simonsobs / map_based_simulations

Map based simulations for the Simons Observatory
4 stars 1 forks source link

Simulations with ideal bandpasses #12

Closed zonca closed 5 years ago

zonca commented 5 years ago

For the next TOD simulation, we would like to use top-hat bandpasses. All the map based simulations we have are delta-bandpass, I would like to run:

with top-hat bandpasses.

2 questions:

mhasself commented 5 years ago

On the first question, please go through @smsimon and the pwg:exm group (they are supposed to track the use of such parameters).

On the second question, I think I disagree with the consensus reached on that S4 page. The purpose of going from delta function bands to non-zero bandwidth is fully achieved by having almost any bandpass function with realistic width, because the point is just to make sure that you can handle the fact that different spectra pick-up differently across the band. The added realism of modeling idealized feed response (nu^-2) is, if I understand it, completely uninteresting.

Furthermore, saying that the bandpass is "top hat", and then defining that bandpass as nu^-2, seems like obfuscation. When we say "top hat", we conventionally mean a function that is 0, then 1 for a while, then 0. With the nu^-2 weight included in the definition of top-hat, if someone asks what the central frequency of the band-pass is, you have to say, "uh, what frequency?" because even if you've "centered" your top hat at nu_0, i.e. it's a band that's 0 except for inside [nu_0 - delta/2, nu_0 + delta/2]... the centroid (and thus plausibly the central frequency) is... and forgive-me-if-I-get-this-wrong-even-if-it-supports-my-point-to-do-so, ...:

ln( (nu + delta) / (nu - delta) ) * (nu^2 - delta^2) / 2 delta.

Blech.

zonca commented 5 years ago

thanks @mhasself would you do a top-hat response in K_RJ or in K_CMB?

smsimon commented 5 years ago

@zonca The bandpasses are here: http://simonsobservatory.wikidot.com/detbandpasses The bolded designs are the final designs.

zonca commented 5 years ago

Thanks @smsimon, I should do the weighting in K_RJ, right?

mhasself commented 5 years ago

@smsimon I'd like the top-hat approximations to the real bands to be "owned" by exm. So whatever the results are, let's make sure they're tracked on the exm page (and referenced by sim groups).

@zonca I think the right weighting is in spectral radiance (SR), i.e. Jy/sr. The K_RJ has a pesky nu^2 (but not the same nu^2 as before) in the conversion to I_nu.

zonca commented 5 years ago

@smsimon to provide a table of bandwidths

zonca commented 5 years ago

ok, @smsimon produced a table of center frequencies and bandwidths, @keskitalo suggested they should go into sotodlib, anyone volunteers to create a pull request to sotodlib for those? Or point me where they should go.

I can then import sotodlib from mapsims to run the map based simulations.

smsimon commented 5 years ago

@zonca done

zonca commented 5 years ago

very good thanks, reference: https://github.com/simonsobs/sotodlib/pull/22 I have everything I need to proceed with this.

zonca commented 5 years ago

results available, see https://github.com/simonsobs/map_based_simulations/pull/20