def rescale_to_netsnr(det1_TimeSeries, det2_TimeSeries, targetsnr):
flen = len(det1_TimeSeries.to_frequencyseries())
delta_f = \
np.diff(det1_TimeSeries.to_frequencyseries().sample_frequencies)[0]
f_low=10.0
psd = aLIGOZeroDetHighPower(flen, delta_f, f_low)
sigma1sq = pycbc.filter.sigmasq(det1_TimeSeries,
low_frequency_cutoff=f_low, psd=psd)
sigma2sq = pycbc.filter.sigmasq(det2_TimeSeries,
low_frequency_cutoff=f_low, psd=psd)
snr_ratio = np.sqrt(targetsnr**2 / (sigma1sq + sigma2sq))
det1_TimeSeries.data *= snr_ratio
det2_TimeSeries.data *= snr_ratio
sigma1 = pycbc.filter.sigma(det1_TimeSeries,
low_frequency_cutoff=f_low, psd=psd)
sigma2 = pycbc.filter.sigma(det2_TimeSeries,
low_frequency_cutoff=f_low, psd=psd)
return det1_TimeSeries, det2_TimeSeries, sigma1, sigma2
def test_likelihood_evaluator_init(self):
# data args
seglen = 4
sample_rate = 2048
N = seglen * sample_rate/2 + 1
fmin = 30.
# setup waveform generator and signal parameters
m1, m2, s1z, s2z, tsig = 38.6, 29.3, 0., 0., 3.1
ra, dec, pol, dist = 1.37, -1.26, 2.76, 3*500.
generator = waveform.FDomainDetFrameGenerator(
waveform.FDomainCBCGenerator, 0., variable_args=["tc"],
detectors=["H1", "L1"], delta_f=1./seglen,
f_lower=fmin, approximant="TaylorF2",
mass1=m1, mass2=m2, spin1z=s1z, spin2z=s2z, ra=ra,
dec=dec, polarization=pol, distance=dist)
signal = generator.generate(tsig)
# get PSDs
psd = pypsd.aLIGOZeroDetHighPower(N, 1./seglen, 20.)
psds = {"H1": psd, "L1": psd}
# get a prior evaluator
uniform_prior = distributions.Uniform(tc=(tsig-0.2,tsig+0.2))
prior_eval = inference.prior.PriorEvaluator(["tc"], uniform_prior)
# setup likelihood evaluator
likelihood_eval = inference.GaussianLikelihood(generator, signal, fmin,
psds=psds, prior=prior_eval, return_meta=False)
def create_noise_curve(noise_curve, f_low, delta_f, flen):
if noise_curve=='aLIGO':
from pycbc.psd import aLIGOZeroDetHighPower
psd = aLIGOZeroDetHighPower(flen, delta_f, f_low)
elif noise_curve=='iLIGO':
from pycbc.psd import iLIGOSRD
psd = iLIGOSRD(flen, delta_f, f_low)
elif noise_curve=='eLIGO':
from pycbc.psd import eLIGOModel
psd = eLIGOModel(flen, delta_f, f_low)
elif noise_curve=='adVirgo':
from pycbc.psd import AdvVirgo
psd = AdvVirgo(flen, delta_f, f_low)
else:
print >> sys.stderr, "error: noise curve (%s) not"\
" supported"%noise_curve
sys.exit(-1)
return psd
def match_inc(inc, spin_1, mass2):
# Allow masses to vary as parameters
m1_1 = mass1
m2_1 = mass2
m1_2 = mass1
m2_2 = mass2
# Phases
phase1 = 0
phase2 = 0
# Convert to precessing coords
inc_1, s1x, s1y, s1z, s2x, s2y, s2z = SimInspiralTransformPrecessingNewInitialConditions(
inc, #theta_JN
phi_JL, #phi_JL
theta_z1, #theta1
theta_z2, #theta2
phi12, #phi12
spin_1, #chi1 - this parameter varies
spin_2, #chi2
m1_1,
m2_1,
f_low,
phiRef=0)
#This is our 'spin1=0' waveform that we match the precessing one with
inc_2, s1x_2, s1y_2, s1z_2, s2x_2, s2y_2, s2z_2 = SimInspiralTransformPrecessingNewInitialConditions(
inc, #theta_JN
phi_JL, #phi_JL
theta_z1, #theta1
theta_z2, #theta2
phi12, #phi12
0, #chi1
spin_2, #chi2
m1_2,
m2_2,
f_low,
phiRef=0)
# Generate the two waveforms to compare
hp, hc = get_td_waveform(approximant=approx1,
mass1=m1_1,
mass2=m2_1,
spin1y=s1y,
spin1x=s1x,
spin1z=s1z,
spin2y=s2y,
spin2x=s2x,
spin2z=s2z,
f_lower=f_low,
inclination=inc_1,
coa_phase=phase1,
delta_t=1.0 / sample_rate)
sp, sc = get_td_waveform(approximant=approx2,
mass1=m1_2,
mass2=m2_2,
spin1y=s1y_2,
spin1x=s1x_2,
spin1z=s1z_2,
spin2y=s2y_2,
spin2x=s2x_2,
spin2z=s2z_2,
f_lower=f_low,
inclination=inc_2,
coa_phase=phase2,
delta_t=1.0 / sample_rate)
# Resize the waveforms to the same length
tlen = max(len(sc), len(hc))
sc.resize(tlen)
hc.resize(tlen)
# Generate the aLIGO ZDHP PSD
delta_f = 1.0 / sc.duration
flen = tlen / 2 + 1
psd = aLIGOZeroDetHighPower(flen, delta_f, f_low)
# Note: This takes a while the first time as an FFT plan is generated
# subsequent calls are much faster.
m, i = match(hc, sc, psd=psd, low_frequency_cutoff=f_low)
#print 'The match is: %1.3f' % m
return m
def assign_noise_curve(self):
ligo3_curves = [
'base', 'highNN', 'highSPOT', 'highST', 'highSei', 'highloss',
'highmass', 'highpow', 'highsqz', 'lowNN', 'lowSPOT', 'lowST',
'lowSei'
]
if self.noise_curve == 'aLIGO':
from pycbc.psd import aLIGOZeroDetHighPower
self.psd = aLIGOZeroDetHighPower(self.flen, self.delta_f,
self.f_low)
elif self.noise_curve == 'adVirgo':
from pycbc.psd import AdvVirgo
self.psd = AdvVirgo(self.flen, self.delta_f, self.f_low)
elif self.noise_curve == 'Green' or self.noise_curve == 'Red':
# Load from ascii
pmnspy_path = os.getenv('PMNSPY_PREFIX')
psd_path = pmnspy_path + '/ligo3/PSD/%sPSD.txt' % self.noise_curve
psd_data = np.loadtxt(psd_path)
target_freqs = np.arange(0.0, self.flen * self.delta_f,
self.delta_f)
target_psd = np.zeros(len(target_freqs))
# Interpolate existing psd data to target frequencies
existing = \
np.concatenate(np.argwhere(
(target_freqs<=psd_data[-1,0]) *
(target_freqs>=psd_data[0,0])
))
target_psd[existing] = \
np.interp(target_freqs[existing], psd_data[:,0],
psd_data[:,1])
# Extrapolate to higher frequencies assuming f^2 for QN
fit_idx = np.concatenate(np.argwhere((psd_data[:,0]>2000)*\
(psd_data[:,0]<=psd_data[-1,0])))
p = np.polyfit(x=psd_data[fit_idx,0], \
y=psd_data[fit_idx,1], deg=2)
target_psd[existing[-1]+1:] = \
p[0]*target_freqs[existing[-1]+1:]**2 + \
p[1]*target_freqs[existing[-1]+1:] + \
p[2]
# After all that, reset everything below f_low to zero (this saves
# significant time in noise generation if we only care about high
# frequencies)
target_psd[target_freqs < self.f_low] = 0.0
# Create psd as standard frequency series object
self.psd = pycbc.types.FrequencySeries(
initial_array=target_psd, delta_f=np.diff(target_freqs)[0])
elif self.noise_curve in ligo3_curves:
# Load from ascii
pmnspy_path = os.getenv('PMNSPY_PREFIX')
psd_path = pmnspy_path + '/ligo3/PSD/BlueBird_%s-PSD_20140904.txt' % self.noise_curve
psd_data = np.loadtxt(psd_path)
target_freqs = np.arange(0.0, self.flen * self.delta_f,
self.delta_f)
target_psd = np.zeros(len(target_freqs))
# Interpolate existing psd data to target frequencies
existing = \
np.concatenate(np.argwhere(
(target_freqs<=psd_data[-1,0]) *
(target_freqs>=psd_data[0,0])
))
target_psd[existing] = \
np.interp(target_freqs[existing], psd_data[:,0],
psd_data[:,1])
# Extrapolate to higher frequencies assuming f^2 for QN
fit_idx = np.concatenate(np.argwhere((psd_data[:,0]>2000)*\
(psd_data[:,0]<=psd_data[-1,0])))
p = np.polyfit(x=psd_data[fit_idx,0], \
y=psd_data[fit_idx,1], deg=2)
target_psd[existing[-1]+1:] = \
p[0]*target_freqs[existing[-1]+1:]**2 + \
p[1]*target_freqs[existing[-1]+1:] + \
p[2]
# After all that, reset everything below f_low to zero (this saves
# significant time in noise generation if we only care about high
# frequencies)
target_psd[target_freqs < self.f_low] = 0.0
# Create psd as standard frequency series object
self.psd = pycbc.types.FrequencySeries(
initial_array=target_psd, delta_f=np.diff(target_freqs)[0])
else:
print >> sys.stderr, "error: noise curve (%s) not"\
" supported"%self.noise_curve
sys.exit(-1)
def reconstruct_freqseries(self,
freqseries,
npcs=1,
this_fpeak=None,
wfnum=None):
"""
Reconstruct the waveform in freqseries using <npcs> principal components
from the catalogue
Procedure:
1) Reconstruct the centered spectra (phase and mag) from the
beta-weighted PCs
2) Un-center the spectra (add the mean back on)
"""
#print "Analysing reconstruction with %d PCs"%npcs
if this_fpeak == None:
# Locate fpeak
# Note: we'll assume the peak we're aligning to is >2kHz. This
# avoids any low frequency stuff.
high_idx = self.sample_frequencies >= 2000
high_freq = self.sample_frequencies[high_idx]
high_spec = freqseries[high_idx]
this_fpeak = high_freq[np.argmax(abs(high_spec))]
# Get projection:
fd_projection = self.project_freqseries(freqseries)
fd_reconstruction = dict()
fd_reconstruction['fd_projection'] = fd_projection
#
# Original Waveforms
#
orimag = abs(freqseries)
oriphi = phase_of(freqseries)
orispec = orimag * np.exp(1j * oriphi)
fd_reconstruction['original_spectrum'] = unit_hrss(orispec,
delta=self.delta_f,
domain='frequency')
fd_reconstruction['sample_frequencies'] = np.copy(
self.sample_frequencies)
#
# Magnitude and phase reconstructions
#
# Initialise reconstructions
recmag = np.zeros(shape=np.shape(orimag))
recphi = np.zeros(shape=np.shape(oriphi))
# Sum contributions from PCs
for i in xrange(npcs):
recmag += \
fd_projection['magnitude_betas'][i]*\
self.pca['magnitude_pca'].components_[i,:]
recphi += \
fd_projection['phase_betas'][i]*\
self.pca['phase_pca'].components_[i,:]
#
# De-center the reconstruction
#
recmag += self.pca['magnitude_pca'].mean_
recphi += self.pca['phase_pca'].mean_
# --- Raw reconstruction quality
idx = (self.sample_frequencies>self.low_frequency_cutoff) \
* (orimag>0.01*max(orimag))
fd_reconstruction['magnitude_euclidean_raw'] = \
euclidean_distance(recmag[idx], fd_projection['magnitude_cent'][idx])
fd_reconstruction['phase_euclidean_raw'] = \
euclidean_distance(recphi[idx], fd_projection['phase_cent'][idx])
#
# Move the spectrum back to where it should be
#
recmag = shift_vec(recmag,
self.sample_frequencies,
fcenter=this_fpeak,
fpeak=self.fcenter).real
# XXX: phase_align
recphi = shift_vec(recphi,
self.sample_frequencies,
fcenter=this_fpeak,
fpeak=self.fcenter).real
fd_reconstruction['recon_mag'] = np.copy(recmag)
fd_reconstruction['recon_phi'] = np.copy(recphi)
#
# Fourier spectrum reconstructions
#
recon_spectrum = recmag * np.exp(1j * recphi)
# --- Unit norm reconstruction
fd_reconstruction['recon_spectrum'] = unit_hrss(recon_spectrum,
delta=self.delta_f,
domain='frequency')
fd_reconstruction['recon_timeseries'] = \
fd_reconstruction['recon_spectrum'].to_timeseries()
# --- Match calculations for mag/phase reconstructions
recon_spectrum = np.copy(fd_reconstruction['recon_spectrum'].data)
# --- Match calculations for full reconstructions
idx = (self.sample_frequencies>self.low_frequency_cutoff) \
* (orimag>0.01*max(orimag))
fd_reconstruction['magnitude_euclidean'] = \
euclidean_distance(recmag[idx], orimag[idx])
fd_reconstruction['phase_euclidean'] = \
euclidean_distance(recphi[idx], oriphi[idx])
# make psd
flen = len(self.sample_frequencies)
psd = aLIGOZeroDetHighPower(flen,
self.delta_f,
low_freq_cutoff=self.low_frequency_cutoff)
fd_reconstruction['match_aligo'] = \
pycbc.filter.match(fd_reconstruction['recon_spectrum'],
fd_reconstruction['original_spectrum'], psd = psd,
low_frequency_cutoff = self.low_frequency_cutoff)[0]
fd_reconstruction['match_noweight'] = \
pycbc.filter.match(fd_reconstruction['recon_spectrum'],
fd_reconstruction['original_spectrum'],
low_frequency_cutoff = self.low_frequency_cutoff)[0]
return fd_reconstruction
def generate(file_path, duration, seed=0, signal_separation=200,
signal_separation_interval=20, min_mass=1.2, max_mass=1.6,
f_lower=20, srate=4096, padding=256, tstart=0):
"""Function that generates test data with injections.
Arguments
---------
file_path : str
The path at which the data should be stored.
duration : int or float
Duration of the output file in seconds.
seed : {int, 0}, optional
A seed to use for generating injection parameters and noise.
signal_separation : {int or float, 200}, optional
The average duration between two injections.
signal_separation_interval : {int or float, 20}, optional
The duration between two signals will be signal_separation + t,
where t is drawn uniformly from the interval
[-signal_separation_interval, signal_separation_interval].
min_mass : {float, 1.2}, optional
The minimal mass at which injections will be made (in solar
masses).
max_mass : {float, 1.6}, optional
The maximum mass at which injections will be made (in solar
masses).
f_lower : {int or float, 20}, optional
Noise will be generated down to the specified frequency.
Below they will be set to zero. (The waveforms are generated
with a lower frequency cutofff of 25 Hertz)
srate : {int, 4096}, optional
The sample rate at which the data is generated.
padding : {int or float, 256}, optional
Duration in the beginning and end of the data that does not
contain any injections.
tstart : {int or float, 0}, optional
The inital time of the data.
"""
np.random.seed(seed)
size = (duration // signal_separation)
#Generate injection times
random_time_samples = int(round(float(signal_separation_interval) * float(srate)))
signal_separation_samples = int(round(float(signal_separation) * float(srate)))
time_samples = randint(signal_separation_samples - random_time_samples, signal_separation_samples + random_time_samples, size=size)
time_samples = time_samples.cumsum()
times = time_samples / float(srate)
times = times[np.where(np.logical_and(times > padding, times < duration - padding))[0]]
size = len(times)
#Generate parameters
cphase = uniform(0, np.pi*2.0, size=size)
ra = uniform(0, 2 * np.pi, size=size)
dec = np.arccos(uniform(-1., 1., size=size)) - np.pi/2
inc = np.arccos(uniform(-1., 1., size=size))
pol = uniform(0, 2 * np.pi, size=size)
dist = power(3, size) * 400
m1 = uniform(min_mass, max_mass, size=size)
m2 = uniform(min_mass, max_mass, size=size)
#Save parameters to file.
stat_file_path, ext = os.path.splitext(file_path)
stat_file_path = stat_file_path + '_stats' + ext
with h5py.File(stat_file_path, 'w') as f:
f['times'] = times
f['cphase'] = cphase
f['ra'] = ra
f['dec'] = dec
f['inc'] = inc
f['pol'] = pol
f['dist'] = dist
f['mass1'] = m1
f['mass2'] = m2
f['seed'] = seed
p = aLIGOZeroDetHighPower(2 * int(duration * srate), 1.0/64, f_lower)
#Generate noise
data = {}
for i, ifo in enumerate(['H1', 'L1']):
data[ifo] = colored_noise(p, int(tstart),
int(tstart + duration),
seed=seed + i,
low_frequency_cutoff=f_lower)
data[ifo] = resample_to_delta_t(data[ifo], 1.0/srate)
# make waveforms and add them into the noise
for i in range(len(times)):
hp, hc = get_td_waveform(approximant="TaylorF2",
mass1=m1[i],
mass2=m2[i],
f_lower=25,
delta_t=1.0/srate,
inclination=inc[i],
coa_phase=cphase[i],
distance=dist[i])
hp.start_time += times[i] + int(tstart)
hc.start_time += times[i] + int(tstart)
for ifo in ['H1', 'L1']:
ht = Detector(ifo).project_wave(hp, hc, ra[i], dec[i], pol[i])
time_diff = float(ht.start_time - data[ifo].start_time)
sample_diff = int(round(time_diff / data[ifo].delta_t))
ht.prepend_zeros(sample_diff)
ht.start_time = data[ifo].start_time
data[ifo] = data[ifo].add_into(ht)
#Save the data
for ifo in ['H1', 'L1']:
data[ifo].save(file_path, group='%s' % (ifo))
# Scale this to the smallest OBSERVED chirp mass
mtot_target = bnru.mtot_from_mchirp(smallest_observed_chirpmass,
simulations.simulations[w]['q'])
print 'scaling to ', mtot_target
hplus_NR = bnru.scale_wave(hplus_NR, mtot_target, init_total_mass)
hplus_NR = pycbc.types.TimeSeries(hplus_NR, delta_t=SI_deltaT)
Hplus_NR = hplus_NR.to_frequencyseries()
axspecSI.loglog(Hplus_NR.sample_frequencies,
2*abs(Hplus_NR)*np.sqrt(Hplus_NR.sample_frequencies), color='grey')
delta_f = Hplus_NR.delta_f
flen = len(Hplus_NR)
psd = aLIGOZeroDetHighPower(flen, delta_f, 0.1)
axspecSI.loglog(psd.sample_frequencies, np.sqrt(psd), label='aLIGO',
color='k', linestyle='--')
axspecSI.set_xlim(5, 512)
axspecSI.axvline(30, color='r')
axspecSI.minorticks_on()
axspecSI.set_xlabel('Frequency [Hz]')
axspecSI.set_ylabel('2|H$_+$($f$)|$\sqrt{f}$ & $\sqrt{S(f)}$')
pl.tight_layout()
pl.show()
sys.exit()
def __init__(self, ts, time_step=0.25, batch_size=32, dt=None):
self.batch_size = batch_size
self.time_step = time_step
if not isinstance(ts, list):
ts = [ts]
self.ts = []
self.dt = []
for t in ts:
if isinstance(t, TimeSeries):
self.dt.append(t.delta_t)
self.ts.append(t)
elif isinstance(t, type(np.array([]))):
if dt == None:
msg = 'If the provided data is not a pycbc.types.TimeSeries'
msg += 'a value dt must be provided.'
raise ValueError(msg)
else:
self.dt.append(dt)
self.ts.append(TimeSeries(t, delta_t=dt))
else:
msg = 'The provided data needs to be either a list or a '
msg += 'single instance of either a pycbc.types.TimeSeries'
msg += 'or a numpy.array.'
raise ValueError(msg)
for delta_t in self.dt:
if not delta_t == self.dt[0]:
raise ValueError('All data must have the same delta_t.')
#Number samples in each channel
self.final_data_samples = 2048
#delta_t of all TimeSeries
self.dt = self.dt[0]
#How big is the window that is shifted over the data
#(64s + 8s for cropping when whitening)
self.window_size_time = 72.0
#Window size in samples
self.window_size = int(self.window_size_time / self.dt)
#How many points are shifted each step
self.stride = int(self.time_step / self.dt)
#total number of window shifts
self.window_shifts = int(
np.floor(
float(len(self.ts[0]) - self.window_size + self.stride) /
self.stride))
#Different parts of the signal are re-sampled to different
#delta_t. This lists the target delta_t.
self.resample_dt = [
1.0 / 4096, 1.0 / 2048, 1.0 / 1024, 1.0 / 512, 1.0 / 256, 1.0 / 128
]
#The inverse of the re-sample delta_t
self.resample_rates = [4096, 4096, 2048, 1024, 512, 256, 128]
#PSD used to whiten the data. Calculate once to save
#computational resources.
self.psd = aLIGOZeroDetHighPower(self.window_size // 2 + 1,
delta_f=1. / self.window_size_time,
low_freq_cutoff=18.)
def generate_psd(**kwargs):
DELTA_F = 1.0 / kwargs['t_len']
F_LEN = int(2.0 / (DELTA_F * kwargs['delta_t']))
return (aLIGOZeroDetHighPower(length=F_LEN,
delta_f=DELTA_F,
low_freq_cutoff=kwargs['f_lower']))
def Noise(flow, delta_f, delta_t, tlen):
flen = int(1.0 / delta_t / delta_f) / 2 + 1
p_s_d = psd.aLIGOZeroDetHighPower(flen, delta_f, flow)
Nt = int(tlen / delta_t)
return noise.noise_from_psd(Nt, delta_t, p_s_d, seed=127)
test_catalogue = bwave.waveform_catalogue(test_simulations, ref_mass=total_mass,
SI_deltaT=SI_deltaT, SI_datalen=SI_datalen, distance=distance)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# Do the PCA
#
print '~~~~~~~~~~~~~~~~~~~~~'
print 'Performing PCA'
print ''
pca1 = bpca.waveform_pca(train1_catalogue, test_catalogue)
pca2 = bpca.waveform_pca(train2_catalogue, test_catalogue)
# Characterise reconstructions
# XXX build a PSD and call projection_fidelity()
psd = aLIGOZeroDetHighPower(pca1.SI_flen, pca1.SI_deltaF, pca1.fmin)
euclidean_distances1, projections1, matches1 = pca1.compute_projection_fidelity(psd=psd)
euclidean_distances2, projections2, matches2 = pca2.compute_projection_fidelity(psd=psd)
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#
# Plots
#
print '~~~~~~~~~~~~~~~~~~~~~'
print 'Plotting results'
print ''
#
# Matches
#
# Adopt the same approach as for explained variance here
fixed_parameters = injection_parameters.copy()
for key in priors:
fixed_parameters.pop(key)
# These lines generate the `model` object - see
# https://gwin.readthedocs.io/en/latest/api/gwin.models.gaussian_noise.html
generator = FDomainDetFrameGenerator(FDomainCBCGenerator,
0.,
variable_args=variable_parameters,
detectors=['H1', 'L1'],
delta_f=1. / seglen,
f_lower=fmin,
approximant='IMRPhenomPv2',
**fixed_parameters)
signal = generator.generate(**injection_parameters)
psd = pypsd.aLIGOZeroDetHighPower(int(N), 1. / seglen, 20.)
psds = {'H1': psd, 'L1': psd}
model = gwin.models.GaussianNoise(variable_parameters,
signal,
generator,
fmin,
psds=psds)
model.update(**injection_parameters)
# This create a dummy class to convert the model into a bilby.likelihood object
class GWINLikelihood(bilby.core.likelihood.Likelihood):
def __init__(self, model):
""" A likelihood to wrap around GWIN model objects
Parameters
hp, hc = pycbc.waveform.waveform.get_td_waveform(**q) # Generte GR waveform tlen = len(hp) # Create parameter tlen tlen = math.log(tlen, 2) # Take the log base 2 of tlen tlen = math.ceil(tlen) # Round up to the nearest integer (result is a float) tlen = 2.0**tlen # Raise 2 to the nearest integer to help FFT go faster tlen = 2.0 * tlen tlen = int(tlen) hp.resize(tlen) # Resize hp f_low = 20 # Lowest frequency # Generate the aLIGO ZDHP PSD delta_f = 1.0 / hp.duration # Frequency incirment flen = tlen / 2 + 1 psd = aLIGOZeroDetHighPower(flen, delta_f, f_low) # Creating PSD ######################################################### ##--------------------NON GR PARAMS--------------------## ######################################################### p = default_args_ngr p['mass1'] = 20 p['mass2'] = 30 #p['spinx1'] = 0.5 #p['spinx2'] = -0.5 p['delta_t'] = 1. / 4096 p['f_lower'] = 20 p['approximant'] = 'IMRPhenomPv2' nGR = ['dalpha4']
def whiten_data(strain_list,
low_freq_cutoff=20.,
max_filter_duration=4.,
psd=None):
"""Returns the data whitened by the PSD.
Arguments
---------
strain_list : list of pycbc.TimeSeries or pycbc.TimeSeries
The data that should be whitened.
low_freq_cutoff : {float, 20.}
The lowest frequency that is considered during calculations. It
must be >= than the lowest frequency where the PSD is not zero.
Unit: hertz
max_filter_duration : {float, 4.}
The duration to which the PSD is truncated to in the
time-domain. The amount of time is removed from both the
beginning and end of the input data to avoid wrap-around errors.
Unit: seconds
psd : {None or pycbc.FrequencySeries, None}
The PSD that should be used to whiten the data. If set to None
the pycbc.psd.aLIGOZeroDetHighPower PSD will be used. If a PSD
is provided which does not fit the delta_f of the data, it will
be interpolated to fit.
Returns
-------
list of pycbc.TimeSeries or TimeSeries
Depending on the input type it will return a list of TimeSeries
or a single TimeSeries. The data contained in this time series
is the whitened input data, where the inital and final seconds
as specified by max_filter_duration are removed.
"""
org_type = type(strain_list)
if not org_type == list:
strain_list = [strain_list]
ret = []
for strain in strain_list:
df = strain.delta_f
f_len = int(len(strain) / 2) + 1
if psd is None:
psd = aLIGOZeroDetHighPower(length=f_len,
delta_f=df,
low_freq_cutoff=low_freq_cutoff - 2.)
else:
if not len(psd) == f_len:
msg = 'Length of PSD does not match data.'
raise ValueError(msg)
elif not psd.delta_f == df:
psd = interpolate(psd, df)
max_filter_len = int(
max_filter_duration *
strain.sample_rate) #Cut out the beginning and end
psd = inverse_spectrum_truncation(psd,
max_filter_len=max_filter_len,
low_frequency_cutoff=low_freq_cutoff,
trunc_method='hann')
f_strain = strain.to_frequencyseries()
kmin = int(low_freq_cutoff / df)
f_strain.data[:kmin] = 0
f_strain.data[-1] = 0
f_strain.data[kmin:] /= psd[kmin:]**0.5
strain = f_strain.to_timeseries()
ret.append(strain[max_filter_len:len(strain) - max_filter_len])
if not org_type == list:
return (ret[0])
else:
return (ret)
spin2x=s2x,
spin2y=s2y,
delta_t=delta_t,
distance=distance,
f_lower=f_lower_he,
inclination=inclination)
### multiply by stupid pycbc factor
hep = hep * pycbc_factor
hec = hec * pycbc_factor
#Generate the aLIGO ZDHP PSD
hep.resize(300000)
delta_f = 1.0 / hep.duration
t_len = len(hep)
f_len = t_len / 2 + 1
if psd_type == 'aLIGOZeroDetHighPower':
psd = aLIGOZeroDetHighPower(f_len, delta_f, f_low)
elif psd_type != 'aLIGOZeroDetHighPower':
psd = pycbc.psd.read.from_txt(psd_path,
f_len,
delta_f,
f_low,
is_asd_file=True)
else:
sys.exit('ERROR: enter vaild psd type')
sigma_sqr_he = matchedfilter.sigmasq(hep,
psd=psd,
low_frequency_cutoff=f_low,
high_frequency_cutoff=f_high)
print sigma_sqr_he
he_l2_norm = np.linalg.norm(np.array(hep))
func_compare_EOS1 = partial(compare, psd, psd_path, hep, hec, distance,
# Now make a catalogue at 350 solar masses and then compute the overlap
#
catalogue350 = bhex.waveform_catalogue(catalogue_name=catalogue_name, fs=2048,
catalogue_len=catlen, mtotal_ref=350, Dist=1., theta=theta)
oriwave350 = np.copy(catalogue350.aligned_catalogue[0,:])
# Finally, compute the match between the reconstructed 350 Msun system and the
# system we generated at that mass in the first place
recwave350_pycbc = pycbc.types.TimeSeries(np.real(recwave350), delta_t=1./2048)
oriwave250_pycbc = pycbc.types.TimeSeries(np.real(oriwave250), delta_t=1./2048)
oriwave350_pycbc = pycbc.types.TimeSeries(np.real(oriwave350), delta_t=1./2048)
psd = aLIGOZeroDetHighPower(len(recwave350_pycbc.to_frequencyseries()),
recwave350_pycbc.to_frequencyseries().delta_f, low_freq_cutoff=10.0)
match_cat = pycbc.filter.match(oriwave250_pycbc.to_frequencyseries(),
oriwave350_pycbc.to_frequencyseries(), psd=psd,
low_frequency_cutoff=10)[0]
match_rec = pycbc.filter.match(recwave350_pycbc.to_frequencyseries(),
oriwave350_pycbc.to_frequencyseries(), psd=psd,
low_frequency_cutoff=10)[0]
print 'Match between 250 and 350 Msun catalogue waves: ', match_cat
print 'Match between 350 reconstruction and 350 catalogue wave: ', match_rec
#
def Noise(flow,delta_f,delta_t): flen = int(2048/delta_f) + 1 p_s_d = psd.aLIGOZeroDetHighPower(flen, delta_f, flow) tlen = int(2048/delta_t) return noise.noise_from_psd(tlen, delta_t, p_s_d, seed=127)
f_low = 30
sample_rate = 4096
# Generate the two waveforms to compare
hp, hc = get_td_waveform(approximant="EOBNRv2",
mass1=10,
mass2=10,
f_lower=f_low,
delta_t=1.0/sample_rate)
sp, sc = get_td_waveform(approximant="TaylorT4",
mass1=10,
mass2=10,
f_lower=f_low,
delta_t=1.0/sample_rate)
# Resize the waveforms to the same length
tlen = max(len(sp), len(hp))
sp.resize(tlen)
hp.resize(tlen)
# Generate the aLIGO ZDHP PSD
delta_f = 1.0 / sp.duration
flen = tlen/2 + 1
psd = aLIGOZeroDetHighPower(flen, delta_f, f_low)
# Note: This takes a while the first time as an FFT plan is generated
# subsequent calls are much faster.
m, i = match(hp, sp, psd=psd, low_frequency_cutoff=f_low)
print 'The match is: %1.3f' % m
def reconstructed_SineGaussianF(posterior, waveform, flow=1000, fupp=4096,
nrec=100):
"""
return the reconstructed F-domain sine-Gaussians from the posterior samples
in posterior, as well as the max-posterior reconstruction and the matches
with the target waveform
"""
wlen=16384
# Get a zero-padded version of the target waveform
htarget=np.zeros(wlen)
htarget[0:len(waveform.hplus)]=waveform.hplus.data
htarget=pycbc.types.TimeSeries(htarget, delta_t = waveform.hplus.delta_t)
# Normalise so that the target has hrss=1
#hrssTarget = pycbc.filter.sigma(htarget, low_frequency_cutoff=flow,
# high_frequency_cutoff=fupp)
#htarget.data /= hrssTarget
# Get frequency series of target waveform
H_target = htarget.to_frequencyseries()
# Make psd for matches
flen = len(H_target)
delta_f = np.diff(H_target.sample_frequencies)[0]
psd = aLIGOZeroDetHighPower(flen, H_target.delta_f, low_freq_cutoff=flow)
# -----------
# MAP waveform
# XXX: Time-domain is a little easier, since we don't have to figure out
# which frequencies to populate in a pycbc object
hp, _ = lalsim.SimBurstSineGaussian(posterior.maxP[1]['quality'],
posterior.maxP[1]['frequency'], posterior.maxP[1]['hrss'], 0.0, 0.0,
waveform.hplus.delta_t)
#hp, _ = lalsim.SimBurstSineGaussian(posterior.maxP[1]['quality'],
# posterior.maxP[1]['frequency'], 1.0, 0.0, 0.0,
# waveform.hplus.delta_t)
# zero-pad
h_MAP = np.zeros(wlen)
# populate
h_MAP[:hp.data.length]=hp.data.data
# pycbc objects
h_MAP_ts = pycbc.types.TimeSeries(h_MAP,waveform.hplus.delta_t)
H_MAP = h_MAP_ts.to_frequencyseries()
MAP_match = pycbc.filter.match(H_target, H_MAP, low_frequency_cutoff=flow,
high_frequency_cutoff=fupp)[0]
# -------------------------
# Waveforms for all samples
# Pre-allocate:
#nrec=500
if len(posterior['frequency'].samples)>nrec:
decidx = np.random.randint(0, len(posterior['frequency'].samples),
nrec)
else:
decidx = xrange(nrec)
reconstructions = np.zeros(shape=(nrec, len(H_target)), dtype=complex)
matches = np.zeros(nrec)
# All samples!
for idx in xrange(nrec):
# let's just use hplus for now...
hp, _ = lalsim.SimBurstSineGaussian(
np.squeeze(posterior['quality'].samples)[idx],
np.squeeze(posterior['frequency'].samples)[idx],
np.squeeze(posterior['hrss'].samples)[idx], 0.0, 0.0,
waveform.hplus.delta_t)
#hp, _ = lalsim.SimBurstSineGaussian(
# np.squeeze(posterior['quality'].samples)[idx],
# np.squeeze(posterior['frequency'].samples)[idx],
# 1.0, 0.0, 0.0,
# waveform.hplus.delta_t)
# zero pad
hcurrent = np.zeros(wlen)
# populate
hcurrent[:hp.data.length] = hp.data.data
# pycbc object
hcurrent_ts = pycbc.types.TimeSeries(hcurrent, delta_t=hp.deltaT)
# populate array of all reconstructions
reconstructions[idx, :] = hcurrent_ts.to_frequencyseries().data
# compute match for this sample
matches[idx] = pycbc.filter.match(H_target,
hcurrent_ts.to_frequencyseries(), low_frequency_cutoff=flow,
high_frequency_cutoff=fupp)[0]
# -----------
reconstruction = {}
reconstruction['FrequencyAxis'] = H_MAP.sample_frequencies
reconstruction['TargetSpectrum'] = H_target
reconstruction['MAPSpectrum'] = H_MAP
reconstruction['MAPMatch'] = MAP_match
reconstruction['SampledReconstructions'] = reconstructions
reconstruction['SampledMatches'] = matches
return reconstruction
