- Version 2 and 3 differences
in the input file: The input file has changed between
Version 3 and Version 2. Parameters 8, 9, and 10 in Version 2
are now called 9, 10, and C0 in Version 3. C0 is now
an optional parameter, so it is only visible when specified explicitly
and given a value, just like Fourier modes, bending modes, spiral
rotation functions, and truncation models.
- Three image header keywords that
GALFIT looks for: EXPTIME,
GAIN (or ATODGAIN), and NCOMBINE.
If missing, the defaults are: EXPTIME=1, GAIN=7, NCOMBINE=1.
Note that in Version 3, RDNOISE is no longer an important parameter
unless one tells GALFIT to estimate a sigma-image by specifying a sky
value manually (see the GALFIT manual for details). RDNOISE is now
automatically estimated along with the background noise by GALFIT
because for the sigma image it is not important to differentiate
Poisson noise in the sky from detector RDNOISE.
Note: Sometimes an image may have units of [counts/sec] but
the EXPTIME keyword does not show 1 second. This is misleading, but
is unfortunately a common practice. Make sure that the EXPTIME
keyword means what GALFIT expects: if the image is in flux units,
EXPTIME = 1; if the image is in counts, EXPTIME should be the total
exposure time; if the image is averaged over N images, each of
which is T seconds, then NCOMBINE = N, and EXPTIME =
T. Or else your magnitudes will be off by 2.5
log10(EXPTIME) mags.
- When one doesn't have an
input sigma image: If an user doesn't provide a sigma
image, or if one isn't found, GALFIT will create it internally. The
sigma image is needed for estimating χ2. However, if
you're letting GALFIT do this, be aware that the ADUs must be in
COUNTS , not counts/sec.
Furthermore, GALFIT needs to know how to convert from ADUs into
electrons -- the meaningful unit for estimating Poisson noise -- via
the GAIN of the CCD, and the NCOMBINE (the number of images
averaged together) image header parameters in the image header.
If GALFIT doesn't find these info, then it defaults to GAIN=7 and
NCOMBINE=1, which are appropriate for WFPC2. Furthermore, if NCOMBINE
is not 1, then GALFIT assumes that the image ADUs are averaged, rather
than summed, over NCOMBINE number of images.
Note: for GALFIT to determine the sigma image correctly, the
sky also ought to dominate the field of view.
- The correct way to fit a
point source: The best way to fit a point source is to
specify in Item D a PSF image (this is the same one used in
convolution), then in the object list, specify ``psf'' for the object
type. The free parameters for the PSF function are simply (x,
y) and magnitude. GALFIT figures out the total magnitude by
integrating over the PSF image and using the zeropoint in the input
parameter file. Note that this method assumes the input PSF image
contains 100% of the PSF flux. If this is not true, i.e. the PSF
image cuts off the wings, then one would either need to make an
aperture correction to account for the missing flux, or make the PSF
image bigger.
This procedure for fitting a point source now differs from an earlier
suggestion that the user should use a narrow Gaussian function,
convolved with the PSF. Even though one can still do this it is not
advisable because there likely will be numerical convergence issues,
having to do with the gradient fitting within a single pixel, when the
FWHM falls below 0.4 to 0.5 pixels. This may cause some or all of the
parameters/components to be frozen. This technique, however, is still
useful to test to see if an object is truly a point source. If one
wants to do this, just make sure the FWHM does not dip below 0.4-0.5
pixels.
- How the PSF should be
centered in the PSF image: If the PSF image has
N=odd number of pixels along an axis, the centroid along that
axis should be at the geometric center, i.e. N/2+0.5. If the
PSF image has N=even number of pixels along an an axis, the
centroid should be at pixel N/2+1. If the PSF is not centered
as described, there will be a systematic offset in the measured
position from actual. The offset is the same for all fitted
components.
- Also, please see notes on
Top 10 rules of thumb for galaxy fitting and
Technical GALFIT FAQ .
)