1. Pipeline tasks¶
1.1. Conventions¶
We will store one-dimensional power spectra in 1d numpy arrays whose index corresponds to the multipole (so the first indices starting with zero store multipoles 0,1,2…etc.).
All CMB power spectra are in units of \((\mu K-{\rm rad})^2\) and do not contain any factors of \(2 \pi\) or \(\ell (\ell+1)\).
We use lensing potential internally (not convergence).
The
falafelcode returns unnormalized quadratic estimators andtempurareturns full-sky normalizations. Withinsolenspipe, functions that return quadratic estimator reconstructions will return normalized estimators.We use the estimator normalization convention that results in the noise power \(N_L^{\phi \phi}\) for an optimal estimator being equal to its normalization \(A_L^{\phi}\).
1.2. Preparation¶
1.2.1. Data access¶
The ACT and SO map-makers provide sets of maps with mutually exclusive data; each set consists of completely independent TOD samples. This constitutes some splitting of the data. For historical reasons, we refer to each such set as an `array’. This terminology is derived from the fact that the TODs are primarily split by which detector array they originate from, though since 2015, ACTpol and its successors (including Advanced ACT and SO) use multi-chroic arrays, which means each hardware array will provide us multiple (almost always two) `array’ map sets even in the same season/year and region. We will now stop using quotes around `array’ under the understanding that it applies to some unit of splitting closely related to what is used in ACT.
Within ACT, these arrays typically come from some region or scan (though since 2016 there has primarily been just a wide scan each for day and night) for a particular season and particular frequency band (since the ACTpol PA3 array, one of two within a dichroic hardware array). For SO, under the current simulation design, there will be two array maps for each optics tube because of the dichroic hardware array in the tube.
We will also be combining with Planck, for which we define a Planck array as a particular frequency band, reprojected to the CAR pixelization and subtracted of sources (see planck_reproj).
1.2.2. Planck reprojection¶
1.3. Simulation¶
1.4. Co-addition¶
1.5. Filtering and data spectra¶
This stage introduces a choice of cosmology.
In this stage, we also calculate the power spectra of the filtered maps and store these. These will be used as inputs for approximate bias subtraction. The spectra can also be used to verify accuracy of the simulations in a manner consistent with the blinding policy.
1.6. Normalization and theory N0¶
This stage also assumes a choice of cosmology (which is more important downstream), but we constrain it to be the same cosmology used in the filtering stage.