A method for constructing a low-density parity-check (ldpc) code using a structured base parity check matrix with permutation matrix, pseudo-permutation matrix, or zero matrix as constituent sub-matrices; and expanding the structured base parity check matrix into an expanded parity check matrix. A method for constructing a ldpc code using a structured base parity check matrix H=[Hd|Hp], Hd is the data portion, and Hp is the parity portion of the parity check matrix; the parity portion of the structured base parity check matrix is such so that when expanded, an inverse of the parity portion of the expanded parity check matrix is sparse; and expanding the structured base parity check matrix into an expanded parity check matrix. A method for encoding variable sized data by using the expanded ldpc code; and applying shortening, puncturing.

Patent
   RE46692
Priority
Oct 12 2004
Filed
Oct 29 2014
Issued
Jan 30 2018
Expiry
Oct 12 2025

TERM.DISCL.
Assg.orig
Entity
Large
2
25
all paid
2. A method of decoding low-density parity-check (ldpc) encoded data, comprising:
wirelessly receiving, by a receiver, encoded data from a data source; and
applying the, by an ldpc decoder, a following expanded parity check matrix to the encoded data to generate decoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96).
1. A method of low-density parity-check (ldpc) encoding data, comprising:
receiving input data from a data source; and
applying the, by an ldpc encoder, a following expanded parity check matrix to the input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92
71 −1 55 −1 12 66 45 79 −1 78 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43
−1 −1 −1 −1 32 52 55 80 95 22 6 51
−1 63 31 88 20 −1 −1 −1 6 40 56 16
37 38 4 11 −1 46 48 0 −1 −1 −1 −1
−1 45 24 32 30 −1 −1 0 0 −1 −1 −1
10 −1 22 55 70 82 −1 −1 0 0 −1 −1
−1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96); and
wirelessly transmitting the encoded data.
8. Apparatus for decoding low-density parity-check (ldpc) encoded data, comprising:
an input port a receiver operable to receive wirelessly transmitted encoded data from a data source; and
means for applying the a following expanded parity check matrix to the encoded data to generate decoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96).
5. Apparatus for low-density parity-check (ldpc) encoding data, comprising:
an input port operable to receive input data from a data source; and
means for applying the a following expanded parity check matrix to the input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96); and
a transmitter operable to wirelessly transmit the encoded data.
6. Apparatus for decoding low-density parity-check (ldpc) encoded data, comprising:
an input port a receiver operable to receive wirelessly transmitted encoded data from a data source; and
circuitry an ldpc decoder coupled to the input port and operable to apply the a following expanded parity check matrix to the encoded data to generate decoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96).
7. Apparatus for decoding low-density parity-check (ldpc) encoded data, comprising:
an input port a receiver operable to receive wirelessly transmitted encoded data from a data source; and
a matrix application element of an ldpc decoder operable to apply the a following expanded parity check matrix to the encoded data to generate decoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96).
4. Apparatus for low-density parity-check (ldpc) encoding data, comprising:
an input port operable to receive input data from a data source; and
a matrix application element of an ldpc encoder operable to apply the a following expanded parity check matrix to the input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96); and
a transmitter operable to wirelessly transmit the encoded data.
3. Apparatus for low-density parity-check (ldpc) encoding data, comprising:
an input port operable to receive input data from a data source; and
circuitry an ldpc encoder coupled to the input port and operable to apply the a following expanded parity check matrix to the input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96); and
a transmitter operable to wirelessly transmit the encoded data.
10. A method of operating a telecommunications network, comprising:
applying the, by an ldpc encoder, a following expanded parity check matrix to input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96);
transmitting the encoded data over a transmission medium;
receiving the transmitted encoded data; and
applying said expanded parity check matrix to the encoded data to recover the input data.
12. A method of operating a transceiver, comprising:
applying the, by an ldpc encoder, a following expanded parity check matrix to input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96);
transmitting the encoded data over a transmission medium;
receiving encoded data from the transmission medium; and
applying said expanded parity check matrix to the received encoded data to generate decoded data.
9. A telecommunications network, comprising:
an ldpc encoder operable to apply the a following expanded parity check matrix to input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96);
a transmitter operable to transmit the encoded data over a transmission medium;
a receiver operable to receive the transmitted encoded data; and
an ldpc decoder operable to apply said expanded parity check matrix to the encoded data to recover the input data.
11. A transceiver, comprising:
an ldpc encoder operable to apply the a following expanded parity check matrix to input data to generate encoded data:
6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0
wherein an expansion factor, L, is between 24 and 96, −1 represents an L×L all-zero square matrix, and
any other integer, Sij, represents an L×L identity matrix circularly right shifted by a shift amount equal to floor ((L×Sij)/96);
a transmitter operable to transmit the encoded data over a transmission medium;
a receiver operable to receive encoded data from the transmission medium; and
an ldpc decoder operable to apply said expanded parity check matrix to the received encoded data to generate decoded data.

This application

More preferably, the threshold for a qnormalized is set to be in the range of 1.2-1.5.

More preferably, the threshold for qnormalized is set to be equal to 1.2.

In accordance with another aspect of the present invention there is provided a method of shortening low-density parity-check (LDPC) code comprising the steps of: a) selecting variable nodes in a parity check matrix; b) ensuring a uniform or a close to uniform row weight distribution after removing the selected variable nodes; and c) ensuring a new column weight distribution as close as possible to an original column weight distribution after removing the columns corresponded to the selected variable nodes.

Preferably, the method further comprises the step of selecting variable nodes that belongs to consecutive columns in the parity check matrix.

Preferably, the method further comprises the step of prearranging columns of the data portion of parity check matrix.

In accordance with another aspect of the present invention there is provided a method of puncturing a low-density parity-check (LDPC) code comprising the steps of: a) selecting variable nodes in a parity check matrix; b) ensuring that each of the selected variable nodes is connected to fewest possible check nodes; and c) ensuring that all of the selected variable nodes are connected to most possible check nodes.

Preferably, the method further comprises the step of selecting variable nodes belonging to consecutive columns in the parity check matrix.

The invention and the illustrated embodiments may be better understood, and the numerous objects, advantages, and features of the present invention and illustrated embodiments will become apparent to those skilled in the art by reference to the accompanying drawings, and wherein:

FIG. 1 shows a typical system in which embodiments of the present invention may be practiced;

FIG. 2 is an example of a structured LDPC code parity check matrix;

FIG. 3 depicts an example of a parity check matrix with dual diagonal;

FIG. 4 illustrates an example of unstructured data portion in a base parity check matrix;

FIG. 5 is an example of the expanded unstructured base parity check matrix of FIG. 4;

FIG. 6 is an example of a parity check matrix expansion;

FIG. 7 is another example showing a base parity check matrix and an expanded parity check matrix;

FIG. 8 depicts a general form of the base parity check matrix of the present invention;

FIG. 9 gives examples of parity portion Hp of the general base parity check matrix allowing efficient encoding;

FIG. 10 is an example of a fully structured base parity check matrix, showing sub-matrices arranged as blocks;

FIG. 11 is an expanded parity check matrix from sub-matrices of the fully structured base parity check matrix of FIG. 10;

FIG. 12 illustrates the general form of the parity check matrix of the present invention;

FIG. 13 shows a parity check matrix with outlined layers for the layered belief propagation decoding;

FIG. 14 gives the high level hardware architecture implementing existing method of layered belief propagation decoding;

FIG. 15 is an example of the high level hardware architecture implementing layered belief propagation decoding in accordance with one embodiment of the present invention;

FIG. 16 shows a sub-matrix construction by concatenation of permutation matrices;

FIG. 17 is an example of parallel decoding with concatenated permutation matrixes;

FIG. 18 is an example of modifications supporting parallel decoding when sub-matrices are built from concatenated smaller permutation matrices;

FIG. 19 illustrates short and long information blocks processing;

FIG. 20 illustrates encoding of data packets, using puncturing and shortening;

FIG. 21 illustrates a data encoding procedure in accordance with another embodiment of the present invention;

FIG. 22 illustrates rearranging of the columns in Hd in order to enable efficient shortening;

FIG. 23 shows a bipartite graph of an LDPC code with emphasis on a punctured bit;

FIG. 24 illustrates puncturing impact on the performance;

FIG. 25 is an example of a parity check matrix suited for both puncturing and shortening operation and used to obtain the results illustrated in FIG. 24; and

FIGS. 26a, 26b and 26c are matrices for use in relevant encoding methods and systems.

Reference will now be made in detail to some specific embodiments of the invention including the best modes contemplated by the inventors for carrying out the invention. Examples of these specific embodiments are illustrated in the accompanying drawings. While the invention is described in conjunction with these specific embodiments, it will be understood that it is not intended to limit the invention to the described embodiments. On the contrary, it is intended to cover alternatives, modifications, and equivalents as may be included within the spirit and scope of the invention as defined by the appended claims. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. The present invention may be practiced without some or all of these specific details. In other instances, well-known process operations have not been described in detail in order not to unnecessarily obscure the present invention.

Efficient decoder architectures are enabled by designing the parity check matrix, which in turn defines the LDPC code, around some structural assumptions: structured LDPC codes.

One example of this design is that the parity check matrix comprises sub-matrices in the form of binary permutation or pseudo-permutation matrices.

The term “permutation matrix” is intended to mean square matrices with the property that each row and each column has one element equal to 1 and other elements equal to 0. Identity matrix, a square matrix with ones on the main diagonal and zeros elsewhere, is a specific example of permutation matrix. The term “pseudo-permutation matrix” is intended to include matrices that are not necessarily square matrices, and matrices may have row(s) and/or column(s) consisting of all zeros. It has been shown, that using this design, significant savings in wiring, memory, and power consumption are possible while still preserving the main portion of the coding gain. This design enables various serial, parallel, and semi-parallel hardware architectures and therefore various trade-off mechanisms.

This structured code also allows the application of layered decoding, also referred to as layered belief propagation decoding, which exhibits improved convergence properties compared to a conventional sum-product algorithm (SPA) and its derivations. Each iteration of the layered decoding consists of a number of sub-iterations that equals the number of blocks of rows (or layers). FIG. 2 shows a matrix having three such layers 21, 22, 23.

LDPC code parity check matrix design also results in the reduction in encoder complexity. Classical encoding of LDPC codes is more complex than encoding of other advanced codes used in FEC, such as turbo codes. In order to ease this complexity it has become common to design systematic LDPC codes with the parity portion of the parity check matrix containing a lower triangular matrix. This allows simple recursive decoding. One simple example of a lower triangular matrix is a dual diagonal matrix as shown in FIG. 3.

Referring to FIG. 3, the parity check matrix 30 is partitioned as H=[Hd|Hp]. Data portion Hd 31 is an M×K matrix that corresponds to the data bits of the codeword. The design of the Hd 31 matrix ensures high coding gain. Parity portion Hp 32 is in this example an M×M dual diagonal matrix and corresponds to the parity bits of the codeword. These codes are systematic block codes. The codeword vector for these systematic codes has the structure:

c = [ d p ]
where d=[d0 . . . dk−1]T is the block of data bits and p=[p0 . . . pM−1]T are the parity bits. A codeword is any binary, or in general, non-binary, N-vector c that satisfies:
Hc=Hdd+Hpp=0

Thus, a given data block d is encoded by solving binary equation Hdd=Hpp for the parity bits p. In principle, this involves inverting the M×M matrix Hp to resolve p:
p=Hp−1Hdd   [equation 1]

Hp is assumed to be invertible. If the inverse of Hp, Hp−1 is also low density then the direct encoding specified by the above formula can be done efficiently. However, with the dual diagonal structure of Hp 32 encoding can be performed as a simple recursive algorithm:

p 0 = k 0 n = 1 h 0 , i n 0 d i n 0 ,
where in0 is the index of the column in which row 0 contains a “1”

p 1 = p 0 + k 1 n = 1 h 1 , i n 1 d i n 1 ,
where in1 is the index of the column in which row 1 contains a “1”

p M - 1 = p M - 2 + k M - 1 n = 1 h M - 1 , i n M - 1 d i n M - 1 ,
where inM−1 is the index of the column in which row M−1 contains a “1”.

In these recursive expressions hr,c are non-zero elements (1 in this exemplary matrix) of the data portion of the parity check matrix, Hd 31. The number of non-zero elements in rows 0, 1, . . . , M−1, is represented by k0, k1, . . . , kM−1, respectively.

One desirable feature of LDPC codes is that they support various required code rates and block sizes. A common approach is to have a small base parity check matrix defined for each required code rate and to support various block sizes by expanding the base parity check matrix. Since it is usually required to support a range of block sizes, a common approach is to define expansion for the largest block size and then apply other algorithms which specify expansion for smaller block sizes. Below is an example of a base parity check matrix:

11 0 10 6 3 5 1 0 −1 −1 −1 −1
10 9 2 2 3 0 −1 0 0 −1 −1 −1
7 9 11 10 4 7 −1 −1 0 0 −1 −1
9 2 4 6 5 3 0 −1 −1 0 0 −1
3 11 2 3 2 11 −1 −1 −1 −1 0 0
2 7 1 0 10 7 1 −1 −1 −1 −1 0

In this example the base parity check matrix is designed for the code rate R=½ and its dimensions are (Mb×Nb)=(6×12). Assume that the codeword sizes (lengths) to be supported are in the range N=[72, 144], with increments of 12, i.e. N=[72, 84, . . . , 132, 144]. In order to accommodate those block lengths the parity check matrix needs to be of the appropriate size (i.e. the number of columns match N, the block length). The number of rows is defined by the code rate: M=(1−R) N. The expansion is defined by the base parity check matrix elements and the expansion factor L, which results in the maximum block size. The conventions used in this example, for interpreting the numbers in the base parity check matrix, are as follows:

The following example shows a rotated identity matrix where the integer specifying rotation is 5:

0 0 0 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 0 0 0 0 1
1 0 0 0 0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0

Therefore, for the largest block (codeword) size of N=144, the base parity check matrix needs to be expanded by an expansion factor of 12. That way the final expanded parity check matrix to be used for encoding and generating the codeword of size 144, is of the size (72×144). In other words, the base parity check matrix was expanded Lmax=12 times (from 6×12 to 72×144). For the block sizes smaller than the maximum, the base parity check matrix is expanded by a factor L<Lmax. In this case expansion is performed in the similar fashion except that now matrices IL and 0L, are used instead of ILmax and 0Lmax, respectively. Integers specifying the amount of rotation of the appropriate identity matrix, IL, are derived from those corresponding to the maximum expansion by applying some algorithm. For example, such an algorithm may be a simple modulo operation:
rL=(rLmax)modulo L

An example of such a matrix is shown in FIG. 4 where the data portion Hd 61 and the parity portion Hp 62 of a matrix 60. The corresponding expanded parity check matrix is shown in FIG. 5 also having a data portion Hd 71 and the parity portion Hp 72 of the matrix 70. Each of the shaded squares 73 indicates a L×L small permutation matrix that is placed on the position of the 1's in the base parity check matrix, where L is the expansion factor. In other words, if the size of the base parity check matrix was Mb×Nb, the size of expanded parity check matrix is now M×N=LMb×LNb.

The expansion may be done for example by replacing each non-zero element with a permutation matrix of the size of the expansion factor. One example of performing expansion is as follows.

Hp is expanded by replacing each “0” element by an L×L zero matrix, 0L×L, and each “1” element by an L×L identity matrix, IL×L, where L represent the expansion factor.

Hd is expanded by replacing each “0” element by an L×L zero matrix, 0L×L, and each “1” element by a circularly shifted version of an L×L identity matrix, IL×L. The shift order, s (number of circular shifts, for example, to the right) is determined for each non-zero element of the base parity check matrix.

It should be apparent to a person skilled in the art that these expansions can be implemented without the need to significantly change the base hardware wiring.

FIG. 6 shows an example of a base parity check matrix 41 and a corresponding expanded parity check matrix 42 using 3×3 sub-matrices of which that labeled 43 is an example.

The simple recursive algorithm described earlier can still be applied in a slightly modified form to the expanded parity check matrix. If hi,j represent elements of the Hd portion of the expanded parity check matrix, then parity bits can be determined as follows:
p0=h0,0d0+h0,1d1+h0,2d2+ . . . +h0,11d11
p1=h1,0d0+h1,1d1+h1,2d2+ . . . +h1,11d11
p2=h2,0d0+h2,1d1+h2,2d2+ . . . +h2,11d11
p3=p0+h3,0d0+h3,1d1+h3,2d2+ . . . +h3,11d11
p4=p1+h4,0d0+h4,1d1+h4,2d2+ . . . +h4,11d11
p5=p2+h5,0d0+h5,1d1+h5,2d2+ . . . +h5,11d11
p6=p3+h6,0d0+h6,1d1+h6,2d2+ . . . +h6,11d11
p7=p4+h7,0d0+h7,1d1+h7,2d2+ . . . +h7,11d11
p8=p5+h8,0d0+h8,1d1+h8,2d2+ . . . +h8,11d11
p9=p6+h9,0d0+h9,1d1+h9,2d2+ . . . +h9,11d11
p10=p7+h10,0d0+h10,1d1+h10,2d2+ . . . +h10,11d11
p11=p8+h11,0d0+h11,1d1+h11,2d2+ . . . +h11,11d11

However, when the expansion factor becomes large, then the number of columns with only one non-zero element, i.e. 1 in the example here, in the Hp becomes large as well. This may have a negative effect on the performance of the code.

One remedy for this situation is to use a slightly modified dual diagonal Hp matrix. This is illustrated with reference to FIG. 7 where the modified base parity check matrix 51 produces the expanded parity check matrix 52.

The parity check equations now become:
h0,0d0h0,1d1+ . . . +h0,11d11+p0+p3=0   [Equation 2]
h1,0d0h1,1d1+ . . . +h1,11d11+p1+p4=0   [Equation 3]
h2,0d0h2,1d1+ . . . +h2,11d11+p2+p5=0   [Equation 4]
h3,0d0h3,1d1+ . . . +h3,1d11+p0+p3+p6=0   [Equation 5]
h4,0d0h4,1d1+ . . . +h4,11d11+p1+p4+p7=0   [Equation 6]
h5,0d0h5,1d1+ . . . +h5,11d11+p2+p5+p8=0   [Equation 7]
h6,0d0h6,1d1+ . . . +h6,11d11+p6+p9=0   [Equation 8]
h7,0d0h7,1d1+ . . . +h7,11d11+p7+p10=0   [Equation 9]
h8,0d0h8,1d1+ . . . +h8,11d1+p8+p11=0   [Equation 10]
h9,0d0h9,1d1+ . . . +h9,11d11+p0+p9=0   [Equation 11]
h10,0d0h10,1d1+ . . . +h10,11d11+p1+p10=0   [Equation 12]
h11,0d0h11,1d1+ . . . +h11,11d11+p2+p11=0   [Equation 13]

Now by summing up equations 2, 5, 8, and 11, the following expression is obtained:
(h0,0+h3,0+h6,0+h9,0)d0+(h0,1+h3,1+h6,1+h9,0)d1 . . . +(h0,11+h3,11+h6,11+h9,11)d11p0+p3+p0+p3+p6+p6+p9+p0+p9=0

Since only p0 appears an odd number of times in the equation above, all other parity check bits cancel except for p0, and thus:
p0=(h0,0+h3,0+h6,0+h9,0)d0+(h0,1+h3,1+h6,1+h9,1)d1+. . . +(h0,11+h3,11+h6,11+h9,11)d11

Likewise:
p1=(h1,0+h4,0+h7,0+h10,0)d0+(h1,1+h4,1+h7,1+h10,1)d0+(h1,11+h4,11+h7,11+h10,11)d11
p2=(h2,0+h5,0+h8,0+h11,0)d0+(h2,1+h5,1+h87,1+h11,1)d1+. . . +(h2,11+h5,11+h8,11+h11,11)d11

After determining p0, p1, p2 the other parity check bits are obtained recursively:
p3=h0,0d0+h0,1d1+. . . +h0,11d11+p0
p4=h1,0d0+h1,1d1+. . . +h1,11d11+p1
p5=h2,0d0+h2,1d1+. . . +h2,11d11+p2
p6=h3,0d0+h3,1d1+. . . +h3,11d11+p0+p3
p7=h4,0d0+h4,1d1+. . . +h4,11d11+p1+p4
p8=h5,0d0+h5,1d1+. . . +h5,11d11+p2+p5
p9=h6,0d0+h6,1d1+. . . +h6,11d11+p6
p10=h7,0d0+h7,1d1+. . . +h7,11d11+p7
p11=h8,0d0+h8,1d1+. . . +h8,11d11+p8   [Equation 14]

The present invention provides method and system enabling high throughput, low latency implementation of LDPC codes, and preserving the simple encoding feature at the same time.

In accordance with one embodiment of the present invention, a general form is shown in FIG. 8 where the matrix 80 has a data portion Hd 81 and a parity portion Hp 82. Each of the shaded blocks 83 represents a sub-matrix. A sub-matrix may be, for example, but not limited to, a permutation matrix, a pseudo-permutation matrix or a zero matrix. When efficient encoding is required, the parity portion Hp 82 of the matrix is designed such that its inverse is also a sparse matrix. Elements of the base parity check matrix and its sub-matrices may be binary or non-binary (belonging to elements of a finite Galois Field of q elements, GF(q)).

The data portion (Hd) may also be placed on the right side of the parity (Hp) portion of the parity check matrix. In the most general case, columns from Hd and Hp may be interchanged.

Parity check matrices constructed according to the embodiments of the present invention supports both regular and irregular types of the parity check matrix. Not only the whole matrix may be irregular (non-constant weight of its rows and columns) but also that its constituents Hd and Hp may be irregular, if such a partition is desired.

If the base parity check matrix is designed with some additional constraints, then base parity check matrices for different code rates may also be derived from one original base parity check matrix in one of two ways:

Row-combining or row-splitting, with the specific constraints defined above, allow efficient coding of a new set of expanded derived base parity check matrices. In these cases the number of layers may be as low as the minimum number of block rows (layers) in the original base parity check matrix.

FIG. 9 shows examples for parity portion Hp's of these base parity check matrices allowing more efficient encoding. In each example, zero sub-matrices 91 are shown lightly shaded with a 0, and permutation (or pseudo-permutation) sub-matrices 90 are shown cross-hatched. In FIG. 9, parity portions with sub-matrices 901, 902 are examples of the embodiments of the invention. Parity portions with sub-matrices 903, 904 represent particularly interesting cases of the generalized dual diagonal form. The first column 95 of the sub-matrix of the parity portion 903, and the last column 96 of the parity portion 904 contain an odd number of sub-matrices in order to ensure the existence of the inverse matrix of Hp. The other columns each contain two sub-matrices in pairs, forming a “staircase”, which ensures efficient recursive encoding. These constructions of Hp can be viewed as generalized cases, where Hp consists of a dual diagonal and an odd-weight column (such as the matrix 60 in FIG. 4), which may be symbolically expressed as:
Hp,present_invention(m)=T (Hp,existing,m),

Where T is the transform describing the base parity check matrix expansion process and m is the size of the permutation matrices. For m=1, Hp of the present invention defines the form of the prior art Hp (dual diagonal with the odd-weight column), i.e.

Hp,present_invention(l)=T (Hp,existing,l)=Hp,existing

A further pair of parity portions with sub-matrices 905, 906 illustrate cases where these first and last columns, respectively, have only one sub-matrix each.

The two parity portions with sub-matrices 907, 908 in FIG. 9 illustrate lower and upper triangular structure, also thus permit efficient recursive encoding. However, in order to solve the weight-1 problem, the sub-matrices 99 (shown hatched) in each example have the weight of all columns equal to 2, except the last one, which has weight equal to 1.

One of the characteristics of the base parity check matrix expansion of the present invention is that the expanded base parity check matrix inherits structural features from the base parity check matrix. In other words, the number of blocks (rows or columns) that can be processed in parallel (or serial, or in combination) in the expanded parity check matrix equals the number of blocks in the base parity check matrix.

Referring to FIG. 10, matrix 100 representing one possible realization of a rate R=½ base parity check matrix based on the structure of the matrix 80 of FIG. 8. In the matrix 100, only non-zero elements (1's in this example) are indicated and all blanks represent 0's. The parity check matrix comprises D rows that may be processed in parallel (D=4 for matrix 100). Each row can be further split into a number of columns, B (B=8 for matrix 100). Therefore, the whole base parity check matrix 100 can be envisioned as comprising D rows and B columns of sub-matrices. Examples of the sub-matrices may be, but not limited to, permutation sub-matrices, pseudo-permutation sub-matrices or zero sub-matrices. Furthermore, it is not necessary that the sub-matrices are square sub-matrices, although in this example all sub-matrices are m×m square sub-matrices.

The base parity check matrix 100 of FIG. 10 is shown as an expanded parity check matrix 110 in FIG. 11. In this example, each non-zero element is replaced by a L×L sub-matrix, for example a permutation sub-matrix, and each zero is replaced by an L×L zero sub-matrix of which the smaller square 111 is an example.

It can be seen that expanded parity check matrix 110 has inherited structural properties of its base parity check matrix 100 from FIG. 10. That means that in the expanded sub-matrix blocks (of which 112 is an example) can be considered as having the same sub-matrices as before the expansion, for example, permutation, or all zero sub-matrices. This property offers implementation advantages.

The sub-matrices of the present invention are not limited to permutation sub-matrices, pseudo-permutation sub-matrices or zero sub-matrices. In other words, the embodiments of the present invention are not restricted to the degree distribution (distribution of column weights) of the parity check matrix, allowing the matrix to be expanded to accommodate various information packet sizes and can be designed for various code rates. This generalization is illustrated through following examples.

FIG. 12 shows a general form of a parity check matrix 120. Cross-hatched blocks, of which 121 is an example, represent sub-matrices S, which may be, in the most general form, rectangular. In addition, these sub-matrices 121 may further comprise a set of smaller sub-matrices of different size. As discussed before, elements of the parity check matrix and its sub-matrices may be binary or non-binary (belonging to elements of a finite Galois Field of q elements, GF(q)).

FIG. 13 shows a parity check matrix 130 having rows corresponding to different layers l to D, of which 132, 133, 134 are examples.

In the context of parallel row processing, layered belief propagation decoding is next briefly described with reference to FIGS. 13 and 14.

A high level architectural block diagram is shown in FIG. 14 for the parallel row processing scenario comprising memory modules 141, connected to a read network 142 (using permuters). These permuters are in turn connected to a number of processing units 143 whose outputs are directed to a write network 144 (using inverse permuters). In this scenario, each iteration of the belief propagation decoding algorithm consists of processing D layers (groups of rows). This approach therefore updates the decoding variables corresponding to a particular layer depends on the equivalent variables corresponding to all other layers.

In order to support a more general approach in accordance with an embodiment of the present invention, the architecture of FIG. 14 may be modified. One example of such a modification is depicted in FIG. 15 where the extra inter-layer storage element 155 is shown. In this new architecture additional storage of inter-layer variables is also required—the function being provided by the element 155. This change enables an increased level of parallelism beyond the limits of existing approach.

By exercising careful design of the parity check matrix, the additional inter-layer storage 155 in FIG. 15 can be implemented with low complexity. One such approach is discussed below.

Iterative parallel decoding process is best described as read-modify-write operation. The read operation is performed by a set of permuters, which deliver information from memory modules to corresponding processing units. Parity check matrices, designed with the structured regularity described earlier, allow efficient hardware implementations (e.g., fixed routing, use of simple barrel shifters) for both read and write networks. Memory modules are organized so as to provide extrinsic information efficiently to processing units.

Processing units implement block (layered) decoding (updating iterative information for a block of rows) by using any known iterative algorithms (e.g. Sum Product, Min-Sum, Bahl-Cocke-Jelinek-Raviv (BCJR)).

Inverse permuters are part of the write network that performs the write operation back to memory modules.

Such parallel decoding is directly applicable when the parity check matrix is constructed based on permutation, pseudo-permutation or zero sub-matrices.

To encode using sub-matrices other than permutation, pseudo-permutation or zero sub-matrices, one embodiment of the present invention uses special sub-matrices. A sub-matrix can also be constructed by concatenation of smaller permutation or pseudo-permutation matrices. An example of this concatenation is illustrated in FIG. 16, in which the four small sub-matrices 161, 162, 163, 164 are concatenated into the sub-matrix 165.

Parallel decoding is applicable with the previously described modification to the methodology; that is, when the parity check matrix includes sub-matrices built by concatenation of smaller permutation matrices.

FIG. 17 illustrates such a base matrix 170. The decoding layer 171 includes permutation sub-matrices 172 S21, S22, S23, S26, S27, sub-matrix S24 (built by concatenation of smaller permutation matrices), and zero sub-matrices S25, and S28. The decoding layer 171 is shown 174 with the sub-matrix S24 split vertically into S124 176 and S224 177.

It can be seen that for the decoding layer 171 a first processing unit receives information in the first row 179 from bit 1 (according to S21), bit 6 (S22), bit 9 (S23), bit 13 (S124), bit 15 (S224), bit 21 (S28), and bit 24 (S29). Other processing units are loaded in a similar way.

For layered belief propagation type decoding algorithms, the processing unit inputs extrinsic information accumulated, by all other layers, excluding the layer currently being processed. Thus, the prior art implementation described using FIG. 14 presents the processing unit with all the results accumulated by other decoding layers. The only bits that require modification in order to satisfy this requirement are bits from S124 176 and S224 177, which are referred to as special bits. To provide complete extrinsic information about these special bits to a processing unit, an output must be added from other variable nodes within the current layer (inter-layer results) as described previously with respect to FIG. 15 where the interlayer storage element 155 was introduced.

This is illustrated in FIG. 18 in which additional memory modules 155 used for interlayer storage are shown, and which provide interlayer extrinsic information to permuters 1821, 1822, 1829. For the special bits the additional storage for inter-layer information comprises the delay lines 186. Processing units 184a-184d, each programmed to correspond with a row 179 of the current decoding layer 174, provide inputs to delay lines 186. A first further permuter 1851 is applied to choose a pair of processing units 184 that operate with same special bit. A second further permuter 1852 chooses a processing unit's “neighbor”—namely one that operates with same special bit at the current decoding layer. Adders 1831a-1831d combine intra-layer information with inter-layer results from the second further permuter 1852. Outputs from the first further permuter 1851 are combined using adders 1835a and 1835b whose outputs enter the inverse permuters 187 as well as all other “normal” (i.e. non-special bits) bits output from each processing unit 184. The outputs from the inverse permuters 187 are written back to the memory modules 155 (intra-layer storage). Processing continues for the complete code matrix 170, taking each layer 174 in turn.

For simplicity, FIG. 18 shows details of modifications for special bits coming from S124 176. The analogous modifications for S224177 are also included in embodiments of the invention.

Improvement in throughput, and reduction in latency in accordance to an embodiment of the present invention is further illustrated by the following example.

The LDPC codes can be decoded using several methods. In general, iterative decoding is applied. The most common is the sum-product algorithm (SPA) method. Each iteration in SPA comprises two steps:

It has been shown that better performance, in terms of the speed of convergence, can be achieved with layered decoding. In layered decoding only row variables are updated for a block of rows, one block row at a time. The fastest approach is to process all the rows within a block of rows simultaneously.

The following is a comparison of the achievable throughput (bit rate) of two LDPC codes: one based on the existing method for expanding matrix, as described in FIG. 5, and the other based on the matrix of the present invention as described in FIG. 17. Throughput in bits per second (bps) is defined as:
T=(K×F)/(C×I),
where K is number of info bits, F is clock frequency, C is number of cycles per iteration, and I is the number of iterations. Assuming that K, F, and I are fixed and, for example, equal: K=320 bits, F=100 MHz, and I=10, the only difference between the existing method and the present invention is derived from C, the factor which is basically a measure of the level of allowed parallelism. It can be seen, by comparing FIG. 5 and FIG. 17, that the number of rows is the same in both cases (Mb=16). Assuming that the expanded parity check matrices are also the same size, and the same payload of, for example, K=320 bits can also be handled, an expansion factor of 20 (L=320/16) will be required. The maximum number of rows that can be handled in parallel is L (=20) for the matrix of FIG. 5, whereas the number of rows for parallel operation in the case of FIG. 17 is 4×20=80. Thus the number of cycles per iteration, C, is given as follows:
Cexisting=16 and Cpresent_invention=4.

Using these numbers in the formula gives:
Tmax,existing=200 Mbps
Tmax,present_invention=800 Mbps

As expected, the maximum throughput is 4 times greater. All the desirable features of the code design in terms of efficient encoding are preserved. For example, without degradation in performance, the encoding algorithm as described earlier with respect to FIG. 2, and the corresponding efficient encoder architecture still apply.

Furthermore, when a scaleable solution is desired, the size of the expanded LDPC parity check matrix is designed to support the maximum block size. The existing solutions do not scale well with respect to the throughput for various block sizes. For example, using the existing method for layered decoding, processing of short and long blocks takes the same amount of time. This is caused by the fact that for shorter blocks, not all processing units are used, resulting proportionally lower achieved throughput.

The following example is based on the same example as before by comparing matrices as described earlier in FIG. 5 and FIG. 17. One embodiment of the present invention allows the splitting of a block of rows into smaller stripes, and still has a reasonably low number of cycles per layered decoding iteration. The existing architecture does not allow this splitting without increasing decoding time beyond a reasonable point.

FIG. 19 illustrates the difference between the prior art method and the method in accordance with one embodiment of the present invention. The blocks 191 and 192 represent short blocks as processed by one embodiment of the present invention and in the existing method, respectively. In the case using the embodiment of the present invention only 4 cycles are required per iteration, whereas prior art implementations require 16 cycles. This represents a considerable savings in processing. For comparison, blocks 193 and 194 represent long blocks as processed by the present invention and in the prior art respectively, where as expected the savings are not made.

The following table compares the computed results.

Number of
Codeword processing Throughput
size C units (Mbps)
Existing (FIG. 5) 320 16 20 200
1280 16 80 800
Embodiment of 320 4 80 800
present invention (FIG. 17) 1280 16 80 800

It can be seen from the table that the embodiment of the present invention provides constant throughput independent on the codeword size, whereas in the case of the existing method the throughput for the smaller blocks drops considerably. The reason is that while the embodiment of the present invention fully utilizes all available processing resources irrespective of block size, the existing method utilizes all processing units only in the case of the largest block, and a fraction of the total resources for other cases.

The example here illustrating the throughput improvement for shorter blocks, leads also to the conclusion that reduced latency is also achieved with the embodiment of the present invention. When large blocks of data are broken into smaller pieces, the encoded data is split among multiple codewords. If one places a shorter codeword at the end of series of longer codewords, then the total latency depends primarily on the decoding time of the last codeword. According to the table above, short blocks require proportionally less time to be decoded (as compared to the longer codewords), thereby allowing reduced latency to be achieved by encoding the data in suitably short blocks.

In addition to the full hardware utilization illustrated above, embodiments of the present invention allow hardware scaling, so that short blocks can use proportionately less hardware resources if an application requires it.

Furthermore, utilization of more efficient processing units and memory blocks is enabled. Memory can be organized to process a number of variables in parallel. The memory can therefore, be partitioned in parallel.

The present invention provides new LPDC base parity matrices, and expanded matrices based on the new base parity matrices, and method for use thereof.

The locations of non-zero matrices for rate R in an exemplary matrix are chosen, so that:

An example of R=¾ base parity check matrix design using criteria a) to d) is:

1 1 1 1 0 0 0 1 1 0 1 0 1 1 1 1 0 1 1 1 0 0 0 0
1 1 1 1 0 1 1 0 1 0 0 1 0 1 1 1 1 0 0 1 1 0 0 0
1 0 1 0 1 1 1 1 0 1 0 0 1 0 1 1 1 1 0 0 1 1 0 0
1 1 0 1 1 1 1 1 0 1 1 1 0 0 0 0 1 1 1 0 0 1 1 0
0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 1 1
0 1 1 1 1 0 0 0 1 1 1 1 1 1 0 0 1 1 1 0 0 0 0 1

The rate R=¾ matrix definition built based on such base parity check matrix covers expansion factors in the range L between 24 and Lmax=96 in increments of 4. Right circular shifts of the corresponding L×L identity matrix s′ij, are determined as follows:

s ij = { floor ( L × s ij L m ax ) , s ij > 0 s ij , otherwise ,
where sij is specified in the matrix definition below:

6 38 3 93 −1 −1 −1 30 70 −1 86 −1 37 38 4 11 −1 46 48 0 −1 −1 −1 −1
62 94 19 84 −1 92 78 −1 15 −1 −1 92 −1 45 24 32 30 −1 −1 0 0 −1 −1 −1
71 −1 55 −1 12 66 45 79 −1 78 −1 −1 10 −1 22 55 70 82 −1 −1 0 0 −1 −1
38 61 −1 66 9 73 47 64 −1 39 61 43 −1 −1 −1 −1 95 32 0 −1 −1 0 0 −1
−1 −1 −1 −1 32 52 55 80 95 22 6 51 24 90 44 20 −1 −1 −1 −1 −1 −1 0 0
−1 63 31 88 20 −1 −1 −1 6 40 56 16 71 53 −1 −1 27 26 48 −1 −1 −1 −1 0

The present invention further enables flexible rate adjustments by the use of shortening, or puncturing, or a combination thereof. Block length flexibility is also enabled through expansion, shortening, or puncturing, or combinations thereof.

Any of these operations can be applied to the base or expanded parity check matrices.

Referring to FIG. 20, a data packet of length L is required to be encoded using an LDPC code (N, K), as previously presented, K 202 is the information block size, N is the length of the codeword, and M 203 is the number of parity check bits, M=N−K. The encoded data is to be transmitted using a number of modulated symbols, each carrying S bits.

The data packet 201 of length L is divided into segments 208. These segments are in turn encoded using an LDPC code (N, K). The information block K 202 may be optionally pruned to K′ 204; and the parity check bits M may be pruned to M′ 205. The term “pruning” is intended to mean applying code shortening by sending less information bits than possible with a given code, (K′<K). The term “puncturing” is intended to mean removing some of the parity bits and/or data bits prior to sending the encoded bits to the modulator block and subsequently over the channel. Pruned codewords may be concatenated 206 in order to accommodate the encoded data packet, and the resulting stream 207 is padded with bits 209 to match the boundaries 210 of modulated symbols before being sent to the modulator. The amount of shortening and puncturing may be different for the constituent pruned codewords. The objectives here are:

From objective (a) above it follows that in order to use a small number of codewords, an efficient shortening and puncturing operation needs to be applied. However, those operations have to be implemented in a way that would neither compromise the coding gain advantage of LDPC codes, nor lower the overall transmit efficiency unnecessarily. This is particularly important when using the special class of LDPC parity check matrices that enable simple encoding operation, for example, as the one describe in the previous embodiments of the present invention. These special matrices employ either a lower triangular, a dual-diagonal, or a modified dual-diagonal in the parity portion of the parity check matrix corresponding. An example of a dual-diagonal matrix is described earlier in FIG. 3 in which the parity portion Hp 32 corresponds to the parity bits, and the data portion Hd 31 to the information data bits.

Work to achieve efficient puncturing has been done using the “rate compatible” approach. One or more LDPC parity check matrix is designed for the low code rate application. By applying the appropriate puncturing of the parity portion, the same matrix can be used for a range of code rates which are higher than the original code rate as the data portion in relation to the codeword increases. These methods predominantly target applications where adaptive coding (e.g. hybrid automatic repeat request, H-ARQ) and/or unequal bit protection is desired.

Puncturing may also be combined with code extension to mitigate the problems associated with “puncturing only” cases. The main problem that researchers are trying to solve here is to preserve an optimum degree distribution through the process of modifying the original parity check matrix.

However, these methods do not directly address the problem described earlier: apply shortening and puncturing in such a way that the code rate is approximately the same as the original one, and the coding gain is preserved.

One method attempting to solve this problem specifies shortening and puncturing such that the code rate of the original code is preserved. The following notation is used:

Npunctured—Number of punctured bits,

Nshortened—Number of shortened bits.

Shortening to puncturing ratio, q, is defined as: q=Nshortened/Npunctured.

In order to preserve the same code rate, q has to satisfy the following equation:
qrate_preserved=R/(1−R)

Two approaches are prescribed for choosing which bits to shorten and which to puncture to reach a shortening and a puncturing pattern.

Two approaches for shortening and puncturing of the expanded matrices are described in Dale Hocevar and Anuj Batra, “Shortening and Puncturing Scheme to Simplify LDPC Decoder Implementation,” Jan. 11, 2005, a contribution to the informal IEEE 802.16e LDPC ad-hoc group, the entirely of the document is incorporated herein by reference. These matrices are generated from a set of base parity check matrices, one base parity check matrix per code rate. The choice depends on the code rate, i.e. on the particular parity check matrix design.

The method may preserve the column weight distribution, but may severely disturb the row weight distribution of the original matrix. This, in turn, causes degradation when common iterative decoding algorithms are used. This adverse effect strongly depends on the structure of the expanded matrix.

This suggests that this approach fails to prescribe general rules for performing shortening and puncturing, and has an unnecessary restriction for a general case such as the one described in FIG. 20. Namely, accepting some reduction in the code rate in order to keep the performance in terms of the coding gain at a certain level.

In general, the amount of puncturing needs to be limited. Extensive puncturing beyond certain limits paralyzes the soft decision decoder. Prior art methods, none of which specify a puncturing limit or alternatively offer some other way for mitigating the problem, may potentially compromise the performance significantly.

In accordance with another embodiment of the present invention, above described shortcomings may be addressed by:

This embodiment of the present invention may be beneficially applied to both the transmitter and the receiver. Although developed for wireless systems, embodiments of the invention can be applied to any other communication system which involves encoding of variable size data packets by a fixed error correcting block code.

The advantage of this invention can be summarized as providing an optimal solution to the above described problem given the range of the system parameters such as the performance, power consumption, and complexity. It comprises the following steps:

At step 213, the minimum number of modulated symbols Nsym_min is calculated. Next at step 214, the codeword size N is selected, and the number of codewords to be concatenated Ncwords is computed. At step 216 the required shortening and puncturing are computed, and performance estimated. If the performance criterion are met 217, the number of bits required to pad the last modulated symbol is computed 218 and the process ends 219. Where the performance criterion are not met 217, an extra modulated symbol is added 215 and the step 214 is reentered.

Both the encoder and the decoder may be presented with the same input parameters in order to be able to apply the same procedure and consequently use the same codeword size, as well as other relevant derived parameters, such as the amount of shortening and puncturing for each of the codewords, number of codewords, etc.

In some cases only the transmitter (encoder) has all the parameters available, and the receiver (decoder) is presented with some derived version of the encoding procedure parameters. For example, in some applications it is desirable to reduce the initial negotiation time between the transmitter and the receiver. In such cases the transmitter initially informs the receiver of the number of modulated symbols it is going to use for transmitting the encoded bits rather than the actual data packet size. The transmitter performs the encoding procedure differently taking into consideration the receiver's abilities (e.g. using some form of higher layer protocol for negotiation). Some of the requirements are relaxed in order to counteract deficiencies of the information at the receiver side. For example, the use of additional modulated symbols to enhance performance may always be in place, may be bypassed altogether, or may be assumed for the certain ranges of payload sizes, e.g. indirectly specified by the number of modulated symbols.

One example of such an encoding procedure is an OFDM based transceiver, which may be used in IEEE 802.11n. In this case the reference to the number of bits per modulated symbol translates into the number of bits per OFDM symbol. In this example, the AggregationFlag parameter specified in 801.11n is used to differentiate between the case when both the encoder and the decoder are aware of actual data packet size (AggregationFlag=0) and the case when the packet size is indirectly specified by the number of required OFDM symbols (AggregationFlag=1).

An exemplary algorithm in accordance with one embodiment of the present invention is with following parameters are now described:

Algorithm Parameters

Algorithm Input

Algorithm Output:

Algorithm Procedure

if(AggregationFlag == 0) {
  NInfobits=8×HT_LENGTH;
 //in non-aggregation case HT_LENGTH is the number of payload octets
  NOFDM=ceil(NInfobits/ (NCBPS×R));
 // minimum number of OFDM symbols
}
else {
  NOFDM= HT_LENGTH;
 // in aggregation case HT_LENGTH is the number of OFDM symbols
  NInfoBits=NOFDM×NCBPS ×R;
 // number of info bits includes padding;MAC will use its own delineation
  //method to recover an aggregate payload
}
NCodeWords = ceil(NCBPS× NOFDM/NNmax);
  //number of codewords is based on maximum codeword length
NN = ceil(NCBPS× NOFDM/(NCodeWords×NNinc))× NNinc;
  // codeword length will be the larger of the closest one
  // to NCBPS× NOFDM/NCodeWords
KK=NN×R;
  // number of information bits in codeword chosen
MM=NN−KK;
  // number of parity bits in codeword chosen
NParityBits_requested=NCodeWords× MM;
  // total number of parity bits allocated in NOFDM symbols
NParityBits =min(NOFDM× NCBPS− NInfoBits,NParityBits_requested);
  //in non-aggregation case allow adding extra OFDM symbol(s) to limit
//puncturing
if(AggregationFlag==0) {
while(100×(NParityBits_requested−NParityBits)/
NParityBits_requested>Pmax) {
NOFDM= NOFDM+1;
// extra OFDM symbol(s) are used to carry parity
NParityBits =min(NParityBits + NCBPS,NParityBits_requested);
}
}
// Finding number of information bits to be sent per codeword(s),
//KKS, KKS_Last, and number of bits the codeword(s) will be punctured
NP,
 //and NP_Last. Making sure that last codeword may only be shortened
 // more then others, and punctured less then others.
KKS=ceil(NInfoBits/ NCodeWords);
KKS_Last =NInfoBits − KKS ×( NCodeWords −1);
MMP =min(MM, floor(NParityBits/CodeWords);
MMP_Last = min(MM, NParityBits − MMP ×(NCodeWords −1));
NP =MM − MMP;
NP_Last =MM− MMP_Last;
// Finally, calculating number of padding bits in last OFDM symbol
NPaddingBits = NOFDM × NCBPS − NInfoBits − NParityBits;

Each of those features will be now described in more detail.

(a) General Rules for Shortening and Puncturing

Much effort has been spent to come up with designs of LDPC parity check matrices such that the derived codes provide optimum performance. Examples include: T. J. Richardson et al., “Design of Capacity-Approaching Irregular Length Low-Density Parity-Check Codes,” IEEE Transactions on Information Theory, vol. 47, February 2001 and S. Y. Chung, et al., “Analysis of Sum-Product Decoding of Low-Density Parity-Check Codes Using a Gaussian Approximation,” IEEE Transactions on Information Theory, vol. 47, February 2001, both of which are incorporated herein by reference, are examples. These papers show that, in order to provide optimum performance, a particular variable nodes degree distribution should be applied. Degree distribution refers here to the distribution of the column weights in a parity check matrix. This distribution, in general, depends on the code rate and the size of the parity check matrix, or codeword. It is desirable that the puncturing and shortening pattern, as well as the number of punctured/shortened bits, are specified in such a way that the variable nodes degree distribution is preserved as much as possible. However, since shortening and puncturing are qualitatively different operations, different rules apply to them, as will now be explained.

(b) Rules for Shortening

Shortening of a code is defined as sending less information bits than possible with a given code, K′<K. The encoding is performed by: taking K′ bits from the information source, presetting the rest (K-K′) of the information bit positions in the codeword to a predefined value, usually 0, computing M parity bits by using the full M×N parity check matrix, and finally forming the codeword to be transmitted by concatenating K′ information bits and M parity bits. One way to determine which bits to shorten in the data portion of the parity check matrix, Hd (31 in FIG. 3), is to define a pattern which labels bits to be shortened, given the parity check matrix, H=[Hd|Hp]. This is equivalent to removing corresponding columns from Hd. The pattern is designed such that the degree distribution of the parity check matrix after shortening, i.e. removing appropriate columns from Hd, is as close as possible to the optimal one for the new code rate and the codeword length. To illustrate this, consider a matrix having the following sequence of weights (each number corresponds to a column weight):

3 3 3 8 3 3 3 8 3 3 3 8

When discarding columns, the aim is to ensure that the ration of ‘3’s to ‘8’s remains close to optimal, say 1:3 in this case. Obviously it cannot be 1:3 when one to three columns are removed. In such circumstances, the removal of 2 columns might result in e.g.:

3 3 8 3 3 8 3 3 3 8

giving a ratio of ˜1:3.3 and the removal of a third column—one with weight ‘8’—might result in:

3 3 3 3 8 3 3 3 8

thus preserving a ratio of 1:3.5, which is closer to 1:3 than would be the case where the removal of the third column with weight ‘3’, which results in:

8 3 3 3 8 3 3 3 8

giving a ratio of 1:2.

It is also important to preserve approximately constant row weight throughout the shortening process.

An alternative to the above-described approach is to prearrange columns of the part of the parity check matrix, such that the shortening can be applied to consecutive columns in Hd. Although perhaps suboptimal, this method keeps the degree distribution of Hd close to the optimum. However, the simplicity of the shortening pattern, namely taking out the consecutive columns of Hd, gives a significant advantage by reducing complexity. Furthermore, assuming the original matrix satisfies this condition, approximately constant row weight is guaranteed. An example of this concept is illustrated in FIG. 22 where the original code rate R=½ matrix 220 is shown. In FIG. 22 (and FIG. 25) the white squares represent a z x z zero matrix, whereas the gray squares represent a z×z identity matrix shifted circularly to the right a number of times specified by the number written in the middle of the corresponding gray square. In this particular case, the maximum expansion factor is: zmax=72.

After rearranging the columns of the Hd part of the original matrix, the new matrix takes on the form 221 shown in FIG. 22. It can be seen that if the shortening is performed as indicated (to the left from the Hd|Hp boundary) the density of the new Hd will slightly increase until it reaches a “heavy” weight column (such as the block column 222). At that point the density of the new Hd will again approach the optimum one. A person skilled in the art will note that the rearranging of the columns in Hd does not alter the properties of the code.

In the case of a regular column parity check matrix, or more generally, approximately regular, or regular and approximately regular only in the data part of the matrix, Hd, the method described in the previous paragraph is still preferred compared to the existing random or periodic/random approach. The method described here ensures approximately constant row weight, which is another advantage from the performance and the implementation complexity standpoint.

(c) Puncturing

Puncturing of a code is defined as removing parity bits from the codeword. In a wider sense, puncturing may be defined as removing some of the bits, either parity bits or data bits or both, from the codeword prior to sending the encoded bits to the modulator block and subsequently over the channel. The operation of puncturing, increases the effective code rate. Puncturing is equivalent to a total erasure of the bits by the channel. The soft iterative decoder assumes a completely neutral value corresponding to those erased bits. In case that the soft information used by the decoder is the log-likelihood ratio, this neutral value is zero.

Puncturing of LDPC codes can be given an additional, somewhat different, interpretation. An LDPC code can be presented in the form of the bipartite graph of FIG. 23, in which the codeword bits are presented by the variable nodes 231, and parity check equations by the check nodes 232.

Each variable node 231 is connected 234 by edges, for example 233, to all the check nodes 232 in which that particular bit participates. Similarly, each check node (corresponding to a parity check equation) is connected by a set of edges 237 to all variable nodes corresponding to bits participating in that particular parity check equation. If a bit is punctured, for example node 235, then all the check nodes connected to it, those connected by thicker lines 236, are negatively affected. Therefore, if a bit chosen for puncturing participates in many parity check equations, the performance degradation may be very high. On the other hand, since the only way that the missing information (corresponding to the punctured bits) can be recovered is from the messages coming from check nodes those punctured bits participate in, the more of those the more successful recovery may be. Faced with contradictory requirements, the optimum solution can be found somewhere in the middle. These general rules can be stated as following:

Some of these trade-offs can be observed from FIG. 24 showing the frame error probability 240 for various situations.

FIG. 25 illustrates the base parity check matrix 250 used for obtaining the results in FIG. 24. The codeword size is 1728, which is obtained by expanding the base parity check matrix by the factor of z=72.

In FIG. 24, the curves are shown for six examples,

It can be seen from the FIG. 24 that puncturing bits corresponding to heavy-weight, or strong columns has a catastrophic effect on performance (241). On the other hand, puncturing block columns that do not participate in very many parity check equations does not provide very good performance (244) either. The best results are obtained when both criteria are taken into account represented by curves 242, 243, 245, 246. Among all of those, it appears that for the particular matrix structure (irregular Hd part with the modified dual diagonal in the Hp part) the best results were obtained when the punctured bits were selected from those corresponding to the weak columns of the data part of the parity check matrix, Hd, (242). If the parity check matrix is arranged as in 221 of FIG. 22, then the puncturing bits can be selected by starting from the leftmost bit of Hd and continuing with consecutive bits towards the parity portion of the matrix.

The matrix in FIG. 25 has undergone column rearrangement such that all the light-weight data columns have been put in the puncturing zone, i.e. leftmost part of the Hd part of the parity check matrix.

As discussed previously, in the case where the preservation of the exact code rate is not mandatory, the shortening-to-puncturing ratio can be chosen such that it guarantees preservation of the performance level of the original code.

Normalizing the shortening-to-puncturing ratio, q, as follows:
qnormalized=(Nshortened/Npunctured)/[R/(1−R)],

means that q becomes independent of the code rate, R. Therefore, qnormalized=1, corresponds to the rate preserving case of combined shortening and puncturing. However, if the goal is to preserve performance, this normalized ratio must be greater than one: qnormalized>1. It was found through much experimentation that one: qnormalized in the range of 1.2-1.5 complies with the performance preserving requirements.

In the case of a column regular parity check matrix, or more generally, approximately regular, or regular and approximately regular only in the data part of the matrix, Hd the method described above in accordance with one embodiment of the present invention is still preferred compared to the existing random or periodic/random approach since the present invention ensures approximately constant row weight, which provides another advantage from both the performance and the implementation complexity standpoints.

A large percentage of punctured bits paralyzes the iterative soft decision decoder. In the case of LDPC codes this is true even if puncturing is combined with some other operation such as shortening or extending the code. One could conclude this by studying the matrix 250 of FIG. 25. Here, it can be seen that as puncturing progresses it is more and more likely that a heavy weight column will be hit. This is undesirable and has a negative effect on the code performance. Defining the puncturing percentage as:
Ppuncture=100×(Npuncture/M),

then it can be seen that the matrix 250 from FIG. 25 cannot tolerate puncturing in excess of Ppuncture_max=33.3%. Therefore, this parameter Ppuncture_max must be set and taken into account when performing the combined shortening and puncturing operation.

Some of the embodiments of the present invention may include the following characteristics:

The system, apparatus, and method as described above are preferably combined with one or more matrices shown in the FIGS. 26a, 26b and 26c that have been selected as being particularly suited to the methodology. They may be used alone, or with other embodiments of the present invention.

The matrices in FIGS. 26a, 26b and 26c have been derived and tested, and have proven to be at least as efficient as prior art matrices in correcting errors.

A first group of matrices (FIG. 26a #1-#5) cover expansion factors up to 72 using rates R=½, ⅔, ¾, ⅚, and ⅞, respectively. The matrices may be utilized as they are specified or with the columns in the data portion of any of the matrices (first R*24 columns on the left side) reordered in anyway. The parity portion ((1−R)*24 rightmost columns) of the matrices is designed to allow simple encoding algorithms. They may be used in standards, such as wireless standards IEEE 802.11, and IEEE 802.16.

A further matrix (FIG. 26b #6) covers expansion factors up to 96 for rate R=¾. The matrix may be utilized as it is or with the columns in the data portion (first R*24 columns on the left side) reordered in any way. The parity portion ((1−R)*24 rightmost columns) of the matrix is designed to allow simple encoding algorithms.

The rate R=¾ matrices (FIG. 26b #7-#9) cover expansion factors in the range between 24 and 96 in increments of 4.

The rate R=⅚ matrix (FIG. 26b #10) may be used to cover expansion factors in the range between 24 and 96 in increments of 4.

The two rate R=⅚ matrices (FIGS. 26c #11 and #12) cover expansion factors up to Lmax=96. The matrices may be utilized as they are or with the columns in the data portion (first R*24 columns on the left side) reordered in any way. The parity portion ((1−R)*24 rightmost columns) of the matrix is designed to allow simple encoding algorithms. These particular matrices can accommodate codeword sizes in the range 576 to 2304 in increments of 96. Consequently, the expansion factors, L are in the range 24 to 96 in increments of 4. Right circular shifts of the corresponding L×L identity matrix (as explained in the previous section), s′, are determined as follows:
s′=floor{s. (L/96)},

where s is the right circular shift corresponding to the maximum codeword size (for L=Lmax=96), and it is specified in the matrix definitions.

The invention can be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or in combinations thereof. Apparatus of the invention can be implemented in a computer program product tangibly embodied in a machine-readable storage device for execution by a programmable processor; and method actions can be performed by a programmable processor executing a program of instructions to perform functions of the invention by operating on input data and generating output. The invention can be implemented advantageously in one or more computer programs that are executable on a programmable system including at least one programmable processor coupled to receive data and instructions from, and to transmit data and instructions to, a data storage system, at least one input device, and at least one output device. Each computer program can be implemented in a high-level procedural or object oriented programming language, or in assembly or machine language if desired; and in any case, the language can be a compiled or interpreted language. Suitable processors include, by way of example, both general and special purpose microprocessors. Generally, a processor will receive instructions and data from a read-only memory and/or a random access memory. Generally, a computer will include one or more mass storage devices for storing data files. Storage devices suitable for tangibly embodying computer program instructions and data include all forms of non-volatile memory, including by way of example semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM disks. Any of the foregoing can be supplemented by, or incorporated in, ASICs (application-specific integrated circuits). Further, a computer data signal representing the software code which may be embedded in a carrier wave may be transmitted via a communication network. Such a computer readable memory and a computer data signal are also within the scope of the present invention, as well as the hardware, software and the combination thereof.

While particular embodiments of the present invention have been shown and described, changes and modifications may be made to such embodiments without departing from the true scope of the invention.

Livshitz, Michael, Purkovic, Aleksandar, Burns, Nina, Sukhobok, Sergey, Chaudhry, Muhammad

Patent Priority Assignee Title
10784895, Aug 08 2016 Nokia Technologies Oy Inter-block modifications to generate sub-matrix of rate compatible parity check matrix
RE49225, Oct 12 2004 Malikie Innovations Limited Structured low-density parity-check (LDPC) code
Patent Priority Assignee Title
7178080, Aug 15 2002 Texas Instruments Incorporated Hardware-efficient low density parity check code for digital communications
7203897, Aug 12 2004 Google Technology Holdings LLC Method and apparatus for encoding and decoding data
7263651, Jan 12 2004 Apple Inc Method and apparatus for varying lengths of low density party check codewords
7313752, Aug 26 2003 Samsung Electronics Co., Ltd.; SAMSUNG ELECTRONICS CO , LTD Apparatus and method for coding/decoding block low density parity check code in a mobile communication system
7581157, Jun 24 2004 LG Electronics Inc.; LG Electronics Inc Method and apparatus of encoding and decoding data using low density parity check code in a wireless communication system
7607063, May 30 2003 Sony Corporation Decoding method and device for decoding linear code
7747934, Oct 12 2004 Malikie Innovations Limited Method for selecting low density parity check (LDPC) code used for encoding of variable length data
7752521, Oct 12 2004 Malikie Innovations Limited Low density parity check (LDPC) code
7917829, Oct 12 2004 Malikie Innovations Limited Low density parity check (LDPC) code
7996746, Oct 12 2004 Malikie Innovations Limited Structured low-density parity-check (LDPC) code
8024641, Oct 12 2004 Malikie Innovations Limited Structured low-density parity-check (LDPC) code
8099646, Oct 12 2004 Malikie Innovations Limited Low density parity check (LDPC) code
8301975, Oct 12 2004 Malikie Innovations Limited Structured low-density parity-check (LDPC) code
20010048744,
20020009199,
20020037014,
20020045428,
20040034828,
20050050435,
20050283707,
20050289437,
20060015791,
20080022191,
20110307755,
WO2006039801,
//////////////////
Executed onAssignorAssigneeConveyanceFrameReelDoc
Nov 19 1996LIVSHITZ, MICHAELBAY NETWORKS, INC ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670670 pdf
Nov 19 1996PURKOVIC, ALEKSANDARBAY NETWORKS, INC ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670670 pdf
Apr 29 1999BAY NETWORKS, INC NORTEL NETWORKS NA INC CHANGE OF NAME SEE DOCUMENT FOR DETAILS 0437430813 pdf
Dec 18 2000NORTEL NETWORKS NA INC NORTEL NETWORKS INC MERGER SEE DOCUMENT FOR DETAILS 0434670717 pdf
Oct 16 2007CHAUDHRY, MOHAMMADNortel Networks LimitedASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670749 pdf
Oct 18 2007BURNS, NINANortel Networks LimitedASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670749 pdf
Apr 05 2008SUKHOBOK, SERGEYNortel Networks LimitedASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670749 pdf
Jul 29 2011Nortel Networks LimitedRockstar Bidco, LPASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670773 pdf
Feb 29 2012Rockstar Bidco, LP2256355 ONTARIO LIMITEDASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670810 pdf
Mar 02 20122256355 ONTARIO LIMITEDResearch In Motion LimitedASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0434670850 pdf
Mar 12 2012NORTEL NETWORKS INC Rockstar Bidco, LPASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0439300931 pdf
Jul 09 2013Research In Motion LimitedBlackBerry LimitedCHANGE OF NAME SEE DOCUMENT FOR DETAILS 0380870963 pdf
Oct 29 2014BlackBerry Limited(assignment on the face of the patent)
Mar 20 2023BlackBerry LimitedOT PATENT ESCROW, LLCASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS 0634710474 pdf
Mar 20 2023BlackBerry LimitedOT PATENT ESCROW, LLCCORRECTIVE ASSIGNMENT TO CORRECT THE COVER SHEET AT PAGE 50 TO REMOVE 12817157 PREVIOUSLY RECORDED ON REEL 063471 FRAME 0474 ASSIGNOR S HEREBY CONFIRMS THE ASSIGNMENT 0648060669 pdf
May 11 2023OT PATENT ESCROW, LLCMalikie Innovations LimitedNUNC PRO TUNC ASSIGNMENT SEE DOCUMENT FOR DETAILS 0640150001 pdf
May 11 2023OT PATENT ESCROW, LLCMalikie Innovations LimitedCORRECTIVE ASSIGNMENT TO CORRECT 12817157 APPLICATION NUMBER PREVIOUSLY RECORDED AT REEL: 064015 FRAME: 0001 ASSIGNOR S HEREBY CONFIRMS THE ASSIGNMENT 0648070001 pdf
May 11 2023BlackBerry LimitedMalikie Innovations LimitedNUNC PRO TUNC ASSIGNMENT SEE DOCUMENT FOR DETAILS 0640660001 pdf
Date Maintenance Fee Events
Apr 30 2020M1552: Payment of Maintenance Fee, 8th Year, Large Entity.
Apr 09 2024M1553: Payment of Maintenance Fee, 12th Year, Large Entity.


Date Maintenance Schedule
Jan 30 20214 years fee payment window open
Jul 30 20216 months grace period start (w surcharge)
Jan 30 2022patent expiry (for year 4)
Jan 30 20242 years to revive unintentionally abandoned end. (for year 4)
Jan 30 20258 years fee payment window open
Jul 30 20256 months grace period start (w surcharge)
Jan 30 2026patent expiry (for year 8)
Jan 30 20282 years to revive unintentionally abandoned end. (for year 8)
Jan 30 202912 years fee payment window open
Jul 30 20296 months grace period start (w surcharge)
Jan 30 2030patent expiry (for year 12)
Jan 30 20322 years to revive unintentionally abandoned end. (for year 12)