#
# SPDX-FileCopyrightText: Copyright (c) 2021-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
"""
Layer for applying OFDM channel: single-tap channel response in the frequency
domain
"""
import tensorflow as tf
from sionna.utils import expand_to_rank
from .awgn import AWGN
[docs]
class ApplyOFDMChannel(tf.keras.layers.Layer):
# pylint: disable=line-too-long
r"""ApplyOFDMChannel(add_awgn=True, dtype=tf.complex64, **kwargs)
Apply single-tap channel frequency responses to channel inputs.
This class inherits from the Keras `Layer` class and can be used as layer
in a Keras model.
For each OFDM symbol :math:`s` and subcarrier :math:`n`, the single-tap channel
is applied as follows:
.. math::
y_{s,n} = \widehat{h}_{s, n} x_{s,n} + w_{s,n}
where :math:`y_{s,n}` is the channel output computed by this layer,
:math:`\widehat{h}_{s, n}` the frequency channel response (``h_freq``),
:math:`x_{s,n}` the channel input ``x``, and :math:`w_{s,n}` the additive noise.
For multiple-input multiple-output (MIMO) links, the channel output is computed for each antenna
of each receiver and by summing over all the antennas of all transmitters.
Parameters
----------
add_awgn : bool
If set to `False`, no white Gaussian noise is added.
Defaults to `True`.
dtype : tf.DType
Complex datatype to use for internal processing and output. Defaults to
`tf.complex64`.
Input
-----
(x, h_freq, no) or (x, h_freq):
Tuple:
x : [batch size, num_tx, num_tx_ant, num_ofdm_symbols, fft_size], tf.complex
Channel inputs
h_freq : [batch size, num_rx, num_rx_ant, num_tx, num_tx_ant, num_ofdm_symbols, fft_size], tf.complex
Channel frequency responses
no : Scalar or Tensor, tf.float
Scalar or tensor whose shape can be broadcast to the shape of the
channel outputs:
[batch size, num_rx, num_rx_ant, num_ofdm_symbols, fft_size].
Only required if ``add_awgn`` is set to `True`.
The noise power ``no`` is per complex dimension. If ``no`` is a
scalar, noise of the same variance will be added to the outputs.
If ``no`` is a tensor, it must have a shape that can be broadcast to
the shape of the channel outputs. This allows, e.g., adding noise of
different variance to each example in a batch. If ``no`` has a lower
rank than the channel outputs, then ``no`` will be broadcast to the
shape of the channel outputs by adding dummy dimensions after the
last axis.
Output
-------
y : [batch size, num_rx, num_rx_ant, num_ofdm_symbols, fft_size], tf.complex
Channel outputs
"""
def __init__(self, add_awgn=True, dtype=tf.complex64, **kwargs):
super().__init__(trainable=False, dtype=dtype, **kwargs)
self._add_awgn = add_awgn
def build(self, input_shape): #pylint: disable=unused-argument
if self._add_awgn:
self._awgn = AWGN(dtype=self.dtype)
def call(self, inputs):
if self._add_awgn:
x, h_freq, no = inputs
else:
x, h_freq = inputs
# Apply the channel response
x = expand_to_rank(x, h_freq.shape.rank, axis=1)
y = tf.reduce_sum(tf.reduce_sum(h_freq*x, axis=4), axis=3)
# Add AWGN if requested
if self._add_awgn:
y = self._awgn((y, no))
return y