The first step consists in creating all files
and directories
that ESpRIT will use later on to process data and store calibration information and stellar
spectra. To do this, go to the directory corresponding to your run (by typing
cd /users/mission/ref where ref stands for your mission reference number, e.g.
L98S14) and type:
- setup_mus date
where date is a 7 character keyword coding the date of observing, for example 05feb98 for 1998 February 05. This procedure should create (within directory /users/mission/ref) subdirectories esprit, esprit/raw, esprit/spec (if these first three are not already here), esprit/raw/date and esprit/spec/date). In addition to this, it should also create files geom.in, wcal.in, night_int and night_pol in directory esprit/spec/date (if not already present) by carbon-copying the default version or the most recent ones available in your directory. Check that all these directories and files are now present (by typing successively ls, ls esprit, ls esprit/raw, ls esprit/spec and ls esprit/spec/date).
The second step is to reformat the raw data
that you have recorded (stored in subdirectory BRUT) and copy the new files into
the newly created subdirectory esprit/raw/date. To do this, you must note the
numbers assigned by the CCD system to the files you want to
reformat (which should have names finishing by something like 0168.mt). It is
the number 168 that you are interested in here, and it is usually a good idea to
note these numbers in the log as soon as the corresponding files
are collected. Armed with the numbers of the files you want to work on, execute the
command:
- cfits date num1 num2
where date has the usual meaning while num1 and num2 are the lower and upper limits in file number for the files you wish to reformat. Check that the reformatted files (that should have names such as 05feb0168, in case date is 05feb98 and the file number is 168) appeared in subdirectory esprit/raw/date (by typing ls esprit/raw/date).
The third step is to create an average flat field
frame (with high signal to noise ratio) from the 15 flat field exposures you
obtained at the start of the night. This is done by executing:
- addff date num1 num2 ff
where num1 and num2 are the first and last file numbers of the series of flat field exposures, and ff is the name of the composite file you will create in esprit/spec/date. Check that file ff appeared (by typing ls esprit/spec/date).
The fourth step in preparing for on-line data reduction is to
perform 2D fits to the shape of the various orders of the spectrum,
and also to the tilt of the projected slit throughout the image. To do this, go to
directory esprit/spec/date (i.e. type cd esprit/spec/date) and edit file
geom.in with the text editor of your choice. In this file
(see format description),
the first line contains the name of the average flat field image you have
constructed in the previous step (e.g. ff). The second line is the file name of a
ThAr comparison frame (with its pathname relative to directory
esprit/spec/date, i.e. ../../raw/date), and the fourth line is that of a
bias frame. Change them by the relevant filenames and save
your edited file geom.in. Be sure that the correct date is inserted in
all relevant places in this file! If you do not have a pgplot graphics
window running already, type pgp to get one and display the results of the
following operations. Then, run the command:
- geometry < geom.in
You will get a lot of screen output from this command. One of the first things you should start to spot is a line looking like this:
- Accuracy of 2d offset fit : 0.60 ADU; Readout noise : 4.82 ADU
indicating what the readout noise in ADUs is. It should not be
much more than 5.0 or so. If it reaches values larger than 10.0,
warn the night assistant quickly.
Then, you get a list of
the vertical position (in pixel) of each of about 40 orders of the
spectrum (and a 2D fit to these positions, displayed on the graphics window), as well as a
second listing that looks like this:
- Looking for columns with more than 5.00% of maximum flux...
- Order # 86 : columns -6.0 to 7.5
- Order # 87 : columns -6.0 to 7.5
- Order # 88 : columns -6.0 to 7.5
- Order # 89 : columns -6.0 to 7.5....
- Order # 87 : columns -6.0 to 7.5
In this part of the listing, the negative and positive numbers do not have pretty nearly equal absolute value. It indicates that the estimated centre point of first order (that you provided to the code as the first number in the seventh line of geom.in) is slightly off: the order extends from -6.0 to 7.5 pixels with respect to the user-provided centre point, implying that your estimate is thus slightly shifted to the left of the true centre point of the order. You thus need to increase this parameter slightly (by editing file geom.in) to get things right (e.g. as in the template output). You then get a 2D graphical fit to the slit shape as tracked by ESpRIT. You should soon learn what all outputs look like, and thus be able to recognize the existence of a problem when they look odd.
The final step is to obtain a wavelength calibration polynomial
for each order (through a 2D fit to all orders simultaneously). To do this, run the
command:
- wcal < wcal.in
The code starts with a user-provided linear dispersion relation
for one of the orders to be calibrated (fourth line of input file
wcal.in).
On the very first night of the run, you may need to adapt these parameters slightly. To do
this, display file th.s with disp (see below how to use disp),
display order #93 (by using disp option k with zooming parameters
7000 7999 -0.001 0.1), identify the lines with the help of the ThAr atlas
(available in the console room), find out the position and wavelength of at least two lines (by
choosing disp option g and clicking on both sides of each line of
interest) and work out an approximate linear dispersion relation for this order
(which should be close to that of the template wcal.in
file).
In principle, no further corrections to the input file wcal.in
should be needed throughout the whole run.
As the code runs, it prints out considerable
output
and displays the quality of the fits in a series of graphs. From these graphs and a
comparison with the template output (checking in particular
lines giving the rms accuracy of preliminary calibration polynomial
fits, which should always be lower than 1 pm = 10 mA), you can
easily recognize if everything has gone well to this point.