COMICS FITS files are 4-D (X x Y x Z x W).

X = x-axis, normally 320 pixels
Y = y-axis, normally 240 pixels
Z = time sequence of data (beamA, chop to beamB, chop back to beamA, etc.) 
W = detectors (COMICS has 6 detectors but ignore this dimension
               for imaging as only one is used)

For example,

COMA00040941.fits[320,240,200,1][long]: object

COMQ00040941.fits[320,240,1,1][long]: object

COMA file contains a series of A1,B1,A2,B2,A3,B3,...
i.e. (co-added) images taken at each chop position (in each beam)
in turn.  In the above example, there are 100 images each taken at
each chop position (beamA and beamB) and so there are 200 images in
total.  At mid-infrared wavelengths, sky is usually much brighter
than an object and each of these images will probably not show
anything.  That is why we employ the chopping technique and what we
should do now is subtract the beamB image from each beamA image.

i.e. (A1-B1), (A2-B2), (A3-B3), etc.

COMQ file has in fact been already through this process and contains
the result of

COMQ = (A1-B1) + (A2-B2) + (A3-B3) + ....

This is what we display on the quick-look viewer.  We provide COMA
files for more thorough data reduction.