def setUp(self):
self.receiver = CorrRx(port=8888)
start_thread_with_cleanup(self, self.receiver, start_timeout=1)
self.correlator = correlator_fixture.correlator
self.corr_fix = correlator_fixture
self.corr_freqs = CorrelatorFrequencyInfo(self.correlator.configd)
dsim_conf = self.correlator.configd['dsimengine']
dig_host = dsim_conf['host']
self.dhost = FpgaDsimHost(dig_host, config=dsim_conf)
self.dhost.get_system_information()
# Increase the dump rate so tests can run faster
self.correlator.xeng_set_acc_time(0.2)
self.addCleanup(self.corr_fix.stop_x_data)
self.corr_fix.start_x_data()
self.corr_fix.issue_metadata()
class test_CBF(unittest.TestCase):
def setUp(self):
self.receiver = CorrRx(port=8888)
start_thread_with_cleanup(self, self.receiver, start_timeout=1)
self.correlator = correlator_fixture.correlator
self.corr_fix = correlator_fixture
self.corr_freqs = CorrelatorFrequencyInfo(self.correlator.configd)
dsim_conf = self.correlator.configd['dsimengine']
dig_host = dsim_conf['host']
self.dhost = FpgaDsimHost(dig_host, config=dsim_conf)
self.dhost.get_system_information()
# Increase the dump rate so tests can run faster
self.correlator.xeng_set_acc_time(0.2)
self.addCleanup(self.corr_fix.stop_x_data)
self.corr_fix.start_x_data()
self.corr_fix.issue_metadata()
# TODO 2015-05-27 (NM) Do test using get_vacc_offset(test_dump['xeng_raw']) to see if
# the VACC is rotated. Run this test first so that we know immediately that other
# tests will be b0rked.
def test_channelisation(self):
"""TP.C.1.19 CBF Channelisation Wideband Coarse L-band"""
test_chan = 1500
expected_fc = self.corr_freqs.chan_freqs[test_chan]
init_dsim_sources(self.dhost)
self.dhost.sine_sources.sin_0.set(frequency=expected_fc, scale=0.25)
# The signal source is going to quantise the requested freqency, so see what we
# actually got
source_fc = self.dhost.sine_sources.sin_0.frequency
# Get baseline 0 data, i.e. auto-corr of m000h
test_baseline = 0
test_data = self.receiver.get_clean_dump(DUMP_TIMEOUT)['xeng_raw']
b_mag = normalised_magnitude(test_data[:, test_baseline, :])
# find channel with max power
max_chan = np.argmax(b_mag)
# self.assertEqual(max_chan, test_chan,
# 'Channel with max power is not the test channel')
requested_test_freqs = self.corr_freqs.calc_freq_samples(
test_chan, samples_per_chan=101, chans_around=5)
# Placeholder of actual frequencies that the signal generator produces
actual_test_freqs = []
# Channel magnitude responses for each frequency
chan_responses = []
last_source_freq = None
for i, freq in enumerate(requested_test_freqs):
# LOGGER.info('Getting channel response for freq {}/{}: {} MHz.'.format(
# i+1, len(requested_test_freqs), freq/1e6))
print ('Getting channel response for freq {}/{}: {} MHz.'.format(
i+1, len(requested_test_freqs), freq/1e6))
if freq == expected_fc:
# We've already done this one!
this_source_freq = source_fc
this_freq_result = b_mag
else:
self.dhost.sine_sources.sin_0.set(frequency=freq, scale=0.125)
this_source_freq = self.dhost.sine_sources.sin_0.frequency
if this_source_freq == last_source_freq:
LOGGER.info('Skipping channel response for freq {}/{}: {} MHz.\n'
'Digitiser frequency is same as previous.'.format(
i+1, len(requested_test_freqs), freq/1e6))
continue # Already calculated this one
else:
last_source_freq = this_source_freq
this_freq_data = self.receiver.get_clean_dump(DUMP_TIMEOUT)['xeng_raw']
this_freq_response = normalised_magnitude(
this_freq_data[:, test_baseline, :])
actual_test_freqs.append(this_source_freq)
chan_responses.append(this_freq_response)
self.corr_fix.stop_x_data()
# Convert the lists to numpy arrays for easier working
actual_test_freqs = np.array(actual_test_freqs)
chan_responses = np.array(chan_responses)
def plot_and_save(freqs, data, plot_filename):
df = self.corr_freqs.delta_f
fig = plt.plot(freqs, data)[0]
axes = fig.get_axes()
ybound = axes.get_ybound()
yb_diff = abs(ybound[1] - ybound[0])
new_ybound = [ybound[0] - yb_diff*1.1, ybound[1] + yb_diff * 1.1]
plt.vlines(expected_fc, *new_ybound, colors='r', label='chan fc')
plt.vlines(expected_fc - df / 2, *new_ybound, label='chan min/max')
plt.vlines(expected_fc - 0.8*df / 2, *new_ybound, label='chan +-40%',
linestyles='dashed')
plt.vlines(expected_fc + df / 2, *new_ybound, label='_chan max')
plt.vlines(expected_fc + 0.8*df / 2, *new_ybound, label='_chan +40%',
linestyles='dashed')
plt.legend()
plt.title('Channel {} ({} MHz) response'.format(
test_chan, expected_fc/1e6))
axes.set_ybound(*new_ybound)
plt.grid(True)
plt.ylabel('dB relative to VACC max')
# TODO Normalise plot to frequency bins
plt.xlabel('Frequency (Hz)')
plt.savefig(plot_filename)
plt.close()
graph_name = '{}.{}.channel_response.svg'.format(strclass(self.__class__),
self._testMethodName)
plot_data_all = loggerise(chan_responses[:, test_chan], dynamic_range=90)
plot_and_save(actual_test_freqs, plot_data_all, graph_name)
# Get responses for central 80% of channel
df = self.corr_freqs.delta_f
central_indices = (
(actual_test_freqs <= expected_fc + 0.8*df) &
(actual_test_freqs >= expected_fc - 0.8*df))
central_chan_responses = chan_responses[central_indices]
central_chan_test_freqs = actual_test_freqs[central_indices]
# Test responses in central 80% of channel
for i, freq in enumerate(central_chan_test_freqs):
max_chan = np.argmax(np.abs(central_chan_responses[i]))
self.assertEqual(max_chan, test_chan, 'Source freq {} peak not in channel '
'{} as expected but in {}.'
.format(freq, test_chan, max_chan))
# TODO Graph the central 80% too.
self.assertLess(
np.max(np.abs(central_chan_responses[:, test_chan])), 0.99,
'VACC output at > 99% of maximum value, indicates that '
'something, somewhere, is probably overranging.')
max_central_chan_response = np.max(10*np.log10(central_chan_responses[:, test_chan]))
min_central_chan_response = np.min(10*np.log10(central_chan_responses[:, test_chan]))
chan_ripple = max_chan_response - min_chan_response
acceptable_ripple_lt = 0.3
self.assertLess(chan_ripple, acceptable_ripple_lt,
'ripple {} dB within 80% of channel fc is >= {} dB'
.format(chan_ripple, acceptable_ripple_lt))
# from matplotlib import pyplot
# colour_cycle = 'rgbyk'
# style_cycle = ['-', '--']
# linestyles = itertools.cycle(itertools.product(style_cycle, colour_cycle))
# for i, freq in enumerate(actual_test_freqs):
# style, colour = linestyles.next()
# pyplot.plot(loggerise(chan_responses[:, i], dynamic_range=60), color=colour, ls=style)
# pyplot.ion()
# pyplot.show()
# import IPython ; IPython.embed()
def test_product_baselines(self):
"""CBF Baseline Correlation Products: VR.C.19, TP.C.1.3"""
init_dsim_sources(self.dhost)
# Put some correlated noise on both outputs
self.dhost.noise_sources.noise_corr.set(scale=0.5)
test_dump = self.receiver.get_clean_dump(DUMP_TIMEOUT)
# Get list of all the correlator input labels
input_labels = sorted(tuple(test_dump['input_labelling'][:,0]))
# Get list of all the baselines present in the correlator output
present_baselines = sorted(
set(tuple(bl) for bl in test_dump['bls_ordering']))
# Make a list of all possible baselines (including redundant baselines) for the
# given list of inputs
possible_baselines = set()
for li in input_labels:
for lj in input_labels:
possible_baselines.add((li, lj))
test_bl = sorted(list(possible_baselines))
# Test that each baseline (or its reverse-order counterpart) is present in the
# correlator output
baseline_is_present = {}
for test_bl in possible_baselines:
baseline_is_present[test_bl] = (test_bl in present_baselines or
test_bl[::-1] in present_baselines)
self.assertTrue(all(baseline_is_present.values()),
"Not all baselines are present in correlator output.")
test_data = test_dump['xeng_raw']
# Expect all baselines and all channels to be non-zero
self.assertFalse(zero_baselines(test_data))
self.assertEqual(nonzero_baselines(test_data),
all_nonzero_baselines(test_data))
# Save initial f-engine equalisations
initial_equalisations = {input: eq_info['eq'] for input, eq_info
in self.correlator.feng_eq_get().items()}
def restore_initial_equalisations():
for input, eq in initial_equalisations.items():
self.correlator.feng_eq_set(source_name=input, new_eq=eq)
self.addCleanup(restore_initial_equalisations)
# Set all inputs to zero, and check that output product is all-zero
for input in input_labels:
self.correlator.feng_eq_set(source_name=input, new_eq=0)
test_data = self.receiver.get_clean_dump(DUMP_TIMEOUT)['xeng_raw']
self.assertFalse(nonzero_baselines(test_data))
#-----------------------------------
all_inputs = sorted(set(input_labels))
zero_inputs = set(input_labels)
nonzero_inputs = set()
def calc_zero_and_nonzero_baselines(nonzero_inputs):
nonzeros = set()
zeros = set()
for inp_i in all_inputs:
for inp_j in all_inputs:
if inp_i in nonzero_inputs and inp_j in nonzero_inputs:
nonzeros.add((inp_i, inp_j))
else:
zeros.add((inp_i, inp_j))
return zeros, nonzeros
#zero_baseline, nonzero_baseline = calc_zero_and_nonzero_baselines(nonzero_inputs)
def print_baselines():
print ('zeros: {}\n\nnonzeros: {}\n\nnonzero-baselines: {}\n\n '
'zero-baselines: {}\n\n'.format(
sorted(zero_inputs), sorted(nonzero_inputs),
sorted(nonzero_baseline), sorted(zero_baseline)))
#print_baselines()
for inp in input_labels:
old_eqs = initial_equalisations[inp]
self.correlator.feng_eq_set(source_name=inp, new_eq=old_eqs)
zero_inputs.remove(inp)
nonzero_inputs.add(inp)