def test_qpsk_channel(self):
upper_bound = tuple(50.0 * numpy.ones((self.num_data, )))
lower_bound = tuple(0.0 * numpy.zeros((self.num_data, )))
self.cons = cons = digital.constellation_qpsk().base()
self.data = data = [
random.randrange(len(cons.points())) for x in range(self.num_data)
]
self.symbols = symbols = numpy.squeeze(
[cons.map_to_points_v(i) for i in data])
chan = channels.channel_model(noise_voltage=0.1,
frequency_offset=0.0,
epsilon=1.0,
taps=[1.0 + 0.0j],
noise_seed=0,
block_tags=False)
evm = digital.meas_evm_cc(cons, digital.evm_measurement_t_EVM_PERCENT)
vso = blocks.vector_source_c(symbols, False, 1, [])
mc = blocks.multiply_const_cc(3.0 + 2.0j)
vsi = blocks.vector_sink_f()
self.tb.connect(vso, chan, evm, vsi)
self.tb.run()
# check data
output_data = vsi.data()
self.assertLess(output_data, upper_bound)
self.assertGreater(output_data, lower_bound)
def test_qpsk_3tap_lms_training(self):
# set up fg
gain = 0.01 # LMS gain
num_taps = 16
num_samp = 2000
num_test = 500
cons = digital.constellation_qpsk().base()
rxmod = digital.generic_mod(cons, False, self.sps, True, self.eb,
False, False)
modulated_sync_word_pre = digital.modulate_vector_bc(
rxmod.to_basic_block(), self.preamble + self.preamble, [1])
modulated_sync_word = modulated_sync_word_pre[86:(
512 + 86)] # compensate for the RRC filter delay
corr_max = numpy.abs(
numpy.dot(modulated_sync_word, numpy.conj(modulated_sync_word)))
corr_calc = self.corr_thresh / (corr_max * corr_max)
preamble_symbols = self.map_symbols_to_constellation(
self.unpack_values(self.preamble, 8, 2), cons)
alg = digital.adaptive_algorithm_lms(cons, gain).base()
evm = digital.meas_evm_cc(cons, digital.evm_measurement_t.EVM_PERCENT)
leq = digital.linear_equalizer(num_taps, self.sps, alg, False,
preamble_symbols, 'corr_est')
correst = digital.corr_est_cc(modulated_sync_word, self.sps, 12,
corr_calc, digital.THRESHOLD_ABSOLUTE)
constmod = digital.generic_mod(constellation=cons,
differential=False,
samples_per_symbol=4,
pre_diff_code=True,
excess_bw=0.35,
verbose=False,
log=False)
chan = channels.channel_model(noise_voltage=0.0,
frequency_offset=0.0,
epsilon=1.0,
taps=(1.0 + 1.0j, 0.63 - .22j,
-.1 + .07j),
noise_seed=0,
block_tags=False)
vso = blocks.vector_source_b(self.preamble + self.data, True, 1, [])
head = blocks.head(gr.sizeof_float * 1, num_samp)
vsi = blocks.vector_sink_f()
self.tb.connect(vso, constmod, chan, correst, leq, evm, head, vsi)
self.tb.run()
# look at the last 1000 samples, should converge quickly, below 5% EVM
upper_bound = list(20.0 * numpy.ones((num_test, )))
lower_bound = list(0.0 * numpy.zeros((num_test, )))
output_data = vsi.data()
output_data = output_data[-num_test:]
self.assertLess(output_data, upper_bound)
self.assertGreater(output_data, lower_bound)
def test_qpsk_nonzeroevm(self):
# set up fg
expected_result = tuple(numpy.zeros((self.num_data, )))
self.cons = cons = digital.constellation_qpsk().base()
self.data = data = [
random.randrange(len(cons.points())) for x in range(self.num_data)
]
self.symbols = symbols = numpy.squeeze(
[cons.map_to_points_v(i) for i in data])
evm = digital.meas_evm_cc(cons, digital.evm_measurement_t_EVM_PERCENT)
vso = blocks.vector_source_c(symbols, False, 1, [])
mc = blocks.multiply_const_cc(3.0 + 2.0j)
vsi = blocks.vector_sink_f()
self.tb.connect(vso, mc, evm, vsi)
self.tb.run()
# check data
output_data = vsi.data()
self.assertNotEqual(expected_result, output_data)
def test_qpsk_3tap_lms_training(self):
# set up fg
gain = 0.001 # LMS gain
num_taps_fwd = 13
num_taps_rev = 3
num_test = 1000
cons = digital.constellation_qpsk().base()
rxmod = digital.generic_mod(cons, False, self.sps, True, self.eb,
False, False)
modulated_sync_word_pre = digital.modulate_vector_bc(
rxmod.to_basic_block(), self.preamble + self.preamble, [1])
# compensate for the RRC filter delay
modulated_sync_word = modulated_sync_word_pre[86:(512 + 86)]
corr_max = numpy.abs(
numpy.dot(modulated_sync_word, numpy.conj(modulated_sync_word)))
corr_calc = self.corr_thresh / (corr_max * corr_max)
preamble_symbols = self.map_symbols_to_constellation(
self.unpack_values(self.preamble, 8, 2), cons)
alg = digital.adaptive_algorithm_lms(cons, gain).base()
evm = digital.meas_evm_cc(cons, digital.evm_measurement_t.EVM_PERCENT)
dfe = digital.decision_feedback_equalizer(num_taps_fwd, num_taps_rev,
self.sps, alg, True,
preamble_symbols, 'corr_est')
correst = digital.corr_est_cc(modulated_sync_word, self.sps, 12,
corr_calc, digital.THRESHOLD_ABSOLUTE)
constmod = digital.generic_mod(constellation=cons,
differential=False,
samples_per_symbol=4,
pre_diff_code=True,
excess_bw=0.35,
verbose=False,
log=False)
chan = channels.channel_model(
noise_voltage=self.noise_voltage,
frequency_offset=0.0,
epsilon=1.0,
taps=(2.0, -0.459489 + -0.156287j, 0.185799 + 0.022878j, 0.060229 +
0.002171j, -0.023041 + -0.016539j, -0.004507 + 0.011984j,
-0.002597 + 0.002675j, 0.002320 + 0.000621j, -0.001420 +
-0.000126j, -0.000118 + -0.000520j, -0.000029 + -0.000201j,
0.000060 + -0.000002j, 0.169089 + -0.500778j, 0.419112 +
0.042402j, -0.139208 + 0.030027j, -0.080077 + 0.036473j,
0.026689 + 0.000837j, -0.002449 + 0.002320j,
-0.000567 + -0.002068j, 0.001528 + 0.002867j,
0.000944 + -0.000166j, 0.000218 + 0.000007j,
0.000214 + -0.000150j, 0.000004 + 0.000008j),
noise_seed=-44982235,
block_tags=False)
repeating_data = self.preamble + self.data * 200
vso = blocks.vector_source_b(repeating_data, False, 1, [])
head = blocks.head(gr.sizeof_char * 1, 500)
vsi = blocks.vector_sink_f()
self.tb.connect(vso, head, constmod, chan, correst, dfe, evm, vsi)
self.tb.run()
# look at the last 1000 samples, should converge quickly, below 5% EVM
upper_bound = list(20.0 * numpy.ones((num_test, )))
lower_bound = list(0.0 * numpy.zeros((num_test, )))
output_data = vsi.data()
output_data = output_data[-num_test:]
self.assertLess(output_data, upper_bound)
self.assertGreater(output_data, lower_bound)
