Skip to content

Latest commit

 

History

History
148 lines (110 loc) · 8.23 KB

README.md

File metadata and controls

148 lines (110 loc) · 8.23 KB
Raw

Map-based noise simulations based on the LAT scan-s0003 time-domain simulations

Tag: mbs-s0013-20230501

Updates

  • 2023-05-09: first release, lmax=5400 simulations.

Summary

300 realistic realizations of the LAT map noise, given the single realization of the time-domain simulations. Simulations cover the full LAT footprint to a bandlimit of lmax=5400, and utilize the directional wavelet noise model.

Noise model

This release is based on the LAT scan-s0003 time-domain simulations, available on NERSC at:

/global/cfs/cdirs/sobs/sims/scan-s0003/output/combined/processed/fejer1

These time-domain simulations implement the type-1.2.1 high-cadence 3-elevation scan-strategy (see scan-strategy repo and ICD). These maps provide a single realization of the noise in each split of each frequency band. More information is also available on the SO productdb and confluence.

Using mnms (commit d4afb9d), we generate 300 realizations of the noise using map-based methods. Specifically, we use the directional wavelet model, governed by the configured parameters.

Notes:

  • Simulations are bandlimited to lmax=5400. To save space, the simulations are stored in a pixelization that is downgraded by a factor of 4 relative to the time-domain simulations (shape=(..., 2640, 10800), 2 arcmin resolution vs. shape=(..., 10560, 43200) 0.5 arcmin resolution). See examples below for further guidance.
  • The time-domain simulations are produced in the Clenshaw-Curtis CAR variant. The map-based simulations are first built using these Clenshaw-Curtis time-domain maps as inputs. Then, their wcs is adjusted post-hoc (by <1 pixel) to match what they would have been had the time-domain simulations been produced in the Fejer1 CAR variant as specified (see this issue for more detail). We note this for the record; it does not impact users.

Available maps

Maps are available in the CAR (Fejer1 variant) pixelization. As above, the maps have a resolution of 2 arcmin. Maps are in Equatorial Coordinates, uK_CMB units, FITS format.

The maps are located here:

/global/cfs/cdirs/sobs/v4_sims/mbs/mbs-s0013-20230501/sims

and have the following naming convention:

so_scan_s0003_fdw_{bands}_lmax5400_4way_set{split_num}_noise_sim_map{sim_num:04}.fits

where bands is in [lf_f030_lf_f040, mf_f090_mf_f150, uhf_f230_uhf_f290], split_num is in [0-3] and sim_num is in [0000-0299].

Please open an issue here for any data access problems.

Available models

The covariance matrices from which simulations are drawn are available here:

/global/cfs/cdirs/sobs/v4_sims/mbs/mbs-s0013-20230501/models

Due to issues with the Clenshaw-Curtis vs. Fejer1 geometry, they do not currently support drawing additional realizations.

Metadata

The models correlate pairs of frequency bands on the same detector wafer. Therefore, the simulations are stored in a similar paired format. Thus, the shape of a given simulation file is (2, 1, 3, 2640, 10800):

  • The first axis is the frequency band
  • The second (singleton) axis follows mnms convention (reserved for map splits; since each simulation corresponds to one map split, the dimension of this axis is 1)
  • The third axis is Stokes component (I, Q, U)
  • The last axes are the map pixels

The geometry (including shape) of the maps is given in the FITS header:

SIMPLE  =                    T / conforms to FITS standard                      
BITPIX  =                  -32 / array data type                                
NAXIS   =
 5 / number of array dimensions                     
NAXIS1  =                10800                                                  
NAXIS2  =                 2640
NAXIS3  =                    3                                                  
NAXIS4  =                    1                                                  
NAXIS5  =                    2                                                  
WCSAXES =                    2 / Number of coordinate axes                      
CRPIX1  =             5400.625 / Pixel coordinate of reference point            
CRPIX2  =               1890.5 / Pixel coordinate of reference point            
CDELT1  =   -0.033333333333333 / [deg] Coordinate increment at reference point  
CDELT2  =    0.033333333333333 / [deg] Coordinate increment at reference point  
CUNIT1  = 'deg'                / Units of coordinate increment and value        
CUNIT2  = 'deg'                / Units of coordinate increment and value        
CTYPE1  = 'RA---CAR'           / Right ascension, plate caree projection        
CTYPE2  = 'DEC--CAR'           / Declination, plate caree projection            
CRVAL1  =                  0.0 / [deg] Coordinate value at reference point      
CRVAL2  =                  0.0 / [deg] Coordinate value at reference point      
LONPOLE =                  0.0 / [deg] Native longitude of celestial pole       
LATPOLE =                 90.0 / [deg] Native latitude of celestial pole        
MJDREF  =                  0.0 / [d] MJD of fiducial time                       
RADESYS = 'ICRS'               / Equatorial coordinate system                   END

Example usage

We give some minimum working examples:

import numpy as np
from pixell import enmap, curvedsky, enplot
from os.path import join

# first get some basic info like the path and the filename template
fdir = '/global/cfs/cdirs/sobs/v4_sims/mbs/mbs-s0013-20230501/sims'
fbase_template = 'so_scan_s0003_fdw_{bands}_lmax5400_4way_set{split_num}_noise_sim_map{sim_num:04}.fits'

# let's load the sim for e.g. f150, split 2, sim 123
bands = 'mf_f090_mf_f150'
band_idx = 1                # f090 is 0, f150 is 1
split_num = 2
sim_num = 123               # has 4 digits (leading 0's), but padding done for us in template

sim = enmap.read_map(
    join(fdir, fbase_template.format(
        bands=bands,
        split_num=split_num,
        sim_num=sim_num)
        ),
    sel=np.s_[band_idx, 0]  # just load f150, remove split axis
    )
print(sim.geometry, sim.dtype)

This should give:

((3, 2640, 10800), car:{cdelt:[-0.03333,0.03333],crval:[0,0],crpix:[5400.62,1890.50]}) float32

As discussed, the sim only contains noise power to the Nyquist frequency of the pixelization (for 2 arcmin pixels, lmax=5400). But what if we wanted to work with objects in map-space defined at the full resolution of the data, for example a sky mask defined at 0.5 arcmin resolution? We need to project the sim to our desired geometry.

Because the sim is bandlimited, we can do this losslessly with a spherical harmonic transforms:

# first we need the geometry of the pixelization.
# we can get this from one of the time-domain simulations
shape, wcs = enmap.read_map_geometry('/global/cfs/cdirs/sobs/sims/scan-s0003/output/combined/processed/fejer1/lat01_s25_fullfp_f150_1pass_4way_set02_map.fits')

# allocate an empty map to hold the sim at full resolution
sim_fullres = enmap.empty((*sim.shape[:-2], *shape[-2:]), wcs, sim.dtype)

# project sim to alm, then from alm to map
curvedsky.alm2map(curvedsky.map2alm(sim, lmax=5400), sim_fullres)

Finally, we can plot the sim:

# can also plot all three components in one call, but the colorbar will
# be shared for all three plots. this is visually nicer for components
# with very different dynamic range, as in this case
for pol in range(3):
    enplot.pshow(sim_fullres[pol], downgrade=32, colorbar=True, ticks=15)

Known issues

  • Large-scale TT (l < 300) has excess noise power of > 10% in f030-f090, see these slides.

Feedback

If anything looks suspicious in the simulations, please do not hesitate to open an issue here.