# this notebook plots the autospectrum and correlation noise pseudospectrum,
# averaged over splits, for a given array (pair), frequency (pair),
# and polarization (pair) of either act_dr6.01 data (dr6v3) or acr_dr6.02 data 
# (dr6v4). the latter maps were used in the dr6 power spectrum analysis. the
# mask in which these spectra are measured is large and smooth, so the
# shape of the power spectrum will be qualitatively very similar.

# IMPORTANT NOTES -- PLEASE READ

# IMPORTANT NOTE 1: these spectra are really only useful for forecasting
# studies -- THEY CANNOT BE USED FOR COSMOLOGICAL ANALYSIS. as described in
# https://arxiv.org/abs/2303.04180, isotropic power spectra are not a realistic
# model of the ACT noise. these spectra provide a reference for the general 
# "shape" of noise spectra that may be expected for large aperture telescopes
# in Chile

# IMPORTANT NOTE 2: the spectra have all been normalized such that, after adding
# in inverse-quadrature over their split axes, they are unity at high ell in
# the noise TT power spectrum. this sum represents the noise power spectrum of
# the map that has been formed by a simple average of the splits. in other words
# the per-split noise power spectrum in TT has a value of NSPLITS at high ell 
# and approximately 2*NSPLITS for EE and BB noise. 

# IMPORTANT NOTE 3: this means that it is up to the user to normalize the 
# white-noise level of the spectra for their desired forecasting application. 
# for constant-depth white noise of X uK-arcmin, the power spectrum is simply
# C = (X * pi / 10800)^2 (in uK^2-steradians). so users should multiply 
# the spectra by C, then take the mean over the split axis, and further divide
# by NSPLITS to get a representative isotropic full-sky noise power spectrum
# with the shape of the corresponding ACT DR6 array-band

# IMPORTANT NOTE 4: the dr6.01 and dr6.02 maps have different pixel window 
# functions, see https://arxiv.org/abs/2503.14451. the choice to normalize at 
# the highest-ell means that comparing dr6.01 to dr6.02 directly at lower ells
# will be a little not-apples-to-apples. 

# IMPORTANT NOTE 5: the correlation noise spectra are independent of any change
# in the normalization.