Before starting the daily high-cadence observations with a frequency of
usually 1 frame per minute, a dark frame and several flat field frames
are obtained. From these frames a flat field image, a quiet Sun image, and
a contrast enhanced H-alpha image are computed - see the sections
below and the paper by Denker et al. (1999).
The typical exposure time for the H-alpha full-disk images obtained at BBSO is 30 ms. The exposure time of the calibration frames and full-disk images is usually the same during the daily observing run.
Once a day - usually early in the morning - a dark current frame is
taken and a flat field frame is computed from one centered solar image
and 7 to 8 solar images shifted with respect to the center of the camera chip
by applying the method of Kuhn, Lin and Loranz (1991).
The shifted frames are obtained by off-pointing the telescope from Sun center.
The basic idea of the Kuhn-Lin-Loranz technique for calibrating spatially non-uniform image array detectors is to compute the horizontal and vertical derivatives of the logarithm of the gain function from multiple displaced full-disk images. Once the gradient of the log-gain function is determined, integration of this function leads to the logarithm of the gain within an additive constant. Finally, an iterative least-square algorithm is used to solve the pixel gain function for various displacement vectors.
Computing the gain table from the full-sized flat-field frames would take a lot of computer time. Therefore, we reduce the size of the centered and shifted calibration frames, which are corrected for dark current, by a factor of 8 before computing the flat field image. The resulting miniature gain table is subsequently enlarged to original frame size.
In the centered image of the Sun the radius and coordinates of the
solar disk are determined. This is done by using the Sobel function
to determine the limb pixels and a subsequent iterative least-square
circle fit. In this way we can eliminate the influence of active regions
(filaments, plages and spots on the disk, prominences at the limb) on the resulting
radius and center values. In order to calculate the gain table
these values have also to be determined in the shifted flat field frames
Using a 200 x 200 arcsec^2 area at disk center, the median disk center brighness
is derived. Then a bilinear interpolation algorithm is applied to transform
the centered H-alpha image from cartesian coordinates to polar coordinates
resulting in one radial disk profile per 1 degree in azimuth. Computing the
median value at each radial position yields an average radial profile. A
second order least-square polynomial fit to the average disk profile is
used to replace the inner 50% of the average disk profile. The average
disk profile is further smoothed by means of a floating average using wider smoothing
kernels at the flatter part of the profile at disk center. Finally, the smoothed
average disk profile is linearly interpolated onto a two-dimensional cartesian
grid, thus yielding the two-dimensional center-to-limb variation of the
Contrast-enhanced full-disk H-alpha images are obtained by
subtracting the quiet Sun center-to-limb variation from the
dark current and flat field corrected observations.