search

CMB-S4 / s4mapbasedsims

CMB-S4 map based simulations
3 stars 1 forks source link

DC-0 Validation spectra: Synchrotron #29

Closed zonca closed 1 year ago

zonca commented 1 year ago

@giuspugl the other issue I noted, also in https://github.com/simonsobs/map_based_simulations/issues/44#issuecomment-1481586936 is spikes at high-ell in the Synchrotron maps.

image image image

Should check if they are still there when masking.

giuspugl commented 1 year ago

hello, not sure what it might cause this honestly . those are spikes at ell= 3000 with overtunes at its integer multiples, i.e. 6000 and 9000. can it be pixel window function of the template?

giuspugl commented 1 year ago

have identified the cause. it 's not related to the way we inject small scales, but it's due to the fact that we currently employ the real observations inside the Galactic plane and do not inject any small scale in the mask below : image There might be something silly that is making the spectra to explode at around ell=3000, see the light blue line below image It seems that this might have overtunes at integer multipoles, not sure why ... Indeed if i don't apply the mask and inject small scales also inside the Gal. plane, we don't have that jump anymore.

I've tried to narrow the mask above so that we won't have this inconsistency, but it seems something related to the Galactic center, or some pixel effect related to the template input maps .

I would propose to give up on including real observations in the Galactic plane small scales , thoughts?

zonca commented 1 year ago

@giuspugl I agree, it also simplifies the dust models. Alternatively what if we apply just a tiny bit of smoothing to the real data to smooth out pixel effects? Is it confirmed that the problem is ud_grade?

It would be good to have feedback from @seclark and @brandonshensley.

giuspugl commented 1 year ago

Alternatively what if we apply just a tiny bit of smoothing to the real data to smooth out pixel effects?

well, this seems to me a bit as an epicycle approach. We need to convolve the template by a beam of ~5' to avoid that spectra explodes at ell=2000. Although it seems totally feasible , it 's true that few users would look at the ell=2000 power spectrum of Gal. plane...

seclark commented 1 year ago

Recording that we discussed this on the PanEx call today and plan to not make the change for now, but to note it as a caveat to the models in the current paper. We do not expect that this will affect typical use cases since these models are not designed to reproduce the power spectrum of the Galactic center, and we will address this in the future as we update our treatment of small scale structure in bright regions anyway.

zonca commented 1 year ago

documented in PySM, see the linked issues.