Discrepancy between d12 and color corrected HFI353

[erussier]

Hello, I plotted the PySM d12 dust power spectra and compared to the HFI353GHz. There seems to be an offset at large scales that we do not find when comparing the MKD map (without color correction) from here (https://apc.u-paris.fr/Downloads/Planck/PSM-Data/MKD-MODEL-PySM-2048/skyinbands/HFI/detector_F353/) and the HFI353GHz without color correction. This is what I get for Emodes and Bmodes. In principle, the PSM MKD and HFI353 without color correction should be compatible and d12 and color corrected HFI353 should also be compatible at least on the largest scales. Here is the jupyter notebook where I did this comparison: comp_dust_PS_d12_MKD_map.ipynb

[zonca]

thanks @erussier, can you please repeat the test at Nside 2048 and update the notebook?

[zonca]

also to test the implementation I am comparing with this reference notebook with a simple analysis ran at Nside 8: https://gist.github.com/brandonshensley/1dda561b245cda6497eb1b5ce2b7c267

can you check there if you find any errors?

[zonca]

@erussier can you please adapt the notebook to run at NSIDE 2048 instead of 512? In the notebook, when you write outputs, can you write to a local folder (`"./output") so the notebook can be executed by other people more easily? Don't worry about the other thing I suggested.

[erussier]

Hi @zonca, I ran the notebook at NSIDE 2048 except for the MKD map since the native NSIDE of the MKD map I have is 512. So if you want to run this notebook at NSIDE 2048, please change the NSIDE in the section "Plot map of Planck MKD model" and the path to the MKD map. Here it is: https://gist.github.com/erussier/89b656bba5c65331b5764d39754b7d38

[zonca]

the MKD is the most important map to have at 2048, @delabrou ran the MKD model at Nside 2048, those are the maps that I am using as inputs to PySM. I think the easiest for comparison would be to have MKD at Nside 2048 at 353GHz include all layers computed using those Nside 2048 templates and compare it directly to the output of PySM 3, the maps should be identical.

[erussier]

Hi @zonca, I found the MKD map at NSIDE = 2048 and indeed we get the same results between the MKD model at 353 GHz and d12. I guess we can close this issue.

[zonca]

Good news! Thanks

[erussier]

In fact, the mkd map I was using at NSIDE 2048 and used in the PySM pipeline, was generated without the color correction implemented. We need to multiply the mkd map by 0.92 to take into account the color correction. Probably the best thing to do is to change the input maps. @delabrou will generate new maps to use as inputs for PySM.

[zonca]

Sure, I can then replace the inputs

[zonca]

@erussier any news on the new maps?

[zonca]

@erussier @delabru what if I just apply the 0.911 factor (see https://github.com/galsci/pysm/issues/99#issue-1068953759) inside PySM?

[erussier]

Sorry for the late reply, we think that is okay to apply the 0.911 factor inside PySM, thank you.

[zonca]

@erussier @delabru ok, I implemented it in:

https://github.com/galsci/pysm/pull/141

can you please confirm that I just multiply the IQU template, I don't change anything in the spectral index map and the black body temperature?

[delabrou]

Yes, I confirm. You multiply all IQU templates by the colour correction factor (in the PSM I use 0.92).

[delabrou]

Do not do anything to the temperature and spectral index maps.