def eval_scaled_dust_dbmmb_map(model, nu_ref, nu_test, beta0, beta1, nubreak,
nside, fsky, radec_center, T):
#def double-beta dust model
if model == 'd0':
print('one beta model')
analytic_expr = ('(exp(nu0 / temp * h_over_k) -1)'
'/ (exp(nu / temp * h_over_k) - 1)'
'* (nu / nu0)**(1 + beta_d)')
dust = AnalyticComponent(analytic_expr,
nu0=nu_ref,
h_over_k=constants.h * 1e9 / constants.k,
temp=T)
#print(dust)
scaling_factor = dust.eval(nu_test, beta0)
else:
print('double beta model')
analytic_expr = double_beta_dust_FGB_Model(units='K_CMB')
dbdust = AnalyticComponent(analytic_expr,
nu0=nu_ref,
h_over_k=constants.h * 1e9 / constants.k,
temp=T)
scaling_factor = dbdust.eval(nu_test, beta0, beta1, nubreak)
sky = pysm3.Sky(nside=nside, preset_strings=['d0'])
dust_map_ref = np.zeros(
(3, 12 * nside**2)) #this way the output is w/0 units!!
dust_map_ref[0:3, :] = sky.get_emission(
nu_ref * u.GHz, None) * utils.bandpass_unit_conversion(
nu_ref * u.GHz, None, u.uK_CMB)
map_test = dust_map_ref * scaling_factor
#mask = s4bi.get_coverage(fsky, nside, center_radec=radec_center)
return map_test
def compute_fg(self, dust_model, presets_dust=None, NSIDE_PATCH=4):
#print("\nYou're computing thermal dust with model {}\n".format(dust_model))
dustmaps = np.zeros((self.nfreqs, 3, self.npix))
if dust_model == 'd0':
sync_model = 's0'
else:
sync_model = 's1'
settings = [dust_model, sync_model]
self.preset_fg(dustmodel=dust_model, dict_dust=presets_dust)
sky = pysm3.Sky(nside=self.nside, preset_strings=settings)
if dust_model == 'd0' or dust_model == 'd1' or dust_model == 'd2' or dust_model == 'd3' or dust_model == 'd6':
sky.components[0].mbb_temperature = 20 * sky.components[
0].mbb_temperature.unit # fix temp at 20 K across the sky
#print('Downgrade pixelization of spectral indices at nside = {}'.format(NSIDE_PATCH))
if dust_model != 'd0':
#print('Downgrade pixelization of spectral indices at nside = {}'.format(NSIDE_PATCH))
for spectral_param in [
sky.components[0].mbb_index, sky.components[1].pl_index
]:
spectral_param[:] = hp.ud_grade(
hp.ud_grade(spectral_param.value, NSIDE_PATCH),
self.nside) * spectral_param.unit
for j in range(self.nfreqs):
dustmaps[j] = np.array(
sky.get_emission(self.nus[j] * u.GHz) *
utils.bandpass_unit_conversion(self.nus[j] * u.GHz, None,
u.uK_CMB))
return dustmaps
def get_fg_notconvolved(model, nu, nside=256):
sky = pysm3.Sky(nside=nside, preset_strings=[model])
maps = np.zeros(((len(nu), 3, 12 * nside**2)))
for indi, i in enumerate(nu):
maps[indi] = sky.get_emission(i * u.GHz,
None) * utils.bandpass_unit_conversion(
i * u.GHz, None, u.uK_CMB)
return maps
def get_dust_sky(nus, model, nside):
sky = pysm3.Sky(nside=nside, preset_strings=[model])
maps_dust = np.zeros(((len(nus), 3, 12 * nside**2)))
for i_freq, freq in enumerate(nus):
maps_XXXGHz = sky.get_emission(freq * u.GHz,
None) * utils.bandpass_unit_conversion(
freq * u.GHz, None, u.uK_CMB)
maps_dust[i_freq] = maps_XXXGHz.copy()
return maps_dust
def get_simple_sky_map(self):
"""
Create as many skies as the number of input frequencies.
Instrumental effects are not considered.
Return a vector of shape (number_of_input_subfrequencies, npix, 3)
"""
_, nus_edge, nus_in, _, _, Nbbands_in = qubic.compute_freq(self.filter_nu, self.Nfin,
self.filter_relative_bandwidth)
sky = np.zeros((self.Nfin, self.npix, 3))
##### This is the old code by JCH - It has been replaced by what Edgar Jaber has proposed
##### see his presentation on Git: qubic/scripts/ComponentSeparation/InternshipJaber/teleconf_03122020.pdf
# for i in range(Nf):
# # ###################### This is puzzling part here: ############################
# # # See Issue on PySM Git: https://github.com/healpy/pysm/issues/49
# # ###############################################################################
# # # #### THIS IS WHAT WOULD MAKE SENSE BUT DOES NOT WORK ~ 5% on maps w.r.t. input
# # # nfreqinteg = 5
# # # nus = np.linspace(nus_edge[i], nus_edge[i + 1], nfreqinteg)
# # # freqs = utils.check_freq_input(nus)
# # # convert_to_uK_RJ = (np.ones(len(freqs), dtype=np.double) * u.uK_CMB).to_value(
# # # u.uK_RJ, equivalencies=u.cmb_equivalencies(freqs))
# # # #print('Convert_to_uK_RJ :',convert_to_uK_RJ)
# # # weights = np.ones(nfreqinteg) * convert_to_uK_RJ
# # ###############################################################################
# # ###### Works OK but not clear why...
# # ###############################################################################
# # nfreqinteg = 5
# # nus = np.linspace(nus_edge[i], nus_edge[i + 1], nfreqinteg)
# # filter_uK_CMB = np.ones(len(nus), dtype=np.double)
# # filter_uK_CMB_normalized = utils.normalize_weights(nus, filter_uK_CMB)
# # weights = 1. / filter_uK_CMB_normalized
# # ###############################################################################
# # ### Integrate through band using filter shape defined in weights
# # themaps_iqu = self.sky.get_emission(nus * u.GHz, weights=weights)
# # sky[i, :, :] = np.array(themaps_iqu.to(u.uK_CMB, equivalencies=u.cmb_equivalencies(nus_in[i] * u.GHz))).T
# # ratio = np.mean(self.input_cmb_maps[0,:]/sky[i,:,0])
# # print('Ratio to initial: ',ratio)
# #### Here is the new code from Edgar Jaber
for i in range(self.Nfin):
nfreqinteg = 5
freqs = np.linspace(nus_edge[i], nus_edge[i + 1], nfreqinteg)
weights = np.ones(nfreqinteg)
sky[i, :, :] = (
self.sky.get_emission(freqs * u.GHz, weights) * utils.bandpass_unit_conversion(freqs * u.GHz,
weights,
u.uK_CMB)).T
return sky
def give_me_maps_d1_modified(nus,
nubreak,
covmap,
delta_b,
nside,
fix_temp=None,
nside_index=256):
maps_dust = np.ones(((len(nus), 3, 12 * nside**2))) * hp.UNSEEN
ind = np.where(covmap > 0)[0]
sky = pysm3.Sky(nside=nside, preset_strings=['d1'])
maps_dust = sky.get_emission(353 * u.GHz,
None) * utils.bandpass_unit_conversion(
353 * u.GHz, None, u.uK_CMB)
map_index = np.array(sky.components[0].mbb_index)
if fix_temp is not None:
sky.components[0].mbb_temperature = fix_temp
map_temperature = np.array(
np.ones(12 * nside**2) * sky.components[0].mbb_temperature)
else:
map_temperature = np.array(sky.components[0].mbb_temperature)
if nside_index != 256:
map_temperature = hp.pixelfunc.ud_grade(map_temperature, nside_index)
map_index = hp.pixelfunc.ud_grade(map_index, nside_index)
map_temperature = hp.pixelfunc.ud_grade(map_temperature, 256)
map_index = hp.pixelfunc.ud_grade(map_index, 256)
#hp.mollview(map_temperature, sub=(1, 2, 1))
#hp.mollview(map_index, sub=(1, 2, 2))
#print(map_index.shape)
# Evaluation of Mixing Matrix for 2 beta model
comp2b = [fgbuster.component_model.Dust_2b(nu0=353)]
A2b = fgbuster.MixingMatrix(*comp2b)
A2b_ev = A2b.evaluator(nus)
new_dust_map = np.ones(((len(nus), 3, 12 * nside**2))) * hp.UNSEEN
for i in ind:
A2b_maxL = A2b_ev([
np.array(map_index)[i] - delta_b,
np.array(map_index)[i] + delta_b, nubreak,
np.array(map_temperature)[i]
])
for j in range(len(nus)):
new_dust_map[j, :, i] = A2b_maxL[j, 0] * maps_dust[:, i]
return new_dust_map, [map_index, map_temperature]
def eval_scaled_sync_map(nu_ref, nu_test, betapl, nside, fsky, radec_center):
#def double-beta dust model
analytic_expr = '(nu / nu0)**(beta_pl)'
sync = AnalyticComponent(analytic_expr, nu0=nu_ref)
scaling_factor = sync.eval(nu_test, betapl)
sky = pysm3.Sky(nside=nside, preset_strings=['s0'])
sync_map_ref = np.zeros(
(3, 12 * nside**2)) #this way the output is w/0 units!!
sync_map_ref[0:3, :] = sky.get_emission(
nu_ref * u.GHz, None) * utils.bandpass_unit_conversion(
nu_ref * u.GHz, None, u.uK_CMB)
map_test = sync_map_ref * scaling_factor
#mask = s4bi.get_coverage(fsky, nside, center_radec=radec_center)
return map_test
def create_dustd0(nside, nus):
# Create 353 GHz dust maps
sky = pysm3.Sky(nside=nside, preset_strings=['d0'])
#comp=[fgbuster.component_model.Dust(nu0=353, beta_d=1.54, temp=20)]
#A = fgbuster.MixingMatrix(*comp)
#A_ev = A.evaluator(nus)
#A_maxL = A_ev()
new_dust_map = np.zeros(((len(nus), 3, 12 * nside**2)))
for i in range(len(nus)):
maps_XXXGHz = sky.get_emission(nus[i] * u.GHz,
None) * utils.bandpass_unit_conversion(
nus[i] * u.GHz, None, u.uK_CMB)
new_dust_map[i] = maps_XXXGHz.copy()
return new_dust_map
def create_sync(nside, nus):
# Create 353 GHz dust maps
sky = pysm3.Sky(nside=nside, preset_strings=['s0'])
maps_70GHz = sky.get_emission(70 * u.GHz,
None) * utils.bandpass_unit_conversion(
70 * u.GHz, None, u.uK_CMB)
comp = [fgbuster.component_model.Synchrotron(nu0=70)]
A = fgbuster.MixingMatrix(*comp)
A_ev = A.evaluator(nus)
A_maxL = A_ev(np.array([-3]))
new_sync_map = np.zeros(((len(nus), 3, 12 * nside**2)))
for i in range(len(nus)):
new_sync_map[i] = A_maxL[i, 0] * maps_70GHz
return new_sync_map
def getskymaps(self,
same_resol=None,
verbose=False,
coverage=None,
iib=1,
noise=False,
signoise=1.,
beta=[],
fix_temp=None,
nside_index=256):
"""
"""
sky = self.get_sky(coverage)
allmaps = np.zeros(((len(self.nus), 3, self.npix)))
pixok = coverage > 0
if same_resol is not None:
self.fwhmdeg = np.ones(len(self.nus)) * np.max(same_resol)
for i in self.skyconfig.keys():
if i == 'cmb':
cmbmap = self.get_cmb(coverage)
for j in range(len(self.nus)):
allmaps[j] += cmbmap
map_index = []
elif i == 'dust':
dustmaps = np.zeros(((len(self.nus), 3, self.npix)))
if self.skyconfig[i] == 'd0':
if verbose:
print('Model : {}'.format(self.skyconfig[i]))
for i in range(len(self.nus)):
print(
'Integration into bands ({:.0f} bands) -> from {:.2f} to {:.2f} GHz'
.format(iib, self.edges[i][0], self.edges[i][1]))
#print(self.edges[i])
nus_inter = get_freqs_inter(self.edges[i], iib)
#print(nus_inter)
dustmaps_inter = create_dustd0(
self.nside, nus_inter
) #get_fg_notconvolved('d0', self.nus, nside=self.nside)
mean_dust_maps = np.mean(dustmaps_inter, axis=0)
dustmaps[i] = mean_dust_maps.copy()
allmaps += dustmaps
map_index = []
elif self.skyconfig[i] == 'd02b':
if verbose:
print(
'Model : d02b -> Twos spectral index beta ({:.2f} and {:.2f}) with nu_break = {:.2f}'
.format(beta[0], beta[1], beta[2]))
# Create 353 GHz dust maps
sky = pysm3.Sky(nside=self.nside, preset_strings=['d0'])
maps_353GHz = sky.get_emission(
353 * u.GHz, None) * utils.bandpass_unit_conversion(
353 * u.GHz, None, u.uK_CMB)
for i in range(len(self.nus)):
print(
'Integration into bands ({:.0f} bands) -> from {:.2f} to {:.2f} GHz'
.format(iib, self.edges[i][0], self.edges[i][1]))
#print(self.edges[i])
nus_inter = get_freqs_inter(self.edges[i], iib)
#print(nus_inter)
#print(nus_inter)
#add Elenia's definition
dustmaps_inter = create_dust_with_model2beta(
maps_353GHz,
self.nside,
nus_inter,
beta[0],
beta[1],
beta[2],
temp=20,
break_width=beta[3])
#print(dustmaps_inter.shape)
mean_dust_maps = np.mean(dustmaps_inter, axis=0)
dustmaps[i] = mean_dust_maps.copy()
allmaps += dustmaps
map_index = []
elif self.skyconfig[i] == 'd1' or self.skyconfig[
i] == 'd2' or self.skyconfig[
i] == 'd3' or self.skyconfig[i] == 'd4':
skyconf = self.skyconfig[i]
#if verbose:
print('Model : {}'.format(skyconf))
if iib == 1:
dustmaps = get_dust_sky(self.nus, skyconf, self.nside)
else:
dustmaps = np.zeros(((len(self.nus), 3, self.npix)))
for i in range(len(self.nus)):
nus_inter = get_freqs_inter(self.edges[i], iib)
print(nus_inter)
dustmaps_inter = get_dust_sky(
nus_inter, skyconf, self.nside)
dust_maps_i = np.mean(dustmaps_inter, axis=0)
dustmaps[i] = dust_maps_i.copy()
allmaps += dustmaps
else:
print('No dust')
elif i == 'synchrotron':
syncmaps = np.zeros(((len(self.nus), 3, self.npix)))
print('Model : {}'.format(self.skyconfig[i]))
for i in range(len(self.nus)):
nus_inter = get_freqs_inter(self.edges[i], iib)
#print(nus_inter)
sync_maps_inter = create_sync(self.nside, nus_inter)
mean_sync_maps = np.mean(sync_maps_inter, axis=0)
syncmaps[i] = mean_sync_maps.copy()
allmaps += syncmaps
map_index = []
else:
print('No more components')
#pass
#hp.mollview(allmaps[0, 1])
if same_resol != 0:
for j in range(len(self.fwhmdeg)):
if verbose:
print('Convolution to {:.2f} deg'.format(self.fwhmdeg[j]))
allmaps[j] = hp.sphtfunc.smoothing(allmaps[j, :, :],
fwhm=np.deg2rad(
self.fwhmdeg[j]),
verbose=False)
if noise:
noisemaps = create_noisemaps(signoise, self.nus, self.nside,
self.depth_i, self.depth_p, self.npix)
maps_noisy = allmaps + noisemaps
if coverage is not None:
pixok = coverage > 0
maps_noisy[:, :, ~pixok] = hp.UNSEEN
noisemaps[:, :, ~pixok] = hp.UNSEEN
allmaps[:, :, ~pixok] = hp.UNSEEN
return maps_noisy, allmaps, noisemaps
return allmaps
def get_partial_sky_maps_withnoise(self, coverage=None, version_FastSim='01', sigma_sec=None,
Nyears=4., FWHMdeg=None, seed=None,
noise_profile=True, spatial_noise=True, nunu_correlation=True,
noise_only=False, integrate_into_band=True,
verbose=False):
"""
This returns maps in the same way as with get_simple_sky_map but cut according to the coverage
and with noise added according to this coverage and the RMS in muK.sqrt(sec) given by sigma_sec
The default integration time is 4 years but can be modified with optional variable Nyears
Note that the maps are convolved with the instrument beam by default, or with FWHMdeg (can be an array)
if provided.
If seed is provided, it will be used for the noise realization. If not it will be a new realization at
each call.
The optional effective_variance_invcov keyword is a modification law to be applied to the coverage in order to obtain
more realistic noise profile. It is a law for effective RMS as a function of inverse coverage and is 2D array
with the first one being (nx samples) inverse coverage and the second being the corresponding effective variance to be
used through interpolation when generating the noise.
Parameters
----------
coverage: array
Coverage map of the sky.
By default, we load a coverage centered on the galactic center with 10000 pointings.
version_FastSim: str
Version of the FastSimulator files: 01, 02, 03... For now, only 01 exists.
sigma_sec: float
Nyears: float
Integration time for observation to scale the noise, by default it is 4.
FWHMdeg:
seed:
noise_profile:
spatial_noise: bool
If True, spatial noise correlations are added. True by default.
nunu_correlation: bool
If True, correlations between frequency sub-bands are added. True by default.
noise_only: bool
If True, only returns the noise maps and the coverage (without the sky signal).
integrate_into_band: bool
If True, averaging input sub-band maps into reconstruction sub-bands. True by default.
verbose: bool
Returns
-------
maps + noisemaps, maps, noisemaps, coverage
"""
### Input bands
Nfreq_edges, nus_edge, nus, deltas, Delta, Nbbands = qubic.compute_freq(self.filter_nu,
self.Nfin,
self.filter_relative_bandwidth)
### Output bands
# Check Nfout is between 1 and 8.
if self.Nfout < 1 or self.Nfout > 8:
raise NameError("Nfout should be contained between 1 and 8 for FastSimulation.")
Nfreq_edges_out, nus_edge_out, nus_out, deltas_out, Delta_out, Nbbands_out = qubic.compute_freq(self.filter_nu,
self.Nfout,
self.filter_relative_bandwidth)
# First get the convolved maps
if noise_only is False:
maps = np.zeros((self.Nfout, self.npix, 3))
if integrate_into_band:
if verbose:
print('Convolving each input frequency map')
maps_all = self.get_fullsky_convolved_maps(FWHMdeg=FWHMdeg, verbose=verbose)
# Now averaging maps into reconstruction sub-bands maps
if verbose:
print('Averaging input maps from input sub-bands into reconstruction sub-bands:')
for i in range(self.Nfout):
print('doing band {} {} {}'.format(i, nus_edge_out[i], nus_edge_out[i + 1]))
inband = (nus > nus_edge_out[i]) & (nus < nus_edge_out[i + 1])
maps[i, :, :] = np.mean(maps_all[inband, :, :], axis=0)
else:
for i in range(self.Nfout):
freq = nus_out[i]
maps[i, :, :] = (self.sky.get_emission(freq * u.GHz)
* utils.bandpass_unit_conversion(freq * u.GHz,
weights=None,
output_unit=u.uK_CMB)).T
_, maps = self.smoothing(maps, FWHMdeg, self.Nfout, nus_out, verbose=verbose)
##############################################################################################################
# Restore data for FastSimulation ############################################################################
##############################################################################################################
#### Directory for fast simulations
dir_fast = os.path.join(os.path.dirname(__file__), 'data', f'FastSimulator_version{version_FastSim}')
#### Integration time assumed in FastSim files
fastsimfile_effective_duration = 2.
with open(dir_fast + os.sep + 'DataFastSimulator_{}{}_nfsub_{}.pkl'.format(self.dictionary['config'],
str(self.filter_nu),
self.Nfout),
"rb") as file:
DataFastSim = pickle.load(file)
print(file)
# Read Coverage map
if coverage is None:
DataFastSimCoverage = pickle.load(open(dir_fast + os.sep + 'DataFastSimulator_{}{}_coverage.pkl'.format(
self.dictionary['config'],
str(self.filter_nu)), "rb"))
coverage = DataFastSimCoverage['coverage']
# Read noise normalization
if sigma_sec is None:
#### Beware ! Initial End-To-End simulations that produced the first FastSimulator were done with
#### Effective_duration = 4 years and this is the meaning of signoise
#### New files were done with 2 years and as result the signoise needs to be multiplied by sqrt(effective_duration/4)
sigma_sec = DataFastSim['signoise'] * np.sqrt(fastsimfile_effective_duration / 4.)
# # Read Nyears
# if Nyears is None:
# Nyears = DataFastSim['years']
# Read Noise Profile
if noise_profile is True:
effective_variance_invcov = DataFastSim['effective_variance_invcov']
else:
effective_variance_invcov = None
# Read Spatial noise correlation
if spatial_noise is True:
clnoise = DataFastSim['clnoise']
else:
clnoise = None
# Read Noise Profile
if nunu_correlation is True:
covI = DataFastSim['CovI']
covQ = DataFastSim['CovQ']
covU = DataFastSim['CovU']
sub_bands_cov = [covI, covQ, covU]
else:
sub_bands_cov = None
##############################################################################################################
# Now pure noise maps
if verbose:
print('Making noise realizations')
noisemaps = self.create_noise_maps(sigma_sec, coverage, nsub=self.Nfout,
Nyears=Nyears, verbose=verbose, seed=seed,
effective_variance_invcov=effective_variance_invcov,
clnoise=clnoise,
sub_bands_cov=sub_bands_cov)
if self.Nfout == 1:
noisemaps = np.reshape(noisemaps, (1, len(coverage), 3))
seenpix = noisemaps[0, :, 0] != 0
coverage[~seenpix] = 0
if noise_only:
return noisemaps, coverage
else:
maps[:, ~seenpix, :] = 0
return maps + noisemaps, maps, noisemaps, coverage