Data Set Overview
  =================
    As part of the Midcourse Space Experiment (MSX), the Spatial
    Infrared Imaging Telescope (Spirit III) observed the lunar eclipse
    of September 27, 1996 with 30 arcsecond resolution.  This data set
    consists of three FITS image cubes corresponding to three data
    collection events about 50 minutes apart, during which the eclipse
    was observed for a duration of approximately 20 minutes each time.
    The data are stored in each image cube as two bands, A and B, each
    consisting of six planes.  Band B data were acquired in two narrow
    spectral bands (FWHM = 4.22 to 4.36 and 4.24 to 4.45 microns),
    converted to brightness temperatures in Kelvin, and then co-added
    to produce brightness temperature maps centered at 4.3 microns.
    Band A consists of qualitative (uncalibrated) data from 6.8 to
    10.8 microns, with an effective wavelength of 8.3 microns.  The
    file names reflect the naming convention from the various
    observations that MSX performed.  In this case, the observations
    were made in a calibration program in which all were designated
    'DC.'  '35' indicates the specific subprogram involved, and '02' -
    '04' are the Data Collection Event (DCE) numbers for the lunar
    eclipse set of observations.  Hence the files are DC3502, DC3503,
    and DC3504.  Additional details of MSX and the SPIRIT III
    instrument can be found on pages 3-4 of the accompanying
    presentation (Price_Eclipsed_Moon.pdf/Price_Eclipsed_Moon.htm in
    the 'doc' folder).  Page numbers given in this overview refer to
    that presentation (which should be viewed with the 'notesview'
    option).  Further information on other astronomy experiments by
    MSX can also be found in Price et al. (2001); discussion of the
    calibration can be found in Price et al. (2004).

    Each DCE consisted of ten 6 degree long legs (raster scans), in
    inertial coordinates of the Moon, and lasted 24 minutes (see page
    6).  The scan rate was 0.05 degrees per second, and roughly two-
    thirds of the Moon was in the field of view of the arrays on each
    scan.  The Moon was within the field of view near the end of the
    even numbered legs and near the beginning of the odd numbered
    legs.  Due to the scan rate, length, and geometry, the time
    intervals between complete scans were 3 min, 1 min, 3 min, etc.
    Therefore, data from even- and odd-numbered legs were co-added to
    fill data gaps, thereby reducing the ten legs to six segments.
    Segment 1 is the data from leg 1, Segment 2 is the data from legs
    2 and 3, Segment 3 is the data from legs 4 and 5, Segment 4 is the
    data from legs 6 and 7, Segment 5 is the data from legs 8 and 9,
    and Segment 6 is the data from leg 10.  Note that this is slightly
    different from the averaging done for the presentation (page 6),
    in which the scans were simply paired 1 & 2, 3 & 4, etc.  The
    remaining data 'dropouts' are a combination of the filter mask
    that separates the two 4.3 micron bands on the focal plane array
    and a group of dead detectors (which show up as lines due to the
    scanning).

    The spatial resolution is 45 km or 1.5 degrees in selenographic
    coordinates     at the center of the lunar disk.  Off-moon data
    values are representative of the system and processing noise (for
    best results in viewing image planes, limit the data values to
    >0).  The uncertainty in the derived temperatures is ~1 K (see
    pages 5-6).  The 8.3 micron data (band A) are all in at least soft
    saturation and thus could not be calibrated. The worst areas went
    into hard saturation and those data were set to 0.  Also, in
    DC3504, band B went into saturation as the Moon came out of
    eclipse, which affected portions of the still-eclipsed Moon as
    well (this can be seen on page 8).

    The three FITS image cubes each have 600 lines, 600 samples per
    line, and 20 planes, with a standard FITS header embedded at the
    beginning of the file and a detached PDS label. Image pixels are
    4-byte IEEE floating-point values. The first six planes in the
    cube are the six segments corresponding to the Band B wavelength
    (4.3 microns), and planes 7 through 12 are the six segments
    corresponding to the Band A wavelength (8.3 microns).  Planes 13
    through 18 give the data acquisition times in msec for each pixel
    in each of the six segments.  These describe when in the scan the
    data were taken.  The values are such that a time datum plus the
    value for DCESTART (in the FITS header) equals the number of
    milliseconds from the start of 1996.  Planes 19 and 20 give the
    selenographic longitude and latitude in degrees, respectively, for
    each pixel.

    Price et al. (2003) describe the data collection and processing in
    more     detail, although it should be noted that the images in
    the paper were averaged differently from the images in this data
    set.  The presentation accompanying this data set also contains
    much of this information.  Since the data set is a time series of
    the eclipse, a mosaic of the Moon is not quantitatively
    meaningful.  For qualitative purposes, though, averaged images of
    the 4.3 and 8.3 micron data near totality are shown on pages 9 and
    10, respectively.  Thermal profiles of selected lunar regions
    during the eclipse are shown and discussed on page 12.

This data set underwent external peer review from December 19,
    2005 to March 15, 2006.