The UV Tables in the original GILDAS V1 data format, created in 1989,
handled visibilities and their required associated data parameters in
a very similar way as the UVFITS data format used to export data from
the VLA at that epoch. A visibility consisted in
- 7 associated data parameters (daps), in fixed order: u, v, w, date, time, first antenna, second antenna
- for each channel, a real part, and imaginary part, and a weight (a so-called complex visibility element).
The daps precede the complex visibilities elements.
The 7 (daps) were real numbers. As Plateau de Bure is a 2-D
array with small field of view, w was occasionally replaced
by the scan number scan for debugging purpose. The number of
channels was simply derived from the visibility size:
Early 2000, an extension was made to this constraining layout by optionally adding two (real) numbers after the channel values. This was made necessary by the development of the ALMA simulator to study the impact of the compact array (ACA).
With further developments coming on, such as polarization, on-the-fly mosaicing, multi-source UV FITS files from CASA, etc... this scheme was becoming far too constraining. In addition, UV data sets could readily exceed the 2 GByte size limit of the GILDAS V1 data format. We decided to take advantage of the changes required to handle large data sets to implement a more flexible UV data format.