Re: [Bicep2-list] [Bispud-pipeline-list] First steps forward for beam map convolution with flat sky
13 Jun
2013
13 Jun
'13
1:23 p.m.
Hi Chris,
On 6/12/13 6:13 PM, Chris Sheehy wrote:
- Your 1) and 2) are the same thing ... as long as you
set mapopt.deproj=true, you pretty much get
dprojected and undeprojected maps for free.
Yep, thanks for this clarification--procedurally these are not really
separate steps, but they are separate tests.
- We talked about this on the telecon, but unless we
have a way to turn
off A/B offsets, what we're going to see is leakage from A/B offsets and
nothing more. After deprojection we will see residual leakage from A/B
offsets and nothing more.
By "residual leakage from A/B offsets" you mean due to template noise,
right? I think we should turn off template noise. See my comment below.
It will still be interesting to add noise to
the beam maps and see what happens, and once you proceed to real beam
maps, it is possible we will see more than just residual A/B offsets,
which would imply an undeprojected residual that is strong enough to
affect us. If we don't see anything, we can be glad that undeprojected
residuals are below what we care about.
This is the main point of the test, yes. The key at this stage is to
establish the limiting level of the beam map noise in assessing residual
T->P.
And it might be interesting to be able to see
beamwidth / ellipticity
deprojection do something when you simulate real beam maps.
I'm in favor doing one more sim, which is with beam maps that have no
A/B centroid mismatch, and running with simopt.diffpoint='ideal'. I
don't see the downside to doing this and I think it gives is more
information about our beams. Without doing it, even if there are weak
undeprojected residuals in our beams that we have a shot at seeing via
these sims, we'd make ourselves blind to them by always including A/B
offsets. Yes, the fact that we'd be blind to them would mean that
they're so weak we don't care about them, but if they're there it'd
be
nice to know.
What do people think? It seemed to me like you guys were opposed.
I'm not adamantly opposed, but the issue I have with this is that it
confuses the intent of this exercise. We are trying to simulate our
actual beam mismatch in a way that is independent of any modeling or
fitting, and then ask what winds up in our CMB maps before and after the
kinds of actual deprojection operations we apply.
"Turning off" differential pointing means assigning a specific value to
the A-B offsets, which we fit from the main beam. But performing the
shift you are talking about, which doesn't happen to our real data,
affects the main beam and sidelobes, symmetric and asymmetric parts.
If we want to ask the question "what is our best estimate of the
undeprojected residual from beam leakage after ideal removal of the
difpoint mode we deproject" then isn't it better to deproject this
specific mode, whether on the beam maps or on the timestreams? We had
planned to do both, but primarily rely on the results of the latter.
And for those I think we should turn off template noise entirely, for
all of these sims. We can re-include (Planck-level) template noise at
the end, but it should just be adding back in a known effect, and one
which is already accurately debiased.
If you want to ask the question "what would the undeprojected residual
look like if our map making code were able to assign A and B beam
pointings separately, based on some fit to individual centers, rather
than to their common centroid" then I think you want the sim you are
talking about. I don't object to it, but unless we are considering that
kind of change to our map making code I don't think it is the primary
question.
I just talked to Chin Lin and she agrees.
John
-Chris
On Wed, Jun 12, 2013 at 2:13 PM, Chin Lin Wong
<clwong(a)physics.harvard.edu <mailto:clwong@physics.harvard.edu>> wrote:
Hi
I was tasked to propose the immediate next steps for using the beam
maps that I've constructed and convolving it with the flat sky to
quantify T->P leakage.
Immediate plan of work:
1. Tag subset, TnoPol, timestream deproj off for
(a) set of noiseless input beam maps with AB offsets in the
beam map
(b) Stefan's multigauss implementation
Make T,Q,U maps and power spectrum, compare T->P leakage. Is (a) the
same as (b) and the same as what Chris has in his posting? This
would check that I've gotten the simulation to run correctly. Start
with 1 realization, and if that goes easily, do 10 realizations.
2. Tag subset, TnoPol, timestream deproj on for:
(a) set of noiseless input beam maps with Ab offsets in the
beam map
(b) Stefan's multigauss implementation
Did I use the deproj correctly? Translate the recovered deproj
coefficients into beam parameters and compare vs input.
3. Tag subset TnoPol, timestream deproj on for a set of input beam
maps w uber-chopper noise levels.
Make T,Q,U maps and power spectrum. It would be informative to see
at what level we can quantify the T->P leakage with uber chopper
noise in the beam maps.
Next, I think we want to use real beam maps. I plan to first make
composite uber-chopper maps, using uber chopper maps at 4 different
dk angles. I will mask out the ground to make these maps, and
hopefully, this will result in nice, clean maps with no ground/MAPO
contamination in them. The uber chopper maps are our least noisy
maps, and it would be interesting to see in test (3) how well we can
quantify the T->P leakage in the presence of noise.
Feedback will be really welcome.
Thanks!
Chin Lin
_______________________________________________
Bispud-pipeline-list mailing list
Bispud-pipeline-list(a)lists.fas.harvard.edu
<mailto:Bispud-pipeline-list@lists.fas.harvard.edu>
https://lists.fas.harvard.edu/mailman/listinfo/bispud-pipeline-list
--
**********************************************************************
Christopher Sheehy - Ph.D. candidate - University of Chicago
Research Specialist, University of Minnesota, Department of Physics
Room 220 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-625-1802 Fax: 612-624-4578 email: csheehy(a)uchicago.edu
<mailto:csheehy@uchicago.edu>
**********************************************************************
_______________________________________________
Bicep2-list mailing list
Bicep2-list(a)lists.fas.harvard.edu
https://lists.fas.harvard.edu/mailman/listinfo/bicep2-list
--
___________________________________________________________________
John Kovac jmkovac(a)cfa.harvard.edu
Assistant Professor, Astronomy and Physics, Harvard University
160 Concord Ave rm 310, Cambridge MA 02138, 617-496-0611
13 Jun
13 Jun
1:34 p.m.
New subject: [Bispud-pipeline-list] First steps forward for beam map convolution with flat sky
Hi John,
I agree with everything you wrote, particularly if we deproject with
noiseless templates, which I'm also on board with. If we don't use a noisy
sim template, than I think my objections about "residual leakage form A/B
offsets" (by which I do mean due to template noise) are moot anyways. And
even if they weren't moot, then I still agree - and agreed before - that
the primary question of "what is our best estimate of the undeprojected
residual from beam leakage after ideal removal of the difpoint mode we
deproject" is best addressed by the steps Chin-Line outlined.
Thanks.
-Chris
On Thu, Jun 13, 2013 at 12:23 PM, John Kovac <jmkovac(a)cfa.harvard.edu>wrote;wrote:
Hi Chris,
On 6/12/13 6:13 PM, Chris Sheehy wrote:
--
**********************************************************************
Christopher Sheehy - Ph.D. candidate - University of Chicago
Research Specialist, University of Minnesota, Department of Physics
Room 220 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-625-1802 Fax: 612-624-4578 email: csheehy(a)uchicago.edu
**********************************************************************
- Your 1) and 2) are the same thing ... as long
as you set
mapopt.deproj=true, you pretty much get
dprojected and undeprojected maps for free.
Yep, thanks for this clarification--procedurally these are not really
separate steps, but they are separate tests.
- We talked about this on the telecon, but unless we have a way to turn
off A/B offsets, what we're going to see is
leakage from A/B offsets and
nothing more. After deprojection we will see residual leakage from A/B
offsets and nothing more.
By "residual leakage from A/B offsets" you mean due to template noise,
right? I think we should turn off template noise. See my comment below.
It will still be interesting to add noise to
the beam maps and see what happens, and once you
proceed to real beam
maps, it is possible we will see more than just residual A/B offsets,
which would imply an undeprojected residual that is strong enough to
affect us. If we don't see anything, we can be glad that undeprojected
residuals are below what we care about.
This is the main point of the test, yes. The key at this stage is to
establish the limiting level of the beam map noise in assessing residual
T->P.
And it might be interesting to be able to see beamwidth / ellipticity
deprojection do something when you simulate real
beam maps.
I'm in favor doing one more sim, which is with beam maps that have no
A/B centroid mismatch, and running with simopt.diffpoint='ideal'. I
don't see the downside to doing this and I think it gives is more
information about our beams. Without doing it, even if there are weak
undeprojected residuals in our beams that we have a shot at seeing via
these sims, we'd make ourselves blind to them by always including A/B
offsets. Yes, the fact that we'd be blind to them would mean that
they're so weak we don't care about them, but if they're there it'd
be
nice to know.
What do people think? It seemed to me like you guys were opposed.
I'm not adamantly opposed, but the issue I have with this is that it
confuses the intent of this exercise. We are trying to simulate our actual
beam mismatch in a way that is independent of any modeling or fitting, and
then ask what winds up in our CMB maps before and after the kinds of actual
deprojection operations we apply.
"Turning off" differential pointing means assigning a specific value to
the A-B offsets, which we fit from the main beam. But performing the shift
you are talking about, which doesn't happen to our real data, affects the
main beam and sidelobes, symmetric and asymmetric parts.
If we want to ask the question "what is our best estimate of the
undeprojected residual from beam leakage after ideal removal of the
difpoint mode we deproject" then isn't it better to deproject this specific
mode, whether on the beam maps or on the timestreams? We had planned to do
both, but primarily rely on the results of the latter. And for those I
think we should turn off template noise entirely, for all of these sims.
We can re-include (Planck-level) template noise at the end, but it should
just be adding back in a known effect, and one which is already accurately
debiased.
If you want to ask the question "what would the undeprojected residual
look like if our map making code were able to assign A and B beam pointings
separately, based on some fit to individual centers, rather than to their
common centroid" then I think you want the sim you are talking about. I
don't object to it, but unless we are considering that kind of change to
our map making code I don't think it is the primary question.
I just talked to Chin Lin and she agrees.
John
-Chris
On Wed, Jun 12, 2013 at 2:13 PM, Chin Lin Wong
<clwong(a)physics.harvard.edu
<mailto:clwong@physics.**harvard.edu<clwong@physics.harvard.edu>>>
wrote:
Hi
I was tasked to propose the immediate next steps for using the beam
maps that I've constructed and convolving it with the flat sky to
quantify T->P leakage.
Immediate plan of work:
1. Tag subset, TnoPol, timestream deproj off for
(a) set of noiseless input beam maps with AB offsets in the
beam map
(b) Stefan's multigauss implementation
Make T,Q,U maps and power spectrum, compare T->P leakage. Is (a) the
same as (b) and the same as what Chris has in his posting? This
would check that I've gotten the simulation to run correctly. Start
with 1 realization, and if that goes easily, do 10 realizations.
2. Tag subset, TnoPol, timestream deproj on for:
(a) set of noiseless input beam maps with Ab offsets in the
beam map
(b) Stefan's multigauss implementation
Did I use the deproj correctly? Translate the recovered deproj
coefficients into beam parameters and compare vs input.
3. Tag subset TnoPol, timestream deproj on for a set of input beam
maps w uber-chopper noise levels.
Make T,Q,U maps and power spectrum. It would be informative to see
at what level we can quantify the T->P leakage with uber chopper
noise in the beam maps.
Next, I think we want to use real beam maps. I plan to first make
composite uber-chopper maps, using uber chopper maps at 4 different
dk angles. I will mask out the ground to make these maps, and
hopefully, this will result in nice, clean maps with no ground/MAPO
contamination in them. The uber chopper maps are our least noisy
maps, and it would be interesting to see in test (3) how well we can
quantify the T->P leakage in the presence of noise.
Feedback will be really welcome.
Thanks!
Chin Lin
______________________________**_________________
Bispud-pipeline-list mailing list
Bispud-pipeline-list@lists.**fas.harvard.edu<Bispud-pipeline-list@lists.fas.harvard.edu>
<mailto:Bispud-pipeline-list@**lists.fas.harvard.edu<Bispud-pipeline-list@lists.fas.harvard.edu>
https://lists.fas.harvard.edu/**mailman/listinfo/bispud-**
pipeline-list<https://lists.fas.harvard.edu/mailman/listinfo/bispud-pipe…
--
****************************************************************
**********
Christopher Sheehy - Ph.D. candidate - University of Chicago
Research Specialist, University of Minnesota, Department of Physics
Room 220 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-625-1802 Fax: 612-624-4578 email: csheehy(a)uchicago.edu
<mailto:csheehy@uchicago.edu>
****************************************************************
**********
______________________________**_________________
Bicep2-list mailing list
Bicep2-list(a)lists.fas.harvard.**edu <Bicep2-list(a)lists.fas.harvard.edu>
https://lists.fas.harvard.edu/**mailman/listinfo/bicep2-list<https://lis…
--
______________________________**______________________________**_______
John Kovac jmkovac(a)cfa.harvard.edu
Assistant Professor, Astronomy and Physics, Harvard University
160 Concord Ave rm 310, Cambridge MA 02138, 617-496-0611
3992
days inactive
3992
days old
bicep2-list@lists.fas.harvard.edu
1 comments
2 participants
participants (2)
-
Chris Sheehy -
John Kovac