Signal + noise composition interface

[msyriac]

What is the recommended code interface to make a joint signal + noise sim?
I would strongly advocate that the default noise sim is not beam deconvolved. Rather, a composite sim of signal and noise should look like:
beam(signal) + noise.
This makes the result look like what is in a real data map, and makes the noise map well behaved.

So, it would be good to have a signal + noise composition interface that does the following:

bandpass_integrate(beam(cmb + galactic + extragalactic)) + noise

e.g.

sgen = SOSim()
sgen.get_sim(band,bandpass_integrate=True,beam=True,include_noise=True,include_galactic=True,include_extragalactic=True)

This probably exists somewhere already?

[ajvanengelen]

Agreed - as you know that is the approach we have taken for the ACT sims.

…

________________________________ From: Mathew S. Madhavacheril <notifications@github.com> Sent: 13 January 2020 09:23 To: simonsobs/mapsims <mapsims@noreply.github.com> Cc: Subscribed <subscribed@noreply.github.com> Subject: [simonsobs/mapsims] Signal + noise composition interface (#31) What is the recommended code interface to make a joint signal + noise sim? I would strongly advocate that the default noise sim is not beam deconvolved. Rather, a composite sim of signal and noise should look like: beam(signal) + noise. This makes the result look like what is in a real data map, and makes the noise map well behaved. So, it would be good to have a signal + noise composition interface that does the following: bandpass_integrate(beam(cmb + galactic + extragalactic)) + noise e.g. sgen = SOSim() sgen.get_sim(band,bandpass_integrate=True,beam=True,include_noise=True,include_galactic=True,include_extragalactic=True) This probably exists somewhere already? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_simonsobs_mapsims_issues_31-3Femail-5Fsource-3Dnotifications-26email-5Ftoken-3DADYBOHFXZ6E734YI74IEKDTQ5SIONA5CNFSM4KGE65Q2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IFZVV2A&d=DwMCaQ&c=l45AxH-kUV29SRQusp9vYR0n1GycN4_2jInuKy6zbqQ&r=mauVTJTX0aydlVkY5fcx1lQbgv8QnPpRaXAD8fKw_Zs&m=vbzX_d-NeD7q9GhVzFjE6RpR8c-QUXhrCLDxZab3jFs&s=-7-aWYv9QIGoVI3__qaZi3Pwf1cEyZ2JUjd1sukF-3o&e=>, or unsubscribe<https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_ADYBOHB63BID3REERB73KFDQ5SIONANCNFSM4KGE65QQ&d=DwMCaQ&c=l45AxH-kUV29SRQusp9vYR0n1GycN4_2jInuKy6zbqQ&r=mauVTJTX0aydlVkY5fcx1lQbgv8QnPpRaXAD8fKw_Zs&m=vbzX_d-NeD7q9GhVzFjE6RpR8c-QUXhrCLDxZab3jFs&s=WxSDHu0TME3W4wuwlCUter1gFVm00GEQMZAC4pKMZrE&e=>.

[zonca]

this is what MapSim does: https://mapsims.readthedocs.io/en/latest/api/mapsims.MapSim.html#mapsims.MapSim

it can start from a configuration file or from classes, and also has a command line tool, see https://mapsims.readthedocs.io/en/latest/usage.html

[msyriac]

Couple of things I see here:

1. We need to add CAR support
2. It doesn't look like beam convolution is ever done anywhere, instead the noise needs to be obtained with "beam_correction=True"

Is (2) correct?

[zonca]

We ran the last set of noise simulations with no beam correction, see https://github.com/simonsobs/map_based_simulations/blob/master/201906_noise_no_lowell/noise.toml

The next set should instead have it included? simonsobs/map_based_simulations#21

[msyriac]

I think it always makes sense to save or return noise simulations with no beam deconvolution. And similarly signal sims should always be beam convolved. This makes the outputs well behaved / sane. The user can then simply treat:
signal + noise
as a representation of the kind of map SO mapmakers will provide us.

[zonca]

so we will run also the new sims with beam_correction=False?

[msyriac]

Yes, but are you also separately running and saving signal sims? Will you apply beam convolution to those?

[zonca]

this release will be only noise, but previous and next signal releases all have beam convolution.

[zonca]

@msyriac @ajvanengelen here is how it works now:

https://gist.github.com/zonca/5cc4e2d27cbd8422f16e1af3e5e29bfd

It would be nicer to have sim.execute(seed) but I wouldn't want to make too many changes to the pipeline. So I think this is easy enough to use, suggestions?

[zonca]

After this release I'll work on the foreground simulations again, I think at that point it will be easier to test maps+noise simulations and work on improving the interface. I would like to check if I can initialize all the foreground and noise classes without providing NSIDE, channel, and num, then call:

sim.execute(channel="ST0", nside=64, num=10)