Measurement Set conventions

simms follows a few Measurement Set (MS) conventions that are easy to get wrong if you’re scripting around it directly (rather than through telsim – building a Measurement Set/skysim – predicting visibilities).

Telescope metadata has a single source of truth

The per-antenna telescope/type label lives in the ANTENNA table column named by --telescope-name-column (default TELESCOPE_NAME). This is what skysim – predicting visibilities reads to select a primary beam per antenna. It is never inferred from other columns such as DISH_DIAMETER – if the column is absent, tools should fail clearly rather than guess.

Pointing centre vs. phase centre

These are two different directions, and conflating them silently mis-centres the primary beam:

• FIELD.PHASE_DIR is the correlator’s phase-tracking centre. It is arbitrary and can be shifted (e.g. by phase rotation) without changing where the antennas are actually pointing.

• POINTING.DIRECTION is the antenna pointing centre, and is what the primary beam is centred on.

Use simms.skymodel.beams.read_pointing_centre() to get the beam centre rather than reading FIELD.PHASE_DIR directly.

Spectral frame must be set

SPECTRAL_WINDOW.MEAS_FREQ_REF must be 5 (TOPO). casacore defaults this to 0 (REST), which leaves the spectral frame undefined and makes CASA imaging fail with “No MeasFrame specified for conversion of Frequency”. telsim sets this correctly when it creates an MS.

STRING columns

casacore STRING columns are numpy object dtype, and simms writes them in one chunk (da.from_array(values, chunks=n)). Adding a new column to a standard subtable needs an explicit descriptor, e.g.:

xds_to_table(..., "{ms}::ANTENNA", columns=[col], descriptor="mssubtable('ANTENNA')")