def calc_min_smearing(self, verbose=False):
"""Calculate minimum (optimal) smearing.
Inputs:
verbose: Print information to screen (Default = False)
"""
half_dDMmin = 0.5 * ALLOW_DMSTEPS[0]
self.min_chan_smear = psr_utils.dm_smear(self.loDM + half_dDMmin, self.obs.chanwidth, self.obs.fctr)
self.min_bw_smear = psr_utils.dm_smear(half_dDMmin, self.obs.BW, self.obs.fctr)
self.min_total_smear = np.sqrt(2 * self.obs.dt ** 2.0 + self.min_chan_smear ** 2.0 + self.min_bw_smear ** 2.0)
self.best_resolution = max([self.req_resolution, self.min_chan_smear, self.min_bw_smear, self.obs.dt])
self.resolution = self.best_resolution
if verbose:
print
print "Minimum total smearing : %.3g s" % self.min_total_smear
print "--------------------------------------------"
print "Minimum channel smearing : %.3g s" % self.min_chan_smear
print "Minimum smearing across BW : %.3g s" % self.min_bw_smear
print "Minimum sample time : %.3g s" % self.obs.dt
print
print "Setting the new 'best' resolution to : %.3g s" % self.best_resolution
# See if the data is too high time resolution for our needs
if (FF * self.min_chan_smear > self.obs.dt) or (self.resolution > self.obs.dt):
if self.resolution > FF * self.min_chan_smear:
if verbose:
print " Note: resolution > dt (i.e. data is higher resolution than needed)"
self.resolution = self.resolution
else:
if verbose:
print " Note: min chan smearing > dt (i.e. data is higher resolution than needed)"
self.resolution = FF * self.min_chan_smear
def _compute_rating(self, cand):
"""Return a rating for the candidate. The rating value is the
ratio of the width of the narrowest gaussian component
to the DM smearing.
Input:
cand: A Candidate object to rate.
Output:
value: The rating value.
"""
pfd = cand.get_from_cache('pfd')
mgauss = cand.get_from_cache('multigaussfit')
ncomp = len(mgauss.components)
if not ncomp:
raise utils.RatingError("Bad number of components for single " \
"gaussian fit (%d)" % ncomp)
# Get the period
period = pfd.bary_p1 or pfd.topo_p1
if period is None:
raise utils.RatingError("Bad period in PFD file (%f)" % period)
f_ctr = (pfd.hifreq + pfd.lofreq)/2.0
dm_smear = psr_utils.dm_smear(pfd.bestdm, pfd.chan_wid, f_ctr)
width_phs = np.sqrt(dm_smear**2 + pfd.dt**2)/period
minfwhm = min([comp.fwhm for comp in mgauss.components])
return width_phs/minfwhm
def _compute_rating(self, cand):
"""Return a rating for the candidate. The rating value is the
ratio of the width of the narrowest gaussian component
to the DM smearing.
Input:
cand: A Candidate object to rate.
Output:
value: The rating value.
"""
pfd = cand.pfd
mgauss = cand.multigaussfit
ncomp = len(mgauss.components)
if not ncomp:
raise utils.RatingError("Bad number of components for single " \
"gaussian fit (%d)" % ncomp)
# Get the period
period = pfd.bary_p1 or pfd.topo_p1
if period is None:
raise utils.RatingError("Bad period in PFD file (%f)" % period)
f_ctr = (pfd.hifreq + pfd.lofreq) / 2.0
dm_smear = psr_utils.dm_smear(pfd.bestdm, pfd.chan_wid, f_ctr)
width_phs = np.sqrt(dm_smear**2 + pfd.dt**2) / period
minfwhm = min([comp.fwhm for comp in mgauss.components])
return width_phs / minfwhm
def subband_smear(DM, subDM, subBW, fctr):
"""
subband_smear(DM, subDM, subBW, fctr):
Return the smearing in ms caused by subbanding at DM='DM' given
subbands of bandwidth 'subBW' (MHz) at DM='subDM'. All values
are computed at the frequency fctr in MHz.
"""
return 1000.0 * pu.dm_smear(num.fabs(DM-subDM), subBW, fctr)
def fit_powers(self, freqlim=None, use_errors=True, **kwargs):
# Use iminuit to fit power-law + DC to powers
import iminuit
if freqlim is None:
freqlim = np.inf
if self.inf.DM > 0:
tdm = psr_utils.dm_smear(self.inf.DM, self.inf.BW,
self.inf.lofreq + 0.5 * self.inf.BW)
freqlim = 1.0 / tdm
print "Dispersion smearing time: %.2f ms" % (1000.0 * tdm)
freqlim = min(10.0, freqlim)
print "Only fitting using frequencies up to %.2f Hz" % freqlim
iuse = (self.freqs < freqlim)
iuse[0] = 0 # Always ignore zeroth element
if use_errors:
# Compute power errors
self.estimate_power_errors()
# Define function to minimize
def to_minimize(amp, index, dc):
model = power_law(self.freqs[iuse], amp, index, dc)
diff = model - self.powers[iuse]
#plt.figure()
#plt.plot(self.freqs[1:], self.powers[1:], 'r-', alpha=0.5)
#plt.plot(self.freqs[iuse], self.powers[iuse], 'k-', alpha=0.5)
#plt.plot(self.freqs[iuse], model, 'k--', lw=2)
#plt.xscale('log')
#plt.yscale('log')
#plt.xlabel("Freq (Hz)")
#plt.ylabel("Raw Power")
#plt.show()
if use_errors:
diff /= self.errs[iuse]
return np.sum(diff**2)
white = self.estimate_white_power_level(1000)
kwargs.setdefault('amp', 1e14)
kwargs.setdefault('error_amp', 1e12)
kwargs.setdefault('limit_amp', (0, 1e20))
kwargs.setdefault('index', -1.5)
kwargs.setdefault('error_index', .1)
kwargs.setdefault('limit_index', (-10.0, 0.0))
kwargs.setdefault('dc', white)
kwargs.setdefault('error_dc', white * 0.1)
#kwargs.setdefault('limit_dc', (white*0.01, np.max(self.powers[1:])))
kwargs.setdefault('fix_dc', True)
m = iminuit.Minuit(to_minimize,
frontend=iminuit.ConsoleFrontend,
print_level=0,
**kwargs)
# Minimize
m.migrad()
return m.values
def calc_min_smearing(self, verbose=False):
"""Calculate minimum (optimal) smearing.
Inputs:
verbose: Print information to screen (Default = False)
"""
half_dDMmin = 0.5 * ALLOW_DMSTEPS[0]
self.min_chan_smear = psr_utils.dm_smear(self.loDM+half_dDMmin, \
self.obs.chanwidth, self.obs.fctr)
self.min_bw_smear = psr_utils.dm_smear(half_dDMmin, self.obs.BW,
self.obs.fctr)
self.min_total_smear = np.sqrt(2*self.obs.dt**2.0 + \
self.min_chan_smear**2.0 + \
self.min_bw_smear**2.0)
self.best_resolution = max([self.req_resolution, self.min_chan_smear, \
self.min_bw_smear, self.obs.dt])
self.resolution = self.best_resolution
if verbose:
print
print "Minimum total smearing : %.3g s" % self.min_total_smear
print "--------------------------------------------"
print "Minimum channel smearing : %.3g s" % self.min_chan_smear
print "Minimum smearing across BW : %.3g s" % self.min_bw_smear
print "Minimum sample time : %.3g s" % self.obs.dt
print
print "Setting the new 'best' resolution to : %.3g s" % self.best_resolution
# See if the data is too high time resolution for our needs
if (FF*self.min_chan_smear > self.obs.dt) or \
(self.resolution > self.obs.dt):
if self.resolution > FF * self.min_chan_smear:
if verbose:
print " Note: resolution > dt (i.e. data is higher resolution than needed)"
self.resolution = self.resolution
else:
if verbose:
print " Note: min chan smearing > dt (i.e. data is higher resolution than needed)"
self.resolution = FF * self.min_chan_smear
def _compute_rating(self, cand):
"""Return a rating for the candidate. The rating value encodes
how close the candidate's period and DM are to that of a
known pulsar.
Input:
cand: A Candidate object to rate.
Output:
value: The rating value.
"""
info = cand.get_from_cache('info')
candp = info['bary_period']
canddm = info['dm']
ra = info['raj_deg']
dec = info['decj_deg']
pfd = cand.get_from_cache('pfd')
diff_ra = np.abs(self.known_ras - ra)
diff_dec = np.abs(self.known_decs - dec)
ii_nearby = (diff_ra < 0.2) & (diff_dec < 0.2)
knownps = self.known_periods[ii_nearby]
knowndms = self.known_dms[ii_nearby]
dp_smear_phase = np.inf * np.ones(knownps.size)
ddms = np.abs(canddm - knowndms)
bw = pfd.hifreq - pfd.lofreq
fctr = 0.5 * (pfd.hifreq + pfd.lofreq)
ddm_smear_phase = psr_utils.dm_smear(ddms, bw, fctr) / knownps
if knownps.size:
for b in range(1, M):
dp_smear_sec = np.abs(candp * b - knownps) * pfd.T
dp_smear_phase = np.min(np.vstack(
(dp_smear_sec / knownps, dp_smear_phase)),
axis=0)
for rat in self.ratios:
dp_smear_sec = np.abs(candp * b - knownps) * pfd.T
dp_smear_phase = np.min(np.vstack(
(dp_smear_sec / knownps, dp_smear_phase)),
axis=0)
smear_phase = np.min(
np.sqrt(dp_smear_phase**2 + ddm_smear_phase**2))
#if smear_phase > 1:
# print pfd.pfd_filename, smear_phase
smear_phase = np.clip(smear_phase, 0, 10)
else:
# No nearby known pulsars
smear_phase = 10
return smear_phase
def filter_on_effective_time_resolution(candidates, metadata, n_bins):
'''
Reject pulsar candidates with fewer than n_bins bins across their profile.
'''
out_cand_list = []
channel_bandwidth = metadata.channel_bandwidth
central_frequency = ((metadata.n_channels/2) - 0.5) * channel_bandwidth +\
metadata.low_channel_central_freq
for c in candidates:
intra_channel_smearing = psr_utils.dm_smear(c.DM, channel_bandwidth,
central_frequency)
if c.p / intra_channel_smearing > n_bins:
out_cand_list.append(c)
print 'Filtering on effective time resolution:'
print ' Removed %d out %d candidates.' % (
len(candidates) - len(out_cand_list), len(candidates))
return out_cand_list
def _compute_rating(self, cand):
"""Return a rating for the candidate. The rating value encodes
how close the candidate's period and DM are to that of a
known pulsar.
Input:
cand: A Candidate object to rate.
Output:
value: The rating value.
"""
info = cand.get_from_cache('info')
candp = info['bary_period']
canddm = info['dm']
ra = info['raj_deg']
dec = info['decj_deg']
pfd = cand.get_from_cache('pfd')
diff_ra = np.abs(self.known_ras - ra)
diff_dec = np.abs(self.known_decs - dec)
ii_nearby = (diff_ra < 0.2) & (diff_dec < 0.2)
knownps = self.known_periods[ii_nearby]
knowndms = self.known_dms[ii_nearby]
dp_smear_phase = np.inf*np.ones(knownps.size)
ddms = np.abs(canddm-knowndms)
bw = pfd.hifreq-pfd.lofreq
fctr = 0.5*(pfd.hifreq+pfd.lofreq)
ddm_smear_phase = psr_utils.dm_smear(ddms, bw, fctr)/knownps
if knownps.size:
for b in range(1, M):
dp_smear_sec = np.abs(candp*b-knownps)*pfd.T
dp_smear_phase = np.min(np.vstack((dp_smear_sec/knownps, dp_smear_phase)), axis=0)
for rat in self.ratios:
dp_smear_sec = np.abs(candp*b-knownps)*pfd.T
dp_smear_phase = np.min(np.vstack((dp_smear_sec/knownps, dp_smear_phase)), axis=0)
smear_phase = np.min(np.sqrt(dp_smear_phase**2+ddm_smear_phase**2))
#if smear_phase > 1:
# print pfd.pfd_filename, smear_phase
smear_phase = np.clip(smear_phase, 0, 10)
else:
# No nearby known pulsars
smear_phase = 10
return smear_phase
def __init__(self, ddplan, downsamp, loDM, dDM, numDMs=0, numsub=0, smearfact=2.0):
"""DDstep object constructor.
Inputs:
ddplan: DDplan object this DDstep is part of.
downsamp: Downsampling factor.
loDM: Low DM edge of this DDstep (pc cm-3)
dDM: DM step size to use (pc cm-3)
numDMs: Number of DMs. If numDMs=0 compute numDMs
based on DM range and spacing.
(Default: 0)
numsub: Number of subbands to use. (Default: no subbanding).
smearfact: Allowable smearing in a single channel, relative
to other smearing contributions (Default: 2.0)
"""
self.ddplan = ddplan
self.downsamp = downsamp
self.loDM = loDM
self.dDM = dDM
self.numsub = numsub
self.BW_smearing = psr_utils.dm_smear(dDM * 0.5, self.ddplan.obs.BW, self.ddplan.obs.fctr)
self.numprepsub = 0
if numsub:
# Calculate the maximum subband smearing we can handle
DMs_per_prepsub = 2
while True:
next_dsubDM = (DMs_per_prepsub + 2) * dDM
next_ss = psr_utils.dm_smear(next_dsubDM * 0.5, self.ddplan.obs.BW / numsub, self.ddplan.obs.fctr)
# The 0.8 is a small fudge factor to make sure that the subband
# smearing is always the smallest contribution
if next_ss > 0.8 * min(self.BW_smearing, self.ddplan.obs.dt * self.downsamp):
self.dsubDM = DMs_per_prepsub * dDM
self.DMs_per_prepsub = DMs_per_prepsub
self.sub_smearing = psr_utils.dm_smear(
self.dsubDM * 0.5, self.ddplan.obs.BW / self.numsub, self.ddplan.obs.fctr
)
break
DMs_per_prepsub += 2
else:
self.dsubDM = dDM
self.sub_smearing = 0.0
# Calculate the nominal DM to move to the next step
cross_DM = self.DM_for_smearfact(smearfact)
if cross_DM > self.ddplan.hiDM:
cross_DM = self.ddplan.hiDM
if numDMs == 0:
self.numDMs = int(np.ceil((cross_DM - self.loDM) / self.dDM))
if numsub:
self.numprepsub = int(np.ceil(self.numDMs * self.dDM / self.dsubDM))
self.numDMs = self.numprepsub * DMs_per_prepsub
else:
self.numDMs = numDMs
self.hiDM = loDM + self.numDMs * dDM
self.DMs = np.arange(self.numDMs, dtype="d") * self.dDM + self.loDM
# Calculate a few more smearing values
self.chan_smear = psr_utils.dm_smear(self.DMs, self.ddplan.obs.chanwidth, self.ddplan.obs.fctr)
self.tot_smear = np.sqrt(
(self.ddplan.obs.dt) ** 2.0
+ (self.ddplan.obs.dt * self.downsamp) ** 2.0
+ self.BW_smearing ** 2.0
+ self.sub_smearing ** 2.0
+ self.chan_smear ** 2.0
)
counterpart = j
break
# print f1,fwant
if counterpart == False:
print "Didn't find counterpart frequency for %f" % f0
sub.set_weight(i, 0.0)
continue
# else:
# print counterpart,f1
# print f0, f1
# sys.exit()
# print f0,f1
p = sub.get_folding_period()
smear = (psr_utils.dm_smear(dm, chbw, f0) +
psr_utils.dm_smear(dm, chbw, f1)) / p
delay = (psr_utils.delay_from_DM(dm, f0) -
psr_utils.delay_from_DM(dm, f1)) / p
delay -= int(delay)
if delay < -0.5: delay += 1.0
if delay > 0.5: delay -= 1.0
#print i, smear, delay
for pol in range(2):
if pol == 0:
correction = pol0correction / 2.0
else:
def plot(self, fn=None):
"""Generate a plot for this dedispersion plan.
Inputs:
fn: Filename to save plot as. If None, show plot interactively.
(Default: Show plot interactively.)
"""
fig = plt.figure(figsize=(11, 8.5))
ax = plt.axes()
stepDMs = []
# Plot each dedispersion step
for ii, (step, wf) in enumerate(zip(self.DDsteps, self.work_fracts)):
stepDMs.append(step.DMs)
DMspan = np.ptp(step.DMs)
loDM = step.DMs.min() + DMspan * 0.02
hiDM = step.DMs.max() - DMspan * 0.02
midDM = step.DMs.min() + DMspan * 0.5
# Sample time
plt.plot(step.DMs, np.zeros(step.numDMs)+step.ddplan.obs.dt*step.downsamp, \
'#33CC33', label=((ii and "_nolegend_") or "Sample Time (ms)"))
# DM stepsize smearing
plt.plot(step.DMs, np.zeros(step.numDMs)+step.BW_smearing, 'r', \
label=((ii and "_nolegend_") or "DM Stepsize Smearing"))
if self.numsub:
plt.plot(step.DMs, np.zeros(step.numDMs)+step.sub_smearing, '#993399', \
label=((ii and "_nolegend_") or "Subband Stepsize Smearing (# passes)"))
plt.plot(step.DMs, step.tot_smear, 'k', \
label=((ii and "_nolegend_") or "Total Smearing"))
# plot text
plt.text(midDM, 1.1*np.median(step.tot_smear), \
"%d (%.1f%%)" % (step.numDMs, 100.0*wf), \
rotation='vertical', size='small', \
ha='center', va='bottom')
plt.text(loDM, 0.85*step.ddplan.obs.dt*step.downsamp, \
"%g" % (1000*step.ddplan.obs.dt*step.downsamp), \
size='small', color='#33CC33', ha='left')
plt.text(hiDM, 0.85*step.BW_smearing, \
"%g" % step.dDM, size='small', color='r', ha='right')
if self.numsub:
plt.text(midDM, 0.85*step.sub_smearing, \
"%g (%d)" % (step.dsubDM, step.numprepsub), \
size='small', color='#993399', ha='center')
allDMs = np.concatenate(stepDMs)
chan_smear = psr_utils.dm_smear(allDMs, self.obs.chanwidth,
self.obs.fctr)
bw_smear = psr_utils.dm_smear(ALLOW_DMSTEPS[0], self.obs.BW,
self.obs.fctr)
tot_smear = np.sqrt(2*self.obs.dt**2.0 + \
chan_smear**2.0 + \
bw_smear**2.0)
plt.plot(allDMs, tot_smear, '#FF9933', label="Optimal Smearing")
plt.plot(allDMs, chan_smear, 'b', label="Channel Smearing")
# Add text above plot
settings = r"$f_{ctr}$ = %g MHz" % self.obs.fctr
if self.obs.dt < 1e-4:
settings += r", dt = %g $\mu$s" % (self.obs.dt * 1e6)
else:
settings += r", dt = %g ms" % (self.obs.dt * 1000)
settings += r", BW = %g MHz" % self.obs.BW
settings += r", N$_{chan}$ = %d" % self.obs.numchan
if self.numsub:
settings += r", N$_{sub}$ = %d" % self.numsub
if self.obs.numsamp:
settings += r", N$_{samp}$ = %d" % self.obs.numsamp
plt.figtext(0.05, 0.005, settings, \
ha='left', size='small')
plt.yscale('log')
plt.xlabel(r"Dispersion Measure (pc cm$^{-3}$)")
plt.ylabel(r"Smearing (s)")
plt.xlim(allDMs.min(), allDMs.max())
plt.ylim(0.3 * tot_smear.min(), 2.5 * tot_smear.max())
leg = plt.legend(loc='lower right')
plt.setp(leg.texts, size='small')
plt.setp(leg.legendHandles, linewidth=2)
if fn is not None:
# Save figure to file
plt.savefig(fn, orientation='landscape', papertype='letter')
else:
# Show figure interactively
def keypress(event):
if event.key in ['q', 'Q']:
plt.close()
fig.canvas.mpl_connect('key_press_event', keypress)
plt.show()
def plot(self, fn=None):
"""Generate a plot for this dedispersion plan.
Inputs:
fn: Filename to save plot as. If None, show plot interactively.
(Default: Show plot interactively.)
"""
fig = plt.figure(figsize=(11, 8.5))
ax = plt.axes()
stepDMs = []
# Plot each dedispersion step
for ii, (step, wf) in enumerate(zip(self.DDsteps, self.work_fracts)):
stepDMs.append(step.DMs)
DMspan = np.ptp(step.DMs)
loDM = step.DMs.min() + DMspan * 0.02
hiDM = step.DMs.max() - DMspan * 0.02
midDM = step.DMs.min() + DMspan * 0.5
# Sample time
plt.plot(
step.DMs,
np.zeros(step.numDMs) + step.ddplan.obs.dt * step.downsamp,
"#33CC33",
label=((ii and "_nolegend_") or "Sample Time (ms)"),
)
# DM stepsize smearing
plt.plot(
step.DMs,
np.zeros(step.numDMs) + step.BW_smearing,
"r",
label=((ii and "_nolegend_") or "DM Stepsize Smearing"),
)
if self.numsub:
plt.plot(
step.DMs,
np.zeros(step.numDMs) + step.sub_smearing,
"#993399",
label=((ii and "_nolegend_") or "Subband Stepsize Smearing (# passes)"),
)
plt.plot(step.DMs, step.tot_smear, "k", label=((ii and "_nolegend_") or "Total Smearing"))
# plot text
plt.text(
midDM,
1.1 * np.median(step.tot_smear),
"%d (%.1f%%)" % (step.numDMs, 100.0 * wf),
rotation="vertical",
size="small",
ha="center",
va="bottom",
)
plt.text(
loDM,
0.85 * step.ddplan.obs.dt * step.downsamp,
"%g" % (1000 * step.ddplan.obs.dt * step.downsamp),
size="small",
color="#33CC33",
ha="left",
)
plt.text(hiDM, 0.85 * step.BW_smearing, "%g" % step.dDM, size="small", color="r", ha="right")
if self.numsub:
plt.text(
midDM,
0.85 * step.sub_smearing,
"%g (%d)" % (step.dsubDM, step.numprepsub),
size="small",
color="#993399",
ha="center",
)
allDMs = np.concatenate(stepDMs)
chan_smear = psr_utils.dm_smear(allDMs, self.obs.chanwidth, self.obs.fctr)
bw_smear = psr_utils.dm_smear(ALLOW_DMSTEPS[0], self.obs.BW, self.obs.fctr)
tot_smear = np.sqrt(2 * self.obs.dt ** 2.0 + chan_smear ** 2.0 + bw_smear ** 2.0)
plt.plot(allDMs, tot_smear, "#FF9933", label="Optimal Smearing")
plt.plot(allDMs, chan_smear, "b", label="Channel Smearing")
# Add text above plot
settings = r"$f_{ctr}$ = %g MHz" % self.obs.fctr
if self.obs.dt < 1e-4:
settings += r", dt = %g $\mu$s" % (self.obs.dt * 1e6)
else:
settings += r", dt = %g ms" % (self.obs.dt * 1000)
settings += r", BW = %g MHz" % self.obs.BW
settings += r", N$_{chan}$ = %d" % self.obs.numchan
if self.numsub:
settings += r", N$_{sub}$ = %d" % self.numsub
if self.obs.numsamp:
settings += r", N$_{samp}$ = %d" % self.obs.numsamp
plt.figtext(0.05, 0.005, settings, ha="left", size="small")
plt.yscale("log")
plt.xlabel(r"Dispersion Measure (pc cm$^{-3}$)")
plt.ylabel(r"Smearing (s)")
plt.xlim(allDMs.min(), allDMs.max())
plt.ylim(0.3 * tot_smear.min(), 2.5 * tot_smear.max())
leg = plt.legend(loc="lower right")
plt.setp(leg.texts, size="small")
plt.setp(leg.legendHandles, linewidth=2)
if fn is not None:
# Save figure to file
plt.savefig(fn, orientation="landscape", papertype="letter")
else:
# Show figure interactively
def keypress(event):
if event.key in ["q", "Q"]:
plt.close()
fig.canvas.mpl_connect("key_press_event", keypress)
plt.show()
def Multi_Gaussian_Ratings(pfd, debug=False):
"""
Calculate ratings that depend on a fit of multiple Gaussian profile
components.
Parameters
----------
pfd : class
An instance of the prepfold.pfd class
debug : boolean
If True, plot the profile and best-fit Gaussians
Returns
-------
names : list
A list of ratings names
ratings : list
A list of ratings values
"""
# The ratings names
name1 = "Number_of_Gaussians"
name2 = "Pulse_Width_Rating"
# De-disperse the profile at the best DM
pfd.dedisperse(DM=pfd.bestdm, doppler=1)
# Internally rotate the data cube so that is aligned with best fold values
pfd.adjust_period()
# Get the best fold profile
profile = N.sum(N.sum(pfd.profs, axis=1), axis=0)
# Get the period from bestprof, if it exists
if hasattr(pfd.bestprof, "p0"):
P0 = pfd.bestprof.p0
# Otherwise use the barycentric or topocentric folding periods
else:
if pfd.bary_p1 != 0.0: P0 = pfd.bary_p1
elif pfd.topo_p1 != 0.0: P0 = pfd.topo_p1
# Get the center observing frequency
f_ctr = (pfd.hifreq + pfd.lofreq)/2
# Fit up to MAX_GAUSSIANS Gaussians to the profile
gaussian_params, red_chi2, ngaussians = \
PT.fit_gaussians_presto(profile, pfd.avgprof, N.sqrt(pfd.varprof),
MAX_GAUSSIANS, F_STAT_THRESHOLD)
if debug:
phases = N.arange(pfd.proflen, dtype=N.float)/pfd.proflen
gaussians = PT.make_gaussians_presto(gaussian_params, pfd.proflen)
PL.plot(phases, profile)
PL.plot(phases, gaussians)
PL.show()
# Create a list for storing the FWHMs of the Gaussians
fwhms = []
# If we were able to fit at least one Gaussian...
if ngaussians != 0:
# Get and store the FWHM for each Gaussian
for nn in xrange(ngaussians):
fwhms.append(gaussian_params[1+3*nn+1])
# Calculate the DM smearing time
dm_smear_sec = PU.dm_smear(pfd.bestdm, pfd.chan_wid, f_ctr)
# Calculate the expected minimum pulse width
width_phase = N.sqrt(dm_smear_sec**2 + pfd.dt**2)/P0
# Compare the expected minimum and actual pulse widths
width_ratio = width_phase/min(fwhms)
# If no Gaussians could be fit to the profile, store bogus metrics
else:
width_ratio = -1.0
# Store the ratings
rating1 = ngaussians
rating2 = width_ratio
return [name1,name2],[rating1,rating2]
def Multi_Gaussian_Ratings(pfd, debug=False):
"""
Calculate ratings that depend on a fit of multiple Gaussian profile
components.
Parameters
----------
pfd : class
An instance of the prepfold.pfd class
debug : boolean
If True, plot the profile and best-fit Gaussians
Returns
-------
names : list
A list of ratings names
ratings : list
A list of ratings values
"""
# The ratings names
name1 = "Number_of_Gaussians"
name2 = "Pulse_Width_Rating"
# De-disperse the profile at the best DM
pfd.dedisperse(DM=pfd.bestdm, doppler=1)
# Internally rotate the data cube so that is aligned with best fold values
pfd.adjust_period()
# Get the best fold profile
profile = N.sum(N.sum(pfd.profs, axis=1), axis=0)
# Get the period from bestprof, if it exists
if hasattr(pfd.bestprof, "p0"):
P0 = pfd.bestprof.p0
# Otherwise use the barycentric or topocentric folding periods
else:
if pfd.bary_p1 != 0.0: P0 = pfd.bary_p1
elif pfd.topo_p1 != 0.0: P0 = pfd.topo_p1
# Get the center observing frequency
f_ctr = (pfd.hifreq + pfd.lofreq) / 2
# Fit up to MAX_GAUSSIANS Gaussians to the profile
gaussian_params, red_chi2, ngaussians = \
PT.fit_gaussians_presto(profile, pfd.avgprof, N.sqrt(pfd.varprof),
MAX_GAUSSIANS, F_STAT_THRESHOLD)
if debug:
phases = N.arange(pfd.proflen, dtype=N.float) / pfd.proflen
gaussians = PT.make_gaussians_presto(gaussian_params, pfd.proflen)
PL.plot(phases, profile)
PL.plot(phases, gaussians)
PL.show()
# Create a list for storing the FWHMs of the Gaussians
fwhms = []
# If we were able to fit at least one Gaussian...
if ngaussians != 0:
# Get and store the FWHM for each Gaussian
for nn in xrange(ngaussians):
fwhms.append(gaussian_params[1 + 3 * nn + 1])
# Calculate the DM smearing time
dm_smear_sec = PU.dm_smear(pfd.bestdm, pfd.chan_wid, f_ctr)
# Calculate the expected minimum pulse width
width_phase = N.sqrt(dm_smear_sec**2 + pfd.dt**2) / P0
# Compare the expected minimum and actual pulse widths
width_ratio = width_phase / min(fwhms)
# If no Gaussians could be fit to the profile, store bogus metrics
else:
width_ratio = -1.0
# Store the ratings
rating1 = ngaussians
rating2 = width_ratio
return [name1, name2], [rating1, rating2]
counterpart = j
break
# print f1,fwant
if counterpart == False:
print "Didn't find counterpart frequency for %f" % f0
sub.set_weight(i,0.0)
continue
# else:
# print counterpart,f1
# print f0, f1
# sys.exit()
# print f0,f1
p = sub.get_folding_period()
smear = (psr_utils.dm_smear(dm,chbw,f0)
+ psr_utils.dm_smear(dm,chbw,f1)) / p
delay = (psr_utils.delay_from_DM(dm,f0)
- psr_utils.delay_from_DM(dm,f1)) / p
delay -= int(delay)
if delay<-0.5: delay += 1.0
if delay>0.5: delay -= 1.0
#print i, smear, delay
for pol in range(2):
if pol == 0:
correction = pol0correction/2.0
else:
def __init__(self, loDM, hiDM, obs, numsub=0, resolution=0.0, verbose=False):
"""DDplan object constructor.
Inputs:
loDM: Low DM to consider (pc cm-3)
hiDM: High DM to consider (pc cm-3)
obs: Observation object
numsub: Number of subbands to use (Default = Don't subband)
resolution: Acceptable smearing in ms (Default = Least smearing possible)
verbose: Print information to screen (Default = False)
"""
self.loDM = loDM
self.hiDM = hiDM
self.obs = obs
self.numsub = numsub
self.req_resolution = resolution * 0.001 # In seconds
self.current_downfact = self.obs.allow_factors[0]
self.current_dDM = ALLOW_DMSTEPS[0]
self.DDsteps = [] # list of dedispersion steps
# Calculate optimal smearing
self.calc_min_smearing(verbose=verbose)
# Calculate initial downsampling
while (self.obs.dt * self.get_next_downfact()) < self.resolution:
self.current_downfact = self.get_next_downfact()
if verbose:
print " New dt is %d x %.12g s = %.12g s" % (
self.current_downfact,
self.obs.dt,
self.current_downfact * self.obs.dt,
)
# Calculate the appropriate initial dDM
dDM = guess_DMstep(self.obs.dt * self.current_downfact, 0.5 * self.obs.BW, self.obs.fctr)
if verbose:
print "Best guess for optimal initial dDM is %.3f" % dDM
while self.get_next_dDM() < dDM:
self.current_dDM = self.get_next_dDM()
self.DDsteps.append(
DDstep(self, self.current_downfact, self.loDM, self.current_dDM, numsub=self.numsub, smearfact=SMEARFACT)
)
# Calculate the next steps
while self.DDsteps[-1].hiDM < self.hiDM:
# Determine the new downsample factor
self.current_downfact = self.get_next_downfact()
eff_dt = self.obs.dt * self.current_downfact
# Determine the new DM step
while psr_utils.dm_smear(0.5 * self.get_next_dDM(), self.obs.BW, self.obs.fctr) < FF * eff_dt:
self.current_dDM = self.get_next_dDM()
# Get the next step
self.DDsteps.append(
DDstep(
self,
self.current_downfact,
self.DDsteps[-1].hiDM,
self.current_dDM,
numsub=self.numsub,
smearfact=SMEARFACT,
)
)
# Calculate the predicted amount of time that will be spent in searching
# this batch of DMs as a fraction of the total
wfs = [step.numDMs / float(step.downsamp) for step in self.DDsteps]
self.work_fracts = np.asarray(wfs) / np.sum(wfs)
def apply_dm(inprof, period, dm, chan_width, freqs, tsamp, \
do_delay=True, do_smear=True, do_scatter=True,
verbose=True):
"""Given a profile apply DM delays, smearing, and scattering
within each channel as is appropriate for the given params.
Inputs:
inprof: The profile to modify.
period: The profiles period (in seconds)
dm: The DM (in pc cm-3)
chan_width: The width of each channel (in MHz)
freqs: The list of frequencies (in MHz)
tsamp: Sample time of the recipient filterbank file (in seconds).
do_delay: Boolean, if True apply DM delays to each channel.
The highest freq channel is not shifted. (Default: True)
do_smear: Boolean, if True apply DM smearing to each channel.
(Default: True)
do_scatter: Boolean, if True apply scattering to each channel.
(Default: True)
Outputs:
vecprof: The delayed and smeared VectorProfile.
"""
weq = inprof.get_equivalent_width()
nfreqs = len(freqs)
if verbose:
print "Applying DM to profile (DM = %.2f; %d channels)..." % \
(dm, nfreqs)
# A list of profiles, one for each channel
profiles = []
if dm <= 0:
warnings.warn("DM will not be applied because it is 0 (or smaller?!)")
do_delay = False
do_smear = False
do_scatter = False
if do_delay:
phasedelays = get_phasedelays(dm, freqs, period)
else:
phasedelays = np.zeros(nfreqs)
# Prepare for smear campaign
smeartimes = psr_utils.dm_smear(dm, abs(chan_width), freqs) # In seconds
smearphases = smeartimes/period
# Prepare to scatter
scattertimes = psr_utils.pulse_broadening(dm, freqs)*1e-3 # In seconds
scatterphases = scattertimes/period
if DEBUG:
for ichan, (freq, smear, scatt, delay) in \
enumerate(zip(freqs, smearphases, scatterphases, phasedelays)):
print " Chan #%d - Freq: %.3f MHz -- " \
"Smearing, scattering, delay (all in phase): " \
"%g, %g, %g" % (ichan, freq, smear, scatt, delay)
oldprogress = 0
sys.stdout.write(" %3.0f %%\r" % oldprogress)
sys.stdout.flush()
# ylim = None
# ylim2 = None
# ylim3 = None
# ylim4 = None
# ylim5 = None
for ii, (delayphs, smearphs, scattphs) in \
enumerate(zip(phasedelays, smearphases, scatterphases)):
#########
# DEBUG: plot all profiles
# plt.clf()
# ax = plt.subplot(5,1,1)
# inprof.plot()
# if ylim is not None:
# ax.set_ylim(ylim)
# else:
# ylim = ax.get_ylim()
if do_smear and not ((smearphs < 0.2*weq) or (smearphs < (tsamp/period))):
# Only smear if requested and smearing-phase is large enough
# bc = boxcar_factory(smearphs, delayphs)
# ax2 = plt.subplot(5,1,2,sharex=ax)
# bc.plot()
# if ylim2 is not None:
# ax2.set_ylim(ylim2)
# else:
# ylim2 = ax2.get_ylim()
if DEBUG:
print "Smearing"
tmpprof = inprof.smear(smearphs, delayphs, npts=NUMPOINTS)
else:
tmpprof = inprof.delay(delayphs)
phs = np.linspace(0, 1, NUMPOINTS+1)
tmpprof = SplineProfile(tmpprof(phs))
# ax3 = plt.subplot(5,1,3,sharex=ax)
# if ylim3 is not None:
# ax3.set_ylim(ylim3)
# else:
# ylim3 = ax3.get_ylim()
# tmpprof.plot()
if do_scatter and not ((scattphs < 0.2*weq) or (scattphs < (tsamp/period))):
# Only scatter if requested and scattering-phase is large enough
# ex = exponential_factory(scattphs)
# ax4 = plt.subplot(5,1,4,sharex=ax)
# ex.plot()
# if ylim4 is not None:
# ax4.set_ylim(ylim4)
# else:
# ylim4 = ax4.get_ylim()
if DEBUG:
print "Scattering"
tmpprof = tmpprof.scatter(scattphs, npts=NUMPOINTS)
# ax5 = plt.subplot(5,1,5,sharex=ax)
# tmpprof.plot()
# if ylim5 is not None:
# ax5.set_ylim(ylim5)
# else:
# ylim5 = ax5.get_ylim()
profiles.append(tmpprof)
# plt.xlim(0,1)
# plt.xlabel("Phase")
# plt.suptitle("Prof %d (%f MHz)" % (ii, freqs[ii]))
# plt.savefig("prof%d.png" % ii)
#########
# Print progress to screen
progress = int(100.0*ii/nfreqs)
if progress > oldprogress:
sys.stdout.write(" %3.0f %%\r" % progress)
sys.stdout.flush()
oldprogress = progress
sys.stdout.write("Done \n")
sys.stdout.flush()
dispersedprof = DispersedProfile(profiles, dm=dm, freqs=freqs, period=period,
intrinsic=inprof, delayed=do_delay)
return dispersedprof
def __init__(self, loDM, hiDM, obs, numsub=0, resolution=0.0, \
verbose=False):
"""DDplan object constructor.
Inputs:
loDM: Low DM to consider (pc cm-3)
hiDM: High DM to consider (pc cm-3)
obs: Observation object
numsub: Number of subbands to use (Default = Don't subband)
resolution: Acceptable smearing in ms (Default = Least smearing possible)
verbose: Print information to screen (Default = False)
"""
self.loDM = loDM
self.hiDM = hiDM
self.obs = obs
self.numsub = numsub
self.req_resolution = resolution * 0.001 # In seconds
self.current_downfact = self.obs.allow_factors[0]
self.current_dDM = ALLOW_DMSTEPS[0]
self.DDsteps = [] # list of dedispersion steps
# Calculate optimal smearing
self.calc_min_smearing(verbose=verbose)
# Calculate initial downsampling
while (self.obs.dt * self.get_next_downfact()) < self.resolution:
self.current_downfact = self.get_next_downfact()
if verbose:
print " New dt is %d x %.12g s = %.12g s" % \
(self.current_downfact, self.obs.dt, \
self.current_downfact*self.obs.dt)
# Calculate the appropriate initial dDM
dDM = guess_DMstep(self.obs.dt*self.current_downfact, \
0.5*self.obs.BW, self.obs.fctr)
if verbose:
print "Best guess for optimal initial dDM is %.3f" % dDM
while (self.get_next_dDM() < dDM):
self.current_dDM = self.get_next_dDM()
self.DDsteps.append(DDstep(self, self.current_downfact, \
self.loDM, self.current_dDM, \
numsub=self.numsub, \
smearfact=SMEARFACT))
# Calculate the next steps
while self.DDsteps[-1].hiDM < self.hiDM:
# Determine the new downsample factor
self.current_downfact = self.get_next_downfact()
eff_dt = self.obs.dt * self.current_downfact
# Determine the new DM step
while psr_utils.dm_smear(0.5*self.get_next_dDM(), self.obs.BW, \
self.obs.fctr) < FF*eff_dt:
self.current_dDM = self.get_next_dDM()
# Get the next step
self.DDsteps.append(DDstep(self, self.current_downfact, \
self.DDsteps[-1].hiDM, \
self.current_dDM, \
numsub=self.numsub, \
smearfact=SMEARFACT))
# Calculate the predicted amount of time that will be spent in searching
# this batch of DMs as a fraction of the total
wfs = [step.numDMs / float(step.downsamp) for step in self.DDsteps]
self.work_fracts = np.asarray(wfs) / np.sum(wfs)
def apply_dm(inprof, period, dm, chan_width, freqs, tsamp, \
do_delay=True, do_smear=True, do_scatter=True,
verbose=True):
"""Given a profile apply DM delays, smearing, and scattering
within each channel as is appropriate for the given params.
Inputs:
inprof: The profile to modify.
period: The profiles period (in seconds)
dm: The DM (in pc cm-3)
chan_width: The width of each channel (in MHz)
freqs: The list of frequencies (in MHz)
tsamp: Sample time of the recipient filterbank file (in seconds).
do_delay: Boolean, if True apply DM delays to each channel.
The highest freq channel is not shifted. (Default: True)
do_smear: Boolean, if True apply DM smearing to each channel.
(Default: True)
do_scatter: Boolean, if True apply scattering to each channel.
(Default: True)
Outputs:
vecprof: The delayed and smeared VectorProfile.
"""
weq = inprof.get_equivalent_width()
nfreqs = len(freqs)
if verbose:
print("Applying DM to profile (DM = %.2f; %d channels)..." % \
(dm, nfreqs))
# A list of profiles, one for each channel
profiles = []
if dm <= 0:
warnings.warn("DM will not be applied because it is 0 (or smaller?!)")
do_delay = False
do_smear = False
do_scatter = False
if do_delay:
phasedelays = get_phasedelays(dm, freqs, period)
else:
phasedelays = np.zeros(nfreqs)
# Prepare for smear campaign
smeartimes = psr_utils.dm_smear(dm, abs(chan_width), freqs) # In seconds
smearphases = smeartimes/period
# Prepare to scatter
scattertimes = psr_utils.pulse_broadening(dm, freqs)*1e-3 # In seconds
scatterphases = scattertimes/period
if DEBUG:
for ichan, (freq, smear, scatt, delay) in \
enumerate(zip(freqs, smearphases, scatterphases, phasedelays)):
print(" Chan #%d - Freq: %.3f MHz -- " \
"Smearing, scattering, delay (all in phase): " \
"%g, %g, %g" % (ichan, freq, smear, scatt, delay))
oldprogress = 0
sys.stdout.write(" %3.0f %%\r" % oldprogress)
sys.stdout.flush()
# ylim = None
# ylim2 = None
# ylim3 = None
# ylim4 = None
# ylim5 = None
for ii, (delayphs, smearphs, scattphs) in \
enumerate(zip(phasedelays, smearphases, scatterphases)):
#########
# DEBUG: plot all profiles
# plt.clf()
# ax = plt.subplot(5,1,1)
# inprof.plot()
# if ylim is not None:
# ax.set_ylim(ylim)
# else:
# ylim = ax.get_ylim()
if do_smear and not ((smearphs < 0.2*weq) or (smearphs < (tsamp/period))):
# Only smear if requested and smearing-phase is large enough
# bc = boxcar_factory(smearphs, delayphs)
# ax2 = plt.subplot(5,1,2,sharex=ax)
# bc.plot()
# if ylim2 is not None:
# ax2.set_ylim(ylim2)
# else:
# ylim2 = ax2.get_ylim()
if DEBUG:
print("Smearing")
tmpprof = inprof.smear(smearphs, delayphs, npts=NUMPOINTS)
else:
tmpprof = inprof.delay(delayphs)
phs = np.linspace(0, 1, NUMPOINTS+1)
tmpprof = SplineProfile(tmpprof(phs))
# ax3 = plt.subplot(5,1,3,sharex=ax)
# if ylim3 is not None:
# ax3.set_ylim(ylim3)
# else:
# ylim3 = ax3.get_ylim()
# tmpprof.plot()
if do_scatter and not ((scattphs < 0.2*weq) or (scattphs < (tsamp/period))):
# Only scatter if requested and scattering-phase is large enough
# ex = exponential_factory(scattphs)
# ax4 = plt.subplot(5,1,4,sharex=ax)
# ex.plot()
# if ylim4 is not None:
# ax4.set_ylim(ylim4)
# else:
# ylim4 = ax4.get_ylim()
if DEBUG:
print("Scattering")
tmpprof = tmpprof.scatter(scattphs, npts=NUMPOINTS)
# ax5 = plt.subplot(5,1,5,sharex=ax)
# tmpprof.plot()
# if ylim5 is not None:
# ax5.set_ylim(ylim5)
# else:
# ylim5 = ax5.get_ylim()
profiles.append(tmpprof)
# plt.xlim(0,1)
# plt.xlabel("Phase")
# plt.suptitle("Prof %d (%f MHz)" % (ii, freqs[ii]))
# plt.savefig("prof%d.png" % ii)
#########
# Print progress to screen
progress = int(100.0*ii/nfreqs)
if progress > oldprogress:
sys.stdout.write(" %3.0f %%\r" % progress)
sys.stdout.flush()
oldprogress = progress
sys.stdout.write("Done \n")
sys.stdout.flush()
dispersedprof = DispersedProfile(profiles, dm=dm, freqs=freqs, period=period,
intrinsic=inprof, delayed=do_delay)
return dispersedprof
def __init__(self, ddplan, downsamp, loDM, dDM, \
numDMs=0, numsub=0, smearfact=2.0):
"""DDstep object constructor.
Inputs:
ddplan: DDplan object this DDstep is part of.
downsamp: Downsampling factor.
loDM: Low DM edge of this DDstep (pc cm-3)
dDM: DM step size to use (pc cm-3)
numDMs: Number of DMs. If numDMs=0 compute numDMs
based on DM range and spacing.
(Default: 0)
numsub: Number of subbands to use. (Default: no subbanding).
smearfact: Allowable smearing in a single channel, relative
to other smearing contributions (Default: 2.0)
"""
self.ddplan = ddplan
self.downsamp = downsamp
self.loDM = loDM
self.dDM = dDM
self.numsub = numsub
self.BW_smearing = psr_utils.dm_smear(dDM*0.5, self.ddplan.obs.BW, \
self.ddplan.obs.fctr)
self.numprepsub = 0
if numsub:
# Calculate the maximum subband smearing we can handle
DMs_per_prepsub = 2
while True:
next_dsubDM = (DMs_per_prepsub + 2) * dDM
next_ss = psr_utils.dm_smear(next_dsubDM*0.5, \
self.ddplan.obs.BW/numsub, \
self.ddplan.obs.fctr)
# The 0.8 is a small fudge factor to make sure that the subband
# smearing is always the smallest contribution
if next_ss > 0.8*min(self.BW_smearing, \
self.ddplan.obs.dt*self.downsamp):
self.dsubDM = DMs_per_prepsub * dDM
self.DMs_per_prepsub = DMs_per_prepsub
self.sub_smearing = psr_utils.dm_smear(self.dsubDM*0.5,
self.ddplan.obs.BW/self.numsub, \
self.ddplan.obs.fctr)
break
DMs_per_prepsub += 2
else:
self.dsubDM = dDM
self.sub_smearing = 0.0
# Calculate the nominal DM to move to the next step
cross_DM = self.DM_for_smearfact(smearfact)
if cross_DM > self.ddplan.hiDM:
cross_DM = self.ddplan.hiDM
if numDMs == 0:
self.numDMs = int(np.ceil((cross_DM - self.loDM) / self.dDM))
if numsub:
self.numprepsub = int(
np.ceil(self.numDMs * self.dDM / self.dsubDM))
self.numDMs = self.numprepsub * DMs_per_prepsub
else:
self.numDMs = numDMs
self.hiDM = loDM + self.numDMs * dDM
self.DMs = np.arange(self.numDMs, dtype='d') * self.dDM + self.loDM
# Calculate a few more smearing values
self.chan_smear = psr_utils.dm_smear(self.DMs, \
self.ddplan.obs.chanwidth, \
self.ddplan.obs.fctr)
self.tot_smear = np.sqrt((self.ddplan.obs.dt)**2.0 + \
(self.ddplan.obs.dt*self.downsamp)**2.0 + \
self.BW_smearing**2.0 + \
self.sub_smearing**2.0 + \
self.chan_smear**2.0)