TIME_REQUIREMENTS · BUGS · SEE_ALSO
nstar -- fit the PSF to groups of stars simultaneously
nstar image groupfile psfimage nstarfile rejfile
- The list of images containing the stellar groups to be fit.
- The list of input group photometry files containing the group membership information and initial estimates for the positions and magnitudes of the stars to be measured. There must be one group file for every input image. If groupfile is "default", "dir$default", or a directory specification then NSTAR will look for a file with the name image.grp.? where ? is the highest existing version number. Groupfile is usually the output of the DAOPHOT GROUP task, but may also be the output of the NSTAR and PSF tasks. Groupfile may be an APPHOT/DAOPHOT text database or an STSDAS binary table.
- The list of images containing the PSF models computed by the DAOPHOT PSF task. The number of PSF images must be equal to the number of input images. If psfimage is "default", "dir$default", or a directory specification, then PEAK will look for an image with the name image.psf.?, where ? is the highest existing version number.
- The list of output photometry files. There must be one output photometry file for every input image. If nstarfile is "default", "dir$default", or a directory specification, then NSTAR will write an output file with the name image.nst.? where ? is the next available version number. Nstarfile is a text database if the DAOPHOT package parameter text is "yes", an STSDAS table database if it is "no".
- The list of output rejected photometry files containing the positions and sky values of stars that could not be fit. If rejfile is undefined, results for all the stars in photfile are written to nstarfile , otherwise only the stars which were successfully fit are written to nstarfile and the remainder are written to rejfile. If rejfile is "default", "dir$default", or a directory specification NSTAR writes an output file with the name image.nst.? where ? is the next available version number. Otherwise rejfile must specify one output photometry file for every input image. Rejfile is a text database if the DAOPHOT package parameter text is "yes", an STSDAS binary table database if it is "no".
- datapars = ""
- The name of the file containing the data dependent parameters. The parameters scale , datamin , and datamax are located here. If datapars is undefined then the default parameter set in uparm directory is used.
- daopars = ""
- The name of the file containing the daophot fitting parameters. The parameters psfrad and fitrad are located here. If daopars is undefined then the default parameter set in uparm directory is used.
- wcsin = ")_.wcsin", wcsout = ")_.wcsout", wcspsf = ")_.wcspsf"
- The coordinate system of the input coordinates read from groupfile
, of the
psf model psfimage
, and of the output coordinates written to
respectively. The image header coordinate
system is used to transform from the input coordinate system to the "logical"
pixel coordinate system used internally, from the internal logicial system to
the PSF model system, and from the internal "logical" pixel coordinate system
to the output coordinate system. The input coordinate system options are
"logical", tv", "physical", and "world". The PSF model and output coordinate
system options are "logical", "tv", and "physical". The image cursor coordinate
system is assumed to be the "tv" system.
- Logical coordinates are pixel coordinates relative to the current image. The logical coordinate system is the coordinate system used by the image input/output routines to access the image data on disk. In the logical coordinate system the coordinates of the first pixel of a 2D image, e.g. dev$ypix and a 2D image section, e.g. dev$ypix[200:300,200:300] are always (1,1).
- Tv coordinates are the pixel coordinates used by the display servers. Tv coordinates include the effects of any input image section, but do not include the effects of previous linear transformations. If the input image name does not include an image section, then tv coordinates are identical to logical coordinates. If the input image name does include a section, and the input image has not been linearly transformed or copied from a parent image, tv coordinates are identical to physical coordinates. In the tv coordinate system the coordinates of the first pixel of a 2D image, e.g. dev$ypix and a 2D image section, e.g. dev$ypix[200:300,200:300] are (1,1) and (200,200) respectively.
- Physical coordinates are pixel coordinates invariant with respect to linear transformations of the physical image data. For example, if the current image was created by extracting a section of another image, the physical coordinates of an object in the current image will be equal to the physical coordinates of the same object in the parent image, although the logical coordinates will be different. In the physical coordinate system the coordinates of the first pixel of a 2D image, e.g. dev$ypix and a 2D image section, e.g. dev$ypix[200:300,200:300] are (1,1) and (200,200) respectively.
- World coordinates are image coordinates in any units which are invariant with respect to linear transformations of the physical image data. For example, the ra and dec of an object will always be the same no matter how the image is linearly transformed. The units of input world coordinates must be the same as those expected by the image header wcs, e. g. degrees and degrees for celestial coordinate systems.
- cache = ")_.cache"
- Cache the image pixels in memory. Cache may be set to the value of the apphot package parameter (the default), "yes", or "no". By default cacheing is disabled.
- verify = ")_.verify"
- Verify the critical NSTAR task parameters? Verify can be set to the DAOPHOT package parameter value (the default), "yes", or "no".
- update = ")_.update"
- Update the NSTAR task parameters if verify is "yes"? Update can be set to the default daophot package parameter value, "yes", or "no".
- verbose = ")_.verbose"
- Print messages about the progress of the task ? Verbose can be set to the DAOPHOT package parameter value (the default), "yes", or "no".
NSTAR computes x and y centers and magnitudes for all the stellar groups in groupfile by fitting the PSF psfimage to the data in image . NSTAR reads the group membership information along with initial estimates of the centers and magnitudes, and the sky values from the photometry file groupfile . Groupfile is usually the output of the DAOPHOT GROUP task but may also be the output of the PSF and NSTAR tasks. The computed centers and magnitudes are written to nstarfile along with the sky values, the number of iterations it took to fit the star, the goodness of fit statistic chi and the image sharpness statistic sharp. If rejfile is undefine, only stars that are successfully fit are written to nstarfile , and the remainder are written to rejfile . Otherwise all the stars are written to nstarfile . Nstarfile and rejfile are text databases if the DAOPHOT package parameter text is "yes", an STSDAS table database if it is "no".
The coordinates read from groupfile are assumed to be in coordinate system defined by wcsin . The options are "logical", "tv", "physical", and "world" and the transformation from the input coordinate system to the internal "logical" system is defined by the image coordinate system. The simplest default is the "logical" pixel system. Users working on with image sections but importing pixel coordinate lists generated from the parent image must use use the "tv" or "physical" input coordinate systems.
The coordinate system of the PSF model is the coordinate system defined by the wcspsf parameter. Normally the PSF model was derived from the input image and this parameter default to "logical". However if the PSF model was derived from a larger image which is a "parent" of the input image, then wcspsf should be set to "tv" or "physical" depending on the circumstances.
The coordinates written to nstarfile and rejfile are in the coordinate system defined by wcsout with the exception of the psf model center coordinates PSFX and PSFY which are always in the logical system of the input image. The options are "logical", "tv", and "physical". The simplest default is the "logical" system. Users wishing to correlate the output coordinates of objects measured in image sections or mosaic pieces with coordinates in the parent image must use the "tv" or "physical" coordinate systems.
If cache is yes and the host machine physical memory and working set size are large enough, the input image pixels are cached in memory. If cacheing is enabled and NSTAR is run interactively the first measurement will appear to take a long time as the entire image must be read in before the measurement is actually made. All subsequent measurements will be very fast because NSTAR is accessing memory not disk. The point of cacheing is to speed up random image access by making the internal image i/o buffers the same size as the image itself. However if the input object lists are sorted close to row order and sparse cacheing may actually worsen not improve the execution time. Also at present there is no point in enabling cacheing for images that are less than or equal to 524288 bytes, i.e. the size of the test image dev$ypix, as the default image i/o buffer is exactly that size. However if the size of dev$ypix is doubled by converting it to a real image with the chpixtype task then the effect of cacheing in interactive is can be quite noticeable if measurements of objects in the top and bottom halfs of the image are alternated.
By default NSTAR computes new centers for all the stars in groupfile . However if the DAOPARS parameter recenter is "no", NSTAR assumes that the x and y centers in groupfile are the true centers and does not refit them. This option can be quite useful in cases where accurate center values have been derived from an image that has been through some non-linear image restoration algorithm, but the photometry must be derived from the original unrestored image.
By default NSTAR computes the sky value for each group by averaging the individual sky values in groupfile for all the stars in the group. If groupsky is "no" then the sky value for a particular pixel which contributes to the group fit is set to the mean of the sky values of only those stars for which the pixel is within one fitting radius. However if the DAOPARS parameter fitksy is "yes", then NSTAR computes a new group sky value as part of the non-linear least-squares fit. Recomputing the sky can significantly reduce the scatter in the magnitudes in regions where the sky background is varying rapidly, but users may need to increase fitrad to include more sky pixels in the fit. Users should experiment cautiously with this option.
Only pixels within the good data range delimited by the DATAPARS task parameters datamin and datamax are included in the fit. Most users set datamin and datamax so as to exclude pixels outside the linearity regime of the detector. By default all the data is fit. Users are advised to determine accurate values for these parameters and set the appropriate parameters in DATAPARS before beginning any DAOPHOT reductions.
Only pixels within the fitting radius fitrad / scale are included in the fit for each star. Fitrad is located in the DAOPARS task and scale is located in the DATAPARS task. Since the non-linear least-squares fitting algorithm determines three unknowns, the x and y position of the star's centroid and its brightness, the value of fitrad must be sufficiently large to include at least three pixels in the fit for each star. To accelerate the convergence of the non-linear least-squares fitting algorithm pixels within fitrad are assigned weights which are inversely proportional to the radial distance of the pixel from the x and y centroid of the star, falling from a maximum at the centroid to zero at the fitting radius. Fitrad must be sufficiently large to include at least three pixels with non-zero weights in the fit for each star. Values of fitrad close to the full-width at half-maxima of the PSF are recommended. In actual fact NSTAR imposes a minimum number of pixel limit of four.
NSTAR performs a weighted fit to the PSF. The weight of each pixel is computed by combining, the radial weighting function described above, with weights derived from the random errors NSTAR predicts based on the values of the DATAPARS parameters readnoise and epadu , and the flat-fielding and profile interpolation errors specified by the DAOPARS flaterr and proferr parameters. To obtain optimal fits, users are strongly advised to determine those parameters accurately and to enter their values in DATAPARS and DAOPARS before beginning any DAOPHOT reductions.
For each group of stars to be fit, NSTAR extracts a subraster from image which extends approximately psfrad / scale + 1 pixels wide past the limiting values of the x and y coordinates of the stars in the group. Psfrad is the PSF radius specified in the DAOPARS task, and scale is the image scale specified by the DATAPARS task. Psfrad may be less than or equal to but can never exceed the value of the image header parameter "PSFRAD" in psfimage . Psfrad should always be several pixels larger than fitrad to permit the x and y centroids to wander during the fitting process.
As well as the computed x and y centers and magnitudes, NSTAR outputs the number of times the PSF fit had to be iterated before reaching convergence. The minimum number of iterations is four. The maximum number of iteration permitted is specified by the maxiter parameter in the DAOPARS task. Obviously the results for stars which have reached the maximum iteration count should be viewed with suspicion. However since the convergence criteria are quite strict, (the computed magnitude must change by less than .0005 magnitudes or 0.10 sigma whichever is larger, and the x and y centroids must change by less than 0.002 pixels from one iteration to the next), even these stars may be reasonably well measured. It must be emphasized that every star in the group must individually satisfy the convergence criteria in order for the group to be considered adequately reduced.
NSTAR computes a goodness of fit statistic chi which is essentially the ratio of the observed pixel-to-pixel scatter in the fitting residuals to the expected scatter. Since the expected scatter is dependent on the DATAPARS task parameters readnoise and epadu , and the DAOPARS parameters flaterr and proferr it is important for these values to be set correctly. A plot of chi versus magnitude should scatter around unity with little or no trend in chi with magnitude, except at the bright end where saturation effects may be present.
Finally NSTAR computes the statistic sharp which estimates the intrinsic angular size of the measured object outside the atmosphere. Sharp is roughly defined as the difference between the square of the width of the object and the square of the width of PSF. Sharp has values close to zero for single stars, large positive values for blended doubles and partially resolved galaxies and large negative values for cosmic rays and blemishes.
NSTAR implements a highly sophisticated star rejection algorithm. First of all, any group of stars which is more than a certain size is simply not fit. The maximum group size is specified by the maxgroup parameter in the DAOPARS task. Larger groups may run into numerical precision problems during the fits. Users should exercise care in increasing the maxgroup parameter. If two stars in a group have centroids separated by a critical distance, currently set arbitrarily to 0.37 * the FWHM of the stellar core, their photocentric position and combined magnitude is assigned to the brighter of the two stars, and the fainter is eliminated. Any star which converges to 12.5 magnitudes greater than the magnitude of the PSF is considered to be non-existent and eliminated from the group.
After iteration 5, if the faintest star in the group has a brightness less than one sigma above zero, it is eliminated. After iterations 10, if the faintest star in the group has a brightness less than 1.5 sigma above zero, it is eliminated. After iterations 15 to 50 or whenever the solutions 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. After iterations 5, 10 and 15, if two stars are separated by more than 0.37 * FWHM and less than 1.0 * 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 reduction proceeds with a smaller set of stars. Backing up the counter gives the second least certain star in the group two iterations to settle into a new fit before its fate is decided. The star rejection algorithm depends upon the DATAPARS readnoise and gain parameters and the DAOPARS parameter flaterr and proferr . Therefore these parameters should be set to reasonable values before running NSTAR.
NSTAR operates in a very similar manner to PEAK. However because it fits groups of stars simultaneously it is much more accurate than PEAK in crowded regions. The ALLSTAR task also fits groups of stars simultaneously, both grouping the stars dynamically as well as producing a subtracted image. Essentially it replaces GROUP, GRPSELECT, NSTAR and SUBSTAR. However the user has little control over the grouping process and does not know at the end which stars were actually fit together. NSTAR is the task of choice when a user wants to maintain rigorous control over the composition of the stellar groups.
If verbose = yes, a single line is output to the terminal for each star fit or rejected. Full output is written to nstarfile and rejfile . At the beginning of these two files a header listing the current values of the parameters is written. For each star fit/rejected the following quantities are written to the output file.
id group xcenter ycenter mag merr msky niter sharpness chi pier perr
Id is the id number of the star and group is its group number. Xcenter and ycenter are the fitted coordinates in pixels. Mag and merr are the fitted magnitude and magnitude error respectively. Msky is the individual sky value for the star. Niter is the number of iterations it took to fit the star and sharpness and chi are the sharpness and goodness of fit statistic respectively. Pier and perror are the photometry error code and accompanying error message respectively.
If no errors occur during the fitting process then pier is 0. Non-zero values of pier flag the following error conditions.
0 # No error 1 # The star is in a group too large to fit 2 # The sky is undefined 3 # There are too few good pixels to fit the star 4 # The fit is singular 5 # The star is too faint 6 # The star has merged with a brighter star 7 # The star is off the image
1. Fit the PSF to a list stars in the test image dev$ypix. Good stars for making the PSF model can be found at (442,410), (348,189), and (379,67).
da> datapars.epadu = 14.0 da> datapars.readnoise = 75.0 ... set the gain and readout noise for the detector da> daofind dev$ypix default fwhmpsf=2.5 sigma=5.0 threshold=20.0 ... answer verify prompts ... find stars in the image ... answer will appear in ypix.coo.1 da> phot dev$ypix default default annulus=10. dannulus=5. \ apertures = 3.0 ... answer verify prompts ... do aperture photometry on the detected stars ... answer will appear in ypix.mag.1 da> display dev$ypix 1 da> psf dev$ypix default "" default default default psfrad=11.0 \ fitrad=3.0 mkstars=yes display=imdr ... verify the critical parameters ... move the image cursor to a candidate star and hit the a key, a plot of the stellar data appears ... type ? for a listing of the graphics cursor menu ... type a to accept the star, d to reject it ... move to the next candidate stars and repeat the previous steps ... type l to list all the psf stars ... type f to fit the psf ... move cursor to first psf star and type s to see residuals, repeat for all the psf stars ... type w to save the PSF model ... type q to quit, and q again to confirm ... the output will appear in ypix.psf.1.imh, ypix.pst.1 and ypix.psg.1 da> group dev$ypix default default default ... verify the prompts ... the output will appear in ypix.grp.1 da> nstar dev$ypix default default default default ... verify the prompts ... the results will appear in ypix.nst.1 and ypix.nrj.1 da> pdump ypix.nst.1 sharpness,chi yes | graph ... plot chi versus sharpness, the stars should cluster around sharpness = 0.0 and chi = 1.0, note that that the frame does not have a lot of stars da> substar dev$ypix default "" default default ... subtract the fitted stars da> display ypix.sub.1 2 ... note that the psf stars subtract reasonably well but other objects which are not stars don't
2. Run nstar on a section of the input image using the group file and PSF model derived in example 1 for the parent image and writing the results in the coordinate system of the parent image.
da> nstar dev$ypix[150:450,150:450] default default default default \ wcsin=tv wcspsf=tv wcsout=tv ... answer the verify prompts ... fit the stars ... the results will appear in ypix.nst.2 and ypix.nrj.2 da> display dev$ypix[150:450,150:450] 1 ... display the image da> pdump ypix.nst.2 xc,yc yes | tvmark 1 STDIN col=204 ... mark the stars da> substar dev$ypix ypix.nst.2 "" default default ... subtract stars from parent image ... the output images is ypix.sub.2 da> substar dev$ypix[150:450,150:450] ypix.nst.2 "" default default \ wcsin=tv wcspsf=tv wcsout=tv ... subtract stars from the nstar input image ... the output images is ypix.sub.3
3. Run nstar exactly as in example 1 but submit the task to the background. Turn off verify and verbose.
da> nstar dev$ypix default default default default verbose- \ verify- & ... the results will appear in ypix.nst.3 and ypix.nrj.3