def spec2filter(filter, obs_spec, model_spec=None, badpxl_tolerance = 0.5):
'''
Converts a spectrum on AB magnitude given a filter bandpass.
If there are bad pixels on filter interval on a fraction inferior to the badpxl_tolerance, it will
interpolate the missing values or, in case of a model_spec != None, it will use the values from model_spec.
Parameters
----------
filter : dict
Filter transmission curve. Dictionary containing the following entries:
{'wl': array_like
Wavelength (in Angstroms!).
'transm': array_like
Filter transmission response.
}
obs_spec : dict
Observed Spectra. Dictionary containing the following entries:
{'wl': array_like,
'flux': array_like,
'error': array_like, optional,
'flag': array_like, optional,
'model_spec': array_like, optional
}
model_spec : dict, optional
Model Spectra which could be used when there are missing (due to err or flagged). Dictionary containing the following entries:
{ 'wl': array_like
Wavelength (in the same units as filter response curve)
'flux': array_like
Flux on a given wl
}
badpxl_tolerance : float, default: 0.5
Bad pixel fraction tolerance on the spectral interval of the filter.
Returns
-------
m_ab : float
AB magnitude on given filter
e_ab : float
AB magnitude error on given filter
See Also
--------
Notes
-----
'''
log = logging.getLogger('bgpe.photometry.photoconv')
#Cut the spectrum on the filterset range. This improves the velocity of the rest of the accounts.
obs_cut = obs_spec[np.bitwise_and(obs_spec['wl'] >= np.min(filter['wl']), obs_spec['wl'] <= np.max(filter['wl']))]
if model_spec != None:
model_cut = model_spec[:,np.bitwise_and(model_spec['wl'] >= np.min(filter['wl']), model_spec['wl'] <= np.max(filter['wl']))]
if len(model_cut) == 0:
log.warning('No enough MODEL datapoints eval synthetic photometry on this filter. Model will NOT be considered.')
model_spec = None
elif np.any(model_cut['wl'] != obs_cut['wl']):
log.warning('Model is not sampled the same way as the observed spectrum. Interpolating...')
aux = np.copy(model_cut)
model_cut = obs_cut[:len(model_spec)]
model_cut['flux'] = np.interp(obs_cut['wl'],aux['wl'],aux['flux'], left=0.0, right=0.0)
#Check if the filterset is sampled correctly
# - Tip: Resample the filter to your data when reading the filter to avoid unnessessary interpolations.
if(np.any(obs_cut['wl'] != filter['wl'])):
log.warning('Filter is not sampled the same way as the observed spectrum. Interpolating...')
wl_ = obs_cut['wl']
transm_ = np.interp(obs_cut['wl'],filter['wl'],filter['transm'])
else:
wl_ = filter['wl']
transm_ = filter['transm']
#Check for bad pixels. Good pixels should be signaled with flag(lambda) = 0 or 1.
# The recipe is:
# - If there is LESS bad_pixels than badpxl_tolerance:
# And there is a model_spec: use model_spec
# And there is NOT a model_spec: interpolate values
# - If there is MORE bad_pixels than badpxl_tolerance:
# Return a magnitude np.inf and an error of np.inf.
n = len(obs_cut)
if('flag' in obs_cut.dtype.names and 'error' in obs_cut.dtype.names): # First check if there is error AND flag.
good = np.bitwise_and(obs_cut['flag'] <= 1, obs_cut['error'] > 0)
bad = np.invert(good)
n_bad = np.sum(bad)
elif('error' in obs_cut.dtype.names): # Check if there is ONLY error.
good = (obs_cut['error'] > 0)
bad = np.invert(good)
n_bad = np.sum(bad)
elif('flag' in obs_cut.dtype.names): # Check if there is ONLY flag.
good = (obs_cut['flag'] <= 1)
bad = np.invert(good)
n_bad = np.sum(bad)
else: # If there is only spectra, all the pixels are good! =)
good = (obs_cut['wl'] > 0)
bad = np.invert(good)
n_bad = np.sum(bad)
log.debug( 'N_points: %d, N_bad: %d' % (n, n_bad) )
if(n_bad > 0 or n == 0): #If we have problems we have to deal with them... ;)
if (n == 0):
log.debug('# of pixels = 0. m = inf and m_err = inf')
m_ab = np.inf
e_ab = np.inf
return m_ab, e_ab
p_bad = np.float(n_bad)/np.float(len(obs_cut))
if(p_bad > badpxl_tolerance): #If we have # of bad pixels greater than 50%, then make filter flux and error equal to inf.
log.debug('# of bad pixels > badpxl_tolerance. m = inf and m_err = inf')
m_ab = np.inf
e_ab = np.inf
return m_ab, e_ab
else:
#If we have # of bad pixels less than 50%, we simply neglect this point on the error acoounts,
#and make flux = synthetic flux, if available, if not, interpolate values.
if model_spec != None: obs_cut['flux'][bad] = model_cut['flux'][bad]
else: raise 'interpo' #TODO: interpolation!
else: # If our observed obs_spec is ALL ok. =)
log.debug('No bad pixel! =)')
m_ab = -2.5 * np.log10( np.trapz(obs_cut['flux'] * transm_ * wl_, wl_) / np.trapz(transm_ / wl_, wl_) ) - 2.41
if('error' in obs_cut.dtype.names):
e_ab = 1.0857362047581294 * np.sqrt( np.sum(transm_[good]**2 * obs_cut['error'][good]**2 * wl_[good] ** 2 )) / np.sum(obs_cut['flux'][good] * transm_[good] * wl_[good])
else:
e_ab = 0.0
return m_ab, e_ab
def spec2filterset(filterset, obs_spec, model_spec = None, badpxl_tolerance = 0.5):
'''
Run spec2filter over a filterset
Parameters
----------
filterset : object
Filter transmission curves (see: bgpe.io.readfilterset).
obs_spec : dict
Observed Spectra. Dictionary containing the following entries:
{ 'wl': array_like
Wavelength. (in Angstroms!)
'flux': array_like
Flux on a given wl.
'error': array_like, optional
Flux error. If err < 0, the point will be considered as a problem.
'flag': array_like, optional
Bad pixel flag. Pixels are considered bad if flag > 1.
'model_spec': array_like, optional
Model Spectra which could be used when there are missing (due to err or flagged).
}
model_spec : dict, optional
Model Spectra which could be used when there are missing (due to err or flagged). Dictionary containing the following entries:
{ 'wl': array_like
Wavelength (in the same units as filter response curve)
'flux': array_like
Flux on a given wl
}
badpxl_tolerance : float, default: 0.5
Bad pixel fraction tolerance on the spectral interval of the filter.
Returns
-------
mags : array_like
Filterset magnitudes
Dictionary containing the following entries:
{
'm_ab': array_like
AB magnitude on given filter.
'e_ab' : array_like
AB magnitude error on given filter.
}
See Also
--------
spec2filter, bgpe.io.readfilterset
Notes
-----
'''
log = logging.getLogger('bgpe.photometry.photoconv')
filter_ids = np.unique(filterset['ID_filter'])
mags = np.zeros(len(filter_ids), dtype = np.dtype([('m_ab', '<f8'), ('e_ab', '<f8')]))
for i_filter in range(len(filter_ids)):
filter = filterset[filterset['ID_filter'] == filter_ids[i_filter]]
mags[i_filter]['m_ab'], mags[i_filter]['e_ab'] = spec2filter(filter, obs_spec, model_spec, badpxl_tolerance = badpxl_tolerance)
log.debug('Magnitude to filter %s: %3.2f, error: %3.2f' % (filter_ids[i_filter], mags[i_filter]['m_ab'], mags[i_filter]['e_ab']) )
return mags
syn02_file = '%s/sample.F%s.%s.f.Starlight.SYN02.tab.BS.bz2' % (tables_dir, args.d[0], args.sa[0])
syn03_file = '%s/sample.F%s.%s.f.Starlight.SYN03.tab.BS.bz2' % (tables_dir, args.d[0], args.sa[0])
syn04_file = '%s/sample.F%s.%s.f.Starlight.SYN04.tab.BS.bz2' % (tables_dir, args.d[0], args.sa[0])
db_file = args.o[0]
filter_file = args.f[0]
filterid = os.path.basename(filter_file).split('.')[0]
z_from = np.float(args.z_ini[0])
z_to = np.float(args.z_fin[0])
z_step = np.float(args.dz[0])
# 1 - Read files
# 1.1 - Filter
db_f = read_filterhdf5(filter_file)
# 1.2 - Emission lines
log.debug('Reading elines...')
t_start = time.time()
tb = atpy.Table(el_file, type='starlight_el')
log.debug('Took %3.2f seconds.' % (time.time() - t_start))
# 1.3 - SYN0[1-4]
log.debug('Reading SYN files...')
t_start = time.time()
tsyn01 = atpy.Table(syn01_file, type='starlight_syn01', include_names = ('id', 'A_V', 'v0', 'vd', 'SN_w', 'SN_n'))
tsyn02 = atpy.Table(syn02_file, type='starlight_syn02', include_names = ('id', 'at_flux', 'at_mass', 'aZ_flux', 'aZ_mass', 'am_flux', 'am_mass'))
tsyn03 = atpy.Table(syn03_file, type='starlight_syn03', include_names = ('id', 'M2L_r'))
tsyn04 = atpy.Table(syn04_file, type='starlight_syn04', include_names = ('id', 'Mcor_fib', 'Mcor_gal', 'DL_Mpc', 'Mini_fib', 'z'))
log.debug('Took %3.2f seconds.' % (time.time() - t_start))
