I am generating random healpix maps from an input angular power spectrum Cl. If I use healpy.synalm, then healpy.alm2map, and finally test the map by running healpy.anafast on the generated map, the output and input power spectra do not agree, especially at high l, the output power spectrum is above the input (See plot produced by code below). If I directly use healpy.synfast, I get an output power spectrum that agrees with the input. The same applies if I use the alms from healpy.synfast and generate a map from the synfast alms using healpy.alm2map. When I look into the source code of synfast, it seems to just call synalm and alm2map, so I don't understand, why their results disagree. My test code looks like this:
import numpy as np
import matplotlib.pyplot as plt
import classy
import healpy as hp
NSIDE = 32
A_s=2.3e-9
n_s=0.9624
h=0.6711
omega_b=0.022068
omega_cdm=0.12029
params = {
'output': 'dCl, mPk',
'A_s': A_s,
'n_s': n_s,
'h': h,
'omega_b': omega_b,
'omega_cdm': omega_cdm,
'selection_mean': '0.55',
'selection_magnification_bias_analytic': 'yes',
'selection_bias_analytic': 'yes',
'selection_dNdz_evolution_analytic': 'yes'}
cosmo = classy.Class()
cosmo.set(params)
cosmo.compute()
theory_cl=cosmo.density_cl()['dd']
data_map,data_alm=hp.synfast(theory_cl[0],NSIDE,alm=True)
data_cl=hp.anafast(data_map)
plt.plot(np.arange(len(data_cl)),data_cl,label="synfast")
data_map=hp.alm2map(data_alm,NSIDE)
data_cl=hp.anafast(data_map)
plt.plot(np.arange(len(data_cl)),data_cl,label="synfast using alm")
data_alm=hp.synalm(theory_cl[0])
data_map=hp.alm2map(data_alm,NSIDE)
data_cl=hp.anafast(data_map)
plt.plot(np.arange(len(data_cl)),data_cl,label="synalm")
plt.plot(np.arange(min(len(data_cl),len(theory_cl[0]))),theory_cl[0][:len(data_cl)],label="Theory")
plt.xlabel(r'$\ell$')
plt.ylabel(r'$C_\ell$')
plt.legend()
plt.show()
The offset becomes larger for lower NSIDE.
Thank you very much for your help.
