def test_conversion3(self):
g = S.GaussianSource(1, 4000, 100, 'angstrom', 'photlam')
tf1 = g.total_flux
g.convert('nm')
tf2 = g.factor * g.sigma * np.sqrt(2.0 * np.pi)
self.assertEqual(tf1, tf2)
def setUp(self):
self.bb = S.BlackBody(5000)
self.em = S.GaussianSource(3300, 1, 1)
self.flat = S.FlatSpectrum(10)
self.pl = S.PowerLaw(5000, -2)
self.tspec = S.ArraySpectrum(self.bb.wave,
self.bb.flux,
fluxunits=self.bb.fluxunits)
self.pl.writefits('ac_pl.fits')
self.fspec = S.FileSpectrum('ac_pl.fits')
def test_conversion1(self):
g = S.GaussianSource(1, 4000, 100, 'angstrom', 'photlam')
angflux = g.sample(self.wave)
g.convert('nm')
nmflux = g.sample(self.wave / 10.)
self.assertEqualNumpy(nmflux, angflux)
def test_conversion2(self):
g = S.GaussianSource(1, 4000, 100, 'angstrom', 'photlam')
self.assertEqualNumpy(
g.sample(self.wave)[:10], g.sample(self.wave[:10]))
g.convert('nm')
self.assertEqualNumpy(
g.sample(self.wave)[:10], g.sample(self.wave[:10]))
def test_values(self):
ref = np.array([
3.63805721e-123, 9.05875032e-121, 2.13395205e-118, 4.75574568e-116,
1.00269809e-113, 2.00004310e-111, 3.77421053e-109, 6.73799198e-107,
1.13802672e-104, 1.81841090e-102
])
g = S.GaussianSource(1, 4000, 100, 'angstrom', 'photlam')
test = g.sample(self.wave[:10])
self.assertApproxNumpy(test, ref)
def rudy_ic5117():
"""
Make a simple spectrum from the line table from Rudy et al.
"""
from astropy.table import Table as table
import pysynphot as S
t = table.read('rudy_table1.dat.txt', format='ascii.cds')
spec = S.FlatSpectrum(6e-11*1.e-4, fluxunits='photlam')
#spec.convert('photlam')
for line in t:
if line['f_Ratio'] == 'Atmosphere':
continue
#
#spec += S.GaussianSource(line['Ratio']*4.95e-12, line['Wave']*1., line['Wave']*30./3.e5, fluxunits='photlam')
if line['Wave'] == 10830:
f = 1.35
else:
f = 1.
#
spec += S.GaussianSource(line['Ratio']*4.95e-12*f, line['Wave']*1., line['Wave']*800./3.e5, fluxunits='photlam')
ok = (spec.wave > 7000) & (spec.wave < 1.7e4)
np.savetxt('rudy_spec.dat', np.array([spec.wave[ok], spec.flux[ok]]).T, fmt='%.5e')
def _obj_beam_fit(params, beams, x_nodes):
"""
params = [2.953e-03, 1.156e+00, 1.297e+01, 9.747e-01 , 9.970e-01 , 8.509e-01, 1.076e+00 , 1.487e+00 , 7.864e-01, 1.072e+00 , 1.015e+00]
"""
import time
from scipy.interpolate.fitpack import splev, splrep
import pysynphot as S
### Spline continuum + gaussian line
l0 = 6563. * (1 + params[1])
if (l0 < 1.12e4) | (l0 > 1.63e4):
return -np.inf
line = S.GaussianSource(params[2], l0, 10)
tck = splrep(x_nodes, params[3:], k=3, s=0)
xcon = np.arange(0.9e4, 1.8e4, 0.01e4)
ycon = splev(xcon, tck, der=0, ext=0)
spec = S.ArraySpectrum(xcon, ycon, fluxunits='flam', keepneg=True) + line
lnprob = 0
for key in beams.keys():
beam = beams[key]
modelf = beam.compute_model(beam.clip_thumb,
xspec=spec.wave,
yspec=spec.flux,
in_place=False)
lnprob += -0.5 * np.sum(((beam.cutout_scif - params[0] - modelf)**2 /
beam.cutout_varf)[beam.cutout_maskf])
if ~np.isfinite(lnprob):
lnprob = -np.inf
print params, lnprob
#time.sleep(0.2)
return lnprob
def setup_fit():
snlim = 0
snlim = 1. / np.sqrt(len(beams))
Rmax = 5
for key in FLT.keys():
beam = beams[key]
beam.cutout_sci = beam.get_cutout(FLT[key].im['SCI'].data) * 1
beam.cutout_dq = beam.get_cutout(FLT[key].im['DQ'].data) * 1
beam.cutout_err = beam.get_cutout(FLT[key].im['ERR'].data) * 1
beam.cutout_mask = (beam.cutout_dq - (beam.cutout_dq & 512) == 0) & (
beam.cutout_err > 0) & (beam.cutout_err < 10)
sh = beam.cutout_sci.shape
yp, xp = np.indices(beam.thumb.shape)
r = np.sqrt((xp - sh[0] / 2)**2 + (yp - sh[0] / 2)**2)
beam.norm = beam.thumb[r <= Rmax].sum() / 1.e-17
beam.clip_thumb = beam.thumb * (r <= Rmax)
beam.compute_model(beam.clip_thumb)
sn = beam.model / beam.cutout_err
#beam.mask &= (sn > 0.5) & (beam.err > 0)
beam.cutout_mask &= (sn > snlim) & (beam.cutout_err > 0)
beam.cutout_sci[~beam.cutout_mask] = 0
beam.cutout_err[~beam.cutout_mask] = 0
beam.cutout_var = beam.cutout_err**2
beam.cutout_scif = beam.cutout_sci.flatten()
beam.cutout_varf = beam.cutout_var.flatten()
beam.cutout_maskf = beam.cutout_mask.flatten()
ds9.view((beam.cutout_scif - beam.cutout_modelf).reshape(
beam.cutout_sci.shape) * beam.cutout_mask)
##
x_nodes = np.arange(1.0e4, 1.71e4, 0.1e4)
x_nodes = np.arange(1.0e4, 1.71e4, 0.03e4)
x_nodes = np.arange(1.05e4, 1.71e4, 0.075e4)
init = np.append([0, 1.148, 500], np.ones(len(x_nodes)))
init = np.append([0, 1.2078, 500], np.ones(len(x_nodes)))
step_sig = np.append([0.01, 0.1, 50], np.ones(len(x_nodes)) * 0.1)
init = np.append([0, 1.0, 500], np.ones(len(x_nodes)))
step_sig = np.append([0.0, 0.2, 150], np.ones(len(x_nodes)) * 0.1)
### don't fit redshift
#init = np.append([0, 1.0, 0], np.ones(len(x_nodes)))
#step_sig = np.append([0.0,0.,0], np.ones(len(x_nodes))*0.2)
obj_fun = _obj_beam_fit
obj_args = [beams, x_nodes]
ndim, nwalkers = len(init), len(init) * 2
p0 = [(init + np.random.normal(size=ndim) * step_sig)
for i in xrange(nwalkers)]
NTHREADS, NSTEP = 2, 5
sampler = emcee.EnsembleSampler(nwalkers,
ndim,
obj_fun,
args=obj_args,
threads=NTHREADS)
#
t0 = time.time()
result = sampler.run_mcmc(p0, NSTEP)
t1 = time.time()
print 'Sampler: %.1f s' % (t1 - t0)
param_names = ['bg', 'z', 'Ha']
param_names.extend(['spl%d' % i for i in range(len(init) - 3)])
import unicorn.interlace_fit
chain = unicorn.interlace_fit.emceeChain(chain=sampler.chain,
param_names=param_names)
#obj_fun(chain.map, beams)
#obj_fun(init, beams, x_nodes)
params = chain.map * 1.
#params = init
### Show spectra
from scipy.interpolate.fitpack import splev, splrep
# NDRAW=100
# draw = chain.draw_random(NDRAW)
# for i in range(NDRAW):
# line = S.GaussianSource(draw[i,2], 6563.*(1+draw[i,1]), 10)
# tck = splrep(x_nodes, draw[i,3:], k=3, s=0)
# xcon = np.arange(0.9e4,1.8e4,0.01e4)
# ycon = splev(xcon, tck, der=0, ext=0)
# spec = S.ArraySpectrum(xcon, ycon, fluxunits='flam', keepneg=True)+line
# plt.plot(spec.wave, spec.flux, alpha=0.1, color='red')
#
line = S.GaussianSource(params[2], 6563. * (1 + params[1]), 10)
tck = splrep(x_nodes, params[3:], k=3, s=0)
xcon = np.arange(0.9e4, 1.8e4, 0.01e4)
ycon = splev(xcon, tck, der=0, ext=0)
pspec = S.ArraySpectrum(xcon, ycon, fluxunits='flam', keepneg=True) + line
for key in beams.keys():
beam = beams[key]
beam.compute_model(beam.clip_thumb,
xspec=pspec.wave,
yspec=pspec.flux,
in_place=True)
xfull, yfull, mfull, cfull = [], [], [], []
for i, key in enumerate(beams.keys()):
beam = beams[key]
#ds9.frame(i+1)
#ds9.view((beam.sci-beam.model)*beam.mask)
ls = 'steps-mid'
marker = 'None'
#ls = '-'
#ls, marker = 'None', '.'
xfull = np.append(xfull, beam.lam)
ybeam = (beam.cutout_sci - params[0] * beam.cutout_mask).sum(
axis=0)[sh[0] / 2:-sh[0] / 2] / beam.ysens / beam.norm
yfull = np.append(yfull, ybeam)
#plt.plot(beam.lam, ybeam, color='black', alpha=0.5, linestyle=ls, marker=marker)
cbeam = ((beam.cutout_m - beam.omodel) * beam.cutout_mask).sum(
axis=0)[sh[0] / 2:-sh[0] / 2] / beam.ysens / beam.norm
cfull = np.append(cfull, cbeam)
mbeam = ((beam.model) * beam.mask).sum(
axis=0)[sh[0] / 2:-sh[0] / 2] / beam.ysens / beam.norm
mfull = np.append(mfull, mbeam)
plt.plot(beam.lam,
ybeam - cbeam,
color='black',
linestyle=ls,
alpha=0.5)
plt.plot(beam.lam, mbeam, color='blue', linestyle=ls, alpha=0.5)
#nx = 20
kern = np.ones(nx) / nx
so = np.argsort(xfull)
plt.plot(xfull[so][nx / 2::nx],
nd.convolve((yfull - cfull)[so], kern)[nx / 2::nx],
color='orange',
linewidth=2,
alpha=0.8,
linestyle='steps-mid')
plt.plot(xfull[so][nx / 2::nx],
nd.convolve(mfull[so], kern)[nx / 2::nx],
color='green',
linewidth=2,
alpha=0.8,
linestyle='steps-mid')
plt.ylim(0, 5)
plt.xlim(1.e4, 1.78e4)
from scipy.optimize import fmin_bfgs
p0 = fmin_bfgs(_loss,
init[3:],
args=(beams, x_nodes),
gtol=1.e-3,
epsilon=1.e-3,
maxiter=100)
params = np.array(init)
params[3:] = p0
def test_symmetry2(self):
g = S.GaussianSource(1, 400, 100, 'nm', 'flam')
self.assertApproxFP(g.sample(395), g.sample(405))
def test_symmetry1(self):
g = S.GaussianSource(1, 4000, 100, 'angstrom', 'photlam')
self.assertApproxFP(g.sample(3950), g.sample(4050))
def subtract_continuum(line='f673n',
cont='f814w',
target='galaxy',
file_name='galaxy*.fits',
line_name='ha',
z=0.02,
plot=False):
"""
Function for subtracting the continuum in a line emission image
INPUTS:
line: Filter into which the emission line falls.
cont: Filter within which the emission line and broader continuum are contained.
target: The name of the target, to be used in the output files.
file_name: General form of the file names that contain the line and continuum images.
line_name: State 'ha' or 'pab' to subtract Balmer-alpha or Paschen-beta respectively.
z: Redshift of the target object.
KEYWORDS:
PLOT: Set this keyword to produce a plot of the two-dimensional
continuum subtracted image.
OUTPUTS:
Fits file with the continuum subtracted line emission image.
"""
import pysynphot as S
import numpy as np
import glob
from grizli import utils
import astropy.io.fits as pyfits
import matplotlib.pyplot as plt
# restframe wavelength of the emission lines to subtract
wave_pab = 1.2822e4
wave_ha = 6562.8
print('Target =', target)
files = glob.glob(file_name)
files.sort()
images = {}
headers = {}
bandpasses = {}
for file in files:
im = pyfits.open(file)
filt = utils.get_hst_filter(im[0].header).lower()
for ext in [0, 1]:
if 'PHOTMODE' in im[ext].header:
photflam = im[ext].header['PHOTFLAM']
headers[filt.lower()] = im[ext].header
bandpasses[filt.lower()] = S.ObsBandpass(
im[ext].header['PHOTMODE'].replace(' ', ','))
break
flat_flam = S.FlatSpectrum(1., fluxunits='flam')
obs = S.Observation(flat_flam, bandpasses[filt.lower()])
my_photflam = 1 / obs.countrate()
flat_ujy = S.FlatSpectrum(1, fluxunits='ujy')
obs = S.Observation(flat_ujy, bandpasses[filt.lower()])
my_photfnu = 1 / obs.countrate()
images[filt.lower()] = [im['SCI'].data, im['ERR'].data]
# Use PySynphot to compute flux calibration factors
if line_name == 'pab':
# Pa-beta
cont_filter, line_filter, line_wave, name = cont, line, wave_pab, 'pab'
elif line_name == 'ha':
# H-alpha
cont_filter, line_filter, line_wave, name = cont, line, wave_ha, 'ha'
################
# Continuum - flat spectrum
cont = S.FlatSpectrum(1.e-19, fluxunits='flam')
###############
# Continuum - slope spectrum
cont_wave = np.arange(1000, 2.e4)
slope = 0 # flat
slope = 1 # red slope, increasing toward longer wavelengths
cont_flux = (cont_wave / 1.e4)**slope
cont = S.ArraySpectrum(cont_wave, cont_flux, fluxunits='flam')
################
# Continuum, galaxy model
templ = utils.load_templates(full_line_list=[],
line_complexes=False,
alf_template=True)['alf_SSP.dat']
cont = S.ArraySpectrum(templ.wave * (1 + z), templ.flux, fluxunits='flam')
# Gaussian line model
ref_flux = 1.e-17
line_model = S.GaussianSource(ref_flux,
line_wave * (1 + z),
10,
waveunits='angstrom',
fluxunits='flam')
cont_contin_countrate = S.Observation(cont,
bandpasses[cont_filter]).countrate()
line_contin_countrate = S.Observation(cont,
bandpasses[line_filter]).countrate()
line_emline_countrate = S.Observation(line_model,
bandpasses[line_filter]).countrate()
# Continuum-subtracted, flux-calibrated
line_calib = (
images[line_filter][0] -
images[cont_filter][0] * line_contin_countrate / cont_contin_countrate)
line_calib /= line_emline_countrate
# Propagated error of the subtraction
err_sub = np.sqrt((images[line_filter][1]**2) +
(images[cont_filter][1] * line_contin_countrate /
cont_contin_countrate)**2)
err_sub /= line_emline_countrate
if plot:
print("Continuum subtracted image")
plt.figure()
plt.imshow(line_calib, vmin=-0.5, vmax=0.5)
plt.colorbar()
primary_extn = pyfits.PrimaryHDU()
sci_extn = pyfits.ImageHDU(data=line_calib, name='SCI')
err_extn = pyfits.ImageHDU(data=err_sub, name='ERR')
hdul = pyfits.HDUList([primary_extn, sci_extn, err_extn])
hdul.writeto('sub_{0}_{1}.fits'.format(line_name, target),
output_verify='fix',
overwrite=True)
print(line_name, ' Continuum Subtracted')
def setUp(self):
self.sp = S.BlackBody(60000) + S.GaussianSource(1e-12, 5000, 30)
def setUp(self):
self.sp = S.GaussianSource(1e-12, 5000, 30)
def etc_spectra():
"""
Make actual spectra for the ETC for the different He bg cases
"""
import pysynphot as S
fl0 = 1.e-17
fwhm = 2 # Angstroms
heline = S.GaussianSource(fl0, 1.083e4, fwhm, fluxunits='flam')
continuum = S.FlatSpectrum(60, fluxunits='STMag')
model_spec = continuum + heline
model_spec.convert('flam')
bp = S.ObsBandpass('wfc3,ir,f105w')
flux = {'50': 0.1, '75': 0.5, '95': 1.5}
obs = S.Observation(heline, bp)
for key in flux.keys():
fl = fl0 * flux[key] / obs.countrate()
print fl
fl = fl0 * 1.0 / obs.countrate()
#### Normalize to "per arcsec"
fl /= 0.1211 * 0.1355
heline = S.GaussianSource(fl, 1.083e4, fwhm, fluxunits='flam')
obs = S.Observation(heline, bp)
import astropy.io.fits as pyfits
h = pyfits.Header()
h['CONTACT'] = ('G. Brammer', 'Person to contact in case of questions')
h['DESCRIPT'] = ('Airglow line at 10830 A')
h['MAPKEY'] = ('el10830a')
h['FILE_TYP'] = ('Airglow line at 10830 Angstroms')
h['SYSTEMS'] = ('PYETC')
h['PSYNEXPR'] = ('GaussianSource(%.3e,1.083e4,%.1f,fluxunits="flam")' %
(fl, fwhm))
#h.add_history("This file is a spectrum produced by PySynphot to model the He I 10830 airglow line as a Gaussian with FWHM = 2A. The spectrum is normalized to produce a background countrate of 1 e/s/pix in the WFC3/IR F105W filter.")
heline.writefits('el10830a_002.fits', clobber=True)
im = pyfits.open('el10830a_002.fits')
for card in h.cards:
im[0].header[card[0]] = (card[1], card[2])
#
im[0].header.add_history(
"This file is a spectrum produced by PySynphot to model the He I 10830 airglow line as a Gaussian with FWHM = 2A. The spectrum is normalized to produce a background countrate of 1 e-/s/pix in the WFC3/IR F105W filter."
)
im[0].header.add_history("")
im[0].header.add_history(
"The 50, 75, and 95 percentile fluxes of the line background observed in archival F105W observations are 0.1, 0.5, and 1.5 e/s/pix."
)
im[0].header.add_history("")
im[0].header.add_history(
"For more information on the He 10830 component to the IR background, see WFC3/ISR 2014-03: Time-varying Excess Earth-glow Backgrounds in the WFC3/IR Channel (Brammer et al.)."
)
im.writeto('el10830a_002.fits', clobber=True)
model_50 = model_spec.renorm(0.1, 'Counts', bp)
model_75 = model_spec.renorm(0.5, 'Counts', bp)
model_95 = model_spec.renorm(1.5, 'Counts', bp)
#### Demo figure
plt.plot(model_50.wave - 10830,
model_50.flux / 1.e-17,
label='F105W 0.1 e/s, 50%')
plt.plot(model_75.wave - 10830,
model_75.flux / 1.e-17,
label='F105W 0.5 e/s, 75%')
plt.plot(model_95.wave - 10830,
model_95.flux / 1.e-17,
label='F105W 1.5 e/s, 95%')
plt.xlim(-20, 20)
plt.ylabel(
r'$f_\lambda$ ($10^{-17}\,\mathrm{erg}\,\mathrm{s}^{-1}\,\mathrm{cm}^{-2}\,\AA^{-1})$'
)
plt.xlabel(r'$\lambda - 10830\,\AA$')
plt.legend(loc='best', fontsize=12)
plt.savefig('He10830_ETC_spec.png')
header = """# wavelength flux
# Units: (angstroms) (erg/s/cm2/A)
# He I 10830 A background component: 50%, 0.1 e/s in F105W
# Created: G. Brammer, 2014-09-18
# (made with PySynphot)
"""
fp = open('HeI_background_50.dat', 'w')
fp.write(header)
np.savetxt(fp, np.array([model_50.wave, model_50.flux]).T, fmt='%.6e')
fp.close()
#
header = """# wavelength flux
# Units: (angstroms) (erg/s/cm2/A)
# He I 10830 A background component: 75%, 0.5 e/s in F105W
# Created: G. Brammer, 2014-09-18
# (made with PySynphot)
"""
fp = open('HeI_background_75.dat', 'w')
fp.write(header)
np.savetxt(fp, np.array([model_75.wave, model_75.flux]).T, fmt='%.6e')
fp.close()
#
header = """# wavelength flux
# Units: (angstroms) (erg/s/cm2/A)
# He I 10830 A background component: 95%, 1.5 e/s in F105W
# Created: G. Brammer, 2014-09-18
# (made with PySynphot)
"""
fp = open('HeI_background_95.dat', 'w')
fp.write(header)
np.savetxt(fp, np.array([model_95.wave, model_95.flux]).T, fmt='%.6e')
fp.close()
def testRenormSynPysyn():
jv = S.ObsBandpass('johnson,v')
acs = S.ObsBandpass('acs,hrc,f555w')
abox = S.Box(5500,1)
#Removing vegamag from this list because it introduces
# a data dependency. Pysynphot has upgraded the version of
# vega that it uses for vegamag conversions (11/5/2010, r1629)
# but synphot has not. This caused the emission line test
# to fail (since the numbers were so small anyway, the small
# difference exceeded the tolerance).
uset = ['photlam', 'flam', 'photnu', 'fnu', 'jy', 'mjy', 'counts',
'stmag' ,'abmag', 'obmag'] #, 'vegamag'
ucounts = ['counts', 'obmag']
bb = S.BlackBody(5000)
bb.syndescrip = 'bb(5000)'
em = S.GaussianSource(1e-13,5500,250)
em.syndescrip = 'em(5500,250,1e-13,flam)'
foo = {bb:'bb', em:'em'}
rnfuncs = {renorm.StdRenorm:'Std'} #Just this one,
# renorm.renormTweak:'tweak',
# renorm.renormRevEng:'RevEng'
# renorm.renormSVN:'SVN'} - ignore SVN for now, we know it doesn't work
for method in rnfuncs:
for sp in (bb, em):
for u in uset: # ucounts
rnunit = Units(u)
if u == 'counts':
bp = acs
bp.syndescrip = "band(%s) " % bp
elif rnunit.isMag:
bp = jv
bp.syndescrip = 'band(johnson,v)'
else:
bp = abox
bp.syndescrip = 'box(5500,1)'
def renorm_compare(sp, rnval, rnunits, bp, callable):
"""Sp and bp must have a special .syndescrip attribute that
contains the string needed to make it with synphot"""
# Renormalize the spectrum
pysyn = callable(sp, bp, rnval, rnunits)
#pysyn=sp.renorm(rnval,rnunits,bp)
#pysyn = spectrum.StdRenorm(sp,bp,rnval,rnunits)
userdir = tempfile.mkdtemp(suffix='pysynphot')
old_cwd = os.getcwd()
iraf.chdir(userdir)
# Make a wavetable and a wavecat
fname = "%s.fits" % rnunits
pysyn.writefits(fname, clobber=True)
wname = "%s.cat" % rnunits
f = open(wname, 'w')
f.write("box %s\n" % fname)
f.close()
oname = "syn_%s" % fname
try:
#Run countrate
spstring = "rn(%s,%s,%s,%s) " % (sp.syndescrip,
bp.syndescrip, rnval,
rnunits)
iraf.countrate(spectrum=spstring, magnitude="",
instrument="box(15000,30000)",
form=str(pysyn.fluxunits), wavecat=wname,
output=oname)
syn = S.FileSpectrum(oname)
# Check that they have the same shape and fluxunits
assert (syn.flux.shape == pysyn.flux.shape)
assert (type(syn.fluxunits) == type(pysyn.fluxunits))
# Now a real test
idx = np.where(syn.flux != 0)
rat = (syn.flux[idx] / pysyn.flux[idx])
q = abs(1 - rat[2:-2])
qtrunc = (10**4*q).astype(np.int)
assert np.alltrue(qtrunc < 110), \
"Min/max ratio = %f,%f" % (q.min(),q.max())
finally:
iraf.chdir(old_cwd)
shutil.rmtree(userdir)
renorm_compare.description = (
"%s.test%s_%s_%s" % (__name__, rnfuncs[method], foo[sp], u))
yield renorm_compare, sp, 10, u, bp, method
def setUp(self):
self.sp=S.GaussianSource(1e14,10000,10)