Home

User's Manual for DAOPHOT II This manual is intended as a guide

1. 35 table If your values for the readout noise and the photons per ADU are correct then in a plot of CHI against derived magnitude e g Stetson and Harris 1988 A J 96 909 Fig 28 most stars should scatter around unity with little or no trend in CHI with magnitude except at the very bright end where saturation effects may begin to set in The second image peculiarity statistic SHARP is vaguely related to the intrinsic i e outside the atmosphere angular size of the astronomical object effectively SHARP is a zero th order estimate of the square of the quantity actual one sigma half characteristic width of the astronomical object as it would be measured outside the atmosphere in pixels SHARP o observed o point spread function This equation is reasonably valid provided SHARP is not significantly larger than the square of the one sigma Gaussian half width of the core of the PSF see the PSF file in Appendix IV For an isolated star SHARP should have a value close to zero whereas for semi resolved galaxies and unrecognized blended doubles SHARP will be significantly greater than zero and for cosmic rays and some image defects which have survived this far into the analysis SHARP will be significantly less than zero SHARP is most easily interpreted when plotted as a function of apparent magnitude for all objects reduced e g Stetson and Harris 1988 A J 96 909 Fig 27 Upper and lower envelopes bounding t
2. 50 if two stars are separated by more than 0 37x the FWHM and less than 1 0x the FWHM and if the fainter of the two is more uncertain than 1 0 1 5 or 2 0 sigma respectively the fainter one is eliminated Whenever a star is eliminated the iteration counter is backed up by one and the reduction proceeds from that point with the smaller set of stars Backing up the iteration counter gives the second least certain star in the group as much as two full iterations to settle into the new model before it comes up for a tenure decision Since the star rejection formula depends in part upon the user specified values for the readout noise and the number of photons per ADU it is once again important that the values you give for these parameters be reasonable 39 XI SUBSTAR The SUBSTAR command takes the point spread function for a frame and a data file containing a set of x y coordinates and apparent magnitudes for a set of stars shifts and scales the point spread function according to each position and magnitude and then subtracts it from your original frame In the process a new data frame is produced your original picture file is left inviolate In principal this can be done using photometry from PHOTOMETRY as well as from PEAK NSTAR or ALLSTAR but I can t imagine why on earth you d want to In general this star subtraction is done after the photometric reductions have been performed as a check on the quality of the profile
3. polynomial order will produce some complex surface that flows smoothly between the borders and doesn t have discontinuous values or gradients at the edges This is quite a bit slower than a simple surface fit so keep the border narrow only a couple three pixels but I think you ll be pleased with the results Use your fudged image to generate a point spread function free of the cosmic ray spike then delete it before anyone finds out what you have done 51 XX ADDSTAR This routine is used to add synthetic stars either placed at random by the computer or in accordance with positions and magnitudes specified by you to your picture They can then be found by FIND reduced by PHOTOMETRY and the rest and the star finding efficiency and the photometric accuracy can be estimated by comparing the output data for these stars to what was put m Example 1 Random star placement COMPUTER TYPES YOU ENTER Command AD File with the PSF default PSF lt CR gt or filename Seed any integer n Photons per ADU nn n Input data file default RANDOM STARS lt CR gt Magnitude of PSF star is nn nnn Minimum maximum magnitudes desired 15 0 18 0 Number of stars to add to each frame 100 Number of new frames 5 File name stem FAKE This will produce five different new data frames each containing 100 artificial stars with instrumental magnitudes between 15 00 and 18 00 mag added to what was already there The five fram
4. 19 20 21 22 Star ID number X coordinate of stellar centroid same as in COO file Y coordinate of stellar centroid same as in COO file Star s magnitude in aperture 1 measured in magnitudes relative to a zero point of 1 star ADU 25 0 mag Star s magnitude in aperture 2 ditto ditto Star s magnitude in aperture 3 ditto ditto 71 7 Star s magnitude in aperture 4 ditto ditto 8 15 Ditto ditto 16 Estimated modal sky value for the star 17 Standard deviation of the sky values about the mean 18 Skewness of the sky values about the mean 19 Estimated standard error of the star s magnitude in aperture 1 20 Ditto ditto aperture 2 21 Ditto ditto aperture 3 22 Ditto ditto aperture 4 23 30 Ditto ditto Magnitudes for a number of stars are 99 999 9 999 because the aperture extends beyond the boundary of the frame 72 1 2 3 es AA A Sample output from PEAK NSTAR or ALLSTAR a PK NST or ALS file NY LOWBAD HIGHBAD THRESH NL NX 1 284 492 7 200 156 1 2 3 4 168 5 110 6 147 7 64 8 270 14 10 38 11 93 12 139 13 207 co 413 131 454 921 865 759 032 776 925 833 625 818 909 400 0 24000 0 aann o 13 12 13 15 15 15 16 652 807 421 138 732 478 401 748 882 508 354 875 459 18 14 14 16 10 17 15 15 14 16 14 17 14 771
5. 094 629 528 793 330 595 522 982 316 188 009 385 20 0 o0o00O0O00O0O00O0O00o0o0oo0o0o 421 012 018 053 007 128 022 029 015 055 015 081 012 AP1 3 00 464 465 462 463 463 241 009 965 463 462 560 570 463 462 462 460 462 465 PH ADU RNOISE 20 00 618 180 206 292 926 874 127 223 6 50 pa o PwWOROWBWPO WO ADO FRAD 2 0 71 92 09 63 76 86 54 93 58 94 21 63 O0o0o0rOoO0o0O0oO0o0o0o0o0ooroo 399 013 012 056 015 044 021 006 016 075 026 069 009 Star ID number X coordinate of stellar centroid a more accurate value than before Y coordinate of stellar centroid a more accurate value than before Star s magnitude measured in magnitudes relative to the magnitude of the PSF star see discussion of PSF file below Estimated standard error of the star s magnitude Estimated modal sky value for the star from PHOTOMETRY see above Number of iterations required for the non linear least squares to converge CHI a robust estimate of the ratio the observed pixel to pixel scatter from the model image profile DIVIDED BY the expected pixel to pixel scatter from the image profile SHARP another image quality diagnostic see PEAK command above SHARP is most easily interpreted by plotting it as a function of apparent magnitude Objects with SHARP significantly greater than zero are proba
6. d run PEAK GROUP NSTAR or ALLSTAR to get slightly improved positions and magnitudes for the stars and e APPEND this PK NST or ALS file to the one generated in step 6 Using the original picture again run GROUP NSTAR or ALLSTAR on the file created in step 8 e Back in DAOPHOT II ATTACH the original picture and run SUBSTAR specifying the file created in step 6 or in step 8 f as the stars to subtract and the stars in the LST file as the stars to keep You have now created a new picture which has the PSF stars still in it but from which the known neighbors of these PSF stars have been mostly removed If the PSF was made with VARIABLE PSF 1 the neighbors are maybe only 90 or 95 removed but still that s a big gain ATTACH the new star subtracted frame and repeat step 5 to derive a new point spread function This time you should have VARIABLE PSF 0 unless the pure analytic PSF did a great job of erasing the stars Specify the file which you created in step 6 or 8 f as the input file with the stellar photometry since this file has the most recent and best estimates for the positions and magnitudes of the PSF stars and their neighbors If the zits left behind when some of the neighbor stars were subtracted trigger the bad pixel detector answer Y to the question 67 11 Try this one anyway and count on the bad pixel rejection to fudge them away if EXTRA PSF CLEANING
7. 792765E 01 6 512524E 01 6 251689E 01 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 An alphanumeric string uniquely defining the module used to define the analytic first approximation to the PSF Size of the array containing the PSF look up table Values are tabulated at half pixel intervals in a square array which extends somewhat beyond the user specified PSF radius N 2 2 radius 1 1 Here the PSF radius was 12 The number of shape parameters in the analytic function The number of look up tables in the PSF Here a linearly varying PSF was used Fractional pixel expansion NOT IMPLEMENTED The instrumental magnitude corresponding to the point spread function of unit normalization Central height in ADU of the analytic function which is used as the first order approximation to the point spread function in the stellar core X and Y coordinates of the center of the frame used for expanding the variable PSF The shape parameters of the analytic function In this case there are four of them The first two are always the half width at half maximum in x and y 74 14 The look up table of corrections from the analytic first approximation to the true point spread function 79
8. 83218 0 30568 Profile errors 180 0 023 555 0 026 122 0 026 531 0 023 725 0 026 821 0 026 759 0 029 82 0 063 19 0 028 303 0 027 784 0 027 189 0 029 660 0 028 536 0 026 427 0 025 92 0 028 766 0 027 873 0 025 512 0 029 456 0 096 865 0 023 752 0 027 715 0 026 125 0 026 440 0 026 375 0 026 467 0 030 99 0 028 652 0 029 File with PSF stars neighbors NEI If WATCH PROGRESS gt 1 the numbers under the Chi Parameters heading will dance about for a while If you are doing these calculations in a batch job which creates a logfile you should set WATCH PROGRESS 2 This will suppress the dancing which especially under Unix could cause a clog in the log When the dancing stops the computer has fit the analytic function of your choice to the in this case 29 PSF stars The value labeled Chi represents the root mean square residuals of the actual brightness values contained within circles of radius one FITTING RADIUS about the centroids of the PSF stars That is to say the routine has done the best job it could of fitting the analytic function to the pixels within one FITTING RADIUS of the PSF stars one function fitting all the stars The Chi value is the root mean square of the residuals that are left expressed as a fraction of the peak height of the analytic function In this case the analytic first approximation matched the observed stellar profiles to within about 2 8 root mean square on average Par
9. hard copy documentation of the circumstances and nature of the difficulty would be highly desirable This manual has been modified to reflect the new VMS Unix TRAF Midas compatible version of DAOPHOT II The Next Generation If you are still running DAOPHOT Classic you will find many places where this manual is inappropriate 2000 June 5 Peter B Stetson 604 363 0029 stetson dao nrc ca Dominion Astrophysical Observatory Herzberg Institute of Astrophysics 5071 West Saanich Road Victoria British Columbia V8X 4M6 Canada A Introduction B A Typical Run of DAOPHOT II C Descriptions of Main Routines Contents I DAOPHOT II itself 11 IT ATTACH Sii ceed ae eee ee 12 III OPTIONS 0 cortita dai Pee S 14 IV SRY ena shen Sie aeaa a E E E a aes ands eave 19 V FIND Gire baade toate a SRA e 20 VI PHOTOMETRY o aid de ela dt 25 VII PICK sand PSF 3 79 wy ee aa he 29 VILI PEAK unitaria eos 34 IX GROUP A es 36 X NSTAR DA ena e a AS e hs 37 XI SUBSTAR iii e e he Se 29 D Additional Commands cc cece eee eee eee eee XIII MONITOR NOMONITOR 41 XIV SURE sieges aw ds ee tebe ao 42 XV SELECT illo oda aie da dd ld 44 XVI OFFSET aii di a lla 45 XV DEZ APPEND titan Se ced tak 46 AVI DUMP a o dis 47 XIX FUDGE aia te AS duty bee a Meee 49 XX ADDSTAR s sccar ota GW 8S 200 5 58 BAS 51 XXI LIST wat iaa lid 53 AXEL y HELP 4 2290 ited ot ao 54 ARIAS EXI
10. magnitudes of 99 999 and will appear at the end of the list and then it uses the FITTING RADIUS and the PSF RADIUS that you have specified among the optional parameters to eliminate a stars that are too close to the edge of the frame within one FITTING RADIUS and b stars that are too close to brighter stars or possibly saturated stars within one PSF RADIUS plus one FITTING RADIUS Any stars that remain after this culling process will be written out in order of increasing apparent magnitude up to the number that you have specified as a target Stars fainter than the magnitude limit will be included only if otherwise there would be too few PSF stars to map out the degree of spatial variation that you specified with your VARIABLE PSF option If you don t know what magnitude to specify just enter 99 or some such number and learn empirically what looks bright enough to be good With the LST file containing the candidate PSF stars you are ready to run the PSF routine COMPUTER TYPES YOU ENTER Command PSF File with aperture results default AP lt CR gt or filename File with PSF stars default LST lt CR gt or filename File for the PSF default PSF lt CR gt or filename What happens next depends upon the WATCH PROGRESS option If WATCH PROGRESS has been set to 1 or 2 the program will produce little pseudo images of your PSF stars on your terminal one b
11. necessary The PEAK algorithm also makes use of the optional parameter FITTING RADIUS only pixels within a FITTING RADIUS of the centroid of a star will be used in fitting the point spread function Furthermore for reasons related to accelerating the convergence of the iterative non linear least squares algorithm the pixels within this FITTING RADIUS are assigned weights which fall off from a value of unity at the position of the centroid of the star to identically zero at the fitting radius Since the least squares fit determines three unknowns and y position of the star s centroid and the star s brightness it is absolutely essential that the fitting radius not be so small as to include only a few pixels with non zero weight FITTING RADII less than 1 6 pixels that would include as few as four pixels in the fit are explicitly forbidden by the program I suggest that a fitting radius comparable to the FWHM be used as a general rule but suit yourself In addition to an improved estimate of the x y position of the centroid of each star and of its magnitude and the standard error of the magnitude PEAK also produces two image peculiarity indices which can be used to identify a disturbed image The first of these CHI is essentially the ratio of the observed pixel to pixel scatter in the fitting residuals to the expected scatter based on the values of readout noise and the photons per ADU which you specified in your aperture photometry
12. photons ADU and readout noise which were specified as options when FIND was run FRAD is the value of the fitting radius user alterable optimizing parameter which was in effect when the file was created One other thing which the output data files have in common is the format of the first four numbers in the data for each star Always they are Star ID X centroid Y centroid magnitude followed by other stuff with format 1X I5 3F9 3 nF9 In the output from PHOTOMETRY alone this is followed by another data line per star In the pages that follow a sample of each of the output formats is shown An example of the disk file that contains the numerical point spread function is also provided 69 Sample output from FIND a COO file NL NX NY LOWBAD HIGHBAD THRESH 2 284 492 400 0 24000 0 20 0 6 953 200 061 612 0 053 807 3 833 156 254 171 3 306 168 911 318 1 137 1 333 0 546 0 532 2 3 4 5 110 885 9 742 7 080 6 T 8 597 0 000 0 054 653 0 144 0 046 613 0 288 0 182 590 0 040 0 045 489 0 034 0 134 643 0 124 0 120 602 0 074 0 034 627 0 045 0 065 643 0 065 0 045 625 0 203 0 103 599 0 034 0 054 649 0 062 0 162 Nn aNN 147 949 10 208 0 440 64 002 13 161 2 427 270 856 12 738 2 304 9 14 925 13 842 2 976 10 38 813 15 268 1 491 11 93 695 15 164 3 687 12 139 798 15 715 1 156 13 207 929 16 209 3 608 OOTD TROTTO OOO 1 2 3 4 5 6 7 1 Star I
13. program considers the group adequately reduced NSTAR also has a slightly sophisticated star rejection algorithm which is essential to its proper operation A star can be rejected for several reasons 1 If two stars in the same group have their centroids separated by less than a critical distance currently set more or less arbitrarily to 0 37x the FWHM of the stellar core they are presumed to be the same star their photocentric position and combined magnitude is provisionally assigned to the brighter of the two and the fainter is eliminated from the starlist before going into the next iteration 2 Any star which converges to more than 12 5 magnitudes fainter than the point spread function one part in ten to the fifth e g central brightness lt 0 2 ADU pixel if the first PSF star had a 38 central brightness of 20 000 ADU pixel is considered to be non existent and is eliminated from the starlist After iterations 5 6 7 8 and 9 if the faintest star in the group has a brightness less than one sigma above zero it is eliminated after iterations 10 14 if the faintest star in the group has a brightness less than 1 5 sigma above zero it is eliminated after iterations 15 50 or when the solution thinks it has converged whichever comes first if the faintest star in the group has a brightness less than 2 0 sigma above zero it is eliminated Before iterations 5 9 before iterations 10 14 and before iterations 15
14. size 9 Coordinates of central pixel 201 378 l fon o oo nn nnn nn nn ee l l l A Rete Ree a sot pao ge ce OS O ake 383 543 556 600 633 643 630 589 538 514 382 581 644 760 884 930 865 732 623 570 381 651 864 1248 1823 2062 1657 1116 800 626 380 775 1303 2791 5995 7442 4802 2166 1096 732 379 916 1955 5933 16430 22029 11974 4104 1526 846 378 977 2259 6364 16623 23622 13658 4751 1762 933 377 936 1836 3878 7751 10436 7269 3380 1611 949 376 798 1217 1963 3179 3815 3109 2006 1273 861 375 656 847 1138 1563 1778 1621 1300 1003 790 374 602 682 798 962 1090 1073 959 824 698 373 550 587 653 729 801 807 782 705 638 l Minimum median maximum 570 1248 23622 l HOT l To get out l COMPUTER TYPES YOU ENTER l Coordinates of central pixel 0 0 or CTRL Z l I just happened to know that there is a bright star centered near 201 378 The user specifies a box size that will conveniently fill his screen without any wraparound with the column and row ID s 48 across the top and down the left side a box 12 pixels on a side is the largest that can be accomodated on an 80 column terminal screen while a box 21 pixels on a side is the largest that can be fit into a 24 x 132 screen Responding to the Coordinates of central pixel prompt with a position that is outside the picture or with a CTRL Z will return you to DAOPHOT c
15. stars have been preselected to be as isolated as possible and you want the best fits you can get But remember to avoid letting neighbor stars intrude within one fitting radius of the center of any PSF star Run GROUP and NSTAR or ALLSTAR on your NEI file If your PSF stars have many neighbors this may take some minutes of real time Please be patient or submit it as a batch job and perform steps 1 5 on your next frame while you wait After NSTAR is finished run SUBSTAR to subtract the stars in the output file from your original picture This step is unnecessary if you used ALLSTAR And why didn t you EXIT from DAOPHOT and send this new picture to the image display Examine each of the PSF stars and its environs Have all of the PSF stars subtracted out more or less cleanly or should some of them be rejected from further use as PSF stars If so use a text editor to delete these stars from the LST file Have the neighbors mostly disappeared or have they left behind big zits Have you uncovered any faint companions that FIND missed If the latter then a use the cursor on your image display to measure the positions of the new companions b use your system s text editor to create a COO file containing star ID numbers which you invent and the x y coordinates of the new stars FORMAT 1X I5 2F9 c re enter DAOPHOT ATTACH the original image run PHOTOMETRY to get sky values and crude magnitudes for the faint stars
16. this new list to the original star list for the frame so there will be no doubt which star number 1 is referred to in the LST file If you have a list of artificial stars that you ve just added to one frame of a field you may want to offset the positions and instrumental magnitudes so that you can add exactly the same stars into another frame of the same field Just for instance An unformatted READ is used to input these data so the numbers may be separated by spaces or commas and the list can be terminated with a slash if you want all remaining numbers to be zero Use of this routine for transferring the starlist from one frame to another is not recommended in crowded fields particularly when the two frames were taken in different photometric bandpasses You really need to know about all of a star s neighbors in order to reduce it properly and different neighbors may be prominent in frames taken in greatly different colors It would be better simply to run FIND on each frame and to match up the stars you are interested in after the photometry is done 46 XVII APPEND APPEND provides a simple way for the user to concatenate any two of the stellar data files which DAOPHOT has written to the disk A special DAOPHOT command has been written to perform this function because if the user were to leave DAOPHOT and use the operating system s COPY MERGE or APPEND command or equivalent it would then be necessary to use an edit
17. toward the end of your reductions see B above you notice that the subtracted images of your bright stars are surrounded by luminous halos with sharp inner edges then your PSF radius is too small On the other hand the CPU time required for the profile fitting reductions is a strong function of the PSF radius so it is counterproductive to make this parameter too large VARIABLE PSF The degree of complexity with which the point spread function is to be modeled In its infancy DAOPHOT Classic allowed only one form for the model PSF a Gaussian analytic first approximation plus a look up table of empirical corrections from the approximate analytic model to the true PSF This now corresponds to VARIABLE PSF 0 Later on I added the possibility of a point spread function which varies linearly with position in the frame this is VARIABLE PSF 1 DAOPHOT II now allows two more possibilities a point spread function which varies quadratically with position in the frame VARIABLE PSF 2 and a purely analytic model PSF with no empirical lookup table of corrections as in ROMAFOT 15 18 Rear 19 VARIABLE PSF 1 Probably best to leave it at 0 0 Constant until you are sure you know what you re doing FRACTIONAL PIXEL EXPANSION Not implemented Leave it alone ANALYTIC MODEL PSF DAOPHOT Classic always used a Gaussian function as an analytic first approximation to the point spread function DA
18. true intensities The user is also assumed to have a priori knowledge of the following pieces of information 1 the approximate size full width at half maximum of unresolved stellar objects in the frame 2 the number of photons corresponding to one analog to digital conversion unit 3 the readout noise per pixel and 4 the maximum brightness level in analog to digital units at which the detector still operates linearly These conditions being satisfied DAOPHOT II will perform the following primary tasks 1 find star like objects above a certain detection threshold rejecting with a certain degree of reliability bad pixels rows and columns and avoiding multiple hits on individual bright objects although it continues to have some trouble with grossly saturated objects I m still thinking about it 2 derive concentric aperture photometry for these objects estimating a local sky brightness for each star from a surrounding annulus of pixels 3 obtain a point spread function for the frame from one star or from the average of several stars in an iterative procedure intended to fit and subtract faint neighbor stars which contaminate the profile 4 compute precise positions and magnitudes for the program stars by fitting the point spread function to each star either individually or in simultaneous multiple profile fits for up to 60 stars at a time and 5 erase stars from the picture by subtracting appropriately scaled point spread funct
19. 2 characters terminated with a colon at the beginning of a filename as a directory name la VMS Thus if in your cshrc file or some similar location you have a statement like setenv ccd scr nebuchadnezzar mountstromloandsidingspring 1989 ccd data then while running DAOPHOT II in some other directory you can refer to an image or other file obs137 in this directory as ccd obs137 In fact I recommend that you do so because all file names used by DAOPHOT are limited to 30 characters Finally on the VMS side you can create multiple versions of the same filename ad libitum but Unix doesn t allow it Therefore on the Unix side if you try to create a file with the same name as one that already exists DAOPHOT II will type out a warning and give you an opportunity to enter a new filename If you respond to the prompt with a simple carriage return the pre existing file will be deleted before the new one is created If you write command procedures to handle your DAOPHOT II reductions I know I do I recommend that you include in your procedures Unix rm commands to remove any files you know you re going to be creating Otherwise you run the risk of getting unexpected prompts asking whether you want to overwrite pre existing files and your list of commands and filenames can get out of synch with what the program is asking for Chaos could result 13 HI OPTIONS Numerous parameters which optimize the reduction code for the s
20. AA RADIUS OF APERTURE 10 0 00 AB RADIUS OF APERTURE 11 0 00 AC RADIUS OF APERTURE 12 0 00 IS INNER SKY RADIUS 20 00 oS OUTER SKY RADIUS 35 00 When you have the PHO gt prompt you can alter any of the values displayed in the table To do this you first enter the two character identifier of the item that you want to change followed by an sign and the new numerical value for that parameter It works just like the OPTIONS command any characters between the first two and the equals sign will be ignored and anything but a legitimate decimal number after the equals sign will produce an error message For instance to 25 change the radius of the first aperture to 2 5 pixels and change the inner and outer radii of the sky annulus to 10 and 20 pixels respectively COMPUTER TYPES YOU ENTER PHO gt A1 2 5 PHO gt IS 10 PHO gt 0S 20 PHO gt lt CR gt A1 RADIUS OF APERTURE 1 2 50 A2 RADIUS OF APERTURE 2 4 00 A3 RADIUS OF APERTURE 3 5 00 A4 RADIUS OF APERTURE 4 6 00 A5 RADIUS OF APERTURE 5 7 00 A6 RADIUS OF APERTURE 6 8 00 A7 RADIUS OF APERTURE 7 10 00 A8 RADIUS OF APERTURE 8 0 00 A9 RADIUS OF APERTURE 9 0 00 AA RADIUS OF APERTURE 10 0 00 AB RADIUS OF APERTURE 11 0 00 AC RADIUS OF APERTURE 12 0 00 IS INNER SKY RADIUS 10 00 oS OUTER SKY RADIUS 20 00 File with the positions default C00 Note that you are allowed to specify radii for up to twelve concentric aper
21. BS ITIME 150 0 LIST gt lt CR gt or CTRL Z Command If you don t happen to remember the FITS keyword for the particular information you want respond to the LIST gt prompt with just OBS and all the keywords will be typed to your terminal just taken note of the one you want as it flies past I have not yet gotten around to implementing this on the Unix side There the LIST command merely reminds you of the name of the image you are working on and returns you to DAOPHOT Command mode 54 XXII HELP This routine simply produces an alphabetical listing of the currently defined commands on your computer terminal it does not allow you to obtain detailed information on any of the routines It is included primarily as a memory jogger in case you momentarily forget what some routine is called COMPUTER TYPES YOU ENTER Command HE COMPUTER TYPES The commands currently recognized are ADDSTAR APPEND ATTACH DUMP EXIT FIND FUDGE GROUP HELP LIST MONITOR NOMONITOR NSTAR OFFSET OPTION PEAK PHOTOMETRY PICK PSF SELECT SKY SORT SUBSTAR Any command may be abbreviated down to its first two characters 55 XXIII EXIT This command allows you to exit cleanly from DAOPHOT with all files properly closed and everything nice and neat COMPUTER TYPES YOU ENTER Command EX Good bye 56 E ALLSTAR Unlike everything else in this manual ALLSTAR is not a routine within DAOPHOT II whic
22. CPU minutes to a few CPU hours per frame depending upon the number of stars to be reduced the degree of crowding and of course the speed of your machine If computer time is expensive for you you may want to decide just how many stars must be reduced in order to fulfill your basic mission For instance if your goal is to locate the principal sequences of some remote cluster it may not be necessary to reduce every last crowded star on the frame instead only a subset consisting of the less crowded stars might be reduced The derivation of the point spread function can also be performed non interactively with a reasonable degree of success but it may be to your advantage to check the quality of the profile fits visually on an image display before accepting the final product The shape of the point spread function is assumed to be spatially constant or to vary smoothly with position within the frame it is assumed not to depend at all on apparent magnitude If these conditions are not met systematic errors may result Although the star finding algorithm is by itself not sophisticated enough to separate badly blended images two stars whose centers are separated by significantly less than one FWHM by iteratively substracting the known stars and searching for fainter companions it is still possible to identify the separate stars in such a case with a good degree of reliability First one runs the star finding algorithm derives aperture
23. D number 2 X coordinate of stellar centroid 3 Y coordinate of stellar centroid 4 Star s magnitude measured in magnitudes relative to star finding threshold hence never positive since stars fainter than the threshold are rejected obviously 5 Sharpness index see FIND above 6 Roundness index see FIND above 7 Roundness index based on the marginal distributions 70 Sample output from PHOTOMETRY a AP file NL NX NY LOWBAD HIGHBAD THRESH AP1 PH ADU RNOISE 2 284 492 400 0 24000 0 20 0 3 00 20 00 6 50 1 6 953 2 612 99 999 99 999 99 999 99 999 464 618 9 71 0 52 9 999 9 999 9 999 9 999 2 200 061 2 807 99 999 99 999 99 999 99 999 465 180 7 79 0 16 9 999 9 999 9 999 9 999 3 156 254 5 171 14 610 14 537 14 483 14 438 462 206 7 26 0 37 0 013 0 014 0 015 0 016 4 168 911 5 318 16 261 16 056 15 855 15 658 463 292 7 16 0 36 0 055 0 053 0 050 0 048 5 110 885 9 742 10 792 10 728 10 688 10 660 463 926 6 81 0 24 0 001 0 001 0 001 0 001 6 147 949 10 208 17 167 17 084 17 019 16 878 462 241 7 16 0 37 0 124 0 135 0 145 0 144 7 64 002 13 161 15 620 15 569 15 549 15 538 462 009 5 93 0 25 0 025 0 028 0 032 0 035 8 270 856 12 738 15 566 15 527 15 493 15 460 460 965 6 28 0 14 0 026 0 029 0 032 0 035 9 14 925 13 842 14 951 14 863 14 800 14 744 463 874 6 65 0 25 0 016 0 017 0 018 0 019 10 38 813 15 268 16 200 16 113 16 052 16 019 462 127 8 63 0 50 0 062 0 066 0 072 0 079 1 2 3 4 5 6 7 16 17 18
24. OPHOT II allows a number of alternatives which will be discussed below under the PSF command EXTRA PSF CLEANING PASSES DAOPHOT Il is now empowered to recognize and reduce the weight of obviously discrepant pixels while generating the average model point spread function for the frame cosmic rays poorly subtracted neighbors and the like This parameter specifies the number of times you want the program to go through the data and reevaluate the degree to which any given pixel is discordant and the amount by which its weight is to be reduced Set this parameter to 0 if you want every pixel accepted at face value with full weight The amount of time taken by the routine increases with the number of extra passes and in my experience the point spread function has usually converged to a stable final value within five passes so I guess that s a reasonable guess at the largest value you d want to use PERCENT ERROR In computing the standard error expected for the brightness value in any given pixel the program obviously uses the readout noise and the Poisson statistics of the expected number of photons This parameter allows you to specify a particular value for the uncertainty of the fine scale structure of the flat field The readout noise is a constant the Poisson error increases as the square root of the intensity the PERCENT ERROR increases linearly with the intensity of star sky You may think of it as the graininess of
25. PASSES is greater than zero otherwise you re counting on the large number of PSF stars to beat the zits down in the average profile Run GROUP NSTAR or ALLSTAR on the file you created in step 6 or 8 f loop back to step 5 and iterate as many times as necessary to get everything to subtract out as cleanly as possible Remember that each time through the loop you should be obtaining the new point spread function from a frame in which the neighbors but not the PSF stars have been subtracted while NSTAR or ALLSTAR should be run on the original picture with all the stars still in it except when you are trying to get crude data for new stars you have just identified by hand and eye on the image display Increase the VARIABLE PSF option by one per iteration if the neighbors are subtracting out relatively cleanly Once you have produced a frame in which the PSF stars and their neighbors all subtract out cleanly one more time through PSF should produce a point spread function you can be proud of 68 APPENDIX IV DATA FILES DAOPHOT II writes its output data to disk in ordinary ASCII sequential access files which may be TYPEd PRINTed and EDITed with your operating system s utilities and may be shipped freely from one machine to another Files produced by the different routines have some things in common The first of these is a three line file header which in its most elaborate form looks like this NL NX NY LOWBAD HIGHBAD TH
26. PPING RANGE is variable a in the formula given there and the CLIPPING EXPONENT is b The clipping exponent you specify will be rounded to the nearest integer I have given the default values RANGE 2 5 i e a 2 5 o residual gets half weight and EXPONENT 6 0 because in my own experiments they seem to work reasonably well 1 do not have a profound understanding of some fundamental way to obtain best values for these parameters this subject still needs much more experimentation by me and if you want by you Anyway on the basis of whatever religious tenets you hold adopt values for these parameters Experiment with them only if you are prepared to burn up a lot of CPU time If you are thoroughly conservative setting CLIPPING EXPONENT to 0 0 turns the clipping off altogether Your own default parameter values may be set by creating a file named ALLSTAR OPT allstar opt under Unix in your directory It works exactly the same as the DAOPHOT OPT file does in DAOPHOT except of course that it should include only those ten parameters recognizable to ALLSTAR 59 The rest of it s pretty darn trivial COMPUTER TYPES YOU ENTER Input image name filename Object name from file header Picture size nnn nnn File with the PSF default PSF lt CR gt or filename Input file default AP lt CR gt or filename File for results default ALS lt CR gt or filename Name for subtracted image default s DST lt CR gt or fi
27. RACT MAKE A NEW PSF FIT AGAIN loop I recommend that you proceed approximately as follows Invoke DAOPHOT and 1 Run FIND on your frame 2 Run PHOTOMETRY on your frame 3 SORT the output from PHOTOMETRY in order of increasing apparent magnitude decreasing wa nH stellar brightness with the renumbering feature This step is optional but it can be more convenient than not PICK to generate a set of likely PSF stars How many stars you want to use is a function of the degree of variation you expect and the frequency with which stars are contaminated by cosmic rays or neighbor stars I d say you d want a rock bottom minimum of three stars per degree of freedom where the degrees of freedom are 1 constant PSF 3 linearly varying PSF and 6 quadratically varying PSF I m referring here to the number of degrees of freedom you expect you ll need ultimately PSF stars are weighted according to their magnitudes so it doesn t hurt to include many faint but good stars along with a few bright but good ones The more crowded the field the more important it is to have many PSF stars so that the increased noise caused by the neighbor subtraction procedure can be beaten down Furthermore if you intend to use the variable PSF option it is vital that you have PSF stars spread over as much of the frame as possible I don t feel particularly bad if I end up using as many as 25 or 50 PSF stars in such a case but maybe I m to
28. RESH AP1 PH ADU RNOISE FRAD 1 284 492 400 0 24000 0 20 0 3 00 20 00 6 50 2 0 The purpose of this header is to provide a partial record of the analysis which produced the file and to supply some auxiliary information to subsequent routines so you don t have to As the data reduction proceeds through FIND PHOTOMETRY and profile fitting each routine adds more information to the header NL is a historical artifact it started out meaning Number of lines where NL 1 indicated that the file contained one line of data per star output from FIND or PEAK for example and NL 2 flagged output from PHOTOMETRY where two lines of data are generated per star NL has since ceased to have precisely this significance now it serves more as a flag to the program to tell it where in the data record the sky brightness for each star may be found For files produced by FIND PEAK NSTAR and ALLSTAR NL 1 for files produced by PHOTOMETRY NL 2 for files produced by GROUP and SELECT NL 3 SORT OFFSET and APPEND produce output files retaining the form of whatever file was used for input Items NX and NY in the file header are the size of the picture in pixels LOWBAD and HIGHBAD are the good data limits the former calculated in FIND and the latter defined as an optional parameter THRESH is the threshold that was calculated in FIND AP1 is the radius in pixels of the first aperture specified for PHOTOMETRY PH ADU and RNOISE are the numerical values of
29. T ia BG ES Grae 55 Es MALESTAR 0 A A ds a a eae Appendix I Optional Parameters oooooocooooo oo Appendix II The FIND Threshold Appendix III Deriving a PSF in a Crowded Field Appendix IV Data Files A Introduction DAOPHOT II is a computer program for obtaining precise photometric indices and astrometric positions for stellar objects in two dimensional digital images It is intended to run as non interactively as possible and furthermore the possibility that DAOPHOT II would be used at other places than the DAO was kept in mind as it was approaching its present form Therefore DAOPHOT II performs no operations related to the display or manipulation of the digital image on an image display system even though at some stages in the data reduction it is useful to be able to examine the picture visually Picture display operations and some other steps in the reduction procedure such as editing intermediate data files or combining results from different frames to obtain instrumental colors may be done outside of DAOPHOT II using IRAF Midas or whatever software you have handy or feel like writing It is assumed that 1 before running DAOPHOT II the user will have performed all necessary preparation of the images such as flat fielding bias level subtraction and trimming worthless rows and columns from around the perimeter of the picture and 2 the brightness data in the image are linearly related to
30. UM in ADU 32766 50 FWHM OF OBJECT 2 50 THRESHOLD in sigmas 4 00 LS LOW SHARPNESS CUTOFF 0 20 HS HIGH SHARPNESS CUTOFF 1 00 LR LOW ROUNDNESS CUTOFF 1 00 HR HIGH ROUNDNESS CUTOFF 1 00 l WATCH PROGRESS 1 00 FITTING RADIUS 2 00 l PSF RADIUS 11 00 VARIABLE PSF 0 00 FRACTIONAL PIXEL EXPANSION 0 00 ANALYTIC MODEL PSF 1 00 EXTRA PSF CLEANING PASSES 5 00 PERCENT ERROR in 0 75 PROFILE ERROR in 5 00 TR E E RR a SS SS SS SS ee SES eee COMPUTER TYPES YOU ENTER Parameter file default KEYBOARD INPUT lt CR gt OPT gt FW 3 5 l OPT gt HS 0 9 l OPT gt lt CR gt 4 ee eee eS a a SS SS Se EE COMPUTER TYPES l READ NOISE ADU 1 frame 5 00 GAIN e ADU 1 frame 10 00 l LOW GOOD DATUM sigmas 7 00 HIGH GOOD DATUM in ADU 32766 50 FWHM OF OBJECT 3 50 THRESHOLD in sigmas 4 00 LS LOW SHARPNESS CUTOFF 0 20 HS HIGH SHARPNESS CUTOFF 0 90 LR LOW ROUNDNESS CUTOFF 1 00 HR HIGH ROUNDNESS CUTOFF 1 00 l WATCH PROGRESS 1 00 FITTING RADIUS 2 00 PSF RADIUS 11 00 VARIABLE PSF 0 00 FRACTIONAL PIXEL EXPANSION 0 00 ANALYTIC MODEL PSF 1 00 EXTRA PSF CLEANING PASSES 5 00 PERCENT ERROR in 0 75 PROFILE ERROR in 5 00 18 IV SKY The first time you start to work on a new frame with DAOPHOT II you might want to issue the SKY command which will return an es
31. Users Manual for DAOPHOT II This manual is intended as a guide to the use of the digital stellar photometry reduction program DAOPHOT II The Next Generation Brief descriptions of the major routines are provided to aid the user in the use of the program and to serve as an introduction to the algorithms for programmers called upon to make modifications A more complete understanding of the methods used can best be obtained by direct examination of the source code and the comment statements embedded therein DAOPHOT Classic was originally constructed within the framework of a computer program given to Linda Stryker and the Dominion Astrophysical Observatory by Jeremy Mould Caltech The aperture photometry portion of the current program is still somewhat influenced by algorithms and subroutines developed at Kitt Peak National Observatory by way of the POORMAN code developed by Jeremy Mould and Keith Shortridge at Caltech POORMAN was first made to run on the DAO VAX by Linda Stryker and was modified and generalized by Ed Olszewski Over the years I have replaced all of the subroutines with others of my own writing In DAOPHOT II The Next Generation all of the major algorithms and actual FORTRAN coding are my own with the exception of those Figaro IRAF or Midas routines which are used for the image input and output I am solely responsible for bugs and algorithm related problems If any such problems are found please contact me for major problems
32. ame extension type in the filename you want without any extension or period and the default filename s extension will be tacked onto the filename you enter Similarly if you like the filename part but want to change the extension simply type a period and the new extension the filename which was offered to you will be retained but with your extension replacing the one offered I strongly suggest that you use the default filenames unless you have some very good reason for not doing so it reduces the amount of typing that you have to do and thereby reduces the probability of an unnoticed typographical error and it helps you to keep your bookkeeping straight If you have elected to watch the output of the program on your terminal either by using the option WATCH PROGRESS 1 0 or by using the MONITOR command see below you will now see the computer count through the rows of your picture As it is doing this it is convolving your picture with a lowered truncated Gaussian function whose FWHM is equal to the value set by the FWHM option see the OPTIONS command above The Gaussian convolution includes compensation for the local background level and since the Gaussian is symmetric smooth gradients in the sky brightness also cancel out These properties enable the sky finding algorithm to ignore smooth large scale variations in the background level of your frame such as those caused by a sea of unresolved fainter stars the threshold that you
33. ameters OPTIONS MONITOR NOMONITOR may be issued to DAOPHOT II without a prior ATTACH command having been given The program will refuse to let you tell it to perform tasks requiring a picture file e g FIND PHOTOMETRY PSF GROUP NSTAR unless a picture has been ATTACHed In all implementations of DAOPHOT II of which I am aware if the extension part of your picture s filename is the standard one for that image format DST for Caltech data structures imh for IRAF bdf for Midas it may be omitted If it is not the standard extension or on VMS machines if you wish to specify other than the most recent version of the image file the filename extension must be included in the ATTACH command The ATTACH command is the only one in DAOPHOT II which allows the user to include additional information viz a filename on the command line All other commands are issued simply and without modification the routines will then prompt the user for any necessary input This is also a good time to point out that DAOPHOT II commands are case insensitive commands and parameter options see below may be entered using either upper or lower case letters even on Unix machines On VMS machines filenames are also case insensitive FILE EXT and file ext refer to the same file On Unix machines FILE EXT file ext FILE ext FiLe ExT etc are all different Finally for Unix afficionados I have taught DAOPHOT II to recognize a string of 1
34. asonably correct 36 IX GROUP Before performing multiple simultaneous profile fits using the routine NSTAR it is necessary to divide the stars in your frame up into natural groups each group to be reduced as a unit The principle is this if two stars are close enough together that the light of one will influence the profile fit of another they belong in the same group That way as the solution iterates to convergence the influence on each star of all of its relevant neighbors can be explicitly accounted for COMPUTER TYPES YOU ENTER Command GR File with the photometry default AP lt CR gt or filename File with the PSF default PSF lt CR gt or filename Critical overlap n n File for stellar groups default GRP lt CR gt or filename The critical overlap has the following significance When GROUP is examining two stars to see whether they could influence each others fits it first identifies the fainter of the two stars Then it calculates the brightness of the brighter star using the scaled PSF at a distance of one fitting radius plus one pixel from the centroid of the fainter If this brightness is greater than critical overlap times the random error per pixel calculated from the known readout noise sky brightness in ADU and number of photons per ADU then the brighter star is known to be capable of affecting the photometry of the fainter and the two are grouped togethe
35. bly galaxies or unrecognized doubles objects with SHARP significantly less than zero are probably bad pixels or cosmic rays that made it past FIND 73 Sample of a point spread function file a PSF file 1 2 B3 5 6 7 8 9 10 11 12 13 14 PENNY1 51 4 3 O 12 033 33419 105 158 5 255 0 1 02691E 00 1 06132E 00 7 23460E 01 1 44567E 01 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 6 484396E 01 6 400895E 01 6 459602E 01 6 713283E 01 6 921574E 01 6 974660E 01 6 971574E 01 6 964603E 01 7 014906E 01 7 076376E 01 7 028298E 01 6 946377E 01 6 923672E 01 6 868890E 01 6 699422E 01 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 1 348391E 02 6 271802E 01 6 469427E 01 6 637348E 01 6 902106E 01 6 919862E 01 6 852924E 01 6 951303E 01 7 280193E 01 7 551125E 01 7 588057E 01 7 524973E 01 7 459446E 01 7 402506E 01 7 357459E 01 7 343834E 01 7 333043E 01 7 215495E 01 7 010623E 01 6
36. btract the stars just reduced from the data frame Or you could get around the whole thing just by running the APPENDed group file through ALLSTAR EXIT from DAOPHOT II Display the star subtracted picture created in step XIV on your image display system Look for stars that have been missed and for galaxies and bad pixels that have been found and reduced as if they were stars If desired create a file containing the coordinates of stars you wish to add to the solution and run PHOTOMETRY on these coordinates To make one more pass through the data you should run this aperture photometry file through GROUP run the previous NSTAR or ALLSTAR results through GROUP again necessary in order to get both the old and new stars into files with the same format APPEND these files together and return to step XIV Repeat as many times as you like and have the time for EXIT from DAOPHOT and examine your picture on the image display system Choose several minimally crowded bright unsaturated stars Make a copy of the output file from your very last run of NSTAR or ALLSTAR and with a text editor delete from this file the data lines for those bright stars which you have just chosen Run DAOPHOT ATTACH your original picture and invoke SUBSTAR to subtract from your picture all the stars remaining in the edited data file With equal ease you can also create a file containing only the stars you want to retain you could even use the file containing
37. charge overflow column for instance then the brightness distribution as a function of x would be sharply peaked while the distribution as a function of y would be nearly flat the height of the x Gaussian function would have some significant value while that of the y Gaussian would be near zero In this case ROUND would have a value near 2 0 The default acceptance limits for ROUND are 1 0 lt ROUND lt 1 0 i e if the heights of the x and y Gaussian distributions for a brightness enhancement were 60 and 180 ADU difference average 120 120 and the object would be right at the limit of acceptance Note 22 that ROUND discriminates only against objects which are elongated along either rows or columns objects which are elongated at an oblique angle will not be preferentially rejected The numerical values for the limits of the acceptance interval for SHARP and ROUND may be changed by the user see the OPTIONS command above if normal stars in your frame should happen to have unusual shapes due to optical aberrations or guiding problems However I recommend that you take great care in deciding on new values for these cutoffs It might be useful to perform a preliminary run of FIND with very generous limits on the acceptance regions and then to plot up both SHARP and ROUND as functions of magnitude for the objects detected see Stetson 1987 PASP 99 191 Fig 2 Observe the mean values of SHARP and ROUND for well exposed star
38. ctive If the background is flat across the frame then you can set a tight limit maybe 50 or so If there is a strong background gradient you will need to set a more generous limit maybe 100 or more to keep legitimate sky pixels from being rejected in those parts of the frame where the background is faint Intelligent use of the DUMP command and or an image display will help you decide what to do HIGH GOOD DATUM The level in data numbers above which a pixel value is to be considered defective Note that this differs from the LOW GOOD DATUM just discussed The LOW GOOD DATUM is defined as a certain number of standard deviations it below a frame s mean sky value Thus assuming that all your frames have comparably flat backgrounds it needs to be specified only once the actual numerical value used will ride up and down as frames with different mean background levels are considered The HIGH GOOD DATUM is specified as a single fixed number which represents the absolute level in data numbers at which the detector becomes non linear or saturates Note that since your data have been bias level subtracted and flat fielded this number will not be 32767 but will be somewhat lower FWHM The approximate FWHM in pixels of the objects for which the FIND algorithm is to be optimized This parameter determines the width of the Gaussian function and the size of the array with which your picture is numerically convolv
39. e the values of any of the optional reduction parameters This step is itself optional III Use ATTACH to tell the program which picture you want to reduce In the VMS version you do not need to include the filename extension DST when you specify the filename but you may if you like In the Unix IRAF version your life will be simpler if you do not include the imh extension IV You might want to use SKY to obtain an estimate of the average sky brightness in the picture Write this number down in your notes This step is not really necessary because FIND below will do it anyway V Use FIND to identify and compute approximate centroids for small luminous objects in the picture One of the user definable optional parameters which you are permitted to define is the significance level in standard deviations of a luminosity enhancement in your image which is to be regarded as real Two other parameters which you must define are the readout noise and gain in photons or electrons per data number which are appropriate to a single exposure with your detector When you run FIND it will ask you whether this particular image is the average or the sum of several individual exposures From the available information FIND will then compute the actual brightness enhancement in data numbers above the local sky brightness which corresponds to the significance level you have specified See the section on FIND and the Appendix on The FIND T
40. ed by FIND see detailed discussion under FIND command below If conditions during your run were reasonably constant a single value should be adequate for all your frames THRESHOLD The significance level in standard deviations that you want the program to use in deciding whether a given positive brightness enhancement is real Normally somewhere 14 9 10 11 12 13 14 Na wa around 4o is good but you may want to set it a little higher for the first pass Then again maybe not LOW and HIGH SHARPNESS CUTOFF Minimum and maximum values for the sharpness of a brightness enhancement which FIND is to regard as a real star intended to eliminate bad pixels may also help to eliminate low surface brightness galaxies In most cases the default values given in the program are adequate but if you want to fine tune them here they are LOW and HIGH ROUNDNESS CUTOFF Minimum and maximum values for the roundness of a brightness enhancement which FIND is to regard as a real star intended to eliminate bad rows and columns may also reject some edge on galaxies Again I think you ll find that the program assigned default values are adequate for all but special cases WATCH PROGRESS Whether to display results on the computer terminal in real time as they are computed Displaying the results may keep you entertained as the reductions proceed but it may slow down the executio
41. ers along to other routines which will then act accordingly 61 APPENDIX II The FIND Threshold Theory and Practice Assume that a given frame has an average sky brightness of s in units of ADU a gain factor of p photons per ADU and a readout noise per pixel of r in electrons or equivalently photons Note that since electrons and photons are simply number counts they really have no physical dimensions this is inherent in the use of Poisson statistics where o N VN would be meaningless if any physical dimensions were involved The expected random noise per pixel may now be computed as follows IN UNITS OF PHOTONS If sky brightness in photons p x s ADU then for the Poisson statistics of the sky Variance sky 0 sky number of photons Poisson statistics pxs variance is in dimensionless units Readout noise Variance readout 0 readout e dimensionless units by definition Total noise Variance total variance sky variance readout propagation of error pxs r dimensionless units standard deviation y p x s r dimensionless units Note that in this equation p has units of photons ADU s has units of ADU and r has units of electrons or photons This is very clumsy but that s the way we usually think of these numbers IN UNITS OF ADU o photons photons per ADU a ADU propagation of error Let R readout noise in ADU r p Then a tota
42. es all other routines read these numbers from the input files not from the optional parameter table The same goes for the HIGH GOOD DATUM value Therefore it is entirely in your own best interests to provide the program with the most reasonable possible estimates of the readout noise and gain and of the minimum and maximum valid data values doing it correctly at the beginning will save you hassles later For instance if the sky background in your frame is dead flat a random Galactic star field not in any cluster a strict value for the lowest good data value might be say five sigma below the average sky A five sigma or greater deviation has a normal probability of about 3 x 107 so in a 300 x 500 image there would be only about one chance in twenty that even one legitimate datum would be unfairly rejected Of course if the sky background does vary significantly across the frame in a globular cluster H IT region or external galaxy you would want to set the minimum good data value maybe ten or more sigma below the average sky Finally FIND asks you to provide a name for the disk file where it is to store the coordinates of the stars it finds Here as you will find nearly everywhere in DAOPHOT II when asking you for a filename the program will offer you a default If you are satisfied with the default filename you need only type a carriage return and this name will be used if you want to change the filename but keep the default filen
43. es will be named FAKE01 DST FAKE05 DST There will also be created five files named FAKE01 ADD FAKE05 ADD containing the x y positions and the magnitudes of the new stars I have you specify a seed for the random number generator so that a if you lose the images by specifying the same seed you can make them again and b exactly the same artificial image will be created on any computer provided the same seed is used 52 Example 2 Deliberate star placement COMPUTER TYPES YOU ENTER Command AD File with the PSF default PSF lt CR gt or filename Seed any integer n Photons per ADU 14 1 Input data file default RANDOM STARS ARTSTAR MAG Output picture name default ARTSTARa lt CR gt l This presumes that by some means you have already created on the disk a file named ARTSTAR MAG or whatever which contains centroid positions and instrumental magnitudes for the stars you want added to the picture This permits you to become just as sophisticated with your artificial star tests as you want you can simulate any color magnitude diagram luminosity function and spatial distribution in the frame that you want just by writing yourself a program which creates the necessary input files Note that in both examples the code asks for a number of photons per ADU It uses this to add the appropriate Poisson noise to the star images the correct amount of readout noise already exi
44. f the image Alternatively 50 COMPUTER TYPES YOU ENTER l l Command FU l Name for output picture default f DST lt CR gt or filename l Border pixels n l Polynomial order O constant 1 plane etc n l l First last column number 267 268 l First last row number 381 381 or 381 l l First last column number CTRL Z l Command l In this case you don t want to insert a single constant brightness value and or you don t know what value you would like to insert Instead you are asking the routine to consider a border n pixels wide around the rectangular area you have specified and use a least squares polynomial surface to interpolate values into the fudged region Now for the tricky bit The routine does not fit a single polynomial surface to the border and then use that polynomial to predict values for all the pixels in the fudged region Oh no Instead for each pixel in the rectangle to be fudged it fits a different polynomial surface to the border pixels employing 1 r weights Thus pixels in the corners and near the edges of the rectangular fudge region will closely reflect the data values and gradients in the pixels next to them and will be minimally affected by the gross gradient across the gap pixels in a long rectangular region will be affected by the pixels next to them and less by the pixels at the far ends Thus even a constant or a plane
45. fits and to highlight non stellar objects and tight binaries Additional uses for SUBSTAR 1 decrowding bright stars for the generation of an improved point spread function see Appendix on constructing a PSF and 2 decrowding bright stars for establishing the magnitude zero point of the frame by means of aperture photometry through a series of apertures Here s how it goes COMPUTER TYPES YOU ENTER Command SU File with the PSF default PSF lt CR gt or filename Do you have stars to leave in N Name for subtracted image default s DST lt CR gt or filename l l l File with photometry default NST lt CR gt or filename This will subtract from the image that is currently ATTACHed all the stars in the File with photometry The picture produced will have a format identical to your original picture it will be acceptable as input to DAOPHOT if you so desire 40 COMPUTER TYPES YOU ENTER Command SU File with the PSF default PSF lt CR gt or filename Do you have stars to leave in Y File with star list default LST lt CR gt or filename Name for subtracted image default s DST lt CR gt or filename l l l l l l File with photometry default NST lt CR gt or filename l l l l l l This time SUBSTAR will subtract from the image all the stars in the File with photometry except th
46. h can be executed in response to a Command prompt Rather ALLSTAR is a separate stand alone program which one executes directly from the operating system I have done it this way because that makes it easier to conserve virtual memory so that DAOPHOT and ALLSTAR can both operate on the largest possible images In general ALLSTAR works pretty much the same as the NSTAR routine in DAOPHOT fitting multiple overlapping point spread functions to star images in your CCD frames Input and output images and data files are fully compatible with those produced by DAOPHOT Some of the noteworthy differences between ALLSTAR and NSTAR 1 ALLSTAR reduces the entire starlist for a frame at once current maximum 15 000 stars With every iteration ALLSTAR subtracts all the stars from a working copy of your image according to the current best guesses of their positions and magnitudes computes increments to the positions and magnitudes from examination of the subtraction residuals around each position and then checks each star to see whether it has converged or has become insignificant When a star has converged its results are written out and the star is subtracted permanently from the working copy of the image when a star has disappeared it is discarded In either case the program has a smaller problem to operate on for the next iteration Throughout this process ALLSTAR maintains a noise map of the image including knowledge of the Poisson statistic
47. have specified represents the minimum central brightness enhancement over the local background which an object must have in order to be detected After having performed the convolution the program will then go through the convolved data 21 looking for local maxima in the brightness enhancement As the program finds candidates it computes a couple of image shape statistics named SHARP and ROUND which are designed to weed out delta functions bad pixels and brightness enhancements that are elongated in x or y bad rows and columns SHARP is defined as the ratio of the height of the bivariate delta function which best fits the brightness peak in the original image to the height of the bivariate Gaussian function with the user supplied value of the FWHM which best fits the peak If the brightness enhancement which was found in the convolved data is due to a single bright hot pixel then the best fitting delta function will have an amplitude equal to the height of this pixel above the mean local background while the amplitude of the best fitting Gaussian will be pulled down by the surrounding low valued pixels hence SHARP gt 1 On the other hand where there is a cold pixel that is to say where there is an isolated pixel whose brightness value is below the local average but still above the lowest good data value in the convolved data there will tend to be brightness enhancements found approximately 0 5 FWHM away from this p
48. he file MATHSUBS FOR 34 VIII PEAK PEAK is a single star profile fitting algorithm Because it is not to be trusted for the reduction of overlapping images PEAK should never be used for the final photometric reduction in fields where stars of interest may be blended On the other hand for sparsely populated frames aperture photometry is often fine and NSTAR or ALLSTAR photometry is virtually as fast Thus PEAK is largely a historical artifact reminding us that once life was simpler For you archaeology buffs here is how PEAK used to be used I now quote from the original DAOPHOT manual Obviously before PEAK can be run you must have created a point spread function Assuming this to be the case then COMPUTER TYPES YOU ENTER Command PE l l l l File with aperture results default AP lt CR gt or filename l File with the PSF default PSF lt CR gt or filename l l File for PEAK results default PK lt CR gt or filename If you have issued the MONITOR command see below or if WATCH PROGRESS 1 0 see the OPTIONS command above you will then see the results of the peak fitting photometry appear on your screen If WATCH PROGRESS 2 0 then you will also see an eleven level gray scale image of the region around each star as it is being reduced since it is very time consuming to produce these pictures on your terminal screen WATCH PROGRESS should be set to 2 0 only when it is strictly
49. he region of single stars may be drawn by eye or by some automatic scheme of your own devising Finally PEAK tells you the number of times that the profile fit had to be iterated The program gives up if the solution has been iterated 50 times without achieving convergence so stars for which the number of iterations is 50 are inherently more suspect than the rest Frequently however the solution was oscillating by just a little bit more than the convergence criterion which is fairly strict from one iteration to the next the computed magnitude must change by less than 0 0001 mag or 0 05 sigma whichever is larger and the x and y coordinates of the centroid must change by less than 0 001 pixel for the program to feel that convergence has been achieved Therefore in many cases stars which have gone 50 iterations are still moderately well measured One other thing of which you should be aware PEAK uses a fairly conservative formula for automatically reducing the weight of a bad pixel not a bad pixel as defined in FIND PEAK ignores those but rather any pixel which refuses to approach the model profile as the iterative fit proceeds This formula depends in part on the random errors that the program expects based on the values for the readout noise and the photons per ADU which you the user specified in the aperture photometry table It is therefore distinctly to your advantage to see to it that the values supplied are re
50. hotons per ADU for a single frame for the detector you used Specify these as options FIND will 2 compute the readout noise in ADU r 1 frame photons p 1 frame photons ADU R 1 frame ADU 3 correct the ratio of photons per ADU and the readout noise in ADU for the number of frames that were averaged or summed E ese ter ceo o a ores cua A oe a ee a ek PE ae ee photons ADU p N Nx p 1 pN p 1 r o noise R N R 1 SQRT CN R N R 1 SQRT N 4 determine the typical sky brightness s in your data frame by using the DAOPHOT command SKY 5 compute the random error per pixel random error in 1 pixel 4 s py R multiply by the relative error defined above to arrive at the random noise in the sky background a of the convolved data frame background noise relative error x random error per pixel Some multiple of order 3 5 say of this background noise as specified by the user is used as your desired star detection threshold 65 APPENDIX III DERIVING A POINT SPREAD FUNCTION IN A CROWDED FIELD Obtaining a good point spread function in a crowded field is a delicate business so please do not expect to do it quickly plan on spending a couple of hours in the endeavor the first few times you try it After that it gets easier and once you know what you re trying to do it s fairly easy to write command procedures to handle major parts of the FIT SUBT
51. hreshold for further details According to a parameter set by the user FIND will also compute a Lowest good data value any pixel whose brightness value is 7 VI VII VII IX XI XII XIII XIV less than some number of standard deviations below the mean sky value will be regarded as bad and will be ignored by FIND and by all subsequent reduction stages Use PHOTOMETRY to obtain sky values and concentric aperture photometry for all objects found by the star finding routine Use PICK to select a set of reasonable candidates for PSF stars PICK first sorts the stars by magnitude and then rejects any stars that are too close to the edge of the frame or to a brighter star It will then write a user specified number of good candidates to a disk file for use by PSF Use PSF to define a point spread function for the frame In crowded fields this is a subtle iterative procedure requiring an image processing system it is outlined in detail in the Appendix on Obtaining a Point Spread Function Consider then that this step is a self contained loop which you will go through several times GROUP NSTAR and SUBSTAR or ALLSTAR GROUP divides the stars in the aperture photometry file created in step VI above into finely divided natural groups for reduction with the multiple star PSF fitting algorithm NSTAR NSTAR will then produce improved positions and instrumental magnitudes by means of multiple profile f
52. htness be This is a much harder question to answer partly because the answer cannot be obtained by an empirical method as simple as finding the mode of an observed distribution and partly because the criteria for including a star in the star list are hard to define absolutely For instance a faint star is far easier to see in a clear sky zone where its contamination can be identified and then ignored in the sky brightness estimate than it would be if it happened to lie near a much brighter star whose intrinsic brightness we are trying to measure Clearly then when we detect a faint star in the sample sky region the decision whether to include or reject that star s photons in the sky estimate becomes some function of the magnitude of the star we are interested in measuring Further a serious attempt to estimate the sky brightness using probabilistic methods would require an accurate model for the full noise spectrum of the instrument including the arrival rate and energy spectrum of cosmic rays the surface density of stars and galaxies on the sky their apparent luminosity functions and the variations of these quantities across the frame Thus a definitive answer to the question How bright is the sky here is exceedingly hard to obtain with full statistical rigor For the present we must be content with a merely adequate answer I have discussed the problem of sky determination in such philosophical detail as a warning to the user not
53. ian over the area of any given pixel may be evaluated as the product of two one dimensional integrals 2 A Moffat function having three free parameters half width at half maximum in x and y and effectively a position angle for the major axis of the ellipse Since it s necessary to compute the two dimensional integral anyway we may as well let the ellipse be inclined with respect to the cardinal directions In case you don t know it a Moffat function is 1 Ma where 2 is something like 17 03 y 0 asyzy Note not 1Y 0xy SO Ary can be zero In this case 8 1 5 3 A Moffat function having the same three parameters free but with 8 2 5 4 A Lorentz function having three free parameters ditto 5 A Penny function the sum of a Gaussian and a Lorentz function having four free parameters As always half width at half maximum in x and y the fractional amplitude of the Gaussian function at the peak of the stellar profile and the position angle of the tilted elliptical Gaussian The Lorentz function may be elongated too but its long axis is parallel to the x or y direction 6 A Penny function with five free parameters This time the Lorentz function may also be tilted in a different direction from the Gaussian 33 It is possible that these assignments will be changed or augmented in the future If you are worried about the details I suggest you consult the source code the routine PROFIL in t
54. ile for NSTAR results default NST lt CR gt or filename The principal difference is that NSTAR is much more accurate than PEAK in crowded regions although at the cost of requiring somewhat more time per star depending on the degree of crowding NSTAR automatically reduces the weight of bad pixels just as PEAK does so it is highly advisable that your values for the readout noise and the number of photons per ADU be just as correct as you can make them NSTAR also produces the same image peculiarity statistics CHI and SHARP defined as they were in PEAK Plots of these against apparent magnitude are powerful tools for seeing whether you have specified the correct values for the readout noise and the photons per ADU Do most stars have CHI near unity Is there a strong trend of CHI with magnitude for identifying stars for which the profile fits just haven t worked they will have values of CHI much larger than normal for stars of the same derived magnitude for identifying probable and possible galaxies they will have larger values of SHARP than most for identifying bad pixels and cosmic rays that made it through FIND large negative values of SHARP and for seeing whether your brightest stars are saturated the typical values of SHARP and CHI will tend to increase for the very brightest stars The maximum number of iterations for NSTAR is 50 but every star in a group must individually satisfy the convergence criteria see PEAK before the
55. ing a point spread function in a crowded field is still done best with at least a small modicum of common as distinguished from artificial intelligence One possible procedure for performing this task is outlined in an Appendix At this point I will say only that in the beginning you may find it to your advantage to perform this task interactively while you are able to examine your original picture and its offspring with various of the stars subtracted out by procedures described in the Appendix on an image display system After you find that you are performing exactly the same operations in exactly the same sequence for all your frames you may choose to write command procedures to carry out the bulk of this chore with a visual check near the end DAOPHOT Classic assumed that the point spread function of a CCD image could be adequately modeled by the sum of a an analytic bivariate Gaussian function with some half width in the x direction and some half width in the y direction and b an empirical look up table representing corrections from the best fitting Gaussian to the actual observed brightness values within the average profile of several stars in the image This hybrid point spread function seemed to offer both adequate flexibility in modelling the complex point spread functions that occur in real telescopes with some hope of reasonable interpolations for critically sampled or slightly undersampled data This approximation has since turned ou
56. ing to any of the auxiliary information in any output file stars can thus be reordered by their sharpness or roundness indices by their magnitudes in any aperture by their sky brightness by the number of iterations they required to 43 converge in PEAK or NSTAR or whatever The value of n which should be entered to specify one of these items is that which is given at the bottom of each sample output file in Appendix IV below Note data on the second line for a star in an AP file are always designated by 16 17 18 regardless of how many apertures were used Thus if two apertures were used then the magnitude in aperture 1 is number 4 the magnitude in aperture 2 is number 5 the sky brightness is number 16 the standard deviation of the sky brightness is number 17 the skewness of the sky brightness is number 18 the error in the first magnitude is 19 and the error in the second magnitude is 20 Numbers 6 15 and 21 30 are useless in this example 44 XV SELECT When you run GROUP on a photometry file to create a group file suitable for input for NSTAR you may find that some groups are larger than 60 stars which is the current maximum group size allowed The SELECT command allows you to cut out of the group file only those groups within a certain range of sizes and put them in their own file nnn stars in nnn groups COMPUTER TYPES YOU ENTER l l Command SE l Input group file filename l l l Minimum
57. ions corresponding to the positions and magnitudes derived for the stars during the photometric reductions ALLSTAR II is a separate stand alone program which performs a much more sophisticated multiple profile fit to all the stars in a frame simultaneously Hereinafter I will include ALLSTAR II under the generic heading of DAOPHOT II even though it is as I said a separate stand alone program In addition to the aforementioned tasks DAOPHOT II contains routines to perform some bookkeeping operations more easily than may be the case with standard facilities e g estimating an average sky brightness for a frame sorting the stars output data according to their positions in the frame or their apparent magnitudes and dividing the stars in the frame into natural groupings for optimal multiple star reductions with NSTAR There is also a routine for adding artificial stars to the picture at random so that the effectiveness of the star finding and profile fitting routines can be studied quantitatively in the particular circumstances of your own picture A few other global considerations of which you should be aware 1 Although DAOPHOT II is designed to be non interactive in fact many of the operations run 3 quickly enough that they are conveniently executed directly from the terminal or workstation Only the multiple star profile fits take long enough that they are more conveniently performed in batch mode they may require anywhere from a few
58. its and SUBSTAR may then be used to subtract the fitted profiles from the image producing a new image containing the fitting residuals Alternatively you could feed the aperture photometry file directly to ALLSTAR which will reduce all the stars in the image simultaneously and produce the star subtracted picture without further ado Use ATTACH to specify the star subtracted picture created in step IX as the one to work on Use FIND to locate new stars which have become visible now that all the previously known stars have been subtracted out Use ATTACH again this time specifying the original picture as the one to work with and use PHOTOMETRY to obtain sky values and crude aperture photometry for the newly found stars using the coordinates obtained in step XI You are performing this operation on the original picture so that the sky estimates will be consistent with the sky estimates obtained for the original star list Use GROUP on the new aperture photometry file you just created Use GROUP again on the profile fitting photometry file created in step IX this step is unfortunately necessary to put both the old and new photometry into files with the same format so that you can Use APPEND to combine the two group files just created into one GROUP SELECT SELECT GROUP SELECT SELECT NSTAR SUBSTAR or ALLSTAR If for some reason you prefer NSTAR to ALLSTAR I sure don t the file just created in step XIII needs
59. ixel in all directions in such a case the height of the delta function which best fits one of these spurious maxima tends to be close to zero while the height of the best fitting Gaussian is some small positive number SHARP 0 To reject both types of bad pixel the default acceptance region for the SHARP parameter is height of best fitting delta function 20 lt SHARP ete height of best fitting Gaussian function lt 1 00 ROUND is computed from the data in the original picture by fitting one dimensional Gaussian functions to marginal sums of the data in x and y Specifically for each brightness enhancement which passes the SHARP test if the height of either of these one dimensional Gaussian distributions happens to be negative a local minimum in the brightness distribution in that coordinate sometimes happens or zero the object is rejected If both turn out to be positive then the ROUND parameter is computed ROUND difference between the heights of the two one dimensional Gaussians average of the heights of the two one dimensional Gaussians Thus if the two heights are say 90 and 150 ADU then the average is 120 ADU and the difference is 60 ADU so that ROUND would be 0 5 The sense of the difference is such that an object which is elongated in the x direction has ROUND lt 0 and one which is elongated in the y direction has ROUND gt 0 If the brightness enhancement which has been detected is really a
60. l y s p R units are ADU 62 Please note that if the frame you are working on is the average or the sum of several raw data frames the values of the gain factor in photons per ADU and the readout noise will have to be adjusted accordingly SQRT s N p 1 R 2 N SQRT s p 1 N R 1 2 If N frames were averaged If N frames were summed photons ADU p N N p 1 pN p t r o noise R N R 1 SQRT N RCN R 1 SQRT N total std dev N std dev N The value of s which FIND uses in these equations is the number you get when you issue the DAOPHOT command SKY Note that the expectation value of s scales as follows s N s 1 s N N s 1 std dev N SQRT s 1 N p 1 R 1 2 N std dev N SQRT s 1 N p 1 N R 1 p 1 2 std dev 1 SQRT N std dev 1 SQRT N Therefore to ensure that the statistics are being done properly for your frames FIND takes the readout noise per frame IN UNITS OF ADU and the ratio of photons per ADU and corrects them for the number of frames that have been averaged or summed according to the first table above 63 TO ESTABLISH A REASONABLE THRESHOLD FOR YOUR STAR FINDING FIND computes the
61. lename And away we go With each iteration the program will keep you updated on how many stars remain to be reduced how many have disappeared due to insignificance and how many have converged and been written out Finally I would like to reiterate that in the final analysis I wrote ALLSTAR like DAOPHOT for me I make no promise to leave it alone or to notify you of minor changes If I discover some major bug that destroys the science I may remember that I gave you a copy and then again I may not As I get time to play I will certainly be attempting to make the program more powerful and reliable for my applications Therefore in some sense you use this copy of the program and of DAOPHOT ID at your own risk Use it as long as you are happy with the results If you don t like what you are getting stop using the program and complain to me Maybe I will have already fixed your problem or maybe your problem will be interesting or important enough that I will want to fix it for you However I get bent all out of shape when somebody has a problem with my software and publishes complaints in the literature without ever coming to me to give me a chance to fix it for them or to explain some point they may have misunderstood 60 APPENDIX I Optional parameters Routines Affected Permitted values Default value HI FW TH LS HS LR HR WA FI PS VA FR AN EX Readout noise 1 exposure ADU 1 Gain 1 expos
62. magnitudes and local sky values for the objects just found and obtains a point spread function in the manner described in an Appendix to this manual Second one performs a profile fitting reduction run for these objects and they are subtracted from the data frame This new picture with the known stars subtracted out is then subjected to the star finding procedure stars which were previously concealed in the profiles of brighter stars stand out in this frame and are picked up quite effectively by the star finding algorithm Sky values and aperture magnitudes for these new stars are obtained from the original data frame and the output from this reduction is appended to the most recent photometry file for the original star list This augmented set of stars is then run through the profile fitting code and the entire list of fitted stars can be subtracted from the original frame The process through this point can easily be set up in a command procedure or script or whatever your favorite rubrik is and carried out in batch mode while you are home sleeping or drinking beer or whatever Finally if absolute completeness is wanted the star subtracted picture can be examined on an image display Any stars that were still undiscovered by the program can be picked out by eye and added to the star list manually Then one final reduction run may be performed Visual examination is also a reasonable way to identify galaxies among the program objects they a
63. maximum group size nn nn l Output group file default GRP lt CR gt or filename l l l l If you want to try to reduce every star in the frame regardless of the errors then you will need to run SELECT at least twice once with minimum maximum group size 1 60 and again with the same input group file a different output group file and minimum maximum group size 61 9999 say This latter file would then be run through GROUP again with a larger critical overlap Note You d be better off using ALLSTAR 45 XVI OFFSET If you have a set of coordinates for objects found in one frame and want to use these as centroids for aperture photometry in another frame and if there is some arbitrary translational shift between the two frames then OFFSET can be used to add constants to all the x and y coordinates in a stellar data file or n nn nn nn nn Output file name default 0FF lt CR gt or filename COMPUTER TYPES YOU ENTER Command OF Input file name filename l Additive offsets ID DX DY DMAG n nn nn nn nn n nnn l l As you can tell from the example you can also use this routine to add some integer constant to the ID numbers or some real constant to the magnitudes Why would you ever want to do these things Well if you ve just run FIND on a star subtracted image you ll probably want to add 50000 or something to the ID numbers before appending
64. n and the program will then permit you to enter parameter values from the keyboard The syntax for specifying a parameter value either from within a file or from the keyboard is as follows The parameter you wish to define is indicated by two alphanumeric characters it doesn t matter whether they are upper or lower case and any spaces or additional characters except an equals sign and a number after the first two are optional The parameter identifier is followed by an equals sign and this is followed by a number The following commands would all set the FWHM of the objects for which the search is to be optimized to the value 3 0 FW 3 0 fwhm 3 Fwied wice is nice 3 When inputing parameter values from a file one parameter is specified per line Note that only those parameters whose values you want to change from the program supplied defaults need be supplied by the user either in manner la above or in 2a or 2b You exit from OPTIONS by responding to the OPT gt prompt with a carriage return or a CTRL Z CTRL D on some Unix machines Whatever means END OF FILE on your system 17 EXAMPLE CHANGING VALUES FROM THE KEYBOARD To change the estimated full width at half maximum from 2 5 to 3 5 and the high sharpness cutoff from 1 0 to 0 9 COMPUTER TYPES YOU ENTER Command OP COMPUTER TYPES READ NOISE ADU 1 frame 5 00 GAIN e ADU 1 frame 10 00 l LOW GOOD DATUM sigmas 7 00 HIGH GOOD DAT
65. n time and in batch mode it will fill up your logfile excessively FITTING RADIUS Most obviously this parameter defines the circular area within which pixels will actually be used in performing the profile fits in PEAK and NSTAR As the point spread function is shifted and scaled to determine the position and brightness of each of your program stars only those pixels within one fitting radius of the centroid will actually be used in the fit More subtly the same region is also used in fitting the analytic first approximation to the point spread function for the PSF stars Moreover the parameter will also contribute in a minor way to the determination of when stars overlap significantly Under normal circumstances this radius should be of order twice the half width at half maximum of a stellar image which is obviously the same as the FWHM When the crowding is extremely severe however it may be advantageous to use a value somewhat smaller than this On the other hand if the point spread function is known to vary across the frame then increasing the fitting radius beyond the FWHM may improve the photometric accuracy provided of course that the field is not horribly crowded PSF RADIUS The radius in pixels of the circle within which the point spread function is to be defined This should be somewhat larger than the actual radius of the brightest star you are interested in as you would measure it on your image display If
66. nge any you don t like FITTING RADIUS 2 50 CE CLIPPING EXPONENT 6 00 REDETERMINE CENTROIDS 1 00 CR CLIPPING RANGE 2 50 WATCH PROGRESS 1 00 MAXIMUM GROUP SIZE 50 00 PERCENT ERROR in 0 75 PROFILE ERROR in 5 00 IS INNER SKY RADIUS 0 00 OS OUTER SKY RADIUS 0 00 OPT gt The FITTING RADIUS and the WATCH PROGRESS options you are familiar with from DAOPHOT the fitting radius is the size of the region around each star which will actually be used in performing the profile fits The WATCH PROGRESS option determines just how much garbage will be typed onto your terminal screen or written into your batch job s logfile The MAXIMUM GROUP SIZE option has been explained above The REDETERMINE CENTROIDS CLIPPING EXPONENT and CLIPPING RANGE options are new The REDETERMINE CENTROIDS option is how you tell the program whether to improve the stars positions in the profile fits RE 0 means no 1 means yes If you enter no the program will assume that the positions of the stars are known with absolute accuracy and will redetermine only the stars magnitudes this is one way of imposing a star list from a good frame onto a poor frame of the same region if yes you will get the full blown astrometric and photometric reduction you are used to from NSTAR The CLIPPING EXPONENT and CLIPPING RANGE options are explained in Stetson 1987 PASP 99 191 111 D 2 d Resisting bad data The CLI
67. o cautious Run PSF tell it the name of your complete sorted renumbered aperture photometry file the name of the file with the list of PSF stars and the name of the disk file you want the point spread function stored in the default should be fine a If you have the WATCH PROGRESS option equal to 1 or 2 PSF will produce on your terminal a gray scale plot of each star and its environs it will tell you the number of ADU in the brightest pixel within the area displayed and then it will ask whether you wish to include this star in the point spread function to which you should answer Y or N as appropriate If WATCH PROGRESS is 0 it will go ahead and use the star unless it finds a bad pixel more than one fitting radius and less than one PSF radius plus 2 pixels from the star s a el we center in such a case it will ask whether you want the star c If WATCH PROGRESS 1 or 2 it will use the star regardless counting on the bad data rejection to clean out any bad pixels It will report the presence of bad pixels but it 66 10 will use the star anyway If the frame is crowded it is probably worth your while to generate the first PSF with the VARIABLE PSF option set to 1 pure analytic PSF That way the companions will not generate ghosts in the model PSF that will come back to haunt you later You should also have specified a reasonably generous fitting radius these
68. o set IS to at least one or two pixels to keep the enhanced photon noise and the uncertain PSF near the center of a star image from overly influencing the derived sky values As long as OS gt IS before every third iteration starting with iteration 3 after all the stars have been subtracted from the working copy of the image the program will redetermine the median brightness value in the annulus bounded by these two radii This is used as the sky estimate for the star for the next three iterations In my opinion this is just about the best way currently available to get at the ever sought ever elusive SKY UNDER THE STAR I do it this way rather than solving for the sky brightness or some analytic function representing the spatial distribution of the sky brightness as part of the least squares profile fits because it is far far easier faster and more precise to determine the median of several hundred pixels than to fit a robust least squares surface to a few dozen This is discussed at some length in Stetson 1987 PASP 99 101 III D 2 a Perhaps in ALLSTAR II v decimal Roman numerals far out I will include optional fitting of a sky model in the least squares problem If OS lt 58 IS ALLSTAR will continue to use the sky brightness values that were in the input data file Like DAOPHOT ALLSTAR begins by showing you a table of optional parameter settings unlike DAOPHOT but like PHOTOMETRY it immediately gives you a chance to cha
69. of the picture The crude magnitudes which FIND computes and includes in your terminal and disk file output are defined relative to the threshold which you gave it a star with a FIND magnitude of 0 000 is right at the detection limit Since most stars are brighter than the threshold FIND will obviously give them negative magnitudes For the faintest stars the magnitudes may be quantized since if the original image data are stored as integers the convolved image data will be too Thus if your FIND threshold came out to 20 0 ADU the next brighter magnitude a star may have after 0 000 is 2 5 log 21 20 0 053 When the image data are stored as floating point numbers this quantization will not occur 24 VI PHOTOMETRY Before you can do concentric aperture photometry with DAOPHOT you need to have an aperture photometry parameter file At the DAO a prototype parameter file named DAO PHOTO OPT is available The file looks something like this When you give DAOPHOT the PHOTOMETRY command to invoke the aperture photometry routine COMPUTER TYPES YOU ENTER Command PH Enter table name default PHOTO OPT lt CR gt or table name A1 RADIUS OF APERTURE 1 3 00 A2 RADIUS OF APERTURE 2 4 00 A3 RADIUS OF APERTURE 3 5 00 A4 RADIUS OF APERTURE 4 6 00 AS RADIUS OF APERTURE 5 7 00 A6 RADIUS OF APERTURE 6 8 00 A7 RADIUS OF APERTURE 7 10 00 A8 RADIUS OF APERTURE 8 0 00 A9 RADIUS OF APERTURE 9 0 00
70. ommand mode 49 XIX FUDGE Although it is morally wrong someday there may come a time when you just have to fudge some of your image data Suppose for instance that you are trying to derive a point spread function from the sole acceptable star in the frame and way way out in the corner of the box wherein the point spread function is to be defined there is a cosmic ray hit If you do nothing then the cosmic ray will produce a spike in the point spread function which will generate a hole whenever this PSF is used to subtract stars In such a desperate case it may be slightly the lesser of two evils to fudge the bad datum which DAOPHOT s FUDGE routine will do Let us assume that from your image display you have ascertained that the cosmic ray occupies the two pixels 267 381 and 268 381 let us further suppose that you see from the aperture photometry for this star that the sky background brightness in its neighborhood is 893 ADU Then COMPUTER TYPES YOU ENTER Command FU l Name for output picture default f DST lt CR gt or filename Border pixels 0 First last column number 267 268 First last row number 381 381 or 381 Brightness value 893 First last column number CTRL Z Command As may be inferred from the example FUDGE allows you to insert any constant brightness value into any rectangular subsection of an otherwise exact copy o
71. on The derived value of this threshold represents the minimum central height of a star image in ADU above its local sky background which is required for a detection to be regarded as statistically significant The theory behind star detection is discussed briefly below and the method used for computing the optimum threshold is described in the Appendix on The FIND Threshold The routine also computes a lowest good data value corresponding to the number of standard deviations you specified in the LOW GOOD DATUM option relative to the average sky brightness That is having determined that the modal sky level in this frame is 156 8 it determines that the brightness 70 below this is 136 8 20 For this calculation it uses the specified readout noise gain and the fact that this is the average of five frames not the observed o sky 4 16 because this latter value may have been puffed up by stars and defects Then any pixel anywhere in the frame which has an observed brightness value less than 136 8 or greater than the HIGH GOOD DATUM which you specified directly will now and ever after be discarded as defective If you want to see what numerical values the routine gave to the star detection threshold and the lowest good data value you can find them in the second line of the output file created by this routine If you want to change either of these values for subsequent reduction steps you must do it in the file header lin
72. ons until it falls apart into subgroups each containing fewer than some user specified number of stars Other associations which are already smaller than the maximum size will not be regrouped with the smaller critical separation Only the ones that are too big will If the group is so dense that the critical separation drops to less than 1 2 pixels the program 57 will arbitrarily delete the faintest star in the group and proceed from there By increasing the value of the optional parameter MAXIMUM GROUP SIZE the user may reduce the number of faint stars that get rejected by this mechanism on the other hand these stars will be so poorly measured that they may not be worth retaining Of course increasing the MAXIMUM GROUP SIZE will also greatly increase the amount of time required for reducing the frame since the reduction time goes roughly as the cube of the size of the largest associations ALLSTAR will produce the star subtracted image for you directly without need to run the DAOPHOT routine SUBSTAR If you don t want the output picture produced when the program asks you for the star subtracted image filename respond with a CTRL Z or type the words END OF FILE in capitals If you think that the stars positions are very well known ahead of time for instance because you have averaged their observed positions on a number of frames or because you can apply positions derived from an excellent seeing frame to poorer frames
73. or to remove the extra file header from the middle of the newly created file APPEND does this for you automatically COMPUTER TYPES YOU ENTER l Command AP l First input file filename l Second input file filename l l l Output file default CMB lt CR gt or filename l Note that APPEND does no checking to ensure that the input files are of the same type the user is perfectly able to APPEND output from FIND onto output from PHOTOMETRY The resulting hybrid file would be illegible to most other DAOPHOT routines and might cause a crash if DAOPHOT tried to read it in Note further that the GROUP command above makes a point of leaving a blank line at the end of every group file so that the groups will remain distinct when APPENDed If in editing group files you delete that last blank line then when APPENDing those files the last group of the first input file and the first group of the second input file will be joined together in the output file 47 XVIII DUMP DUMP permits you to display on your terminal screen the raw values in a specified square subarray of your picture This is useful for figuring out why DAOPHOT isn t working with your data maybe all the pixels have intensities of 32768 or in deciding whether a particular star has a central intensity above the maximum value where your detector behaves linearly Example COMPUTER TYPES YOU ENTER Command DU Box
74. or the sum of several independent readouts readsout of the chip The program uses this information to adjust the readout noise and the gain appropriately In the example here I have assumed that five individual exposures were averaged together to make the frame being reduced here that is the five frames were added together and the sum was divided by the scalar value 5 If the frames had been added together and not divided by five I would have entered 1 5 instead of 5 1 If I had taken six frames summed them by pairs and then averaged the three different sums I would have entered 3 2 meaning the average of three sums of two If I had taken six frames averaged the first three averaged the second three and then summed the two averages I would also have entered 3 2 averages of three sum two of them One final nuance it happens that from a statistical noise point of view the median of three frames is about as good as the average of two Therefore if the frame you are reducing represents the median of several independent exposures to remove cosmic rays or whatever enter it as if it were the average of two thirds as many frames the median of three images would be entered as 2 1 and the median of five would be entered as 3 3 1 With this information the routine will calculate a star detection threshold corresponding to the number of standard deviations which you entered as the THRESHOLD opti
75. ose that appear in the File with star list This makes it easy to clean out stars around your PSF stars or around the stars for which you wish to perform concentric aperture photometry Note however that stars are cross identified solely on the basis of their ID numbers in the various files If you use the renumber option of the SORT command or if you APPEND together files where ID numbers are duplicated you may find yourself leaving the wrong stars in the image 41 D Additional Commands XI MONITOR NOMONITOR If you want simply to turn off the sending of the results to your terminal screen this can be accomplished without going through all the foofarah of the OPTIONS command by using the NOMONITOR command COMPUTER TYPES YOU ENTER Command NO Similarly after the NOMONITOR command or the WATCH PROGRESS 0 option output to your terminal screen can be restored with the MONITOR command COMPUTER TYPES YOU ENTER l Command MO MONITOR and NOMONITOR set the WATCH PROGRESS parameter to 1 and 0 respectively If you want the special effects produced by setting WATCH PROGRESS to 2 1 or 2 you must set it explicitly in your DAOPHOT OPT file or with the OPTIONS command 42 XIV SORT This routine will take in any stellar data file produced by DAOPHOT viz files produced by FIND PHOTOMETRY PEAK GROUP NSTAR ALLSTAR or any of the other auxiliary routines discussed below and re order the sta
76. pecific properties of your picture may be altered The sigificance of each one is described below in reference to the particular routines affected and a reference table is presented as an Appendix Here the options will be simply enumerated and the procedures available for changing the parameters will be described 1 na At present the user is permitted to specify values for nineteen parameters READ NOISE The readout noise in data numbers of a single exposure made with your detector Later on the software will allow you to specify whether a given data frame is in fact the sum or the average of several individual exposures If you have a separate subdirectory for data from a given night or observing run the readout noise needs to be specified only once in the DAOPHOT OPT file see below GAIN The gain factor of your detector in photons or electrons per data number As with the readout noise you want to specify the gain corresponding to a single exposure allowance can be made later for frames that are in fact the averages or sums of several exposures For both READ NOISE and GAIN the default values are deliberately invalid You must put correct values for these parameters in a DAOPHOT OPT file or the program will hassle you again and again and again and LOW GOOD DATUM The level in standard deviations below the frame s mean sky value below which you want the program to consider a pixel defe
77. r You must determine what value of critical overlap is suitable for your data frame by trial and error if a critical overlap 0 1 divides all of the stars up into groups smaller than 60 then you may be sure that unavoidable random errors will dominate over crowding If critical overlap 1 0 works then crowding will be no worse than the random errors If critical overlap gt gt 1 0 is needed then in many cases crowding will be a dominant source of error After GROUP has divided the stars up and created an output disk file containing these natural stellar groups a little table will be produced on your terminal showing the number of groups as a function of their size If any group is larger than the maximum acceptable to NSTAR currently 60 stars then the critical overlap must be increased or the SELECT command see below should be used to cut the overly large groups out of the file When crowding conditions vary across the frame judicious use of GROUP and SELECT will pick out regions of the data frame where different critical overlaps will allow you to get the best possible photometry for stars in all of the crowding regimes 37 X NSTAR NSTAR is DAOPHOT s multiple simultaneous profile fitting photometry routine It is used in very much the same way as PEAK COMPUTER TYPES YOU ENTER Command NS File with the PSF default PSF lt CR gt or filename File with stellar groups default GRP lt CR gt or filename F
78. r is the typing out on your terminal of a brief message notifying you that some information to your advantage is now available If this message has changed since the last time you ran the program answer the question with a capital or lower case Y lt CR gt The program will then type out the text of the entire message a section at a time It will pause at the end of each section of the message to allow you to read what it s written before it rolls off the screen or to escape to the main program without reading the rest of the message When you reach the main part of the program the current values of the optional reduction parameters will appear on your screen see the Appendix on Optional Parameters and the OPTIONS command below When you see the Command prompt the program is ready to accept the commands described below 11 II ATTACH If you want to work on a digital picture the first thing you should do is specify the disk filename of that picture with the ATTACH command COMPUTER TYPES YOU ENTER Command AT filename Your picture s header comment if any Your picture s header comment if any Picture size nnan nnn 4 5 5 5 5 5 5 5 5 5 or COMPUTER TYPES YOU ENTER Command AT Enter file name filename Some commands the ones which operate only on data files e g SORT OFFSET APPEND and others which set the optional par
79. random error per pixel in units of ADU from s random error in 1 pixel R PN where s is the number you get from SKY and appropriate values of py and Ry have been computed according to the table above When you then run FIND you will discover that it also gives you a number called relative error This is because in trying to establish whether there is a star centered in a certain pixel FIND operates on weighted sums and differences of several adjacent pixels The relative error is merely a scaling parameter it is the number that the standard error of one pixel must be multiplied by to obtain the standard error of the smoothed differenced data in the convolved picture Therefore to arrive at a reasonable threshold FIND computes the standard error per pixel as described above multiplies it by the relative error factor and sets the threshold at the multiple of this number which you have specified with the THRESHOLD option say Threshold 3 5 x relative error x standard error in one pixel for 3 5 0 detections A 3 5 0 excursion occurs about 233 times per one million independent random events In an otherwise empty 300 x 500 frame this would produce about 35 false detections In a frame some non trivial fraction of whose area was occupied by real stars the false detections would be fewer 64 TO SUMMARIZE 1 Ascertain the values of the readout noise in electrons or photons and the number of p
80. re easily recognizable with over subtracted centers surrounded by luminous fuzz My experience is that the number of stars found in the second pass the automatic star finding on the first subtracted frame amounts to of order one third the number of stars found in the first pass The number of stars missed in the first two passes and later picked out by eye is of order one to three percent of the total found in the first two passes This procedure assumes that computer time is cheap for you and your own time is valuable If the converse is the case 4 you may prefer to skip the second or even both automatic star finding passes and go directly to interactive star identification 5 A principal source of photometric error for the faint stars is the difficulty of defining what is meant by the term sky brightness in crowded fields This is not simply the practical difficulty of identifying contaminated pixels in the sky annulus so that they can be omitted from the average although certainly this is a significant part of the problem There is also an underlying philosophical ambiguity For aperture photometry the term sky brightness encompasses not only emission from the terrestrial night sky from diffuse interplanetary and interstellar material and from faint unresolved stars and galaxies It also includes the possibility of a contribution of light from some bright star or galaxy That is to say for aperture photometry the rele
81. rs according to position within the frame apparent magnitude identification number or other available datum e g magnitude error number of iterations CHI or SHARP Of course if a file produced by GROUP or NSTAR is sorted then the association of stars into groups will be destroyed COMPUTER TYPES YOU ENTER Command SO The following sorts are currently possible 1 By increasing decreasing star ID number 2 By increasing decreasing X coordinate 3 By increasing decreasing Y coordinate n By increasing decreasing OTHER n lt 30 Which do you want n Input file name filename Output file name default 7 lt CR gt or filename 4 By increasing decreasing magnitude Do you want the stars renumbered Y or N If you answer the question Which do you want with 4 the stars will be reordered by increasing apparent magnitude if by 4 they will be reordered by decreasing apparent magnitude and so forth If you say that you want the stars renumbered the first star in the new output file will be given the identification number 1 the second star 2 and so on The output file will contain exactly the same data in exactly the same format as the input file the stars will just be rearranged within the file The n option permits you to sort accord
82. s observe the magnitude fainter than which these indices blow up and determine the range of values of SHARP and ROUND spanned by stars above that magnitude limit Having decided on the region of each of the two diagrams occupied by worthwhile stars you could then rerun FIND with a new threshold and new values for the limits on SHARP and ROUND Alternatively you could write yourself a quickie program which goes through the output file produced by FIND and rejects those objects which fall outside your new acceptance limits which could even be functions of magnitude if you so desired Back in FIND Once a brightness enhancement passes muster as a probable star its centroid is computed approximately When the FIND routine has gone through your entire picture it will ask COMPUTER TYPES YOU ENTER Are you happy with this Y or N If you answer Y the program will exit from FIND and return you to DAOPHOT Command mode If you answer N it will ask for a new threshold and output filename and will then search through the convolved picture again A couple more comments on star finding 1 Stars found in the outermost few rows and columns of the image that is where part of their profile hangs over the edge of the frame do not have their positions improved Instead their positions are returned only to the nearest pixel Furthermore the ROUNDNESS index is not computed for such stars although the SHARPNESS
83. s of stars that have previously converged and been permanently subtracted from the working copy Since the entire star list is a group in the sense of NSTAR ALLSTAR does not require that the starlist be GROUPed ahead of time and is perfectly happy to run from your AP files 2 In order to make the problem tractable after all we re potentially dealing with a 45 000x 45 000 matrix with every iteration ALLSTAR does associate stars with one another in order to make the big matrix block diagonal so it can be inverted in a finite amount of time These associations are temporary they are reconsidered every iteration and they do not compromise the full rigorous simultaneous least squares solution of the entire star list ALLSTAR will do the best it can to reduce your data frame regardless of the degree of crowding What happens is the following during the process of each iteration ALLSTAR automatically associates all the stars into the frame into manageable bite sizes on the basis of a critical separation which is calculated from the fitting radius If you have the WATCH PROGRESS option equal to 1 then in a densely packed frame you might notice ALLSTAR typing out messages like Group too large 107 2 50 That means that a group of 107 stars resulted from use of the nominal critical separation 2 50 pixels in this case ALLSTAR will attempt to break up this group by regrouping it with smaller and smaller critical separati
84. sonable modal sky value very rare this latter case will be flagged by O sky 1 3 Very important If after the rejection of the tails of the sky histogram the sky finding algorithm is left with fewer than 20 pixels PHOTOMETRY will refuse to proceed and will return to to DAOPHOT II command mode Check your inner and outer sky annulus radii and your good data value limits In particular if you are reducing a short exposure frame where the mean sky brightness is small compared to the readout noise the lowest good data value computed in 27 FIND see above should have been set to some moderate sized negative number If this is not the case there ll be bad pixels all over the place and you ll never get any photometry done As suggested above I generally set the bad pixel threshold somewhere around five to six sigma below the typical sky brightness obtained from the SKY command see above unless the background is significantly non uniform in which case I set it even lower But if you are absolutely sure that the sky background does not change significantly across your frame and you want to be really clever you can set the threshold to something like 4 35 sigma below the mean sky brightness the one sided tail of the cumulative normal distribution Prob X lt 4 350 7 x 107 or about one legitimate random pixel being spuriously identified as bad in a 300 x 500 picture 28 VII PICK and PSF In my opinion obtain
85. sts in the frame For realistic tests you should specify the correct value for this number If for some reason you would like to have the scaled PSF added into the image without extra Poisson noise just specify some enormous number of photons per ADU such as 99999 To avoid small number statistics in your artificial star analysis you should create a number of different frames each containing only a few extra stars If you try to add too many stars at once your synthetic frames will be significantly more crowded than your original frame making it difficult to apply the artificial star conclusions to your program star results 53 XXI LIST If at your installation the standard disk format for digital images to be reduced with DAOPHOT is the Caltech data structure file as it is on the VMS machines at the DAO then the LIST command will enable you to examine the contents of the image header For instance let us suppose that your images originated as FITS files from Kitt Peak or Cerro Tololo and that you want to learn the right ascension declination sidereal time and integration time of your image It happens that all this information is contained in the OBS substructure of the Caltech data structure so COMPUTER TYPES YOU ENTER Command LI File DST Components OBS Z LIST gt OBS RA 9 23 41 right ascension LIST gt OBS DEC 77 4 48 declination LIST gt OBS ST 12 53 46 sideral time LIST gt O
86. t finds such a file then it will read in the parameter specifications according to the format described below When you run DAOPHOT II the file DAOPHOT OPT acts pretty much the way LOGIN COM does when you log onto a VMS VAX Note that DAOPHOT II will only look for DAOPHOT OPT 16 in the directory from which you are currently running the program so you should have a copy of DAOPHOT OPT in each subdirectory where you are likely to work If you have one subdirectory for every set of matched frames you are working on it is easy to set up optimum parameter values for each set and have them invoked automatically whenever you run DAOPHOT from that set s directory At the very least you should have a DAOPHOT OPT file specifying the READOUT NOISE GAIN FWHM FITTING RADIUS and HIGH GOOD DATA VALUE in every subdirectory where you keep images for DAOPHOT II 1b If no file named DAOPHOT OPT is found then default values for the parameters as defined in the table in the Appendix will be supplied by the program 2a Whenever you have the Command prompt you may use the OPTIONS command The program will type on the terminal the current values for all user definable parameters and then ask you for an input filename You may then enter the name of a file containing values same format as used in a DAOPHOT OPT file you want to specify for the parameters 2b When the OPTIONS command asks you for a filename you may simply type a carriage retur
87. t for images intended for processing with DAOPHOT is the Caltech data structure DST file and routines exist for copying data from a number of different formats including FITS to this type of file On our Unix machines we use IRAF for image manipulation and IRAF image files imh and pix for image storage At ESO there exists a version of DAOPHOT II that operates on Midas format image files on a variety of hardwares If you don t happen to be at the DAO then you will have to check with your local curator of DAOPHOT II to learn how to put your data into the proper format In all that follows I shall assume that you are either at DAO or are using an unadulterated DAO VMS Unix version of DAOPHOT II See your local curator for changes that are specific to your facility I will now talk you quickly through the separate steps in reducing a typical data frame from beginning to end I suggest that you read quickly through this section and the following chapters on the major and minor routines in DAOPHOT II and then come back and reread this section more carefully Words written in boldface CAPITALS will be DAOPHOT II commands which you may issue in response to a Command prompt I From a system prompt run DAOPHOT II Either read or don t read the latest news if it s even offered to you The values of certain user definable parameters will be typed out Check their values You might want to change some of them II Use OPTIONS to chang
88. t of this number presumably represents the noise in the stellar profiles but most of it is due to the fact that the analytic function does not accurately reflect the true stellar profile of the frame It is this systematic difference between the true profile and the analytic first approximation that is to go into the look up table of profile corrections The actual derived parameters of the best fitting analytic function are typed out next In this case the analytic function chosen had three free parameters For all the different analytic first approximations the first two parameters are always the half width at half maximum in x and y Any other parameters the model may have differ from function to function but the first two are as I say always the half width at half maximum in x and y Next the routine types out the individual values for the root mean square residual of the actual stellar profile from the best fitting analytic model star by star Again these values are computed only from those pixels lying within a FITTING RADIUS of the stars centroids cosmic rays or companion stars in the PSF stars outer wings do not contribute to these numbers Any star with an individual profile scatter greater than three times the average is flagged with a a star showing 32 scatter greater than twice the average is flagged with The user may want to consider deleting these stars from the LST file and making another PSF witho
89. t to be not terribly good for significantly undersampled groundbased data and it has turned out to be particularly poor for images from the Hubble Space Telescope To help meet these new requirements DAOPHOT II offers a wider range of choices in modelling the point spread function In particular different numerical models besides the bivariate Gaussian function may be selected for the analytic first approximation These are selected by the ANALYTIC MODEL PSF option as I will explain later But to back up a bit to aid you in your choice of PSF stars I have provided a very simple routine called PICK which does some fairly obvious things First it asks for an input file containing a list of positions and magnitudes by no coincidence at all the aperture photometry file you just created with PHOTOMETRY is ideal COMPUTER TYPES YOU ENTER l Command PICK l Input file name default AP lt CR gt or filename l Desired number of stars faintest magnitude lt some numbers gt Output file name default LST lt CR gt or filename lt Some number of gt suitable candidates were found l You tell the thing how many PSF stars you d like to have and the limiting instrumental magnitude for the faintest star you d like to consider using for the PSF It then sorts the input star list by apparent magnitude note that if you have set the HIGH GOOD DATUM parameter appropriately any saturated stars should have
90. test is still performed 2 Please don t try to set the threshold low enough to pick up every last one sigma detection this will cause you nothing but grief later on The CPU time for NSTAR goes as something like the fourth power of the surface density of detected objects in your frame ALLSTAR isn t quite so bad so including a large number of spurious detections greatly increases the reduction time Even worse as the iterative profile fits progress if these fictitious stars have no real brightness enhancements to anchor themselves to they can migrate around the frame causing real stars to be fit twice in PEAK and NSTAR probably not in ALLSTAR or fitting themselves to noise peaks in the profiles of brighter stars all three routines This will add scatter to your photometry Try to set a reasonable threshold If you want you can experiment by repeatedly 23 replying to Are you happy with this with N and on a piece of graph paper build yourself up a curve showing number of detected objects as a function of threshold This curve will probably have an elbow in it with the number of detected objects taking off as the threshold continues to be lowered The best threshold value will be somewhere near this elbow This method of experimenting with different thresholds is much faster than just running FIND many times because by answering N and giving a new threshold you avoid the need to recompute the convolution
91. the inappropriateness of the flat field frame which you used to calibrate your program images not just the photon statistics but also any fine structure on a scale smaller than a seeing disk in the mismatch between the illumination of your flat field frame and your program images PROFILE ERROR In fitting the point spread function to actual stellar images there will also be some error due to the fact that the point spread function is not known to infinite precision not only will there be interpolation errors due to the finite sampling but the point spread function may vary in a greater than quadratic fashion with position in the frame or with apparent magnitude or with color or with something else This parameter defines the amplitude of this further contribution to the noise model the PROFILE ERROR increases linearly with the intensity of the star alone no sky and inversely as the fourth power of the full width at half maximum Therefore this error grows in importance relative to the PERCENT ERROR as the seeing improves since interpolation becomes harder as the data become more undersampled Leave parameters 18 and 19 alone until much much later There are four ways in which the user can supply values for these parameters to the program la Whenever you run DAOPHOT II the first thing the program will do is look in your current default directory for a file named DAOPHOT OPT daophot opt on Unix machines Tf i
92. the list of PSF stars and you can tell SUBSTAR to subtract all the stars from the image except the ones listed here In either case the stars which you have chosen will now be completely alone and uncrowded in this new picture measure them with the aperture PHOTOMETRY routine using apertures with a range of sizes up to very large These data will serve to establish the absolute photometric zero point of your image This has been the reduction procedure for a program field assumed to contain some hundreds or thousands of stars most of which you are potentially interested in The reduction procedure for 9 a standard star frame which may contain only some tens of objects only a few of which you are interested in may be different It may be that you will want to run FIND on these frames and later on match up the stars found with the ones you want Or perhaps you would rather examine these frames on the image display system use the cursor to measure the coordinates of the stars you are interested in and create your own coordinate file for input into PHOTOMETRY step VI In any case for your standard fields it is possible that you won t bother with profile fits but will just use the aperture photometry employing a growth curve analysis to define the stars instrumental magnitudes 10 C Descriptions of Main Routines In approximate order of use I DAOPHOT II itself When you run DAOPHOT II the first thing that may occu
93. then it is possible to tell ALLSTAR not to attempt to adjust the stellar positions you give it but just to solve for the magnitudes If you are very very careful this can produce more accurate photometry for your stars than otherwise But please be careful Remember that if you have frames taken at different airmasses or in different photometric bandpasses then stars positions in the various frames will not be related by simple zero point offsets There may also be slight rotations scale changes compression along the direction toward the zenith and shifts dependent upon the stars intrinsic colors due to atmospheric dispersion This then is an option to use only if you are sure your predicted positions correctly include these effects Unlike NSTAR ALLSTAR II is now empowered to redetermine sky values for the stars in the frame This is done as follows the user sets the optional parameters OS Outer sky radius gt 0 and IS Inner sky radius lt OS These need not and probably shouldn t be the same as the sky radii that you used in the PHOT routine in DAOPHOT Here OS should be large compared to the FITTING RADIUS if possible comparable to or somewhat larger than the PSF RADIUS but definitely smaller than the spatial scales of any significant variations in the sky background The inner sky radius IS can be quite small comparable to or less than the FITTING RADIUS or even zero It is probably best t
94. timate of the typical sky brightness for the frame COMPUTER TYPES YOU ENTER Command SK Approximate sky value for this frame 156 8 Standard deviation of sky brightness 4 16 Clipped mean and median 157 9 157 5 Number of pixels used after clip 8673 The sky value returned is an estimate of the mode of the intensity values in somewhat fewer than 10 000 pixels scattered uniformly throughout the frame That is it picks 10 000 pixels clips the low and high tails after the fashion of Kitt Peak s Mountain Photometry code and computes the mean and median from what is left The mode is taken as three times the median minus twice the mean The standard deviation is the one sigma width of the peak of the sky brightness histogram about the mean sky brightness not the mode or the median after clipping for all but horrendously crowded frames this distinction is negligible for our present purposes If you don t want to run the SKY command FIND will do it for you anyhow 19 V FIND You are now ready to find stars COMPUTER TYPES YOU ENTER Command FI Approximate sky value for this frame 156 8 4 16 Standard deviation of sky brightness Relative error 1 14 Number of frames averaged summed 5 1 File for the positions default COQ lt CR gt or filename ext The Number of frames averaged summed question is in case the frame you are about to reduce represents the average
95. to be run through GROUP once again to sort the combined starlist into the best groupings for the next pass of NSTAR Watch the table of group sizes that gets created on your terminal very carefully The multiple PSF fitting routine is at present capable of fitting no more than 60 stars at a time If any of the groups created by GROUP is larger than 60 stars the SELECT command can be used to pick out only those groups within a certain range of sizes You would run 8 XV XVI 1 SELECT once to pick out those groups containing from 1 to 60 stars putting them in their own file You could discard all groups larger than 60 if you only wanted a representative as distinguished from a complete sample Alternatively you could run 2 SELECT again to pick out those groups containing 61 or more stars putting them in their own file Then you would run 3 GROUP with a larger critical overlap on the file created in 2 to produce a new group file with smaller groups The photometry for these stars will be poorer than the photometry for the less crowded stars picked out in XIV 1 Return to 1 and repeat until a all stars are in groups containing less than or equal to 60 stars or b preferred and cheaper enough stars are in groups smaller than 60 that you feel you can perform your basic astronomical mission Then 4 NSTAR as many times as necessary to reduce the group files just created and 5 SUBSTAR as many times as necessary to su
96. to regard DAOPHOT II or any other program of which I am aware as the final 5 solution to the problem of accurate stellar photometry in crowded fields As it stands now the aperture photometry routine is the only place in DAOPHOT II proper where sky brightness values are estimated these estimates are based on the modal values observed in annuli around the stars and they are carried along for use by PEAK and NSTAR ALLSTAR II has the capability as an option of iteratively re determining the sky brightness value for each star defined as the median value found in pixels surrounding the star after all known stars have been subtracted from the frame using the current provisional estimates of their position and brightness These sky brightnesses are assumed to be adequate for the purposes of defining and fitting point spread functions but of course this is only approximately true The extent to which this false assumption affects the observational results and the astrophysical conclusions which you derive from your frames can best be estimated at present by the addition of artificial stars of known properties to your pictures with subsequent identification and reduction by procedures identical to those used for the program stars B A Typical Run of DAOPHOT II Before you run DAOPHOT II The Next Generation you must arrange for your pictures to exist on the computer s disk in a format acceptable to the program On the DAO VAXen the standard forma
97. tures These do not need to increase or decrease in any particular order except that only the magnitude in the first aperture will appear on the screen of your terminal as the reductions proceed and the magnitude in the first aperture will be used to define the zero point and to provide starting guesses for the profile fitting photometry Photometric data for all apertures will appear in the disk data file created by this routine The first zero or negative number appearing in the list of aperture radii terminates the list Thus the tables above both specify that photometry through seven apertures is to be obtained and that the first aperture is to be 3 pixels in radius in the first instance 2 5 pixels in radius in the second Items IS and OS are the inner and outer radii respectively of a sky annulus centered on the position of each star Tf you plan to use DAOGROW the aperture radii must increase monotonically 26 Now you can proceed to do the photometry COMPUTER TYPES YOU ENTER l l File with the positions default 7 C00 lt CR gt or filename File for the magnitudes default 7 AP lt CR gt or filename If you are monitoring the progress of the reductions on your terminal the computer will start spitting out the star ID s and coordinates from the star finding results along with the instrumental magnitudes from the first aperture and the sky brightness values for all the stars When all the stars ha
98. ure photons per ADU 1 Low good datum standard deviations 1 High good datum ADU 1 FWHM of objects for which FIND is to be optimized in pixels Significance threshold for detection standard deviations Low sharpness cutoff High sharpness cutoff Low roundness cutoff High roundness cutoff for the profile fits Watch progress of reductions on terminal The fitting radius in pixels PSF radius radius in pixels within which the point spread function is to be defined 2 Degree of variation in the PSF 2 NOT IMPLEMENTED Which analytic formula for PSF 2 How many passes to clean discordant pixels from the PSF table s Percent error e g flat field Profile error inadequate PSF FIND FIND FIND FIND FIND FIND FIND FIND PHOT PEAK PSF NSTAR SUBSTAR SORT PSF PEAK GROUP NSTAR PSF PSF PSF PSF PEAK NSTAR PEAK NSTAR positive positive non negative non negative 0 2 15 0 non negative 0 0 1 0 0 6 2 0 2 0 0 0 0 0 2 0 2S 2 1 0 10 0 1 0 35 0 al 2 1 6 O 9 O 100 O 100 32766 5 2 5 BPrerRO oO OON Notes 1 FIND is the only routine where these values are read from the options table However the file headers carry these numbers along to other routines which will also be affected 2 PSF is the only routine where these values are read from the options table However the PSF file will carry these numb
99. ut them Apart from the two objects flagged the scatter values given in the sample above demonstrate that systematic differences between the true stellar profile and the analytic first approximation dominate over raw noise in the profiles the typical root mean square residual does not increase very much from the beginning of the list to the end I happen to know that these stars were sorted by increasing apparent magnitude in the LST file Finally PSF reminds you that it has created a file named NEI which contains the PSF stars and their recognized neighbors in the frame This file may be run through GROUP and NSTAR or through ALLSTAR for a quickie profile fit and then the neighbors may be selectively subtracted from the original image to help isolate the PSF stars for an improved second generation PSF Unlike DAOPHOT Classic DAOPHOT II The Next Generation is able to use PSF stars that are within a PSF RADIUS of the edge of the frame provided that they are at least a FITTING RADIUS from the edge Finally the analytic first approximations At this particular point in time you like that verbose stuff there are six allowed options 1 A Gaussian function having two free parameters half width at half maximum in x and y The Gaussian function may be elliptical but the axes are aligned with the x and y directions in the image This restriction allows for fast computation since the two dimensional integral of the bivariate Gauss
100. vant sky brightness is defined by the answer to the question If the particular star we are interested in were not in the center of this aperture what intensity would we measure there from all other sources of brightness on or near this line of sight If there is available a set of pixels whose brightness values are uncontaminated by the star we are trying to measure but which are subject to all other sources of emission characteristic of that portion of the frame including the possibility of a contribution from individual bright objects nearby then we may answer the question Most probably the intensity would be such and such The specific value such and such is well predicted by the modal value of the brightnesses in the sample of sky pixels This is why DAOPHOT II uses the mode of the intensities within the sky annulus to define the value that should be subtracted from the intensities inside the star aperture not because it is a robust estimator of the local diffuse sky brightness but because it is a sort of maximum likelihood estimator it yields the most probable value of the brightness of a randomly chosen pixel in this region of the picture In the case of photometry from multiple simultaneous profile fits on the other hand the sky brightness is defined by the answer to a different question altogether If none of the stars included in my star list were in this region of the picture what would the typical brig
101. ve been reduced a line like the following will appear Estimated magnitude limit Aperture 1 nn nn n n per star This is simply a very crude estimate of the apparent instrumental magnitude of a star whose brightness enhancement was exactly equal to the threshold you specified for the star finding algorithm It is intended for your general information only it is not used elsewhere in DAOPHOT II and it has meaning only if all the stars were found by the FIND routine If you have entered some of the coordinates by hand or if you are redoing aperture photometry from data files that have already undergone later stages of reduction the magnitude limit is not meaningfully defined Other things you should know about PHOTOMETRY 1 The maximum outer radius for the sky annulus depends on how much virtual memory your system manager will let the program use and on how big you set the inner radius If you set too large a value the program will complain and tell you the largest radius you can get away with 2 If any of the following conditions is true PHOTOMETRY will assign an apparent magnitude of 99 999 9 999 to your star a if the star aperture extends outside the limits of the picture b if there is a bad pixel either above the High good datum or below the Low good datum inside the star aperture c if the star sky is fainter than the sky or d if for some reason the program couldn t determine a rea
102. y one These images are made up of standard alphanumeric characters so it should work whether you are on a graphics terminal or not After showing you each picture the routine will ask you whether you want this star included in the average PSF If you answer y or Y the star will be included if you answer n or N it won t If the PSF routine discovers that a PSF star has a bad pixel defined as being below the low good data value or above the HIGH GOOD DATUM value inside one FITTING RADIUS of the star s centroid it will refuse to use the star If it discovers that the star has a bad pixel outside one FITTING RADIUS but inside a radius equal to one PSF RADIUS plus two pixels what it does again depends on the WATCH PROGRESS option If WATCH PROGRESS 0 1 or 2 PSF will inform you of the existence of the bad pixels and ask whether you want to use that star anyway If you answer y or Y you are counting on the bad pixel rejection scheme later on in PSF to eliminate the flaw while extracting whatever information it can from the remaining uncontaminated part of the star s profile In my experience this has always worked acceptably well If you have set WATCH 30 PROGRESS to 2 or 1 PSF will type out a message about the bad pixel s but will go ahead and use the star anyway without prompting for a response After PSF has read your input list of PSF stars 31 Chi Parameters 0 0282 0 71847 0

User's Manual for DAOPHOT II This manual is intended as a guide

Contents

Download Pdf Manuals

Related Search

Related Contents