Low density parity check decoder

• US 9,112,530 B2
• Filed: 12/27/2013
• Issued: 08/18/2015
• Est. Priority Date: 05/01/2007
• Status: Active Grant

• Current Assignee: Texas A&M University System
• Original Assignee: Texas A&M University System
• Inventors: Gunnam, Kiran Kumar, Choi, Gwan S.
• Primary Examiner(s): Bonzo, Bryce
• Assistant Examiner(s): Mehrmanesh, Elmira
• Application Number:
  US14/141,508
• Publication Number:
  US 20140181612A1
• Time in Patent Office: 599 Days
• Field of Search: 714/758, 714/801
• US Class Current: 1/1
• CPC Class Codes: H03M 13/1105 Decoding
• CPC Class Codes: H03M 13/1128 Judging correct decoding and iterative stopping criteria other than syndrome check and upper limit for decoding iterations
• CPC Class Codes: H03M 13/116 Quasi-cyclic LDPC [QC-LDPC] codes, i.e. the parity-check matrix being composed of permutation or circulant sub-matrices
• CPC Class Codes: H03M 13/1177 Regular LDPC codes with parity-check matrices wherein all rows and columns have the same row weight and column weight, respectively
• CPC Class Codes: H03M 13/13 Linear codes
• CPC Class Codes: H03M 13/616 Matrix operations, especially for generator matrices or check matrices, e.g. column or row permutations

Abstract

A method and system for decoding low density parity check (“LDPC”) codes. An LDPC decoder includes an R select unit, a Q message first-in first-out (“FIFO”) memory, and a cyclic shifter. The R select unit provides an R message by selecting from a plurality of possible R message values. The Q message memory stores a Q message until an R message is generated by a CNU, the Q message and the R message are combined to provide a P message. The cyclic shifter shifts the P message.

Claims

1. A low density parity check (LDPC) code decoder, comprising:

decoding circuitry configured to;

perform block parallel processing that initiates processing all non-zero block columns of a plurality (M) of rows of a layer of an LDPC matrix in each clock cycle, and wherein M≦

p, and p is a total number of elements in a layer of the LDPC matrix; and

update a P message responsive to determination of a final state for each row of the LDPC matrix;

wherein the LDPC matrix comprises a plurality of layers, each of the layers comprising a plurality (M) of rows that are processed per clock cycle;

wherein each of the plurality of rows of each of the layers is datawise independent of the rows processed during a previous NP_MAX clock cycles and at least one row in each layer is datawise dependent on a row of an immediately preceding layer;

wherein NP_MAX is greater than one.

2. The LDPC code decoder of claim 1, wherein the decoding circuitry is further configured to generate a Q message by combining an R message with a P message.

3. The LDPC code decoder of claim 1, wherein the decoding circuitry comprises a permuter configured to permute the P message.

4. The LDPC code decoder of claim 3, wherein the permuter is configured to permute the P message by a difference of permutation of a block currently being processed and permutation of a block previously processed;

wherein the block currently being processed and the block previously processed are in a same block column of the LDPC matrix.

5. The LDPC code decoder of claim 4, wherein the permuter is configured to perform M×

M permutation, and the decoding circuitry further comprises memory address generation logic that in conjunction with the permuter is configured to permute P messages for M rows of the LDPC matrix, where p is the circulant size, and M<

=p.

6. The LDPC code decoder of claim 1, wherein the decoding circuitry is further configured to subtract an R old message from a permuted P message to generate a Q message.

7. The LDPC code decoder of claim 1, wherein the decoding circuitry is further configured to select one of an updated P message and a channel log-likelihood ratio (LLR) for storage in a P memory, wherein a channel LLR is selected to initialize decoding.

8. The LDPC code decoder of claim 1, wherein the decoding circuitry is further configured to select a R message from a plurality of previously generated possible R messages based on at least a message index value and a sign bit.

9. The LDPC code decoder of claim 1, wherein the LDPC matrix is quasi-cyclic.

10. The LDPC code decoder of claim 1, wherein each of the plurality (M) of rows for which processing is initiated in a given clock cycle comprises non-zero entries only in different columns from non-zero entries of rows for which processing is initiated during the previous NP_MAX clock cycles.

11. A low density parity check (LDPC) code decoder, comprising:

an array of reconfigurable minimum finder units adapted to;

perform block parallel processing of an LDPC matrix;

selectably process in parallel a first plurality (dc1) of block columns of a plurality (M1) of rows of a layer of a first LDPC matrix, where;

dc1 is check node degree of a block row of the first LDPC matrix;

p1 is a total number of elements of the block row of a first LDPC matrix; and

M1<

=p1; and

selectably process in parallel a second plurality (dc2) of block columns of a plurality (M2) of rows of a layer of a second LDPC matrix, where;

dc2 is check node degree of a block row of the second LDPC matrix, and dc1 is different from dc2;

p2 is a total number of elements of the block row of the second LDPC matrix;

M2<

=p2; and

M2 is different from M1.

12. The LDPC code decoder of claim 11, wherein the array of reconfigurable minimum finder units is adapted to:

partition a vector of input values into a plurality of sub-vectors;

determine a first minimum value and a second minimum value for each sub-vector; and

determine a first minimum and a second minimum of the vector using the first minimum value and second minimum value of the sub-vectors.

13. A low density parity check (LDPC) code decoder, comprising:

a Q message generator configured to combine an R message with a P message to produce a Q message;

logic configured to reduce the magnitude a Q message provided to a check node unit of the decoder; and

a permuter configured to permute the P message by a difference of permutation of a block currently being processed and permutation of a block previously processed;

wherein the block currently being processed and the block previously processed are in a same block column of an LDPC matrix.

14. The LDPC code decoder of claim 13, wherein the logic is configured to reduce the magnitude the Q message of a circulant based on a check node degree of a layer to which the circulant belongs.

15. The LDPC code decoder of claim 13, wherein the logic is configured reduce the magnitude the Q message of a circulant based on a variable node degree of a block column to which the circulant belongs.

16. The LDPC code decoder of claim 15, further comprising a Q message memory that stores a Q message until an R message is generated by the R select unit.

17. The LDPC code decoder of claim 15, further comprising a multiplexer that selects one of a channel log-likelihood ratio (LLR) value and a P message to input to the permuter, and wherein the multiplexer selects a channel LLR value to initialize the decoder.

18. The LDPC code decoder of claim 13, wherein the logic is configured reduce the magnitude the Q message of a circulant based on a variable node degree of a block column to which the circulant belongs and a check node degree of a layer to which the circulant belongs.

19. The LDPC code decoder of claim 13, further comprising an R select unit that provides an R message by selecting from a plurality of possible R message values.

20. The LDPC code decoder of claim 13, further comprising a P sum adder that produces a P message by adding a Q message to an R message.

21. The LDPC code decoder of claim 13, wherein the permuter is configured to perform M×

M permutation, and the decoder further comprises memory address generation logic that in conjunction with the permuter is configured to permute P messages for M rows of an LDPC matrix, where p is the circulant size, and M<

=p.

22. The LDPC code decoder of claim 13, further comprising of a hard decision memory that buffers hard decisions of decoded codewords that are processed out-of-order and provides the decoded codewords in-order.

23. The LDPC code decoder of claim 13, further comprising a Q memory configured to buffer channel LLR values from an inter-symbol interference detector and provide internal storage for Q messages the LDPC decoder.

24. The LDPC code decoder of claim 13, further comprising iteration logic configured to determine a maximum number of LDPC decoder iterations for each received codeword based on a number of failing LDPC check node constraints in the received codeword.

25. A low density parity check (LDPC) code decoder, comprising:

a control unit that controls decoder processing, the control unit configured to cause the decoder to process blocks of an LDPC matrix in a sequence defined by an order of non-zero blocks of a given layer of the LDPC matrix;

wherein the LDPC matrix comprises a plurality of layers, each layer having a plurality of blocks ordered such that the sequence of non-zero blocks of the given layer of the LDPC matrix specifies a first set of non-zero blocks of the given layer to be processed at a given time and a second set of non-zero blocks of the given layer to be processed after the first set of non-zero blocks;

wherein the first set specifies only non-zero blocks of the given layer that are not dependent on a result of a previously processed layer and the second set specifies non-zero blocks of the given layer that are dependent on a result of the previously processed layer.

26. The LDPC code decoder of claim 25, wherein the control unit is configured to cause the decoder to process each block of the matrix in processing substeps comprising:

an R new update substep that produces an R new message for a block of a different layer of the matrix from a layer containing a block currently being processed;

an R old update substep that selects an R old message for a layer of the matrix currently being processed;

a P message substep that generates updated P messages; and

a Q message substep that computes variable node messages (Q messages).

27. The LDPC code decoder of claim 26, wherein the control unit is configured to cause the decoder to generate an updated P message based on a determination of a final state for a block row.

28. The LDPC code decoder of claim 25, further comprising:

a P message generator that combines an R message with a Q message to produce a P message; and

a Q message generator that combines an R message with a P message to produce a Q message.

29. The LDPC code decoder of claim 25, further comprising a Q message memory that stores a Q message that is to be combined with an R message to produce a P message.

30. The LDPC code decoder of claim 25, further comprising a P message memory that stores a P message that is to be combined with an R message to produce a Q message.

31. The LDPC code decoder of claim 25, further comprising a permuter that permutes a P message, wherein the permuter permutes the P message by the difference of the permutation of a block currently being processed and the permutation of a block previously processed;

wherein the block currently being processed and the block previously processed are in a same block column.

32. A method for generating a low density parity check (LDPC) matrix, comprising:

generating a first plurality of rows of the LDPC matrix for processing in an LDPC decoder during a first NP_MAX clock cycles, wherein NP_MAX is greater than one; and

generating a second plurality of rows of the LDPC matrix for processing in the LDPC decoder immediately after the first set of rows, each of the rows of the second plurality of rows is datawise independent of each of the rows of the first plurality of rows;

wherein the LDPC matrix comprises a plurality of layers and at least one row in each layer is datawise dependent on a row of a previous layer.

33. The method of claim 32, further comprising computing a shift difference for each circulant in a first layer of the LDPC matrix to a circulant in a previous layer of the LDPC matrix, and the wherein the circulant in the first layer of the LDPC matrix and the circulant in the previous layer of the LDPC matrix are in a same block column.

34. The method of claim 32, further comprising computing a number of stall cycles due to pipeline dependencies in the LDPC decoder while processing the LDPC matrix as a function of the shift difference, parallelization M (where M is a number of rows in each layer of the LDPC matrix), and pipeline depth of the LDPC decoder.

35. The method of claim 32, further comprising computing a number of stall cycles as a function of LDPC matrix structure, parallelization M (where M is a number of rows in each layer of the LDPC matrix), and pipeline depth of the LDPC decoder.