- readaxi=column
- functio=spline3
- order=50
- c1-c3:
- biassec=[2:2046,4098:4158]
- trimsec=[2:2046,2070:4090]
- c4:
- biassec=[2:2037,4098:4158]
- trimsec=[2:2037,2070:4090]
- Divided the average of:
- B-frame, direct image, ambient light, 10x10sec
- B-frame, direct image, QL, 5x5sec
- ccdmask Divc4 Badc4.pl ncmed=11, nlmed=11, ncsig=21, nlsig=21, ngood=25
- Wound up combining all 54 of the flats from all nights using ccdclip. The slit gaps don't exactly cancel but close.
- Tried the usual n=15 spline3 normalization and n=1 cheb (constant). The
n=1 cheb constant was laughably the bet. Here's what F110 looks like using
the two different flats:
- Now, we could divide by a single constant, but why not let the
headers keep track of this for us, and if we need to deal with this
downstream, we'll deal. So, each FlatallcN.fits was run through response
with cheb order 1 and produced nFlatallcN.fits. Then:
- ccdproc iff*c1.fits ccdtype='object' flat='nFlatallc1'
- ccdproc iff*c1.fits ccdtype='comp' flat='nFlatallc1'
- etc...
- For the low dispersion stuff (grating 300) we have the problem that we REALLY would like to use all of the standards, but that the fringing varies with rotator angle and probably something to do with Santa's list.
- For each grating I combined all of the skies (since I had used a single flat series for the run) and then fit it one bin per chip. c4 (bluest) was the fussiest, and I used a 3rd order cubic spline; even so, it was just about 8% from one end to the other. Mostly the corrections were on the order of <1%. I applied the same correction to the comps, for reasons of good hygene.
Extractions and comparison solutions.
Amusing factoids
- The ROTANGLE in the header is wrong by 180 degrees, compare to what we specified in the observing catalog. If we want a PA of 200.00 degrees, we specify 200.00-46.15=153.85 OFF in the catalog. But, the header will report 333.85. Star "A" though is really star A.
- Use the long (15 min) afternoon comps to define the dispersion solution and then do doslit with refit=no with the individual comps (60 sec) to define the zero point shift
- Do the full fit with the short comps at night.
A crucial question is how much the shape of dispersion solution changes as you rotate the instrument. At some level doing the full solution would seem safer.
Here's what I played with: imacsid.dat is the id file with the ArII lines removed and lines cautiously added.
C1:
Residuals from full solution on the long comps is 0.03A, from the short comp it is 0.06A. Had to use 3rd order:
If I do a refit=no after tweaking parameters thresho=10000. cradius=50. then it takes the 3 strongest lines and gives me an rms of 0.048A.
C2:
Now a 2nd order works fine. On the long comps the residuals are 0.02A; on the short comp, 0.06A. If I do a refit=no as per above, then the residuals (from the 2 strongest lines) are 0.02A.
C3:
We get residuals of 0.018A from a 3rd order fit on the long comp. On an individual comp, I'm achieving about 0.05A. The "refit=no" option barely works; with thresh=10000 it correctly identifies two lines and I get 0.07A residuals. With thresh=1000 it misidentifies all lines.
C4:
Look, there are only 3 lines. Nothing I can do can change that. I get residuals of 0.04A from the long comp. On an individual comp, I'm achieving (again) about 0.05A. The "refit=no" option may or may not work; only one line is reidentified, and I don't get the option of going in to check it.
Bottom lines:
N.B.: 0.05A is about 3.5km/sec at mid spectrum.
- c1: Need 3rd order: residuals are 0.06A on an individual short comp, and 0.05A using the long comp to fix the zero-point using only the 3 strongest lines. Use 3rd order on individual comps.
- c2: Need only 2nd order. Residuals are 0.06A on indiidual short; 0.02A for refit using the long comp to fix the zero point using only the 2 strongest lines. Use 2nd order on individual comps.
- c3: Need 3rd order. Residuals are 0.05A on individual shorts; 0.07A for refit using the long comp to fix the zero point using only the strongest 2 lines. Could use 3rd order on individual comps, but preferred spline3 order 1. Residuals were more commponly 0.02A on individual comps, and it makes sense---it isn't a cubic. Spline3 order 1 on individual comps.
- c4: Can only use 2nd order. Residuals are 0.05A on individual shorts, and refit doesn't work. Use 2nd order on dividual comps.
Fixing the titles
What a mess! After we had extracted everything we mostly had *.ms.fites files with aperture 1 being the star furthest to the left which, if there were two stars, is the one that in the title was "+something". If I had had to use A and B, then A was the first ap, and B was the second.
-
I turned everything into oned spectra like this:
- files *c2Ac.ms.fits > allc2
- files *c2Bc.ms.fits > allc2
- files *c2c.ms.fits > allc2
- Hand-edited to put the A and B ones in the right place.
- scopy @allc2 @allc2 format='onedspec'
- files *%.ms.fits.0001.fits%% > aa
- imrename old=@aa//.ms.fits.0001.fits new=@aa//.0001.fits
- files *%.ms.fits.0002.fits%% > bb
- imrename old=@bb//.ms.fits.0002.fits new=@bb//.0002.fits
- files *.0001.fits > ones
- hselect @ones $I,title,APNUM1 yes > onesta
- files *.0002.fits > twos
- hselect @twos $I,title,APNUM1 yes > twosta
- Carefully checked the angles and the files and found that 877-881 were actually the stars MSc047161+SMC047621 and fixed that.
- Edited onesta and twosta to keep the "right" names.
- hedit @ones OLDTITLE '(title)' add+ ver- show+
- list='onesta'
- while(fscan(list,s1,s2) !=EOF){ >>> hedit(s1,'title',s2,ver+,show+) >>> }