generate_reg_ldpc

sionna.phy.fec.utils.generate_reg_ldpc(v: int, c: int, n: int, allow_flex_len: bool = True, verbose: bool = True) → Tuple[numpy.ndarray, int, int, float]

Generates a random regular (v, c) LDPC code.

This function generates a random Low-Density Parity-Check (LDPC) parity-check matrix of length n where each variable node (VN) has degree v and each check node (CN) has degree c. Note that the generated LDPC code is not optimized to avoid short cycles, which may result in a non-negligible error floor. For encoding, the LinearEncoder block can be used, but the construction does not guarantee that the parity-check matrix (pcm) has full rank.

Parameters:

• v (int) – Desired degree of each variable node (VN).

• c (int) – Desired degree of each check node (CN).

• n (int) – Desired codeword length.

• allow_flex_len (bool) – If True, the resulting codeword length may be slightly increased to meet the degree requirements.

• verbose (bool) – If True, prints code parameters.

Outputs:

• pcm – The parity-check matrix.

• k – The number of information bits.

• n – The codeword length.

• coderate – The code rate.

Notes

This algorithm is designed only for regular node degrees. To achieve state-of-the-art bit-error-rate performance, optimizing irregular degree profiles is usually necessary (see [tenBrink]).

Examples

from sionna.phy.fec.coding import generate_reg_ldpc

pcm, k, n, rate = generate_reg_ldpc(3, 6, 100, verbose=False)
print(f"Generated LDPC code: n={n}, k={k}, rate={rate:.3f}")