README.md

Reran of realistic noise version 3.1.1 with MSS1 hitmaps

Tag: mbs-s0011-20200618

Release date 18 Jun 2020

New noise release using version 3.1.1 of the noise curves from so_noise_models.

This release fixes an issue with the tube scaling present in the previous 202002_noise release.

Compared to previous releases:

• Updated noise spectra which include a fix for the atmosphere parameters
• Includes cross-correlations between the 2 bands in the same dichroic tube
• Uses new hitmaps from the MSS-0001 time domain simulations
• Supports noise splits
• Variable Nside based on channels
• Released in HEALPix and (later) in CAR

Input components

All maps are in uK_CMB, IQU, single precision (float32)

• For the exact configuration of the noise simulations refer to the noise.toml configuration file available in this repository. For all available options refer to the SONoiseSimulator class in mapsims.

Available maps

Only 1 realizations of the full mission and 1 realization of 4 splits are made available as maps.

Reference frame for noise maps is assumed to be Equatorial.

Location at NERSC:

/project/projectdirs/sobs/v4_sims/mbs/202006_noise

The naming convention is:

{num:04d}/simonsobs_{content}_uKCMB_{tube}_{band}_nside{nside}_{num:04d}_{split}_of_{nsplits}.fits"

where:

• content is noise

• num is the realization number (seed) and currently only 0

• tube is LT0-LT6 or ST0-ST3

• band is one of the available bandpasses (available inside mapsims.so_utils):

    bands = ("LF1", "LF2", "MFF1", "MFF2", "MFS1", "MFS2", "UHF1", "UHF2")
    frequencies = (27, 39, 93, 145, 93, 145, 225, 280)
  

• nsplits is the number of splits, 1 for full mission, 4 for the splits, split is 1 based split ID.

For example: 0000/simonsobs_noise_uKCMB_ST2_MFS2_nside512_0000_1_of_1.fits

Total disk space used is 548 GB for one realization of full mission and splits.

Backed up to tape in ~zonca/sobs/mbs/202002_noise.

Generate other realizations on the fly

More noise realizations should be generated on-the-fly and never written to disk using the mapsims package, check the documentation at https://mapsims.readthedocs.io for installation instructions.

Then check the run_mapsims_onthefly.ipynb notebook. See also an executed version of the notebook on Gist.

Hitmaps

Hitmaps are used un-normalized to scale the pixel-by-pixel noise. They are the first split (of 20) of the MSS-0001 time domain simulations, see their documentation on the Simons Observatory wiki (restricted) They are available at NERSC in:

/global/project/projectdirs/sobs/www/so_mapsims_data/v0.2

Or via web through NERSC

Or they can be accessed through the get_hitmaps method of mapsims.SONoiseSimulator.

Software

• The noise component and the runner script are available in the mapsims package, version 2.3.0, see the documentation

Spectra plots

For validation purposes, see a comparison between the input spectra as generated by so_noise_models and a power spectrum (simple anafast with no mask deconvolution).

The input spectra assume uniform coverage, so it is expected that the simulated spectra with non-uniform hits are a bit higher overall. I don't know exactly how to quantify this or correct for it, suggestions welcome.

Spectra were generated with the compute_cl_noise_full.py and compute_cl_noise_splits.py scripts and plotted with plot_cl_noise_LAT_full.ipynb and plot_cl_noise_LAT_splits.ipynb

LAT Temperature

Blue is the input spectra by so_noise_models, orange and green are TT and BB/EE spectra. All plots have the same vertical scale.

LAT Polarization

LAT Temperature splits

We ran 4-way splits, noise power spectra are 4 times higher

LAT Polarization splits

SAT Temperature

Blue is the input spectra by so_noise_models, orange and green are TT and BB/EE spectra. All plots have the same vertical scale.

SAT Polarization