Data Set Overview
  =================
    This data set contains raw images of Comet 9P/Tempel 1 obtained
    with MIRSI (the Mid-InfraRed Spectrometer and Imager) at the NASA
    Infrared Telescope Facility (IRTF) on Mauna Kea, Hawaii.  Included
    are dark images used in processing the comet observations.
    Nightly images of HR5340 (Alpha Boo) were obtained as the primary
    flux standard.  Calibration images were also obtained of HR5315
    (on July 13 and 18 only), HR5056 (Spica, one observation on July
    6), and of the asteroid 68 Leto.
 
    Nightly logs giving details of the observations, including sky
    conditions, are provided in the documents directory.
 
    The MIRSI camera is described in a paper by Deutsch et al. 2003
    [DEUTSCHETAL2003].  It utilizes a 320 x 240 Si:As Impurity Band
    Condustion (IBC) array developed by Raytheon/SBRC.  On the IRTF,
    MIRSI has a 85 x 64 arcsec field of view with a pixel scale of
    0.27 arcsec.  The nominal point source sensitivity for a 1-sigma
    detection in a 60 second integration is 20 mJy at 10 microns, and
    100 mJy at 20 microns.
 
    The MIRSI observations of Comet Tempel 1 were taken through a
    series of discrete filters with the following parameters:
 
        Band     Central wavelength     Band pass
                     (microns)
          M             4.9                21%
                        7.7               9.0%
                        8.7               8.9%
                        9.8               9.4%
          N            10.4                46%
                       11.6               9.9%
                       12.5               9.6%
                       18.4               8.0%
 
    Details for the standard stars are listed below.  In the case of
    HR5340 (Alpha Boo) published magnitudes are provided for
    bandpasses close to those provided with MIRSI.
 
    HR 5056  Alpha Vir (Spica)   Spectral type = B1V
               V = 1.04    N (est.) ~ 1.6
 
    HR 5315  Kappa Vir    Spectral type = K3III
               V = 4.19   N = 0.88
 
    HR 5340  Alpha Boo (Arcturus)  Spectral type = K2III
               V = -0.04    N = -3.16
           4.8 microns  =  -2.93   [COHENETAL1995]
           7.8 microns  =  -3.08   [GEZARIETAL1993]
           8.7 microns  =  -3.12   [COHENETAL1995]
           9.8 microns  =  -3.13   [GEZARIETAL1993]
          10.3 microns  =  -3.15   [GEZARIETAL1993]
          11.7 microns  =  -3.16   [COHENETAL1995]
          12.5 microns  =  -3.23   [GEZARIETAL1993]
          18.4 microns  =  -3.20   [GEZARIETAL1993]
 
    The observations were made in chop-nod mode, resulting in 3-D
    image cubes with four image planes of 320 by 240 pixels each:  two
    chop pairs that are offset by a small nod in the telescope
    position.  The dark frames were recorded as 2-D 'grab frame'
    images.
 
    The MIRSI array is read out through 16 parallel readout lines,
    each controlling the output from 20 array columns.  There is
    common-mode (common to each of the 16 outputs) pattern noise that
    repeats every 20 columns, but which is temporally variable from
    frame to frame.  There are also column bleed and level shifts
    which occur adjacent to very bright pixels, as well as row-wise
    repetition of bright pixels in all outputs.  Except for bleeding
    and level shifts, various median filtering techniques can be used
    to minimize the pattern noise in each output region as part of the
    data reduction.
 
    Some HINTS for Data Reduction:
    At mid-infrared wavelengths, the atmosphere and telescope are both
    significant sources of thermal radiation.  As a result, sky images
    taken in the mid-IR have very high background counts, to the level
    where bright point sources are difficult to see in a single,
    non-sky-subtracted image.  For this reason, all of the data
    presented here are taken as series of chop-nod image sets, which
    allow for proper sky subtraction.
 
    Each data file contains four image planes stacked along the third
    dimension.  These image planes, referred to here as A, B, C, and
    D, are the first chop pair (A and B), where the telescope
    secondary is chopped back and forth usually in the N-S direction
    at a frequency of a few Hz, followed by the second chop pair (C
    and D) taken after the telescope position was nodded slightly,
    usually in the E-W direction.
 
    To reduce these data, one of the first steps involves subtracting
    the background sky in each of the corresponding pairs: A - B, B -
    A, C - D, and D - C.  If the telescope chop and nod distances were
    relatively small (smaller than the field of view on the array),
    then a positive - negative pair of images should appear in each of
    these subtractions.
 
    Because the chopping of the secondary offsets the light path in
    the camera, a residual signature of the telescope radiation may
    remain in the subtracted image.  This can be removed by
    subtracting the results from the two nod positions - ie: (A - B) -
    (C - D).  Again, if the chop and nod distances are relatively
    small, the result from this subtraction should contain four
    images, two positive and two negative, often in a square pattern
    (if the N-S secondary mirror chop distance was the same as the E-W
    telescope nod distance).
 
    Depending on the stretch of the image, pattern noise may be
    visible in the background of the chop-nod subtracted image.  This
    repetitive common-mode noise pattern (described above) can be
    mostly removed by: 1) subdividing the image into 16 sub-images,
    each corresponding to 1 readout channel, 20 columns wide, 2)
    subtracting any residual mean sky level from each sub-image (to
    zero out small shifts in the background between the different
    readout channels), and 3) taking the pixel-by-pixel median of the
    sub-images.  This produces a noise pattern image (with mean count
    level of zero) that can be subtracted from each readout channel.
    When observing bright sources, level shifts in the background are
    usually seen in those channels containing
    the brightest pixels (shifts in the background level in that
    20-column wide channel, in those rows either above or below the
    bright source on the array).  These level shifts can be modeled
    and removed in the final stages of data processing.
 
    Summary:
    These data were obtained through a coordinated effort by the
    following observers:
       Diane Wooden
       Carey Lisse
       Neil Dello Russo
       David Harker
       Michael S. Kelley
       Chick Woodward
 
    MIRSI is a PI instrument, made available at the IRTF through an
    agreement with Boston University.  As part of that agreement, the
    MIRSI instrument team should be included in the authorship of any
    publications resulting from this data set.

Confidence Level Overview
  =========================
    The quality of the MIRSI data is limited by various sources of
    electronic noise, including the temporally-variable pattern noise
    that repeats across each output section (the effects of which can
    be minimized using median filtering techniques), and bleeding and
    level shifts associated with bright sources, such as standard
    stars.  The start times recorded in the MIRSI image headers are
    derived from the instrument computer clock.  Because the IC clock
    is not automatically synchronized to a time standard, slow drifts
    in this clock lead to uncertainties of several seconds in the
    exposure start times.  The observation end times are estimated by
    the MIRSI software, based on the total exposure times, readout and
    chop frequency parameters.