IPAC 2MASS Working Group Meeting Minutes #58

IPAC 2MASS Working Group Meeting #58 Minutes, 3/21/95


T.Chester, T.Conrow, T.Evans, J.Fowler, T.Jarrett, G.Kopan, B.Light, C.Lonsdale, J.White


  1. Galaxy Processing Workshop Report
  2. Time-Variable PSF for Protopipeline
  3. Analysis Tasks for April Data


The minutes of meeting number 57 indicated that benchmarks had been run on karloff with CPUs upgraded from 50 MHz to 75 MHz; this is incorrect; the 75 MHz CPUs were on a machine loaned for a few days by Sun.


  1. Galaxy Processing Workshop Report -- T. Chester reported certain highlights from the Galaxy Processing Workshop held last week in Boston. The entire session was nonstop work. Two recommendations by the committee were: (a.) wherever possible, perform scan-to-adjacent- scan smoothing in order to minimize striping in various parameter values (e.g., photometric accuracy, completeness and reliability, etc.); (b.) remove source detection from the galaxy code and use the ProPhot detection list exclusively.

    Point (a.) was suggested to apply to the PSF estimation, in which context it was found to be very difficult to escape single-scan orientation, since the seeing is expected to vary over time scales of the same order as a scan. Furthermore, the scan processing order is essentially random because the design takes advantage of the processing of all calibration scans before any survey scans, and several scans will generally be processed simultaneously on different CPUs and/or machines, making it impossible for one survey scan to depend on any other survey scan's estimate of the PSF (or other parameter). In the case of the PSF, a time history is to be generated in 2MAPPS, so the processing runs after the first could take advantage of previous runs' PSF estimates. It was also noted that scan-to-scan nonuniformities always appear terrifying to science teams but tend to go unnoticed by users, the IRAS ecliptic latitude dependence of position accuracy (primarily in single-band sources) being an outstanding example (apparently no user ever mentioned this well documented and unavoidable effect caused by sun-locked scanning).

    Point (b.) was questioned, but its expected reduction of CPU time was said to make it worth whatever (expectedly small) impact it may have on completeness.

    T. Jarrett reported that the galaxy processor is now ready for integration into the protopipeline.

  2. Time-Variable PSF for Protopipeline -- G. Kopan, B. Light, and J. Fowler reported on the plan for incorporating a time-variable PSF into the protopipeline. Since it was never possible to study seeing variations with previous data (the optical PSF, focus drift, and optical misalignment caused too many problems, and the only good night from June '94 involved very good seeing the entire night), we do not have a data set of KAMPhot PSF models spanning the usable range of seeing. This missing set of PSFs will be simulated by convolving the June '94 KAMPhot PSF model with Moffat functions of varying blur parameter (Moffat functions are the currently fashionable model of the blurring due to seeing; they are similar to Lorentz distributions with denominators raised to variable powers).

    G. Kopan is currently developing the STATS subsystem, from which he will extract code to estimate a seeing parameter to be used by KAMPhot to select one of the PSF models. The parameter and its relation to KAMPhot PSF models will be obtained via simulation of a constant scan observed with variable seeing. KAMPhot will construct a PSF model for each seeing level, and these will be compared to the Moffat-blurred real PSF models to determine which are the most similar and what the corresponding values of the seeing parameter are. Assuming acceptable relationships can be obtained (i.e., a pairing of members from the two sequences of PSF models can be determined), then KAMPhot can be made to use the seeing parameter to select a Moffat-blurred real PSF model, thus obtaining a scan-dependent PSF model. This is the level of time dependence sought for the April run, and the exercise should contribute to our understanding of how to handle seeing in 2MAPPS.

  3. Analysis Tasks for April Data -- C. Lonsdale distributed a preliminary list of analysis tasks identified for the April data and organized by S. Wheelock. Some tasks have names associated with them. Comments should be emailed to C. Lonsdale.