Increase resolution for next runs

[zonca]

I plan to switch to nside 1024 for SAT and 8192 for LAT in the next future, as that agrees with the data interface document.

Just keeping track of this for now, let's discuss in August/September

[zonca]

from f2f, no need for 1024 for SAT, keep 512

[zonca]

@msyriac can you post here a proposal of a varying NSIDE/resolution for each channel/telescope?

[msyriac]

This is my recommendation for the nside (right) based on the FWHM arcmin (left)

SAT
91.0     128
63.0     128
30.0     256
17.0     512
11.0     1024
9.0  1024
LAT
7.4  1024
5.1  2048
2.2  4096
1.4  8192
1.0  8192
0.9  8192

Pinging @mhasself @damonge @jcolinhill @dpole @thibautlouis

[zonca]

thanks @msyriac can you please also add CAR maps resolution?

[jcolinhill]

overall looks good, but I would prefer if the LAT 30 GHz channel (7.4 arcmin FWHM) is at Nside=2048. Planck 143 GHz is at roughly the same resolution, and they use Nside=2048.

[msyriac]

Updated recommendation including CAR pixel size in arcmin, and incorporating Colin's request:

91.0     128     32.00
63.0     128     32.00
30.0     256     15.00
17.0     512     8.00
11.0     1024    4.50
9.0  1024    4.50
7.4  2048    2.00
5.1  2048    2.00
2.2  4096    0.50
1.4  8192    0.50
1.0  8192    0.50
0.9  8192    0.50

@amaurea should probably look at this too!

[mhasself]

Wouldn't it be easier if the CAR resolutions scale by powers of 2, tied to NSIDE? I.e. insist that:

CDELT = C / NSIDE

Mat's suggestion is close to C = 64 degrees... resolutions would be (30,15,7.5, ..., 0.46875) arcminutes. If you prefer (32,16,8,...,0.5), set C = (1048 deg/15).

[msyriac]

Since the CAR maps will likely be used by source/object/cluster people, I would prefer to have the LAT high frequencies to have identical pixelization (as we do in ACT currently), at the cost of oversampling a bit at low freq and undersampling a bit at high freq. That said, powers of two are good, so here's an update:

fwhm     NSIDE  CAR resol
91.0     128     32.00
63.0     128     32.00
30.0     256     16.00
17.0     512     8.00
11.0     1024    4.00
9.0  1024    4.00
7.4  2048    2.00
5.1  2048    2.00
2.2  4096    0.50
1.4  8192    0.50
1.0  8192    0.50
0.9  8192    0.50

[mattyowl]

Yeah - for cluster stuff it makes life a lot easier if all the maps have the same pixelisation. I think we need NSIDE=8192.

[zonca]

thanks @mattyowl, would you like nside 8192 also for the channel at 2.2 arcmin?

[mattyowl]

Yes please, I think so. It's less critical there though than for the higher frequencies.

[msyriac]

Updating based on Matt's suggestion and Simone's array tags:

tel_freq    fwhm    NSIDE   CAR resol
sat_027         91.0    128     32.00
sat_039         63.0    128     32.00
sat_093     30.0    256     16.00
sat_145     17.0    512     8.00
sat_225     11.0    1024    4.00
sat_280         9.0 1024    4.00
lat_027         7.4 2048    2.00
lat_039         5.1 2048    2.00
lat_093         2.2 8192    0.50
lat_145         1.4 8192    0.50
lat_225         1.0 8192    0.50
lat_280         0.9 8192    0.50

[zonca]

Thanks @mattyowl and @msyriac, I think we should also think about ell_max for the harmonic expansions, (used both for reprojections, see #13, or beam smoothing), what about 3*nside-1 for nside<8192 and 2*nside for nside=8192? Or are we interested in ell>16384?

[aiolasimone]

It becomes a bit tricky with localized features, such as point sources or dust clumps close to the galaxy. We should check that 16384 is enough to make ringing around point sources negligible compared to the noise level. Map-space operations like source and cluster finding, staking, and such may care a lot about this.

[zonca]

@aiolasimone who can I ask about operations on point sources?

[msyriac]

Let's keep it at 3*nside-1 throughout if possible?

[zonca]

I'll do some benchmarks and check file sizes, I'll report back here.

[aiolasimone]

@msyriac and @chinoney, it would be amazing if the table above and the WCS definition discussed here #13 get added to the tod2map_docs. I would also include a column with file size for single precision (just a suggestion).

Let's try to have a look at that document before the AGWs all-day meeting in Nov.

[msyriac]

@amaurea and I chatted about additionally requiring that the CAR maps have integer number of pixels in both directions, and having the different maps be convertible between different pixelizations through integer multiples. We came up with the following, which oversamples the SATs a bit further:

tel_freq    fwhm    NSIDE   CAR resol
sat_027         91.0    128     12.00
sat_039         63.0    128     12.00
sat_093     30.0    256     12.00
sat_145     17.0    512     4.00
sat_225     11.0    1024    4.00
sat_280         9.0 1024    4.00
lat_027         7.4 2048    2.00
lat_039         5.1 2048    2.00
lat_093         2.2 8192    0.50
lat_145         1.4 8192    0.50
lat_225         1.0 8192    0.50
lat_280         0.9 8192    0.50

@mhasself @aiolasimone I'll be pushing this to tod2mapdocs soon

[zonca]

@msyriac what if we increase resolution on 93GHz to the same of 145GHz? so that when they are simulated in a dichroic tube they have the same resolution.

tel_freq    fwhm    NSIDE   CAR resol
sat_027         91.0    128     12.00
sat_039         63.0    128     12.00
sat_093     30.0    512     4.00
sat_145     17.0    512     4.00
sat_225     11.0    1024    4.00
sat_280         9.0 1024    4.00
lat_027         7.4 2048    2.00
lat_039         5.1 2048    2.00
lat_093         2.2 8192    0.50
lat_145         1.4 8192    0.50
lat_225         1.0 8192    0.50
lat_280         0.9 8192    0.50

[msyriac]

Yes, that makes sense. Let's go with that.

[zonca]

ok, this is implemented in mapsims and is the default is nside is set to None, see https://github.com/simonsobs/mapsims/blob/master/mapsims/data/so_default_resolution.csv

First release using this is https://github.com/simonsobs/map_based_simulations/tree/master/202002_noise

Next also a foreground simulation natively at these NSIDEs