A User's Guide to Reducing Echelle Spectra With IRAF
Contents
1. A User s Guide to Reducing Echelle Spectra With TRAF Daryl Willmarth Instrument Support Group National Optical Astronomy Observatories P O Box 26732 Tucson AZ 85726 1 Jeannette Barnes Central Computer Services National Optical Astronomy Observatories P O Box 26732 Tucson AZ 85726 1 May 16 1994 Echelle spectra are generated by the use of a high angle typically 63 deg ruled grating usually cross dispersed by a low dispersion grating grism or prism The result is a high resolution closely spaced array of side by side orders of large spectral coverage The amount and nature of the data can be intimidating to observers to reduce and analyze Fortunately some powerful tools have been written in IRAF to make this task manageable This manual describes the use of these tasks Operated by the Association of Universities for Research in Astronomy Inc under cooperative agreement with the National Science Foundation Contents 1 Introduction 2 Overview of Reduction Process 3 Processing Details 3 1 Initial CCD Processing 3 2 Creating a Normalized Flat 3 3 Flat fielding the Data 3 4 Extraction of the Orders doecslit or apall 3 4 1 Aperture and Background Definition 3 4 2 Aperture Tracing 3 5 Comparison Identification and Wavelength Solutions ecid2. E g ET OF a F a 4 A pi tt ee Ske eee 005 Me LE He 4 eS cee ig z ote 01H gt x 5 10 15 20 Figure 15 The final fit 18 line is especially difficult in the early type stars Velocity differences between the object and standard can change the bandpass enough to cause significant errors in the presence of steep line gradients Unless the observations of the objects and standards are made with the spectrograph slit at the parallactic angle significant throughput differences will exist between the extreme orders In regions of the spectra not containing H lines however flux calibration may be a useful way of flattening the rather peaked echelle blaze function For this purpose alone though one may want to consider using the continuum task as it allows more direct interaction with the spectrum itself Flux calibration in doecslit is accomplished through the tasks standard sensfunc and calibrate An extinction correction is also applied during the calibrate phase based on header parameters Standard allows an interactive definition of bandpass regions starting with the initial values found in the sensitivity calibration parameters or uses those from the calibration file if bandwith and bandsep are set to INDEF Unless the orders span a wavelength interval of about 100A or more it will probably work better to specify the above paramaters as 5 or 10A
3. Function Fitting o I Hit oO HH Residual f u 1 56 ee es Laer Order Figure 13 Initial wavelength fitting for an echelle spectrum with residuals versus order num ber shown This is the same data shown in Figure 12 just plotted differently 16 When sufficient lines have been identified type f to fit the default function to the spec trum Now a plot of deviations from the fit vs pixel number is displayed and interactive fitting may be done Figure 12 shows an example of this initial fit using the default function and orders One can see a couple rejected points probably misidentifications and a notice able curve indicating a higher order is needed It may be a little more useful to display the plot with residual vs order number Figure 13 by typing x and o This will make it easier to find a bad identification than just having all the lines displayed as pixel number Other options are also shown when typing x When satisfied with the initial fit type q to return to the line identification mode One can now mark additional lines in other orders to test the fit or type 1 to locate and mark all the possible lines falling within the acceptance criteria If some likely looking candidates are not marked they may be close blends or the mazfeatures parameter may not be set to a large enough value Type f again to enter the two dimensional fitting routine Figure 14 and type d to delete points
4. Since the smallest bandwidths in the standard calibration files are 16A onedstds spec16cal smaller bandwidth values will allow more points along the order The standard calibration data is interpolated to the bandpasses specified or defined graphically Figure 16 shows an order with 5A interpolated bandpasses If editing of bandpasses is selected you may add a or delete d bandpasses The final bandpasses selected for each aperture remain in effect for subsequent images so there is no need to edit each time After all the bandpasses for each aperture have been defined a curve fitting phase is entered sensfunc where the flux values for each order are fit to a smooth curve to relate system sensitivity to wavelength for each aperture Figure 17 shows the default plot for an aperture Points may be deleted the fitting function type or order changed and curves may be shifted grey shift to name a few of the options available See the help pages for this and the previous task for details Finally the task calibrate applies both an extinction correction and the derived cali bration curves for each aperture to produce the final flux calibrated spectra This phase is non interactive 4 Looking at the Reduced Data The splot task can be used to examine the extracted and calibrated spectra If you are new to this task read the help page since splot offers many different options for plotting and manipulating the data 4 1 Other Tasks Other t
5. apscatter where apertures are defined sized and traced The area between the apertures is then fit with deviant pixel rejection perpendicular to the dispersion and then these fits are smoothed in the orthogonal direction This is the most time consuming method and can be shortened by not smoothing the fits usually without too much loss of accuracy This would be similar to using the task background except there apertures are not defined With background the orders are rejected by setting the parameters low_reject and high_reject to low values like 1 or 2 Apscatter can also be invoked from within doecslit but the noise values will not be remembered during weighted extraction Typical parameters for doecslit are shown in Figure 8 Note that it has a nested pa rameter set that is the parameter sparams is actually another parameter set It can be edited either by typing e on that line or with epar sparams One of the groups of parameters in sparams see Figure 9 concerns the manner in which the comparison spectra are assigned to each object spectrum This is a very flexible part of the reductions and can assign and weight comparison spectra based on the relative time between the object and comparisons the position in the sky or even the date to mention a few of the possibilities The default in doecslit is to run the tasks setjd and setairmass on all spectra to compute and add the header parameters for the Julian Date JD the local Juli
6. sigmas hi n lo n n value of cutoff First time users may wish to review the discussion in UGRSSI Section 3 3 2 When the tracing has been finished another q begins the fitting of the 1 dimensional aperture sum intensity along the dispersion to fit the general spectral shape of the flat field light The fit parameters can be adjusted here just like for the aptrace phase Figure 5 demonstrates how closely the fit can be made to the shape of the spectrum In the far red fringing will be significant and only the envelope of the spectrum can be fit Alternatively a first order legendre or chebyshev polynomial may be used if the shape of the flat is to be preserved After running apflatten you may want to plot the output image to see the bizarre re sult You should see alternating smooth and noisy sections if plotting perpendicular to the dispersion as in Figure 6 and a normal looking plot along an order All values should be near 1 0 The regions between the orders have been set to 1 and the data inside the aperture have been normalized to 1 with the large scale fluctuations removed If a first order function was fit to the light distribution then the flat will have the same shape with an average value of 1 PACKAGE TASK input output referen interac find recente resize edit trace fittrac flatten fitspec line nsum thresho backgro pfit clean skybox saturat readnoi gain 1si
7. spectrum with exactly the same aperture as the object Prospective users should read the IRAF document Guide to the Slit Spectra Reduction Task DOECSLIT by Francisco Valdes Since apall is just a subset of doecslit its use will not be discussed here but all of the details except wavelength calibration extinction correction and flux calibration are applicable A careful examination with implot of a typical object image is useful to determine values for setting task parameters These include the base width of the orders order separation maximum signal level number of orders to be extracted and scattered light and or sky background levels There is generally a few percent compared to order maxima of scattered light present in echelle spectra It is often helpful to average 50 100 rows or columns perpendicular to the dispersion to see a more accurate view of the nature of the scattered light as in Figure 7 Here we find about 60 ADUs of inter order light with some structure compared to order maxima of around 4700 ADUs We can only assume conveniently the scattered light under an order is continuous with the inter order light and a smooth fit between the orders will be representative Comparison exposures generally have negligible scattered light There is more than one way to subtract scattered light but the most rigorous method does it during extraction with doecslit or apall This is recommended as only here is the noise of the background
8. taken into account for cosmic ray cleaning and optimal extraction Since the scattered light is highest near the peak intensity of an order the background noise NOAO IRAF V2 10 3BETA willmart jannu Thu 11 20 52 27 Jan 94 verage of lines 0 to 900 of a0018 alpha cma 300 TT TT T 1500 200 1000 100 lal 500 RAA EWES AL A V MW AW And Vad ly IN N WW JU VV WWW w w W y Column pixels Figure 7 plot perpendicular to the dispersion showing the scattered light is also higher here in other words there is a variable background noise level There are several methods for subtracting the background within doecslit and the doecslit guide can be consulted for details Here also the distinction can be made between scattered light and true background from the sky Sky background will usually not be significant at the f ratio of the coude feed telescope but could be with the 4 m echelle If scattered background light is to be removed in doecslit it is important to do a careful job of normalizing the flat in apflatten Fitting a constant to the flat to preserve its spectral shape won t do as this flat would change the values of the object spectra compared to the inter order regions where they are divided by 1 A method to subtract scattered light before extraction involves using
9. variations in the flat orders than their exact level If your version of IRAF is earlier than V2 10 2 the gain parameter should be set to 1 A bug in earlier versions causes the output spectrum to be multiplied by the gain This will generally not affect the result as cleaning will already have been done during flatcombine If your version of IRAF is V2 10 2 or later this should not be a problem Running apflatten apflatten Flat NFlat will first enter the aperture editor Figure 3 where all the relevant apertures can be marked if not already and the aperture widths checked You may want to set the width of the apertures using the y key during apedit for a more consistent aperture size than given by the auto resize when in the presence of a varying background When satisfied with the aperture selections type q to enter the aptrace phase where all of the selected apertures can be traced interactively or at least their traces examined one after another as described in UGRSSI Section 3 3 2 The icfit routine is used here to map closely the position of the aperture along the dis persion axis Figure 4 If the initial fit does not follow the points to the desired closeness the following parameters can be adjusted 1 The fit order can be increased 0 n where n is the order number 2 Deviant points deleted d with cursor on point 3 The number of fit iterations increased niterate n n number of iterations 4 The high and low rejection
10. E NOT I TETE 1 00E5 80000 60000 40000 20000 UUU 0 200 400 600 800 Figure 3 A flat from the echelle grism setup at the 2 1m coude spectrograph with the orders and apertures marked NOAG TRAF V2 10 3BETA willmart jannu Tue 17 36 05 25 Jan 94 func eee order 4 low _rej 3 high_rej 3 niterate 3 grow 0 total 89 samp le 89 rejected 3 deleted 0 RMS 0 02449 perture 10 of Flat l I T gt 368 4 m ct T at E EN dele RE LORS mt ae PK 368 3 yee HE 368 2 A _ x i g 7S 368 1 y _ HAT e 368 L a 7 367 9 7 4 bt 367 8 E u 0 250 500 750 1000 1250 1500 1750 Line Figure 4 Position of an aperture along the dispersion axis The line represents the fit to the data points The fitting information is at the top of the plot NOAG TRAF V2 10 3BETA willmart jannu Tue 17 tea 04 pa Jan 94 func spline3 order 9 low _rej 3 high_rej 3 niterat oy row 0 total 1775 sanp le 4775 rejected 0 deleted 0 RMS 98 99 imh none T 25000 iN 4 22500 if LT 20000 17500 15000 12500 0000 0 250 500 750 1000 1250 1500 1750 Line Figure 5 Fit along the dispersion of the general spectral shape NOAG IRAF V2 10 3BETA Tne 794 of Nat Tue 17 44 07 25 Jan 94 ui Jos 1 f 1 Ht mia H NIT H i
11. Order of coordinate function across dispersion 3 Rejection iterations 3 Lower rejection sigma 3 Upper rejection sigma yes Refit coordinate function when reidentifying AUTOMATIC ARC ASSIGNMENT PARAMETERS interp Selection method for reference spectra jd Sort key 1jd Group key no Is sort key a time 17 Time wrap point for time sorting DISPERSION CORRECTION PARAMETERS yes Linearize interpolate spectra no Logarithmic wavelength scale yes Conserve flux SENSITIVITY CALIBRATION PARAMETERS 10 Bandpass widths 10 Bandpass separation yes Graphic interaction to examine define bandpasses spline3 Fitting function 1 Order of sensitivity function no Create spectra having units of FNU ql Figure 9 Typical parameters for sparams cont 13 NOAG TRAF V2 10 3BETA willmart jannu Thu 16 04 2 6 27 Jan 94 func legendre order 3 low_rej 3 high rej 0 niterate 5 grow 0 total 821 sample 46 rejected 30 deleted 0 RMS 20 49 Set Background Subtraction for Aperture 4 T T T 5000 4000 4 3000 4 4 2000 1000 amp 2 Ka a Bs ic ewer ner _ 206008 ne CR mm ear 20 10 0 10 20 X Figure 10 An expanded plot of an echelle order examining the background regions apedit stage with skybox or set it to the desired value with skybox 50 for example If background subtraction is turned on check the regions for each ord
12. an Date LJD universal time UTMIDDLE and the air mass at the middle of the exposure This allows the reduction of more than one night of data by using LJD to group the data with the correct comparison spectra and UTMIDDLE to assign the correct spectra within a given night Using other criteria for assigning comparison spectra may involve adding new header keywords and com puting values for them The help pages for refspectra describe these options in detail and give several examples What the last paragraph infers is that doecslit is header driven certain keywords are expected to be in the image headers See the help pages for setairmass and setjd for more information The image header keyword IMAGETYP must also be present in the image headers possible values are object OBJECT comp or COMPARISON Data that are not from NOAO may need to be modified with hedit or asthedit before doecslit will run properly on the data 3 4 1 Aperture and Background Definition Having more or less characterized our images we can now run doecslit doecslit Gimagelist It is good practice to review the apertures initially found for the reference object and subse quent images Occasionally an aperture width will not be wide enough due to asymmetry of the profile or a local peak In these cases the y key or the 1 and u keys can be used to resize the aperture Check for proper numbering of the apertures which should be increasing or decreasi
13. asks are also available for extracting plotting and combining orders Some of these are listed below 19 NOAG TRAF V2 10 3BETA willmart jannu Thu 15 09 24 27 Jan 94 a0035 ec imh HR7426 yet Way 8000 m a Vu 4 Fu iol Fi 1 Hi i 6000 be a 2 7 W 4000H w 4 M Ns 2000 kes 4050 4060 4070 4080 4090 Wavelength Angstroms Figure 16 A plot of an order with 5 interpolated bandpasses marked using the task stan dard 29 28 5 28 27 5 27 26 5 004 002 002 004 006 NOAG TRAF V2 10 3BETA willmart jannu Thu 15 21 17 27 Jan 94 Aper ture 23 munethons Vege adr rder 4 Points 11 RMS 0 0029 ensitivity vs Wavelength Ps a nr ii T T FR l l u 4620 4630 4640 4650 4660 4670 Sensitivity Residuals vs Wavelength T T P se as l 4620 4630 4640 4650 4660 4670 Figure 17 The interactive phase of the task sensfunc The units on the left axis are in magnitudes 20 bplot Batch plots of spectra continuum Fit the continuum in spectra deredden Apply interstellar extinction corrections dopcor Doppler correct spectra sarith Spectrum arithmetic scombine Combine spectra scopy Select and copy apertures in different spectral formats slist List spectrum header parameters specplot Stack and plot multiple spectra specshift Shift spectral dis
14. by interpolation with ccdproc First one must decide which if any pixels are bad or are not linear with flat fielding This may be a bit difficult to determine but generally if a pixel drops to near bias level it will not flat field A more empirical method might be used but depends upon obtaining a couple of groups of flats at different exposure levels After de biasing the flats and averaging each group with flatcombine divide one average flat by the other with imarith Any non linear pixels should be obvious One might also be able to determine the faulty pixels from the data itself after imcopying a test image and flat fielding it Cosmic rays may confuse the determination however After the bad pixels have been identified create a bad pixel file by identifying them as rectangular sections one to a line 149 149 1 800 250 252 450 800 Here we have a short file identifying the region as column 149 and all rows lines of the CCD as bad The second line designates the region between columns 250 and 252 and between lines 450 and 800 as bad When cedproc is run on the data set the parameter fixpix to yes and the parameter fixfile to the name of the file you have created e g badcols 3 2 Creating a Normalized Flat Flat fielding is done mainly to remove individual pixel sensitivity variations chip interference fringing if any and possibly the slowly varying grating blaze function It is desirable that the flat fielding proces
15. entify 3 6 The Rest of the Story 3 7 Flux Calibration 2 2 aaa ee ee 4 Looking at the Reduced Data 4 1 Other Tasks 4 2 Image Headers DS GO OO amp ND WD 1 Introduction Echelle spectra are generated by the use of a high angle typically 63 deg coarsely ruled erating usually cross dispersed by a low dispersion grating grism or prism The result is a high resolution closely spaced array of side by side orders of large spectral coverage Figure 1 shows a surface plot of a small section of a typical echelle spectrum viewed parallel to the orders In addition to the well defined orders numerous cosmic ray spikes are visible which will hopefully be cleaned out of the final extracted spectra by the tasks described here As such echelles produce excellent formats for two dimensional arrays such as CCDs However the amount and nature of the data can be intimidating to observers to reduce and analyze Fortunately some powerful tools have been written in IRAF to make this task manageable This manual describes the use of these tasks and is a supplement to the following manuals e A User s Guide to CCD Reductions with IRAF by Philip Massey hereafter UGCRI e A User s Guide to Reducing Slit Spectra with IRAF by Massey et al hereafter UGRSSI e The main extraction task doecslit has rather lengthy parame
16. er by typing b with the cursor centered on the order An expanded plot will be drawn which you may need to expand vertically and horizontally with w and e as in Figure 10 to see the lower levels of the signal accurately It might be easiest to start at the order with the largest separation usually the red end The sample region need not omit the order itself since the sigma rejection or medianing will eliminate the region within the order The existing sample region can be deleted with z if necessary and new ones defined with the s key If the a ll key is set all other orders will now have the same size region centered on them Typing f will fit the background and show the rejected points 3 4 2 Aperture Tracing Once the apertures and background regions have been satisfactorily defined type q to enter the aperture tracing phase Here one can interactively fit the curve defined by the center of each order at points along the order specified by the parameters nsum and t_step An initial aperture trace is displayed after which the order of the fit may be changed 0 n or points deleted d with cursor at point and the curve refit by typing f The function type itself may also be changed func chebyshev legendre spline1 or spline3 and the curve refit A spline3 function has been found to work well for the aperture curves with a different number of pieces for various orders When a satisfactory fit is obtained as in Figure 4 typing q will di
17. from order to order while a different band will be used only if is typed Since this poking around is just splot it will all work the same after the reductions are finished 3 7 Flux Calibration Flux Calibration of echelle spectra is problematic due to the short pieces of spectra and the large bandpasses of most standard star data An order containing or near a hydrogen 17 NOAG TRAF V2 10 3BETA willmart jannu Mon 16 07 27 31 Jan 94 Function legendre xorder 3 yorder 3 slope 1 offset 144 rms 0 0041 Echelle Dispersion Function Fitting ae TH 5h 4 Se Rs 4 lt O z5 mS FFT SS 3 Z 05 b ae hy EE 15 Order Figure 14 A second pass through the fitting of the dispersion function for echelle compari son data more lines have been added to the solution and the fitting parameters have been adjusted NOAG TRAF V2 10 3BETA willmart jannu Mon 16 10 50 31 Jan 94 Function legendre xorder 5 yorder 6 slope 1 offset 144 rms 0 003 Echelle Dispersion Function Fitting S 01 gt A B a Aes E 4 H l t a 005 in Faak a t t T 4 F TE
18. gma usigma functio order sample naverag niterat low_rej high_re grow mode IRAF Image Reduction and Analysis Facility echelle apflatten Flat NFlat yes yes yes yes yes yes yes yes yes INDEF 20 20 none fitid no 1 INDEF 6 2 6 4 4 spline3 1 1 3 3 3 0 ql List of images to flatten List of output flatten images List of reference images Run task interactively Find apertures Recenter apertures Resize apertures Edit apertures Trace apertures Fit traced points interactively Flatten spectra Fit normalization spectra interactively Dispersion line Number of dispersion lines to sum Threshold for flattening spectra Background to subtract Profile fitting type fitid fit2d Detect and replace bad pixels Box car smoothing length for sky Saturation level Read out noise sigma photons Photon gain photons data number Lower rejection threshold Upper rejection threshold Fitting function for normalization spectra Fitting function order Sample regions Average or median Number of rejection iterations Lower rejection sigma High upper rejection sigma Rejection growing radius 5 Figure 2 Typical parameters for the apflatten task NOAG IRAF V2 10 SBETA pire annu Tue 17 25 17 25 Jan 94 Image F Sum of lines 877 896 cee and Edit Apertures 7234567 8 9 101012131415 16 1708 19 20 aT 22 231 TP
19. ik 500 200 400 600 800 Column pixels Figure 6 A plot of the output image from apflatten perpendicular to the dispersion The regions between the orders have been set to 1 the data inside the apertures have been normalized to 1 with the large scale gradients removed The last step in preparing the flat is to reset the value of CCDMEAN in the header to 1 This parameter is added to the header during flatcombine for subsequent use in ccdproc If the value of CCDMEAN is not changed the object spectra would be multiplied by some large number and then divided by a flat whose value is roughly one resulting in spectra with counts many times higher than in the original data hedit Flat ccdmean 1 0 In IRAF versions V2 10 3 and later apflatten will reset CCDMEAN to 1 automatically 3 3 Flat fielding the Data Now that we have our normalized flat we can proceed to flat fielding the spectra with ccdproc Trimming overscan subtraction and zero subtraction can also be done now if not previously With ccdproc parameters set as before ccdproc objects flatcor flat NFlat or epar ccdproc to set the parameters Process the comps in the same manner 3 4 Extraction of the Orders doecslit or apall Generally one will want wavelength scales on the spectra so doecslit is the task to use Apall can be used otherwise or if one just wants to do it step by step Doecslit has the advantage of extracting each comparison
20. ize clean trace backgro splot redo update quicklo batch listonl sparams mode echelle doecslit IRAF Image Reduction and Analysis Facility a0024 a0018 comps 6 2 6 25000 24 7 yes no no yes yes yes fit yes no no no yes no ql List of object spectra Aperture reference spectrum List of arc spectra Arc assignment table optional List of standard star spectra Read out noise sigma photons Photon gain photons data number Max data value cosmic ray threshold Number of orders Width of profiles pixels Dispersion correct spectra Extinction correct spectra Flux calibrate spectra Resize object apertures Detect and replace bad pixels Trace object spectra Background to subtract Plot the final spectra Redo operations if previously done Update spectra if cal data changes Approximate quicklook reductions Extract objects in batch List steps but don t process Algorithm parameters Figure 8 Typical Parameters for doecslit 11 PACKAGE TASK line nsum extras ylevel t_step t_funct t_order t_niter t_low t_high b_funct b_order b_naver b_niter b_low b_high buffer apscati apscat2 weights pfit 1sigma usigma IRAF Image Reduction and Analysis Facility echelle sparams INDEF Default dispersion line 20 Number of dispersion lines to sum no Extract sky sig
21. ma etc AUTOMATIC APERTURE RESIZING PARAMETERS 0 05 Fraction of peak or intensity for resizing TRACE PARAMETERS 20 Tracing step spline3 Trace fitting function 2 Trace fitting function order 3 Trace rejection iterations 3 Trace lower rejection sigma 3 Trace upper rejection sigma BACKGROUND AND SCATTERED LIGHT PARAMETERS legendre Background function 2 Background 5 Background 5 Background function order average or median rejection iterations 3 Background lower rejection sigma 1 Background upper rejection sigma 2 Buffer distance from apertures Fitting parameters across the dispersion Fitting parameters along the dispersion APERTURE EXTRACTION PARAMETERS variance Extraction weights none variance fitid Profile fitting algorithm fitid fit2d 4 Lower rejection threshold 4 Upper rejection threshold Figure 9 Typical parameters for sparams 12 coordli match fwidth cradius i_funct i_xorde C i_yorde i_niter i_low i_high refit select sort group time timewra lineari log flux bandwid bandsep s_inter s_funct s_order fnu mode ARC DISPERSION FUNCTION PARAMETERS linelists thar dat Line list 0 1 Line list matching limit in Angstroms 3 5 Arc line widths in pixels 5 Centering radius in pixels legendre Echelle coordinate function 3 Order of coordinate function along dispersion 3
22. moarc ec DC FLAG O STD FLAG yes EX FLAG O CA FLAG O BUNIT erg cm2 s A 22
23. ng monotonically The numbering should also be consistent with any skipped orders the desired orders are numbered as if all were there A useful option can be used by toggling the all key if set other commands will perform the same operation on all the apertures For example the y key discussed above would resize all the apertures to the same vertical height instead of just the aperture near the cursor Of the various options available for background subtraction one will probably want to use methods that eliminate cosmic rays and take the additional background noise into account Global scattered light fitting apscatter does the former but not the latter It is also very slow The other method involves the definition of background regions on each side of the order by specifying a buffer distance from the edge of each aperture and a range of pixels The background regions are set automatically by the program but can be changed interactively during the background review as discussed below By using either the fit or median options of background cosmic rays can be eliminated and the background noise values will be taken into account If sky background is not a factor and just scattered light is present boxcar smoothing of the background may be useful Check the present value while in the 10 PACKAGE TASK objects apref arcs arctabl standar readnoi gain datamax norders width dispcor extcor fluxcal res
24. persion coordinate systems 4 2 Image Headers The IRAF image headers are updated throughout the reduction process and by the time the echelle reductions have been completed there are a whole spectrum of new keywords added to the image header These keywords can vary slightly depending on which version of IRAF you are using A partial image header of a reduced echelle spectrum is presented below using IRAF Version 2 10 3 Note that the wavelength information for each order is stored in the WAT keywords For a complete description of the headers for spectral data see phelp oned spec specwes IRAF Version 2 10 3 or later or phelp onedspec package in earlier ver sions BANDID1 BANDID2 BANDID3 WCSDIM CTYPE1 CTYPE2 CTYPE3 CD1_1 CD2_2 CD3_3 LTM1_1 LTM2_2 LTM3_3 WATO_001 WAT1_001 WAT2_001 WAT3_001 WAT2_002 spectrum background none weights variance clean no raw background none weights none clean no sigma background none weights variance clean no 3 MULTISPE MULTISPE LINEAR PPPrPrRP PR gt system multispec wtype multispec label Wavelength units Angstroms wtype multispec speci 1 113 0 4955 0239257812 0 06515394896268 gt wtype linear 756 0 23 22 31 27 spec2 2 112 0 4998 9482421875 0 06753599643 21 WAT2_003 3 256 0 46 09 58 44 spec3 3 111 0 5043 6640625 0 06996016949 WAT2_004 2 256 0 69 28 77 89 DCLOG1 REFSPEC1 demostdde
25. rgon Hollow Cathode set the cursor on a known line type m mark and type in the wavelength Usually entering to the nearest tenth of an Angstrom is sufficient to match a line in the linelist Mark several lines across the order and then move to a new aperture with the j k or o key Knowing the wavelength of one order will enable one to predict wavelengths in other orders with a table of central order wavelengths such as in the KPNO Coude Spectrograph Manual Do not type f fit until the three or more apertures have had lines marked Otherwise the derived fit may not be accurate enough to automatically find additional lines from the linelist Occasionally there will be one or more intense lines in the plot which scale the plot such that few other lines are distinguishable In these cases rescaling the plot with w and e will help to see the lower intensity lines 15 NOAG TRAF V2 10 3BETA willmart jannu Mon 16 01 49 31 Jan 94 Function legendre xorder 2 yorder 2 slope 1 offset 144 rms 0 1347 chelle Dispersion Function Fitting I I T b Residual f u 1 56 es L S 250 500 750 1000 1250 1500 1750 Pixel Figure 12 Initial wavelength fitting for an echelle spectrum with residuals versus pixel number shown The fitting parameters are listed at the top of the plot NOAG TRAF V2 10 3BETA willmart jannu Mon 15 58 58 31 Jan 94 Function legendre xorder 2 yorder 2 slope 1 offset 144 rms 0 1347 Echelle Dispersion
26. s does not significantly change the intensity of the pixels or their relative values perpendicular to the dispersion Otherwise the spatial profile would be altered and optimal extraction and cosmic ray removal would not be as effective We therefore need to normalize the average flat by fitting its intensity along the dispersion by a first or higher order fit while setting all points outside the order aperture to 1 The task apflatten can be used to do this a typical parameter set is shown in Figure 2 The parameters are set after examining the particular flat to be used In this case we are using a flat from the echelle grism setup at the 2 1m coude spectrograph as shown in Figure 3 Apflatten uses parameters from apdefault apfind aprecenter apresize apedit and aptrace so check these also Many of these individual tasks have parameters that are redundant with those in apflatten and only the ones specific to that task need be set The other important parameter set is the echelle package parameters Type epar echelle and check the dispaxis parameter For orders roughly parallel to rows set dispaxis 1 and for orders parallel to columns set dispaxis 2 You may also want the verbose parameter set to yes in order to see on the screen all the information on the extraction that is sent to the logfile Notice that even though there is some scattered light present 3 we have background set to none as we are more interested in the
27. splay the next order for fitting 14 NOAC IRAF V2 10 3BETA willmart jannu Mon 12 38 43 31 Jan 94 Aperture 1 Image line er ecidentify aQ112 ec thar 4300A 2500 Fy T T T 3995 09 2000 3950395 1500 3948 0305 3919 07 1000 M 500 0 EEE ETE kore ie LU Armelle ent F L 500 0 250 500 750 1000 1250 1500 1750 Pixel Figure 11 A plot of the comparison spectrum for the first order 3 5 Comparison Identification and Wavelength Solutions ecidentify After q is typed for the last aperture trace of the reference spectrum the first comparison image will be extracted based on the apertures just defined for the reference image The first plot will be of the first aperture of the comparison spectrum Figure 11 It is assumed that an appropriate file is specified for the parameter coordlist linelist thar dat by default The trick now is to identify a few lines in each of an order near the first middle and last apertures In other words you needn t identify lines in all the apertures or necessarily in the first aperture but at least three apertures should have identifications It can be sometimes difficult to make the initial line identifications and perhaps one of the object spectra with known lines can help to pin down the rough area With the help of an appropriate atlas of the comparison lamp e g A CCD Atlas of Comparison Spectra Thorium A
28. ter sets and the docu ment Guide to the Slit Spectra Reduction Task DOECSLIT by Francisco Valdes should be at hand e For users new to IRAF the document A Beginner s Guide to Using IRAF may be a good starting point e Help pages for tasks in the echelle package will also be useful This manual assumes that the user is familiar with IRAF so in this sense it is not a cookbook It is also assumed that the user has already made at least one pass with ccdproc see UGCRI to debias and trim the data Flat fielding requires special considerations and that is where this manual begins after a summary of the reduction process The preliminary reductions are done within the noao imred ccdred package the actual echelle reductions are done using tasks in the noao imred echelle package An attempt is made to use different font types for IRAF tasks parameters and image names e a bold font is used for an IRAF task name e an italic font is used for a task parameter e a computer font is used for an image name or an interactive cursor keystroke 2 Overview of Reduction Process 1 Follow the first six steps as outlined in UGCRI Section 3 1 to combine zeros and flats and to trim and debias the data a0020 350 470 400 500 HD72688 Figure 1 Surface plot of a small section of a typical echelle spectrum viewed parallel to the orders 2 Create a normalized flat from your average flat with apflatten 3 Run ccdproc again
29. that fall outside of whatever value you feel is acceptable The good lines generally are pretty tightly clumped about small residual values Note the value of the order offset shown at the top of the residuals plot if the value does not reflect the true order of the aperture it can be adjusted with the o key Type q when satisfied with the fit and q again when you are satisfied with the whole comparison identification and fitting process Figure 15 shows the residuals from the final fit with a Legendre function having xorder 5 and yorder 6 Order values higher than these did not improve the fit significantly 3 6 The Rest of the Story Unless flux calibration is to be done the rest of the extraction will be done without interaction Then you are queried whether to plot the final spectra or not splot yes You may wish to look at least at the first image to check the results The line aperture number will be queried for followed by the band number If the parameter extras is yes the output image will be 3 dimensional with band number referring to the following sets of apertures band image type 1 cleaned weighted extracted spectra 2 uncleaned unweighted spectra 3 background spectra 4 estimated sigma spectrum Band 1 is what we are really after but the others may be of interest to insure the cleaning and background subtraction did what we thought it should The keys and can be used to move around
30. to flat field the objects and comps 4 Extract the data with apall no wavelength calibration or doecslit with wavelength calibration In the first case wavelength calibration could be done subsequently with ecidentify ecreidentity refspec setairmass and dispcor The output in either case will be a two dimensional image with as many lines as orders extracted 3 Processing Details We now go through the steps listed previously in detail discussing the individual task pa rameters 3 1 Initial CCD Processing Only a few items will be mentioned here considering the wealth of detail given in UGCRI One should implot a few raw biases and flats to check for any fluctuation in the overscan or the lamp If the overscan is seen to vary by more than an ADU or two from one zero image to another one may want to debias and trim the individual images before running zerocombine or flatcombine If flats are taken in more than one group say 10 at the beginning and 10 at the end of the night there may be a slight image shift on the CCD between the two groups In this case it might be preferable to flatcombine each group separately to remove cosmic rays and then run flatcombine again without rejection or scaling to combine the group averages into one final average flat Otherwise there may be very noisy edges on the average flat where the image shift caused excessive pixel rejection One item not discussed in UGCRI is how to fix bad columns or pixels
Download Pdf Manuals
Related Search