def test_phase_rotation(self):
self.vis = create_visibility(
self.lowcore,
self.times,
self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre,
weight=1.0,
polarisation_frame=PolarisationFrame("stokesIQUV"))
self.vismodel = predict_skycomponent_visibility(self.vis, self.comp)
# Predict visibilities with new phase centre independently
ha_diff = -(self.compabsdirection.ra - self.phasecentre.ra).to(
u.rad).value
vispred = create_visibility(
self.lowcore,
self.times + ha_diff,
self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.compabsdirection,
weight=1.0,
polarisation_frame=PolarisationFrame("stokesIQUV"))
vismodel2 = predict_skycomponent_visibility(vispred, self.comp)
# Should yield the same results as rotation
rotatedvis = phaserotate_visibility(
self.vismodel, newphasecentre=self.compabsdirection, tangent=False)
assert_allclose(rotatedvis.vis, vismodel2.vis, rtol=1e-7)
assert_allclose(rotatedvis.uvw, vismodel2.uvw, rtol=1e-7)
def ingest_visibility(self, freq=None, chan_width=None, times=None, reffrequency=None, add_errors=False,
block=True):
if freq is None:
freq = [1e8]
if times is None:
ntimes = 5
times = numpy.linspace(-numpy.pi / 3.0, numpy.pi / 3.0, ntimes)
if chan_width is None:
chan_width = [1e6]
if reffrequency is None:
reffrequency = [1e8]
lowcore = create_named_configuration('LOWBD2-CORE')
frequency = numpy.array([freq])
channel_bandwidth = numpy.array([chan_width])
phasecentre = SkyCoord(ra=+180.0 * u.deg, dec=-60.0 * u.deg, frame='icrs', equinox='J2000')
if block:
vt = create_blockvisibility(lowcore, times, frequency, channel_bandwidth=channel_bandwidth,
weight=1.0, phasecentre=phasecentre,
polarisation_frame=PolarisationFrame("stokesI"))
else:
vt = create_visibility(lowcore, times, frequency, channel_bandwidth=channel_bandwidth,
weight=1.0, phasecentre=phasecentre,
polarisation_frame=PolarisationFrame("stokesI"))
cellsize = 0.001
model = create_image_from_visibility(vt, npixel=self.npixel, cellsize=cellsize, npol=1,
frequency=reffrequency, phasecentre=phasecentre,
polarisation_frame=PolarisationFrame("stokesI"))
flux = numpy.array([[100.0]])
facets = 4
rpix = model.wcs.wcs.crpix - 1.0
spacing_pixels = self.npixel // facets
centers = [-1.5, -0.5, 0.5, 1.5]
comps = list()
for iy in centers:
for ix in centers:
p = int(round(rpix[0] + ix * spacing_pixels * numpy.sign(model.wcs.wcs.cdelt[0]))), \
int(round(rpix[1] + iy * spacing_pixels * numpy.sign(model.wcs.wcs.cdelt[1])))
sc = pixel_to_skycoord(p[0], p[1], model.wcs, origin=1)
comp = create_skycomponent(flux=flux, frequency=frequency, direction=sc,
polarisation_frame=PolarisationFrame("stokesI"))
comps.append(comp)
if block:
predict_skycomponent_blockvisibility(vt, comps)
else:
predict_skycomponent_visibility(vt, comps)
insert_skycomponent(model, comps)
self.model = copy_image(model)
self.empty_model = create_empty_image_like(model)
export_image_to_fits(model, '%s/test_pipeline_bags_model.fits' % self.dir)
if add_errors:
# These will be the same for all calls
numpy.random.seed(180555)
gt = create_gaintable_from_blockvisibility(vt)
gt = simulate_gaintable(gt, phase_error=1.0, amplitude_error=0.0)
vt = apply_gaintable(vt, gt)
return vt
def test_invert_2d(self):
# Test if the 2D invert works with w set to zero
# Set w=0 so that the two-dimensional transform should agree exactly with the model.
# Good check on the grid correction in the vis->image direction
self.actualSetUp()
self.componentvis = create_visibility(
self.lowcore,
self.times,
self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre,
weight=1.0,
polarisation_frame=self.vis_pol)
self.componentvis.data['uvw'][:, 2] = 0.0
self.componentvis.data['vis'] *= 0.0
# Predict the visibility using direct evaluation
for comp in self.components:
predict_skycomponent_visibility(self.componentvis, comp)
dirty2d = create_empty_image_like(self.model)
dirty2d, sumwt = invert_2d(self.componentvis, dirty2d, **self.params)
export_image_to_fits(
dirty2d,
'%s/test_imaging_functions_invert_2d_dirty.fits' % self.dir)
self._checkcomponents(dirty2d)
def calibrate_visibility(vt: Visibility,
model: Image = None,
components=None,
predict=predict_2d,
**kwargs) -> Visibility:
""" calibrate Visibility with respect to model and optionally components
:param vt: Visibility
:param model: Model image
:param components: Sky components
:return: Calibrated visibility
"""
assert model is not None or components is not None, "calibration requires a model or skycomponents"
vtpred = copy_visibility(vt, zero=True)
if model is not None:
vtpred = predict(vtpred, model, **kwargs)
if components is not None:
vtpred = predict_skycomponent_visibility(vtpred, components)
else:
vtpred = predict_skycomponent_visibility(vtpred, components)
bvt = decoalesce_visibility(vt)
bvtpred = decoalesce_visibility(vtpred)
gt = solve_gaintable(bvt, bvtpred, **kwargs)
bvt = apply_gaintable(bvt,
gt,
inverse=get_parameter(kwargs, "inverse", False))
return convert_blockvisibility_to_visibility(bvt)
def ingest_visibility(self,
freq=1e8,
chan_width=1e6,
times=None,
reffrequency=None,
add_errors=False):
if times is None:
times = (numpy.pi / 12.0) * numpy.linspace(-3.0, 3.0, 5)
if reffrequency is None:
reffrequency = [1e8]
lowcore = create_named_configuration('LOWBD2-CORE')
frequency = numpy.array([freq])
channel_bandwidth = numpy.array([chan_width])
phasecentre = SkyCoord(ra=+180.0 * u.deg,
dec=-60.0 * u.deg,
frame='icrs',
equinox='J2000')
vt = create_visibility(lowcore,
times,
frequency,
channel_bandwidth=channel_bandwidth,
weight=1.0,
phasecentre=phasecentre,
polarisation_frame=PolarisationFrame("stokesI"))
cellsize = 0.001
model = create_image_from_visibility(
vt,
npixel=self.npixel,
cellsize=cellsize,
npol=1,
frequency=reffrequency,
phasecentre=phasecentre,
polarisation_frame=PolarisationFrame("stokesI"))
flux = numpy.array([[100.0]])
facets = 4
rpix = model.wcs.wcs.crpix
spacing_pixels = self.npixel // facets
centers = [-1.5, -0.5, 0.5, 1.5]
comps = list()
for iy in centers:
for ix in centers:
p = int(round(rpix[0] + ix * spacing_pixels * numpy.sign(model.wcs.wcs.cdelt[0]))), \
int(round(rpix[1] + iy * spacing_pixels * numpy.sign(model.wcs.wcs.cdelt[1])))
sc = pixel_to_skycoord(p[0], p[1], model.wcs, origin=0)
comp = create_skycomponent(
flux=flux,
frequency=frequency,
direction=sc,
polarisation_frame=PolarisationFrame("stokesI"))
comps.append(comp)
predict_skycomponent_visibility(vt, comps)
insert_skycomponent(model, comps)
self.model = copy_image(model)
export_image_to_fits(model,
'%s/test_bags_model.fits' % (self.results_dir))
return vt
def sagecal_fit_gaintable(evis, theta, gain=0.1, niter=3, tol=1e-3, **kwargs):
"""Fit a gaintable to a visibility i.e. A13
This is the update to the gain part of the window
:param evis:
:param theta:
:param kwargs:
:return:
"""
previous_gt = copy_gaintable(theta[1])
gt = copy_gaintable(theta[1])
model_vis = copy_visibility(evis, zero=True)
model_vis = predict_skycomponent_visibility(model_vis, theta[0])
gt = solve_gaintable(evis,
model_vis,
gt=gt,
niter=niter,
phase_only=True,
gain=0.5,
tol=1e-4,
**kwargs)
gt.data['gain'][...] = gain * gt.data['gain'][...] + \
(1 - gain) * previous_gt.data['gain'][...]
gt.data['gain'][...] /= numpy.abs(previous_gt.data['gain'][...])
return gt
def test_phase_rotation_identity(self):
self.vis = create_visibility(
self.lowcore,
self.times,
self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre,
weight=1.0,
polarisation_frame=PolarisationFrame("stokesIQUV"))
self.vismodel = predict_skycomponent_visibility(self.vis, self.comp)
newphasecenters = [
SkyCoord(182, -35, unit=u.deg),
SkyCoord(182, -30, unit=u.deg),
SkyCoord(177, -30, unit=u.deg),
SkyCoord(176, -35, unit=u.deg),
SkyCoord(216, -35, unit=u.deg),
SkyCoord(180, -70, unit=u.deg)
]
for newphasecentre in newphasecenters:
# Phase rotating back should not make a difference
original_vis = self.vismodel.vis
original_uvw = self.vismodel.uvw
rotatedvis = phaserotate_visibility(phaserotate_visibility(
self.vismodel, newphasecentre, tangent=False),
self.phasecentre,
tangent=False)
assert_allclose(rotatedvis.uvw, original_uvw, rtol=1e-7)
assert_allclose(rotatedvis.vis, original_vis, rtol=1e-7)
def test_visibilitysum(self):
self.vis = create_visibility(self.lowcore, self.times, self.frequency,
channel_bandwidth=self.channel_bandwidth, phasecentre=self.phasecentre, weight=1.0,
polarisation_frame=PolarisationFrame("stokesIQUV"))
self.vismodel = predict_skycomponent_visibility(self.vis, self.comp)
# Sum the visibilities in the correct_visibility direction. This is limited by numerical precision
summedflux, weight = sum_visibility(self.vismodel, self.compreldirection)
assert_allclose(self.flux, summedflux, rtol=1e-7)
def test_sum_visibility(self):
self.vis = create_visibility(self.lowcore, self.times, self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre,
polarisation_frame=PolarisationFrame("linear"),
weight=1.0)
self.vis = predict_skycomponent_visibility(self.vis, self.comp)
flux, weight = sum_visibility(self.vis, self.comp.direction)
assert numpy.max(numpy.abs(flux - self.flux)) < 1e-7
def test_qa(self):
self.vis = create_visibility(self.lowcore, self.times, self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre, weight=1.0,
polarisation_frame=PolarisationFrame("stokesIQUV"))
self.vismodel = predict_skycomponent_visibility(self.vis, self.comp)
qa = qa_visibility(self.vis, context='test_qa')
self.assertAlmostEqual(qa.data['maxabs'], 100.0, 7)
self.assertAlmostEqual(qa.data['medianabs'], 11.0, 7)
assert qa.context == 'test_qa'
def test_predict_2d(self):
# Test if the 2D prediction works
#
# Set w=0 so that the two-dimensional transform should agree exactly with the component transform.
# Good check on the grid correction in the image->vis direction
# Set all w to zero
self.actualSetUp()
self.componentvis = create_visibility(
self.lowcore,
self.times,
self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre,
weight=1.0)
self.componentvis.data['uvw'][:, 2] = 0.0
# Predict the visibility using direct evaluation
predict_skycomponent_visibility(self.componentvis, self.components)
self.modelvis = create_visibility(
self.lowcore,
self.times,
self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre,
weight=1.0,
polarisation_frame=self.vis_pol)
self.modelvis.data['uvw'][:, 2] = 0.0
predict_2d(self.modelvis, self.model, **self.params)
self.residualvis = create_visibility(
self.lowcore,
self.times,
self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre,
weight=1.0,
polarisation_frame=self.vis_pol)
self.residualvis.data['uvw'][:, 2] = 0.0
self.residualvis.data[
'vis'] = self.modelvis.data['vis'] - self.componentvis.data['vis']
self._checkdirty(self.residualvis,
'test_predict_2d',
fluxthreshold=4.0)
def test_insert_skycomponent_dft(self):
sc = create_skycomponent(
direction=self.phasecentre,
flux=numpy.array([[1.0]]),
frequency=self.frequency,
polarisation_frame=PolarisationFrame('stokesI'))
self.vis.data['vis'][...] = 0.0
self.vis = predict_skycomponent_visibility(self.vis, sc)
im, sumwt = invert_2d(self.vis, self.model)
export_image_to_fits(im, '%s/test_skycomponent_dft.fits' % self.dir)
assert numpy.max(numpy.abs(self.vis.vis.imag)) < 1e-3
def test_invert_2d(self):
# Test if the 2D invert works with w set to zero
# Set w=0 so that the two-dimensional transform should agree exactly with the model.
# Good check on the grid correction in the vis->image direction
self.actualSetUp()
self.componentvis.data['uvw'][:, 2] = 0.0
self.componentvis.data['vis'] *= 0.0
# Predict the visibility using direct evaluation
for comp in self.components:
predict_skycomponent_visibility(self.componentvis, comp)
dirty2d = create_empty_image_like(self.model)
dirty2d, sumwt = invert_2d(self.componentvis, dirty2d, **self.params)
export_image_to_fits(
dirty2d,
'%s/test_imaging_functions_invert_2d_dirty.fits' % self.dir)
self._checkcomponents(dirty2d)
def make_e(vis, calskymodel, evis_all):
# Return the estep for a given skymodel
evis = copy_visibility(vis)
tvis = copy_visibility(vis, zero=True)
tvis = predict_skycomponent_visibility(tvis, calskymodel[0].components)
tvis = apply_gaintable(tvis, calskymodel[1])
# E step is the data model for a window plus the difference between the observed data
# and the summed data models or, put another way, its the observed data minus the
# summed visibility for all other windows
evis.data['vis'][...] = tvis.data['vis'][...] + vis.data['vis'][
...] - evis_all.data['vis'][...]
return evis
def test_phase_rotation_inverse(self):
self.vis = create_visibility(self.lowcore, self.times, self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre, weight=1.0,
polarisation_frame=PolarisationFrame("stokesIQUV"))
self.vismodel = predict_skycomponent_visibility(self.vis, self.comp)
there = SkyCoord(ra=+250.0 * u.deg, dec=-60.0 * u.deg, frame='icrs', equinox='J2000')
# Phase rotating back should not make a difference
original_vis = self.vismodel.vis
original_uvw = self.vismodel.uvw
rotatedvis = phaserotate_visibility(phaserotate_visibility(self.vismodel, there, tangent=False,
inverse=True),
self.phasecentre, tangent=False, inverse=True)
assert_allclose(rotatedvis.uvw, original_uvw, rtol=1e-7)
assert_allclose(rotatedvis.vis, original_vis, rtol=1e-7)
def skymodel_cal_e_step(vis: BlockVisibility, evis_all: BlockVisibility, calskymodel, **kwargs):
"""Calculates E step in equation A12
This is the data model for this window plus the difference between observed data and summed data models
:param evis_all: Sum data models
:param csm: csm element being fit
:param kwargs:
:return: Data model (i.e. visibility) for this csm
"""
evis = copy_visibility(evis_all)
tvis = copy_visibility(vis, zero=True)
tvis = predict_skycomponent_visibility(tvis, calskymodel[0].components)
tvis = apply_gaintable(tvis, calskymodel[1])
evis.data['vis'][...] = tvis.data['vis'][...] + vis.data['vis'][...] - evis_all.data['vis'][...]
return evis
def spectral_line_imaging(vis: Visibility,
model: Image,
continuum_model: Image = None,
continuum_components=None,
predict=predict_2d,
invert=invert_2d,
deconvolve_spectral=False,
**kwargs) -> (Image, Image, Image):
"""Spectral line imaging from calibrated (DIE) data
A continuum model can be subtracted, and deconvolution is optional.
If deconvolve_spectral is True then the solve_image is used to deconvolve.
If deconvolve_spectral is False then the residual image after continuum subtraction is calculated
:param vis: Visibility
:param continuum_model: model continuum image to be subtracted
:param continuum_components: mode components to be subtracted
:param spectral_model: model spectral image
:param predict: Predict fumction e.g. predict_2d
:param invert: Invert function e.g. invert_wprojection
:return: Residual visibility, spectral model image, spectral residual image
"""
vis_no_continuum = copy_visibility(vis)
if continuum_model is not None:
vis_no_continuum = predict(vis_no_continuum, continuum_model, **kwargs)
if continuum_components is not None:
vis_no_continuum = predict_skycomponent_visibility(
vis_no_continuum, continuum_components)
vis_no_continuum.data[
'vis'] = vis.data['vis'] - vis_no_continuum.data['vis']
if deconvolve_spectral:
log.info(
"spectral_line_imaging: Deconvolving continuum subtracted visibility"
)
vis_no_continuum, spectral_model, spectral_residual = solve_image(
vis_no_continuum, model, **kwargs)
else:
log.info(
"spectral_line_imaging: Making dirty image from continuum subtracted visibility"
)
spectral_model, spectral_residual = \
invert(vis_no_continuum, model, **kwargs)
return vis_no_continuum, spectral_model, spectral_residual
def test_fit_visibility(self):
# Sum the visibilities in the correct_visibility direction. This is limited by numerical precision
methods = ['CG', 'BFGS', 'Powell', 'trust-ncg', 'trust-exact', 'trust-krylov']
for method in methods:
self.actualSetup()
self.vis = create_visibility(self.lowcore, self.times, self.frequency,
channel_bandwidth=self.channel_bandwidth,
phasecentre=self.phasecentre, weight=1.0,
polarisation_frame=PolarisationFrame("stokesI"))
self.vismodel = predict_skycomponent_visibility(self.vis, self.comp)
initial_comp = Skycomponent(direction=self.comp_start_direction, frequency=self.frequency,
flux=2.0 * self.flux, polarisation_frame=PolarisationFrame("stokesI"))
sc, res = fit_visibility(self.vismodel, initial_comp, niter=200, tol=1e-5, method=method, verbose=False)
# print(method, res)
assert sc.direction.separation(self.comp_actual_direction).to('rad').value < 1e-6, \
sc.direction.separation(self.comp_actual_direction).to('rad')
def actualSetup(self,
sky_pol_frame='stokesIQUV',
data_pol_frame='linear',
f=None,
vnchan=3):
self.lowcore = create_named_configuration('LOWBD2-CORE')
self.times = (numpy.pi / 43200.0) * numpy.linspace(0.0, 30.0, 3)
self.frequency = numpy.linspace(1.0e8, 1.1e8, vnchan)
self.channel_bandwidth = numpy.array(
vnchan * [self.frequency[1] - self.frequency[0]])
if f is None:
f = [100.0, 50.0, -10.0, 40.0]
if sky_pol_frame == 'stokesI':
f = [100.0]
self.flux = numpy.outer(
numpy.array(
[numpy.power(freq / 1e8, -0.7) for freq in self.frequency]), f)
# The phase centre is absolute and the component is specified relative (for now).
# This means that the component should end up at the position phasecentre+compredirection
self.phasecentre = SkyCoord(ra=+180.0 * u.deg,
dec=-35.0 * u.deg,
frame='icrs',
equinox='J2000')
self.compabsdirection = SkyCoord(ra=+181.0 * u.deg,
dec=-35.0 * u.deg,
frame='icrs',
equinox='J2000')
self.comp = Skycomponent(
direction=self.compabsdirection,
frequency=self.frequency,
flux=self.flux,
polarisation_frame=PolarisationFrame(sky_pol_frame))
self.vis = create_blockvisibility(
self.lowcore,
self.times,
self.frequency,
phasecentre=self.phasecentre,
channel_bandwidth=self.channel_bandwidth,
weight=1.0,
polarisation_frame=PolarisationFrame(data_pol_frame))
self.vis = predict_skycomponent_visibility(self.vis, self.comp)
def sagecal_e_all(vis: BlockVisibility, thetas, **kwargs):
"""Calculates E step in equation A12
This is the sum of the data models over all windows
:param vis:
:param thetas:
:param kwargs:
:return:
"""
evis = copy_visibility(vis, zero=True)
tvis = copy_visibility(vis, zero=True)
for i, theta in enumerate(thetas):
tvis.data['vis'][...] = 0.0
tvis = predict_skycomponent_visibility(tvis, theta[0])
tvis = apply_gaintable(tvis, theta[1])
evis.data['vis'][...] += tvis.data['vis'][...]
return evis
def skymodel_cal_e_all(vis: BlockVisibility, calskymodels):
"""Calculates E step in equation A12
This is the sum of the data models over all skymodel
:param vis: Visibility
:param csm: List of (skymodel, gaintable) tuples
:param kwargs:
:return: Sum of data models (i.e. a visibility)
"""
evis = copy_visibility(vis, zero=True)
tvis = copy_visibility(vis, zero=True)
for csm in calskymodels:
tvis.data['vis'][...] = 0.0
tvis = predict_skycomponent_visibility(tvis, csm[0].components)
tvis = apply_gaintable(tvis, csm[1])
evis.data['vis'][...] += tvis.data['vis'][...]
return evis
def sagecal_e_step(vis: BlockVisibility,
evis_all: BlockVisibility,
theta,
beta=1.0,
**kwargs):
"""Calculates E step in equation A12
This is the data model for this window plus the difference between observed data and summed data models
:param vis:
:param theta:
:param kwargs:
:return:
"""
evis = copy_visibility(evis_all)
tvis = copy_visibility(vis, zero=True)
tvis = predict_skycomponent_visibility(tvis, theta[0])
tvis = apply_gaintable(tvis, theta[1])
evis.data['vis'][...] = tvis.data['vis'][...] + \
beta * (vis.data['vis'][...] - evis_all.data['vis'][...])
return evis
def skymodel_cal_fit_gaintable(evis, calskymodel, gain=0.1, niter=3, tol=1e-3, **kwargs):
"""Fit a gaintable to a visibility
This is the update to the gain part of the window
:param evis: Expected vis for this ssm
:param calskymodel: csm element being fit
:param gain: Gain in step
:param niter: Number of iterations
:param kwargs: Gaintable
"""
previous_gt = copy_gaintable(calskymodel[1])
gt = copy_gaintable(calskymodel[1])
model_vis = copy_visibility(evis, zero=True)
model_vis = predict_skycomponent_visibility(model_vis, calskymodel[0].components)
gt = solve_gaintable(evis, model_vis, gt=gt, niter=niter, phase_only=True, gain=0.5,
tol=1e-4, **kwargs)
gt.data['gain'][...] = gain * gt.data['gain'][...] + \
(1 - gain) * previous_gt.data['gain'][...]
gt.data['gain'][...] /= numpy.abs(previous_gt.data['gain'][...])
return gt
def spectral_line_imaging(vis: BlockVisibility,
model: Image,
continuum_model: Image = None,
continuum_components=None,
context='2d',
**kwargs) -> (Image, Image, Image):
"""Spectral line imaging from calibrated (DIE) data
A continuum model can be subtracted, and the residual image deconvolved.
:param vis: Visibility
:param model: Image specify details of model
:param continuum_model: model continuum image to be subtracted
:param continuum_components: mode components to be subtracted
:param spectral_model: model spectral image
:param predict: Predict fumction e.g. predict_2d
:param invert: Invert function e.g. invert_wprojection
:return: Residual visibility, spectral model image, spectral residual image
"""
vis_no_continuum = copy_visibility(vis)
if continuum_model is not None:
vis_no_continuum = predict_function(vis_no_continuum,
continuum_model,
context=context,
**kwargs)
if continuum_components is not None:
vis_no_continuum = predict_skycomponent_visibility(
vis_no_continuum, continuum_components)
vis_no_continuum.data[
'vis'] = vis.data['vis'] - vis_no_continuum.data['vis']
log.info(
"spectral_line_imaging: Deconvolving continuum subtracted visibility")
return ical(vis_no_continuum.data,
model,
components=None,
context=context,
do_selfcal=False,
**kwargs)
def test_predict_2d(self):
# Test if the 2D prediction works
#
# Set w=0 so that the two-dimensional transform should agree exactly with the component transform.
# Good check on the grid correction in the image->vis direction
# Set all w to zero
self.actualSetUp()
self.componentvis.data['uvw'][:, 2] = 0.0
# Predict the visibility using direct evaluation
self.componentvis.data['vis'][...] = 0.0
self.componentvis = predict_skycomponent_visibility(
self.componentvis, self.components)
self.modelvis = copy_visibility(self.componentvis, zero=True)
self.modelvis.data['uvw'][:, 2] = 0.0
self.modelvis = predict_2d(self.modelvis, self.model, **self.params)
self.residualvis = copy_visibility(self.componentvis, zero=True)
self.residualvis.data['uvw'][:, 2] = 0.0
self.residualvis.data[
'vis'] = self.modelvis.data['vis'] - self.componentvis.data['vis']
self._checkdirty(self.residualvis, 'predict_2d')
def rcal(vis: BlockVisibility, components, **kwargs) -> GainTable:
""" Real-time calibration pipeline.
Reads visibilities through a BlockVisibility iterator, calculates model visibilities according to a
component-based sky model, and performs calibration solution, writing a gaintable for each chunk of
visibilities.
:param vis: Visibility or Union(Visibility, Iterable)
:param components: Component-based sky model
:param kwargs: Parameters
:return: gaintable
"""
if not isinstance(vis, collections.Iterable):
vis = [vis]
from arl.calibration.solvers import solve_gaintable
for ichunk, vischunk in enumerate(vis):
vispred = copy_visibility(vischunk, zero=True)
vispred = predict_skycomponent_visibility(vispred, components)
gt = solve_gaintable(vischunk, vispred, **kwargs)
yield gt
def actualSetup(self, sky_pol_frame='stokesIQUV', data_pol_frame='linear'):
self.lowcore = create_named_configuration('LOWBD2-CORE')
self.times = (numpy.pi / 43200.0) * numpy.arange(0.0, 300.0, 30.0)
vnchan = 3
self.frequency = numpy.linspace(1.0e8, 1.1e8, vnchan)
self.channel_bandwidth = numpy.array(
vnchan * [self.frequency[1] - self.frequency[0]])
# Define the component and give it some spectral behaviour
f = numpy.array([100.0, 20.0, -10.0, 1.0])
self.flux = numpy.array([f, 0.8 * f, 0.6 * f])
self.phasecentre = SkyCoord(ra=+180.0 * u.deg,
dec=-35.0 * u.deg,
frame='icrs',
equinox='J2000')
self.compabsdirection = SkyCoord(ra=+181.0 * u.deg,
dec=-35.0 * u.deg,
frame='icrs',
equinox='J2000')
if sky_pol_frame == 'stokesI':
self.flux = self.flux[:, 0][:, numpy.newaxis]
self.comp = Skycomponent(
direction=self.compabsdirection,
frequency=self.frequency,
flux=self.flux,
polarisation_frame=PolarisationFrame(sky_pol_frame))
self.vis = create_blockvisibility(
self.lowcore,
self.times,
self.frequency,
phasecentre=self.phasecentre,
channel_bandwidth=self.channel_bandwidth,
weight=1.0,
polarisation_frame=PolarisationFrame(data_pol_frame))
self.vis = predict_skycomponent_visibility(self.vis, self.comp)
def create_blockvisibility_iterator(
config: Configuration,
times: numpy.array,
frequency: numpy.array,
channel_bandwidth,
phasecentre: SkyCoord,
weight: float = 1,
polarisation_frame=PolarisationFrame('stokesI'),
integration_time=1.0,
number_integrations=1,
predict=predict_timeslice,
model=None,
components=None,
phase_error=0.0,
amplitude_error=0.0,
sleep=0.0,
**kwargs):
""" Create a sequence of Visibilities and optionally predicting and coalescing
This is useful mainly for performing large simulations. Do something like::
vis_iter = create_blockvisibility_iterator(config, times, frequency, channel_bandwidth, phasecentre=phasecentre,
weight=1.0, integration_time=30.0, number_integrations=3)
for i, vis in enumerate(vis_iter):
if i == 0:
fullvis = vis
else:
fullvis = append_visibility(fullvis, vis)
:param config: Configuration of antennas
:param times: hour angles in radians
:param frequency: frequencies (Hz] Shape [nchan]
:param weight: weight of a single sample
:param phasecentre: phasecentre of observation
:param npol: Number of polarizations
:param integration_time: Integration time ('auto' or value in s)
:param number_integrations: Number of integrations to be created at each time.
:param model: Model image to be inserted
:param components: Components to be inserted
:param sleep_time: Time to sleep between yields
:return: Visibility
"""
for time in times:
actualtimes = time + numpy.arange(
0, number_integrations) * integration_time * numpy.pi / 43200.0
bvis = create_blockvisibility(config,
actualtimes,
frequency=frequency,
phasecentre=phasecentre,
weight=weight,
polarisation_frame=polarisation_frame,
integration_time=integration_time,
channel_bandwidth=channel_bandwidth)
if model is not None:
vis = predict(bvis, model, **kwargs)
bvis = convert_visibility_to_blockvisibility(vis)
if components is not None:
bvis = predict_skycomponent_visibility(bvis, components)
# Add phase errors
if phase_error > 0.0 or amplitude_error > 0.0:
gt = create_gaintable_from_blockvisibility(bvis)
gt = simulate_gaintable(gt=gt,
phase_error=phase_error,
amplitude_error=amplitude_error)
bvis = apply_gaintable(bvis, gt)
import time
time.sleep(sleep)
yield bvis
def test_peel_skycomponent_blockvisibility(self):
df = 1e6
frequency = numpy.array([1e8 - df, 1e8, 1e8 + df])
channel_bandwidth = numpy.array([df, df, df])
# Define the component and give it some spectral behaviour
f = numpy.array([100.0, 20.0, -10.0, 1.0])
flux = numpy.array([f, 0.8 * f, 0.6 * f])
phasecentre = SkyCoord(0 * u.deg, -60.0 * u.deg)
config = create_named_configuration('LOWBD2-CORE')
peeldirection = SkyCoord(+15 * u.deg, -60.0 * u.deg)
times = numpy.linspace(-3.0, 3.0, 7) * numpy.pi / 12.0
# Make the visibility
vis = create_blockvisibility(
config,
times,
frequency,
phasecentre=phasecentre,
weight=1.0,
polarisation_frame=PolarisationFrame('linear'),
channel_bandwidth=channel_bandwidth)
vis.data['vis'][...] = 0.0
# First add in the source to be peeled.
peel = Skycomponent(direction=peeldirection,
frequency=frequency,
flux=flux,
polarisation_frame=PolarisationFrame("stokesIQUV"))
vis = predict_skycomponent_visibility(vis, peel)
# Make a gaintable and apply it to the visibility of the peeling source
gt = create_gaintable_from_blockvisibility(vis, timeslice='auto')
gt = simulate_gaintable(gt,
phase_error=0.01,
amplitude_error=0.01,
timeslice='auto')
gt.data['gain'] *= 0.3
vis = apply_gaintable(vis, gt, timeslice='auto')
# Now create a plausible field using the GLEAM sources
model = create_image_from_visibility(
vis,
cellsize=0.001,
frequency=frequency,
polarisation_frame=PolarisationFrame('stokesIQUV'))
bm = create_low_test_beam(model=model)
sc = create_low_test_skycomponents_from_gleam(
flux_limit=1.0,
polarisation_frame=PolarisationFrame("stokesIQUV"),
frequency=frequency,
kind='cubic',
phasecentre=phasecentre,
radius=0.1)
sc = apply_beam_to_skycomponent(sc, bm)
# Add in the visibility due to these sources
vis = predict_skycomponent_visibility(vis, sc)
assert numpy.max(numpy.abs(vis.vis)) > 0.0
# Now we can peel
vis, peel_gts = peel_skycomponent_blockvisibility(vis, peel)
assert len(peel_gts) == 1
residual = numpy.max(peel_gts[0].residual)
assert residual < 0.7, "Peak residual %.6f too large" % (residual)
im, sumwt = invert_timeslice(vis, model, timeslice='auto')
qa = qa_image(im)
assert numpy.abs(qa.data['max'] - 14.2) < 1.0, str(qa)
def ical(block_vis: BlockVisibility,
model: Image,
components=None,
context='2d',
controls=None,
**kwargs):
""" Post observation image, deconvolve, and self-calibrate
:param vis:
:param model: Model image
:param components: Initial components
:param context: Imaging context
:param controls: Calibration controls dictionary
:return: model, residual, restored
"""
nmajor = get_parameter(kwargs, 'nmajor', 5)
log.info("ical: Performing %d major cycles" % nmajor)
do_selfcal = get_parameter(kwargs, "do_selfcal", False)
if controls is None:
controls = create_calibration_controls(**kwargs)
# The model is added to each major cycle and then the visibilities are
# calculated from the full model
vis = convert_blockvisibility_to_visibility(block_vis)
block_vispred = copy_visibility(block_vis, zero=True)
vispred = convert_blockvisibility_to_visibility(block_vispred)
vispred.data['vis'][...] = 0.0
visres = copy_visibility(vispred)
vispred = predict_function(vispred, model, context=context, **kwargs)
if components is not None:
vispred = predict_skycomponent_visibility(vispred, components)
if do_selfcal:
vis, gaintables = calibrate_function(vis,
vispred,
'TGB',
controls,
iteration=-1)
visres.data['vis'] = vis.data['vis'] - vispred.data['vis']
dirty, sumwt = invert_function(visres, model, context=context, **kwargs)
log.info("Maximum in residual image is %.6f" %
(numpy.max(numpy.abs(dirty.data))))
psf, sumwt = invert_function(visres,
model,
dopsf=True,
context=context,
**kwargs)
thresh = get_parameter(kwargs, "threshold", 0.0)
for i in range(nmajor):
log.info("ical: Start of major cycle %d of %d" % (i, nmajor))
cc, res = deconvolve_cube(dirty, psf, **kwargs)
model.data += cc.data
vispred.data['vis'][...] = 0.0
vispred = predict_function(vispred, model, context=context, **kwargs)
if do_selfcal:
vis, gaintables = calibrate_function(vis,
vispred,
'TGB',
controls,
iteration=i)
visres.data['vis'] = vis.data['vis'] - vispred.data['vis']
dirty, sumwt = invert_function(visres,
model,
context=context,
**kwargs)
log.info("Maximum in residual image is %s" %
(numpy.max(numpy.abs(dirty.data))))
if numpy.abs(dirty.data).max() < 1.1 * thresh:
log.info("ical: Reached stopping threshold %.6f Jy" % thresh)
break
log.info("ical: End of major cycle")
log.info("ical: End of major cycles")
restored = restore_cube(model, psf, dirty, **kwargs)
return model, dirty, restored