hem         j      h      k
	     n        0.035  0.035  0.035
	     s        0.035  0.075  0.035
                

• Added capability to blank only between streak endpoints (plus metwid) if length is flagged (<0) in *.met file. This provides restricted blanking for slow satellites and other things causing streaks in consecutive frames that can result in coverage holes.
• Modifed to call streakc to flag streaks found in 3 or more consectutive frames so that pixphot v.010523 can blank only the identified streak length for those frames.
• Program to identify and flag streaks found in 3 or more consecutive frames. These streaks are flagged by negating the length parameter, so that pixphot will blank only the area between the endpoints in the *.met file (plus metwid). This is a patch for slow moving satellites which sometimes cause holes in the coadds due to consecutive frames being blanked all the way across the frames. This fix will probably occasionally leave pieces of satellite tracks in the frames, as streakr does not reliably identify the exact ends of tracks.
• Modified to add the date/scan to the header of the *.met output file.

• New southern PSFs delivered. These are essential "v2.x" style PSFs, which will exhibit the chi-squared distribution offsets. here
• The previous "fix" to the PSF weighting became obsolete when new PSFs were generated, and was removed.
• The adjustment of the blend flag is to allow for the fact that some sources which are being processed as single-component fits may have a stronger previously-subtracted source in their vicinity. The latter would have included the current source as a passive deblend. The new blend flag therefore acknowledges the presence of a close neighbor in these cases.
• Improvement of the confusion error (really background dispersion). estimation. Previously, it was estimated from a trimmed average of the sky values in an annulus surrounding the source. This procedure was susceptible to contamination from nearby bright sources and to sky gradients in general. A symptom of the problem was the presence of anomalously low chi squared values for faint sources.

  The new procedure uses an estimate whose statistical properties are representative of the pixels within an area corresponding to the data circle used in PSF fitting. The procedure is as follows:

  (1) Construct an annulus surrounding the source, and divide into sectors of similar size to the PSF-fitting data circle.
  (2) Calculate the variance of sky background in each sector using a straight (untrimmed) average.
  (3) Take the median value of the sector variances as the best estimate of the background variance.

• Namelist changes to eliminate problems with very brightest srcs and their artifacts (1st item), and r1 srcs with mags=99.999 and saturated r2 matches (2nd item).
  
  • namelist max photometric confusion radius (confph_rmax) was set too low, effectively limiting delta mag to ~16 -- raised to allow for delta mag up to ~20.
  • namelist r1 position mag threshold (r1pos_thr) set to m=18, so r1 srcs with m=99.999 would use r2 position where available -- eliminated the r1pos_thr check to force using both r1 position and mag values when r2 nsat>0, and set namelist value to 109.
• Set delta_mag = 0 when possible pers src is fainter than predicted pers mag, to improve pers probabilities for pers from very bright srcs especially -- r2 pers srcs had mags of ~10-11 no matter what the parent's mag
• Do not use previous pers position when searching for next pers src -- keep parent position, since it's now a good position in almost all cases
• Change r1 artifact conf radius calculation back to log-linear as before version 010418 -- artifact conf radii needed to be fixed (enlarged) slightly for r1, and enormously for r2, for very bright parents; now r1 radii calculation same as r2 but m0 shifted by 3.5 (pipe.cal values are 3.49 mag apart for r1 and r2)
• Namelist changes: diff_lmax increased -- was much too small for very bright parents, confar* params redone (increased) using ver.3 RTB nights for r2 srcs, and r1 parameters set the same but m0 shifted from 6.0 to 2.5 (pipe.cal values are 3.49 mag apart for r1 and r2)

• Don't apply bmg-to-pfmp correction to Read1's. Input select for bmg-to-pfmp computation changed.
  

   
      Problem:
    
      There has been a recurring problem with slight mismatches between
      Read1 and Read2 positions because of their different paths through
      the pipeline.  Attempts to remove the difference open-loop by
      adjustments in Frexas and ProPhot have been only partly successful.
     
      Solution:
      
      Since the Read1 positions originally come from the PosFrm solution
      (as represented by the pfmp file) it was decided not to apply the 
      bmg-to-pfmp correction to these sources. For sources which are
      a mix of Read1 and Read2 inputs, the correction is weighted
      according to the number of Read1's and Read2's by band.
      Band fills are dropped from the weighting computation.
  

• Previously, clean sources with mk>8.5 were used to compute the linear bmg-to-pfmp correction. Now, since the Read1:Read2 status by band is available, only clean sources with no Read1's in any band are used.
• Fixes a problem with the exr1 values of msig when only one "good" measurement is available within the band, but one (or more) saturated measurements are available. My thanks to Gene Kopan for noting the msig values were incorrect.

• The first is to lower the floor on the isophotal radius from 7" to 5".
• The second mod is to the bright galaxy background removal process. Instead of fitting across the entire 6-deg scan, I fit across adjacent coadds. This means that very large galaxies (e.g., m31), that cross coadds, will have an over-estimated background (resulting in a lower integ. flux). This disadavantage is, however, counter-weighed by the greater overall robustness against wildly incorrect background removal, giving much better reliability (not to mention the fact that >95% of all nearby galaxies are smaller than a 2MASS coadd; e.g., NGC 891).

• Minor modification to correct "field overflow" problem in output "scale" parameter.
• The background noise value, for the purposes of the photometry noise model, is now based on the estimated value from the annulus surrounding the source rather than from the assumed values of gain and read noise. This desensitizes the chi squared values to errors in the gain and read noise.
• Flux desensitization. The flux solution is weighted so as to de-emphasize the wings of the PSF. The algorithm is now essentially back to how it was during the mid-May RTBs.
• The switch to turn off the desensitization is in the form of an extra parameter, desens, in the prophot namelist files. If desens=.false., the desensitization will be turned off. The default is .true.
• Degeneracy removal step in passive deblending. This step is designed to prevent the flux of a source from being split into two components at the same location. It involves a chi squared test which was not being applied correctly. This fix may correct the completeness dip at a separation of about 5 arcsec.
• Modify PSF term in active deblending noise model. Previously, the PSF error contribution to active deblending noise model had been based on the flux obtained by coarse-grid search. This led to underestimates in some cases, resulting in a spurious low-sigma branch in plots of sigma v. magnitude. The PSF error contribution is now based on the aperture magnitude, which will somewhat overestimate the error, but at least the errors will be conservative.
• Fix Poisson noise term in noise model. A long standing error was present, in which the Poisson noise had been based on the pixel values of the subtracted image rather than the variance-added image.
• Restore variance floor, but with rescaling. The floor on the range of pixel variance used in the PSF-fitting noise model was removed earlier because of some undesirable side effects in the form of bifurcations in plots of sigma v. magnitude. Recent tests showed that the floor was necessary to alleviate numerical instability due to the large dynamic range of the variance map. However, it is now accompanied by a scaling which restores the proper chi squared behavior. The latter effect has made the sigma values of bright sources more reasonable.

• Correct namelist parameter that was allowing "far" optical associations (dr>5") to be passed. Also set the default condition in the code to not pass the far associations in case namelists are missed.
• Use x-scan scale factor corrections. GenSFP has been modified to accept x-scan scale factor corrections in the dxcor input file. It has also been modified to accept uncertainty inputs directly as sigma values from sigx and sigy files. Originally it was set up to accept uncertainty inputs as variance factors from vfacx and vfacy files which were then converted to sigma values. Having the input files contain sigmas makes it much easier to do manual checking.
• Use sigx/sigy inputs rather than vfacx/vfacy. Scan-segment position corrections for many more scans are now available and for the first time scan-segment position uncertainties are also available.
• In order to make use of the changes described above, changes have also been made to the PosFrm script as well as PosFrm namelist used when scan-segment corrections are available.
• dxcor, dycor, sigx & sigy files augmented. The pos. corrections and the corresponding uncertainties as well as x-scan scale corrections for all scans in Dec bands 15-30 (everything < +6 degrees) are now available. In addition 840 scans in Dec band 14 also have full corrections. The scans with full corrections can be identified using either the sigx or sigy files. Scans contained in the dxcor and dycor files, but not the sigx and sigy files, are from the 2nd Release prototype Martinizing.

• Change faint r1 src rejection criteria to msig > 9.990, instead of 9.9985 -- rejected srcs must be r1, have no r2 match, no saturated pixels, msig > 9.990, and be fainter than r2 saturation mag
• New namelist parameters -- slightly larger artifact confusion radius, defaulting to the lookup tbl values for all r1 normalizations
• New namelist parameters -- slightly different predicted persistence mag offsets. These will slightly improve the persistence probability calculations, mainly in the south. (We don't yet have enough data for the north to get improvements in the higher-order persistence src mag offsets, e.g. 3,4,... frames beyond the parent.) The persistance probabilities for the vast majority of srcs should not change drastically.
• Updated southern namelist for mapcor which has different dens_slp paremeters. These are the parameters which shorten the diffraction spikes as a function of increasing source density. Recent plots of many southern bright stars (more than in the original analysis of the density effect) revealed that we were making the spikes too short by about a factor of two, and thus we were not flagging the ends of the spikes. The northern namelist will very likely need the same treatment, and we will perform the same analysis.

• Synchronized version numbering on scripts for debugging/configuration purposes. For some unknown reason in the qadeltas script sm failed to generate a gif file using "sm input temp999zp.mac" so it is now called using "sm < temp999zp.mac" which does seem to work.
• New astrometric wander diagnostic plot delivered. Scansci code updated to use it.
• Bug fixed in qasat_read2.csh that now correctly leaves plot in datequal directory.
• Scansci now only reports "unexpected scan" errors for the updatesit file; this should match the list of scans skipped for v3 production. Redundant extra scan reports have been suppressed.
• The template QAreview file has been reorganized slightly to match workflow.
• The order of plots on the PSF-AP plots web page have been reordered.
• Index page error list now lists v2 grades for scans in the updatesit file that were not processed for v3 (as a sanity check against which scans were excluded from v3 prod).
• Changed "missing scan" output messages in input.c
• Bug introduced in last version (by datestamps) corrupted headers of .qacat and .qagrade files; now fixed.
• Workfiles in night root directory generated by Roc's various science plot scripts are cleaned up by scansci.csh.
• qaastwndr.csh needed a full path to /2massc/utl/bin/stf (not in 2mops path).
• Both scansci and scansci.csh have been modified to add links to the observer's log and the version 2 QA review.

• As originally coded the ds2dxy and ds2sig routines, which interpolate position corrections and uncertainties, returned the end-segment values once past the Dec of the end-segment. This is sometimes resulting in a rapid error growth at extreme end points. With this mod the correction is extrapolated past the end point.
• Fix bias in sfpos ra/dec values. During delivery testing for the above change a bias was found in the sfpos ra/dec positions. This problem has been masked by the fact that although PosFrm does not allow the relative J-band frame-to-frame positions to change from the sfpos values, it does allow the connected set to float in order to get the best Tycho-2 fit.

• Misplaced comment line fixed in .qagrade & .qacat tables.
• The minor planet summary file now includes RA & Dec to make it easier to track down missed objects.
• Astronomical wander plot now has night in plot title.
• SCANSCI code has a clarified error message for bad cumtab files and now implements the photometric downgrade for the old observing strategy of #cal set downgrade = sqrt(n/6).
• Reverted to earlier form of .mpsum file to fix unknown awk bug introduced in last update. This bug also prevented the web pages from updating properly.
• RA & Dec now in .mpsum file
• .kmax based on .ptscat
• Fixed a bug that would prevent the plot from being made if a scan was missing from v2 that was present in v3 (as with mystery mapcor losses).
• Overlay of total number of v2 source counts is added to the source difference panel.
• makeplots.pl fixes a minor axis label bug (makes the 'p' in 'ap' subscripted); Hess plots are now filtered to reduce junk. hessplot2 calls /2massd/qatools/overlay2.pnm instead of version in John's home directory
• qar1r2stats.pl superscedes qasat_read2.csh (which is no longer used and should be removed from the pipeline. This updated version is faster and filters out artifact-tagged sources. scansci.csh cleans up after this script better deleting the many sizable temp files.
• qaastwndr.csh now does a better job cleaning up after itself, deleting work files

• Use original v3 PSFs, generated in early March
• Default namelist parameters for PSF weighting:
  desens=.true.
  totvarfloor=.true.
  
• Active deblending turned off in all scans
• Namelist change to make gain and readnoise values for two subintervals of the third northern hardware made the same (gain=7.3, readnoise=56.0)

• Delivery of scan segment corrections and sigmas derived from global Martinizing.

  Comments on Positions - Analysis of overlap, Tycho-2 and UCAC residuals after the Martinizing indicates a large improvement (>2x on overlap sigma's). There are a few scans (~0.1% to 0.2%) where inconsistencies remain. For Dec <6 degrees all scans with radial correction differences w.r.t. UCAC >0.5 asec on one or more segments, dropped back to UCAC corrections for the whole scan. A total of 62 scans out of the 38,470 in that Dec range were so adjusted, UCAC corrections were not used in general because of small biases between UCAC and Tycho-2. Not sure which is right, but the 2MASS mandate is to reconstruct to Tycho.

  Comments on Uncertainties - The sigma's as delivered are somewhat conservative (0.6 < chisq < 0.7). Given more time, the sigma's could be reduced to bring the chi-squares up to 1.0. There is another factor, however, pushing them in the opposite direction. Additional errors are added when the fixed frame-to-frame solution from the sfpos is allowed to float to the best Tycho-2 fit. Recent pipeline tests have shown this to be particularly true for low density scans. Without a fix to the "float" problem it is expected than the frame position uncertainties may be considerably underestimated for low density scans.




• Modify date and time of processing tags in Martinizing flags for 3 nights that were reprocessed after initial preliminary processing.

• Brightest Ks source in each scan search removed from SCANSCI
• Galaxy diagnostics now include two additional plots (col-col & col-mag) which are now grouped on a web page.
• Global plot documentation is now linked to the nightly index pages
• The plot html pages in qasciplots now have title tags to make the windows easier to identify.
• Fix path to "stf" for galaxy diagnostic plots

• Changes to filters for selecting sources for scan overlap statistics. SNR threshold changed to magnitude threshold (J<15.5, H<14.5 and Ks<13.7 - which correspond to the nominal SNR=20:1 levels). This compensates for elevated photometric uncertainties from certain PSF families. Probability of persistence limit changed to <0.2 (was 0.5).
• Chi-squared limit for selecting sources for overlap statistics raised to 5.0 (from 3.0) for northern data taken before the first summer shutdown. This is to accomodate extremely elongated images on some nights.
• Array holding science scan data for overlap plots was too small for unusual science scan blocks of 10 scans or more. Increased to hold up to 50 scans.
• Changed differential and cumulative overlap plot titles
• Output mean position offset ellipse parameters for overlap scans. These will provide long-term chi-squared diagnostics to help characterize Martinizing residuals

• Changed coadd_dat bkgnd_sig_rej and bkgnd_sig_his <= 0 to be nulls (not an error in format)
• Fixed pt_src default mag to be null if rd = 9
• Fixed pt_src snr calc -- do if rd > 0 and rd < 6
• Fixed mp_mch id_mp read -- was 1 char to left of where should be
• Changed MP *.aster file read to use new (catform-type) format