LayerDemapper#

class sionna.phy.nr.LayerDemapper(layer_mapper: sionna.phy.nr.layer_mapping.LayerMapper, num_bits_per_symbol: int = 1, *, precision: str | None = None, device: str | None = None, **kwargs)[source]#

Bases: sionna.phy.block.Block

Demaps MIMO layers to coded transport block(s) by following Sec. 6.3.1.3 and Sec. 7.3.1.3 in [3GPPTS38211].

This block must be associated to a LayerMapper and performs the inverse operation.

It is assumed that num_bits_per_symbol consecutive LLRs belong to a single symbol position. This allows to apply the LayerDemapper after demapping symbols to LLR values.

If the layer mapper is configured for dual codeword transmission, a list of both transport block streams is returned.

Parameters:

  • layer_mapper (sionna.phy.nr.layer_mapping.LayerMapper) – Associated LayerMapper.

  • num_bits_per_symbol (int) – Modulation order. Defines how many consecutive LLRs are associated to the same symbol position.

  • precision (str | None) – Precision used for internal calculations and outputs. If None, precision is used.

  • device (str | None) – Device for computation (e.g., ‘cpu’, ‘cuda:0’). If None, device is used.

Inputs:

inputs – […, num_layers, n/num_layers], torch.float. MIMO layer data sequences.

Outputs:

llr – […, n] or [[…, n1], […, n2]], torch.float. Sequence of bits after layer demapping. If num_codewords = 2, a list of two transport blocks is returned.

Notes

As it is more convenient to apply the layer demapper after demapping symbols to LLRs, this block groups the input sequence into groups of num_bits_per_symbol LLRs before restoring the original symbol sequence. This behavior can be deactivated by setting num_bits_per_symbol = 1.

Examples

import torch
from sionna.phy.nr import LayerMapper, LayerDemapper

mapper = LayerMapper(num_layers=2)
demapper = LayerDemapper(mapper, num_bits_per_symbol=4)

symbols = torch.randn(10, 100, dtype=torch.complex64)
mapped = mapper(symbols)
# After channel + demapping to LLRs...
llrs = torch.randn(10, 2, 200)  # num_bits_per_symbol=4, so 50*4=200
demapped = demapper(llrs)
print(demapped.shape)
# torch.Size([10, 400])

Methods

build(input_shapes)[source]#

Test input shapes for consistency.