Hi Chao-Lin,
I think what Jamie's post shows is that the leakage which causes the
(supposed) TB pattern is leakage of some function of the true T pattern. It is
clearly not the T pattern itself or it ought to deproject. However the funny
thing is that the form of the pattern in TB does look similar to a simple
leakage.
Clem
--
**********************************************************************
Clem Pryke - Associate Professor - Physics
University of Minnesota,
Room 313 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-624-7578 Fax: 612-624-4578 email: pryke(a)physics.umn.edu
**********************************************************************
Hi Clem,
Thanks, this seems like a good summary of where we are with B2's
investigation of our apparent B contamination.
I am eager to see the results of the sims of monopole and quadrupole
sidelobes that Chris and Jamie T are performing. Regardless, I think it
is important that now we try an empirical model of T->P leakage. I'll
just repeat what I said before on that:
> To me the most interesting thing in the pagers was the apparent
> correlation in morphology between the detailed patterns in TB and BB
> power. It suggested to me that most of the false B may be in fact
> direct T->B, local in the 2d Fourier plane...anyway, I hope so.
> I suggested a couple weeks ago that someone (Jamie or Chris?) try
> constructing an ad-hoc model of T->B leakage in the 2d Fourier plane,
> e.g. a function in the 2d Fourier plane smooth on ell=30(?) scales
> that predicts the B based on T, and try subtracting this B from the
> data. I think this is worth trying, whatever the outcome of the toy
> sims of degree-scale leakage. Whether or not we'd be satisfied with
> this ad hoc Fourier-plane based deprojection for final analysis, if
> it does remove a large fraction of our false B (more so than it
> removes real B in the r=0.1 sims) then that will be very telling.
Specifically, if this works it would tell us that the T signal in our
own maps predicts the false B, which would rule out a huge class of
other explanations. Furthermore, the (complex) 2d aps T->B template
could be ifft'ed to shed light on the amplitude and angular scale of the
implied residual beam mismatch (yes, dk rotation and E/B-Q/U will
complicate this somewhat). On the other hand, if it is impossible to
construct a smooth, complex function on the 2d Fourier plane that
predicts our excess B based on our T, then we are left pursuing more
complicated explanations for the false B, e.g. sidelobes, buddies,
out-of-band response, poor ground subtraction...things that may produce
some TB but a false B with only partial (and perhaps spatially variant)
correlation with T.
Is anyone currently taking this on? Chris? Jamie T?
(writing on the long CHC-MCM LC130 flight)
John
On 1/16/13 7:36 AM, Clem Pryke wrote:
> With no deprojection deck type jackknifes fail.
>
> Adding A/B offset deprojection jackknifes considered so far basically pass.
>
> However there remains is a high s/n nominal detection of B-modes concentrated
> to low ell.
>
> We know this is not real because:
> - it seems too large versus foreground projections.
> - the nominal signal is less, and has different form in the 2d F-plane in each
> Keck receiver.
>
> Adding additional deprojection modes (relgain, beam width, ellipticity) has
> essentially no effect. (No effect for B2 - however note that for Keck rx2 and
> rx3 relgain is very much required to make the jacks pass.)
>
> We have therefore been thrashing around trying to find a pattern of
> differential beams which evades the deck jackknife:
> - we realized that the two fold rotational symmetry of the focal plane means
> that the deck jackknife might pass for any pattern of difference beams which
> is fixed wrt the tile orientation.
>
> We therefore tried a mixed tile/deck jack which splits the data by tile
> orientation on the sky. Due to the particular deck angles which we have used
> this covers only a small "letterbox" in the middle of the field. This
> jackknife ought to throw a tile fixed effect into stark relief but in fact it
> shows no strong failure.
>
> We are now experimenting with additional postulated patterns of difference
> beams. Fully uniform monopoles and quadrupoles might still evade the deck
> jack. However seems like they ought also to deproject at least a bit...
>
> An additional empirical fact is that there is a pattern of TB correlation in
> the 2d F-plane. It seems like we are dealing with a relgain like effect - but
> one which for some reason does not deproject.
>
> The next step may be to develop an empirical model of T-pol mixing which
> attempts to reproduce the observed TB modes. Subtracting this would force TB
> to zero and would be a major capitulation - however the form of the model may
> lead us to understand what is really going on.
>
--
___________________________________________________________________
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
Hi Walt,
That makes sense - witime is always less that itime - the act of weighting
makes some time periods more important than others reducing the effective
integration time. But for a pixel which uniformly sucks the weather doesn't
matter and the two will be close to equal.
Because we weight separately for pair-sum and pair-diff and the weather
effects the former much more than the latter Pwitime>>Twitime. The T maps are
actually constructed from quite small amounts of time during the best weather
while the pol maps are much more evenly weighted over the run. One sees this
very clearly in the diagnostic plots. This is formally correct though one
might question if we really want it that way.
I'm surprised there is a single stand out aberant pixel on your plot - must be
that all others have been cut one way or another.
Best,
Clem
--
**********************************************************************
Clem Pryke - Associate Professor - Physics
University of Minnesota,
Room 313 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-624-7578 Fax: 612-624-4578 email: pryke(a)physics.umn.edu
**********************************************************************
Hi Chao-Lin and all,
> Is there evidence for excess TB?
I think there is. TB is often plotted on a massively saturated color scale
which can make it hard to see what is going on. See this post:
http://bicep.caltech.edu/~spuder/analysis_logbook/analysis/20121217_leakage/
Are you still unconvinced?
at the end of that post I show that applying relgain deproj *does not* appear to alter the TB result.
[Unfortunately in that post I mis-quoted the color limits for TB and TE and Chris copied these in his most recent posts - sorry for that.]
Clem
--
**********************************************************************
Clem Pryke - Associate Professor - Physics
University of Minnesota,
Room 313 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-624-7578 Fax: 612-624-4578 email: pryke(a)physics.umn.edu
**********************************************************************
> [ Clem, can you lead this discussion from Pole? ]
I will try - basically we will go through whatever new postings are up.
Clem
--
**********************************************************************
Clem Pryke - Associate Professor - Physics
University of Minnesota,
Room 313 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-624-7578 Fax: 612-624-4578 email: pryke(a)physics.umn.edu
**********************************************************************
Hi all,
As agreed with Pole last week, BICEP2 team meeting today (Tuesday) will be back to
its usual time of 9:00 am Pacific (12:00 eastern, 6:00am Pole).
Phone: 1-866-890-3820 (toll: 1-334-323-7229) Passcode:59702175
Walt, could you pleasetake notes <https://docs.google.com/a/cfa.harvard.edu/document/d/1iVrfX19cXQr8fUgRYOUos…>.
John
current agenda outline (add edits/details directly to google doc notes):
1) Pole Operations
a) Report on last week at Pole
b) Status of each "final calibration"
1) thermal testing [ Jon ]: are we done?
2) FTS planning [ Jon et al ]: are we ready to test systematic repeatability?
3) near sidelobes: whopper data quality check, who?
4) Far sidelobe mapping [ Abby, Randol ]
5) transfer function [ Justus ]: any additional data needed?
6) RPS, yukical pol angle analysis
c) Plan for the coming week
2) Analysis To-do Overview
Thread “coordinators” please fill in sections directly in the Google Doc,
to serve as an overview of tasks remaining before we can publish B2 initial
results.
3) Pipeline telecon