Standard sequence: read and store pixels
Purpose: acquire and store readouts in a uniform way,
independent of the type of camera or its mode.
Typical course of events:
-
System reads pixels from the camera and writes them directly to the
image area of a FITS file. The readout is placed at the
start of the image area and fills it, leaving a fraction of a FITS
block which the system pads with zeros.
-
System applies those windows in the range 0..10 that are
enabled (one is always enabled). System reads from the camera only
those pixels that lie within an enabled window, and disposes of the
other pixels with the minimum possible delay. In the FITS image,
the pixels are rearranged to remove any rows or columns that lie
entirely between windows.
-
System applies binning. All windows are reduced in each dimension
by the binning factor for that dimension. Where a window's
dimensions
are not multiples of its binning factor, the window is treated
as if it had the next smallest dimensions that are exact
multiples.
Any window that has a
dimension smaller than the corresponding binning factor produces
no pixels. In the latter two cases, the image-section recorded
for the window is not updated.
-
System applies transformations. In the camera's profiles there
is a list, possible empty, of rotations and/or reflections to be
applied to the image; the system applies this list.
-
System applies pixel-value corrections. 16-bit unsigned CCD data
have 32,678 subracted to carry them in a 16-bit signed format.
IR data are converted to 32-bit floating-point numbers. The engineering
profile decides which conversion is used.
Variations:
-
The camera has more than one readout channel. Each channel is
treated separately and directed to its own file.
-
In general, the image area of the FITS file can be a plane, a cube
or a hypercube of 4, 5, 6 or 7 dimensions.
Any number of readouts may be combined to fill the image.
A single readout always goes entirely into one plane, but otherwise
there are no restrictions on how successive readouts are placed.
Any unused parts of the image are filled with zeros.
-
Instead of replacing the values in an image plane, the pixels from
a readout may be co-added to the existing numbers. This is much
more common for IR data.