def dice_csi_tags(self, data, type, num_inputs, num_tags, tag_pos, vlen=1):
tags = []
expected_result = np.empty([np.size(data, 0)*np.size(data,1)], dtype=complex)
if type == 'MMSE':
# Add an SNR tag at the start of the stream for MMSE.
tags.append(gr.tag_utils.python_to_tag((0,
pmt.string_to_symbol("snr"),
pmt.make_f32vector(num_inputs, 1e8),
pmt.from_long(0))))
for i in range(0, num_tags):
# Randomly generate CSI for one symbol.
csi = (np.random.randn(vlen, num_inputs, num_inputs) + 1j * np.random.randn(vlen, num_inputs, num_inputs))
# Assign the CSI vector to a PMT vector.
csi_pmt = pmt.make_vector(vlen, pmt.make_vector(num_inputs, pmt.make_c32vector(num_inputs, 1.0)))
for k, carrier in enumerate(csi):
carrier_vector_pmt = pmt.make_vector(num_inputs, pmt.make_c32vector(num_inputs, csi[k][0][0]))
for l, rx in enumerate(csi[k]):
line_vector_pmt = pmt.make_c32vector(num_inputs, csi[k][l][0])
for m, tx in enumerate(csi[k][l]):
pmt.c32vector_set(v=line_vector_pmt, k=m, x=csi[k][l][m])
pmt.vector_set(carrier_vector_pmt, l, line_vector_pmt)
pmt.vector_set(csi_pmt, k, carrier_vector_pmt)
# Append stream tags with CSI to data stream.
tags.append(gr.tag_utils.python_to_tag((tag_pos[i],
pmt.string_to_symbol("csi"),
csi_pmt,
pmt.from_long(0))))
# Calculate expected result.
expected_result[tag_pos[i]*num_inputs::] = np.reshape(np.transpose(np.dot(np.linalg.inv(csi), data[::, tag_pos[i]::])), (np.size(data, 0)*(np.size(data,1)-tag_pos[i])))
return tags, expected_result
def test_001_t (self):
# Define test params.
data_length = 20
repetitions = 3
num_tags = 3
vlen = 1
result = np.empty(shape=[data_length*vlen], dtype=complex)
for n in range(repetitions):
# Generate random input data.
data = np.random.randn(data_length*vlen) + 1j * np.random.randn(data_length*vlen)
# Generate random tag positions.
tag_pos = np.random.randint(low=0, high=data_length/2, size=num_tags)*2
# Add tag pos 0 for initial channel state.
tag_pos = np.append(tag_pos, 0)
tag_pos = np.sort(tag_pos)
# Iterate over tags.
for i in range(0, num_tags+1):
# Randomly generate CSI for one symbol. (Flat channel)
csi = (np.random.randn(1, 1, 2) + 1j * np.random.randn(1, 1, 2))
# Assign the CSI vector to a PMT vector.
csi_pmt = pmt.make_vector(vlen, pmt.make_vector(1, pmt.make_c32vector(2, 1.0)))
for k, carrier in enumerate(csi):
carrier_vector_pmt = pmt.make_vector(1, pmt.make_c32vector(2, csi[0][0][0]))
for l, rx in enumerate(csi[k]):
row_vector_pmt = pmt.make_c32vector(2, csi[0][l][0])
for m, tx in enumerate(csi[k][l]):
pmt.c32vector_set(v=row_vector_pmt, k=m, x=csi[0][l][m])
pmt.vector_set(carrier_vector_pmt, l, row_vector_pmt)
pmt.vector_set(csi_pmt, k, carrier_vector_pmt)
# Append stream tags with CSI to data stream.
tags = [(gr.tag_utils.python_to_tag((0,
pmt.string_to_symbol("csi"),
csi_pmt,
pmt.from_long(0))))]
# Build up the test flowgraph.
subdata = data[tag_pos[i]*vlen::]
src = blocks.vector_source_c(data=subdata,
repeat=False,
tags=tags)
alamouti_encoder = digital.alamouti_encoder_cc()
# Simulate channel with matrix multiplication.
channel = blocks.multiply_matrix_cc_make(csi[0])
alamouti_decoder = digital.alamouti_decoder_cc(vlen=vlen)
sink = blocks.vector_sink_c()
self.tb.connect(src, alamouti_encoder, channel,
blocks.stream_to_vector(gr.sizeof_gr_complex, vlen),
alamouti_decoder, sink)
self.tb.connect((alamouti_encoder, 1), (channel, 1))
# Run flowgraph.
self.tb.run()
result[tag_pos[i]*vlen::] = sink.data()
'''
Check if the expected result (=the data itself, because
we do a loopback) equals the actual result. '''
self.assertComplexTuplesAlmostEqual(data, result, 4)
def notest_debug(self):
src = blocks.vector_source_c(range(100), False, 1, [])
separator = inspector_test.signal_separator_c(32000, firdes.WIN_HAMMING, 0.1, 100)
msg = pmt.make_vector(1, pmt.PMT_NIL)
flanks = pmt.make_f32vector(2, 0.0)
pmt.f32vector_set(flanks, 0, 12500)
pmt.f32vector_set(flanks, 1, 20)
pmt.vector_set(msg, 0, flanks)
msg_src = blocks.message_strobe(msg, 100)
def test_001_t(self):
src = blocks.vector_source_c(list(range(10000)), False, 1, [])
separator = inspector_test.signal_separator_c(
32000, firdes.WIN_HAMMING, 0.1, 100, False,
inspector_test.map_float_vector({0.0: [0.0]}))
vec_sink = blocks.vector_sink_c(1)
ext = inspector_test.signal_extractor_c(0)
snk = blocks.vector_sink_c(1)
# pack message
msg = pmt.make_vector(1, pmt.PMT_NIL)
flanks = pmt.make_f32vector(2, 0.0)
pmt.f32vector_set(flanks, 0, 12500)
pmt.f32vector_set(flanks, 1, 20)
pmt.vector_set(msg, 0, flanks)
msg_src = blocks.message_strobe(msg, 100)
taps = filter.firdes.low_pass(1, 32000, 500, 50, firdes.WIN_HAMMING,
6.76)
self.tb.connect(src, separator)
self.tb.connect(src, vec_sink)
self.tb.msg_connect((msg_src, 'strobe'), (separator, 'map_in'))
self.tb.msg_connect((separator, 'sig_out'), (ext, 'sig_in'))
self.tb.connect(ext, snk)
self.tb.start()
time.sleep(0.3)
self.tb.stop()
self.tb.wait()
data = vec_sink.data()
sig = numpy.zeros(len(vec_sink.data()), dtype=complex)
for i in range(len(vec_sink.data())):
sig[i] = data[i] * numpy.exp(
-1j * 2 * numpy.pi * 12500 * i * 1 / 32000)
taps = filter.firdes.low_pass(1, 32000, 900, 90, firdes.WIN_HAMMING,
6.76)
sig = numpy.convolve(sig, taps, 'full')
out = numpy.empty([0])
decim = 32000 / 20 / 100
j = 0
for i in range(len(sig) / decim):
out = numpy.append(out, sig[j])
j += decim
data = snk.data()
for i in range(min(len(out), len(data))):
self.assertComplexAlmostEqual2(out[i], data[i], abs_eps=0.01)
def handle_msg(self, msg):
cell_id = pmt.to_long(msg)
if self.cell_id == cell_id:
return
else:
self.cell_id = cell_id
print "received cell_id = " + str(cell_id)
seqs = pmt.make_vector(10, pmt.make_vector(32, pmt.from_double(0.0)))
for ns in range(10):
scr = utils.get_pcfich_scrambling_sequence(cell_id, ns)
scr = utils.encode_nrz(scr)
pmt_seq = pmt.make_vector(len(scr), pmt.from_double(0.0))
for i in range(len(scr)):
pmt.vector_set(pmt_seq, i, pmt.from_double(scr[i]))
pmt.vector_set(seqs, ns, pmt_seq)
self.message_port_pub(self.msg_buf_out, seqs)
def handle_msg(self, msg):
cell_id = pmt.to_long(msg)
if self.cell_id == cell_id:
return
else:
self.cell_id = cell_id
print "received cell_id = " + str(cell_id)
seqs = pmt.make_vector(10, pmt.make_vector(32, pmt.from_double(0.0)))
for ns in range(10):
scr = utils.get_pcfich_scrambling_sequence(cell_id, ns)
scr = utils.encode_nrz(scr)
pmt_seq = pmt.make_vector(len(scr), pmt.from_double(0.0))
for i in range(len(scr)):
pmt.vector_set(pmt_seq, i, pmt.from_double(scr[i]))
pmt.vector_set(seqs, ns, pmt_seq)
self.message_port_pub(self.msg_buf_out, seqs)
def no_test_sebastian (self):
src = blocks.vector_source_c(range(100), False, 1, [])
separator = inspector_test.signal_separator_c(32000, firdes.WIN_HAMMING, 0.1, 100)
# pack message
msg = pmt.make_vector(1, pmt.PMT_NIL)
flanks = pmt.make_f32vector(2, 0.0)
pmt.f32vector_set(flanks, 0, 12490)
pmt.f32vector_set(flanks, 1, 12510)
pmt.vector_set(msg, 0, flanks)
msg_src = blocks.message_strobe(msg, 100)
self.tb.connect(src, separator)
self.tb.msg_connect((msg_src, 'strobe'), (separator, 'map_in'))
self.tb.run()
def test_001_t (self):
src = blocks.vector_source_c(range(10000), False, 1, [])
separator = inspector_test.signal_separator_c(32000, firdes.WIN_HAMMING, 0.1, 100, False, inspector_test.map_float_vector({0.0:[0.0]}))
vec_sink = blocks.vector_sink_c(1)
ext = inspector_test.signal_extractor_c(0)
snk = blocks.vector_sink_c(1)
# pack message
msg = pmt.make_vector(1, pmt.PMT_NIL)
flanks = pmt.make_f32vector(2, 0.0)
pmt.f32vector_set(flanks, 0, 12500)
pmt.f32vector_set(flanks, 1, 20)
pmt.vector_set(msg, 0, flanks)
msg_src = blocks.message_strobe(msg, 100)
taps = filter.firdes.low_pass(1, 32000, 500, 50, firdes.WIN_HAMMING, 6.76)
self.tb.connect(src, separator)
self.tb.connect(src, vec_sink)
self.tb.msg_connect((msg_src, 'strobe'), (separator, 'map_in'))
self.tb.msg_connect((separator, 'sig_out'), (ext, 'sig_in'))
self.tb.connect(ext, snk)
self.tb.start()
time.sleep(0.3)
self.tb.stop()
self.tb.wait()
data = vec_sink.data()
sig = numpy.zeros(len(vec_sink.data()), dtype=complex)
for i in range(len(vec_sink.data())):
sig[i] = data[i]*numpy.exp(-1j*2*numpy.pi*12500*i*1/32000)
taps = filter.firdes.low_pass(1, 32000, 900, 90, firdes.WIN_HAMMING, 6.76)
sig = numpy.convolve(sig, taps, 'full')
out = numpy.empty([0])
decim = 32000/20/100
j = 0
for i in range(len(sig)/decim):
out = numpy.append(out, sig[j])
j += decim
data = snk.data()
for i in range(min(len(out), len(data))):
self.assertComplexAlmostEqual2(out[i], data[i], abs_eps=0.01)
def test_001_t (self):
# set up fg
src = blocks.vector_source_c(range(1024), False, 1, [])
# pack message
msg = pmt.make_vector(1, pmt.PMT_NIL)
flanks = pmt.make_f32vector(2, 0.0)
pmt.f32vector_set(flanks, 0, 12500)
pmt.f32vector_set(flanks, 1, 8)
pmt.vector_set(msg, 0, flanks)
msg_src = blocks.message_strobe(msg, 100)
separator = inspector_test.signal_separator_c(1024, firdes.WIN_HAMMING, 0.1, 1)
extractor = inspector.signal_extractor_c(0)
taps = firdes.low_pass(1, 1024, 2, 0.1*2, firdes.WIN_HAMMING, 6.76)
reference = numpy.convolve(range(1024), taps, 'same')
#xlator = filter.freq_xlating_fir_filter_ccf(160, taps, 12500, 32000)
stv1 = blocks.stream_to_vector(gr.sizeof_gr_complex, 128)
#stv2 = blocks.stream_to_vector(gr.sizeof_gr_complex, 128)
snk1 = blocks.vector_sink_c(128)
#snk2 = blocks.vector_sink_c(128)
null1 = blocks.null_sink(gr.sizeof_gr_complex)
# connect this
self.tb.connect(src, separator)
self.tb.msg_connect((msg_src, 'strobe'), (separator, 'map_in'))
#self.tb.connect(src, xlator)
#self.tb.connect(xlator, null1)
self.tb.msg_connect((separator, 'msg_out'), (extractor, 'sig_in'))
self.tb.connect(extractor, stv1)
#self.tb.connect(xlator, stv2)
self.tb.connect(stv1, snk1)
#self.tb.connect(stv2, snk2)
self.tb.start()
time.sleep(0.5)
self.tb.stop()
# check data
data1 = snk1.data()
#data2 = snk2.data()
data2 = reference[0::8]
for i in range(min(len(data2), len(data1))):
self.assertComplexAlmostEqual(data1[i], data2[i], 4)
def build_and_run_flowgraph(self,
repetitions,
data_length,
num_inputs_min,
num_inputs_max,
equalizer_type,
vlen=[1, 1]):
for a in range(repetitions):
num_inputs = np.random.randint(low=num_inputs_min,
high=num_inputs_max + 1)
# Generate random input data.
data = np.random.randn(
data_length * num_inputs * vlen[a]) + 1j * np.random.randn(
data_length * num_inputs * vlen[a])
# Randomly generate CSI for one symbol.
csi = (np.random.randn(vlen[a], num_inputs, num_inputs) +
1j * np.random.randn(vlen[a], num_inputs, num_inputs))
# Assign the CSI vector to a PMT vector.
csi_pmt = pmt.make_vector(
vlen[a],
pmt.make_vector(num_inputs,
pmt.make_c32vector(num_inputs, 1.0)))
for k, carrier in enumerate(csi):
carrier_vector_pmt = pmt.make_vector(
num_inputs, pmt.make_c32vector(num_inputs, csi[k][0][0]))
for l, rx in enumerate(csi[k]):
line_vector_pmt = pmt.make_c32vector(
num_inputs, csi[k][l][0])
for m, tx in enumerate(csi[k][l]):
pmt.c32vector_set(v=line_vector_pmt,
k=m,
x=csi[k][l][m])
pmt.vector_set(carrier_vector_pmt, l, line_vector_pmt)
pmt.vector_set(csi_pmt, k, carrier_vector_pmt)
# Append stream tags with CSI to data stream.
tags = [(gr.tag_utils.python_to_tag(
(0, pmt.string_to_symbol("csi"), csi_pmt, pmt.from_long(0))))]
if equalizer_type == 'MMSE':
# Add an SNR tag at the start of the stream for MMSE.
tags.append(
gr.tag_utils.python_to_tag(
(0, pmt.string_to_symbol("snr"),
pmt.make_f32vector(num_inputs,
1e8), pmt.from_long(0))))
# Build up the test flowgraph.
src = blocks.vector_source_c(data=data, repeat=False, tags=tags)
vblast_encoder = digital.vblast_encoder_cc(num_inputs)
demux = []
channels = []
for i in range(0, vlen[a]):
for j in range(0, num_inputs):
demux.append(
blocks.keep_m_in_n(gr.sizeof_gr_complex, 1, vlen[a],
i))
# Simulate channel with matrix multiplication.
channels.append(blocks.multiply_matrix_cc_make(csi[i]))
mux = []
s2v = []
for i in range(0, num_inputs):
mux.append(
blocks.stream_mux(gr.sizeof_gr_complex, [1] * vlen[a]))
s2v.append(
blocks.stream_to_vector(gr.sizeof_gr_complex, vlen[a]))
vblast_decoder = digital.vblast_decoder_cc(num_inputs,
equalizer_type, vlen[a])
sink = blocks.vector_sink_c()
self.tb.connect(src, vblast_encoder)
for n in range(0, num_inputs):
for i in range(0, vlen[a]):
self.tb.connect((vblast_encoder, n),
demux[i * num_inputs + n],
(channels[i], n))
self.tb.connect((channels[i], n), (mux[n], i))
self.tb.connect(mux[n], s2v[n], (vblast_decoder, n))
self.tb.connect(vblast_decoder, sink)
# Run flowgraph.
self.tb.run()
'''
Check if the expected result (=the data itself, because
we do a loopback) equals the actual result. '''
self.assertComplexTuplesAlmostEqual(data, sink.data(), 2)
def dice_csi_tags(self, input, num_tags, tag_pos, vlen=1):
tags = []
# Calculate initial behaviour before first tag.
output = np.empty(shape=[len(input)], dtype=complex)
output[::2] = (input[::2] + np.conj(input[1::2])) / 2.0
output[1::2] = (input[::2] - np.conj(input[1::2])) / 2.0
# Iterate over tags and update the calculated output according to the diced CSI.
for i in range(0, num_tags):
# Randomly generate CSI for one symbol.
csi = (np.random.randn(vlen, 1, 2) +
1j * np.random.randn(vlen, 1, 2))
# Assign the CSI vector to a PMT vector.
csi_pmt = pmt.make_vector(
vlen, pmt.make_vector(1, pmt.make_c32vector(2, 1.0)))
for k, carrier in enumerate(csi):
carrier_vector_pmt = pmt.make_vector(
1, pmt.make_c32vector(2, csi[k][0][0]))
for l, rx in enumerate(csi[k]):
row_vector_pmt = pmt.make_c32vector(2, csi[k][l][0])
for m, tx in enumerate(csi[k][l]):
pmt.c32vector_set(v=row_vector_pmt,
k=m,
x=csi[k][l][m])
pmt.vector_set(carrier_vector_pmt, l, row_vector_pmt)
pmt.vector_set(csi_pmt, k, carrier_vector_pmt)
# Append stream tags with CSI to data stream.
tags.append(
gr.tag_utils.python_to_tag(
(tag_pos[i], pmt.string_to_symbol("csi"), csi_pmt,
pmt.from_long(0))))
# Calculate the expected result.
if vlen == 1:
total_branch_energy = np.sum(np.square(np.abs(csi[0][0])))
if tag_pos[i] % 2 == 0:
output[tag_pos[i]::2] = (
np.conj(csi[0][0][0]) * input[tag_pos[i]::2] +
csi[0][0][1] * np.conj(input[tag_pos[i] + 1::2])
) / total_branch_energy
output[tag_pos[i] + 1::2] = (
np.conj(csi[0][0][1]) * input[tag_pos[i]::2] -
csi[0][0][0] * np.conj(input[tag_pos[i] + 1::2])
) / total_branch_energy
else:
output[tag_pos[i] + 1::2] = (
np.conj(csi[0][0][0]) * input[tag_pos[i] + 1::2] +
csi[0][0][1] * np.conj(input[tag_pos[i] + 2::2])
) / total_branch_energy
output[tag_pos[i] + 2::2] = (
np.conj(csi[0][0][1]) * input[tag_pos[i] + 1::2] -
csi[0][0][0] * np.conj(input[tag_pos[i] + 2::2])
) / total_branch_energy
else:
if tag_pos[i] % 2 == 0:
for j in range(tag_pos[i] * vlen, len(input), 2):
total_branch_energy = np.sum(
np.square(np.abs(csi[j % vlen][0])))
output[j] = (np.conj(csi[j % vlen][0][0]) * input[j] +
csi[j % vlen][0][1] * np.conj(
input[j + 1])) / total_branch_energy
output[j +
1] = (np.conj(csi[j % vlen][0][1]) * input[j] -
csi[j % vlen][0][0] * np.conj(
input[j + 1])) / total_branch_energy
else:
for j in range((tag_pos[i] + 1) * vlen, len(input), 2):
total_branch_energy = np.sum(
np.square(np.abs(csi[j % vlen][0])))
output[j] = (np.conj(csi[j % vlen][0][0]) * input[j] +
csi[j % vlen][0][1] * np.conj(
input[j + 1])) / total_branch_energy
output[j +
1] = (np.conj(csi[j % vlen][0][1]) * input[j] -
csi[j % vlen][0][0] * np.conj(
input[j + 1])) / total_branch_energy
return tags, output