Reduction Steps for the Swope
CCD reductions
All of the 1-m SITe CCD data were taken by collaborator
Nidia Morrell
and her colleagues at Las Campanas.
Some basic facts:
- Gain 3.0 e/ADU; read-noise about 7e.
- Scale 0.435"/pixels
- East is on the left, and North is at the bottom. At the "tfinder" level
the rotation is +1.5 degrees.
Some nitty gritty:
Exposures were primarily in "V", but with some U, B, and R, and sometimes
additional filters. Kept only the UBVR.
Dome flats and sky flats agreed to better than 1% in V, so used dome
flats for BVR. For "U" there are only sky flats.
Typical overscan level is 580, with "good" values extending to about 620.
Among our nearly 10,000 or so frames there were a few bad ones, with bias levels
of 0, 200, or 700+. These were eliminated.
There were three different sized images. This is what we finally
adopted:
- BIG: 2080 x 3150. Adopted trimsec=[1:2028,4:3147] biassec=[2052:2078,4:3147]
- MEDIUM: 1233 x 3001. Adopted trimsec=[1:1195,1:2999] biassec=[1205:1232,1:2999]
- SMALL: 1233 x 1201. Adopted trimsec=[1:1200,2:1199] biassec=[1205:1232,2:1199]
Combined zero's night-by-night but not much structure.
Header weirdnesses:
- UTSTART in header is in form of "HH MM SS" rather than "HH:MM:SS"; created all sorts of havoc, which I fixed in a script.
- DATE-OBS in header contains a BLANK rather than a "0" in the case of
single-digit days; i.e., "2005-03- 4". (One wonders why this isn't "2005- 3- 4"
but it isn't.) Fixed that in script as well.
Outline
In the following, I have included some minor helping scripts not because
they will be useful per se, but that they might serve as examples for
others. I had some help from the ledgendary Ed Anderson (NAU) in
these.
- Organized and cleaned up headers:
- Renamed all images n1ccdXXX.fits,n2ccdXXX.fits from various nights, and
combined into one directory. Used rename.cl
- Got rid of all non-zero frame filters which were not "V","B", etc.
Here is a simple script:. Used filtsort.cl
- Fixed up headers: corrected "imagtyp" in 20031129_30 (n7) for some
biases, and so on.
- Made any non-U images containing "sky" or "Sky" in their headers into
imagetyp="object" rather than "flat" by using hselect(img,"$I",(i_title ?= '*sky*'),>"stuff") and so on.
- Sorted into subdirectories "small", "med", and "big" depending upon the
value of NAXIS2.
- Overscan and trim (used parameters above)
- Combined the biases and subtracted:
-
Mario Hamuy have made careful measurements using varying exposure times
of dome flats.
and decided that the SITe3 chip is non-linear.
He derived the following values:
- c1=1.0
- c2=-0.133722484
- c3=0.0498351082
The effect is a few percent over 4 magnitudes. We have carefully experimented
comparing the relative photometry of bright and faint stars on short and
long exposurs. If the correction were correct, we would see good agreement.
In fact, we see good agreement (<1-2%) when we make NO correction; making
the corection results in worse agreement (2-3%). For consistency with
the other R136 reductions, we made the correction on those data, but not
on the others.
- Used dome flats for B, V, and R, and sky flats for
U; we wound that the agreement between sky and dome flats at V was better than
1%. Used cheb order 1 (minmax gave spurious results) on the overscan
- Set bad pixels to 50000 as follows:
- Constructed big and small bad pixel masks by comaring
short sky flats with heavily exposed dome flats, and doing a visual
inspection after running ccdmask with the default parameters.
- Then, in order to set the bad pixels to a Very Large Value, we did the following:
files n16ccd*.fits > fred
list="fred"
while (fscan(list,s1) !=EOF) {
imexpr("b !=0 ? 50000 : a","c"//s1,s1,"med.pl")
}
