Method and apparatus for a low-density parity-check decoder

#This invention relates generally to low-density parity-check (LDPCs) decoders, and more specifically to a method and apparatus for an LDPC decoder.

#LDPC codes are linear block codes. The codeword space and the encoding procedure of LDPC codes are specified by a generator matrix G, given by: x=uG

#where G is a K×N matrix with full-row rank, u is a 1×K vector representing information bits and x is a 1×N vector for the codeword. Usually, the generator matrix can be written as follows: G=└IK×K PK×(N−K)┘

#FIG. 1 is a block diagram of a prior art Tanner graph for a low-density parity-check (LDPC) decoder.

#FIG. 2 is a block diagram of a prior art LDPC encoder with a code rate of K/N.

#FIG. 3 depicts a flowchart of a method operating in a prior art LDPC decoder.

#The conventional dual-max method of equation (1) in the aforementioned prior art approximates an LLR bit by calculating all possible likelihoods and selecting the largest one. However, if additional information is available about which constellation points should be used to determine an LLR bit, an approximation is not necessary. FIG. 4 depicts a constellation integrating teachings of the present disclosure. In conventional dual-max calculations, to calculate the LLR of a first bit of a received soft symbol 102 depicted as a circle with rough edges in FIG. 4, a distance from point 102 to all gray points 104 is calculated to determine a point having a minimum distance to point 102, which in this example is point 11-11. A distance of point 102 is then calculated to all uncolored points 106 to determine a point having a minimum distance thereto, which in this illustration is point −11-11.

#If additional information about bits 2, 3 and 4 are available, say b2=b3=b4=−1, then only two constellation points (1-1-1-1) colored in gray in FIG 5 as point 108, and (−1-1-1-1) uncolored point 110 should be used for LLR calculation for the first bit b1. That is, the LLR of bit b1 is the difference between the distances of point 102 to point 108 (i.e., 1-1-1-1) and point 102 to point 110 (i.e., −1-1-1-1). These calculations are the true LLR of bit b1 without approximation.

#Unfortunately, the additional information about bits 2, 3 and 4 are generally not available before the information bits are decoded in a conventional decoder. However, in an LDPC decoder intermediate results can be used to update the decoder input such that the input to the decoder is approaching a true LLR for each bit. As described earlier, an LDPC decoder can calculate an LLR or a soft decision for each bit iteratively. The sign of the soft decision determines the value of an associated bit (1 or −1), while the magnitude of a soft decision indicates the confidence of the decoded bit. The larger the soft decision magnitude, the higher the confidence for the decoded bit.