Wavelength calibration

The parameter WIDTH2 is the width of the calibration lines. The parameter THRES2 defines the threshold above the local background to detect the comparison lines. A way to estimate this value is to load the WLC on the display. Then choose an order with few bright lines and extract this order using EXTRACT/TRACE. Produce a plot via PLOT/IMAGE TRACE and choose the threshold level such that 7-12 lines (per order) will be above this level. To produce a hardcopy of the WLC give the commands:

LOAD/IMAGE image
COPY/DISPLAY

Now we can execute

CALIBRATE/ECHELLE
or
CALIBRATE/ECHELLE DONE ! If the order definition is already done.

Note For EFOSC reduction, the recommended method is TWO-D. In this method a bivariate polynomial is fitted to the initial identifications, instead of the monovariate polynomial involved in methods PAIR and ANGLE.

This command is the first step in the reduction. It finds the positions of the orders and extracts the comparison lines. The user is prompted for the identification of two calibration lines on the image display, repeated in overlapped regions of adjacent orders. For a detailed description of the wavelength calibration, see Section . This identification is performed with the cursor. In case the orders are not overlapped, use the method ANGLE:

SET/ECHELLE WLCMTD=ANGLE
CALIBRATE/ECHELLE DONE

which performs the wavelength calibration differently, this time prompting for a minimum of four identifications anywhere in the spectrum.

Press in the control box of the display and provide the asked order number and wavelengths.

If the automatic identification procedure does not converge to a satisfactory solution, use the level 1 command:

IDENTIFY/ECHELLE [wlc] [lincat] [dc] [tol] [wlcloop] [wlcmtd] [guess]

to compute new dispersion coefficients. Use commands:

PLOT/RESIDUAL
SEND/PLOT VERSATEC
READ/HISTOGRAM line.tbl :RESIDUAL 0.08 -0.04,0.04

to display the results. You have to specify all parameters, because you are dealing with level 1 commands. The definition of the orders should give a standard deviation of less than 0.5 pixel and the dispersion coefficients should be fitted to better than 0.2 pixel.

The extracted orders in the pixel domain can be resampled into wavelengths with the command REBIN/ECHELLE. A method to verify the wavelength calibration consists of extracting and resampling the wavelength calibration frame and displaying the resampled image with a large scaling factor on the Y-axis. The variation of wavelength coverage from one order to the next should be smooth. Different regions of the resampled image and in particular the overlaps can be verified with the command PLOT/SPECTRUM.

EXTRACT/ECHELLE {WLC} &w
REBIN/ECHELLE &w &r
PLOT/SPECTRUM &r 4000,4050

Solutions are computed by default for each independent order ( SET/ECH WLCOPT=1D). They can also be computed using a global bivariate polynomial ( SET/ECH WLCOPT=2D).

Once a solution has been obtained for a particular set-up, using either of the echelles, you should save the results via

SAVE/ECHELLE name

For all the observations done with the same instrumental set-up, one can now use in CALIBRATE/ECHELLE this dispersion relation as a first guess to the final coefficients by specifying the parameter GUESS:

SET/ECHELLE WLCMTD=GUESS GUESS=name

The main advantage is that when using the GUESS option, no interaction is required to identify the lines in CALIBRATE/ECHELLE. After searching for the lines which are to be used for the determination of the dispersion relation the program applies as a first guess the coefficients used in the session previously saved and then starts the automatic identification.

The command CALIBRATE/ECHELLE has proved to be extremely flexible and reliable. However, the first-time users might run into some problems. The most frequent problems are listed below and a suggestion for their cure is given:

• The wavelength calibration in a certain wavelength interval is wrong: There is probably a wrong identification in the table. Give the command LOAD/IDENTIFICATIONS and proceed as indicated above.
• The wavelength calibration is wrong and apparently no identification error was made: There are too many lines per order. Try a higher threshold in the line detection.
• There are *** NOT ENOUGH LINES *** per order to fit a polynomial of order three to get the dispersion coefficients: The threshold was set too high. Try a lower one and use the GUESS option with the previous solution as the starting point. If, however, this problem is restricted to one or two orders this does not have to be serious because in those orders the dispersion coefficients will be deduced from a global fit to the whole set of orders. It is also possible to determine a bivariate solution ( WLCOPT=2D)