% @(#)identieche.hlq 17.1.1.1 (ESO-IPG) 01/25/02 17:50:36 %++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ %.COPYRIGHT (c) 1990 European Southern Observatory %.IDENT identieche.hlq %.AUTHOR PB, IPG/ESO %.KEYWORDS MIDAS, help files, IDENT/ECHELLE %.PURPOSE On-line help file for the command: IDENT/ECHELLE %.VERSION 1.0 08-FEB-1991 : Creation, PB %---------------------------------------------------------------- \se SECTION./ECHE \es\co IDENT/ECHELLE 10-NOV-1994 PB \oc\su IDENT/ECHEL [wlc] [lincat] [dc] [tol] [wlcloop] [wlcmtd] [guess,[shift]] [ccdbin] perform wavelength calibration of echelle spectra \us\pu Purpose: Implement an instrument independent wavelength calibration for echelle spectra. Line positions are assumed to be in the auxiliary table line.tbl, as created by, e.g., SEARCH/ECHELLE. \\ The command works interactively in the display, assumed to be existent. The command computes dispersion coefficients if the line identification has been performed successfully. \up\sy Syntax: IDENT/ECH [wlc] [lincat] [dc] [tol] [wlcloop] [wlcmtd] [guess,[shift]] [ccdbin] \ys\pa wlc = Name of wavelength calibration frame \ap\pa lincat = Name of line catalog table, which must include at least a column called :WAVE, providing laboratory wavelengths of the calibration lamp used to expose Wlc. The column :ORDER, required by command IDENT/ECH is not used anymore. \ap\pa dc = Degree of polynomial defining dispersion relation. Maximum value is 3 and also the recommended value for CASPEC. \ap\pa tol = Tolerance on rms error of global relation, Default value is 0.2. \\ Tol is interpreted as follows:\\ If Tol > 0, Tol is considered in pixel units.\\ If Tol < 0, abs(Tol) is in wavelength units. \\ Tol is involved at different steps of the loop:\\ - Standard deviation of the very first echelle relation must be smaller than Tol. Otherwise, the user is prompted whether he wants to "Start anyway".\\ - Tol controls the smallest wavelength window when identifying lines using a global dispersion relation.\\ - Before computing single-order dispersion relations, lines with residuals larger than Tol will be rejected. Relaxing Tol value means accepting less accurate global dispersion relation in the automatic line identification and increases the risk of misiden- tifications. Values larger than 1. pixel should not be used. \ap\pa wlcloop = initial tolerance, next neighbour, maxi error. These three parameters control the loop of lines identification at the global step. \\ Initial tolerance (pixels) is used to check the very first dispersion relation obtained from initial identifications. If the accuracy of this starting relation is not better than the specified tolerance, the user is prompted whether it is possible to start anyway. \\ The second element of wlcloop controls the identification of lines by comparison of the residual of a line (computed wavelength minus position of the next catalog line) and the distance of the next neighbours in the line catalog and in the line table. If residual is less than wlcloop(2) times the distance of the next neighbour, the identification is confirmed. \\ Maximal error (pixels) controls the ungraceful exit of the identification loop. If the rms error of the global relation becomes larger than the specified limit (due to initial misidentifications or distortions of the wavelength calibration frame), the loop will be interrupted \ap\pa wlcmtd = Starting method. Can be PAIR, ANGLE, TWO-D, GUESS, RESTART or ORDER. Default is PAIR. See note 1. \ap\pa guess = Name of a previously saved session (cf. SAVE/ECH). The global solution of this session will be used to perform the initial line identification. Used only with method Guess. \\ An additional parameter shift can be added to the name of the guess session, separated by a coma. Its value indicates the shift in pixels between the two sets of calibration lines. If no value is provided, the shift is estimated by cross-correlation. \ap\pa ccdbin = Bin factor of the CCD. This keyword is updated by the command ROTATE/ECHELLE. \ap\sa See also: SEARCH/ECHELLE, PLOT/RESIDUAL, PLOT/IDENT, LOAD/IDENT,\\ CORRELATE/LINE, PLOT/CALIBRATE, LOAD/CALIBRATE \as\no Note: 1. Presentation of the methods. The methods PAIR, ANGLE and TWO-D are interactive and use the command CENTER/MOMENT with option CURSOR to point the lines. It is recommended to beginners to read the related help and use this command before identify/echelle. The three methods can be used if the disperser is a grating. Only the method TWO-D can apply if the disperser is a grism. In case of initial misidentifications, methods PAIR and ANGLE invoke a wavelength calibration diagnosis (see Note 3). In interactive modes, one must first point all the positions to identify, then provide order number and wavelengths. In method PAIR, one must recognize two pairs of lines in overlapped regions of the spectrum and provide the absolute order number of the first pointed line and wavelengths for each of the two pairs. The geometry of the pairs provide the angle of orientation of the spectrum. In method ANGLE, one identify at least four lines anywhere and provide absolute order number of the first pointed line and the wavelengths of all lines. The angle of orientation of the spectrum is computed analytically. At the difference of PAIR and ANGLE which involve the echelle relation (product order by wavelength as a function of the x position), the method TWO-D fits from the beginning a bivariate polynomial to the identifications. Therefore, it requires more initial identifications than the above methods. The minimum number is given by (dc + 1)**2 + 1, where dc is the degree of the polynomial used to fit the dispersion relation. Method RESTART must be used only IMMEDIATELY AFTER the methods PAIR, ANGLE or TWO-D. It avoids the interactive pointing of the lines and enters the program at the step of the identifications. Method GUESS is not interactive and can be used after any calibration by an interactive method. The method GUESS requires the name of a previously saved echelle session, which process tables will be used to start automatically the lines identification. The shift between the two sets of calibration lines is estimated by cross-correlation, unless its value is provided (see parameter guess). Method ORDER allows to restart only the single-order solutions with a different parameter wlctol. 2. Input parameters Echelle keywords are shown in section 6. Wavelength calibration of the SHOW/ECHELLE).If not provided, parameters are read from echelle keywords. One additional parameter is only read from the echelle session:\\ - number of orders (fixed by DEFINE/ECH). 3. Initial identifications, Error diagnosis The command first try to fit an accurate global dispersion relation from the minimum four identifications provided by the user (either two times two lines in overlapped regions of successive orders (PAIR), or four lines anywhere (ANGLE). If it appears that the initial relation is not accurate enough, the user will be prompted for corrective actions, which can consist of:\\ - starting anyway\\ - perform initial identifications again. \on\no Note: In methods PAIR and ANGLE, a wavelength calibration diagnosis is invoked. The diagnosis makes use of the redundancy in the input data to check whether one of the values is possibly wrong. Therefore, no more than one error can be recovered. By solving equations, all sets of (n-1) parameters are used to compute the optimal n-th parameter and the resulting rms obtained by replacement of the actual value by the optimal value is computed. The interpretation of the diagnosis is normally done as follows: A linear fit of the echelle relation with the original values is computed. Read the rms of this fit and take it as reference value. The optimum for each parameter is computed. A diagnosis is possible if one of the values improves dramatically the rms. Then use the mathematical optimum as an indication of the correct value. A very common mistake in the initial identifications is the order number wrong by one. 4. Identification loop The initial solution being accepted, an iterative loop improves the global solution by identifying more lines. The convergence is achieved when no new lines can be identified. Finally, order by order solutions are either fitted, or estimated from the global solution, depending on the number of lines available for each order. Residuals are computed for every order (of course unless there is no identified line in that order). 5. Final solution All fits are performed using polynomials. Calibration results are written in the table line.tbl and will be used in next reduction steps or as a guess for another calibration. The command IDENTIFY/ECHELLE accepts an additional option controlled by the session keyword WLCOPT, which value can be 1D or 2D. This keyword controls the method used to compute the coefficients of the single order coefficients: in 1D mode the coefficients are computed independently for each order, in 2D mode the coefficients are derived from a global 2D solution. The default value of WLCOPT is 1D. \on\exs Examples: \ex IDENT/ECHELLE mywlc mycat 2 0.25 3.0 ANGLE \xe\ex IDENT/ECHELLE P6=GUESS P7=oldses,-1. \xe\ex IDENT/ECH \xe\sxe