def get_spectra_raw(self, start, stop):
'''Wrapper for get_spectra for waterfaller compatibility
'''
data = self.get_spectra(start, stop)
return spectra.Spectra(self.freqs, self.tsamp, data.transpose(), \
starttime=self.tsamp*start, dm=0)
def get_spectra(self, startsamp, N, nsub=None):
"""Return 2D array of data from PSRFITS file.
Inputs:
startsamp, Starting sample
N: number of samples to read
Output:
data: 2D numpy array
"""
# Calculate starting subint and ending subint
startsub = int(startsamp / self.nsamp_per_subint)
skip = startsamp - (startsub * self.nsamp_per_subint)
endsub = int((startsamp + N) / self.nsamp_per_subint)
trunc = ((endsub + 1) * self.nsamp_per_subint) - (startsamp + N)
if nsub:
print "Will subband the data to : %i channels" % nsub
nchan_per_sub = self.nchan / nsub
sub_hifreqs = self.freqs[np.arange(nsub) * nchan_per_sub]
sub_lofreqs = self.freqs[(1 + np.arange(nsub)) * nchan_per_sub - 1]
sub_ctrfreqs = 0.5 * (sub_hifreqs + sub_lofreqs)
sub_freqs = sub_ctrfreqs
# Read data
data = []
for isub in xrange(startsub, endsub + 1):
if nsub:
sub_data = self.read_subint(isub)
# Subband
sub_data = np.transpose(np.array([np.sum(sub, axis=1) for sub in \
np.hsplit(sub_data, nsub)]))
data.append(sub_data)
else:
data.append(self.read_subint(isub))
if len(data) > 1:
data = np.concatenate(data)
else:
data = np.array(data).squeeze()
data = np.transpose(data)
#Truncate data to desired interval
if trunc > 0:
data = data[:, skip:-trunc]
elif trunc == 0:
data = data[:, skip:]
else:
raise ValueError("Number of bins to truncate is negative: %d" %
trunc)
if not self.specinfo.need_flipband:
# for psrfits module freqs go from low to high.
# spectra module expects high frequency first.
data = data[::-1, :]
freqs = self.freqs[::-1]
else:
freqs = self.freqs
if nsub:
freqs = sub_freqs
return spectra.Spectra(freqs, self.tsamp, data, \
starttime=self.tsamp*startsamp, dm=0)
def get_spectra(self, start, nspec):
stop = min(start+nspec, self.nspec)
pos = self.header_size+start*self.bytes_per_spectrum
# Compute number of elements to read
nspec = int(stop) - int(start)
num_to_read = nspec*self.nchans
num_to_read = max(0, num_to_read)
self.filfile.seek(pos, os.SEEK_SET)
spectra_dat = np.fromfile(self.filfile, dtype=self.dtype,
count=num_to_read)
spectra_dat.shape = nspec, self.nchans
spec = spectra.Spectra(self.freqs, self.tsamp, spectra_dat.T,
starttime=start*self.tsamp, dm=0.0)
return spec
def get_spectra(self, startsamp, N):
"""Return 2D array of data from filterbank file.
Inputs:
startsamp: Starting sample
N: number of samples to read
Output:
data: 2D numpy array
"""
self.seek_to_sample(startsamp)
data = self.read_Nsamples(N)
data.shape = (N, self.header['nchans'])
return spectra.Spectra(self.frequencies, self.tsamp, data.T, \
starttime=self.tsamp*startsamp, dm=0)
def random_gaussian_radius_spectra(self, true_spectrum):
""" Smears the radius of a spectra object by generating a
number of random points from a Gaussian pdf generated for
that bin. The number of points generated is equivalent
to the number of entries in that bin.
Args:
true_spectrum (spectra): spectrum to be smeared
Returns:
A smeared spectra object.
"""
raw_events = true_spectrum._raw_events
energy_step = (true_spectrum._energy_high -
true_spectrum._energy_low) / true_spectrum._energy_bins
time_step = (true_spectrum._time_high -
true_spectrum._time_low) / true_spectrum._time_bins
radial_step = (true_spectrum._radial_high -
true_spectrum._radial_low) / true_spectrum._radial_bins
smeared_spectrum = spectra.Spectra(
true_spectrum._name + str(self._position_resolution) +
"_position_resolution", true_spectrum._num_decays)
for time_bin in range(true_spectrum._time_bins):
mean_time = time_bin * time_step + 0.5 * time_step
for energy_bin in range(true_spectrum._energy_bins):
mean_energy = energy_bin * energy_step + 0.5 * energy_step
for radial_bin in range(true_spectrum._radial_bins):
mean_radius = radial_bin * radial_step + 0.5 * radial_step
entries = true_spectrum._data[energy_bin, radial_bin,
time_bin]
for i in range(int(entries)):
try:
smeared_spectrum.fill(
mean_energy,
np.fabs(
np.random.normal(
mean_radius,
self._position_resolution)), mean_time)
except ValueError:
# Occurs when smeared radius is outside bin range
print "Warning: Smeared radius out of bounds. Skipping."
continue
smeared_spectrum._raw_events = raw_events
return smeared_spectrum
def get_spectra(self, startsamp, N):
"""Return 2D array of data from PSRFITS file.
Inputs:
startsamp, Starting sample
N: number of samples to read
Output:
data: 2D numpy array
"""
# Calculate starting subint and ending subint
startsub = int(startsamp / self.nsamp_per_subint)
skip = startsamp - (startsub * self.nsamp_per_subint)
endsub = int((startsamp + N) / self.nsamp_per_subint)
trunc = ((endsub + 1) * self.nsamp_per_subint) - (startsamp + N)
# Read data
data = []
for isub in xrange(startsub, endsub + 1):
data.append(self.read_subint(isub))
if len(data) > 1:
data = np.concatenate(data)
else:
data = np.array(data).squeeze()
data = np.transpose(data)
# Truncate data to desired interval
if trunc > 0:
data = data[:, skip:-trunc]
elif trunc == 0:
data = data[:, skip:]
else:
raise ValueError("Number of bins to truncate is negative: %d" %
trunc)
if not self.specinfo.need_flipband:
# for psrfits module freqs go from low to high.
# spectra module expects high frequency first.
data = data[::-1, :]
freqs = self.freqs[::-1]
else:
freqs = self.freqs
return spectra.Spectra(freqs, self.tsamp, data, \
starttime=self.tsamp*startsamp, dm=0)
def dm_transform_filobj(data,
dt,
freqs,
ref_freq=np.inf,
dm_min=-10,
dm_max=10,
ndm=50):
""" Transform freq/time data to dm/time data.
"""
import spectra
nt = data.shape[1]
ref_freq = 0.5 * (freqs[0] + freqs[-1])
datafilobj = spectra.Spectra(freqs, dt, data, starttime=0, dm=0)
dms = np.linspace(dm_min, dm_max, 250)
dmtarr = np.empty([250, nt])
for ii, dm in enumerate(dms):
datafilobj.dedisperse(dm, ref_freq=ref_freq)
dmtarr[ii] = datafilobj.data.mean(0)
return dmtarr
def weight_gaussian_radius_spectra(self, true_spectrum, num_sigma=5.):
""" Smears the radius of a spectra object by calculating a Gaussian pdf
for each bin and applies a weight to the bin and corresponding bins a
default 5 sigma apart.
Args:
true_spectrum (spectra): spectrum to be smeared
num_sigma (float): Width of window to apply the weight method.
Default is 5.
Returns:
A smeared spectra object.
"""
raw_events = true_spectrum._raw_events
energy_step = (true_spectrum._energy_high -
true_spectrum._energy_low) / true_spectrum._energy_bins
time_step = (true_spectrum._time_high -
true_spectrum._time_low) / true_spectrum._time_bins
radial_step = (true_spectrum._radial_high -
true_spectrum._radial_low) / true_spectrum._radial_bins
smeared_spectrum = spectra.Spectra(
true_spectrum._name + str(self._position_resolution) +
"_position_resolution", true_spectrum._num_decays)
for time_bin in range(true_spectrum._time_bins):
mean_time = time_bin * time_step + 0.5 * time_step
for energy_bin in range(true_spectrum._energy_bins):
mean_energy = energy_bin * energy_step + 0.5 * energy_step
for radial_bin in range(true_spectrum._radial_bins):
mean_radius = radial_bin * radial_step + 0.5 * radial_step
entries = float(true_spectrum._data[energy_bin, radial_bin,
time_bin])
if entries == 0:
continue # Bin Empty
lower_bin = self.floor_to_bin(
mean_radius - num_sigma * self._position_resolution,
radial_step) + 0.5 * radial_step
upper_bin = self.ceil_to_bin(
mean_radius + num_sigma * self._position_resolution,
radial_step) - 0.5 * radial_step
if upper_bin > true_spectrum._radial_high:
upper_bin = true_spectrum._radial_high - 0.5 * energy_step
if lower_bin < true_spectrum._radial_low:
lower_bin = true_spectrum._radial_low + 0.5 * energy_step
weights = []
for radius in np.arange(lower_bin, upper_bin, radial_step):
weights.append(
self.calc_gaussian(radius, mean_radius,
self._position_resolution))
weight_tot = np.array(weights).sum()
i = 0
for radius in np.arange(lower_bin, upper_bin, radial_step):
try:
smeared_spectrum.fill(
mean_energy, radius, mean_time,
entries * weights[i] / weight_tot)
except ValueError:
# Occurs when smeared radius is outside bin range
print "Warning: Smeared radius out of bounds. Skipping."
continue
i += 1
smeared_spectrum._raw_events = raw_events
return smeared_spectrum
