def get_ml_toa(fits_fn, prof_mod, parfile, scope='swift', print_offs=None,
frequency=None, epoch=None, sim=False, bg_counts=0, Emin=None,
Emax=None, gauss_err=False, tempo2=False, debug=False,
correct_pf=False, split_num=None, split_orbits=False):
print_timings = False # if want to print summary of runtime
fits = pyfits.open(fits_fn)
t = smu.fits2times(fits_fn, scope=scope, Emin=Emin, Emax=Emax)
#if scope != 'chandra':
# exposure = fits[0].header['EXPOSURE']
try:
obsid = fits[0].header['OBS_ID']
except KeyError:
obsid = os.path.basename(fits_fn)
if bg_counts < 0:
bg_scale = -1.0*bg_counts
bg_fits_fn = fits_fn.replace('reg','bgreg')
bg_fits = pyfits.open(bg_fits_fn)
bg_counts = int(bg_fits[1].header['NAXIS2'] * bg_scale)
print 'BG Counts:',bg_counts
bg_fits.close()
if frequency and epoch:
par = lambda: None
par.epoch = epoch
par.f0 = frequency
par.fdots = [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0]
else:
par = PSRpar(parfile)
# split times into multiple arrays if needed
if split_orbits:
dt = t[1:] - t[:-1]
splits = np.where(dt > 0.0116)[0] # 1 ks in days
if len(splits):
ts = [ t[:splits[0]] ]
for i in range(len(splits)-1):
ts.append(t[splits[i]+1:splits[i+1]])
ts.append(t[splits[-1]+1:])
else:
ts = np.atleast_2d(t)
elif split_num:
remainder = len(t) % split_num
if remainder:
sys.stderr.write("Warning: Number of events in %s not divisable by %d. " \
"Dropping last %d events.\n" % (obsid, split_num, remainder))
ts = np.split(t[:-remainder],split_num)
else:
ts = np.split(t,split_num)
else:
ts = np.atleast_2d(t)
if len(ts) > 1 and debug:
plt.figure()
for t in ts:
nbins = int((t[-1] - t[0]) * 8640.0)
hist = np.histogram(t,bins=nbins)
plt.plot(hist[1][:-1],hist[0],c='b')
plt.axvline(t[0],ls='--',c='k',lw=2)
plt.axvline(t[-1],ls='-',c='k',lw=2)
plt.show()
for i,t in enumerate(ts):
sys.stderr.write('Measuring TOA #%d for %s\n' % (i+1,obsid))
phases = smu.times2phases(t, parfile)
if correct_pf:
old_model, new_model, corr_folded = correct_model(phases,prof_mod)
maxoff, error = calc_toa_offset(phases,prof_mod.prof_mod,sim_err=sim,bg_counts=bg_counts, gauss_err=gauss_err, debug=debug)
midtime = (t[-1]+t[0])/2.0
p_mid = 1.0/psr_utils.calc_freq(midtime, par.epoch, par.f0, par.fdots[0], par.fdots[1], par.fdots[2], par.fdots[3],
par.fdots[4], par.fdots[5], par.fdots[6], par.fdots[7], par.fdots[8])
t0 = psr_utils.calc_t0(midtime, par.epoch, par.f0, par.fdots[0], par.fdots[1], par.fdots[2], par.fdots[3],
par.fdots[4], par.fdots[5], par.fdots[6], par.fdots[7], par.fdots[8])
t0i = int(t0)
t0f = t0 - t0i
toaf = t0f + maxoff*p_mid / SECPERDAY
newdays = int(np.floor(toaf))
if tempo2:
psr_utils.write_tempo2_toa(t0i+newdays, toaf-newdays, error*p_mid*1.0e6, 0000, 0.0, name=obsid)
else:
psr_utils.write_princeton_toa(t0i+newdays, toaf-newdays, error*p_mid*1.0e6, 0000, 0.0, name=obsid)
if print_offs:
offs_file = open(print_offs,'a')
#print "\t",error*p_mid*1.0e6,"\t",exposure # this was for checking uncertainties scaling with exposure time
offs_file.write(fits_fn + "\t" + str(maxoff) + "\t" + str(error) + "\n")
#print obsid,"\tOffset:",maxoff,"+/-",error
offs_file.close()
fits.close()
#double check PF correction with measuring binned model pulsed fraction
if correct_pf and debug:
plt.figure()
nbins = len(corr_folded[0])
uncertainties = np.sqrt(corr_folded[0])
area = np.sum(corr_folded[0],dtype='float')/nbins
plt.step(corr_folded[1][:-1],np.roll(corr_folded[0]/area,int(1.0-maxoff*nbins)),where='mid')
plt.errorbar(corr_folded[1][:-1],np.roll(corr_folded[0]/area,int(1.0-maxoff*nbins)),uncertainties/area,fmt='ko')
model_x = np.linspace(0,1,100)
plt.plot(model_x,old_model(model_x),label='uncorrected')
plt.plot(model_x,new_model(model_x),label='corrected')
plt.legend()
plt.show()
if print_timings:
global calcprobtime
global logsumtime
global integratetime
sys.stderr.write('\tCalc Prob: %f s\n' % calcprobtime)
sys.stderr.write('\tLog Sum: %f s\n' % logsumtime)
sys.stderr.write('\tIntegrate Norm: %f s\n' % integratetime)
# period.
dd_phs_2 = subdelays2[jj] * (1.0 / p - 1.0 / p_dedisp)
# Sum up several phase shifts
tau_tot = Num.fmod(tau + sumsubdelays_phs[jj] + dd_phs_2 + 3.0,
1.0)
if (tau_tot > 0.5): tau_tot -= 1.0
# Send the TOA to STDOUT
toaf = t0f + (tau_tot * p + offset) / SECPERDAY
newdays = int(Num.floor(toaf))
if t2format:
psr_utils.write_tempo2_toa(t0i + newdays,
toaf - newdays,
tau_err * p * 1000000.0,
sumsubfreqs[jj],
0.0,
name=fold_pfd.pfd_filename,
obs=obs)
else:
psr_utils.write_princeton_toa(t0i + newdays,
toaf - newdays,
tau_err * p * 1000000.0,
sumsubfreqs[jj],
0.0,
obs=obs)
if (otherouts):
sys.stderr.write(
"FFTFIT results: b = %.4g +/- %.4g SNR = %.4g +/- %.4g"
% (b, errb, snr, esnr))
tau = psr_utils.measure_phase_corr(prof, template)
# This needs to be changed
tau_err = 0.1/len(prof)
# Calculate correction for dedispersion to true channel
# center freqs that used a slightly different pulse
# period.
dd_phs_2 = subdelays2[jj] * (1.0/p - 1.0/p_dedisp)
# Sum up several phase shifts
tau_tot = Num.fmod(tau+sumsubdelays_phs[jj]+dd_phs_2+3.0, 1.0)
if (tau_tot > 0.5): tau_tot -= 1.0
# Send the TOA to STDOUT
toaf = t0f + (tau_tot*p + offset)/SECPERDAY
newdays = int(Num.floor(toaf))
if t2format:
psr_utils.write_tempo2_toa(t0i+newdays, toaf-newdays,
tau_err*p*1000000.0,
sumsubfreqs[jj], 0.0, name=fold_pfd.pfd_filename, obs=obs)
else:
psr_utils.write_princeton_toa(t0i+newdays, toaf-newdays,
tau_err*p*1000000.0,
sumsubfreqs[jj], 0.0, obs=obs)
if (otherouts):
sys.stderr.write("FFTFIT results: b = %.4g +/- %.4g SNR = %.4g +/- %.4g" %
(b, errb, snr, esnr))
except ValueError, fftfit.error:
pass