def write_ip_NDF(data, bad_pixel_ref):
"""
This function writes out the array ip parameter data to an ndf_file.
Invocation:
result = write_ip_NDF(data,bad_pixel_ref)
Arguements:
data = The array ip parameter data
bad_ref = A NDF with bad pixel values to copy over.
Returned Value:
Writes NDF and returns handle.
"""
ndf_name_orig = NDG(1)
indf = ndf.open(ndf_name_orig[0], 'WRITE', 'NEW')
indf.new('_DOUBLE', 2, numpy.array([1, 1]), numpy.array([32, 40]))
ndfmap = indf.map('DATA', '_DOUBLE', 'WRITE')
ndfmap.numpytondf(data)
indf.annul()
# Copy bad pixels
ndf_name = NDG(1)
invoke("$KAPPA_DIR/copybad in={0} ref={1} out={2}".format(
ndf_name_orig, bad_pixel_ref, ndf_name))
return ndf_name
def force_flat(ins, masks):
"""
Forces the background regions to be flat in a set of Q or U images.
Invocation:
result = force_flat( ins, masks )
Arguments:
in = NDG
An NDG object specifying a group of Q or U images from which
any low frequency background structure is to be removed.
masks = NDG
An NDG object specifying a corresponding group of Q or U images
in which source pixels are bad. These are only used to mask the
images specified by "in". It should have the same size as "in".
Returned Value:
A new NDG object containing the group of corrected Q or U images.
"""
# How many NDFs are we processing?
nndf = len(ins)
# Blank out sources by copy the bad pixels from "mask" into "in".
msg_out(" masking...")
qm = NDG(ins)
invoke("$KAPPA_DIR/copybad in={0} ref={1} out={2}".format(ins, masks, qm))
# Smooth the blanked NDFs using a 3 pixel Gaussian. Set wlim so that
# small holes are filled in by the smoothing process.
msg_out(" smoothing...")
qs = NDG(ins)
invoke("$KAPPA_DIR/gausmooth in={0} out={1} fwhm=3 wlim=0.5".format(
qm, qs))
# Fill remaining big holes using artifical data.
msg_out(" filling...")
qf = NDG(ins)
invoke("$KAPPA_DIR/fillbad in={0} out={1} niter=10 size=10 variance=no".
format(qs, qf))
# Subtract the filled low frequency data form the original to create the
# returned images.
msg_out(" removing low frequency background structure...")
result = NDG(ins)
invoke("$KAPPA_DIR/sub in1={0} in2={1} out={2}".format(ins, qf, result))
return result
def run_calcqu(input_data, config, harmonic):
# The following call to SMURF:CALCQU creates two HDS container files -
# one holding a set of Q NDFs and the other holding a set of U NDFs. Create
# these container files in the NDG temporary directory.
qcont = NDG(1)
qcont.comment = "qcont"
ucont = NDG(1)
ucont.comment = "ucont"
msg_out("Calculating Q and U values for each bolometer...")
invoke(
"$SMURF_DIR/calcqu in={0} config=\"{1}\" lsqfit=no outq={2} outu={3} "
"harmonic={4} fix".format(input_data, starutil.shell_quote(config),
qcont, ucont, harmonic))
return (qcont, ucont)
ff = indata
except starutil.StarUtilError:
pass
# Flat field the supplied template data
if not ff:
ff = NDG.load("FF")
if ff:
msg_out("Re-using old flatfielded template data...")
if not ff:
ffdir = NDG.subdir()
msg_out("Flatfielding template data...")
invoke("$SMURF_DIR/flatfield in={0} out=\"{1}/*\"".format(
indata, ffdir))
ff = NDG("{0}/\*".format(ffdir))
ff.save("FF")
# Output files. Base the modification on "ff" rather than "indata",
# since "indata" may include non-science files (flatfields, darks etc)
# for which no corresponding output file should be created.
gexp = parsys["OUT"].value
outdata = NDG(ff, gexp)
# If required, create new artificial I, Q and U maps.
if newart:
msg_out("Creating new artificial I, Q and U maps...")
# Get the parameters defining the artificial data
artform = parsys["ARTFORM"].value
ipeak = parsys["IPEAK"].value
else:
fcf_i = 491.0
elif filter == 850:
fcf_qu = 725.0
if ipol2:
fcf_i = 725.0
else:
fcf_i = 537.0
else:
raise starutil.InvalidParameterError("Invalid FILTER header value "
"'{0} found in {1}.".format( filter, qin[0] ) )
# Remove any spectral axes
qtrim = NDG(qin)
invoke( "$KAPPA_DIR/ndfcopy in={0} out={1} trim=yes".format(qin,qtrim) )
utrim = NDG(uin)
invoke( "$KAPPA_DIR/ndfcopy in={0} out={1} trim=yes".format(uin,utrim) )
itrim = NDG(iin)
invoke( "$KAPPA_DIR/ndfcopy in={0} out={1} trim=yes".format(iin,itrim) )
# Rotate them to use the same polarimetric reference direction.
qrot = NDG(qtrim)
urot = NDG(utrim)
invoke( "$POLPACK_DIR/polrotref qin={0} uin={1} like={2} qout={3} uout={4} ".
format(qtrim,utrim,qtrim[0],qrot,urot) )
# Mosaic them into a single set of Q, U and I images, aligning them
# with the first I image.
qmos = NDG( 1 )
# Get a flag indicating if the tile's master NDF existed before the
# above invocation of "tileinfo".
existed = starutil.get_task_par("exists", "tileinfo")
# Get the 2D spatial pixel index bounds of the master tile.
tlbnd = starutil.get_task_par("lbnd", "tileinfo")
tubnd = starutil.get_task_par("ubnd", "tileinfo")
# If the NDFs are not gridded using the JSA all-sky grid appropriate to
# the specified instrument, then we need to resample them onto that grid
# before coadding the new and old data. We only need do this for the
# first tile for each input NDF, since all tiles are aligned on the same
# pixel grid.
if aligned is None:
if not jsa:
aligned = NDG(1)[0]
invoke("$KAPPA_DIR/wcsalign in={0} ref={1} out={2} lbnd=! "
"method=bilin".format(ndf, tilendf, aligned))
else:
aligned = ndf
# Get the pixel index bounds of the aligned NDF.
invoke("$KAPPA_DIR/ndftrace ndf={0}".format(aligned))
nlbnd = starutil.get_task_par("lbound", "ndftrace")
nubnd = starutil.get_task_par("ubound", "ndftrace")
# Get the 2D spatial pixel index bounds of the overlap of the current tile
# and the aligned NDF.
olbnd = [1, 1]
oubnd = [0, 0]
for i in (0, 1):
def remove_corr(ins, masks):
"""
Masks the supplied set of Q or U images and then looks for and removes
correlated components in the background regions.
Invocation:
result = remove_corr( ins, masks )
Arguments:
ins = NDG
An NDG object specifying a group of Q or U images from which
correlated background components are to be removed.
masks = NDG
An NDG object specifying a corresponding group of Q or U images
in which source pixels are bad. These are only used to mask the
images specified by "in". It should have the same size as "in".
Returned Value:
A new NDG object containing the group of corrected Q or U images.
"""
# How many NDFs are we processing?
nndf = len(ins)
# Blank out sources by copy the bad pixels from "mask" into "in". We refer
# to "q" below, but the same applies whether processing Q or U.
msg_out(" masking...")
qm = NDG(ins)
invoke("$KAPPA_DIR/copybad in={0} ref={1} out={2}".format(ins, masks, qm))
# Find the most correlated pair of imagtes. We use the basic correlation
# coefficient calculated by kappa:scatter for this.
msg_out(" Finding most correlated pair of images...")
cmax = 0
for i in range(0, nndf - 1):
for j in range(i + 1, nndf):
invoke("$KAPPA_DIR/scatter in1={0} in2={1} device=!".format(
qm[i], qm[j]))
c = starutil.get_task_par("corr", "scatter")
if abs(c) > abs(cmax):
cmax = c
cati = i
catj = j
if abs(cmax) < 0.3:
msg_out(" No correlated images found!")
return ins
msg_out(" Correlation for best pair of images = {0}".format(cmax))
# Find images that are reasonably correlated to the pair found above,
# and coadd them to form a model for the correlated background
# component. Note, the holes left by the masking are filled in by the
# coaddition using background data from other images.
msg_out(" Forming model...")
# Form the average of the two most correlated images, first normalising
# them to a common scale so that they both have equal weight.
norm = "{0}/norm".format(NDG.tempdir)
if not normer(qm[cati], qm[catj], 0.3, norm):
norm = qm[cati]
mslist = NDG([qm[catj], norm])
ave = "{0}/ave".format(NDG.tempdir)
invoke(
"$CCDPACK_DIR/makemos in={0} method=mean genvar=no usevar=no out={1}".
format(mslist, ave))
# Loop round each image finding the correlation factor of the image and
# the above average image.
temp = "{0}/temp".format(NDG.tempdir)
nlist = []
ii = 0
for i in range(0, nndf):
c = blanker(qm[i], ave, temp)
# If the correlation is high enough, normalize the image to the average
# image and then include the normalised image in the list of images to be
# coadded to form the final model.
if abs(c) > 0.3:
tndf = "{0}/t{1}".format(NDG.tempdir, ii)
ii += 1
invoke(
"$KAPPA_DIR/normalize in1={1} in2={2} out={0} device=!".format(
tndf, temp, ave))
nlist.append(tndf)
if ii == 0:
msg_out(" No secondary correlated images found!")
return ins
msg_out(
" Including {0} secondary correlated images in the model.".format(
ii))
# Coadded the images created above to form the model of the correlated
# background component. Fill any remaining bad pixels with artificial data.
model = "{0}/model".format(NDG.tempdir)
included = NDG(nlist)
invoke(
"$CCDPACK_DIR/makemos in={0} method=mean usevar=no genvar=no out={1}".
format(included, temp))
invoke("$KAPPA_DIR/fillbad in={1} variance=no out={0} size=10 niter=10".
format(model, temp))
# Now estimate how much of the model is present in each image and remove it.
msg_out(" Removing model...")
temp2 = "{0}/temp2".format(NDG.tempdir)
qnew = NDG(ins)
nbetter = 0
for i in range(0, nndf):
# Try to normalise the model to the current image. This fails if the
# correlation between them is too low.
if normer(model, qm[i], 0.3, temp):
# Remove the scaled model form the image.
invoke("$KAPPA_DIR/sub in1={0} in2={1} out={2}".format(
ins[i], temp, temp2))
# We now check that removing the correlated background component has in
# fact made the image flatter (poor fits etc can mean that images that
# are poorly correlated to the model have a large amount of model
# removed and so make the image less flat). FInd the standard deviation
# of the data in the original image and in the corrected image.
invoke("$KAPPA_DIR/stats {0} quiet".format(ins[i]))
oldsig = get_task_par("sigma", "stats")
invoke("$KAPPA_DIR/stats {0} quiet".format(temp2))
newsig = get_task_par("sigma", "stats")
# If the correction has made the image flatter, copy it to the returned NDG.
if newsig < oldsig:
nbetter += 1
invoke("$KAPPA_DIR/ndfcopy in={1} out={0}".format(
qnew[i], temp2))
else:
invoke("$KAPPA_DIR/ndfcopy in={0} out={1}".format(
ins[i], qnew[i]))
# If the input image is poorly correlated to the model, return the input
# image unchanged.
else:
invoke("$KAPPA_DIR/ndfcopy in={0} out={1}".format(ins[i], qnew[i]))
msg_out(" {0} out of {1} images have been improved.".format(
nbetter, nndf))
# Return the corrected images.
return qnew
if starutil.get_task_par("numgood", "stats") > 0:
# If so, append the section to the list of NDFs to be included in the output.
tilendf.append(sec)
itilelist.append(itile)
# Raise an exception if no data is available for the tiles overlap
msg_out(" ")
if len(tilendf) == 0:
raise starutil.StarUtilError(
"No JSA {0} data is available "
"for the requested region.".format(instrument))
# Otherwise, paste the sections together to form the output NDF.
else:
tiles = NDG(tilendf)
invoke("$KAPPA_DIR/paste in={0} out={1}".format(tiles, outdata))
msg_out("Created output NDF {0} from tiles {1}".format(
outdata, itilelist))
# Remove temporary files.
cleanup()
# If an StarUtilError of any kind occurred, display the message but hide the
# python traceback. To see the trace back, uncomment "raise" instead.
except starutil.StarUtilError as err:
# raise
print(err)
cleanup()
# This is to trap control-C etc, so that we can clean up temp files.
def get_filtered_skydip_data(qarray, uarray, clip, a):
"""
This function takes q and u array data (output from calcqu), applies ffclean to remove spikes
and puts in numpy array variable
It borrows (copies) heavily from pol2cat.py (2015A)
Invocation:
( qdata_total,qvar_total,udata_total,uvar_total,elevation,opacity_term,bad_pixel_ref ) = ...
get_filtered_skydip_data(qarray,uarray,clip,a)
Arguments:
qarray = An NDF of Q array data (output from calcqu).
uarray = An NDF of U array data (output form calcqu).
clip = The sigma cut for ffclean.
a = A string indicating the array (eg. 'S8A').
Returned Value:
qdata_total = A numpy array with the cleaned qarray data.
qvar_total = A numpy array with the qarray variance data.
udata_total = A numpy array with the cleaned uarray data.
uvar_total = A numpy array with the uarray variance data.
elevation = A numpy array with the elevation data
opacity_term = A numpy array with the opacity brightness term (1-exp(-tau*air_mass))
Here tau is calculated using the WVM data as input.
"""
# Remove spikes from the Q images for the current subarray. The cleaned NDFs
# are written to temporary NDFs specified by the new NDG object "qff", which
# inherit its size from the existing group "qarray"".
msg_out("Removing spikes from {0} bolometer Q values...".format(a))
qff = NDG(qarray)
qff.comment = "qff"
invoke("$KAPPA_DIR/ffclean in={0} out={1} genvar=yes box=3 clip=\[{2}\]".
format(qarray, qff, clip))
# Remove spikes from the U images for the current subarray. The cleaned NDFs
# are written to temporary NDFs specified by the new NDG object "uff", which
# inherit its size from the existing group "uarray"".
msg_out("Removing spikes from {0} bolometer U values...".format(a))
uff = NDG(uarray)
uff.comment = "uff"
invoke("$KAPPA_DIR/ffclean in={0} out={1} genvar=yes box=3 clip=\[{2}\]".
format(uarray, uff, clip))
elevation = []
opacity_term = []
for stare in range(len(qff[:])):
# Stack Q data in numpy array
# Get elevation information
elevation.append(
numpy.array(
float(
invoke(
"$KAPPA_DIR/fitsmod ndf={0} edit=print keyword=ELSTART"
.format(qff[stare])))))
# Get Tau (Opacity) information
tau_temp = numpy.array(
float(
invoke(
"$KAPPA_DIR/fitsmod ndf={0} edit=print keyword=WVMTAUST".
format(qff[stare]))))
# Convert to obs band.
if '4' in a:
tau_temp = 19.04 * (tau_temp - 0.018) # Eq from Dempsey et al
elif '8' in a:
tau_temp = 5.36 * (tau_temp - 0.006) # Eq from Dempsey et al.
opacity_term.append(1 -
numpy.exp(-1 * tau_temp /
numpy.sin(numpy.radians(elevation[-1]))))
invoke("$KAPPA_DIR/ndftrace {0} quiet".format(qff[stare]))
nx = get_task_par("dims(1)", "ndftrace")
ny = get_task_par("dims(2)", "ndftrace")
qdata_temp = numpy.reshape(Ndf(qff[stare]).data, (ny, nx))
qdata_temp[numpy.abs(qdata_temp) > 1e300] = numpy.nan
if stare == 0:
qdata_total = qdata_temp
else:
qdata_total = numpy.dstack((qdata_total, qdata_temp))
qvar_temp = numpy.reshape(Ndf(qff[stare]).var, (ny, nx))
qdata_temp[numpy.abs(qvar_temp) > 1e300] = numpy.nan
if stare == 0:
qvar_total = qvar_temp
else:
qvar_total = numpy.dstack((qvar_total, qvar_temp))
# Stack U data in numpy array
invoke("$KAPPA_DIR/ndftrace {0} quiet".format(uff[stare]))
nx = get_task_par("dims(1)", "ndftrace")
ny = get_task_par("dims(2)", "ndftrace")
udata_temp = numpy.reshape(Ndf(uff[stare]).data, (ny, nx))
udata_temp[numpy.abs(udata_temp) > 1e300] = numpy.nan
if stare == 0:
udata_total = udata_temp
else:
udata_total = numpy.dstack((udata_total, udata_temp))
uvar_temp = numpy.reshape(Ndf(uff[stare]).var, (ny, nx))
udata_temp[numpy.abs(uvar_temp) > 1e300] = numpy.nan
if stare == 0:
uvar_total = uvar_temp
else:
uvar_total = numpy.dstack((uvar_total, uvar_temp))
# Create bad pixel reference.
bad_pixel_ref = NDG(1)
invoke("$KAPPA_DIR/copybad in={0} ref={1} out={2}".format(
qff, uff, bad_pixel_ref))
return (qdata_total, qvar_total, udata_total, uvar_total, elevation,
opacity_term, bad_pixel_ref)
chi2Vals[row_val, col_val] = ipprms.fun
else:
returnCode[row_val, col_val] = False
# Write NDFs.
out_p0 = write_ip_NDF(ip_prms['Pf_' + a[-1]], bad_pixel_ref)
out_p1 = write_ip_NDF(ipprms_pol_screen, bad_pixel_ref)
out_c0 = write_ip_NDF(ipprms_Co, bad_pixel_ref)
out_angc = write_ip_NDF(ip_prms['Theta_ip_' + a[-1]],
bad_pixel_ref)
# Fill any bad pixels with smooth function to match surrounding pixels
msg_out(
"Filling in bad pixel values for {0} bolometer IP parameters..."
.format(a))
out_p0_filled = NDG(1)
invoke(
"$KAPPA_DIR/fillbad in={0} out={1} variance=true niter=10 size=15"
.format(out_p0, out_p0_filled))
out_p1_filled = NDG(1)
invoke(
"$KAPPA_DIR/fillbad in={0} out={1} variance=true niter=10 size=15"
.format(out_p1, out_p1_filled))
out_c0_filled = NDG(1)
invoke(
"$KAPPA_DIR/fillbad in={0} out={1} variance=true niter=10 size=15"
.format(out_c0, out_c0_filled))
out_angc_filled = NDG(1)
invoke(
"$KAPPA_DIR/fillbad in={0} out={1} variance=true niter=10 size=15"
.format(out_angc, out_angc_filled))
# Create a list holding the paths to the tile NDFs that intersect
# the required region.
ntile = 0
used_tile_list = []
for jsatile in jsatile_list:
key = str(jsatile)
if key in tile_dict and tile_dict[key]:
used_tile_list.append(tile_dict[key])
ntile += 1
# Create an NDG holding the group of tile NDFs.
if ntile > 0:
msg_out(
"{0} of the supplied tiles intersect the requested region.".
format(ntile))
used_tiles = NDG(used_tile_list)
else:
raise starutil.InvalidParameterError(
"None of the supplied JSA tiles "
"intersect the requested region")
# If we are using all tiles, just use the supplied group of tiles. Use
# the middle supplied tile as the reference.
else:
used_tiles = tiles
jsatile = int(len(tiles) / 2)
jsatile = starutil.get_fits_header(tiles[jsatile], "TILENUM")
# Paste these tile NDFs into a single image. This image still uses the
# JSA all-sky pixel grid. If we have only a single tile, then just use
# it as it is.
# south pole). The above call to jsatileinfo will have determined the
# appropriate projection to use, so get it.
proj = starutil.get_task_par("PROJ", "jsatilelist")
# Create a file holding the FITS-WCS header for the first tile, using
# the type of projection determined above.
head = "{0}/header".format(NDG.tempdir)
invoke("$SMURF_DIR/jsatileinfo itile={0} instrument={1} header={2} "
"proj={3} quiet".format(tiles[0], instrument, head, proj))
# Get the lower pixel index bounds of the first tile.
lx = int(starutil.get_task_par("LBND(1)", "jsatileinfo"))
ly = int(starutil.get_task_par("LBND(2)", "jsatileinfo"))
# Create a 1x1 NDF and store the tile headers in the FITS extension.
ref = NDG(1)
invoke("$KAPPA_DIR/creframe out={0} mode=fl mean=0 lbound=\[{1},{2}\] "
"ubound=\[{1},{2}\]".format(ref, lx, ly))
invoke("$KAPPA_DIR/fitstext ndf={0} file={1}".format(ref, head))
# Get the nominal spatial pixel size of the supplied NDF.
invoke("$KAPPA_DIR/ndftrace ndf={0} quiet".format(inndf))
pixsize1 = float(starutil.get_task_par("FPIXSCALE(1)", "ndftrace"))
pixsize2 = float(starutil.get_task_par("FPIXSCALE(2)", "ndftrace"))
pixsize_in = math.sqrt(pixsize1 * pixsize2)
# Get the nominal tile pixel size.
invoke("$KAPPA_DIR/ndftrace ndf={0} quiet".format(ref))
pixsize1 = float(starutil.get_task_par("FPIXSCALE(1)", "ndftrace"))
pixsize2 = float(starutil.get_task_par("FPIXSCALE(2)", "ndftrace"))
pixsize_tile = math.sqrt(pixsize1 * pixsize2)
fd.write("flt.filt_edgehigh_last=<undef>\n") # final iteration. We
fd.write("flt.filt_edgelow_last=<undef>\n") # reset them here in
fd.write("flt.whiten_last=<undef>\n") # case they are set in
fd.write("com.perarray_last=<undef>\n") # the supplied config.
if precleaned:
fd.write("downsampscale = 0\n") # Cleaned data will have been downsampled already.
fd.write("downsampfreq = 0\n")
fd.close() # Close the config file.
# Get the name of a temporary NDF that can be used to store the first
# iteration map. This NDF is put in the NDG temp directory. If we are
# only doing one iteration, used the supplied output NDF name.
if niter == 1:
newmap = outdata
else:
newmap = NDG(1)
prevmap = None
# Start a list of these maps in case we are creating an output itermap cube.
maps = []
maps.append(newmap)
# If we are restarting, check if the NDF already exists and is readable.
# If so, we do not re-create it.
msg_out( "Iteration 1...")
gotit = False
if restart != None:
try:
invoke("$KAPPA_DIR/ndftrace ndf={0} quiet=yes".format(newmap))
msg_out( "Re-using existing map {0}".format(newmap) )
gotit = True
# Erase any NDFs holding cleaned data, exteinction or pointing data from
# previous runs.
for path in glob.glob("*_con_res_cln.sdf"):
myremove(path)
base = path[:-16]
myremove("{0}_lat.sdf".format(base))
myremove("{0}_lon.sdf".format(base))
myremove("{0}_con_ext.sdf".format(base))
# Use sc2concat to concatenate and flatfield the data.
msg_out("Concatenating and flatfielding...")
concbase = NDG.tempfile("")
invoke("$SMURF_DIR/sc2concat in={0} outbase={1} maxlen=360".format(
indata, concbase))
concdata = NDG("{0}_*".format(concbase))
# Use makemap to generate quality, extinction and pointing info.
confname = NDG.tempfile()
fd = open(confname, "w")
fd.write("^$STARLINK_DIR/share/smurf/dimmconfig.lis\n")
fd.write("numiter=1\n")
fd.write("exportclean=1\n")
fd.write("exportndf=ext\n")
fd.write("exportlonlat=1\n")
fd.write("dcfitbox=0\n")
fd.write("noisecliphigh=0\n")
fd.write("order=0\n")
fd.write("downsampscale=0\n")
if fakemap != None:
fd.write("fakemap={0}\n".format(fakemap))
fd.write("flt.filt_edgelow_last=<undef>\n") # reset them here in
fd.write("flt.whiten_last=<undef>\n") # case they are set in
fd.write("com.perarray_last=<undef>\n") # the supplied config.
if precleaned:
fd.write("downsampscale = 0\n"
) # Cleaned data will have been downsampled already.
fd.write("downsampfreq = 0\n")
fd.close() # Close the config file.
# Get the name of a temporary NDF that can be used to store the first
# iteration map. This NDF is put in the NDG temp directory. If we are
# only doing one iteration, used the supplied output NDF name.
if niter == 1:
newmap = outdata
else:
newmap = NDG(1)
prevmap = None
# Start a list of these maps in case we are creating an output itermap cube.
maps = []
maps.append(newmap)
# If we are restarting, check if the NDF already exists and is readable.
# If so, we do not re-create it.
msg_out("Iteration 1...")
gotit = False
if restart is not None:
try:
invoke("$KAPPA_DIR/ndftrace ndf={0} quiet=yes".format(newmap))
msg_out("Re-using existing map {0}".format(newmap))
gotit = True
