Sky-model calibration with holography beams

The continuum and RACS-All continuum flows can calibrate each measurement set against a model of the sky before imaging starts. A reference catalogue is turned into a list of components. Each component is weighted by the measured ASKAP primary beam from a holography cube, predicted into the measurement set with addmodel, and used for one phase-only round of gaincal. In the RACS-All flow the low, mid and high bands are calibrated separately, which ties all three to the same external phase reference.

This page explains how that model is built, with a focus on how the holography beam is applied. That part follows linmos exactly, so the beam used for calibration is the same beam used to correct the mosaics.

Turning it on

The step is off by default. Enable it on the flint_flow_racs_all_continuum_pipeline command line:

flint_flow_racs_all_continuum_pipeline ... \
    --run-skymodel-calibration \
    --calibrate-container /path/to/calibrate.sif \
    --casa-container /path/to/casa.sif \
    --reference-catalogue-directory /path/to/catalogues \
    --low-holofile /path/to/akpb.iquv.closepack36.54.943MHz.SB55219.cube.fits \
    --mid-holofile /path/to/akpb.iquv.closepack36.54.1295MHz.SB66096.cube.fits \
    --high-holofile /path/to/akpb.iquv.closepack36.54.1655MHz.SB34422.cube.fits

  • --calibrate-container provides addmodel and is required.

  • --casa-container provides gaincal and is required.

  • --reference-catalogue-directory holds the reference catalogues (NVSS, SUMSS, ICRF, RACS-low) and is required. Before the flow starts, every one of them is loaded to check it is there and readable. This holds even when the model uses a catalogue of your own, because the flow also makes its validation plots from these catalogues. If no catalogue is named in the strategy file, one is picked by declination: SUMSS below −75°, RACS-low below +26°, and NVSS otherwise.

  • The three holography cubes are optional. Without one, a band falls back to a Gaussian primary beam with a FWHM of λ/D for a 12 m dish.

The flint_flow_continuum_pipeline takes the same options, with the single holography cube that is also used to linmos the field:

flint_flow_continuum_pipeline ... \
    --run-skymodel-calibration \
    --calibrate-container /path/to/calibrate.sif \
    --casa-container /path/to/casa.sif \
    --reference-catalogue-directory /path/to/catalogues \
    --holofile /path/to/akpb.iquv.closepack36.54.943MHz.SB55219.cube.fits

The calibrated measurement sets are named *.round0.ms, and the self-calibration rounds that follow carry on from them.

The options that shape the model itself live in the skymodel section of the strategy file. See Configuration.

What happens, step by step

For each band’s measurement set:

  1. Pick sources. Catalogue sources brighter than flux_cutoff and closer to the phase centre than fwhm_scale_cutoff × λ/D (at the band’s lowest frequency) are kept.

  2. Load the beam once. The beam number comes from the measurement set’s name (beam12 gives beam 12). That beam’s Stokes I response is read from the band’s holography cube and normalised. See Normalisation.

  3. Find each source’s offset. The offset from the phase centre is measured as direction cosines (l to the east, m to the north).

  4. Rotate the offset by α. α corrects for the field’s third-axis orientation differing from the orientation the holography was measured at. See The beam frame and rotation.

  5. Look up the beam. The response is interpolated in the cube at the rotated offset, on every usable frequency plane, then interpolated onto the measurement set’s channels.

  6. Fit the apparent spectrum. The catalogue spectrum is multiplied by the beam response and fitted with a curved power law. Sources whose apparent flux falls below flux_cutoff are dropped.

  7. Write the model. The components are written in the BBS-style text format that addmodel reads. See Gaussian components and position angles.

  8. Predict and calibrate. addmodel predicts the model into MODEL_DATA, and gaincal solves for phases only (calmode: p) and applies them. Its other settings come from the strategy file’s round 0; see Configuration.

The beam frame and rotation

An ASKAP holography cube has axes (beam, stokes, freq, dec, ra). Although its spatial axes are labelled RA and Dec, they are not positions on the sky. The reference value (CRVAL) is (0, 0), and each beam’s slice is that beam’s pattern centred on its own pointing. The cube is a grid of offsets.

linmos reads it that way, ignoring CRVAL and using only CRPIX and CDELT, and flint does the same:

  • The beam centre is the measurement set’s phase centre. Each ASKAP beam is phased to its own pointing. linmos uses each image’s reference pixel, which is the same direction.

  • The offset is rotated before the lookup. For a source at distance d and position angle PA (east of north) from the beam centre, the cube is read at

    x = sin(d) · sin(PA − α)
y = sin(d) · cos(PA − α)

    then converted to pixels with x / CDELT1 + (CRPIX1 − 1), and likewise for y. These are the expressions in linmos’s ASKAP_PB primary beam.

  • α is the same value flint gives linmos. It is the field’s third-axis angle pol_axis, read from the measurement set’s FEED table, plus 45°, the angle the holography is assumed to have been measured at. A field with pol_axis = −45° has α = 0, as in the RACS observations checked while writing this page.

Left, a holography beam slice with a source and its rotated lookup point. Right, the response along the diagonal, comparing the measured beam, the matched Gaussian taper and a plain Gaussian.

Beam 12 of the low-band holography cube SB55219 at 943 MHz. Left: the beam in its offset grid, with east to the left. With an illustrative α of 30°, a source at PA 60° is looked up at PA 30°. Contours are at 0.5 and 0.1. Right: the response along the diagonal at PA 45°, out through the cube’s corner. The measured beam is used all the way to the corner, including the sidelobe beyond 2°. With holography_taper on, a matched Gaussian takes over where the cube ends. The plain λ/D Gaussian, used when there is no holography, is narrower than the measured beam.

Left, a holography beam slice with a source and its rotated lookup point. Right, the response along the diagonal, comparing the measured beam, the matched Gaussian taper and a plain Gaussian.

Beam 12 of the low-band holography cube SB55219 at 943 MHz. Left: the beam in its offset grid, with east to the left. With an illustrative α of 30°, a source at PA 60° is looked up at PA 30°. Contours are at 0.5 and 0.1. Right: the response along the diagonal at PA 45°, out through the cube’s corner. The measured beam is used all the way to the corner, including the sidelobe beyond 2°. With holography_taper on, a matched Gaussian takes over where the cube ends. The plain λ/D Gaussian, used when there is no holography, is narrower than the measured beam.

Frequency. Planes with no response at the beam centre are dropped. These are usually flagged holography channels stored as NaN; in the mid-band cube SB66096, for example, everything below 1296 MHz is NaN. The remaining planes are put in increasing frequency and interpolated linearly onto the measurement set’s channels. linmos uses the nearest plane instead, which makes no practical difference with 1 MHz planes. Channels more than half a plane beyond the usable planes take the nearest plane’s beam, and a warning says how many.

Normalisation

Each frequency plane is divided by its value at the beam centre, so the response is exactly 1 there. That matches the flux scale of the data: the bandpass is calibrated on PKS B1934−638 placed at each beam’s centre.

The cubes are not quite normalised as stored. They read about 1.005 to 1.016 at the beam centre, and peak a further 0.5–1% higher, 2–6′ away from it. linmos uses the stored values, so mosaic fluxes and this model can differ by 0.5–1.5%. A phase-only solution is not sensitive to that.

Sources beyond the edge of the cube

The cubes are small. They span about ±1.7° in the low and mid bands and ±0.9° in the high band. By default a source off the cube gets a response of 0 and is dropped, which is what linmos assumes.

With holography_taper: true, those sources are kept. Inside the cube the measured beam is still used everywhere, exactly as linmos uses it. That includes the corners, which reach about 1.4 times further out than the edges and can hold real sidelobes: a few percent at 2.1–2.3° on the diagonals of the low-band cube. For a source off the cube, follow the line from the beam centre through the source to where it leaves the cube, one pixel inside the edge. The response is then:

  • the measured value at that exit point, clipped at 0;

  • multiplied by how much a λ/D Gaussian falls off between the exit point and the source.

The response is therefore continuous at the edge and falls off smoothly in the source’s own direction. It is 0 where the measured value at the exit point is 0 or below. The exit point is used rather than the nearest pixel because the beam is not circular: the nearest pixel can lie in a different direction, and so a different part of the pattern. Only a Gaussian is used for the fall-off. Airy and sinc² patterns have nulls inside the cube edge in some bands, so they cannot be matched to the measured value safely.

To bring in more distant bright sources, also raise fwhm_scale_cutoff, which sets how far out sources are looked for.

When something goes wrong

Each case below either stops with an error or logs a warning:

  • The MS name has no beam number. The beam is read from the MS’s FEED1 column instead, with a warning starting BEAM NUMBER NOT IN NAME. If FEED1 holds more than one beam, it stops with an error.

  • The beam has no usable response. That is, every plane is NaN or zero at the beam centre. A warning names the cube and the beam, and that measurement set falls back to the Gaussian beam. For example, beam 12 of the high-band cube SB34422 is entirely zero. The rest of the flow carries on as normal.

  • The cube’s beam centre is at its edge. If the reference pixel (CRPIX) is within a pixel of the cube’s edge, it stops with an error. ASKAP holography cubes are centred on the beam, so this means the cube is not the kind expected.

  • Channels fall outside the usable planes. A warning gives the number of channels and the usable range; those channels take the nearest plane’s beam.

  • No sources are left to model. If none survive the flux and radial cuts, the flow stops with a NoFieldSourceError. Lowering flux_cutoff, raising fwhm_scale_cutoff or turning on holography_taper may help.

Gaussian components and position angles

What is written for each source

Each component goes into the model file as one line, with its position, the fitted apparent spectrum and its shape:

  • POINT if both axes are under 1″, or if it has no position angle.

  • GAUSSIAN otherwise, with the major axis and minor axis in arcseconds and the orientation in degrees.

A missing position angle needs care. In NVSS, the position angle is blank exactly when both axes are only upper limits, which is 84% of the catalogue. Those sources are unresolved, so they are written as points. A nan orientation would be worse than useless: addmodel would predict NaN for every visibility, not just that source’s.

The sizes must be intrinsic (deconvolved), because the instrument’s resolution comes from the visibilities themselves. NVSS and SUMSS list deconvolved sizes. For a catalogue of your own, say which kind you have with catalogue_sizes_deconvolved. If the sizes are as observed, name the per-source PSF columns (catalogue_psf_maj_col, catalogue_psf_min_col, catalogue_psf_pa_col) so they can be deconvolved.

The rotation α does not move sources

α only changes where in the cube a source’s beam response is read. It turns the beam pattern, not the sky. The sources’ shapes stay as the catalogue gives them, for three reasons:

  • addmodel predicts with the measurement set’s own (u, v, w), which are fixed by the baselines and do not depend on how the phased array feed is turned.

  • ASKAP’s third axis holds the feed at a fixed angle on the sky for the whole observation, so the beam does not turn during a track either. flint checks that the angle is constant.

  • fixms changes the polarisation basis, not the Stokes I geometry.

North at the source versus north at the phase centre

There is one real frame difference, and it has nothing to do with α. Catalogue position angles are measured east of north at the source. addmodel builds each Gaussian directly in the (u, v) plane of the phase centre, so it measures the angle from north at the phase centre. Away from the phase centre, meridians converge, and the two norths differ by an angle η:

η = atan2(−sin δ · sin Δα, cos δ · cos δ₀ + sin δ · sin δ₀ · cos Δα)

Here δ is the source’s declination, δ₀ the phase centre’s, and Δα the difference in right ascension (source minus phase centre). η is measured east from north at the phase centre to north at the source. It is zero for sources due north or south of the phase centre. It is negative east of a northern field and positive east of a southern one, and grows roughly as Δα · sin δ.

Phase centre Dec

1° east

1.7° east

1.7° at PA 45°

+18.3°

−0.33°

−0.56°

−0.42°

0°

0.00°

0.00°

−0.03°

−30°

+0.58°

+0.98°

+0.66°

−60°

+1.73°

+2.94°

+1.98°

−80°

+5.65°

+9.55°

+6.05°
A source 1.7 degrees east of a phase centre at Dec −80°, with arrows for north at the source and north at the phase centre, the angle η between them, and the catalogue shape against the shape addmodel would draw.

A source 1.7° east of a phase centre at Dec −80°, with east to the left. North at the source (orange) is turned 9.5° from north at the phase centre (blue). A catalogue position angle of 35° is measured from the orange arrow. Without the correction, addmodel measures the same 35° from the blue arrow, so the source is drawn 9.5° off. The shapes are enlarged for clarity.

A source 1.7 degrees east of a phase centre at Dec −80°, with arrows for north at the source and north at the phase centre, the angle η between them, and the catalogue shape against the shape addmodel would draw.

A source 1.7° east of a phase centre at Dec −80°, with east to the left. North at the source (orange) is turned 9.5° from north at the phase centre (blue). A catalogue position angle of 35° is measured from the orange arrow. Without the correction, addmodel measures the same 35° from the blue arrow, so the source is drawn 9.5° off. The shapes are enlarged for clarity.

With rotate_pa_to_phase_centre: true, η is added to each Gaussian’s position angle as it is written, so addmodel draws the source the way the catalogue describes it.

When to turn it on. It matters only for elongated sources in fields near the poles. Near the equator η is a fraction of a degree. A compact source looks much the same at any position angle, and phase-only calibration is barely affected by source shapes at all. It is off by default.

What it changes. Only the calibrate model file passed to addmodel. The ds9 region file and the hyperdrive model keep the catalogue’s own position angles, which is what those tools expect.

Configuration

The RACS-All flow reads the sky-model step’s options from the strategy file, for self-calibration round 0: the model from the skymodel section, and the calibration from the gaincal section. Values under defaults apply first, and values under selfcal: 0 override them. Every option on this page:

defaults:
  skymodel:
    # Catalogue to use; unset picks SUMSS, RACS-low or NVSS by declination
    reference_name: NVSS
    # Minimum brightness, in Jy, before and after the beam is applied
    flux_cutoff: 0.02
    # How far out to look for sources, in units of λ/D at the lowest frequency
    fwhm_scale_cutoff: 1.0
    # Keep sources beyond the edge of the holography cube with a matched taper
    holography_taper: false
    # Measure position angles from north at the phase centre, as addmodel expects
    rotate_pa_to_phase_centre: false

selfcal:
  0:
    skymodel:
      holography_taper: true
    # Calibration against the sky model; always phase-only
    gaincal:
      solint: 120s
      minsnr: 3.0
      uvrange: ">200m"
      nspw: 1

The sky-model calibration is always phase-only. A calmode other than p in its options is ignored, with a warning. Without a gaincal section, gaincal runs on flint’s defaults: solint: 60s, minsnr: 0.0 and uvrange: ">200m". Values under defaults: gaincal: also apply to the regular self-calibration rounds, which each read the gaincal section of their own round.

The flow always writes the calibrate model, and takes the catalogue directory from --reference-catalogue-directory. To build the model from a catalogue of your own instead, see Using your own catalogue.

Using your own catalogue

The model can be built from any catalogue astropy can read, instead of the built-in reference catalogues. Point catalogue_path at the file in the skymodel section of the strategy file and name its columns. Nothing is guessed: every column and unit the model needs must be given. With catalogue_path set, reference_name is ignored and the model does not use the reference catalogue directory. --reference-catalogue-directory is still required and checked, because the flow’s validation plots use it.

Each beam still gets its own model. From the one file, only sources within fwhm_scale_cutoff × λ/D of that beam’s phase centre and brighter than flux_cutoff are kept.

The options

Option

Needed

What it gives

catalogue_path

always

The file. Its format comes from the extension (.fits, .vot, .ecsv, .csv). Use a path every worker can read.

catalogue_ra_col, catalogue_dec_col

always

Position columns

catalogue_radec_unit

if not degrees

Unit of the position columns

catalogue_flux_col

always

Flux of each component at the reference frequency

catalogue_flux_unit

if not Jy

Unit of the flux column

catalogue_freq or catalogue_freq_col

one of them

The reference frequency in Hz for every source, or a column holding one per source (with catalogue_freq_unit if not Hz)

catalogue_name_col

no

Component names, used only as labels. Without it, components are named src0, src1, …

catalogue_maj_col, catalogue_min_col, catalogue_pa_col

no, but all three or none

Shape: major and minor axis (FWHM) and position angle, east of north in degrees. Without them, every component is a point.

catalogue_shape_unit

if not arcsec

Unit of the axis columns, including the PSF columns below

catalogue_sizes_deconvolved

when shapes are given

true if the axes are already intrinsic (deconvolved), false if they are as observed

catalogue_psf_maj_col, catalogue_psf_min_col, catalogue_psf_pa_col

when catalogue_sizes_deconvolved is false

Each component’s restoring beam, used to deconvolve the observed axes

catalogue_alpha_col, catalogue_beta_col

no

Spectral index and curvature. Without them, assumed_alpha (−0.83) and assumed_beta (0) are used for every source.

Things to check in your catalogue

  • Units are labels, not conversions. Each *_unit option is stamped onto its column, replacing whatever unit the file declares. If a FITS column holds mJy and catalogue_flux_unit is left as Jy, every flux is 1000 times too bright. Set each unit to what the file actually holds.

  • Use the per-component flux (integrated flux of each Gaussian), not a peak flux or a whole-source total.

  • One unit for all the shape columns. catalogue_shape_unit covers the axis columns and the PSF columns together, so they must share a unit. If they don’t, use deconvolved axes instead.

  • Blank shapes are fine. A component with no position angle, or with both axes under 1″, is written as a point. That matches how PyBDSF and NVSS mark unresolved sources.

  • Every row is used. There is no option to filter rows, so remove anything you don’t want modelled (flagged, spurious or duplicated components) before pointing the flow at the file.

  • Large catalogues are read in full for every beam and band. That is 108 reads for a 36-beam, three-band run. Write a copy that keeps only the columns the model uses, and if you can, only the sky around your fields:

    from astropy.table import Table

columns = ["RA", "DEC", "Total_flux", "DC_Maj", "DC_Min", "DC_PA", "Gaussian_ID"]
table = Table.read("survey_components.fits")
table[columns].write("survey_components_skymodel.fits")

Example: a PyBDSF Gaussian list

PyBDSF writes RA and Dec in degrees, Total_flux per Gaussian in Jy, and deconvolved sizes (DC_Maj, DC_Min, DC_PA) in degrees, with both axes set to 0 for unresolved Gaussians. The reference frequency is the FREQ0 keyword in the table header. Use the sky-plane columns, not the _img_plane ones. For a RACS-low3 Gaussian list at 943.491 MHz:

defaults:
  skymodel:
    catalogue_path: /path/to/RACS-low3_flint_components.fits
    catalogue_ra_col: RA
    catalogue_dec_col: DEC
    catalogue_radec_unit: deg
    catalogue_flux_col: Total_flux
    catalogue_flux_unit: Jy
    catalogue_freq: 943491000.0
    catalogue_name_col: Gaussian_ID
    catalogue_maj_col: DC_Maj
    catalogue_min_col: DC_Min
    catalogue_pa_col: DC_PA
    catalogue_shape_unit: deg
    catalogue_sizes_deconvolved: true
    assumed_alpha: -0.83
    assumed_beta: 0.0

The as-observed columns (Maj, Min, PA) can’t be paired with PyBDSF’s PSF columns here: the axes are in degrees and the PSF columns in arcseconds, and catalogue_shape_unit has to cover both.

Building a sky model from the command line

flint_skymodel builds the same model for one measurement set. Pass --holofile to take the beam from a holography cube; without it, the Gaussian beam is used. It accepts every option above:

Create a calibrate compatible sky-model for a given measurement set.

usage: flint_skymodel [-h] [--holofile HOLOFILE]
                      [--reference-catalogue-directory REFERENCE_CATALOGUE_DIRECTORY]
                      [--reference-name REFERENCE_NAME]
                      [--assumed-alpha ASSUMED_ALPHA]
                      [--assumed-beta ASSUMED_BETA]
                      [--flux-cutoff FLUX_CUTOFF]
                      [--fwhm-scale-cutoff FWHM_SCALE_CUTOFF]
                      [--write-hyperdrive-model] [--write-calibrate-model]
                      [--write-ds9-region] [--holography-taper]
                      [--rotate-pa-to-phase-centre]
                      [--catalogue-path CATALOGUE_PATH]
                      [--catalogue-freq CATALOGUE_FREQ]
                      [--catalogue-freq-col CATALOGUE_FREQ_COL]
                      [--catalogue-freq-unit CATALOGUE_FREQ_UNIT]
                      [--catalogue-ra-col CATALOGUE_RA_COL]
                      [--catalogue-dec-col CATALOGUE_DEC_COL]
                      [--catalogue-flux-col CATALOGUE_FLUX_COL]
                      [--catalogue-name-col CATALOGUE_NAME_COL]
                      [--catalogue-maj-col CATALOGUE_MAJ_COL]
                      [--catalogue-min-col CATALOGUE_MIN_COL]
                      [--catalogue-pa-col CATALOGUE_PA_COL]
                      [--catalogue-sizes-deconvolved CATALOGUE_SIZES_DECONVOLVED]
                      [--catalogue-psf-maj-col CATALOGUE_PSF_MAJ_COL]
                      [--catalogue-psf-min-col CATALOGUE_PSF_MIN_COL]
                      [--catalogue-psf-pa-col CATALOGUE_PSF_PA_COL]
                      [--catalogue-alpha-col CATALOGUE_ALPHA_COL]
                      [--catalogue-beta-col CATALOGUE_BETA_COL]
                      [--catalogue-radec-unit CATALOGUE_RADEC_UNIT]
                      [--catalogue-flux-unit CATALOGUE_FLUX_UNIT]
                      [--catalogue-shape-unit CATALOGUE_SHAPE_UNIT]
                      ms

Positional Arguments

ms

Path to the measurement set to create the sky-model for

Named Arguments

--holofile

Holography cube to take the primary beam from. If unset, a Gaussian beam is used

Inputs for SkyModelOptions

--reference-catalogue-directory

The reference catalogue directory that contains the known flint reference catalogues

Default: .

--reference-name
--assumed-alpha

Assume this to be the typical spectral index if it is not recorded in the reference catalogue

Default: -0.83

--assumed-beta

Assume this to be the typical amount of spectral curvature should they not be in the reference catalogue

Default: 0.0

--flux-cutoff

The intrinsic brightness a source needs to be for it to be included in the sky model

Default: 0.02

--fwhm-scale-cutoff

A source needs to be within this many FWHM units from the direction of interest for it to be included

Default: 1

--write-hyperdrive-model

Should the model for hyperdrive be created. The output will have .hypderdrive.yaml suffix appended to the MS path.

Default: False

--write-calibrate-model

Should the model for calibrate be created. The output will have .calibrate.txt suffix appended to the MS path.

Default: False

--write-ds9-region

Should a DS9 region file be created. The output will have .ds9.reg suffix appended to the MS path.

Default: False

--holography-taper

Keep sources beyond the edge of the holography cube, continuing the beam past the edge as a Gaussian fall-off from the edge value. Off drops them, as linmos does.

Default: False

--rotate-pa-to-phase-centre

Measure position angles in the calibrate model from north at the phase centre, as addmodel expects, instead of from north at each source

Default: False

--catalogue-path

Path to a user-supplied catalogue file. If set, takes precedence over reference_catalogue_directory/reference_name. Column names/units below are never guessed.

--catalogue-freq

Reference frequency, in Hz, of catalogue_path’s flux column, used for every source. Ignored if catalogue_freq_col is set. One of catalogue_freq/catalogue_freq_col is required whenever catalogue_path is set.

--catalogue-freq-col

Per-source reference-frequency column in catalogue_path. If set, overrides catalogue_freq with a per-row value instead of one value for the whole catalogue.

--catalogue-freq-unit

Astropy unit string for catalogue_freq_col

Default: 'Hz'

--catalogue-ra-col

RA column in catalogue_path. Required whenever catalogue_path is set – never guessed

--catalogue-dec-col

Dec column in catalogue_path. Required whenever catalogue_path is set

--catalogue-flux-col

Flux column in catalogue_path. Required whenever catalogue_path is set

--catalogue-name-col

Source name column in catalogue_path. If unset, names are synthesised (src0, src1, …)

--catalogue-maj-col

Major-axis column in catalogue_path. None disables shape info (point sources). Must be set together with catalogue_min_col/catalogue_pa_col

--catalogue-min-col

Minor-axis column in catalogue_path, paired with catalogue_maj_col

--catalogue-pa-col

Position-angle column in catalogue_path, paired with catalogue_maj_col. Always degrees

--catalogue-sizes-deconvolved

Whether catalogue_maj_col/catalogue_min_col are already intrinsic (PSF-deconvolved) sizes, ready to use directly in the sky model, or as-observed (PSF-convolved) sizes that must be deconvolved first. Required whenever catalogue_maj_col is set – never guessed

--catalogue-psf-maj-col

Per-source PSF major-axis column in catalogue_path, used to deconvolve catalogue_maj_col/catalogue_min_col when catalogue_sizes_deconvolved is False. Required in that case – there is no pipeline restoring beam to fall back on here

--catalogue-psf-min-col

Per-source PSF minor-axis column, paired with catalogue_psf_maj_col

--catalogue-psf-pa-col

Per-source PSF position-angle column, paired with catalogue_psf_maj_col. Always degrees

--catalogue-alpha-col

Spectral index column in catalogue_path. If unset, falls back to assumed_alpha

--catalogue-beta-col

Spectral curvature column in catalogue_path. If unset, falls back to assumed_beta

--catalogue-radec-unit

Astropy unit string for catalogue_ra_col/catalogue_dec_col. Always applied explicitly, never inferred from the file

Default: 'deg'

--catalogue-flux-unit

Astropy unit string for catalogue_flux_col

Default: 'Jy'

--catalogue-shape-unit

Astropy unit string for catalogue_maj_col/catalogue_min_col/catalogue_psf_maj_col/catalogue_psf_min_col

Default: 'arcsec'