POINTING table is not filled

[bennahugo]

As discussed elsewhere in calibration implementation the (compulsory) POINTING table is not being generated. This means that the pointing information (necessary for applying the far field response to a model) is lost on datasets that were rephased.

I suspect the following is never actually added into the generation of the dataset:

katdal/katdal/ms_extra.py

Lines 855 to 900 in 4898eaf

	def populate_pointing_dict(num_antennas, observation_duration, start_time, phase_center, pointing_name='default'):
	"""Construct a dictionary containing the columns of the POINTING subtable.
	The POINTING subtable contains data on individual antennas tracking a target.
	It has one row per pointing/antenna?
	Parameters
	----------
	num_antennas : integer
	Number of antennas
	observation_duration : float
	Length of observation, in seconds
	start_time : float
	Start time of observation, as a Modified Julian Date in seconds
	phase_center : array of float, shape (2,)
	Direction of phase center, in ra-dec coordinates as 2-element array
	pointing_name : string, optional
	Name for pointing
	Returns
	-------
	pointing_dict : dict
	Dictionary containing columns of POINTING subtable
	"""
	phase_center = phase_center.reshape((2, 1, 1))
	pointing_dict = {}
	# Antenna Id (integer)
	pointing_dict['ANTENNA_ID'] = np.arange(num_antennas, dtype=np.int32)
	# Antenna pointing direction as polynomial in time (double, 2-dim)
	pointing_dict['DIRECTION'] = np.repeat(phase_center, num_antennas)
	# Time interval (double)
	pointing_dict['INTERVAL'] = np.tile(np.float64(observation_duration), num_antennas)
	# Pointing position name (string)
	pointing_dict['NAME'] = np.array([pointing_name] * num_antennas)
	# Series order (integer)
	pointing_dict['NUM_POLY'] = np.zeros(num_antennas, dtype=np.int32)
	# Target direction as polynomial in time (double, -1-dim)
	pointing_dict['TARGET'] = np.repeat(phase_center, num_antennas)
	# Time interval midpoint (double)
	pointing_dict['TIME'] = np.tile(np.float64(start_time), num_antennas)
	# Time origin for direction (double)
	pointing_dict['TIME_ORIGIN'] = np.tile(np.float64(start_time), num_antennas)
	# Tracking flag - True if on position (boolean)
	pointing_dict['TRACKING'] = np.ones(num_antennas, dtype=np.uint8)
	return pointing_dict

[bennahugo]

Relevant to issues ratt-ru/codex-africanus#227 and ratt-ru/QuartiCal#37

[ludwigschwardt]

Thanks Ben, I just discovered this again for myself while helping out with reference pointing via CASA 😂

I guess I would like someone to tell me that I filled the table correctly. Any pointers to the correct values appreciated!

[bennahugo]

@ludwigschwardt to first order I think this needs to be filled with requested pointing values per scan - this was quite a problem for the Lunar data (we manually built the table for the AIPS data to get the PA calculations working). I understand it is probably low on the rung so to speak - I just documented the issue so that we don't forget about it