- While a high quality pipeline would provide a non-fail
wavelength solution, I fear that is not yet the case. At present
I am quite confident about data taken with ECHANGL=0 with HIRESb.
Aside from that, I consider the code to be a work in progress.
It is improving rapidly however.
- The source of all the trouble is that the resolution actually
changes with ECHANGL. This is well understood and expected, but a
pain to deal with nevertheless. Basically, it amounts to a
stretching or compression of the spectrum relative to the ECHANGL=0
case.
- With HIRESb, I'd say one has very little reason to use
an ECHANGL value other than 0 with the 3 CCD mosaic. With HIRESr,
however, there is still plenty of reason to do so.
- Anyhow, the obvious source of trouble is that the code
yields an RMS for the fit in a given order that exceeds 0.2 pixels.
I wouldn't worry for the reddest orders where there are very few
lines, but for anywhere else this is a likely problem. I generally
check the QA plots for at least one
arc in each of my setups. I suggest the same for now.
Running hires_tweakarc
- Minor change :: Most of the problems will be quite minor and
hires_tweakarc should be used in this fashion. The code will
launch x_identify and the current line identifications. You will
then re-identify and re-fit.
IDL> hires_tweakarc, 'Arcs/Fits/Arc_C####)fit.idl', orders
--> Here, C is the chip (B,G,R), #### is the frame number, and orders
is an array of the echelle order(s) you wish to tweak.
I now detail the steps to take.
- Delete outliers and re-fit
- Type 'F' -- This will launch a fitting window and the default will
be a 2nd order polynomial. In most cases, there will be a
few lines misidentified at the edges of the CCD.
- Delete outliers with the RMB (right mouse button)
- Type 'R' -- This will show the residuals of the fit
- Type 'u' -- This will increase the fit to 3rd order (go no higher)
- Type 'q' -- This will quit the fit and apply it
- Option 1: The fit was good
- Type 'L' -- This will identify all good lines according to the new fit
- Type 'F' -- This will launch the Fit window again, but to fit the
new set.
- Type 'R' -- Examine the residuals
- Delete outliers (RMB) -- For now, delete the pink triangles too
- Keep the order# at 3 unless you think 4 is much better
(toggle with 'u' and 'd')
- Type 'q' -- Quit and apply the fit
- Type 'q' -- This will quit the re-identify process, save the info
to your file and proceed to the next order to tweak as
desired.
- Option 2: The fit was no good. For now, I suggest you type 'D' to
delete all of the lines and start from scratch. Alternatively,
read how to do the tweak process below.
- Major change :: If it appears that most of the lines identified
are bad, you should reidentify them all. A quick way of doing this
is to use one of the archived solutions as guidance. Here is the
call to hires_tweakarc
IDL> hires_tweakarc, 'Arcs/Fits/Arc_C####)fit.idl', orders, 'templatefile'
--> Here, C is the chip (B,G,R), #### is the frame number, orders
is an array of the echelle order(s) you wish to tweak, and 'templatefile'
is the name of a good archived solution that includes the echelle orders
you wish to tweak. Here is a
list of the filenames and info on the
orders they cover.
I now detail the steps to take.
- Stretch and shift
- The black plot is your arc and the green is the archived arc.
- In the top/center you can type in values to shift and stretch the
archived data.
- When the two overlap closely (within 2-3 pixels everywhere), click
on Apply.
- Identify and Fit
- Type 'L' -- This will identify the arc lines using the archived
wavelength solution
- Type 'F' -- This will launch a Fit window
- Type 'R' -- Examine the residuals
- Type 'u' -- Increase the order number to 3 if necessary
- Delete outliers (RMB) -- For now, delete the pink triangles too
(toggle with 'u' and 'd')
- Type 'q' -- Quit and apply the fit
- Type 'q' -- This will quit the re-identify process, save the info
to your file and proceed to the next order to tweak as
desired.
Continuing and Reprocessing
- It is quite likely that you have only noticed the problem to
your arcs after running through most or all of the steps.
- WARNING: If you try to restart after tweaking (e.g. buy using the
/CLOBBER flag), you will overwrite your tweaked fits.
- To avoid this, run as follows (which will skip the hires_fitarc
file:
- One file
IDL> print,
hires_allarc_sngl('Raw/file.fits', hires, setup, chip, /NOONED, /CLOBBER)
- All files
IDL>
hires_allarc, hires, setup, [chip], /NOONED, /CLOBBER)
Reporting
- Finally, I'd like to know which orders went wrong and with what
setup. I'd also like your fit solution and QA file. Please
send me this info.
- It is quite possible I will add your solution to the archive.
Last modified 2005-10-10