When trying the Part 4: Toward Learned Receivers example at the step Training the Neural Receiver I get an error about TensorShape to a Tensor: (None, None) while i did not alter the example in any way. why is this? The full error is bellow:

Traceback (most recent call last): File "/cm/local/apps/slurm/var/spool/job6555120/slurm_script", line 308, in loss = model(BATCH_SIZE, ebno_db) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 885, in call result = self._call(*args, **kwds) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 933, in _call self._initialize(args, kwds, add_initializers_to=initializers) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 760, in _initialize *args, **kwds)) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 3066, in _get_concrete_function_internal_garbage_collected graph_function, _ = self._maybe_define_function(args, kwargs) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 3463, in _maybe_define_function graph_function = self._create_graph_function(args, kwargs) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 3308, in _create_graph_function capture_by_value=self._capture_by_value), File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/framework/func_graph.py", line 1007, in func_graph_from_py_func func_outputs = python_func(*func_args, **func_kwargs) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 668, in wrapped_fn out = weak_wrapped_fn().wrapped(*args, **kwds) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 3990, in bound_method_wrapper return wrapped_fn(*args, **kwargs) File "/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/framework/func_graph.py", line 994, in wrapper raise e.ag_error_metadata.to_exception(e) ValueError: in user code:

/cm/local/apps/slurm/var/spool/job6555120/slurm_script:280 __call__  *
    llr = self.rg_demapper(llr) # Extract data-carrying resource elements. The other LLrs are discarded
/home/tue/20184223/.local/lib/python3.6/site-packages/sionna/ofdm/resource_grid.py:470 call  *
    y = flatten_dims(y, 2, 2)
/home/tue/20184223/.local/lib/python3.6/site-packages/sionna/utils/tensors.py:76 flatten_dims  *
    flat_dim = tf.reduce_prod(tensor.shape[axis:axis+num_dims])
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/util/dispatch.py:206 wrapper  **
    return target(*args, **kwargs)
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:2744 reduce_prod
    input_tensor, _ReductionDims(input_tensor, axis), keepdims,
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:2094 _ReductionDims
    return range(0, array_ops.rank(x))
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/util/dispatch.py:206 wrapper
    return target(*args, **kwargs)
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/ops/array_ops.py:839 rank
    return rank_internal(input, name, optimize=True)
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/ops/array_ops.py:859 rank_internal
    input = ops.convert_to_tensor(input)
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/profiler/trace.py:163 wrapped
    return func(*args, **kwargs)
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/framework/ops.py:1566 convert_to_tensor
    ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
/home/tue/20184223/.local/lib/python3.6/site-packages/tensorflow/python/framework/constant_op.py:363 _tensor_shape_tensor_conversion_function
    "Cannot convert a partially known TensorShape to a Tensor: %s" % s)

ValueError: Cannot convert a partially known TensorShape to a Tensor: (None, None)

Hi!

I wanted to compare different Bit Error Rates for different modulations (QPSK, 8PSK and 16QAM) for the uncoded and encoded case (LDPC). The system is is the same, just the modulation differs...

I'm not sure if the 8PSK plot is correct. Especially in the encoded case, as the BER doesnt seem to improve with the encoder. For the QAM and 16QAM the encoded plot looks better, as an improvement is visible, in contrast to the uncoded case. Where I think the mistake may be, is during the set up of the constellation diagram for the mapper and demapper:

8PSK:

NUM_BITS_PER_SYMBOL = 3
BLOCK_LENGTH = 3**8
n=12*3

real = 1 * np.cos(np.pi/4)
imag = 1 * np.sin(np.pi/4)
CONST_SHAPE = np.array([-1, 1, 1j, -1j, complex(real, imag), complex(real, -imag), complex(-real, imag), complex(-real, -imag)]) # set the shape of the constellation diagram
constellation = sn.mapping.Constellation("custom", NUM_BITS_PER_SYMBOL, CONST_SHAPE, normalize=True, center=True)
constellation.show()

16QAM:

NUM_BITS_PER_SYMBOL = 4
BLOCK_LENGTH = 2**8
n=2**8

constellation = sn.mapping.Constellation("qam", NUM_BITS_PER_SYMBOL, normalize=True, center=True)
constellation.show();

QPSK:

NUM_BITS_PER_SYMBOL = 2
BLOCK_LENGTH = 2**8
n=2**8

constellation = sn.mapping.Constellation("qam", NUM_BITS_PER_SYMBOL, normalize=True, center=True)
constellation.show();

Any help or ideas would be greatly appriciated, thanks!

I have a test case where I compare it with results from matlab:

A = 32
P = 24
K = A+P
N = 512 # calculated according to Section 5.3.1 of 3GPP TS 38.212
decIn = np.array([98.0994,   100.2936,  -100.1615,  -111.2452,  -101.7249,  -101.5950,    78.5901,   -77.7265,   -76.4686,   -81.3223,  -115.8069,    89.6047,    85.9367,   104.1048,   -98.2053,  -106.7203,    83.1207,    88.2861,    83.8204,  -109.2175,  -124.4919,    94.6728,    91.5959,   -89.5352,  -101.8143,  -123.6915,  -103.1579,  -108.0879,   100.9417,   -95.5231,   122.5312,    99.6845,   -78.9441,   115.9421,    65.5815,    90.9532,   -88.9573,   -85.7132,   -83.5216,   -78.7969,   -54.1358,    75.8672,    84.9239,   -89.5731,   -92.8101,   -76.0732,    72.9594,   -74.3997,    93.8353,   -80.0016,    83.2706,  -104.8245,    75.5381,   -89.3799,    86.4390,    84.6055,    43.8157,  -100.4702,   -79.0332,   -89.0215,   109.9606,   -99.6110,   111.9779,  -101.8772,   -68.2989,    93.4047,    95.1260,    79.8283,    80.2065,   -89.3344,    68.9841,   -77.7767,    95.8681,   -79.4709,    62.2583,   -63.4266,   -91.1614,    81.0331,   -83.1753,   -95.7477,    79.3811,   -78.2227,   109.3239,   -70.7376,   -90.1096,   -93.8385,   -83.5296,    79.8664,  -107.2777,    86.7808,   -83.8104,   -89.5405,    83.9844,    88.8928,  -114.4005,   -81.6010,    68.2572,   100.1189,   -94.9682,  -106.6306,    94.3960,   -75.7935,   -80.0656,   -80.4422,    80.0088,    78.5461,  -101.9442,   101.2641,    79.4489,   -82.3328,    85.9272,   -93.2934,    96.1777,    88.5216,    84.3977,   -82.6021,    84.0117,    95.0405,   -86.1691,   -74.1916,   108.7456,    86.4657,   -75.5564,   -99.4773,    74.5251,    98.5329,   -83.2432,    70.5688,   -68.4671,   -79.3403,   -62.6724,   -60.8476,    67.8379,    91.1800,    87.8499,  -101.2957,   -69.6481,   -74.5603,    70.1766,   -99.2101,   101.3056,    57.1986,    87.5917,    84.1423,   -66.9824,   -62.7680,    60.1339,   -78.9674,    65.3749,   -64.6940,    50.0190,   -89.6008,   -59.3167,   -63.1129,    61.5982,    48.8499,    65.2957,   -44.2573,   -67.7464,   -87.8192,    73.3080,   -55.5333,    61.1872,    83.1372,    72.6622,    56.7458,   -90.9942,   -99.5641,   -62.1492,    75.1305,   -67.2248,    82.1078,   -71.9184,   -96.0142,   -82.4769,    63.0060,   -78.2155,    88.4656,    78.1025,   -62.1181,   -66.5374,    74.3818,    84.9285,    65.4854,    60.3368,   -59.9013,    62.8409,    45.3090,    58.8132,   -90.1745,   -80.5019,    76.9129,    53.0226,   -60.5902,    68.6878,    71.6496,   -79.5789,    68.2251,    69.2261,   -57.2262,    67.4926,   -56.8821,   -78.9174,    54.9742,   -77.1147,    72.3281,    45.8284,    74.1518,   -96.2469,    74.1641,    71.7959,   -84.9770,    75.2968,    43.3046,   -75.7478,    79.3113,   -67.3357,    83.5540,    88.1043,    92.9397,    60.5756,    96.0108,    68.6091,    56.8510,   -81.4292,   -84.7288,   -61.2944,   -77.7179,   -74.2870,    73.8435,   -68.7351,   -50.0336,    55.0304,    84.1329,    68.3470,   -42.6786,    51.2943,   -66.9083,   -66.1926,    79.1121,   -59.2044,   -69.6405,    89.3530,   -68.6773,   -78.4004,   -57.7046,   -56.9396,   -56.8815,    84.8120,    85.3458,    72.9297,    90.6085,   110.8704,    81.9232,    78.7975,   -75.6312,   -76.4349,   -65.0490,    66.5703,    77.3265,    70.3819,    58.6750,   -61.8244,    94.9379,    68.1455,    80.8718,    71.4065,   -89.3790,   -42.6856,   -64.9363,    76.2369,    69.3416,   -64.3863,   -64.0232,   -58.3138,    79.5344,    67.9879,   -79.1206,   -79.1714,    60.2715,    72.0637,    87.6743,    87.1759,    81.4093,   -74.1212,   100.5560,   -82.5535,   -56.7645,    90.7906,  -100.7988,   -64.6227,   -98.9736,    89.3586,    78.5604,    99.6034,    55.0377,    83.2926,   -72.7276,   -98.3970,    71.7933,    61.7939,    67.9014,   -69.5731,    73.2906,   -70.8013,    63.1397,   -52.9220,    80.8111,   -76.1652,    76.7023,   -66.0006,   -52.8898,   -67.6308,    85.2930,    78.6230,    75.2349,   -75.5770,    61.5977,   -82.0592,    56.5244,    73.6083,   -91.9549,   -48.5210,   -68.8685,   -87.2434,    55.9644,   -90.2183,    76.1801,   -85.3346,    66.5699,   -93.3345,    46.6941,    74.9255,   -85.1393,    65.0381,    59.5640,   -65.9826,    84.1422,   -68.0985,    74.0803,    66.0246,    83.8283,    78.7515,   -80.7086,    52.4481,   -66.9362,    56.3773,    65.2047,   -60.8237,   -76.4423,    79.9544,    76.8582,   -76.3870,   -74.6423,    50.1182,    43.6267,   -69.2405,    65.5949,    78.9524,    64.1765,   -55.5729,   -64.7979,    61.4107,   -74.2137,   -74.3217,    62.0394,   -74.1537,    69.4664,   -96.5727,   -98.9427,   -77.9188,    81.6283,   -92.4196,   -81.5016,   116.7946,   111.1792,  -104.2775,  -102.8753,   105.5464,    73.3699,  -121.8863,   -83.9218,    99.9105,   -76.7338,    71.2199,    84.3709,   123.7199,    78.3144,  -110.0666,  -102.4246,   -94.3307,    82.9367,    96.7602,    97.7912,   -85.7501,    88.8595,   -71.4018,   -73.5545,   -58.8523,  -103.0843,    97.3005,    78.5062,  -109.6570,    85.2691,   -91.4737,    87.4685,   -87.5806,   106.0338,    76.4113,   112.0384,   -78.9545,   -67.7884,   -70.9985,    84.6570,    98.7297,    91.5841,   -83.0778,   108.2129,   -94.8490,   -87.3756,  -111.9341,   -86.7001,    95.5885,   115.4110,    72.8125,  -106.4446,    80.6854,   -90.9581,    87.6573,    68.6028,   103.3637,   -92.1900,    73.5025,   -53.7392,   -64.1919,   106.1772,    79.7842,   -69.1110,   -79.4048,   -98.8703,   -57.1228,    62.4514,   -93.6077,   -45.7700,   -68.4287,    71.8675,    66.7427,   -63.6768,   -74.7351,    68.3605,   -64.5109,   -86.4599,    92.2079,   -62.3135,   -47.5140,    74.8112,   -55.6607,    72.2295,    98.2756,   -54.4462,   104.4787,   -76.8929,   -84.0509,   -87.5470,    88.1545,   -38.5373,   -73.9172,    86.9245,   -65.0378,   -62.6790,    76.2379,   -54.8815,    70.4215,   -64.9437,   -64.9268,   -78.1091,   -52.9684,   -53.5112,   -67.4531,   -47.4951,    54.0182,    62.7692,    47.2771,    64.6934,    62.8590,   -63.9107,    52.5923,    69.3743,   -64.8862,   -76.7852,   -60.9050,    51.3035,   -61.0244,    75.3462,    66.0729,   -50.1459,    61.4202,    80.4490,   -65.3638,    65.9481,    66.2290,    68.6534,    48.4402,    76.8827,   -88.6572,  -102.7420,   -94.8729,  -128.7776,  -122.5039,  -111.4672,  -105.0354,   107.5132])
from sionna.fec.polar import PolarEncoder, Polar5GEncoder, PolarSCLDecoder, Polar5GDecoder, PolarSCDecoder
from sionna.fec.polar.utils import generate_5g_ranking

# decode
frozen_pos, info_pos = generate_5g_ranking(K, N)
decoder = PolarSCLDecoder(frozen_pos, N, list_size=8, crc_degree='CRC24C', cpu_only=True, use_fast_scl=False)
decoded = np.array(decoder(np.expand_dims(decIn, 0))[0], 'int')

def interleave(c):
# 38.212 Table 5.3.1.1-1
    p_IL_max_table = [0, 2, 4, 7, 9, 14, 19, 20, 24, 25, 26, 28, 31, 34, 42, 45, 49, 50, 51, 53, 54, 56, 58, 59, 61, 62, 65, 66, 67, 69, 70, 71, 72, 76, 77, 81, 82, 83, 87, 88, 89, 91, 93, 95, 98, 101, 104, 106, 108, 110, 111, 113, 115, 118, 119, 120, 122, 123, 126, 127, 129, 132, 134, 138, 139, 140, 1, 3, 5, 8, 10, 15, 21, 27, 29, 32, 35, 43, 46, 52, 55, 57, 60, 63, 68, 73, 78, 84, 90, 92, 94, 96, 99, 102, 105, 107, 109, 112, 114, 116, 121, 124, 128, 130, 133, 135, 141, 6, 11, 16, 22, 30, 33, 36, 44, 47, 64, 74, 79, 85, 97, 100, 103, 117, 125, 131, 136, 142, 12, 17, 23, 37, 48, 75, 80, 86, 137, 143, 13, 18, 38, 144, 39, 145, 40, 146, 41, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163]
    k = 0
    K_IL_max = 164
    p = np.empty(c.shape[0], 'int')
    for p_IL_max in p_IL_max_table:
        if p_IL_max >= (K_IL_max - K):
            p[k] = p_IL_max - (K_IL_max - K)
            k += 1
    return p

# deinterleave
p_IL = interleave(decoded)
decoded2 = np.empty(decoded.shape, 'int')
np.put(decoded2, p_IL, decoded)

the result in python is [0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1]

the result in matlab when using nrPolarDecode() is [0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0]

which is the same as in python, except the last bit. When I put the matlab result into nrCRCDecode() it says checksum ok, with the python result it says checksum wrong.

input = [0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1]';
[data, cs] = nrCRCDecode(input,'24C')

gives cs = 1, which means error

input = [0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0]';
[data, cs] = nrCRCDecode(input,'24C')

gives cs = 0, which means ok

Describe the bug Passing a Constellation object as the constellation parameter to MaximumLikelihoodDetector gives a TypeError

To Reproduce

See this example

import sionna
from sionna.mapping import Constellation
from sionna.mimo import MaximumLikelihoodDetector

print(sionna.__version__)

mld1 = MaximumLikelihoodDetector('bit', 'app', 6, constellation_type='qam', num_bits_per_symbol=2)
mld2 = MaximumLikelihoodDetector('bit', 'app', 6, constellation=Constellation(constellation_type='qam', num_bits_per_symbol=2))

which produces this output:

0.10.0
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Input In [6], in <cell line: 8>()
      5 print(sionna.__version__)
      7 mld1 = MaximumLikelihoodDetector('bit', 'app', 6, constellation_type='qam', num_bits_per_symbol=2)
----> 8 mld2 = MaximumLikelihoodDetector('bit', 'app', 6, constellation=Constellation(constellation_type='qam', num_bits_per_symbol=2))

File ~/anaconda3/envs/noma/lib/python3.10/site-packages/sionna/mimo/detection.py:237, in MaximumLikelihoodDetector.__init__(self, output, demapping_method, k, constellation_type, num_bits_per_symbol, constellation, hard_out, dtype, **kwargs)
    234 self._c = tf.cast(c, tf.int32)
    236 if output == 'bit':
--> 237     self._logits2llr = SymbolLogits2LLRs(
    238                             method=demapping_method,
    239                             num_bits_per_symbol=num_bits_per_symbol,
    240                             hard_out=hard_out,
    241                             dtype=dtype.real_dtype,
    242                             **kwargs)

File ~/anaconda3/envs/noma/lib/python3.10/site-packages/sionna/mapping.py:641, in SymbolLogits2LLRsWithPrior.__init__(self, method, num_bits_per_symbol, hard_out, dtype, **kwargs)
    639 self._hard_out = hard_out
    640 self._num_bits_per_symbol = num_bits_per_symbol
--> 641 num_points = int(2**num_bits_per_symbol)
    643 # Array composed of binary representations of all symbols indices
    644 a = np.zeros([num_points, num_bits_per_symbol])

TypeError: unsupported operand type(s) for ** or pow(): 'int' and 'NoneType'

Expected behavior No TypeError

To Fix: Replace line 239 of detection.py with

num_bits_per_symbol=self._constellation.num_bits_per_symbol,

Describe the bug While plotting the panel array geometry using the PanelArray.show() function, the same antenna array is shown for both polarization type 'VH' and 'cross'. Ideally, the antenna orientation for cross polarization should be rotated by 45 degree w.r.t. polarization type 'VH.

To Reproduce

from sionna.channel.tr38901 import PanelArray

array = PanelArray(num_rows_per_panel = 4,
                   num_cols_per_panel = 4,
                   polarization = 'dual',
                   polarization_type = 'cross',
                   antenna_pattern = '38.901',
                   carrier_frequency = 3.5e9)

array.show()

Expected behavior The plot when polarization type is set to 'cross' should be as below:

Proposed solution The markers for antenna, when plotting figure with polarization type 'cross', should be rotated by 45 degrees. Please find the modified PanelArray.show() function:

def show(self):
    """Show the panel array geometry"""
    marker_vert = MarkerStyle("|")
    marker_horz = MarkerStyle("_")
    if self._polarization == 'dual' and self._polarization_type == 'cross':
        marker_vert = MarkerStyle("|")
        marker_vert._transform.rotate_deg(-45)
        marker_horz = MarkerStyle("_")
        marker_horz._transform.rotate_deg(-45)

    fig = plt.figure()
    pos_pol1 = self._ant_pos_pol1
    plt.plot(pos_pol1[:,1], pos_pol1[:,2], marker = marker_vert,
        markeredgecolor='red', markersize="20", linestyle="None",
        markeredgewidth="2")
    for i, p in enumerate(pos_pol1):
        fig.axes[0].annotate(self._ant_ind_pol1[i].numpy()+1, (p[1], p[2]))
    if self._polarization == 'dual':
        pos_pol2 = self._ant_pos_pol2
        plt.plot(pos_pol2[:,1], pos_pol2[:,2], marker = marker_horz,
            markeredgecolor='black', markersize="20", linestyle="None",
            markeredgewidth="1")
    plt.xlabel("y (m)")
    plt.ylabel("z (m)")
    plt.title("Panel Array")
    plt.legend(["Polarization 1", "Polarization 2"], loc="upper right")

Describe the bug The class OFDMModulator() returns a time-domain OFDM signal with size = None while Training the Keras model. (only while training the model)

To Reproduce Please find a gist link : "https://gist.github.com/kassankar/70d62384d05b00fa8b9486ed862a4784" to reproduce the error.

The bug occurred while running block number 7 due to OFDMModulator() in block number 6.

Expected behavior Based on the OFDMModulator description, the function should return an output size = num_ofdm_symbols*(fft_size+cyclic_prefix_length)

Screenshots

Additional context The problem is directly connected to the function flatten_last_dims() used in the OFDMModulator() class. More precisely, in the last two lines new_shape = tf.concat([shape[:-num_dims], [-1]], 0), return tf.reshape(tensor, new_shape). I think the problem is occuring due to the use of [-1] in the tf.reshape() function. In the training the batch_size is equal to None, and based on the description of the tf.reshape() function with an axis shape = [-1], the dimension is computed so that the total size remains constant. However, due to batch_size = None, the tf.reshape() return another 'None' dimension for the last output.

Note: tf.reshape() with [-1] is also used in OFDMDemodulator(), in new_shape = tf.concat([tf.shape(inputs)[:-1], [-1], [self.fft_size + self.cyclic_prefix_length]], 0)

Hi! First of all thank you for this software. I was trying to build a simulation where 2 MIMO BS receive data from a single-antenna receiver. Accordingly, I set the number of tx_streams to be equal to 1; however, the StreamManagement object returns that the _detection_desired_ind is empty, because the num_streams_per_rx is 0. Therefore, the stream results undesired to both receiver and the LMMSEEqualizer tries to output an empty x_hat estimated from an empty h_dt_desired tensor. This generated an error in line 241 of ofdm/equalizer.py, when x_hat is reshaped.

Is this a bug, or is it intended? If the latter, did I make some mistake in the set up? In the follows, I paste the code for the setup of the Stream Management, while is it attach the whole script (only 86 lines of code). Thank you!

self._num_ut = 1
self._num_bs = 2
self._num_ut_ant = 1
self._num_bs_ant = 64
self._num_streams_per_tx = self._num_ut_ant
self._perfect_csi = perfect_csi

# Setup Stream Management
self._rx_tx_association = np.zeros([self._num_bs, self._num_ut])
self._rx_tx_association[:, 0] = 1
self._stream_management = sn.mimo.StreamManagement(self._rx_tx_association, self._num_streams_per_tx)

rayleigh_experiment.zip

Screenshots Added the screenshot of the error:

System:

• OS: Ubuntu 20.04 and Linux Mint (last)
• NVIDIA Driver version: 510.85.02
• CUDA version: 11.6
• Python version: 3.7 and 3.8
• TensorFlow version: 2.6.2 Tested also in Google collab, same issue there.

Problem: In the beginner tutorial part 3: advanced link level simulations, when executing the command "UT_ARRAY.show()" the following error occurs: "Type Error: float() argument must be a string or a number, not 'MarkStyle'".

Possible solution: The error happens because in the show() function of the PanelArray() class, marker_vert and marker_horz are objects and the argument 'marker' must be a string. The possible solution is to use the get_marker() function which returns a string. This way: marker = marker_vert.get_marker() marker = marker_horz.get_marker()

Best regards, Pedro

I am in the process of setting up the environment on Mac OS and explore Sionna using the examples provided. The Discover Sionna notebook throws-up error half-way into the execution. I am attaching the complete pdf version of the executed notebook. Further, I have added code in the notebook to print the version of OS, python, TensorFlow etc installed in my environment to help debugging.

Please help me setup the environment correctly. Discover_Sionna-3.pdf

Describe the bug The layer ResourceGridDemapper does not work properly.

To Reproduce Please, find below a minimal code to reproduce the error:

import tensorflow as tf
from tensorflow.keras import Model
from sionna.ofdm import ResourceGrid,ResourceGridDemapper
from sionna.mimo import StreamManagement
from sionna.utils import QAMSource
import numpy as np

class E2ESystem(Model):
    def __init__(self):
        super().__init__()
        self.cp_length = 0
        self.fft_size = 72
        self.num_ofdm_symbols = 14
        self.num_streams_per_tx = 1
        self.subcarrier_spacing = 15e3
        self.stream_manager = StreamManagement(np.array([[1]]),1)
        self.rg = ResourceGrid(num_ofdm_symbols = self.num_ofdm_symbols,
                  fft_size = self.fft_size,
                  subcarrier_spacing = self.subcarrier_spacing,
                  num_tx = 1,
                  num_streams_per_tx = self.num_streams_per_tx)
        self.qam = QAMSource(4)
        self.rg_demap  = ResourceGridDemapper(self.rg,self.stream_manager)

    
    @tf.function
    def call(self, batch_size):
        x_rg = self.qam([batch_size, 1, 1, self.num_ofdm_symbols, self.fft_size])
        print(x_rg.shape)
        x_rg_dem = self.rg_demap(x_rg)
        return x_rg_dem

e2e = E2ESystem()
e2e(128)

Expected behavior The ResourceGridDemapper layer should return the data mapped from the qam layer.

Proposed solution In the ResourceGridDemapper layer replace : 1. if tf.rank(y)==5: --> if len(y.shape)==5: 2. if tf.shape(y)[-1]==1: --> if y.shape[-1]==1: Please, find below the new version of the layer:

class _ResourceGridDemapper(Layer):
    # pylint: disable=line-too-long
    r"""ResourceGridDemapper(resource_grid, stream_management, dtype=tf.complex64, **kwargs)

    Extracts data-carrying resource elements from a resource grid.

    This layer takes as input an OFDM :class:`~sionna.ofdm.ResourceGrid` and
    extracts the data-carrying resource elements. In other words, it implements
    the reverse operation of :class:`~sionna.ofdm.ResourceGridMapper`.

    Parameters
    ----------
    resource_grid : ResourceGrid
        An instance of :class:`~sionna.ofdm.ResourceGrid`.

    stream_management : StreamManagement
        An instance of :class:`~sionna.mimo.StreamManagement`.

    dtype : tf.Dtype
        Datatype for internal calculations and the output dtype.
        Defaults to `tf.complex64`.

    Input
    -----
    : [batch_size, num_rx, num_streams_per_rx, num_ofdm_symbols, fft_size, data_dim]
        The full OFDM resource grid in the frequency domain.
        The last dimension `data_dim` is optional. If `data_dim`
        is used, it refers to the dimensionality of the data that should be
        demapped to individual streams. An example would be LLRs.

    Output
    ------
    : [batch_size, num_rx, num_streams_per_rx, num_data_symbols, data_dim]
        The data that were mapped into the resource grid.
        The last dimension `data_dim` is only returned if it was used for the
        input.
    """
    def __init__(self,
                 resource_grid,
                 stream_management,
                 dtype=tf.complex64,
                 **kwargs):
        super().__init__(dtype=dtype, **kwargs)
        self._stream_management = stream_management
        self._resource_grid = resource_grid

        # Precompute indices to extract data symbols
        mask = resource_grid.pilot_pattern.mask
        num_data_symbols = resource_grid.pilot_pattern.num_data_symbols
        data_ind = tf.argsort(flatten_last_dims(mask), direction="ASCENDING")
        self._data_ind = data_ind[...,:num_data_symbols]

    def call(self, y): # pylint: disable=arguments-renamed

        # y has shape
        # [batch_size, num_rx, num_streams_per_rx, num_ofdm_symbols,...
        # ..., fft_size, data_dim]

        # If data_dim is not provided, add a dummy dimension
        if len(y.shape)==5:
            y = tf.expand_dims(y, -1)

        # Remove nulled subcarriers from y (guards, dc). New shape:
        # [batch_size, num_rx, num_rx_ant, ...
        #  ..., num_ofdm_symbols, num_effective_subcarriers, data dim]
        y = tf.gather(y, self._resource_grid.effective_subcarrier_ind, axis=-2)

        # Transpose tensor to shape
        # [num_rx, num_streams_per_rx, num_ofdm_symbols,...
        #  ..., num_effective_subcarriers, data_dim, batch_size]
        y = tf.transpose(y, [1, 2, 3, 4, 5, 0])

        # Merge num_rx amd num_streams_per_rx
        # [num_rx * num_streams_per_rx, num_ofdm_symbols,...
        #  ...,num_effective_subcarriers, data_dim, batch_size]
        y = flatten_dims(y, 2, 0)

        # Put first dimension into the right ordering
        stream_ind = self._stream_management.stream_ind
        y = tf.gather(y, stream_ind, axis=0)

        # Reshape first dimensions to [num_tx, num_streams] so that
        # we can compared to the way the streams were created.
        # [num_tx, num_streams, num_ofdm_symbols, num_effective_subcarriers,...
        #  ..., data_dim, batch_size]
        num_streams = self._stream_management.num_streams_per_tx
        num_tx = self._stream_management.num_tx
        y = split_dim(y, [num_tx, num_streams], 0)

        # Flatten resource grid dimensions
        # [num_tx, num_streams, num_ofdm_symbols*num_effective_subcarriers,...
        #  ..., data_dim, batch_size]
        y = flatten_dims(y, 2, 2)

        # Gather data symbols
        # [num_tx, num_streams, num_data_symbols, data_dim, batch_size]
        y = tf.gather(y, self._data_ind, batch_dims=2, axis=2)

        # Put batch_dim first
        # [batch_size, num_tx, num_streams, num_data_symbols]
        y = tf.transpose(y, [4, 0, 1, 2, 3])

        # Squeeze data_dim
        if y.shape[-1]==1:
            y = tf.squeeze(y, -1)

        return y

I want to decode the PBCH. In matlab I can do it like this

    iBIL = false;
    iIL = true;
    crcLen = 24;
    nMax = 9;
    A = 32;
    P = 24;
    K = A+P;
    N = 512;
    decIn = nrRateRecoverPolar(pbchBits_csi,K,N,iBIL);
    decBits = nrPolarDecode(decIn,K,E,polarListLength,nMax,iIL,crcLen);

However when I try it with sionna like this

    from sionna.fec.polar import PolarEncoder, Polar5GEncoder, PolarSCLDecoder, Polar5GDecoder, PolarSCDecoder
    enc = Polar5GEncoder(k=K, n=864)
    dec = Polar5GDecoder(enc, dec_type="SCL", list_size=8)
    dec(np.expand_dims(pbchBits_csi, 0))

I get a wrong (different than matlab) result. I know the matlab result is correct, because the CRC gives 0. I found something suspicious with enc._init_rate_match(k, n), when I call it with k=54 and n=512 (or n=864) it gives CRC order 11, but it should be 24 (24C CRC polynomial). Is this a bug?

Description

Until now it is not possible to decode 5G downstream signals, because deinterleaving according to TS 38.212 Table 5.3.1.1 (I_IL) is not implemented. This PR implements:

• optional deinterleaving (I_IL) in PolarSCLDecoder and PolarEncoder and Polar5GEncoder
• optional channel interleaving (B_IL) in Polar5GDecoder and Polar5GEncoder
• optional specification of CRC polynomial to Polar5GEncoder
• make payload incl crc available as attribute in Polar5GDecoder

This PR is still WIP, but I already open it for discussion. All changes preserve the old API and should not affect existing use cases.

Decoding of 5G PBCH signal with deinterleaving works (the checksum inside the SCL decoder becomes True) The example with real 5G data can be found here: https://github.com/catkira/py3gpp/blob/master/examples/test_py3gpp.ipynb

I will squash all commits in this PR when its done. It solves this issue: https://github.com/NVlabs/sionna/issues/77

Checklist

[x] Detailed description [x] Added references to issues and discussions [ ] Added / modified documentation as needed [x] Added / modified unit tests as needed [ ] Passes all tests [x] Lint the code [ ] Performed a self review [ ] Ensure you Signed-off the commits. Required to accept contributions! [x] Co-authored with someone? Add Co-authored-by: [email protected] and ensure they signed off their commits too.

Description

Added:

• dual polarization for the SSFM by the implementation of the corresponding Manakov equation.
• dual polarization for the EDFA.
• test cases for the latter two classes.

Fixed:

• noise bug as proposed by sebastian-xyz.
• typos in documentation.

Additional commits due to "sign"-requirement.

Checklist

[x] Detailed description [ ] Added references to issues and discussions [x] Added / modified documentation as needed [x] Added / modified unit tests as needed [x] Passes all tests [x] Lint the code [x] Performed a self review [x] Ensure you Signed-off the commits. Required to accept contributions! [x] Co-authored with someone? Add Co-authored-by: [email protected] and ensure they signed off their commits too.

Description

Fixes the calculation of the amplifier noise applied in each step of the SSFM if the Raman amplifier is used. The calculation for the noise power of the imaginary part was incorrect, since a division was used instead of a multiplication. To further simplify the code, the calculation of the noise power is only done once per step.

Checklist

• [x] Detailed description
• [ ] Added references to issues and discussions
• [x] Added / modified documentation as needed
• [x] Added / modified unit tests as needed
• [x] Passes all tests
• [x] Lint the code
• [x] Performed a self review
• [x] Ensure you Signed-off the commits. Required to accept contributions!
• [ ] Co-authored with someone? Add Co-authored-by: [email protected] and ensure they signed off their commits too.

Describe the bug _llr_max attribute is set in both LDPC5GDecoder and LDPCBPDecoder. Since the initialization of the latter is done after the setting in the child class, this might create some inconsistency.

To Reproduce Static code analysis, so not required IMO.

Expected behavior Set the value only in one place (e.g. in parent class and passing it as argument).

Screenshots N/A

System: N/A

GPUs: N/A

Additional context N/A