FixUVW

Note

fixuvw has changed its name to fixvis because it will eventually be a general task for performing MS fixes not handled by split or calibration. The page name is staying the same for now, though.

Scope

fixvis will be the CASA equivalent of AIPS' UVFIX. It may need to do these things, in roughly decreasing order of immediate need:

1. Generate (u, v, w)s for an ms that does not have them.
2. Change the phase tracking center.
3. Correct for differential aberration.
4. Change the equinox (i.e. B1950 <-> J2000 <-> 2006.78).
5. Change the projection (i.e. (-)NCP <-> SIN).
6. Refocus.
7. Differential refraction? (R_Jones)

CASA already has functions which do most of 1, 2, 3, 4, and 6. measures/Measures/UVWMachine has a flag for requesting an EW coordinate system, but it has not been implemented.

To delineate the scope of #1 in more detail I need to know what range of SDMs have to be supported.

Spectral manipulations (like AIPS' CVEL) will probably go in split or a separate CVEL like task.

Multiple Fields in an MS

UVFIX operates on one field at a time, with 2 runs necessary for changing equinoxes. fixvis will attempt to operate on all fields in an MS, and if that is unwanted the MS should be split first.

Parameters

This is a possible list of parameters. If a default is blank, it means that the corresponding action will not be taken or the value in the MS will be left unchanged. Some parameters have multiple forms (i.e. scalar, list, and dictionary), not all of which will be implemented if they are not needed. In particular, the antenna dependent values have multiple ways of being specified, including a list of values which must match the order as the antennas in the MS, or a dictionary keyed by antenna name. Since the order or names of antennas may not match those of an external file, list, or dictionary, it is safest to get them from the MS. Conceivably a dictionary could be used to correct values for just one or a few antennas.

Parameter	Default	Description	Comments
vis	None	The MS filename	This accounts for 8 of UVFIX's parameters, partly because of how AIPS specifies file locations, but more importantly because in fixvis the input MS is also the output.
phase_dir	''	The new phase tracking center.
		A single direction string, or a list or dictionary of them.
		If a list, then there must be a direction for each field.
		A dictionary keyed by source name would usually be better.
xyzs	''	Antenna positions	Along with the coordinate system that UVFIX uses, we might need to support UTM as used by ALMA.
		Non-blank string: filename for almasimmos-type coordinates	Update: the MSes from ASDMs that I've seen which need UVWs come with antenna positions, making the correction of a few positions the most likely use-case for xyzs.
		or: a list or dictionary of triplets.
array_center	''	Triplet: it gets used.	See the bit about coordinate systems above, and consider whether the centroid of xyzs should have any effect.
		String: Observatory name that CASA will use to look it up.
subarray	'' (all)	(list of) Int: subarray(s) to work on.
k_terms	[]	List or dictionary of floats: the distances between the elevation axis and the phase reference center for each antenna.	Should soon start appearing as a column in the antenna table of MSes from heterogeneous arrays. There already is an "offsets" column, but so far I've only seen zeroes, even in CARMA data.
crct_df_abr	True?	Correct for differential aberration?	The parameter name could maybe use a few more vowels, but it should stay under 15 chars.
			Apparently EVLA data will come already corrected for D.A., but VLA data does not. Should fixvis try guessing whether to correct for D.A. based on the telescope? In either case it should have an option to explicitly remove, ignore, or perhaps (if a mistake has already been made) even restore D.A..
			The other reason I can think of to not correct is if it turns out to be slow and negligable. The order of magnitude for the maximum correction is twice the telescope's velocity relative to the speed of light, i.e. positions 100" away from the phase tracking center could be off by 0.01".
clock_err	0.0	Data filled with FILLM versions older than 31DEC07 are off by by half of the MODCOMP online averaging time.

General notes:

UVFIX uses up to 29 parameters, some of which are arrays. The antenna positions use 6 arrays, as a way to get more precision out of 7 character strings. CASA doesn't have that problem.

UVFIX parameters that I am explicitly dropping:

Dropped Parameter	Reason
output filename	It seems safe enough to add a UVW column to an existing MS, or make a copy with cp -r first.
Besselian epoch of the data	Use imhead to make the epoch match the visibilities before running fixvis.
UT1-UTC, IAT-UTC	Can't these always be derived from the observation time of the MS? (Simulations of observations in the future are a known exception I'm happy to ignore.)
Frequency	Fix it with imhead.
Do not compute UVWs	UVFIX has this option so the phase tracking center of data in (-)NCP can be changed without converting it to SIN projection. The imaging routines take care of the phase tracking center "on the fly" without needing fixvis, so it seems strange to call fixvis without changing the UVWs.

I have not said much about selecting data, since my feeling is that leaving an MS partially fixed should only be done if there is a dire need to save time or space. Split might be a better way to do fix part of a dataset without leaving MSes that have inconsistent internal states.

Does CASA have any plans to support orbiting antennas? (My guess is no.) A potential speedup for correcting differential aberration, which would not work for VLBI, is to pretend that every antenna in the array is moving at the same velocity.

Switching the visibilities to an exotic frame (i.e. the CMB) would amplify the effect of differential aberration, but that would be, well, exotic.