There have been some occurences of underlying offsets,
in some image frames, showing a clear banded structure.
It seems that this is a random LORAL3 controller issue.
A display/diagnosis tool is provided here to check the
smoothness of the offset of any image, and set the offset
correction method accordingly.
This is done by column-averaging a strip situated
in the CCD overscan zone; it has been found that the strips
polluting the offset are parallel to CCD lines, and extend all
the way through the CCD including the overscan zone.
The average spectrum N = N(y) of the overscan strip is thus a
good image of the underlying offset [y] profile.
According to the aspect of this overscan strip spectrum,
the user has to choose between two offset correction methods
: Polynomial fit
for smooth shape spectrum (that is sloping, bent, but not
stepped; often present, but to a very low level),
Median filtering for a stepped spectrum (yes, it
happens; the biggest step we have ever seen was approximately
2-3 counts high, versus a 1152 counts regular offset level...);
The default setting is polynomial fit.
Use
Click on [CCD] in the main menu, then on [Check frame
overscan].
The [Check frame overscan] window pops up.
Enter the name of the frame to be checked. You can type it
in directly, or use the browse
icon at the end of the field, or drag and drop it from the
Reduction
folder.
The resulting [y] offset spectrum of the frame
is displayed.
Options description
Overscan limits :
The user has to choose here the [x] limits of the overscan
zone to be used for the spectrum evaluation,
that is the left [X1] and right [X2] limits of the
sub-frame. The full height is used,
and the default [X1,X2] limits are [2055,2080].
Offset correction method :
The user has to choose between Polynomial fit and
Median filter for the smoothing method to be used
on the offset [y] spectrum. This method will be applied
during the Build bias phase and the
Preprocess phase, later on.
Save values :
All the input values (files names, coordinates) are saved,
and become the new default values for this user. They can
be recalled at will, and are used each time the Check frame
offset job window is opened.
Recall values :
The values (files names, coordinates) saved by the user,
are loaded to the various input fields.
Default values :
The input fields are set to the general defaults values;
for instance, the file names are set to blank.
Set defaults
Principles
This function is used to set :
the default method to be used to smooth the [y] offset
spectrum of the currently preprocessed frame. Use the
[Check frame overscan] function described above to choose
between polynomial fit and median filter, or let things
as they are (polynomial fit) if you trust blindly the
CCD controller electronics; not always wise with the present
LORAL3.
the default x limits of the overscan zone used to compute
the [y] offset spectrum. The default defaults are X1=2060,
X2=2080, which is usually OK.
The present defaults will be used for all subsequent frame
preprocesses.
Use
Click on CCD in the main menu, then on Offset,
then on Set defaults. The [Set defaults] window pops
up.
Check the method you prefer (see
Check frame overscan section for details, and enter the
limits of the overscan zone to be used (see same preceding
section for details).
You must have 2049 <= X1 < X2 <= 2088.
Options description
Save values :
All the input values are saved,
and become the new default values for this user. They can
be recalled at will, and are used each time the [Set defaults]
window is opened.
Recall values :
The values saved by the user,
are loaded to the various input fields.
Default values :
The input fields are set to the general defaults values;
for instance, the method is set to Polynomial fit
and the overscan limits are set to [2060,2088].
The [y] offset spectrum is evaluated over the [X1,X2]
overscan zone strip (see section
Offset correction above).
This spectrum is smoothed using the method set in the same
preceding section.
The smoothed spectrum is subtracted from every [1...2048]
column of the raw bias frame.
The resulting frame is clipped from the [2049,2088]
overscan strip. The new frame is named
B{original bias name}.
Use
Click on [CCD] in the main menu, then on [Build bias].
In the [Build bias] window which pops up, enter the name
of the raw bias frame. It should have been already checked
using the [Offset correction] function,
subfunction [Check frame overscan], which allows the user to set
the offset correction method; if not, the default
(Polynomial fit) method is used. Regarding the name
of the bias frame, you can type it in directly, or use the browse
icon at the end of the field, or drag and drop it from the
Reduction
folder...
We encountered some data encoding problem on CFHT
(that is IRAF in fact) medians, with image values being strangely
quantified in the integer-to-floating point conversion at the end of
the median process. One way of going around this is to convert, from
within XOasis, the median bias file to some non-FITS format
(MIDAS for instance), then back
to FITS format, using the Import/Export,
File format conversion function.
In the process, the image data is correctly converted to floating point
format, and some descriptors are set to the right value. You may delete
the intermediate MIDAS file after completion. This procedure is
strongly recommended if you wish to use CFHT medians.
Options description
Debug :
No action; forget it...
Save values :
All the input values (file name, coordinates) are saved,
and become the new default values for this user. They can
be recalled at will, and are used each time the [Build bias]
window is opened.
Recall values :
The values (file name, coordinates) saved by the user,
are loaded to the various input fields.
Default values :
The input fields are set to the general defaults values;
for instance, the file name is set to blank.
Remember that to use this function without any S/N
degradation, you must use a continuum which is in fact a median of
a large number of well-exposed individual continuum frames; something
like 50 units. As far as we know, nobody has ever tried to obtain a
median of 50 continua :-( ... So this paragraph is usually
pointless; but, just to be complete :
The continuum frame, after bias subtraction, is scanned column
after column.
On each column, a spline fit is performed and the result is
recorded.
Each column is divided by the spline fit, to isolate the high
frequencies, and the result used to
build the high-frequency correction frame.
Use
Click on [CCD] in the main menu, then on [Build HF Flatfield].
In the [Flat HF] window which pops up, enter the name
of the [Bias-subtracted continuum frame], and the name of the
[HF correction frame] to be computed. You may type in the names, or
use the browse icon at the end of the field, or drag and drop them
from the
Reduction
folder.
Click on [Accept].
Options description
Smoothing parameter :
This value controls the smoothness of the spline fit. The default
is 3, which means that two consecutive nodes are separated by at
least 3 pixels.
Debug :
If you check this option, computations are done on a single column,
which the user specifies in the [Column number] input field; this
may be used to check the dependency of the results on the smoothing
parameter value.
Save values :
All the input values (file name, coordinates) are saved,
and become the new default values for this user. They can
be recalled at will, and are used each time the [Build bias]
window is opened.
Recall values :
The values (file name, coordinates) saved by the user,
are loaded to the various input fields.
Default values :
The input fields are set to the general defaults values; for instance,
the file names are set to blank.
The offset is subtracted from the input frame. For that :
The [y] offset spectrum of the object frame to be
processed is obtained from the [X1,X2] overscan strip.
This spectrum is smoothed using the method set in the
Offset correction
preceeding section, subfunction [Check frame overscan].
If this is not set, the default (Polynomial fit method
is used.
The smoothed spectrum is subtracted from every [1...2048]
column of the raw object frame.
The bias is subtracted from the offset-corrected object frame.
The bias used is the one which has been offset-corrected in the
Build bias section.
If no bias frame is available, this step is skipped;
this is not perfect, but on the examples we have at hand, the
bias value is typically 1 (one) which is quite negligible against
the offset value.
For that, enter none as the bias frame name.
The frame is clipped according to the limits given.
The frame is corrected from high-frequency flat-field response
irregularities (applicable only to image frames, not to
spectroscopic TIGER frames).
The frame is dark current-subtracted, that is the
specific dark current constant of the CCD used, scaled to the
integration time achieved, is subtracted from the frame.
This solution has been chosen because nobody can afford obtaining
real dark current frames, owing to observing time and ambient
light constraints, specially for long exposure times.
Use
Click on [CCD] in the main menu, then on [Preprocess]. The
[Preprocess] window pops up.
Enter the name(s) of the [Frame to be preprocessed], of the
[bias frame], of the high frequency [Flat frame] (only for
image frames). You can type it in directly, or use the
browse icon at the end of the field, or drag and drop
it from the
Reduction folder.
Options description
Clip & Bias selected means that the input frame will
be clipped according to the values given in the four fields
to the right of the window, at least to get rid of the
offset strip. This will be done after the bias
correction.
Flat selected means that you want your frame to be
corrected from high frequency flat field errors. This
option is applicable to image frames only. If the
frame to be processed has not yet been clipped and
de-biased, it will be before beeing flat-fielded.
You must fill in the [Flat frame] name zone if you want to
use this option.
Dark selected means that you want your frame to be
corrected from CCD dark signal. If the
frame to be processed has not yet been clipped and
de-biased, it will be before beeing dark-subtracted.
The dark is taken as a
scalar constant, as it is extremely low
(0.85 e.hour-1 for instance for the LORAL3 CCD),
to avoid introducing noise into
the reduction process; medians of several tens of
one-hour dark's would of course be better, but are clearly
unobtainable. The value used is shown in the [Dark current]
input field, and should not be changed unless you get some
very good reason to do so.
Debug :
This switches the program to verbose mode, and more
informations are recorded into the history file (see section
[Start] in the left reduction menu).
Clipping :
This gives the lower left (X1,Y1) and upper right (X2,Y2)
corners of the frame area which will be kept after clipping.
Some do
not trust the edge pixels of the CCD, and like to clip off
a few pixels on three edges, and of course the
offset strip on the fourth. Default values keep the full
2048x2048 frame, which is OK.
Save values :
All the input values (files names, coordinates) are saved,
and become the new default values for this user. They can
be recalled at will, and are used each time the [Preprocess]
window is opened.
Recall values :
The values (files names, coordinates) saved by the user,
are loaded to the various input fields.
Default values :
The input fields are set to the general defaults values;
for instance, the file names are set to blank.
Back to top of chapter.
Back to OASIS home page.