# *** set up calibration
calarr = {}
cal_fi = arange(0.0, nchan / 2, 0.5)
fi = arange(freqs.shape[0])
cal_freqs = NP.interp(cal_fi, fi, freqs)
curr_gains = 0.25 * NP.ones((n_antennas, len(cal_freqs)), dtype=NP.complex64)
freq_ave = bchan
for pol in pols:
calarr[pol] = EPICal.cal(cal_freqs,
antpos_info['positions'],
pol=pol,
sim_mode=False,
n_iter=cal_iter,
damping_factor=0.7,
inv_gains=False,
sky_model=sky_model,
freq_ave=bchan,
exclude_autos=True,
phase_fit=False,
curr_gains=curr_gains,
ref_ant=5,
flatten_array=True,
n_cal_sources=1)
if scramble_gains > 0:
for i in NP.arange(NP.ceil(NP.float(nchan) / freq_ave)):
mini = i * freq_ave
maxi = NP.min((nchan, (i + 1) * freq_ave))
calarr[pol].curr_gains[:, mini:maxi] += (
NP.random.normal(0, NP.sqrt(scramble_gains), (n_antennas, 1)) +
1j * NP.random.normal(0, NP.sqrt(scramble_gains),
(n_antennas, 1)))
sky_model[:, :, 0:3] = skypos.reshape(n_src, 1, 3)
sky_model[:, :, 3] = src_flux.reshape(n_src, 1)
#### set up calibration
calarr = {}
calarr_full = {}
calarr_one = {}
ant_pos = ant_info[:, 1:] # I'll let the cal class put it in wavelengths.
for pol in ["P1", "P2"]:
calarr_full[pol] = EPICal.cal(
freqs,
antpos_info["positions"],
pol=pol,
sim_mode=True,
n_iter=cal_iter,
damping_factor=damping_factor,
inv_gains=False,
sky_model=sky_model,
exclude_autos=True,
n_cal_sources=1,
)
calarr_full[pol].sim_gains = NP.ones(calarr_full[pol].sim_gains.shape, dtype=NP.complex64)
calarr_one[pol] = EPICal.cal(
freqs,
antpos_info["positions"],
pol=pol,
sim_mode=True,
n_iter=cal_iter,
damping_factor=damping_factor,
inv_gains=False,
sky_model=sky_model[-1, :, :].reshape(1, nchan, 4),
sky_model = NP.zeros((n_src, nchan, 4))
sky_model[:, :, 0:3] = skypos.reshape(n_src, 1, 3)
sky_model[:, :, 3] = src_flux.reshape(n_src, 1)
# sky_model=sky_model[0,:,:].reshape(1,nchan,4)
sky_model = sky_model[-ind:, :, :]
# *** set up calibration
calarr = {}
ant_pos = ant_info[:, 1:] # I'll let the cal class put it in wavelengths.
# auto_noise_model = rxr_noise
for pol in ['P1', 'P2']:
calarr[pol] = EPICal.cal(freqs, ant_pos, pol=pol, sim_mode=True, n_iter=cal_iter,
damping_factor=0.35, inv_gains=False, sky_model=sky_model,
exclude_autos=True, conv_max_try=1)
# Create array of gains to watch them change
ncal = itr / cal_iter
cali = 0
gain_stack = NP.zeros((ncal + 1, ant_info.shape[0], nchan), dtype=NP.complex64)
for i in xrange(itr):
print i
# simulate
E_timeseries_dict = SIM.stochastic_E_timeseries(f_center, nchan / 2, 2 * channel_width,
flux_ref=src_flux, skypos=skypos,
antpos=antpos_info['positions'],
tshift=False)
sky_model = NP.zeros((n_src,nchan,4))
sky_model[:,:,0:3] = skypos.reshape(n_src,1,3)
sky_model[:,:,3] = src_flux.reshape(n_src,1)
#sky_model=sky_model[0,:,:].reshape(1,nchan,4)
sky_model=sky_model[-ind:,:,:]
#### set up calibration
calarr={}
ant_pos = ant_info[:,1:] # I'll let the cal class put it in wavelengths.
auto_noise_model = rxr_noise
for pol in ['P1','P2']:
#calarr[pol] = EPICal.cal(ant_pos,freqs,n_iter=cal_iter,sim_mode=True,sky_model=sky_model,gain_factor=0.5,pol=pol,cal_method='multi_source',inv_gains=False)
calarr[pol] = EPICal.cal(freqs,antpos_info['positions'],pol=pol,sim_mode=True,n_iter=cal_iter,damping_factor=0.35,inv_gains=False,sky_model=sky_model,exclude_autos=True,n_cal_sources=1)
# Create array of gains to watch them change
ncal=itr/cal_iter
cali=0
gain_stack = NP.zeros((ncal+1,ant_info.shape[0],nchan),dtype=NP.complex64)
for i in xrange(itr):
print i
# simulate
E_timeseries_dict = SIM.stochastic_E_timeseries(f_center, nchan/2, 2*channel_width,
flux_ref=src_flux, skypos=skypos, antpos=antpos_info['positions'],tshift=False)
timestamp = str(DT.datetime.now())
update_info={}
skypos = NP.hstack((lmrad * NP.cos(lmang), lmrad * NP.sin(lmang)))
nvect = NP.sqrt(1.0-NP.sum(skypos**2, axis=1)).reshape(-1,1)
skypos = NP.hstack((skypos,nvect))
sky_model = NP.zeros((n_src,nchan,4))
sky_model[:,:,0:3] = skypos.reshape(n_src,1,3)
sky_model[:,:,3] = src_flux.reshape(n_src,1)
#### set up calibration
calarr={}
calarr_full={}
calarr_one={}
ant_pos = ant_info[:,1:] # I'll let the cal class put it in wavelengths.
for pol in ['P1','P2']:
calarr_full[pol] = EPICal.cal(freqs,antpos_info['positions'],pol=pol,sim_mode=True,n_iter=cal_iter,damping_factor=damping_factor,inv_gains=False,sky_model=sky_model,exclude_autos=True,n_cal_sources=1)
calarr_full[pol].sim_gains = NP.ones(calarr_full[pol].sim_gains.shape,dtype=NP.complex64)
calarr_one[pol] = EPICal.cal(freqs,antpos_info['positions'],pol=pol,sim_mode=True,n_iter=cal_iter,damping_factor=damping_factor,inv_gains=False,sky_model=sky_model[-1,:,:].reshape(1,nchan,4),exclude_autos=True,n_cal_sources=1)
calarr_one[pol].sim_gains = NP.ones(calarr_one[pol].sim_gains.shape,dtype=NP.complex64)
visdata = NP.zeros((n_antennas,n_antennas,nchan),dtype=NP.complex64)
for i in xrange(itr):
if i == 1:
print 'rxr_noise=',rxr_noise
print 'cal_iter=',cal_iter
# simulate
if i == 0:
E_timeseries_dict = SIM.stochastic_E_timeseries(f_center, nchan/2, 2*channel_width,
flux_ref=src_flux, skypos=skypos, antpos=antpos_info['positions'],tshift=False,verbose=False)
sky_model[:,:,0:3] = skypos.reshape(n_src,1,3)
sky_model[:,:,3] = src_flux.reshape(n_src,1)
#### set up calibration
calarr={}
cal_fi=arange(0.0,nchan/2,0.5)
fi=arange(freqs.shape[0])
cal_freqs = NP.interp(cal_fi,fi,freqs)
# auto noise term
auto_noise_model = 0.25 * NP.sum(sky_model[:,0,3]) # roughly rxr:sky based on Ellingson, 2013
#auto_noise_model=0.0
curr_gains = 0.25*NP.ones((n_antennas,len(cal_freqs)),dtype=NP.complex64)
freq_ave = bchan
for pol in pols:
calarr[pol] = EPICal.cal(cal_freqs,antpos_info['positions'],pol=pol,sim_mode=False,n_iter=cal_iter,damping_factor=0.7,inv_gains=False,sky_model=sky_model,freq_ave=bchan,exclude_autos=True,phase_fit=False,curr_gains=curr_gains,ref_ant=5,flatten_array=True,n_cal_sources=1)
if scramble_gains > 0:
for i in NP.arange(NP.ceil(NP.float(nchan)/freq_ave)):
mini = i*freq_ave
maxi = NP.min((nchan,(i+1)*freq_ave))
calarr[pol].curr_gains[:,mini:maxi] += NP.random.normal(0,NP.sqrt(scramble_gains),(n_antennas,1)) + 1j * NP.random.normal(0,NP.sqrt(scramble_gains),(n_antennas,1))
if initial_gains_file is not None:
calarr['P1'].curr_gains = NP.load(initial_gains_file)
# Create array of gains to watch them change
ncal=max_n_timestamps/cal_iter
cali=0
gain_stack = NP.zeros((ncal+1,ant_info.shape[0],nchan),dtype=NP.complex64)
amp_stack = NP.zeros((ncal+1,nchan),dtype=NP.float64)
amp_full_stack = NP.zeros((max_n_timestamps,nchan),dtype=NP.float64)
frac_flux=NP.sum(src_flux[-ind:])/tot_flux
sky_model = NP.zeros((n_src,nchan,4))
sky_model[:,:,0:3] = skypos.reshape(n_src,1,3)
sky_model[:,:,3] = src_flux.reshape(n_src,1)
#sky_model=sky_model[0,:,:].reshape(1,nchan,4)
sky_model=sky_model[-ind:,:,:]
#### set up calibration
calarr={}
ant_pos = ant_info[:,1:] # I'll let the cal class put it in wavelengths.
#auto_noise_model = rxr_noise
for pol in ['P1','P2']:
calarr[pol] = EPICal.cal(freqs,ant_pos,pol=pol,sim_mode=True,n_iter=cal_iter,damping_factor=0.35,inv_gains=False,sky_model=sky_model,exclude_autos=False)
# Create array of gains to watch them change
ncal=itr/cal_iter
cali=0
gain_stack = NP.zeros((ncal+1,ant_info.shape[0],nchan),dtype=NP.complex64)
PLT.ion()
PLT.show()
for i in xrange(itr):
print i
# simulate
E_timeseries_dict = SIM.stochastic_E_timeseries(f_center, nchan/2, 2*channel_width,
flux_ref=src_flux, skypos=skypos, antpos=antpos_info['positions'],tshift=False)
