def simulate_AETnoise_from_arbitrary_SpH(duration, stime, t0, GWSpectralSlope,
H0, Ppath, lmax, seed,
N_previous_draws, Nvar,
compute_ORF_SpHs, *orfpaths):
"""
Simulates noise(s) in the freq-domain from SpHs of ORF and SkyMap.
Partly hardwired to simulate for A,E and T.
tdiORF_SpHs' IJs are set from AET explicitly, so they can be used for analysis too.
INPUT:
duration -- duration of the time-series
stime -- sampling time
t0 --- initial time
GWSpectralSlope --- GW background spectral slope
H0 --- normalisation constant for H(f)
Ppath --- file path to SkyMap of P_{lm}s
lmax --- maximum degree l in SpH considered
seed --- seed for the random number generator: 'random' or a positive integer
N_previous_draws --- numuber of random number draws to discard first
Nvar --- number of random variables (or time-series)
compute_ORF_SpHs --- True (to compute tdiORF_SpHs) or False (to load from disk)
*orfpaths --- file paths to save or to look for tdiORF_SpHs
OUTPUT:
t --- time labels of the time-series (N x 1 numpy array)
n --- time-series (Nvar x N numpy array)
"""
Nvar = 3 #3 variables: A, E and T
IJs = [
'AA', 'AE', 'AT', 'EE', 'ET', 'TT'
] #Explicitly label tdiORF_SpHs (Not really needed here, but this way they can be used for the analysis pipeline.)
freqdict = mufls.get_freqs_from_duration_and_stime(stime, duration)
if compute_ORF_SpHs:
Nindies = int(Nvar * (Nvar + 1) / 2)
if len(orfpaths) < Nindies:
raise InputError, 'You must enter at least %d orfpaths!' % Nindies
[
make_arbitrary_tdiORF_SpHs(orfpaths[k],
freqdict['f'],
IJ=IJs[k],
lmax=lmax) for k in range(Nindies)
]
f, comatrix = mufls.get_CovarMatrix(Nvar, Ppath, GWSpectralSlope, H0,
*orfpaths)
else:
f, comatrix = mufls.get_CovarMatrix(Nvar, Ppath, GWSpectralSlope, H0,
*orfpaths)
if not np.allclose(f, freqdict['f']):
raise InputError, 'Frequencies of loaded covariance matrix do not match those of the fft. Try computing the tdiORF_SpHs first with compute_ORF_SpHs=True'
t, n = mufls.get_noise_freq_domain_CovarMatrix(
comatrix,
freqdict['df'],
t0,
freqdict['parityN'],
seed=seed,
N_previous_draws=N_previous_draws)
return t, n
def simulate_AETnoise_from_arbitrary_SpH( duration , stime , t0 ,
GWSpectralSlope, H0,
Ppath , lmax ,
seed , N_previous_draws ,
Nvar , compute_ORF_SpHs ,
*orfpaths ) :
"""
Simulates noise(s) in the freq-domain from SpHs of ORF and SkyMap.
Partly hardwired to simulate for A,E and T.
tdiORF_SpHs' IJs are set from AET explicitly, so they can be used for analysis too.
INPUT:
duration -- duration of the time-series
stime -- sampling time
t0 --- initial time
GWSpectralSlope --- GW background spectral slope
H0 --- normalisation constant for H(f)
Ppath --- file path to SkyMap of P_{lm}s
lmax --- maximum degree l in SpH considered
seed --- seed for the random number generator: 'random' or a positive integer
N_previous_draws --- numuber of random number draws to discard first
Nvar --- number of random variables (or time-series)
compute_ORF_SpHs --- True (to compute tdiORF_SpHs) or False (to load from disk)
*orfpaths --- file paths to save or to look for tdiORF_SpHs
OUTPUT:
t --- time labels of the time-series (N x 1 numpy array)
n --- time-series (Nvar x N numpy array)
"""
Nvar = 3 #3 variables: A, E and T
IJs = [ 'AA' , 'AE' , 'AT' , 'EE' , 'ET' , 'TT' ] #Explicitly label tdiORF_SpHs (Not really needed here, but this way they can be used for the analysis pipeline.)
freqdict = mufls.get_freqs_from_duration_and_stime( stime , duration )
if compute_ORF_SpHs :
Nindies = int( Nvar*(Nvar+1) / 2 )
if len( orfpaths ) < Nindies :
raise InputError , 'You must enter at least %d orfpaths!' % Nindies
[ make_arbitrary_tdiORF_SpHs( orfpaths[k] , freqdict['f'] , IJ=IJs[k] , lmax=lmax )
for k in range( Nindies ) ]
f , comatrix = mufls.get_CovarMatrix( Nvar , Ppath , GWSpectralSlope, H0 , *orfpaths )
else :
f , comatrix = mufls.get_CovarMatrix( Nvar , Ppath , GWSpectralSlope, H0 , *orfpaths )
if not np.allclose( f , freqdict['f'] ) :
raise InputError , 'Frequencies of loaded covariance matrix do not match those of the fft. Try computing the tdiORF_SpHs first with compute_ORF_SpHs=True'
t , n = mufls.get_noise_freq_domain_CovarMatrix( comatrix , freqdict['df'] , t0 ,
freqdict['parityN'] ,
seed=seed ,
N_previous_draws=N_previous_draws )
return t , n
comatrix = get_comatrix( freqdict['f'] )
for day in options.days :
print 'Day %d' % day
t0 = ( day - 1 )*dayinsecs
t = t0 + options.stime * np.arange( N )
Npd_before_today = (day - 1) * Nseg * 2 * Nvar * freqdict['Nf']
s = 1 ; tails = [ None ]*Nvar ; TSs = [ [] for v in range( Nvar ) ]
while s <= Nseg :
print 'Segments (%d|%d)' % ( s , s+1 )
Npd_before_segl = Npd_before_today + (s-1)*2*Nvar*freqdict['Nf']
Npd_before_segr = Npd_before_today + s*2*Nvar*freqdict['Nf']
tl , tsl = mufls.get_noise_freq_domain_CovarMatrix(comatrix,freqdict['df'],
t0,freqdict['parityN'],
options.seed,
Npd_before_segl )
tr , tsr = mufls.get_noise_freq_domain_CovarMatrix(comatrix,freqdict['df'],
t0,freqdict['parityN'],
options.seed,
Npd_before_segr )
for v in range( Nvar ) :
print 'v = ' , v
if s == 1 :
print 'tails[v]' , tails[v]
else :
print 'tails[v][:3]' , tails[v][:3]
ts , tail = mufls.window_and_join( tsl[v] , tsr[v] , tails[v] )
TSs[ v ] += list( ts ) ; tails[ v ] = np.copy( tail )
print 'tails[v][:3]' , tails[v][:3]
( options , args ) = parser.parse_args()
if len( args ) < 1 :
parser.error( "You must specify a TSPATH! Type './x_simulate_defined_stationary_noise.py' " )
else :
tspath = args[ 0 ] ; tsdir = os.path.dirname( tspath )
stime , duration , inittime , N_previous_draws = float(options.stime) , float(options.duration) , float(options.inittime) , int( options.N_previous_draws )
N = np.round( duration / stime )
if N % 2 == 0 :
Nf = int( N/2 - 1 ) ; parityN = 'Even'
else :
Nf = int( ( N-1 ) / 2 ) ; parityN = 'Odd'
df = 1 / (N*stime)
f = df * np.arange( 1 , Nf + 1 )
comatrix = get_comatrix( f )
t , n = mufls.get_noise_freq_domain_CovarMatrix( comatrix , df , inittime , parityN , options.seed , N_previous_draws )
tscale = { 'Cadence1':stime , 'Offset1':inittime }
tsdict = { 't':AS.Coarsable( t , **tscale ) ,
'1':AS.Coarsable( n[0] , **tscale ) , '2':AS.Coarsable( n[1] , **tscale ) , '3':AS.Coarsable( n[2] , **tscale ) }
if tsdir not in glob.glob( tsdir ) :
os.system( 'mkdir %s' % tsdir )
print 'saving time-series to disk...'
file = open( tspath , 'wb' ) ; cpkl.dump( tsdict , file , -1 ) ; file.close()
for day in options.days:
print 'Day %d' % day
t0 = (day - 1) * dayinsecs
t = t0 + options.stime * np.arange(N)
Npd_before_today = (day - 1) * Nseg * 2 * Nvar * freqdict['Nf']
s = 1
tails = [None] * Nvar
TSs = [[] for v in range(Nvar)]
while s <= Nseg:
print 'Segments (%d|%d)' % (s, s + 1)
Npd_before_segl = Npd_before_today + (s -
1) * 2 * Nvar * freqdict['Nf']
Npd_before_segr = Npd_before_today + s * 2 * Nvar * freqdict['Nf']
tl, tsl = mufls.get_noise_freq_domain_CovarMatrix(
comatrix, freqdict['df'], t0, freqdict['parityN'], options.seed,
Npd_before_segl)
tr, tsr = mufls.get_noise_freq_domain_CovarMatrix(
comatrix, freqdict['df'], t0, freqdict['parityN'], options.seed,
Npd_before_segr)
for v in range(Nvar):
print 'v = ', v
if s == 1:
print 'tails[v]', tails[v]
else:
print 'tails[v][:3]', tails[v][:3]
ts, tail = mufls.window_and_join(tsl[v], tsr[v], tails[v])
TSs[v] += list(ts)
tails[v] = np.copy(tail)
print 'tails[v][:3]', tails[v][:3]
