in parallel” means that the correction values are determined at least partly simultaneously (that is to say at least partly with overlapping timing), preferably using different hardware and/or software.

Various bit positions to be corrected

i₀,i₁, . . . , i_r can be reversibly uniquely assigned various valuesβ_i0, . . . , β_ir in pairs. The bit positions can be denoted by0, 1, . . . , rand the values can be denoted byβ₀, . . . , β_r in simplified notation.

The first comparison value for the bit position i_j to be corrected, with j=0, . . . , r can be determined as

(s₁β_ij)^k and the second comparison value for the bit position to be corrected can be determined ass₂β_ij^k,where s₁ denotes the first partial error syndrome, s₂ denotes the second partial error syndrome and k denotes a specified value.

Here, k is not equal to 0; s₁, s₂and β may be determined as values of a Galois field. If the Galois field is GF(2^m) with m>2, s₁, s₂ and β can be described as m-component binary vectors in their vector representation. Then, formation of exponentials is likewise determined in the corresponding Galois field.

In one option, all bit positions to be corrected

i₀,i₁,i₂, . . . i_n−1 are assigned pairwise different valuesβ_i0,β_i1, β_i2, . . . , β_in−1 of the Galois field. If a is a generator or primitive element of the Galois field,

${\beta }_{i_0}={\alpha }^{i_0},{\beta }_{i_1}={\alpha }^{i_1},{\beta }_{i_2}={\alpha }^{i_2},⋮$ ${\beta }_{i_{n-1}}={\alpha }^{i_{n-1}}$

is defined in one option.

A correction value Δ_r for the r-th bit can be determined as 1 if

(s₁+α^ir)^k=s₂+α^k·ir   (1) applies, and a correction value Δ_r can be determined as 0 if

(s₁+α^ir)^k≠s₂+α^k·ir   (2)

applies.

An advantageous variant consists of the term s₁+α^ir being realized by virtue of the components of si in which air equals 1 being inverted and the components of s₁ in which α^ir equals 0 remaining unchanged.

The addition of s₁ and α^ir in the Galois field GF(2^m), which corresponds to a component-by-component XOR operation on si and α^ir in their vector representation, can be realized by means of inverters which invert the components of s₁ in which air equals 1 and the components of s₁ in which α^ir equals 0 remain unchanged. Thus, to implement s₁+α^ir in this example, only inverters and no XOR gates are required.

Accordingly, the term s₂+α^k·ir can be realized by virtue of the components of s₂ in which α^k·ir equals 1 being inverted and the components of s₂ in which α^k·ir equals 0 remaining unchanged.

A bit error-correcting code which can correct at least two bit errors can be used as an error code. By way of example, an appropriate BCH code may be used. An option consists of incorporating a parity (that is to say at least one parity bit) into the error detection.

An H-matrix

$\begin{array}{cc}H=\left(\begin{array}{c}{H}_1\\ H_2\end{array}\right)=\left(\begin{array}{cccccc}{\alpha }^{i_0}& {\alpha }^{i_1}& {\alpha }^{i_2}& {\alpha }^{i_3}& \dots & {\alpha }^{i_{n-1}}\\ {\alpha }^{3i_0}& {\alpha }^{3i_1}& {\alpha }^{3i_2}& {\alpha }^{3i_3}& \dots & {\alpha }^{3i_{n-1}}\end{array}\right)& \left(3\right)\end{array}$

may be used to describe a 2-bit error-correcting BCH code. The values α^j are elements of the Galois field GF(2^m) in their vector representation as m-component binary vectors, wherein the exponents should be interpreted modulo 2^m−1. Furthermore, n≤2^m−1 applies.

A 2-bit error-correcting and 3-bit error-detecting BCH code with included parity may be described by an H-matrix

$\begin{array}{cc}H=\left(\begin{array}{c}{H}_1\\ H_2\\ P\end{array}\right)=\left(\begin{array}{cccccc}{\alpha }^0& {\alpha }^1& {\alpha }^2& {\alpha }^3& \dots & {\alpha }^{n-1}\\ {\alpha }^0& {\alpha }^3& {\alpha }^6& {\alpha }^9& \dots & {\alpha }^{3·\left(n-1\right)}\\ 1& 1& 1& 1& \dots & 1\end{array}\right)& \left(4\right)\end{array}$

for example. In this case, the following applies:

$i_0=0,i_1=1,i_2=2,⋮$ $i_r=r,⋮$ $i_{n-1}=n-1.$

If

x′=x′₀, . . . ,x′_n−1

is a binary vector to be corrected (that is to say a bit sequence which might be erroneous), then the error syndrome s is determined as follows:

s=H·x′  (5)

For an H matrix with

$H=\left(\begin{array}{c}{H}_1\\ H_2\end{array}\right)$

the first partial error syndrome s1 may be determined as

s ₁ =H ₁ ·x′  (6)

and the second partial error syndrome s2 may be determined as

s ₂ =H ₂ ·x′  (7)

with the matrix H₂ also being referred to as matrix H₃ and the partial error syndrome s₂ also being referred to as partial error syndrome s₃ in the case of a binary BCH code.

In the case of a 2-bit error, the partial error syndromes s₁ and s₂ of the error syndrome are determined as:

s ₁=αⁱ+α^j   (8)

s ₂=α^3·i+α^3·j   (9)

For the partial error syndrome

s _P=[1, . . . , 1]·x′

the following applies in the case of a 2-bit error:

s _P=1+1=0

For the i-th bit, the correction value Δ_i for a 2-bit error is as follows:

$\begin{array}{cc}{\Delta }_i=\{\begin{array}{c}1\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3=s_2+{\alpha }^{3·i}\\ 0\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3\ne s_2+{\alpha }^{3·i}\end{array}& \left(10\right)\end{array}$

If a 2-bit error is present in bit positions i and j, then taking account of equations (8) and (9) above

s ₁+αⁱ+αⁱ+α^j+αⁱ=α^j

s ₂+α³ⁱ=α³ⁱ+α^3j+α³ⁱ=α^3j

and hence

(s₁+αⁱ)³=α^3j=s₂+α³ⁱ

with the correction value Δ_i=1.

For r≠i,j, once again taking account of equations (8) and (9), the following applies:

s ₁+α^r=αⁱ+a^j+α^r

s ₂+α^3r=α³ⁱ+α^3j+α^3r

and hence

(s₁+α^r)³≠s₂+α^3r

with the correction value Δ_i=0.

The bit positions

$i_0=0,i_1=1,i_2=2,⋮$ $i_{n-1}=n-1$

are reversibly uniquely assigned the elements α⁰, α¹, α², αⁿ⁻¹ of the Galois field GF(2^m) in their vector representation as m-component binary vectors, for example.

For the bit position i to be corrected,

• • a first comparison value(s₁+αⁱ)³ is determined on the basis of the first partial error syndrome Si and the bit position i and
  • a second comparison values₂[αⁱ]³=s₂+α³ⁱ is determined on the basis of the second partial error syndrome s₂ and the bit position i.

If the first and second comparison values correspond, Δ_i=1 arises as correction value and the i-th bit position is corrected. If the two comparison values do not correspond, Δ_i=0 arises as correction value and there is no correction at the i-th bit position.

For a 1-bit error at the error position j, the following applies:

s₁=α^j

s₂=α^3·j

Hence, the first comparison value is determined as

(s₁+αⁱ)³=(α^j+αⁱ)³

and the second comparison value is determined as

(s₁+α^3·i=α^3·j+α^3·i

for for a bit position i to be corrected. If the error position j and the bit position i to be corrected correspond (that is to say i=j), then the following applies:

(s₁+αⁱ)³=(αⁱ+αⁱ)³=0³=0=s₂+α^3·i=α^3·i+α^3·i=0.

The first comparison value and the second comparison value are the same even in the case of the 1-bit error in position i and there is a correction of the bit position i.

If the error position j and the bit position i to be corrected do not correspond (that is to say i≠j), then the following applies:

(s₁+αⁱ)³=(α^j+αⁱ)³≠s₂+α^3·i+α^3·j+α^3·i

In this case, the first comparison value and the second comparison value are not the same, and so there is no correction at the bit position i since the error is present at another bit position j.

A comparison of the first comparison value and of the second comparison value, which are each determined as m-component binary vectors, can be implemented for example by a component-by-component exclusive or operation (XOR operation) and a subsequent or operation (OR operation).

EXAMPLE: H-MATRIX WITH PARITY

A further example of an H-matrix of a code for 2-bit error correction and 3-bit error detection is as follows:

$\begin{array}{cc}H=\left(\begin{array}{c}{H}_1\\ H_{-1}\\ P\end{array}\right)=\left(\begin{array}{cccccc}{\alpha }^0& {\alpha }^1& {\alpha }^2& {\alpha }^3& \dots & {\alpha }^{n-1}\\ {\alpha }^0& {\alpha }^{-1}& {\alpha }^{-2}& {\alpha }^{-3}& \dots & {\alpha }^{-\left(n-1\right)}\\ 1& 1& 1& 1& \dots & 1\end{array}\right).& \left(11\right)\end{array}$

The partial error syndromes s₁ and s₂ of the error syndrome arise as follows for the 2-bit error:

s ₁=αⁱ+α^j   (12)

s ₂=α⁻ⁱ+α^−j   (13)

The values α^j again are elements of the Galois field GF(2^m) in their vector representation as m-component binary vectors, wherein the exponents should be interpreted modulo 2^m−1. Furthermore, n≤2^m−1 applies.

For the i-th bit, the correction value Δ_i for the 2-bit error is determined as follows:

$\begin{array}{cc}{\Delta }_i=\{\begin{array}{c}1\mathrm{for}{\left(s_1+{\alpha }^i\right)}^{-1}=s_2+{\alpha }^{-1}\\ 0\mathrm{for}{\left(s_1+{\alpha }^i\right)}^{-1}\ne s_2+{\alpha }^{-1}\end{array}& \left(14\right)\end{array}$

The first comparison value for the bit position i is

(s₁+αⁱ)⁻¹

and the second comparison value for this bit position is

s₂+α⁻ⁱ

Further Exemplary Embodiments

FIG. 1 shows an exemplary diagram for r bit positions 0 to r−1 to be corrected.

An m-bit wide partial error syndrome Si is applied to an m-bit wide input 113, which is connected to the inputs of r first comparison value generators 11 to 13. The comparison value generators 11 to 13 are suited to determining a first comparison value per bit position on the basis of the partial error syndrome s₁ and on the basis of the bit position.

An m-bit wide partial error syndrome s₂ is applied to an m-bit wide input 114, which is connected to the inputs of r second comparison value generators 14 to 16. The comparison value generators 14 to 16 are suited to determining a second comparison value per bit position on the basis of the partial error syndrome s₂ and on the basis of the bit position.

Now, a comparison of the first comparison value with the second comparison value is carried out for each bit position by means of a comparator. To this end, r comparators 17 to 19 are provided, the inputs of which are connected to the appropriate outputs of the comparison value generators.

A respective correction value Δ₀ to Δ_{r −1} for the respective bit position is provided at the output of the comparator 17 to 19. By way of example, the correction value is 1 if the comparison value generators connected to the comparator supply the same output signals and the correction value is 0 if the comparison value generators supply unequal output signals.

The output of the respective comparator 17 to 19 is connected to the input of an XOR gate 110 to 112. A possibly erroneous bit x₀¹ to x_r−1¹ is applied to the other input of the respective XOR gate 110 to 112. A corrected bit

$x_0^{\mathrm{cor}}=x_0^{\prime }+{\Delta }_0,⋮$ $x_{r-1}^{\mathrm{cor}}=x_{r-1}^{\prime }+{\Delta }_{r-1}$

is provided at the output of the XOR gate 110 to 112. Here, + denotes addition modulo 2.

In this case, a bit is referred to as corrected bit if a correction value has been added to a possibly erroneous bit. Here, the correction value can equal 1 if the bit was erroneous and the correction value can equal 0 if the bit was not erroneous. Independently of whether or not the bit was erroneous, the resultant bit following the addition of the correction value is referred to as corrected bit in this case. The addition is implemented modulo 2 and can be realized by an XOR operation.

As already explained above, the determination of the corrected bit is implemented for all bit positions. By way of example, the following applies to the bit position 0: The output of the comparison value generator 11 is connected to the first input of the comparator 17 and the output of the comparison value generator 14 is connected to the second input of the comparator 17. The output of the comparator 17 supplies the correction value Δ₀ to the first input of the XOR gate 110, at the second input of which the possibly erroneous bit x′₀ is applied. At its output, the XOR gate 110 consequently supplies the corrected bit x₀^corr. This is implemented accordingly for the further bit positions 1 to r−1.

FIG. 2 shows an exemplary arrangement on the basis of the representation shown in FIG. 1. An AND gate 21 to 23 is arranged between the output of the respective comparator 17 to 19. The first input of the AND gate 21 to 23 is connected to the output of the comparator 17 to 19 and the second input of the AND gate 21 to 23 is connected to an image error coding signal 201. The output of the AND gate 21 to 23 is connected to the XOR gate 110 to 112.

The arrangement shown in FIG. 2 once again applies per bit position. By way of example, for the bit position 0, the output of the comparator 17 is connected to the first input of the AND gate 21. The second input of the AND gate is connected to the bit error coding signal 201 The output of the AND gate 21 is connected to the first input of the XOR gate 110. These explanations apply accordingly to the further bit positions 1 to r−1, with the bit error coding signal 201 being applied to all second inputs of the AND gates 21 to 23.

The bit error coding signal 201 is for example a binary value 1 if a 1-bit error or a 2-bit error was determined on account of the error syndrome, and a binary value 0 if the error syndrome equals 0, and hence no errors are indicated by the error syndrome. Consequently, what follows from the link to the respective AND gate 21 to 23 is that no change of the value of the bit x′₀ to x′_r−1 applied to the XOR gate 110 to 112 is implemented, that is to say no correction is carried out.

FIG. 3 shows an exemplary configuration of the comparison value generators 11 to 13 of FIG. 1.

By way of example, a 2-bit error-correcting BCH code with an H-matrix as per equation (4) is used. The possibly to be corrected bit positions are the bit positions 0 to r−1.

Each comparison value generator 11 to 13 comprises a series connection of an adding member 31 to 33 and a third power generator 34 to 36.

The structure of the comparison value generator 11 for the bit position 0 can be described as follows: The comparison value generator 11 comprises the series connection of the adding member 31 and the third power generator 34. The adding member 31 facilitates a component-by-component XOR operation on the first partial error syndrome s₁ and α⁰ in its vector representation. It is an option to realize the XOR operation on the partial error syndrome s₁ and α⁰ by virtue of inverting those components of s₁ where α⁰ equals 1 and leaving the components of s₁ in which α⁰ equals 0 unchanged. In a Galois field GF(2^m)

${\alpha }^0=\left(\underset{\underset{m}{︸}}{0\mathrm{\dots 01}}\right),$

and so the first component of s₁ is inverted while all other components of s₁ remain unchanged.

The third power generator 34 is provided to form and output, in the Galois field GF(2^m), the third power (s₁+α⁰)³ of the value (s₁+α⁰) applied to its input. The third power generator 34 can be realized as a combinational circuit.

This is implemented accordingly for the further bit positions 1 to r−1, wherein the respective adding member correspondingly carries out an addition with α¹ to α^r−1.

FIG. 4 shows an exemplary configuration of the comparison value generators 14 to 16 of FIG. 1.

Each comparison value generator 14 to 16 comprises an adding member 41 to 43.

The structure of the comparison value generator 14 for the bit position 0 can be described as follows: The comparison value generator 14 comprises the adding member 41 in order to XOR-link, component-by-component, the partial error syndrome s₂ and α⁰ functionally in its vector representation. It is an option to realize the XOR operation on the partial error syndrome s₂ and α⁰ by virtue of inverting those components of s₂ where α⁰ equals 1 and leaving the components of s₂ where α⁰ equals 0 unchanged.

This is implemented accordingly for the further bit positions j (with j=1, . . . , r−1), wherein the respective adding member accordingly carries out an addition with α^3j.

The Third Power Generator

A third power generator in a Galois field GF(2^m) can be realized as a combinational circuit with m binary inputs and m binary outputs. It is one option to describe the combinational functionality by a value table and to realize the value table as a circuit using a synthesis tool.

For i=0, . . . , 2^m−1, the input value αⁱ in its vector representation as a m-component binary vector is assigned the corresponding output value α^3·i in such a value table. The input value

$\left(\underset{\underset{m}{︸}}{0\mathrm{\dots 0}}\right)$

can be assigned the output value

$\left(\underset{\underset{m}{︸}}{0\mathrm{\dots 0}}\right).$

The exponents should be interpreted modulo 2^m−1 and a is a generator of the Galois field.

For elucidating purposes, a Galois field GF(2³) with the modular polynomial

M(x)=x³+x+1

is considered, for which the elements α⁰ to α⁶ are represented in their vector representation in the following table “Exponential representation and vector representation”, wherein the element 0 is shown as 000:

Exponential representation Vector representation
0                          000
α0                         001
α1                         010
α2                         100
α3                         011
α4                         110
α5                         111
α6                         101

Value tables for Galois fields up to GF(2¹⁰) are shown in [Lin, S.; Costello, D.: Error Control Coding, Prentice Hall, 1983, pages 561 to 577] for example.

The third power generator outputs α^3·i on the basis of an input of αⁱ, wherein the exponents should be interpreted modulo 2³−1=7 in this case. That is to say, the following relationships arise between input and output:

Input Output
α0    α0
α1    α3
α2    α6
α3    α2
α4    α5
α5    α1
α6    α4

A corresponding binary value table on the basis of the vector representations of the values αⁱ shown above is elucidated by the following table of the third power generator in GF(2³):

Input value Output value
000         000
001         001
010         011
011         100
100         101
101         110
110         111
111         010

Example: Comparison Value Generator with Galois Field Inverter

FIG. 5 shows an exemplary alternative configuration of the comparison value generators 11 to 13 of FIG. 1.

A 2-bit error-correcting code with an H-matrix as per equation (11) is used in exemplary fashion. The bit positions possibly to be corrected are the bit positions 0 to r−1.

Each comparison value generator 11 to 13 comprises a series connection of an adding member 51 to 53 and a Galois field inverter 54 to 56.

The structure of the comparison value generator 11 for the bit position 0 can be described as follows: The comparison value generator 11 comprises the series connection of the adding member 51 and the Galois field inverter 54. The adding member 51 facilitates a component-by-component XOR operation on the first partial error syndrome s₁ and α⁰ in its vector representation. The Galois field inverter 54 is provided to form and output, in the Galois field GF(2^m), (s₁+α⁰)⁻¹ on the basis of the value (s₁+α⁰) applied to its input. The Galois field inverter 54 can be realized as a combinational circuit.

This is implemented accordingly for the further bit positions 1 to r−1, wherein the respective adding member correspondingly carries out an addition with α¹ to α^r−1.

FIG. 6 shows an exemplary configuration of the comparison value generators 14 to 16 of FIG. 1.

Each comparison value generator 14 to 16 comprises an adding member 61 to 63.

The structure of the comparison value generator 14 for the bit position 0 can be described as follows: The comparison value generator 14 comprises the adding member 61 in order to XOR-link, component-by-component, the partial error syndrome s₂ and α⁰ functionally in its vector representation.

This is implemented accordingly for the further bit positions j (with j=1, . . . , r−1), wherein the respective adding member correspondingly carries out an addition with α^j.

The Galois field inverter in a Galois field GF(2^m) can be realized as a combinational circuit with m binary inputs and m binary outputs. It is one option to describe the combinational functionality thereof by a value table and to realize the value table as a circuit using a synthesis tool.

For i=0, . . . ,2^m−1, the input value αⁱ in its vector representation as a m-component binary vector is assigned the corresponding output value α⁻ⁱ in such a value table. The input value

$\left(\underset{\underset{m}{︸}}{0\mathrm{\dots 0}}\right)$

can be assigned the output value

$\left(\underset{\underset{m}{︸}}{0\mathrm{\dots 0}}\right)$

The exponents should be interpreted modulo 2^m−1 and α is a generator of the Galois field. Furthermore, α⁰=α⁻⁰ applies.

For elucidating purposes, the Galois field GF(2³) with the modular polynomial

M(x)=x³+x+1

is considered, for which the elements α⁰ to α⁶ are represented in the table labeled “Exponential representation and vector representation” above.

The Galois field inverter outputs α⁰ when α⁰ is input, and outputs α⁷⁻ⁱ when αⁱ is input, wherein the exponents should be interpreted modulo 2³−1=7 in this case. That is to say, the following applies:

Input Output
α0    α0
α1    α6
α2    α5
α3    α4
α4    α3
α5    α2
α6    α1

A corresponding binary value table on the basis of the vector representations of the values αⁱ shown above is elucidated by the following table of the Galois field inverter in GF(2³):

Input value Output value
000         000
001         001
010         101
011         110
100         111
101         010
110         011
111         100

Fast correction of 3-bit errors

A fast correction of 3-bit errors is presented below in exemplary fashion. To this end, a 3-bit error-correcting BCH code may be used.

A 3-bit error-correcting BCH code may be described by an H-matrix

$\begin{array}{cc}H=\left(\begin{array}{c}{H}_1\\ H_2\\ H_3\end{array}\right)=\left(\begin{array}{cccccc}{\alpha }^{i_0}& {\alpha }^{i_1}& {\alpha }^{i_2}& {\alpha }^{i_3}& \dots & {\alpha }^{i_{n-1}}\\ {\alpha }^{3i_0}& {\alpha }^{3i_1}& {\alpha }^{3i_2}& {\alpha }^{3i_3}& \dots & {\alpha }^{3i_{n-1}}\\ {\alpha }^{5i_0}& {\alpha }^{5i_1}& {\alpha }^{5i_2}& {\alpha }^{5i_3}& \dots & {\alpha }^{5i_{n-1}}\end{array}\right)& \left(15\right)\end{array}$

By way of example, the values a′ are elements of the Galois field GF(2^m) in their vector representation as m-component binary vectors, wherein the exponents should be interpreted modulo (2^m−1). The following apply: n≤2^m−1 and m>2.

Additionally, a 3-bit error-correcting BCH code may be described by an H-matrix:

$\begin{array}{cc}H=\left(\begin{array}{c}{H}_1\\ H_2\\ H_3\end{array}\right)=\left(\begin{array}{cccccc}{\alpha }^0& {\alpha }^1& {\alpha }^2& {\alpha }^3& ...& {\alpha }^{n-1}\\ {\alpha }^0& {\alpha }^3& {\alpha }^6& {\alpha }^9& ...& {\alpha }^{3·\left(n-1\right)}\\ {\alpha }^0& {\alpha }^5& {\alpha }^{10}& {\alpha }^{15}& ...& {\alpha }^{5·\left(n-1\right)}\end{array}\right)& \left(16\right)\end{array}$

In this case, the following apply:

$\begin{array}{c}{i}_0=0,\\ i_1=1,\\ i_2=2,\\ ...\\ i_r=r,\\ ⋮\\ i_{n-1}=n-1\end{array}$

If

x′=x′₀, . . . , x′_n−1

is a binary vector to be corrected, then the error syndrome s is determined as

s=H·x′  (17)

For the H-matrix with

$H=\left(\begin{array}{c}{H}_1\\ H_2\\ H_3\end{array}\right)$

the first partial error syndrome s₁ can be determined as

s ₁ =H ₁ ·x′  (18)

the second partial error syndrome s₂ can be determined as

s ₂ =H ₂ ·x′  (19)

and the third partial error syndrome s₃ can be determined as

s _x =H ₃ ·x′  (20)

In the case of a binary BCH code, the matrix H₂ can be referred to as matrix H₃, the partial error syndrome s₂ can be referred to as partial error syndrome s₃, the matrix H₃ can be referred to as matrix H₅ and the partial error syndrome s₃ can be referred to as partial error syndrome s₅.

While two partial error syndromes, for example the partial error syndromes s₁ and s₂, are used for the correction of a 2-bit error, the partial error syndrome s₃ is additionally used to correct a 3-bit error.

One option is to use the parity of bits for error detection and/or error correction. A 3-bit error-correcting BCH code with parity can be described by an H-matrix.

$\begin{array}{cc}H=\left(\begin{array}{c}{H}_1\\ H_2\\ H_3\\ H_P\end{array}\right)=\left(\begin{array}{cccccc}{\alpha }^{i_0}& {\alpha }^{i_1}& {\alpha }^{i_2}& {\alpha }^{i_3}& ...& {\alpha }^{i_{n-1}}\\ {\alpha }^{3i_0}& {\alpha }^{3i_1}& {\alpha }^{3i_2}& {\alpha }^{3i_3}& ...& {\alpha }^{3i_{n-1}}\\ {\alpha }^{5i_0}& {\alpha }^{5i_1}& {\alpha }^{5i_2}& {\alpha }^{5i_3}& ...& {\alpha }^{5i_{n-1}}\\ 1& 1& 1& 1& ...& 1\end{array}\right)& \left(21\right)\end{array}$

where H_P denotes a parity matrix. Alternatively, a 3-bit error-correcting BCH code with parity may also be described by the H-matrix

$\begin{array}{cc}H=\left(\begin{array}{c}{H}_1\\ H_2\\ H_3\\ H_P\end{array}\right)=\left(\begin{array}{cccccc}{\alpha }^0& {\alpha }^1& {\alpha }^2& {\alpha }^3& ...& {\alpha }^{n-1}\\ {\alpha }^0& {\alpha }^3& {\alpha }^6& {\alpha }^9& ...& {\alpha }^{3·\left(n-1\right)}\\ {\alpha }^0& {\alpha }^5& {\alpha }^{10}& {\alpha }^{15}& ...& {\alpha }^{5·\left(n-1\right)}\\ 1& 1& 1& 1& ...& 1\end{array}\right)& \left(22\right)\end{array}$

In this respect, reference is made to equations (15) and (16) above, which were augmented by the matrix H_P.

With the binary vector x′ to be corrected, the error syndrome s arises as

s=H·x′  (23)

With

$H=\left(\begin{array}{c}{H}_1\\ H_2\\ H_3\\ H_P\end{array}\right)$

the following follows for the partial error syndromes s1, s2, s3 and s4:

s ₁ =H ₁ ·x′

s ₂ =H ₂ ·x′

s ₃ =H ₃ ·x′

s ₄ =s _P =H _P ·x′=x ₀ +x ₁ +. . . +x _n−1.

In this case, “+” denotes addition modulo 2, corresponding to an XOR operation. The partial error syndrome s_P is the parity of the bits x₀ to x_n−1.

The partial error syndromes s₁, s₂ and s₃ of the error syndrome arise for the 3-bit error in bit positions i, j and k as

s ₁=αⁱ+α^j+α^k   (24)

s ₂=α³ⁱα^3jα^3k   (25)

s ₃=α⁵ⁱ+α^5j+α^5k   (26)

The bit positions i, j and k may also be described, for example, by the values aⁱ, a^j, a^k, by elements of the Galois field GF(2^m), by the binary values x′_i, x′_j, x′_k to be corrected or by i, j and k.

For a 3-bit error in bit positions i, j and k, the following applies:

(s₁+αⁱ)³=s₂+α³ⁱ+α^j·α^k)α^j+α^k)   (27)

This emerges by recalculation. In this case, α^j·α^k is the product of the elements α^j and α^k in the Galois field GF(2^m) and α^j+α^k denotes their sum in the Galois field, which may be determined for example in the vector representation of α^j and α^k as m-component of binary vectors by the component-by-component XOR operation on these binary vectors.

In the case of a 3-bit error in bit positions aⁱ, α^j and α^k, the partial error syndromes s₁ and s₂ are determined by equations (24) and (25).

With

s ₁+αⁱ=α^j+α^k   (28)

s ₂+α³ⁱ=α^3j+α^3k   (29)

it is possible to verify by insertion into equation (27) that:

$\begin{array}{c}{\left(s_1+{\alpha }^i\right)}^3={\left(s_1+{\alpha }^i\right)}^2·\left(s_1+{\alpha }^i\right)=\\ ={\left({\alpha }^j+{\alpha }^k\right)}^2·\left({\alpha }^j+{\alpha }^k\right)=\\ =\left({\alpha }^{2j}+{\alpha }^j·{\alpha }^k+{\alpha }^j·{\alpha }^k+{\alpha }^{2k}\right)·\left({\alpha }^j+{\alpha }^k\right)=\\ =\left({\alpha }^{2j}+{\alpha }^{2k}\right)·\left({\alpha }^j+{\alpha }^k\right)=\\ ={\alpha }^{3j}+{\alpha }^{2j}·{\alpha }^k+{\alpha }^{2k}·{\alpha }^j+{\alpha }^{3k}=\\ ={\alpha }^{3j}+{\alpha }^{3k}+{\alpha }^j·{\alpha }^k·\left({\alpha }^j+{\alpha }^k\right)=\\ =s_2+{\alpha }^{3i}+{\alpha }^j·{\alpha }^k·\left({\alpha }^j+{\alpha }^k\right).\end{array}$

In this case, it should be observed that addition modulo 2 of two equal values yields 0.

In the case of a 3-bit error in bit positions i, j and k, a comparison of a first value (s₁+αⁱ)³ with a second value s₂+α³ⁱ yields the specified value α^jα^k(α^j+α^k) if an error is corrected in the bit position i. If (s₁+αⁱ)³ and s₂+α³ⁱ differ by a different value than α^jα^k(α^j+α^k), there is no correction.

To be able to determine for the i-th bit position whether a 3-bit error, where the i-bit is erroneous and corrected, is present, for example the value

α^jα^k(α^j+α^k)   (30)

is determined on the basis of the partial error syndromes and only the bit position i. The two further erroneous bit positions j and k need not be determined for the correction of the i-th bit position.

The value (α^j+α^k) from equation (30) of the two further erroneous bit positions can be determined on the basis of the i-th bit position αⁱ and on the basis of the partial error syndrome s₁:

α^j+α^k=s₁+αⁱ

This follows immediately from equation (24).

The value α^jα^k from equation (30) can be represented on the basis of the i-th bit position using the symmetric function σ₂

σ₂=αⁱα^j+αⁱα^k+α^jα^k   (31)

as

$\begin{array}{cc}\begin{array}{c}{\alpha }^j{\alpha }^k={\sigma }_2+{\alpha }^i{\alpha }^j+{\alpha }^i{\alpha }^k=\\ ={\sigma }_2+{\alpha }^i·\left({\alpha }^j+{\alpha }^k\right)=\\ ={\sigma }_2+{\alpha }^i·\left(s_1+{\alpha }^i\right)\end{array}& \left(32\right)\end{array}$

Hence, the following arises from equation (27) in conjunction with equation (32) and equation (28):

(s₁+αⁱ)³=s₂+α³ⁱ+(σ₂+αⁱ(s₁+αⁱ))·(s₁+αⁱ)   (33)

For σ₂, the following furthermore emerges by means of the partial error syndromes s₁, s₂ and s₃ as per equations (24), (25) and (26) which in turn are symmetric functions of the bit positions to be corrected:

${\sigma }_2=\frac{s_1^2·s_2+s_3}{s_1^3+s_2}.$

This can be verified by insertion into equations (24), (25) and 26.

Consequently, a common value of the symmetric function σ₂, which is determined by the syndrome components s₁, s₂ and s₃, can be used for all bit positions to be corrected.

By way of example, a binary correction value Δ_i for the i-th bit position in the case of a 3-bit error is determinable as per

$\begin{array}{cc}{\Delta }_i=1\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3=s_2+{\alpha }^{3i}+\left({\sigma }_2+{\alpha }^i\left(s_1+{\alpha }^i\right)\right)\left(s_1+{\alpha }^i\right),{\Delta }_i=0\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3\ne s_2+{\alpha }^{3i}+\left({\sigma }_2+{\alpha }^i\left(s_1+{\alpha }^i\right)\right)\left(s_1+{\alpha }^i\right).& \left(34\right)\end{array}$

A first comparison value (s₁+αⁱ)³ is determined on the basis of the bit position i and the first partial error syndrome s₁. A second comparison value s₂+α³ⁱ is determined on the basis of the bit position i and the second partial error syndrome s₂.

The bit position i is corrected with Δ_i=1 if the comparison of the first comparison value to the second comparison value assumes a specified value, in this case the specified value

(α^jα^k)(α^j+α^k)=(σ₂αⁱ(s₁+αⁱ))(s₁+60ⁱ)   (35)

The bit position i is not corrected, or is corrected with Δ_i=0, if the comparison of the first comparison value to the second comparison value does not assume the specified value as per equation (35).

An advantage of the method presented here for forming a correction value for the i-th bit is, for example, that values that depend on the syndrome components can be formed and provided only once (e.g., centrally). Thus, for example the value

s₁σ₂

can be determined once and can be provided for a plurality of bit positions. Advantageously, the formation of the term s₁+αⁱ does not require a component-by-component XOR operation. Instead, inverters can be used in simplified fashion to this end: The components of s₁ in which αⁱ equals 1 are inverted; components in which αⁱ equals 0 remain unchanged. Corresponding explanations apply to the addition of α³ⁱ.

Likewise, the value

s₁²+σ₂

need only be formed once and can be provided as an m-bit wide value for all bit positions to be corrected. In each bit position i to be corrected, all that still is required is a multiplication by the constant αⁱ. For the i-th bit position, the value s₁+αⁱ can advantageously be realized by inverting the bits of the m-bit vector s₁ in which the bits of αⁱ equal 1. It is consequently possible to dispense with a component-by-component XOR operation, significantly reducing the outlay.

In an alternative embodiment, the rewritten form

(σ₂+αⁱ(s₁+αⁱ))(s₁+αⁱ)=s₁σ₂+αⁱ(s₁²+σ₂)+α³ⁱ

can be used. What then follows from equation (34) is that

$\begin{array}{cc}{\Delta }_i=1\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3=s_2+{\alpha }^{3i}+s_1{\sigma }_2+{\alpha }^i\left(s_1^2+{\sigma }_2\right)+{\alpha }^{3i},{\Delta }_i=0\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3\ne s_2+{\alpha }^{3i}+s_1{\sigma }_2+{\alpha }^i\left(s_1^2+{\sigma }_2\right)+{\alpha }^{3i}.& \left(36\right)\end{array}$

A further option consists, in equation (27), of representing the value α^j·α^k on the basis of the i-th bit position using a symmetric function σ₃

σ₃=αⁱα^jα^k   (37)

as

α^jα^k=σ₂·α⁻ⁱ   (38)

The following then emerges from equation (27):

(s₁+αⁱ)³=s₂+α³ⁱ+σ₃·(s₁+αⁱ)·α⁻ⁱ   (39)

Using the partial error syndromes s₁, s₂and the symmetric function σ₂, the symmetric function σ₃ can be represented by:

σ₃=s₁³+s₂s₁·σ₂   (40)

This can be verified by inserting the values s₁, s₂ and σ₂.

If 1-bit errors, 2-bit errors and 3-bit errors should be corrected, it is an option, if a 1-bit error or a 2-bit error occurs, to determine a correction value Δ_i for the i-th bit such that

$\begin{array}{cc}{\Delta }_i=1\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3=s_2+{\alpha }^{3i}{\Delta }_i=0\mathrm{for}{\left(s_1+{\alpha }^i\right)}^3\ne s_2+{\alpha }^{3i}& \left(41\right)\end{array}$

applies, and that equation (34), for example, applies in the case of a 3-bit error.

In the case of a 1-bit error or a 2-bit error, a comparison value between

(s₁+αⁱ)³

and

s₂+α³ⁱ

is constant and identical for all bit positions. In the case of a 3-bit error, this comparison value is determined by the bit position i, and the values of the symmetric function σ₂ and the component of the error syndrome s₁, which are the same for all bit positions, because, in accordance with the explanations above, the additional term (σ₂+αⁱ(s₁+αⁱ))(s₁+αⁱ) is also added.

Even if a 2-bit error is present, the correction value Δ_i for the i-th bit can be determined according to equation (34), because for the 2-bit error in bit positions i and j with

${\sigma }_2=\frac{s_3+s_1^2{s}_3}{s_1^3+s_2},s_1={\alpha }^i+{\alpha }^j,s_2={\alpha }^{3i}+{\alpha }^{3j},s_3={\alpha }^{5i}+{\alpha }^{5j}$

σ₂ is determined as σ₂=αⁱ·α^j

Hence,

(σ₂+αⁱ(s₁+αⁱ))(s₁+αⁱ)=0

and in the case of a 2-bit error, the correction value is determinable by

(s₁+αⁱ)³+s₂+α³ⁱ

as required.

Whether there is no error or whether there is a 1-bit error or whether there is a 2-bit error or whether there is a 3-bit error can be determined using the error syndrome components s₁, s₂, s₃ and s₄=s_P:

• • If no error is present, all error syndrome components s₁, s₂, s₃ and s₄=s_P are equal to 0. Otherwise an error is present.
  • If a 1-bit error is present, s₁³=s₂ and s_P=1 then and applies.
  • If a 2-bit error is present, s_P=0 then s₁³≠s₂ and applies.
  • If a 3-bit error is present, s_P=1 and s₁³≠s₂ applies.

Even though the invention was illustrated more closely and described in detail by the at least one exemplary embodiment shown, the invention is not restricted thereto, and other variations can be derived herefrom by a person skilled in the art without departing from the scope of protection of the invention.

Claims

1. A method for processing a bit sequence, determining a first partial error syndrome s1 of an error syndrome and a second partial error syndrome s2 of the error syndrome for the bit sequence,determining a first comparison value based on a bit position and the first partial error syndrome,determining a second comparison value based on the bit position and the second partial error syndrome, andchanging a bit value at the bit position to a corrected bit value when a comparison of the first comparison value with the second comparison value is a specified value, and otherwise leaving the bit value at the bit position when the comparison of the first comparison value with the second comparison value is not the specified value.

2. The method of claim 1, wherein the changing of the bit value to the corrected bit value uses an error code capable of correcting at least a 2-bit error.

3. The method of claim 2, wherein the error code is a —-Chaudhuri—Hocquenghem (BCH) code.

4. The method of claim 1, wherein the bit value is changed to the corrected bit value when the first comparison value equals the second comparison value.

5. The method of claim 1, wherein the first partial error syndrome s1 and the second partial error syndrome s2 each have m components of the error syndrome and m>2.

6. The method of claim 1, wherein the bit value is assigned a value β,wherein the first comparison value is determined as (s1β)k and the second comparison value is determined as s2+βk,wherein k≠0, and s1, s2 and β are elements of a Galois field.

7. The method of claim 6, wherein the Galois field is GF(2m)and m>2.

8. The method of claim 6, wherein r bit positions i0, i1, . . . , ir−1 of the bit sequence are assigned r values αi0, αi1, αi2, αir−1,wherein first comparison values

9. The method of claim 6, wherein k=3 or k=−1.

10. The method of claim 1 for error correction of a 2-bit error at bit positions i and j, wherein the first partial error syndrome is determined as s1=αi+αj wherein the second partial error syndrome is determined as s2=αk·i+αk·j wherein the comparison supplies a correction signal for correcting the i-th bit,wherein the correction signal is determined such that it assumes a first value if (s1+αi)k=s2+αk·i applies, and assumes a second value if (s1+αi)k≠s2+αk·i applies,where αi, αj, αk·i, αk·j are elements of a Galois field GF(2m), m≥3 and k is an integer not equal to 0, and + denotes addition in the Galois field.

11. The method of claim 10, further comprising: determining a plurality of correction signals, in each case one for a different bit, and wherein the plurality of correction signals are at least partially determined with overlapping timing.

12. The method of claim 10, wherein different correction signals are determined by means of different components.

13. The method of claim 10, wherein the first comparison value and the second comparison value are at least partially determined simultaneously.

14. The method of claim 1, further comprising: determining a third partial error syndrome.

15. The method of claim 14, wherein the third partial error syndrome is a syndrome of a 3-bit error-correcting —-Chaudhuri—Hocquenghem (BCH) code.

16. The method of claim 14, wherein the specified value in correction of a 2-bit error is constant.

17. The method of claim 15, wherein, in correction of a 3-bit error, the specified value is determined based on the respective bit position to be corrected.

18. The method of claim 14, wherein, when the corrected bit value corrects an erroneous first bit of a 3-bit error, the specified value equals a value of a product of: a second bit position, a third bit position, and a sum of the second bit position and the third bit position; where the second bit position represents a position of a second bit error and the third bit position represents a position of a third bit error, and where the product and the sum of the bit positions are determined in a Galois field GF(2m).

19. The method of claim 18, wherein the specified value is determined based on a symmetric function of the erroneous bit positions.

20. The method of claim 19, wherein the symmetric function is determined using partial error syndromes of a Bose—Chaudhuri—Hocquenghem (BCH) code.

21. The method of claim 14, further comprising: determining a parity for the bit sequence.

22. The method of claim 21, wherein the specified value is constant if the parity of the bit sequence equals 0 and the first partial error syndrome si is not equal to 0.

23. The method of claim 21, wherein the specified value is constant if the parity of the bit sequence equals 1 and s13=s2, where s1 is the first partial error syndrome and s2 is the second partial error syndrome.

24. A device for processing a bit sequence, configured to: determine a first partial error syndrome s1 of an error syndrome and a second partial error syndrome s2 of the error syndrome for the bit sequence,determine a first comparison value based on a bit position and the first partial error syndrome,determine a second comparison value based on the bit position and the second partial error syndrome,correct a bit value at the bit position when a comparison between the first comparison value and the second comparison value assumes a specified value,not correct the bit value at the bit position when the comparison between the first comparison value and the second comparison value do not assume the specified value.

25. The device of claim 24, wherein the device comprises a plurality of processing units and/or circuits configured to correct or not correct bit values at corresponding bit positions at least in part in simultaneous fashion.

26. The device of claim 25, wherein the device is part of a memory or a memory system or is separate from the memory or the memory system.