Add self cal algorithm and example script

Author: matmanc

lofarimaging/lofarimaging.py

@@ -1,12 +1,14 @@
 """Functions for working with LOFAR single station data"""
 
+import logging
 from typing import Dict, List
-import numpy as np
-from numpy.linalg import norm, lstsq
-import numexpr as ne
+
 import numba
+import numexpr as ne
+import numpy as np
+import scipy.optimize
 from astropy.coordinates import SkyCoord, SkyOffsetFrame, CartesianRepresentation
-
+from numpy.linalg import norm, lstsq
 
 __all__ = ["nearfield_imager", "sky_imager", "ground_imager", "skycoord_to_lmn", "calibrate", "simulate_sky_source",
            "subtract_sources"]
@@ -170,6 +172,70 @@ def calibrate(vis, modelvis, maxiter=30, amplitudeonly=True):
     return residual, gains
 
 
+def compute_calibrated_model(vis, model_vis, maxiter=30):
+    n_ant = vis.shape[0]
+    gain_phase = np.zeros((n_ant), dtype=complex)
+
+    def gains_difference(gain_phase, vis, model_vis):
+        gains = np.exp(1.j * gain_phase)
+        gains_mat = np.diag(gains)
+        gains_mat_star = np.diag(np.conj(gains))
+        predicted = gains_mat_star @ model_vis @ gains_mat
+        residual = np.sum(np.abs(vis - predicted))
+        return residual
+
+    result = scipy.optimize.minimize(gains_difference, gain_phase, args=(vis, model_vis), options={'maxiter': maxiter})
+    gain_phase = result.x
+    gains_mat = np.diag(np.exp(1.j * gain_phase))
+
+    calibrated = np.conj(gains_mat) @ model_vis @ gains_mat
+
+    return calibrated, gains_difference(gain_phase, vis, model_vis)
+
+
+def compute_pointing_matrix(sources_positions, baselines, frequency):
+    n_baselines = baselines.shape[0]
+    n_sources = sources_positions.shape[0]
+
+    pointing_matrix = np.zeros(shape=(n_baselines, n_baselines, n_sources), dtype=complex)
+
+    for q in range(n_sources):
+        pointing_matrix[:, :, q] = simulate_sky_source(sources_positions[q, :], baselines, frequency)
+    return pointing_matrix
+
+
+def estimate_sources_flux(visibilities, pointing_matrix):
+    def residual_flux(signal, vis, point_matrix):
+        residual = vis - point_matrix @ signal
+        residual = np.sum(np.abs(residual))
+        return residual
+
+    n_antennas = visibilities.shape[0]
+    n_sources = pointing_matrix.shape[2]
+    lin_visibilities = visibilities.reshape((n_antennas * n_antennas))
+    lin_pointing_matrix = pointing_matrix.reshape((n_antennas * n_antennas, n_sources))
+    fluxes, *_ = np.linalg.lstsq(lin_pointing_matrix, lin_visibilities)
+
+    result = scipy.optimize.minimize(residual_flux, fluxes, args=(visibilities, pointing_matrix))
+    return result.x
+
+
+def estimate_model_visibilities(sources_positions, visibilities, baselines, frequency):
+    pointing_matrix = compute_pointing_matrix(sources_positions, baselines, frequency)
+    fluxes = estimate_sources_flux(visibilities, pointing_matrix)
+    model = pointing_matrix @ fluxes
+
+    return model
+
+
+def self_cal(visibilities, expected_sources, baselines, frequency, iterations=10):
+    model = estimate_model_visibilities(expected_sources, visibilities, baselines, frequency)
+    for i in range(iterations):
+        model, residual = compute_calibrated_model(visibilities, model, maxiter=50)
+        logging.debug('residual after iteration %d: %s', i, residual)
+    return model
+
+
 def simulate_sky_source(lmn_coord: np.array, baselines: np.array, freq: float):
     """
     Simulate visibilities for a sky source

lofarimaging/singlestationutil.py

@@ -1,36 +1,32 @@
 """Functions for working with LOFAR single station data"""
 
-import os
 import datetime
+import os
 from typing import List, Dict, Tuple, Union
 
-import numpy as np
-from packaging import version
-import tqdm
+import astropy.units as u
 import h5py
-
-import matplotlib.pyplot as plt
+import lofarantpos
+import lofargeotiff
 import matplotlib.animation
-from matplotlib.ticker import FormatStrFormatter
-from matplotlib import cm
-from matplotlib.figure import Figure
-from matplotlib.colors import ListedColormap, Normalize
-from matplotlib.patches import Circle
 import matplotlib.axes as maxes
-from mpl_toolkits.axes_grid1 import make_axes_locatable
-
+import matplotlib.pyplot as plt
+import numpy as np
+import tqdm
 from astropy.coordinates import SkyCoord, GCRS, EarthLocation, AltAz, get_sun, get_moon
-import astropy.units as u
 from astropy.time import Time
-
-import lofargeotiff
 from lofarantpos.db import LofarAntennaDatabase
-import lofarantpos
+from matplotlib import cm
+from matplotlib.colors import ListedColormap, Normalize
+from matplotlib.figure import Figure
+from matplotlib.patches import Circle
+from matplotlib.ticker import FormatStrFormatter
+from mpl_toolkits.axes_grid1 import make_axes_locatable
+from packaging import version
 
-from .maputil import get_map, make_leaflet_map
-from .lofarimaging import nearfield_imager, sky_imager, skycoord_to_lmn, subtract_sources
 from .hdf5util import write_hdf5
-
+from .lofarimaging import nearfield_imager, sky_imager, skycoord_to_lmn, subtract_sources
+from .maputil import get_map, make_leaflet_map
 
 __all__ = ["sb_from_freq", "freq_from_sb", "find_caltable", "read_caltable",
            "rcus_in_station", "read_acm_cube", "get_station_pqr", "get_station_type",
@@ -331,7 +327,7 @@ def get_station_pqr(station_name: str, rcu_mode: Union[str, int], db):
             'intl': GENERIC_INT_201512, 'remote': GENERIC_REMOTE_201512, 'core': GENERIC_CORE_201512
         }
         selected_dipoles = selected_dipole_config[station_type] + \
-            np.arange(len(selected_dipole_config[station_type])) * 16
+                           np.arange(len(selected_dipole_config[station_type])) * 16
         station_pqr = db.hba_dipole_pqr(full_station_name)[selected_dipoles]
     else:
         raise RuntimeError("Station name did not contain LBA or HBA, could not load antenna positions")
@@ -340,7 +336,7 @@ def get_station_pqr(station_name: str, rcu_mode: Union[str, int], db):
 
 
 def make_ground_plot(image: np.ndarray, background_map: np.ndarray, extent: List[float], title: str = "Ground plot",
-        subtitle: str = "", opacity: float = 0.6, fig: Figure = None, draw_contours: bool = True, **kwargs) \
+                     subtitle: str = "", opacity: float = 0.6, fig: Figure = None, draw_contours: bool = True, **kwargs) \
         -> Tuple[Figure, np.ndarray]:
     """
     Make a ground plot of an array with data
@@ -585,7 +581,6 @@ def make_xst_plots(xst_data: np.ndarray,
         opacity: Opacity for map overlay. Defaults to 0.6.
         hdf5_filename: Filename where hdf5 results can be written. Defaults to outputpath + '/results.h5'
         outputpath: Directory where results can be saved. Defaults to 'results'
-        subtract: List of sources to subtract. Defaults to None
 
 
     Returns:
@@ -662,14 +657,14 @@ def make_xst_plots(xst_data: np.ndarray,
     marked_bodies = {
         'Cas A': SkyCoord(ra=350.85 * u.deg, dec=58.815 * u.deg),
         'Cyg A': SkyCoord(ra=299.86815191 * u.deg, dec=40.73391574 * u.deg),
-        'Per A': SkyCoord(ra=49.95066567*u.deg, dec=41.51169838 * u.deg),
-        'Her A': SkyCoord(ra=252.78343333*u.deg, dec=4.99303056*u.deg),
-        'Cen A': SkyCoord(ra=201.36506288*u.deg, dec=-43.01911267*u.deg),
-        'Vir A': SkyCoord(ra=187.70593076*u.deg, dec=12.39112329*u.deg),
-        '3C295': SkyCoord(ra=212.83527917*u.deg, dec=52.20264444*u.deg),
+        'Per A': SkyCoord(ra=49.95066567 * u.deg, dec=41.51169838 * u.deg),
+        'Her A': SkyCoord(ra=252.78343333 * u.deg, dec=4.99303056 * u.deg),
+        'Cen A': SkyCoord(ra=201.36506288 * u.deg, dec=-43.01911267 * u.deg),
+        'Vir A': SkyCoord(ra=187.70593076 * u.deg, dec=12.39112329 * u.deg),
+        '3C295': SkyCoord(ra=212.83527917 * u.deg, dec=52.20264444 * u.deg),
         'Moon': get_moon(obstime_astropy, location=station_earthlocation).transform_to(GCRS),
         'Sun': get_sun(obstime_astropy),
-        '3C196': SkyCoord(ra=123.40023371*u.deg, dec=48.21739888*u.deg)
+        '3C196': SkyCoord(ra=123.40023371 * u.deg, dec=48.21739888 * u.deg)
     }
 
     marked_bodies_lmn = {}
@@ -752,7 +747,7 @@ def make_xst_plots(xst_data: np.ndarray,
             "pixels_per_metre": pixels_per_metre}
     tags.update(calibration_info)
     lon_min, lon_max, lat_min, lat_max = extent_lonlat
-    lofargeotiff.write_geotiff(ground_img[::-1,:], os.path.join(outputpath, f"{fname}_nearfield_calibrated.tiff"),
+    lofargeotiff.write_geotiff(ground_img[::-1, :], os.path.join(outputpath, f"{fname}_nearfield_calibrated.tiff"),
                                (lon_min, lat_max), (lon_max, lat_min), as_pqr=False,
                                stationname=station_name, obsdate=obstime, tags=tags)
 
@@ -769,7 +764,7 @@ def make_sky_movie(moviefilename: str, h5file: h5py.File, obsnums: List[str], vm
     """
     Make movie of a list of observations
     """
-    fig = plt.figure(figsize=(10,10))
+    fig = plt.figure(figsize=(10, 10))
     for obsnum in tqdm.tqdm(obsnums):
         obs_h5 = h5file[obsnum]
         skydata_h5 = obs_h5["sky_img"]
@@ -787,7 +782,28 @@ def make_sky_movie(moviefilename: str, h5file: h5py.File, obsnums: List[str], vm
 
     # Thanks to Maaijke Mevius for making this animation work!
     ims = fig.get_children()[1:]
-    ims = [ims[i:i+2] for i in range(0, len(ims), 2)]
+    ims = [ims[i:i + 2] for i in range(0, len(ims), 2)]
     ani = matplotlib.animation.ArtistAnimation(fig, ims, interval=30, blit=False, repeat_delay=1000)
     writer = matplotlib.animation.writers['ffmpeg'](fps=5, bitrate=800)
     ani.save(moviefilename, writer=writer, dpi=fig.dpi)
+
+
+def compute_baselines(station_name, rcu_mode):
+    # Setup the database
+    db = LofarAntennaDatabase()
+
+    station_pqr = get_station_pqr(station_name, rcu_mode, db)
+
+    station_name = get_full_station_name(station_name, rcu_mode)
+
+    # Rotate station_pqr to a north-oriented xyz frame, where y points North, in a plane through the station.
+    rotation = db.rotation_from_north(station_name)
+
+    pqr_to_xyz = np.array([[np.cos(-rotation), -np.sin(-rotation), 0],
+                           [np.sin(-rotation), np.cos(-rotation), 0],
+                           [0, 0, 1]])
+
+    station_xyz = (pqr_to_xyz @ station_pqr.T).T
+
+    baselines = station_xyz[:, np.newaxis, :] - station_xyz[np.newaxis, :, :]
+    return baselines

self_cal_demo.py

@@ -0,0 +1,66 @@
+import h5py
+import numpy
+import lofarimaging.hdf5util as h5utils
+from lofarimaging.singlestationutil import compute_baselines, make_sky_plot
+from lofarimaging.lofarimaging import compute_pointing_matrix, compute_calibrated_model, sky_imager,\
+    self_cal, estimate_model_visibilities
+import matplotlib.pyplot as plt
+
+test_data_path = '/home/mmancini/Documents/Projects/SingleStationImager/test_dataset.h5'
+test_station = 'CS103'
+test_data = h5py.File(test_data_path, 'r')
+obs_names = h5utils.get_obsnums(test_data, station_name=test_station)
+an_obs = test_data[obs_names[10]]
+
+a_frequency = an_obs.attrs['frequency']
+a_source_lmn = an_obs.attrs['source_lmn']
+a_source_name = an_obs.attrs['source_names']
+a_rcu_mode = an_obs.attrs['rcu_mode']
+a_dataset = an_obs['calibrated_data']
+
+a_dataset_xx = a_dataset[::2, ::2]
+a_dataset_yy = a_dataset[1::2, 1::2]
+
+a_dataset_i = a_dataset_xx + a_dataset_yy
+a_dataset_q = a_dataset_xx - a_dataset_yy
+a_dataset_u = 2 * (a_dataset_xx * a_dataset_yy.conj()).real
+a_dataset_v = -2 * (a_dataset_xx * a_dataset_yy.conj()).imag
+
+baselines = compute_baselines(test_station, a_rcu_mode)
+sky_image = sky_imager(a_dataset_i, baselines, a_frequency, 100, 100)
+
+model = estimate_model_visibilities(a_source_lmn, a_dataset_i, baselines, a_frequency)
+
+model_image = sky_imager(model, baselines, a_frequency, 100, 100)
+
+pointing_matrix = compute_pointing_matrix(a_source_lmn, baselines, a_frequency)
+calibrated_model, _ = compute_calibrated_model(a_dataset_i, model, maxiter=50)
+self_cal_model = self_cal(a_dataset_i, a_source_lmn, baselines, a_frequency)
+
+sky_without_calibrated_model = sky_imager(a_dataset_i - self_cal_model, baselines, a_frequency, 100, 100)
+calibrated_model_image = sky_imager(calibrated_model, baselines, a_frequency, 100, 100)
+sky_without_model = sky_imager(a_dataset_i - model, baselines, a_frequency, 100, 100)
+
+
+f, (ax1, ax2, ax3, ax4) = plt.subplots(1, 4)
+min, max = numpy.min(sky_image), numpy.max(sky_image)
+
+ax1.scatter(a_source_lmn[:, 0], a_source_lmn[:, 1], marker='o', color='r', s=100, facecolors='none')
+im = ax1.imshow(sky_image[::-1,:], vmin=min, vmax=max, extent=(1, -1, -1, 1))
+ax1.set_title('initial sky')
+
+ax2.scatter(a_source_lmn[:, 0], a_source_lmn[:, 1], marker='o', color='r', s=100, facecolors='none')
+ax2.imshow(sky_without_model[::-1, :], extent=(1, -1, -1, 1), vmin=min, vmax=max)
+ax2.set_title('sky without model')
+
+ax3.scatter(a_source_lmn[:, 0], a_source_lmn[:, 1], marker='o', color='r', s=100, facecolors='none')
+ax3.imshow(calibrated_model_image[::-1, :], extent=(1, -1, -1, 1))
+ax3.set_title('calibrated model image')
+
+ax4.imshow(sky_without_calibrated_model[::-1, :], extent=(1, -1, -1, 1), vmin=min, vmax=max)
+to_plot = {name: pos for pos, name in zip(a_source_lmn, a_source_name)}
+ax4.set_title('sky without calibrated model')
+
+make_sky_plot(sky_image, to_plot)
+
+plt.show()