During October 1998 observing time was awarded on the Jacobus Kapteyn Telescope in La Palma for early February 1999. This presented an ideal opportunity to test the prototype. For this reason the prototype was constructed to operate on, and optimized for, this particular data set. The data was RIZ observations of low mass stars in the Praesepe cluster, with each target object was observed in all three filters.
The prototype PL was written using several programming languages and pre-written astronomical data reduction packages, to increase speed of production. The aim of the prototype was to find difficulties not apparent when reducing data manually. The prototype was written in several sections or “pipes” which, when connected together formed the PL.
Outlined here are each of the pipes specifying, the programming languages, any pre-written packages, difficulties and the solutions employed. Each of the prototype pipes have been written on a cluster of DEC alpha workstation running Digital Unix V 4.0F and a SUN Sparc Ultra 5 workstation running Solaris Unix 7. See Figure 4.1 for a schematic view of the pipe-line.
Once identified the individual calibration frames must be combined to form the master versions. This is achieved through use of the CCDPACK package (Draper, 1999). The master frames are achieved by median filtering the relevant files together. The locating of self-similar files and automation of this pipe is achieved through use of a C shell script.
De-biasing is performed by the methods described in Section 2.4.1. First a mean bias level is calculated from bias strips of the object frame. This value is then used to normalise the bias frame’s bias strips, and hence the bias frame, to the same bias level as the object frame. This normalised bias frame is then subtracted from the object frames.
Flat-fielding is implemented by the method described in Section 2.4.3. It is important that the correct pass band is used to flat-field the object frames. To this end the automation of this script is again carried out using a C shell script with in-script use of the programming languages awk, a pattern scanning and processing language, and sed, a stream editor.
This pipe also normalises all object frames to a one second time-scale, by interrogating the HDUs for the exposure time, as described above and dividing the object frame by this value. Photometry may now be performed on the reduced files.
The pipe’s automation was carried out using C shell scripts on SLs routines and complemented with in-script use of awk and sed commands.
above the average background level of the frame, this gave a 99.7% probability that the
feature detected was real (Bevington and Robinson, 1992). It will be seen later that this was a
reasonable level of detection, as all target objects were detected. The background level was
computed by KAPPA using a median algorithm. A true image median would need to use all
10242 values contained in the image. However this is impractical, the more values used the
longer the computation time will be. The number of values used is a tradeoff between
accuracy and processing time. Currie and Berry (1999) note that: “It is hard to
quantify the tradeoff precisely; testing with a typical CCD image of stars and galaxies
increases the process time approximately as the log of the number of bins, the
accuracy of the pixel sum for varying values of histogram bins used was: 100, 2.8%;
1000, 0.1%; 10000, 0.03% and 100000, 0.002%”. The default value of 2048 bins,
giving an accuracy of 0.02%, was the value used, using the default value increased
production of the prototype by speeding coding whilst the accuracy produced was
sufficient.
A single value above this detection level can not be considered an object, a true object detection is judged by a group of contiguous pixels above the specified level. A target object must be integrated upon long enough to gain a suitable signal to noise and the spread of a point source is governed by the seeing. If a good signal to noise had been achieved then most objects are found to have spread to occupy at least 6 pixels. However inspecting the object frames it was found that some of the targets occupied at least 8 pixels. This larger value was used as the pixel threshold to reduce objects being detected that were not real.
Having calculated the necessary inputs the software used to do this work was PISA, automated through the C shell and complemented with in-script use of awk and sed commands. For each object frame a file was generated which contained the integrated intensity of each detected object, its peak intensity, its location within the frame the number of pixels above the threshold value, the object’s ellipticity, object orientation and an object index number. The generated file then needs to be reconstructed back to a 32 bit data range. This was achieved by an external awk script which divides the columns, peak intensity and integrated intensity, by the normalising constant calculated in pipe 4.
Once the target object has been located its data can be extracted from the file generated by PISA. The data may then be input, along with a unique identification tag which indicates the where data was extracted from, to a file created to record the target object data from a full nights observations. Also written to this file are the airmass and filter band, these data being extracted by cross-referencing the identification tag with the object frames and interrogating the HDUs as above.
If the images are of standard objects, e.g., objects with known magnitudes, then a separate file is created in the same format as the object file, however the magnitude must also be recorded to the data file to enable construction of airmass curves. The standard objects for the RIZ data used during the construction of the PL were extracted from Cossburn et al. (in prep).
Much of the work carried out in this pipe is manual. The automation that could be carried out was accomplished through use of FORTRAN 77, the C shell and complemented with in-script use of awk and sed commands.
In the Z band the standard stars used were those presented by Cossburn et al. (in prep), which were again observed with same telescope and instrumentation used to observe the target objects. Therefore the Z band observations are on the Cossburn et al. (in prep) system which is defined using the Zrgo filter by setting the I-Z colour of an un-reddened A0 star to be 0.00.
This pipe was coded using FORTRAN 77. The file containing data for the standard stars is appended with the converted magnitudes.
level by the code
generating the fit. These data however may still be correct. Such points are checked
both against the observing logs, to ensure nothing was noted at the time by the
observer, (e.g., telescope vignetted), and against the original image frame (e.g., a
cosmic-ray may have occurred). If a scientific reason can be found for removing
the data point then this is carried out by hand. The fitting routine may then be
re-attempted.
The code for this pipe was written using FORTRAN 77 and complemented manually through use of a text editor.
 | (4.1) |
To calculate the actual magnitude of the object the airmass correction calculated in pipe 7 must be used. Each straight line equation calculated by pipe 7 is unique to its filter band. In calculating the true target magnitude, m, the correct equation must be selected and the appropriate airmass used to calculate the zeropoint, zp, correction factor. Equation (4.1) may then be re-written as:
 | (4.2) |
A second magnitudes file is produced by this pipe in the same configuration as the first but with Rc and Ic magnitudes, the Z magnitudes being untouched.
Print
Email
News!
