SHORT_MODEL

        SHORT_MODEL

    This task simulates observations performed with single-dish antennas, to
    be used later on as short spacings information. The model image is  con-
    volved  by the antenna lobe (via Fourier Transform) and the intensity is
    then estimated at (i) the position of the mosaic fields, and (ii) on  an
    externally  defined  grid (see IMAGE_SAMPLING documentation). If the re-
    quired position does not coincide with a pixel center, a bilinear inter-
    polation from the neighbor pixels is performed.

    Pointing  errors  can  be  simulated  by  estimating  the intensity at a
    slightly wrong position. This task can generate simple kind of  pointing
    errors  but  it can also read an input pointing error table (enabling to
    simulate much more complex kind of pointing errors).

    Thermal noise can then be added  to  the  data,  and  the  corresponding
    weight is stored. Finally, a calibration error can be simulated: the ob-
    served intensities are multiplied by a random factor  whose  mean  value
    (different from 1 if a systematic error is present) and rms can be spec-
    ified. In this case, we obtain: results = cal_rr*(model+thermal0ise).

    The error on the amplitude gain is modeled as the sum of an offset and a
    drift  with time. The offset and drift values are randomly reset at each
    calibration for each antennas. In addition, an offset common to all  the
    antennas can also be added to the amplitude gain.

    Intensity unit of input image is supposed to be Kelvin. The output table
    has the following format: 4 columns x Number of antenna  x  (NFIELDS$  +
    Number  of observed positions). Column 1 is the X offset in radian, col-
    umn 2 the Y offset in radian, column 3 the weight and column 4 the  flux
    in  Jy.   The  first NFIELDS$ lines of the model are the position of the
    mosaic for use by UV_ZERO and all the other ones are the nod of the  ex-
    ternally  defined  grid  (by  IMAGE_SAMPLING) for use by UV_SINGLE. This
    complex line layout comes from historical reasons (as usual)...

    Limitation: only image can be processed (i.e. *no*  data  cube)  meaning
    that this task can not (yet) handle spectra cubes.