Report on Data Format Session at 2001 IAU WG Meeting

The preliminary version of documentation of the format for the interchange of calibrated optical/IR data being developed by Tom Pauls of NRL and John Young of COAST was presented and a number of issues discussed.

The proposed format is in the general form of a relational database with different components stored in FITS binary tables. The different components are related by means of indices into other tables. For example the target being observed is given as an index in the target table.

The discussion points were the following:

  1. Should there be a single data table type that includes all data types, or multiple table types, one per data type?
  2. What data types are needed? Currently considered are Complex visibilities, Squared visibilities and triple products.
  3. Should Virtual columns be used? If all entries in a column are the same, they can be replaced by a keyword. This trades storage aginst complexity. If this is to be implemented, how will it be done?
  4. Array geometry (OI_ARRAY) table. Should this be a single table with all possible, nominal, locations or a separate table per configuration giving the actual locations. The difference is mostly of use for astrometry.
  5. Should the allowable visibility units include physical units such as Janskys as well as normalized units. For circumstellar disks and nonstellar objects normalized visibilities are hard to define.

Discussion

Items 1 and 2

These were discussed jointly. It was generally concluded that there should be a separate table type for each data type. There was no general sentiment in favor of additional types at present but it will be straightforward to implement new data types as new table types. It was also concluded that the names of the tables needed to be explicitly for Optical/IR interferometry data. The new names are the following:

OI_ARRAY specifies array geometry
OI_TARGET specifies targets, positions etc.
OI_WAVELENGTH gives the effective wavelengths of the spectral channels
OI_VIS gives complex visibilities
OI_VIS2 gives visibility squared
OI_T3 gives triple product amplitudes and phases

Item 3

It was concluded that there was no strong argument for using virtual keywords and the simplicity of not using them more than made up for possible increases in storage space.

Item 4

The discussion of this issue was inconclusive.

Item 5

There was a general consensus that physical units as well as normalized units should be allowed by the format definition.

Other issues

Preben Grosbol (ESO) expressed concern that the relational database aspect of the format might be disrupted if constituent tables were moved to separate files. A possible solution is to include a unique identifier in each data set to establish which sets of tables go together.

Action Items

Bill Cotton


Data Exchange Standard | OLBIN

John Young <jsy1001 (at) cam.ac.uk>
Last modified: Mon Mar 18 13:43:32 GMT 2002