Date: Thu, 14 Feb 2008 15:15:58 -1000
Hi,
Here's a short summary of the "Q97" analysis: the Smith et al. SDSS
standards located in the Landolt fields gathered from early 2003 up
to today. Even though we are working on using tertiary standards from
the Deep fields, for timing reasons and values of comparison for
gradual evolutions in the processing chain, it was decided to stick
to the Smith et al. standards for T0005. This new analysis of Q97
consists of 7000 standards exposures!
The new Elixir flats for the analysis are called "B4". Previous
generations that have been distributed at CADC since 2003 are B1,
B2, and B3 - corresponding to various iterations of the photometric
grids as we were attempting to address the photometric flatness
issues. The B4 version is the dual-color derived grid by Nicolas
Regnault for SNLS. We adopted 4 grids for the lifetime of the camera,
matching the time when an instrument change caused a significant change on
the illumination pattern of the camera (grids are typically captured
after any instrument evolution, and sometimes we see no change, such
as the light baffle which oddly enough did not have any impact).
Grid Range (Runs) Event
---------------------------------------------------------
03B 03Am01 -> 04Bm04 Initial setup
04B 04Bm05 -> 05Am06 L3 lens flip
05B 05Am07 -> 06Am03 Lift of WFC
06B 06Am04 -> 06Bm05 WFC contamination
05B 06Bm06 -> present WFC cleaning
For matching the dates or QSO runs, please consult the following page:
http://www.cfht.hawaii.edu/Instruments/Imaging/MegaPrime/megaprimeschedule.html
The situation with T0004 is that it had a mix of data generated with
flats of various generations (B1/2/3). The idea for T0005 is that all
LS data are processed following the exact same recipe for the photometric
grid for all epochs.
The first plot (ZPoffsets_Data.B3.png) presents the old collection
of zero points from Elixir. I plotted various known events that did
have an impact (optics cleaning, mirror coating) and some less (mirror
washing). There are many more jumps visible, and we can see the u*
calibration is poor (a result mainly of Elixir getting lost on attempting
an astrometry solution on single CCDs with only 2 stars visible on those
defocused 2sec exposures, not so surprising).
The second plot shows the same study, this time using the new collection
of flat fields (ZPoffsets_Data.B4.png). There are still many jumps but
the overall envelope has shrunk in amplitude and g' tends to be more
decoupled that riz. Also the strange increase of zero point detected on
in 03B/04A on the old processing is almost gone (we had no explanation
for that one at the time). The u* band is still rather dreadful, but at
least more balanced (Terapix had reported serious issues indeed with the
B3 version in 06A pointing to an analysis issue by Elixir).
All things considered, the final set of zero points was set from
deriving the median value over the whole time period for each filter
(which match to a great precision the ZPs of 05Bm05 - so I anchored
all data set on that run which had a ton of Q97 observations, stats
are great on that run). I then measured the offset to that median
value for the g, r, and i, applied the g offset to the u band, and
applied the average of the offset of r and i to r, i, and z. The idea
here is to reject some noise (based on the previous plot, you can see
indeed that riz behave pretty much the same, while g is quite off, hence
it was better not to mix everything). I recreated the whole photometric
Elixir database and plotted again the offsets to the 05Bm05 ZPs, and
this is the third plot attached (ZPoffsets_Data.B4.OffsetTo05B_Final.png).
As you can see, u and g are on top of each other (u invisible) and same
for r, i and z. Looking at this plot, I wonder if I was given the correct
dates for the mirror washing - they all seem to correlate by one run offset...
Note that the signature of the calendar of B4 grids provided above is not
detectable on this data set: it is a good sign that photometric continuity
is ensured throughout the life of the camera despite instrument/recipe
changes.
The last plot is simply the absolute value of all 5 ZPs plotted
together, with again the ug and riz having the same behavior from
run to run (ZP_Data.B4.png).
The global CFHTLS reprocessing we are about to start will make use of
these zero points. Terapix will conduct a general photometric analysis
and we'll see how much better we are over T004 for fields that overlap
with Sloan patches.
The future sits within the four Deep fields: the next top priorities for
Elixir (for T0006) is to integrate their hundreds of stars as tertiary
standards (and recalibrate all MegaCam data since 2003). Since last summer,
the Q97 program includes medium exposures (~2mn) on Deep fields. Even when
SNLS is over, we'll keep going back to these fields over the next years:
they will be the reference field for the lifetime of the camera. And now
that flats are (presumed) final, a focus on the defringing recipe is needed
as well.
Time to release the data and confront them to some analysis, please let
me know if you have any comment/ideas on this!
Thank you so much to Nicolas Regnault for his contribution on this effort!
Jean-Charles.
ps: for the record, Elixir photometric equations for megacam are matched
to the SNLS ones:
u* = uSDSS - 0.214 * (u-g)SDSS
g' = gSDSS - 0.156 * (g-r)SDSS
r' = rSDSS - 0.000 * (g-r)SDSS
i' = iSDSS - 0.094 * (r-i)SDSS
z' = zSDSS + 0.050 * (i-z)SDSS
(image/png attachment: ZPoffsets_Data.B3.png)
(image/png attachment: ZPoffsets_Data.B4.png)
(image/png attachment: ZPoffsets_Data.B4.OffsetTo05B_Final.png)
(image/png attachment: ZP_Data.B4.png)