The function implements the BP algorithm based on the decoding algorithm presented in [2]. For transmitted LDPC-encoded codeword $$c=\left(c_0,c_1,\dots ,c_{n-1}\right)$$, the input to the LDPC decoder is the LLR given by

$$L\left(c_i\right)=\log \left(\frac{\mathrm{Pr}\left(c_i=0|\text{channel output for }{c}_i\right)}{\mathrm{Pr}\left(c_i=0|\text{channel output for }{c}_i\right)}\right)$$.

In each iteration, the function updates the key components of the algorithm based on these equations:

$$L\left(r_{ji}\right)=2\text{atanh}\left({\displaystyle \prod _{i^{\prime }\in V_j\backslash \left\{i\right\}}\tanh \left(\frac{1}{2}L\left(q_{i^{\prime }j}\right)\right)}\right)$$,

$$L\left(q_{ij}\right)=L\left(c_i\right)+{\displaystyle \sum _{j\text{'}\in C_i\backslash \left\{j\right\}}L\left(r_{j^{\prime }i}\right)}$$, initialized as $$L\left(q_{ij}\right)=L\left(c_i\right)$$ before the first iteration, and

$$L\left(Q_i\right)=L\left(c_i\right)+{\displaystyle \sum _{j^{\prime }\in C_i}L\left(r_{j^{\prime }i}\right)}$$.

At the end of each iteration, $$L\left(Q_i\right)$$ is an updated estimate of the LLR value for the transmitted bit, $$c_i$$. The value $$L\left(Q_i\right)$$ is the soft-decision output for $$c_i$$. If $$L\left(Q_i\right)$$ is negative, the hard-decision output for $$c_i$$ is 1. Otherwise, the output is 0.

Index sets $$C_i\backslash \left\{j\right\}$$ and $$V_j\backslash \left\{i\right\}$$ are based on the PCM such that the sets $$C_i$$ and $$V_j$$ correspond to all nonzero elements in column i and row j of the PCM, respectively.

This figure demonstrates how to compute these index sets for PCM $$H$$ for the case i = 5 and j = 3.

To avoid infinite numbers in the algorithm equations, atanh(1) and atanh(–1) are set to 19.07 and –19.07, respectively. Due to finite precision, MATLAB^® returns 1 for tanh(19.07) and –1 for tanh(–19.07).

When you specify the 'EarlyTermination' name-value pair argument as 0 (false), the decoding terminates after the number of iterations specified by the 'MaximumLDPCIterationCount' name-value pair argument. When you specify the 'EarlyTermination' name-value pair argument as 1 (true), the decoding terminates when all parity checks are satisfied ($$H{c}^T=0$$) or after the number of iterations specified by the 'MaximumLDPCIterationCount' name-value pair argument.

The function implements the layered BP algorithm based on the decoding algorithm presented in Section II.A of [3]. The decoding loop iterates over subsets of rows (layers) of the PCM.

For each row, m, in a layer and each bit index, j, the implementation updates the key components of the algorithm based on these equations.

(1) $$L\left(q_{mj}\right)=L\left(q_j\right)-R_{mj}$$

(2) $$\Psi (x)=\log \left(\left|\tanh \left(x/2\right)\right|\right)$$

(3) $$A_{mj}={\displaystyle \sum _{n\in N(m)\backslash \left\{j\right\}}\Psi }\left(L\left(q_{mn}\right)\right)$$

(4) $$s_{mj}={\displaystyle \prod _{n\in N(m)\backslash \left\{j\right\}}\mathrm{sgn}}\left(L\left(q_{mn}\right)\right)$$

(5) $$R_{mj}=-s_{mj}\Psi \left(A_{mj}\right)$$

(6) $$L\left(q_j\right)=L\left(q_{mj}\right)+R_{mj}$$

For each layer, the decoding equation (6) works on the combined input obtained from the current LLR inputs, $$L\left(q_{mj}\right)$$, and the previous layer updates, $$R_{mj}$$.

Because the layered BP algorithm updates only a subset of the nodes in a layer, this algorithm is faster than the BP algorithm. To achieve the same error rate as attained with BP decoding, use half the number of decoding iterations when using the layered BP algorithm.

The function implements the normalized min-sum decoding algorithm by following the layered BP algorithm with equation (3) replaced by

$$A_{mj}={\min }_{n\in N(m)\backslash \left\{j\right\}}\left(\alpha \left|L\left(q_{mn}\right)\right|\right)$$,

where α is the scaling factor specified by the 'MinSumScalingFactor' name-value pair argument. This equation is an adaptation of equation (4) presented in [4].

The function implements the offset min-sum decoding algorithm by following the layered BP algorithm with equation (3) replaced by

$$A_{mj}=\max \left({\min }_{n\in N(m)\backslash \left\{j\right\}}\left(\left|L\left(q_{mn}\right)\right|-\beta \right), 0\right)$$,

where β is the offset specified by the 'MinSumOffset' name-value pair argument. This equation is an adaptation of equation (5) presented in [4].