apflatten input output
input
List of input flat field observations.
output =
List of output flat field images. If no output name is given then the
input name is used as a root with the extension ".flat".
references =
List of reference images to be used to define apertures for the input
images. When a reference image is given it supersedes apertures
previously defined for the input image. The list may be null, "", or
any number of images less than or equal to the list of input images.
There are three special words which may be used in place of an image
name. The word "last" refers to the last set of apertures written to
the database. The word "OLD" requires that an entry exist
and the word "NEW" requires that the entry not exist for each input image.
interactive = yes
Run this task interactively? If the task is not run interactively then
all user queries are suppressed and interactive aperture editing and trace
fitting are disabled.
find = yes
Find the spectra and define apertures automatically? In order for
spectra to be found automatically there must be no apertures for the
input image or reference image defined in the database.
recenter = yes
Recenter the apertures?
resize = yes
Resize the apertures?
edit = yes
Edit the apertures? The interactive parameter must also be yes.
trace = yes
Trace the apertures?
fittrace = yes
Interactively fit the traced positions by a function? The interactive
parameter must also be yes.
flatten = yes
Remove the profile shape and flat field spectrum leaving only
sensitivity variations?
fitspec = yes
Fit normalization spectrum interactively? The interactive
parameter must also be yes.
line = INDEF, nsum = 1
The dispersion line (line or column perpendicular to the dispersion
axis) and number of adjacent lines (half before and half after unless
at the end of the image) used in finding, recentering, resizing,
and editing operations. For tracing this is the starting line and
the same number of lines are summed at each tracing point. A line of
INDEF selects the middle of the image along the dispersion axis.
A positive nsum sums the lines and a negative value takes the median.
However, for tracing only sums are allowed and the absolute value
is used.
threshold = 10.
Division threshold. If a pixel in the two dimensional normalization spectrum
is less than this value then a flat field value of 1 is output.
The following parameters control the profile and spectrum fitting.
background = none
Type of background subtraction. The choices are "none" for no
background subtraction, "average" to average the background within the
background regions, or "fit" to fit across the dispersion using the
background within the background regions. Note that the "average"
option does not do any medianing or bad pixel checking; it is faster
than fitting however. Background subtraction also requires that the
background fitting parameters are properly defined. For the "average"
option only the background sample regions parameter is used.
pfit = fit1d (fit1d|fit2d)
Profile fitting algorithm to use with variance weighting or cleaning.
When determining a profile the two dimensional spectrum is divided by
an estimate of the one dimensional spectrum to form a normalized two
dimensional spectrum profile. This profile is then smoothed by fitting
one dimensional functions, "fit1d", along the lines or columns most closely
corresponding to the dispersion axis or a special two dimensional
function, "fit2d", described by Marsh (see approfile).
clean = no
Detect and replace deviant pixels?
skybox = 1
Box car smoothing length for sky background when using background
subtraction. Since the background noise is often the limiting factor
for good extraction one may box car smooth the sky to improve the
statistics in smooth background regions at the expense of distorting
the subtraction near spectral features. This is most appropriate when
the sky regions are limited due to a small slit length.
saturation = INDEF
Saturation or nonlinearity level. During variance weighted extractions
wavelength points having any pixels above this value are excluded from the
profile determination.
readnoise = 0.
Read out noise in photons. This parameter defines the minimum noise
sigma. It is defined in terms of photons (or electrons) and scales
to the data values through the gain parameter. A image header keyword
(case insensitive) may be specified to get the value from the image.
gain = 1
Detector gain or conversion factor between photons/electrons and
data values. It is specified as the number of photons per data value.
A image header keyword (case insensitive) may be specified to get the value
from the image.
lsigma = 3., usigma = 3.
Lower and upper rejection thresholds, given as a number of times the
estimated sigma of a pixel, for cleaning.
The following parameters are used to fit the normalization spectrum using
the ICFIT routine.
function = legendre
Fitting function for the normalization spectra. The choices are "legendre"
polynomial, "chebyshev" polynomial, linear spline ("spline1"), and
cubic spline ("spline3").
order = 1
Number of polynomial terms or number of spline pieces for the fitting function.
sample = *
Sample regions for fitting points. Intervals are separated by "," and an
interval may be one point or a range separated by ":".
naverage = 1
Number of points within a sample interval to be subaveraged or submedianed to
form fitting points. Positive values are for averages and negative points
for medians.
niterate = 0
Number of sigma clipping rejection iterations.
low_reject = 3. , high_reject = 3.
Lower and upper sigma clipping rejection threshold in units of sigma determined
from the RMS sigma of the data to the fit.
grow = 0.
Growing radius for rejected points (in pixels). That is, any rejected point
also rejects other points within this distance of the rejected point.
I/O parameters and the default dispersion axis are taken from the package parameters, the default aperture parameters from apdefault, automatic aperture finding parameters from apfind, recentering parameters from aprecenter, resizing parameters from apresize, parameters used for centering and editing the apertures from apedit, and tracing parameters from aptrace.
It is sometimes the case that it is undesirable to simply divide two dimensional format spectra taken through fibers, aperture masks with small apertures such as holes and slitlets, or small slits in echelle formats by a flat field observation of a lamp. This is due to the sharp dropoff of the flat field and object profiles and absence of signal outside of the profile. Slight shifts or changes in profile shape introduce bad edge effects, unsightly "grass" is produced where there is no signal (which may also confuse extraction programs), and the division will also remove the characteristic profile of the object which might be needed for tracking the statistical significance, variance weighted extraction, and more. A straight flat field division also has the problem of changing the shape of the spectrum in wavelength, again compromising the poisson statistics and artificially boosting low signal regions.
There are three approaches to consider. First, the flat field correction can be done after extraction to one dimension. This is valid provided the flat field and object profiles don't shift much. However, for extractions that depend on a smooth profile, such as the variance weighting algorithms of this package, the sensitivity corrections must remain small; i.e. no large fringes or other small scale variations that greatly perturb the true photon profile. The second approach is to divide out the overall spectral shape of the flat field spectrum, fill regions outside of the signal with one and leave the profile shape intact. This will still cause profile division problems described earlier but is mentioned here since it implemented in a related task called apnormalize. The last approach is to model both the profile and overall spectrum shape and remove it from the flat field leaving only the sensitivity variations. This is what the task apflatten does.
The two dimensional flat field spectra within the defined apertures of the input images are fit by a model having the profile of the data and a smooth spectral shape. This model is then divided into the flat field image within the aperture, replacing points of low signal, set with the threshold parameter, within the aperture and all points outside the aperture by one to produce an output sensitivity variation only flat field image.
A two dimensional normalized profile is computed by dividing the data within the aperture by the one dimensional spectrum and smoothing with low order function fits parallel to the dispersion axis if the aperture is well aligned with the axis or parallel to the traced aperture center if the trace is tilted relative to the dispersion axis. The smooth profile is then used to improve the spectrum estimate using variance weighting and to eliminate deviant or cosmic ray pixels by sigma tests. The profile algorithm is described in detail in approfiles and the variance weighted spectrum is described in apvariance.
The process of determining the profile and variance weighted spectrum, and hence the two dimensional spectrum model, is identical to that used for variance weighted extraction of the one dimensional spectra in the tasks apall or apsum and in making a two dimensional spectrum model in the task apfit. Most of the parameters in this task are the same in those tasks and so further information about them may be found in their descriptions. In fact, up to this point the task is the same as apfit and, if the flat field were normalized by this model it would produce the "ratio" output of that task.
This task deviates from apfit in that the final variance weighted one dimensional spectrum of the flat field is subjected to a smoothing operation. This is done by fitting a function to the spectrum using the icfit routine. This may be done interactively or noninteractively depending on the interactive parameter. The default fitting parameters are part of this task. The goal of the fitting is to follow the general spectral shape of the flat field light (usually a lamp) but not the small bumps and wiggles which are the one dimensional projection of sensitivity variations. When the fitted function is multiplied into the normalize profile and then the two dimensional model divided into the data the sensitivity variations not part of the fitted spectrum are what is left in the final output flat field.
The remainder of this description covers the basic steps defining the apertures to be used. These steps and parameter are much the same as in any of the other apextract tasks.
Aperture definitions may be inherited from those of other images by specifying a reference image with the references parameter. Images in the reference list are matched with those in the input list in order. If the reference image list is shorter than the number of input images, the last reference image is used for all remaining input images. Thus, a single reference image may be given for all the input images or different reference images may be given for each input image. The special reference name "last" may be used to select the last set apertures used in any of the apextract tasks.
If an aperture reference image is not specified or no apertures are found for the specified reference image, previously defined apertures for the input image are sought in the aperture database. Note that reference apertures supersede apertures for the input image. If no apertures are defined they may be created automatically, the find option, or interactively in the aperture editor, if the interactive and edit options are set.
The functions performed by the task are selected by a set of flag parameters. The functions are an automatic spectrum finding and aperture defining algorithm (see apfind) which is ignored if apertures are already defined, automatic recentering and resizing algorithms (see aprecenter and apresize), an interactive aperture editing function (see apedit), a spectrum position tracing and trace function fit (see aptrace), and the main function of this task, the flat field profile and spectral shape modeling and removal.
Each function selection will produce a query for each input spectrum if the interactive parameter is set. The queries are answered by "yes", "no", "YES", or "NO", where the upper case responses suppress the query for following images. There are other queries associated with tracing which first ask whether the operation is to be done interactively and, if yes, lead to queries for each aperture. If the interactive parameter is not set then aperture editing interactive trace fitting, and interactive spectrum shape fitting are ignored.
1. To make a two dimensional flat field from a lamp observation:
cl> apflatten fiber1 flat read=3 gain=1 back=fit Yes find No resize No edit Yes trace Yes trace interactively NO Yes flatten Yes fit interactively
SEE ALSO, apbackground, approfile, apvariance, apfit, icfit, , apdefault, apfind, aprecenter, apresize, apedit, aptrace, apsum,