This page gives information about the fits files that are available at web address:
Each file has a fits header with kewords and valued given. The header includes an abbreviated description of the keywords. A sample fits header with an expanded explanation each keyword can be examined as a text file. This sample includes both the abbreviated description along with an expanded description. Some quantities are also given in the summary files described below.
File names are generated automatically according to the following rules:
Each file is given a name in the format PPYYYYMMDD-SS-yyyymmdd-ss.fits where:
PP -- designates the program series to which the image belongs. Two programs are planned for inclusion at this time:
K and tK indicate image in the CaK spectroheliogram series and thumbnail version of these;
I and tI will be intensity and thumbnail intensity images of the broadband-direct image sequence.
YYYYMMDD -gives the date of the observation in year month day format.
SS - is a sequence number assigned during the logbook entry process. It is generally the ordinal number of the observation within the month.
yyyymmdd - gives the date the image was scanned in year month day format. Occasionally an image will be scanned more than once and our designation then provides a distinction between the images. The plate setup typically differs between the multiple scans in such cases of duplicate scans of a single image.
ss - gives the ordinal number for the scan during the indicated day. Typically each scan consists of three or four plates with each plate having two to four solar images, not all of which are part of the program sequence.
THE SUMMARY FILES
Each year is accompanied by a summary file which gives a few parameters for each reduced fits file with 1 line of text 134 characters long for each file. The files provided in the above directory have been rebinned a factor of 3 in each coordinate in order to facilitate their transfer and use. The image coordinates in the summary files are the unreduced values and need to be divided by 3 for application to the reduced size files. The following list gives the character position range and the field these characters represent:
For summary files written after Sep. 27, 2005 (applies to all currently available summary files) the format of the last columns is:
Prior to April 17, 2006, the entries in the above table were all too large by 1 except for column 1.
THE PLATES AND SCANS
The photographic plates have been scanned in groups of three with each plate having up to 4 images. Most early plates have two pairs with each pair being exposed simultaneously. One image of the pair is at the center of the Ca K line and includes the light of Ca K2V, K3 and K2R. This is the image of interest to us. The secone image of the pair is in the light of the nearby part of the K1 line and does not usually include other spectral features. The separation between the two images is a function of the wavelength chosen for this second image and this separation is not constant, sometimes producing overlapping exposures. We do not use the data from the second images although they are included in the general scans. The scans are being carried out by Liz Werden using an Eskographics F14 scanner. The solar images have radii of about 1000 pixels and the digitization provides about 12 bits of significant data. The scanner output images are in a 16-bit tif format with the intensity values being stored as positive integers. Thus only 15-bits are available to hold the intensities. Since the scanner returns only 12-bits of significant data and noise is not added, the 3 least significant bits are not used and the data values are not uniformly distributed over the 16-bit integer values. We refer to these initial scans as fulldeck scans since all images are scanned simultaneously using the full deck of the scanner. Each full deck scan has a size of approximately 220 Mbytes. Positions on the full deck scan of the plates according to their plate number are recorded on a hand written sheet and each full deck is stored on the basis of the date of the scan.
LOGBOOK INFORMATION AND SCAN IDENTIFICATION
Data from the original log books has been entered manually by Pam Gilman and examined for validity by Larry Webster. These digitized log files are downloaded to a archive database. The solar images on the full deck scans are then tagged with corresponding digital log book entries by Javaraiah Javaraiah by visually examining each image on the scan for written marking giving the date, time and observation ID number (sigma numbers or C numbers depending on the era of the observation). This tagging step links the time and date of observation to the image position on the full deck scan. Using the tag information an extraction is done with software written specifically for this purpose by Ferenc Varadi. Plate geometry and solar image geometry are essential in this step. Each extracted image is stored as a separate 16-bit tif file and the information from the database is recorded as the figure caption of the image in a format that can be used during subsequent processing steps. This initial extraction provides the first indication of the location of the solar image since the extraction software finds a circular structure in the expected position and then picks a 2601 by 2601 pixel array with this circle of roughly 1000 pixel radius at its center.
The scanner calibration/setup sets the digitization so that the value returned where nothing is between the lamp and the detector is 32768 and when a completely opaque object is between the lamp and detector the number is 0. Since only 12 bits of information are provided, the returned values are not smoothly distributed over all possible 16-bit integers.
DUST AND PIT REMOVAL
In addition a Laplacian filter is applied to the image to locate discordant groups of points whose scale is smaller than the spatial resolution of the observing system. This filter focuses on each pixel and starts with the average of all pixels distant from the target pixel by less than or equal to two pixels. Those pixels whose value differs from this average by more than Itest digitization units are considered discordant. The data for the discordant pixels is then replaced by that of the surrounding average. This replacement is vetoed if the adjacent pixel is more discordant than is the target pixel. This process is applied to all pixels on the image further from the image edge than 2 pixels and carried out iteratively 4 times with the value of Itest taken to be 1000 digitization units (out of 32768) for the first iteration, 800 for the second and 700 for the third and fourth. Typically 4000 to 100000 pixels out of 6.7 Mpix have their values replaced in this process.
INITIAL GUESS OF IMAGE CENTER
The solar image is identified in the 2601 by 2601 pixel array by first calculating the rms deviation of square sub-arrays of 5 by 5 pixels. The rms value of these 25 pixels is assigned to the center pixel of this sub-array and this calculation is repeated for all pixels at least 2 pixels from the edge of the array. Using the average of all these rms values as a dividing point, the centroid of all pixels having larger than average rms values is taken as the center of the solar image.
Due to the nature of the scans where the dispersion direction is in the direction of the scan and the cross-dispersion direction is parallel to the slit, a strategy was adopted whereby the search in the image is for radii and image image centers associated with the x (scan/dispersion direction) and y (cross-dispersion/slit direction) independently. Four separate locations are sought: x+, x-, y+, and y-. The center point in each direction is then the average of the + and - limb positions along each of these axes. With a tentative center, four quadrants of the image correspond to each of these limb positions. Furthermore, we know from averages over many images what the expected solar image radius should be as a function of the time of year. This expectation value is used to constrain the solutions for the radii within bounds that are tightened successively during an iterative process.
INTENSITY GRADIENT DETERMINATION
Based on these image constraints, the next step in finding the solar image limb is to calculate a gradient image using the first guess image center. For each pixel on the image outside of a central square, the direction toward and away from the first guess center is known. It is also known if the pixel is nearly along a diagonal. Averages are calculated for two 5 by 7 blocks of pixels which are symmetrically centered on the target pixel. One block is in a direction away from the trial center and the other block is toward the image center. The 7 pixel side of the block is along the line toward the image center and the 5 pixel side is centered on the target pixel and is along the axis perpendicular to the direction toward the image center. The inner block average is subtracted from the outer block average to form a smoothed measure of the intensity slope at each point. The differences are calculated in each portion of the image in such a way that when the intensity drops abruptly in going away from the trial center, a positive slope will be found. These gradient images have a very clear circular signature of the image edge.
THE REFERENCE GRADIENT
Due to exposure variations and optical system vignetting, the intensity and its gradient are highly variable from one image to another. In order to use the gradient images in spite of these variations, the first step is to determine typical values of the gradient for each quadrant by averaging the absolute value of the gradient in a strip of width 20 percent of the array dimension centered on the axes which go through the trial image center. These typical values serve as references for the search for a point where the gradient reaches a large enough value that it can be identified as the image limb.
CIRCULAR ARC AVERAGING - EDGE DEFINITION
Still using the trial image center, a series of circular arcs can be defined as a function of trial image radii in each quadrant. Within each quadrant, x+ say, and for each trial image center and trial image radius the gradient is averaged along arcs extending in the y+ and y- directions for this case with between 0.15 R and 0.45 R. The series of trial radii begins at a distance larger than the expected solar radius and is successively decreased until the gradient averaged along this arc segment exceeds 20 percent of the reference gradient. This detected radius is slightly larger than would be identified by locating the point of maximum slope.
SCATTERED LIGHT CORRECTION
Many of the images have scattered light primarily due to the condition of the dispersion grating which was in a pit below the observing floor and subject to conditions of variable and high humidity. This scattered light is primarily in the cross-dispersion direction since the only light hitting the optics comes through the entrance slit. During each sequence of decreasing trial radius, the average of the gradients for the trial values more than two steps previously and less than 10 steps previously is calculated. This average is taken to be representative of the scattered light and is subtracted from the gradient at the trial radius under consideration. The scattered light correction is limited to 50 percent of the reference gradient.
In addition to the basic criterion above, a number of veto conditions are tested and the trial radii will be decreased further if any of the veto conditions are found to be true. These veto conditions are: Plate scratches produce paired and opposing gradients which are either vertical or horizontal, plate edges produce strong gradients which are either horizontal or vertical and which are also not curved. If the gradient maximum is found to have one of these properties, it is identified as either a line (paired positive and negative) or an edge (unpaired) and is vetoed. Very strong dust spots or plate pits produce very high, paired slopes of opposite sign. If a dust spot is found along the averaging arc, the gradient maximum is vetoed. An area off the plate has a constant value equal to the clear scan result so that its rms is zero. Such areas are identified as blank and if the radius scan stops at a boundary of a blank area, it is vetoed. Finally, if the trial radius becomes smaller than 95 percent of the expected radius, the search is stopped. After the search is stopped, the test causing the previous point to be unacceptable is examined and its nature is used to calculate an image demerit value. The proper test to have failed is that the prior gradient was too small.
IMAGE CENTER ITERATION
After radii are determined in the four quadrants, a new trial center is taken having its x and y coordinates as the averages of (x+ and x-) and (y+ and y-) respectively. The quadrant radius searches are repeated iteratively until the trial center no longer changes.
After the sequence converges, a demerit value is computed based on veto conditions and the deviation of the image from its expected size and its deviation from being circular. The demerit calculation is given below. Each of the calculations for vetos, scattered light and limits is done for every quadrant while the shape calculation applies to the final converged set of x and y radii. When each radius scan is stopped the following conditions lead to the specified demerits. The code indicated is included in the summary file.
Outer Radius Limit: r-scan stops at the first possible point, 5 demerits. Code R
Inner Radius Limit: r-scan stops at inner limit, 5 demerits. Code r
Scattered Light 50% limitation applied: 2 demerits. Code S left of other codes.
Line, Edge, Dust, Blank Veto: 1 demerit. Codes respectively: L, E, D, B
Gradient condition: r-scan passes gradient test, 0 demerits. Code G
If the Gradient condition and either the Line or Edge condition are simultaneously responsible for the previous point not being accepted, the code is given as l or e respectively.
The demerits from image shape are calculated as:
1 demerit for every 10 pixels by which the radius deviates from the expected value.
1 demerit for every 10 pixels of difference between the x and y radii (an offset of 15 is used to compensate for typical image conditions in the early years).