def quick_analytic_cov(l, Clth_dict, window, binning_file, lmax):
bin_lo, bin_hi, bin_c, bin_size = pspy_utils.read_binning_file(
binning_file, lmax)
nbins = len(bin_size)
fsky = steve_effective_fsky(window)
prefac = 1 / ((2 * bin_c + 1) * fsky * bin_size)
cov = {}
cov["TT"] = Clth_dict["TaTc"] * Clth_dict["TbTd"] + Clth_dict[
"TaTd"] * Clth_dict["TbTc"]
cov["TE"] = Clth_dict["TaTc"] * Clth_dict["EbEd"] + Clth_dict[
"TaEd"] * Clth_dict["EbTc"]
cov["ET"] = Clth_dict["EaEc"] * Clth_dict["TbTd"] + Clth_dict[
"EaTd"] * Clth_dict["TbEc"]
cov["EE"] = Clth_dict["EaEc"] * Clth_dict["EbEd"] + Clth_dict[
"EaEd"] * Clth_dict["EbEc"]
mat_diag = np.zeros((4 * nbins, 4 * nbins))
for count, spec in enumerate(["TT", "TE", "ET", "EE"]):
lb, cov[spec] = pspy_utils.naive_binning(l, cov[spec], binning_file,
lmax)
cov[spec] *= prefac
for i in range(nbins):
mat_diag[i + count * nbins, i + count * nbins] = cov[spec][i]
return fsky, mat_diag
def theory_for_covariance(ps_dict,
spec_name_list,
spectra,
lmax,
beam=None,
binning_file=None,
force_positive=True):
ps_dict_for_cov = deepcopy(ps_dict)
if force_positive:
for name in spec_name_list:
m1, m2 = name.split("x")
for spec in spectra:
X, Y = spec
ps_dict_for_cov["%sx%s" % (m1, m1)][X + X] = np.abs(
ps_dict_for_cov["%sx%s" % (m1, m1)][X + X])
ps_dict_for_cov["%sx%s" % (m1, m1)][Y + Y] = np.abs(
ps_dict_for_cov["%sx%s" % (m2, m2)][Y + Y])
if beam is not None:
beam_data = np.loadtxt(beam)
l, bl = beam_data[:, 0], beam_data[:, 1]
lb, bb = pspy_utils.naive_binning(l, bl, binning_file, lmax)
for name in spec_name_list:
for spec in spectra:
ps_dict_for_cov[name][spec] *= bb**2
return ps_dict_for_cov
surveys = d["surveys"]
lmax = d["lmax"]
type = d["type"]
binning_file = d["binning_file"]
lth = np.arange(2, lmax + 2)
for sv in surveys:
arrays = d["arrays_%s" % sv]
for id_ar1, ar1 in enumerate(arrays):
for id_ar2, ar2 in enumerate(arrays):
if id_ar1 > id_ar2: continue
l, bl_ar1 = pspy_utils.read_beam_file(d["beam_%s_%s" % (sv, ar1)])
l, bl_ar2 = pspy_utils.read_beam_file(d["beam_%s_%s" % (sv, ar2)])
lb, bb_ar1 = pspy_utils.naive_binning(l, bl_ar1, binning_file,
lmax)
lb, bb_ar2 = pspy_utils.naive_binning(l, bl_ar2, binning_file,
lmax)
nsplits = len(d["maps_%s_%s" % (sv, ar1)])
spec_name_noise = "%s_%s_%sx%s_%s_noise" % (type, sv, ar1, sv, ar2)
print(spec_name_noise)
lb, nbs = so_spectra.read_ps("%s/%s.dat" %
(spectra_dir, spec_name_noise),
spectra=spectra)
nl_dict = {}
for spec in spectra:
nbs_mean = nbs[spec] * bb_ar1 * bb_ar2
plt.figure(figsize=(12, 12))
print("use kspace tf from file %s" % sv)
_, _, kf_tf, _ = np.loadtxt(ks_f["tf"], unpack=True)
tf_survey *= np.sqrt(np.abs(kf_tf[:len(lb)]))
if deconvolve_pixwin:
# extra pixel window function deconvolution for healpix and planck projected on CAR
pixwin_l = np.ones(2 * lmax)
if sv == "Planck":
print("Deconvolve Planck pixel window function")
pixwin_l = hp.pixwin(2048)
if template.pixel == "HEALPIX":
pixwin_l = hp.pixwin(template.nside)
# this should be checked with simulations since maybe this should be done at the mcm level
_, pw = pspy_utils.naive_binning(np.arange(len(pixwin_l)), pixwin_l,
binning_file, lmax)
tf_survey *= pw
for id_ar, ar in enumerate(d["arrays_%s" % sv]):
tf_array[sv, ar] = tf_survey.copy()
if d["deconvolve_map_maker_tf_%s" % sv]:
print("deconvolve map maker tf %s %s" % (sv, ar))
_, mm_tf = np.loadtxt("mm_tf_%s_%s.dat" % (sv, ar), unpack=True)
tf_array[sv, ar] *= mm_tf[:len(lb)]
np.savetxt(spec_dir + "/tf_%s_%s.dat" % (sv, ar),
np.transpose([lb, tf_array[sv, ar]]))
# compute the power spectra
arrays_1 = d["arrays_%s" % sv1]
nsplits_1 = nsplit[sv1]
if d["tf_%s" % sv1] is not None:
print("will deconvolve tf of %s" %sv1)
_, _, tf1, _ = np.loadtxt(d["tf_%s" % sv1], unpack=True)
tf1 = tf1[:len(lb)]
else:
tf1 = np.ones(len(lb))
if deconvolve_pixwin:
# we have an extra correction for the 1d healpix pixel window function
# this should be checked with simulations since maybe this
# step should be done at the mcm level
l_pw = np.arange(len(pixwin_l[sv1]))
_, pw1 = pspy_utils.naive_binning(l_pw, pixwin_l[sv1], binning_file, lmax)
tf1 *= pw1
for id_ar1, ar1 in enumerate(arrays_1):
for id_sv2, sv2 in enumerate(surveys):
arrays_2 = d["arrays_%s" % sv2]
nsplits_2 = nsplit[sv2]
if d["tf_%s" % sv2] is not None:
print("will deconvolve tf of %s" %sv2)
_, _, tf2, _ = np.loadtxt(d["tf_%s" % sv2], unpack=True)
tf2 = tf2[:len(lb)]
else:
tf2 = np.ones(len(lb))
