def get_sacc_windows(self, wsp):
"""
Get window functions for each bandpower so they can be stored into the final SACC files.
"""
#Compute window functions
nbands = wsp.wsp.bin.n_bands
l_arr = np.arange(self.lmax + 1)
if not os.path.isfile(self.get_output_fname('windows_l', ext='npz')):
print("Computing window functions")
windows = np.zeros([nbands, self.lmax + 1])
t_hat = np.zeros(self.lmax + 1)
for il, l in enumerate(l_arr):
t_hat[il] = 1.
windows[:, il] = wsp.decouple_cell(
wsp.couple_cell(l_arr, [t_hat]))
t_hat[il] = 0.
np.savez(self.get_output_fname('windows_l'), windows=windows)
else:
print("Reading window functions")
windows = np.load(self.get_output_fname('windows_l',
ext='npz'))['windows']
windows_sacc = []
#i_x=0
for i in range(self.nbins):
for j in range(i, self.nbins):
for b in range(nbands):
windows_sacc.append(sacc.Window(l_arr, windows[b]))
return windows_sacc
def get_sacc_binning(self, with_windows=False):
typ, ell, t1, q1, t2, q2 = [], [], [], [], [], []
l_eff = self.bins.get_effective_ells()
windows = None
if with_windows:
windows_wsp = {}
for b1 in range(self.n_bpss):
for b2 in range(b1, self.n_bpss):
name = self.get_workspace_label(b1, b2)
windows_wsp[name] = {}
wsp = self.workspaces[name]
bpw_win = wsp.get_bandpower_windows()
windows_wsp[name]['EE'] = bpw_win[0, :, 0, :]
windows_wsp[name]['EB'] = bpw_win[1, :, 1, :]
windows_wsp[name]['BE'] = bpw_win[2, :, 2, :]
windows_wsp[name]['BB'] = bpw_win[3, :, 3, :]
windows = []
for b1, b2, s1, s2, l1, l2 in self.get_cell_iterator():
if (b1 == b2) and (s1 == s2):
types = ['EE', 'EB', 'BB'] # Do not double-count EB
else:
types = ['EE', 'EB', 'BE', 'BB']
name_wsp = self.get_workspace_label(b1, b2)
for ty in types:
for il, l in enumerate(l_eff):
ell.append(l)
typ.append(ty)
t1.append(b1 * self.nsplits + s1)
t2.append(b2 * self.nsplits + s2)
q1.append('C')
q2.append('C')
if with_windows:
windows.append(
sacc.Window(self.larr_all,
windows_wsp[name_wsp][ty][il]))
return sacc.Binning(typ, ell, t1, q1, t2, q2, windows=windows)
axis=0) / area
ec = {
pn: np.sum(np.array([
np.array([t.extra_cols[pn]
for t in sc[f].tracers]) * sc[f].meta['Area_rad']
for f in fields
]),
axis=0) / area
for pn in pz_codes
}
# Windows
wins = [
sacc.Window(
sc['GAMA09H'].binning.windows[i_w].ls,
np.sum(np.array([
sc[f].binning.windows[i_w].w * sc[f].meta['Area_rad']
for f in fields
]),
axis=0) / area)
for i_w, ww in enumerate(sc['GAMA09H'].binning.windows)
]
# Bins
s_bn = sc['GAMA09H'].binning
s_bn.windows = wins
# Tracers
s_tr = []
for i_t in range(4):
T = sacc.Tracer('bin_%d' % i_t,
'point',
if pb == "T":
tb_name = "%s_%s_s0" % (eb, fb)
else:
tb_name = "%s_%s_s2" % (eb, fb)
cl_type = "cl_" + map_types[pa] + map_types[pb]
lbin = data["%s_%s" % (ea, fa), "%s_%s" % (eb, fb)]["lbin"]
cb = data["%s_%s" % (ea, fa), "%s_%s" % (eb, fb)][pa + pb]
spec_sacc.add_ell_cl(cl_type, ta_name, tb_name, lbin, cb)
if isim == 0:
bbl = get_bbl(ea, fa, pa, eb, fb, pb)
ls_w = np.arange(2, bbl.shape[-1] + 2)
wins = sacc.Window(ls_w, bbl.T)
cov_sacc.add_ell_cl(cl_type,
ta_name,
tb_name,
lbin,
cb,
window=wins,
window_id=range(len(lbin)))
if isim == 0:
cov_sacc.add_covariance(cov_full)
cov_sacc.save_fits("%s/data_sacc_w_covar_and_Bbl.fits" % sacc_dir,
overwrite=True)
spec_sacc.save_fits("%s/data_sacc_%s.fits" % (sacc_dir, sim_suffix),
overwrite=True)
q1_arr = []
q2_arr = []
w_arr = []
for ic, c in enumerate(corr_ordering):
s1, s2 = c
tn1 = s1[:-2]
q1 = s1[-1]
t1 = np.where(tracer_names == tn1)[0][0]
tn2 = s2[:-2]
q2 = s2[-1]
t2 = np.where(tracer_names == tn2)[0][0]
typ = q1 + q2
for b in range(nells):
w = windows[b, ic]
lmean = np.sum(ls * w) / np.sum(w)
win = sacc.Window(ls, w)
ls_arr.append(lmean)
w_arr.append(win)
q1_arr += nells * [q1]
q2_arr += nells * [q2]
t1_arr += nells * [t1]
t2_arr += nells * [t2]
typ_arr += nells * [typ]
bins = sacc.Binning(typ_arr,
ls_arr,
t1_arr,
q1_arr,
t2_arr,
q2_arr,
windows=w_arr)
# at 100 ell bins for density correlations
lvals = np.arange(100, 1000, 100)
Ntracer = len(tracers)
type, ell, t1, q1, t2, q2, val, err, wins = [], [], [], [], [], [], [], [], []
for t1i in range(Ntracer):
for t2i in range(t1i, Ntracer):
for l in lvals:
## we have Fourier space measurement
type.append('FF')
## at this nominal ell
ell.append(l)
## but in detail the measurement
## is a Gaussian window around central ell +/- 50
wins.append(
sacc.Window(
np.arange(l - 50, l + 50),
np.exp(-(1.0 * np.arange(-50, 50))**2 / (2 * 20.**2))))
## We refer to tracers by their index
t1.append(t1i)
t2.append(t2i)
## Here we have density cross-correlations so "P" as point
## except for CMB where
q1.append('S')
q2.append('S')
## values and errors
val.append(np.random.uniform(0, 10))
err.append(np.random.uniform(1, 2))
binning = sacc.Binning(type, ell, t1, q1, t2, q2, windows=wins)
mean = sacc.MeanVec(val)
cl = np.exp(lcl_i(np.log(l_all)))
nl = nl_yy_f(l_all)
# Beam
beam_fwhm_amin = 1.4
beam = np.exp(-0.5 * l_all * (l_all + 1) *
(beam_fwhm_amin * np.pi / 180 / 60)**2)
cl *= beam**2
# Bandpower window
d_ell = 100
n_bands = 10000 // d_ell
windows = np.zeros([n_bands, lmax - 1])
for i in range(n_bands):
windows[i, i * d_ell + 2:(i + 1) * d_ell + 2] = 1. / d_ell
win = sacc.Window(l_all, windows.T)
# Convolve and Gaussian covariance
bpws = np.dot(windows, cl)
bpws_n = np.dot(windows, nl)
leff = np.dot(windows, l_all)
fsky = 0.1
cov = np.diag((bpws + bpws_n)**2 / ((leff + 0.5) * fsky * d_ell))
# Add statistical noise
#bpws = np.random.multivariate_normal(bpws, cov)
s = sacc.Sacc()
s.add_tracer('Map', 'SO_y', quantity='cmb_tSZ', spin=0, ell=l_all, beam=beam)
s.add_ell_cl('cl_00',
'SO_y',
from packaging import version
# Dummy arrays
vecsize = 100
x = np.linspace(0, 100, vecsize)
y = 1. / (x + 50)**2
# Create file
s = sacc.Sacc()
# Add tracers
s.add_tracer('NZ', 'nz', x, y)
s.add_tracer('NZ', 'nz_b', x, y)
wth = sacc.TopHatWindow(0., 1.)
ww = sacc.Window(x, y)
s.add_ell_cl('galaxy_density_cl', 'nz', 'nz', x, y, window=[wth] * vecsize)
s.add_theta_xi('galaxy_density_xi', 'nz', 'nz', x, y, window=[ww] * vecsize)
nvec = 2
if version.parse(sacc.__version__) > version.parse('0.2.0'):
wlth = sacc.LogTopHatWindow(-2., 0.)
s.add_ell_cl('galaxy_density_cl',
'nz_b',
'nz_b',
x,
y,
window=[wlth] * vecsize)
nvec += 1
if version.parse(sacc.__version__) > version.parse('0.4.0'):
t2_arr = []
q1_arr = []
q2_arr = []
w_arr = []
for ic, c in enumerate(corr_ordering):
s1, s2 = c
tn1 = s1[:-2]
q1 = s1[-1]
t1 = np.where(tracer_names == tn1)[0][0]
tn2 = s2[:-2]
q2 = s2[-1]
t2 = np.where(tracer_names == tn2)[0][0]
typ = q1 + q2
for ib, w in enumerate(windows):
lmean = leff[ib]
win = sacc.Window(larr_all, w)
ls_arr.append(lmean)
w_arr.append(win)
q1_arr += nbands * [q1]
q2_arr += nbands * [q2]
t1_arr += nbands * [t1]
t2_arr += nbands * [t2]
typ_arr += nbands * [typ]
bins = sacc.Binning(typ_arr,
ls_arr,
t1_arr,
q1_arr,
t2_arr,
q2_arr,
windows=w_arr)
print("Reading window functions")
windows=np.load(o.fname_mcm+".windows.npz")['windows']
#Compute all cross-correlations
print("Computing all cross-power specra")
cls_all=[]
windows_sacc=[]
ordering=np.zeros([nbins,nbins],dtype=int)
i_x=0
for i in range(nbins) :
for j in range(i,nbins) :
cl_coupled=nmt.compute_coupled_cell_flat(tracers[i].field,tracers[j].field,bpws)
cls_all.append(wsp.decouple_cell(cl_coupled)[0])
for b in range(nbands) :
windows_sacc.append(sacc.Window(l_arr,windows[b]))
ordering[i,j]=i_x
if j!=i :
ordering[j,i]=i_x
i_x+=1
cls_all=np.array(cls_all)
n_cross=len(cls_all)
n_ell=len(ell_eff)
#Get guess power spectra
if o.guess_cell=='data' :
#Interpolate measured power spectra
lth=np.arange(2,lmax+1)+0.
clth=np.zeros([n_cross,len(lth)])
for i in range(n_cross) :
clf=interp1d(ell_eff,cls_all[i],bounds_error=False,fill_value=0,kind='linear')
def get_binnings(self, with_windows=True):
# Get windows if needed
win_coadd = None
win_nulls = None
if with_windows:
win_coadd = []
win_nulls = []
ls_win = self.s_splits.binning.windows[0].ls.copy()
nls = len(ls_win)
windows = np.zeros(
[self.nbands, 2, self.nbands, 2, self.n_bpws, nls])
for t1, t2, typ, ells, ndx in self.sorting:
b1, s1 = self.tracer_number_to_band_split(t1)
b2, s2 = self.tracer_number_to_band_split(t2)
typ = typ.decode()
p1 = self.index_pol[typ[0]]
p2 = self.index_pol[typ[1]]
if (s1 == 0) and (s2 == 0):
for b, i in enumerate(ndx):
w = self.s_splits.binning.windows[i].w
windows[b1, p1, b2, p2, b, :] = w
if not ((b1 == b2) and (p1 == p2)):
windows[b2, p2, b1, p1, b, :] = w
# Binnings for coadds
typ, ell, t1, q1, t2, q2 = [], [], [], [], [], []
for i1 in range(2 * self.nbands):
b1 = i1 // 2
p1 = i1 % 2
for i2 in range(i1, 2 * self.nbands):
b2 = i2 // 2
p2 = i2 % 2
ty = self.pol_names[p1] + self.pol_names[p2]
for il, ll in enumerate(self.ells):
ell.append(ll)
typ.append(ty)
t1.append(b1)
t2.append(b2)
q1.append('C')
q2.append('C')
if with_windows:
win_coadd.append(
sacc.Window(ls_win, windows[b1, p1, b2, p2, il]))
self.bins_coadd = sacc.Binning(typ,
ell,
t1,
q1,
t2,
q2,
windows=win_coadd)
# Binnings for nulls
typ, ell, t1, q1, t2, q2 = [], [], [], [], [], []
for i_null, (i, j, k, l) in enumerate(self.pairings):
for b1 in range(self.nbands):
tr1 = self.ind_nulls['band%d_null%dm%d' %
(b1 + 1, i + 1, j + 1)]
for p1 in range(2):
for b2 in range(self.nbands):
tr2 = self.ind_nulls['band%d_null%dm%d' %
(b2 + 1, k + 1, l + 1)]
for p2 in range(2):
ty = self.pol_names[p1] + self.pol_names[p2]
for il, ll in enumerate(self.ells):
ell.append(ll)
typ.append(ty)
t1.append(tr1)
t2.append(tr2)
q1.append('C')
q2.append('C')
if with_windows:
win_nulls.append(
sacc.Window(
ls_win, windows[b1, p1, b2, p2,
il]))
self.bins_nulls = sacc.Binning(typ,
ell,
t1,
q1,
t2,
q2,
windows=win_nulls)