Examples of Weighting Across Frequency Plots

[CarloBaccigalupi]

Examples of plots showing the weighiting of a given component separation procedure of the various frequency channels across frequency are Figures B.2, C.2, C.3 of the Planck 2018 Component Separation Paper.

In B.2, the sky average of the weights for Needlet are plotted across frequencies, in different colors for different bands in l, showing the different weighting of the foregrounds across the l space, with the associated huge variation of weights.

In C.2 and C.3, methods fitting in the l space, and l by l, show the weightings across the l space for the different frequency channels.

[damonge]

The weights for the C_ell pipeline could be obtained using Eq. 2 of the paper above

[CarloBaccigalupi]

Grazie!!!

---

From: David Alonso @.> Sent: Wednesday, May 18, 2022 1:14 PM To: simonsobs/BBPipe_paper @.> Cc: Carlo Baccigalupi @.>; Author @.> Subject: Re: [simonsobs/BBPipe_paper] Examples of Weighting Across Frequency Plots (Issue #1)

The weights for the C_ell pipeline could be obtained using Eq. 2 of the paper above [smica_weights]https://user-images.githubusercontent.com/8809493/169026039-89055f29-9d74-40ef-a451-d107849292b2.png

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1129879099, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACVV4LDPZWB7GDO5CSRVE6TVKTGHTANCNFSM5WB655HQ. You are receiving this because you authored the thread.Message ID: @.***>

[CarloBaccigalupi]

Very interesting to study the channel relevances!!!

---

From: Carlo Baccigalupi Sent: Wednesday, May 18, 2022 2:48 PM To: David Alonso @.>; simonsobs/BBPipe_paper @.>; simonsobs/BBPipe_paper @.> Cc: Author @.> Subject: Re: [simonsobs/BBPipe_paper] Examples of Weighting Across Frequency Plots (Issue #1)

Grazie!!!

---

From: David Alonso @.> Sent: Wednesday, May 18, 2022 1:14 PM To: simonsobs/BBPipe_paper @.> Cc: Carlo Baccigalupi @.>; Author @.> Subject: Re: [simonsobs/BBPipe_paper] Examples of Weighting Across Frequency Plots (Issue #1)

The weights for the C_ell pipeline could be obtained using Eq. 2 of the paper above [smica_weights]https://user-images.githubusercontent.com/8809493/169026039-89055f29-9d74-40ef-a451-d107849292b2.png

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1129879099, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACVV4LDPZWB7GDO5CSRVE6TVKTGHTANCNFSM5WB655HQ. You are receiving this because you authored the thread.Message ID: @.***>

[kwolz]

These are the results for the Cross-CL pipeline. The first plot contains just the average over the whole map, allowing to combine all pipeline results in a single plot like so. map_averaged_bandpass_weights_d1s1.pdf The second plot deconstructs the bandpass weights as a function of multipole (solid line is goal rms + optimistic 1/f noise, dashed line is baseline rms + pessimistic 1/f noise). bandpass_weights_cross_cl_d1s1.pdf

[CarloBaccigalupi]

Wow Kevin,

this are really interesting results, and give feedback on how optimal the distribution of frequencies and sensitivities actually is. We certainly need some time to digest these results, but at a first glance, the first plot shows that no weight is small, which is good, as it tells that the B solution is absorbing good information for all frequency channels (it wasn't the case for Planck for instance, in one of the plots on the githab you could see how the 30 GHz, despite of the very good location, was actually not important in the solution because of limited sensitivity). And second, the weights look veeeeeeeeeeeeeeery similar across methods, validating the whole apparatus.

The second plot is also very interesting, potentially showing a (rather) flat behavior over multipoles...reflecting of course the sky setup...with the exception of the lower resolution channels dying out in weighting the CMB solution at high multipoles...

Very well done!!!

Carlo

---

From: Kevin Wolz @.> Sent: Friday, June 17, 2022 6:06 PM To: simonsobs/BBPipe_paper @.> Cc: Carlo Baccigalupi @.>; Author @.> Subject: Re: [simonsobs/BBPipe_paper] Examples of Weighting Across Frequency Plots (Issue #1)

These are the results for the Cross-CL pipeline. The first plot contains just the average over all multipoles which we could nicely combine in a single plot like so. map_averaged_bandpass_weights_d1s1.pdfhttps://github.com/simonsobs/BBPipe_paper/files/8929021/map_averaged_bandpass_weights_d1s1.pdf The second plot deconstructs the bandpass weights as a function of multipole. bandpass_weights_cross_cl_d1s1.pdfhttps://github.com/simonsobs/BBPipe_paper/files/8929024/bandpass_weights_cross_cl_d1s1.pdf

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1159023620, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACVV4LGKMC6FI5URTCSJMW3VPSO6RANCNFSM5WB655HQ. You are receiving this because you authored the thread.Message ID: @.***>

[kwolz]

Thanks Carlo! I should mention that the first plot is preliminary and does not contain the results from pipelines B and C yet, it was just a copy and paste of the first one (with shifts in the x axis to help the reader). This being said, I agree that these visualizations may allow us to have a better understanding of our component separation methods!

On 17 Jun 2022, at 18:17, Carlo Baccigalupi @.***> wrote:



Wow Kevin,

this are really interesting results, and give feedback on how optimal the distribution of frequencies and sensitivities actually is. We certainly need some time to digest these results, but at a first glance, the first plot shows that no weight is small, which is good, as it tells that the B solution is absorbing good information for all frequency channels (it wasn't the case for Planck for instance, in one of the plots on the githab you could see how the 30 GHz, despite of the very good location, was actually not important in the solution because of limited sensitivity). And second, the weights look veeeeeeeeeeeeeeery similar across methods, validating the whole apparatus.

The second plot is also very interesting, potentially showing a (rather) flat behavior over multipoles...reflecting of course the sky setup...with the exception of the lower resolution channels dying out in weighting the CMB solution at high multipoles...

Very well done!!!

Carlo

---

From: Kevin Wolz @.> Sent: Friday, June 17, 2022 6:06 PM To: simonsobs/BBPipe_paper @.> Cc: Carlo Baccigalupi @.>; Author @.> Subject: Re: [simonsobs/BBPipe_paper] Examples of Weighting Across Frequency Plots (Issue #1)

These are the results for the Cross-CL pipeline. The first plot contains just the average over all multipoles which we could nicely combine in a single plot like so. map_averaged_bandpass_weights_d1s1.pdfhttps://github.com/simonsobs/BBPipe_paper/files/8929021/map_averaged_bandpass_weights_d1s1.pdf The second plot deconstructs the bandpass weights as a function of multipole. bandpass_weights_cross_cl_d1s1.pdfhttps://github.com/simonsobs/BBPipe_paper/files/8929024/bandpass_weights_cross_cl_d1s1.pdf

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1159023620, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACVV4LGKMC6FI5URTCSJMW3VPSO6RANCNFSM5WB655HQ. You are receiving this because you authored the thread.Message ID: @.***>

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1159033003, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ARU5ARHRABKHV53KKOADJN3VPSQLBANCNFSM5WB655HQ. You are receiving this because you commented.Message ID: @.***>

[CarloBaccigalupi]

Okok Kevin thank you, weights did look to similar, indeed.

In the paper, I would certainly advocate the inclusion of the weighting across l...indeed, as you see on the left of the plot, and as it is comprehensible given the foreground power rising at low l, the low frequency channels in particular are essential for the cleaning. They then are even contributing to the CMB solution (39 GHz) and that has to be checked against the foreground models, as for more complex synchrotron models, with reasonable power at intermediate and high ls, the cleaning could be absorbed by the 39 GHz also at l>100 or so.

It would be extremely interesting to see the weighting also for the other pipelines!

Cheers, Carlo

---

From: Kevin Wolz @.> Sent: Friday, June 17, 2022 6:31 PM To: simonsobs/BBPipe_paper @.> Cc: Carlo Baccigalupi @.>; Author @.> Subject: Re: [simonsobs/BBPipe_paper] Examples of Weighting Across Frequency Plots (Issue #1)

Thanks Carlo! I should mention that the first plot is preliminary and does not contain the results from pipelines B and C yet, it was just a copy and paste of the first one (with shifts in the x axis to help the reader). This being said, I agree that these visualizations may allow us to have a better understanding of our component separation methods!

On 17 Jun 2022, at 18:17, Carlo Baccigalupi @.***> wrote:



Wow Kevin,

this are really interesting results, and give feedback on how optimal the distribution of frequencies and sensitivities actually is. We certainly need some time to digest these results, but at a first glance, the first plot shows that no weight is small, which is good, as it tells that the B solution is absorbing good information for all frequency channels (it wasn't the case for Planck for instance, in one of the plots on the githab you could see how the 30 GHz, despite of the very good location, was actually not important in the solution because of limited sensitivity). And second, the weights look veeeeeeeeeeeeeeery similar across methods, validating the whole apparatus.

The second plot is also very interesting, potentially showing a (rather) flat behavior over multipoles...reflecting of course the sky setup...with the exception of the lower resolution channels dying out in weighting the CMB solution at high multipoles...

Very well done!!!

Carlo

---

From: Kevin Wolz @.> Sent: Friday, June 17, 2022 6:06 PM To: simonsobs/BBPipe_paper @.> Cc: Carlo Baccigalupi @.>; Author @.> Subject: Re: [simonsobs/BBPipe_paper] Examples of Weighting Across Frequency Plots (Issue #1)

These are the results for the Cross-CL pipeline. The first plot contains just the average over all multipoles which we could nicely combine in a single plot like so. map_averaged_bandpass_weights_d1s1.pdfhttps://github.com/simonsobs/BBPipe_paper/files/8929021/map_averaged_bandpass_weights_d1s1.pdf The second plot deconstructs the bandpass weights as a function of multipole. bandpass_weights_cross_cl_d1s1.pdfhttps://github.com/simonsobs/BBPipe_paper/files/8929024/bandpass_weights_cross_cl_d1s1.pdf

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1159023620, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACVV4LGKMC6FI5URTCSJMW3VPSO6RANCNFSM5WB655HQ. You are receiving this because you authored the thread.Message ID: @.***>

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1159033003, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ARU5ARHRABKHV53KKOADJN3VPSQLBANCNFSM5WB655HQ. You are receiving this because you commented.Message ID: @.***>

— Reply to this email directly, view it on GitHubhttps://github.com/simonsobs/BBPipe_paper/issues/1#issuecomment-1159042964, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ACVV4LA2LROOKQTOAN2MBDLVPSR5JANCNFSM5WB655HQ. You are receiving this because you authored the thread.Message ID: @.***>

[chervias]

Hi all, I attach the weights per channel for the nilc pipeline. Since nilc works per needlet window, I have to plot the 5 windows separately. This is for r=0 Alens=1 d1s1, goal op noise. The windows from large to small scales. You can see them in the draft of the paper. What do we think about plotting each window separately ? weights_per_channel_nilc.pdf