Hi Chris,
I am forwarding this to the bicep2 list--I'm not sure why you sent it to
bispud (the joint B2/Keck pipeline list) and to keckarray, but not to
bicep2.
John
-------- Original Message --------
Subject: [Bispud-pipeline-list] proposal for new BICEP2/Keck simset
Date: Mon, 6 May 2013 22:15:11 -0500
From: Chris Sheehy <csheehy(a)uchicago.edu>
To: bispud-pipeline-list <bispud-pipeline-list(a)lists.fas.harvard.edu>du>,
Keckarray <keckarray(a)mailman.stanford.edu>
Hi,
We've come to the realization that we want to run a new, joint BICEP2
3-yr / Keck 2012 simset. This email is an attempt to start a discussion
as to what this simset should include. Please look it over and make
suggestions as needed.
-----------------
BICEP2 3-yr:
TODs = the latest ones currently on disk. Is there a need to re-run the
low level reduction? I seem to remember someone saying last week there
might be.
tags = all 3 years (tags with the incorrect DSL GPS coordinates are now
flagged as bad and will not be included)
realizations = 50 at first, expanding to 100 later. (I bet Clem will
want 100 right away.)
ukpervolt = 3168? Walt?
noise = normal noise model sim
EnoB = Gaussian smoothed 2048 synfast maps, based on WMAP 5-yr param
CAMB spectra (could update to WMAP 9-yr or, God forbid, Planck.) Current
assumed B2 beam is 31.22'. Is this an okay beam size to use, or should
we update it? I don't see any evidence that we need to change our sim
input maps at all.
BnoEnoT = r=0.1 only
additional signal = some atmospheric model?
beamcenters = from beams_cmb_3yr.csv (should we change this name?)
diffpoint = from beams_cmb_3yr.csv
polpar = leave ideal (?)
deproj. template = simulated Planck 143 templates, using Planck beam
window function and random realizations of white noise scaled by Planck
noise maps. WMAP forced us to use nside 512 maps for deprojection.
Planck allows us to use 2048 maps. However, after smoothing to the B2
beam this should not make a big difference, and in fact I've shown that
the additional residuals after deprojection are very small when
downgrading from 2048 to 512. Thus I propose we use a downgraded Planck
deprojection template, and correspondingly use 512 synfast maps with
Planck noise added in sim deprojection. I think things will slow down
considerably if we start using 2048 maps as deprojection templates. We
will accumulate the templates for the standard 6 Gaussian modes: 1
relgain, 2 A/B offsets, 1 beamwidth, 2 ellipticity.
round1/round2 cuts = same as we've been using, though please speak up if
there are any further tweaks that need to happen
channel cuts = same as 1234 "b" and "c" coadds, described here:
http://bicep0.caltech.edu/~spuder/analysis_logbook/analysis/20121112_sim123….
Although these can be changed in daughter coadds, it's not 100% trivial
with many realizations. The only difference between 1234 "b" and "c"
is
that "c" excludes some tags that "b" does not. The channel cuts are
the
same. We could look into these channel cuts more and try to relax them,
but I don't know that it's necessary. These cuts also have a few binary
cuts that I incorporated from Jamie's list of contaminated pairmaps in
2010 and 2011. The 2012 cut will default to the 2011 cut. (I believe the
pairmap contamination cuts get all of "Randol's worst nine" plus some.)
------------
Keck 2012:
TODs = should re-run low level reduction to get the elnod cut params
tags = 2012
realizations = 50 at first, expanding to 100 later
ukpervolt = 3400? Sarah?
noise = sign flip noise (needs more investigation) simultaneously with
BICEP2, so we get B2 sign flip as well. The sign flip noise will be the
only model that captures the correlation between B2 2012 and Keck 2012
noise.
EnoB = same input maps as B2 but smoothed to Keck beam. Have been using
29.5' up to now, but this will probably change. Comments?
BnoEnoT = r=0.1 only
additional signal = same as BICEP2, whatever we decide
beamcenters = what should we use? Presumably CMB derived beam centers.
aux_data/beams/beamcen.csv symlink points to the 4param model, which we
don't want.
diffpoint = presumably same CMB derived CSV file as for beamcenters
polpar = leave ideal (is this our only option?)
deproj template = same as B2, but smoothed to Keck beam
round1/round2 cuts = okay for 2012? Jamie, Sarah? I think you guys
concluded cuts were functioning well for 2012, if not 2013.
channel cuts = need to work this out, but sims can start running before
defining this
--------------------
In both B2 and Keck, the EnoB sims will simulate only T->pol leakage
from our observed differential pointing. We could also simulate BICEP2's
(and probably Keck's) measured per-channel abscals if we wanted.
In both, the BnoE sims will be BnoEnoT and will contain no T->pol or
E->B leakage.
We should decide on any additional signal only sims we want.
We can perhaps run the BnoEnoT and additional signal sims as tag
subsets. I hesitate to run the EnoB sims as tag subsets, but with a
little investigation we might be able to conclude that this is okay. And
if we decide it's okay but it later turns out not to be, nothing is
stopping us from turning the tag subset sims into full up sims.
It should be noted that only sign flip noise sims will model the
correlation between B2 2012 and Keck 2012 noise. We should still run the
3-yr full up noise sims to (1) compare the B2 sign flip noise against
and (2) give us the ability to produce B2 bandpowers with an analysis
that is as minimally different from B1 and QUaD's.
-Chris
--
**********************************************************************
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>
**********************************************************************