---------------------------------------------------------------------------- The data is naturally divided into 49 parts for fields BUL_SC1 -- BUL_SC49. Reference images for each field (2k x 8k FITS frames) are stored in the subdirectory reference_frames. Information is available in two formats: plain text (subdirectory plain_text) and binary FITS tables (subdirectory fits_tables). FITS format is an astronomical standard and the ease of its use with programs like IDL is remarkable. There are two types of files for each field: the catalog of candidate variables, and the database of light curves. The catalog contains a single entry per object with the overall parameters of the light curve and identifying information, like coordinates. ---------------------------------------------------------------------------- Format of *_cat.dat files and *_cat.fts. Names in brackets give the names of columns in the first binary FITS table. Below is a sample record with the explanation of fields: 3764 207.14 6721.60 271.272315 -28.578460 18:05:05.35 -28:34:42.5 17.407 0.783 359.5 23.38 0.13 1 152 0 181 181 1 1. [omitted] number of candidate variable as returned by the pipeline 2. [X_TPL] x template coordinate (0.0 is the middle of the bottom left pixel) 3. [Y_TPL] y template coordinate (0.0 is the middle of the bottom left pixel) 4. [RA] RA in decimal degrees 5. [DEC] DEC in decimal degrees 6. [RA_STR] RA in sexagesimal hours 7. [DEC_STR] DEC in sexagesimal degrees 8. [MEAN_MAG] mean of all magnitude values which could be determined 9. [MAG_SCAT] scatter of all magnitudes used in mean calculation 10. [REF_FLUX] reference flux 11. [MAG_0] magnitude zero point 12. [ID_RAD] distance between the centroid of the variable and the nearest DoPhot star in the reference image in pixels (for the calculation of the reference flux and conversion to magnitudes) 13. [VTYPE] type of variable coded in bits of a 2 byte integer: 1st bit -- ``transient'', 2nd bit -- ``continuous''. Therefore the value of the integer will be 1 for ``transient'', 2 for ``continuous variable '', and 3 for both 14. [N_FRAMES] number of frames used in centroid determination 15. [N_BAD] number of bad pixels in the fitting radius on the reference image 16. [NGOOD] number of ``good'' flux measurements. A ``good'' point is the one for which non of the several types of problems monitored by the pipeline occurred (flags 1--11 are set to 0, see below). 17. [NMAG] number of magnitude values which could be determined (the ones which are not determined come from non-positive fluxes) 18. [FLAG] flags, see the explanation below ---------------------------------------------------------------------------- Several kinds of problematic situations are reported as flags in the last column of the catalog file. Flags are explained in Appendix A. The database of light curves includes all measurements for all detected objects. Light curves in plain text format are stored one per file and grouped by the field. For example, subdirectory BUL_SC1 in plain_text contains 4597 bul_sc1_*.dat.gz files, compressed to save space and transfer time. ------------------------------------------------------------------------------ The columns in light curve files (*_db.fts and *.dat): 1. [OBS_TIME] Heliocentric Julian Day of the observation, offset by 2450000.0 2. [DIFF_FLUX] difference flux (-99.00 for error code) 3. [FLUX_ERR] difference flux error (-99.00 for error code) 4. [MAG] I band magnitude (-1.0 for error code) 5. [MAG_ERR] I band magnitude error (-1.0 for error code) 6. [FLAG] flags, explained below ------------------------------------------------------------------------------ The flags are explained in Appendix B. They provide a lot of information on whether the measurement is valid or not and common problems which may have affected its reliability. To stay on the conservative side, only measurements with no flags should be used (integer value 0). In plain_text subdirectory there are also 49 bul_sc*_db.tar files with all light curve files for a given field grouped together for convenient transfers. In binary FITS format all light curves for each field are stored in a single table. The first extension contains the names of the frames and starting times of the drift-scan exposures in Heliocentric Julian Days shifted by 2450000.0. These time stamps are identical for all objects in a single image. However, the effective time of mid exposure varies depending on the position of the object along the scan. The corrected times of observations are provided for each star separately in the second FITS extension with the actual photometry. All measurements for all stars are stored in the same columns and identified by their index within the column. The number of observations per star is fixed and given by the length of the time vector from the first extension. Ordering is such that the number of the individual observation within a single light curve is ascending fastest along the column of the binary table. If, for example, the number of dates in the first extension of bul_sc1_db.fts is 197, the first 197 rows of the second FITS extension correspond to the first light curve, the next 197 rows are the second light curve and so on. The total number of rows is 197x4597, where 4597 is the number of candidate variables in BUL_SC1 field, the same as the number of rows in catalog files bul_sc1_cat.dat and bul_sc1_cat.fts. This information, along with several other useful numbers, is stored in headers. In Appendix C we include the explanation and values of the pipeline parameters which were important for detection of variables. ---------------------------------------------------------------------------- APPENDIX A Catalog flags are coded as single bits of a 4 byte integer and listed below (the least significant bit first). This provides a very efficient way of storing information about up to 32 conditions that occurred or not by setting 1 or 0 for individual bits of the binary representation. Integer value 12, e.g., means that flags 3 and 4 are true and the rest are false. For pipeline parameters relevant to setting the flags we quote their names and actual values used in our analysis. 1. crowding flag, set if within +/- 4 pixels of the maximum pixel with flux f0 there is a secondary local maximum with pixel flux f>0.15*f0*r, where r is the distance from the star centroid in pixels 2. fewer than N_FRAMES = 4 used in centroid finding 3. more than N_BAD = 0 bad pixels on the reference image within the fitting radius of 3.0 pixels 4. fraction of less than MIN_GFRA = 0.5 difference flux measurements are ``good'' from the total number of frames taken for the field. A ``good'' point is the one for which non of the several types of problems monitored by the pipeline occurred (flags 1--11 in Appendix B are set to 0). 5. mean magnitude and its scatter could not be calculated because fewer than 2 individual magnitudes were defined ---------------------------------------------------------------------------- \section{APPENDIX B} Light curve flags are coded as single bits of a 4 byte integer and listed below (the least significant bit first). This provides a very efficient way of storing information about up to 32 conditions that occurred or not by setting 1 or 0 for individual bits of the binary representation. Integer value 12, e.g., means that flags 3 and 4 are true and the rest are false. For pipeline parameters relevant to setting the flags we quote their names and actual values used in our analysis. 1. pipeline returned error code for difference flux 2. pipeline returned error code for flux error 3. chi2 per pixel of the difference subframe larger than MAXCHI2I = 6.0 4. chi2 per pixel of the PSF fit larger than MAXCHI2N = 1.0e32 (effectively no cut) 5. FWHM of the PSF fit larger than MAX_FWHM = 3.4 pix 6. number of bad pixels within the fitting radius larger than MAX_NBAD = 3 7. correlation coefficient with the PSF lower than MIN_CORR = 0.0 (effectively no cut) 8. currently unused (always 0 = false) 9. star in the rejected region of the CCD (currently empty) 10. flux error NSIGERR = 10 times larger than percentile ERRFRAC = 0.5 of all individual flux errors (0.5 corresponds to median) 11. per pixel of the PSF fit NSIGCHI2 = 10 times larger than percentile CHI2FRAC = 0.5 of all individual values (0.5 corresponds to median) 12. magnitude could not be calculated due to missing or non positive fluxes 13. magnitude error could not be calculated due to missing or non-positive values ---------------------------------------------------------------------------- APPENDIX C Explanation of light curve cleaning parameters. For pipeline parameters relevant to setting the flags in Appendices A and B we quote their names and actual values used in our analysis. 1. [MAXNBAD0 = 0] max number of bad pixels on the reference image within the fitting radius 2. [MINNFRM0 = 4] min number of frames used in centroid calculation 3. [MAX\_NBAD = 3] max number of bad pixels on a given image within the fitting radius 4. [MIN\_GFRA = 0.5] min fraction of good points within entire sequence of frames 5. [BAD\_FLUX = -99.0] error code for difference flux 6. [BAD\_ERR = -99.0] error code for flux error 7. [MAXCHI2N = 1.0e32] max chi2 per pixel for PSF fit 8. [MAXCHI2I = 6.0] max chi2 per pixel for difference subframe 9. [MIN\_CORR = 0.0] min correlation coefficient with the PSF 10. [MAX\_FWHM = 3.4] max FWHM in pixels 11. [ERRCLN = 1] is flagging of large error bars on ? (1 = yes) 12. [NSIGERR = 10.0] base threshold is multiplied by this factor to get the final threshold for error bar 13. [ERRFRAC = 0.5] percentile of the error distribution for base threshold 14. [CHI2CLN = 1] is flagging poor PSF fits on ? (1 = yes ) 15. [NSIGCHI2 = 10.0] base threshold is multiplied by this factor to get the final threshold for chi2 of the PSF fit 16. [CHI2FRAC = 0.5] percentile of the chi2 per pix distribution for base threshold ----------------------------------------------------------------------------