def process(self):
'''execute the loading of raw data from disk
'''
t2 = self.t2
table = self.table
loop = self.loop
# first file requires special synchronization
# this is the rule that determines that it is the first file
first = 'first' if all([t2 == 0, table == 0, loop == 0]) else False
# Load everything for the given t2, table, and loop value
analogs = nudie.load_analogtxt(self.batch_name, self.batch_path, t2,
table, loop)
cdata = nudie.load_camera_file(self.batch_name,
self.batch_path,
t2,
table,
loop,
force_uint16=True)
# Synchronize it, trimming some frames in the beginning and end
data, (a1, a2) = nudie.trim_all(*nudie.synchronize_daq_to_camera(\
cdata, analog_channels=analogs, which_file=first),
trim_to=self.trim_to)
data = data.astype(float) # convert to float64 before manipulating it!
start_idxs, period = nudie.detect_table_start(a1)
# determine where the shutter is
shutter_open, shutter_closed, duty_cycle = nudie.determine_shutter_shots(
data)
shutter_info = {
'last open idx': shutter_open,
'first closed idx': shutter_closed
}
tags = nudie.tag_phases(start_idxs,
period,
tags=self.phase_cycles,
nframes=data.shape[1],
shutter_info=shutter_info)
self.tags = tags
self.camera_data = data
achan = namedtuple('analog_channels', ['a1', 'a2'])
self.analog_channels = achan(a1, a2)
return self
def test_general_case(offset, repeat, nwave):
''' tests whether many waveforms, several phase case works'''
duty_cycle = 0.5
transition_width = 3
nframes = 100
waveform_repeat = repeat
nwaveforms = nwave
tags = ['one', 'two', 'three']
nphases = len(tags)
data = np.zeros((nframes, 4), dtype=object)
analog1 = np.zeros((nframes, ), dtype=float)
states = it.cycle([(rep, phase, waveform) \
for waveform in range(nwaveforms) \
for phase in tags \
for rep in range(waveform_repeat)])
#states, tmp = it.tee(states, 2)
#test = list(it.islice(tmp, offset, 100))
# set shutter
last_on = int(nframes * duty_cycle)
shutter_it = it.chain(it.repeat('shutter open', last_on),
it.repeat(None, transition_width),
it.repeat('shutter closed'))
for i, (rep, phase, waveform), shut in \
zip(range(nframes), it.islice(states, offset, None), shutter_it):
data[i] = (rep, phase, waveform, shut)
if waveform == 0 and phase == tags[0]: analog1[i] = 2.5
start, period = nudie.detect_table_start(analog1, waveform_repeat)
shutter_info = {
'last open idx': slice(None, last_on + 1),
'first closed idx': slice(last_on + transition_width, None)
}
tagged = nudie.tag_phases(start, period, tags, nframes, \
waveform_repeat=waveform_repeat, shutter_info=shutter_info)
for i, (r, p, w, s) in enumerate(data):
if s is None: continue # skip transition frames
assert i in tagged[r][w][p][s]
def test_nrepeat(inp):
nphases = 6
nwaveforms = 10
nframes = 1000
duty_cycle = 0.5
reps = inp[0]
offset = 2
period = nphases * nwaveforms * reps
table_start = np.array(
[period * i + offset for i in range(nframes // period)])
shutter_info = {
'last open idx': slice(None, int(nframes * duty_cycle)),
'first closed idx': slice(int(nframes * duty_cycle) + 5, None)
}
tags = nudie.tag_phases(table_start, period, range(nphases), \
nframes, waveform_repeat=reps, shutter_info=shutter_info)
assert len(tags) == reps, 'incorrect detected reps'
def test_exact_period():
# Fixed bug where an assertion would fail if table_start_detect[0]
# indicated that the first camera frame was the first phase in the
# waveform
period = 4
num_phases = 4
nframes = 100
duty_cycle = 0.5
table_start = np.array([4 * i for i in range(nframes // period)],
dtype=int)
shutter_info = {
'last open idx': slice(None, int(nframes * duty_cycle)),
'first closed idx': slice(int(nframes * duty_cycle) + 5, None)
}
tags = nudie.tag_phases(table_start, period, range(num_phases), \
nframes, shutter_info=shutter_info)
# check that the first phase of the only waveform with the shutter open
# starts at index 0
assert tags[0][0][0]['shutter open'][0] == 0
def test_several_waveforms_one_phase_offset(period, offset):
num_phases = 1
nframes = 1000
duty_cycle = 0.5
table_start = np.array(
[period * i + offset for i in range(nframes // period)])
shutter_info = {
'last open idx': slice(None, int(nframes * duty_cycle)),
'first closed idx': slice(int(nframes * duty_cycle) + 5, None)
}
tags = nudie.tag_phases(table_start, period, range(num_phases), \
nframes, shutter_info=shutter_info)
assert len(tags) == 1, 'waveform repeat detected incorrectly'
assert len(tags[0]) == period, 'nwaveforms detected incorrectly'
assert len(tags[0][0]) == num_phases, 'num_phases detected incorrectly'
assert tags[0][0][0]['shutter open'][0] == offset
# first repeat, offset'th waveform, first phase, shutter open, and first
# index
assert tags[0][-offset % period][0]['shutter open'][0] == 0, \
'incorrect first waveform'
def run(tg_name,
tg_batch,
when='today',
wavelengths=None,
plot=False,
pad_to=2048,
prd_est=850.,
lo_width=200,
dc_width=200,
gaussian_power=2.,
analysis_path='./analyzed',
datapath=None):
if plot:
import matplotlib as mpl
import matplotlib.pyplot as plt
mpl.rcParams['figure.figsize'] = (16, 12)
mpl.use('Qt4Agg')
nrepeat = 1 # how many times each waveform is repeated in the camera file. Assumed to be one
waveforms_per_table = 1
npixels = 1340
trim_to = 3, -3
# change datapath
if datapath is None:
datapath = nudie.data_folder
# load up tg data to use
tg_info = next(
nudie.load_job(job_name=tg_name,
batch_set=[tg_batch],
when=when,
data_path=datapath))
phase_cycles = ['none1', 'zero', 'none2', 'pipi']
# set current batch directory
current_path = Path(tg_info['batch_path'])
# generate hdf filename based on data date
analysis_folder = Path(analysis_path)
# create folder if it doesn't exist
if not analysis_folder.exists():
analysis_folder.mkdir(parents=True)
save_path = analysis_folder / (tg_info['batch_name'] + '.h5')
# remove data file if it exists
if save_path.exists(): save_path.unlink()
with h5py.File(str(save_path), 'w') as sf:
# initialize groups
sf.create_group('axes')
shape = (tg_info['nt2'], npixels)
sf.create_dataset('raw transient-grating', shape, dtype=complex)
try:
wl = load_wavelengths(current_path.parent / wavelengths)
except FileNotFoundError as e:
nudie.log.error('Could not load wavelength calibration file!')
raise e
# define gaussian
def gaussian2(w, x0, x):
c = 4 * np.log(2)
ξ = x - x0
return np.exp(-c * (ξ / w)**(2 * gaussian_power))
for loop, t2, table in it.product(tg_info['loop_range'],
tg_info['t2_range'],
tg_info['table_range']):
if any([loop != 0, table != 0]):
raise NotImplementedError('Code is not setup to handle multiple ' +\
'loops or tables.')
# first file requires special synchronization
# this is the rule that determines that it is the first file
first = 'first' if all([t2 == 0, table == 0, loop == 0]) else False
# Load everything for the given t2, table, and loop value
analogs = nudie.load_analogtxt(tg_info['job_name'], current_path, t2,
table, loop)
cdata = nudie.load_camera_file(tg_info['job_name'],
current_path,
t2,
table,
loop,
force_uint16=True)
# Synchronize it, trimming some frames in the beginning and end
data, (a1, a2) = nudie.trim_all(*nudie.synchronize_daq_to_camera(\
cdata, analog_channels=analogs, which_file=first),
trim_to=slice(*trim_to))
data = data.astype(float) # convert to float64 before manipulating it!
start_idxs, period = nudie.detect_table_start(a1)
f, data, df = nudie.wavelen_to_freq(wl, data, ret_df=True, ax=0)
# determine where the shutter is
shutter_open, shutter_closed, duty_cycle = nudie.determine_shutter_shots(
data)
# tag the phases
shutter_info = {
'last open idx': shutter_open,
'first closed idx': shutter_closed
}
tags = nudie.tag_phases(start_idxs,
period,
tags=phase_cycles,
nframes=data.shape[1],
shutter_info=shutter_info)
nudie.remove_incomplete_t1_waveforms(tags, phase_cycles)
data_t = np.zeros((data.shape[1], data.shape[0]), dtype=float)
# subtract the average of closed shutter shots from shutter open data
for t1, k in it.product(range(waveforms_per_table), phase_cycles):
idx_open = tags[nrepeat - 1][t1][k]['shutter open']
idx_closed = tags[nrepeat - 1][t1][k]['shutter closed']
data_t[idx_open, :] = data[:, idx_open].T \
- data[:, idx_closed].mean(axis=1)
t = np.fft.fftfreq(pad_to, df)
lo_window = gaussian2(lo_width, prd_est, t)
dc_window = gaussian2(dc_width, 0, t)
# fft everything
fdata = np.fft.fft(data_t, axis=1, n=pad_to)
if all([plot, table == 0, t2 == 0]):
tmp_fft = fdata[tags[0][0][phase_cycles[1]]['shutter open'][0], :]
plot_windows(t, lo_window, dc_window, tmp_fft, prd_est)
# spectral interferometry
rIprobe = np.fft.ifft(fdata * dc_window, axis=1)[:, :npixels]
rIlo = np.fft.ifft(fdata * lo_window, axis=1)[:, :npixels]
rEprobe = np.sqrt(np.abs(rIprobe))
'''
tag_none1 = tags[0][0]['none1']['shutter open']
nnone1 = rEprobe[tag_none1].mean(axis=0)
none1 = rIlo[tag_none1].mean(axis=0)
tag_zero = tags[0][0]['zero']['shutter open']
nzero = rEprobe[tag_zero].mean(axis=0)
zero = rIlo[tag_zero].mean(axis=0)
tag_none2 = tags[0][0]['none2']['shutter open']
nnone2 = rEprobe[tag_none2].mean(axis=0)
none2 = rIlo[tag_none2].mean(axis=0)
tag_pipi = tags[0][0]['pipi']['shutter open']
npipi = rEprobe[tag_pipi].mean(axis=0)
pipi = rIlo[tag_pipi].mean(axis=0)
'''
tag_none1 = tags[0][0]['none1']['shutter open']
nnone1 = rEprobe[tag_none1]
none1 = rIlo[tag_none1]
tag_zero = tags[0][0]['zero']['shutter open']
nzero = rEprobe[tag_zero]
zero = rIlo[tag_zero]
tag_none2 = tags[0][0]['none2']['shutter open']
nnone2 = rEprobe[tag_none2]
none2 = rIlo[tag_none2]
tag_pipi = tags[0][0]['pipi']['shutter open']
npipi = rEprobe[tag_pipi]
pipi = rIlo[tag_pipi]
#TG = 1/(nzero + npipi)*(zero/nzero + pipi/npipi) \
# - (1/(nnone1 + nnone2)*(none1/nnone1 + none2/nnone2))
#TG = ((zero + pipi - none1 - none2)/(nzero + npipi)).mean(axis=0)
TG = ((zero + pipi - none1 - none2) / (nzero + npipi)).mean(axis=0)
if all([plot, table == 0, t2 == 0]):
plot_phasing_tg(f, TG)
plot_phasing_tg(f, (nzero + npipi).mean(axis=0))
with h5py.File(str(save_path), 'a') as sf:
# save data at current t2
sf['raw transient-grating'][t2] = TG
with h5py.File(str(save_path), 'a') as sf:
# write out meta data
sf.attrs['batch_name'] = tg_info['batch_name']
sf.attrs['batch_no'] = tg_info['batch_no']
sf.attrs['batch_path'] = tg_info['batch_path']
sf.attrs['job_name'] = tg_info['job_name']
sf.attrs['nt2'] = tg_info['nt2']
sf.attrs['when'] = tg_info['when']
sf.attrs['detection axis pad to'] = pad_to
sf.attrs['probe lo delay estimate'] = prd_est
sf.attrs['analysis timestamp'] = arrow.now().format('DD-MM-YYYY HH:mm')
sf.attrs['nudie version'] = nudie.version
sf.attrs['experiment type'] = 'transient-grating'
# write out axes
gaxes = sf.require_group('axes')
freq_dataset = gaxes.create_dataset('detection frequency', data=f)
freq_dataset.attrs['df'] = df
gaxes.create_dataset('detection wavelength',
data=nudie.spectrometer.speed_of_light / f)
gaxes.create_dataset('t2', data=tg_info['t2'][:, 1])
# add dimension scales and attach them
rdata = sf['raw transient-grating']
rdata.dims.create_scale(gaxes['t2'], 'population time / fs')
rdata.dims[0].label = 'population time'
rdata.dims[0].attach_scale(gaxes['t2'])
rdata.dims.create_scale(gaxes['detection frequency'],
'frequency / 1000 THz')
rdata.dims.create_scale(gaxes['detection wavelength'],
'wavelength / nm')
rdata.dims[1].label = 'detection axis'
rdata.dims[1].attach_scale(gaxes['detection frequency'])
rdata.dims[1].attach_scale(gaxes['detection wavelength'])
def run(stark_name, stark_batch, when='today', wavelengths=None, plot=False, analysis_path='analyzed', datapath = None):
'''run the batch job to analyze the linear-stark data
Performs a shot-to-shot linear stark analysis.
```wavelengths``` is a path to a wavelengths file
```plot``` is a Boolean parameter controlling whether to show matplotlib
plots
'''
analysis_folder = Path(analysis_path)
# create folder if it doesn't exist
if not analysis_folder.exists():
analysis_folder.mkdir()
# change datapath
if datapath is None:
datapath = nudie.data_folder
# load up pp data to use for linear stark
stark_info = next(nudie.load_job(job_name=stark_name, batch_set=[stark_batch], when=when, data_path=datapath))
# set current batch directory
current_path = Path(stark_info['batch_path'])
if wavelengths is not None:
wl = load_wavelengths(current_path.parent / wavelengths)
# tags for phase cycles
phase_cycles = ['none1', 'zero', 'none2', 'pipi']
nrepeat = 1 # how many times each waveform is repeated in the camera file. Assumed to be one
nwaveforms = 1 # number of dazzler waveforms in pump probe is 1, assumed to be one
# used to store data generated by iterations
saved_data = deque()
# loop over available data
for t2,table,loop in it.product(stark_info['t2_range'], stark_info['table_range'], stark_info['loop_range']):
# first file requires special synchronization
# this is the rule that determines that it is the first file
first = 'first' if all([t2 == 0, table == 0, loop == 0]) else False
# Load everything for the given t2, table, and loop value
analogs = nudie.load_analogtxt(stark_info['job_name'], current_path, t2, table, loop)
cdata = nudie.load_camera_file(stark_info['job_name'], current_path, t2, table, loop, force_uint16=True)
# Synchronize it, trimming some frames in the beginning and end
data, (a1, a2) = nudie.trim_all(*nudie.synchronize_daq_to_camera(\
cdata, analog_channels=analogs, which_file=first), trim_to=slice(2, None))
data = data.astype(float) # convert to float64 before manipulating it!
start_idxs, period = nudie.detect_table_start(a1)
# determine where the shutter is
shutter_open, shutter_closed, duty_cycle = nudie.determine_shutter_shots(data)
# tag the phases
shutter_info = {'last open idx': shutter_open, 'first closed idx': shutter_closed}
tags = nudie.tag_phases(start_idxs, period, tags=phase_cycles, nframes=data.shape[1], shutter_info=shutter_info)
# verify synchronization of applied field to Dazzler
for i,k in enumerate(phase_cycles):
# check that every second frame in analog channel 2 is either high field, or low field, but not both
# Offset 0, every second frame has high field
ao = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter open']][0::2] > 0.2
# Offset 0, every second frame has low field
bo = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter open']][0::2] <= 0.2
# Offset 1, every second frame has high field
co = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter open']][1::2] > 0.2
# Offset 1, every second frame has low field
do = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter open']][1::2] <= 0.2
# Offset 0, every second frame has high field
ac = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter closed']][0::2] > 0.2
# Offset 0, every second frame has low field
bc = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter closed']][0::2] <= 0.2
# Offset 1, every second frame has high field
cc = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter closed']][1::2] > 0.2
# Offset 1, every second frame has low field
dc = a2[tags[nrepeat-1][nwaveforms-1][k]['shutter closed']][1::2] <= 0.2
assert np.logical_xor(np.all(ao), np.all(bo)), 'detected stark synchronization error, offset 0, phase %s' % k
assert np.logical_xor(np.all(co), np.all(do)), 'detected stark synchronization error, offset 1, phase %s' % k
assert np.logical_xor(np.all(ac), np.all(bc)), 'detected stark synchronization error, offset 0, phase %s' % k
assert np.logical_xor(np.all(cc), np.all(dc)), 'detected stark synchronization error, offset 1, phase %s' % k
# determine stark-on indexes
stark_idx, nostark_idx = np.where(a2 > 0.2)[0], np.where(a2 <= 0.2)[0]
tmp = []
# remove the probe scatter phase by phase
for i,k in enumerate(phase_cycles):
so = np.intersect1d(tags[nrepeat-1][nwaveforms-1][k]['shutter open'], stark_idx, assume_unique=True)
sc = np.intersect1d(tags[nrepeat-1][nwaveforms-1][k]['shutter closed'], stark_idx, assume_unique=True)
no = np.intersect1d(tags[nrepeat-1][nwaveforms-1][k]['shutter open'], nostark_idx, assume_unique=True)
nc = np.intersect1d(tags[nrepeat-1][nwaveforms-1][k]['shutter closed'], nostark_idx, assume_unique=True)
# remove probe scatter for stark shots
data[:, so] -= data[:, sc].mean(axis=-1)[:, np.newaxis]
# and for no-stark shots
data[:, no] -= data[:, nc].mean(axis=-1)[:, np.newaxis]
data = ma.masked_less(data, 1e-7*data.max())
stark_spec = np.log10(np.mean(data[:, so], axis=-1)/np.mean(data[:, no], axis=-1))
stark_spec = ma.filled(stark_spec, 0)
tmp.append(stark_spec)
# average phase cycles together to get slight improvement in S/N
px = max([len(i) for i in tmp]) # num pixels
m = np.zeros((px, len(phase_cycles)))
for i,k in enumerate(tmp):
m[:, i] = k
avg_spectrum = m.mean(axis=1)
mean_voltage = a2[stark_idx].mean()
saved_data.append((avg_spectrum, "loop {}".format(loop)))
# interpret `when` parameter
if when == 'today':
timestamp = arrow.now()
elif when is not None:
# try to interpret when as a string
timestamp = arrow.get(when, 'YY-MM-DD')
else:
raise NotImplemented('don\'t pass a direct path to the ' +\
'filename please')
# Write out information generated
filename = '{timestamp:s}-{name:s}-{voltage:1.1f}kV.hdf5'.format(
timestamp=timestamp.format('YYYY-MM-DD'),
name=stark_info['batch_name'], voltage=mean_voltage)
with h5py.File(str(analysis_folder / filename), mode='w') as f:
for data, j in saved_data:
f[j] = data
f.attrs['applied voltage'] = mean_voltage
f.attrs['file path'] = stark_info['batch_path']
f.attrs['job name'] = stark_info['job_name']
f.attrs['batch_no'] = stark_batch # FIXME: workaround for nudie bug
if wavelengths is not None:
f['wavelength axis'] = wl
if all([plot, loop == stark_info['loop_range'].stop-1]):
for data, loop in saved_data:
pyplot.plot(wl, data, label=loop)
pyplot.xlabel('wavelength / nm')
pyplot.ylabel('OD')
pyplot.show()
def run(tg_name,
tg_batch,
when='today',
wavelengths=None,
plot=False,
pad_to=2048,
prd_est=850.,
lo_width=200,
dc_width=200,
gaussian_power=2.,
analysis_path='./analyzed',
min_field=0.2,
datapath=None):
if plot:
import matplotlib as mpl
import matplotlib.pyplot as plt
mpl.rcParams['figure.figsize'] = (16, 12)
mpl.use('Qt4Agg')
# change datapath
if datapath is None:
datapath = nudie.data_folder
nrepeat = 1 # how many times each waveform is repeated in the camera file. Assumed to be one
waveforms_per_table = 1
npixels = 1340
trim_to = 3, -3
nstark = 2 # number of stark items
# load up tg data to use
tg_info = next(
nudie.load_job(job_name=tg_name,
batch_set=[tg_batch],
when=when,
data_path=datapath))
phase_cycles = ['none1', 'zero', 'none2', 'pipi']
# set current batch directory
current_path = Path(tg_info['batch_path'])
# generate hdf filename based on data date
analysis_folder = Path(analysis_path)
# create folder if it doesn't exist
if not analysis_folder.exists():
analysis_folder.mkdir(parents=True)
save_path = analysis_folder / (tg_info['batch_name'] + '.h5')
# remove data file if it exists
if save_path.exists(): save_path.unlink()
with h5py.File(str(save_path), 'w') as sf:
# initialize groups
sf.create_group('axes')
sf.attrs['nstark'] = nstark
shape = (tg_info['nt2'], nstark, npixels)
sf.create_dataset('raw transient-grating', shape, dtype=complex)
try:
wl = load_wavelengths(current_path.parent / wavelengths)
except FileNotFoundError as e:
nudie.log.error('Could not load wavelength calibration file!')
raise e
# define gaussian
def gaussian2(w, x0, x):
c = 4 * np.log(2)
ξ = x - x0
return np.exp(-c * (ξ / w)**(2 * gaussian_power))
for loop, t2, table in it.product(tg_info['loop_range'],
tg_info['t2_range'],
tg_info['table_range']):
if any([loop != 0, table != 0]):
raise NotImplementedError('Code is not setup to handle multiple ' +\
'loops or tables.')
# first file requires special synchronization
# this is the rule that determines that it is the first file
first = 'first' if all([t2 == 0, table == 0, loop == 0]) else False
# Load everything for the given t2, table, and loop value
analogs = nudie.load_analogtxt(tg_info['job_name'], current_path, t2,
table, loop)
cdata = nudie.load_camera_file(tg_info['job_name'],
current_path,
t2,
table,
loop,
force_uint16=True)
# Synchronize it, trimming some frames in the beginning and end
data, (a1, a2) = nudie.trim_all(*nudie.synchronize_daq_to_camera(\
cdata, analog_channels=analogs, which_file=first),
trim_to=slice(*trim_to))
data = data.astype(float) # convert to float64 before manipulating it!
start_idxs, period = nudie.detect_table_start(a1)
f, data, df = nudie.wavelen_to_freq(wl, data, ret_df=True, ax=0)
# determine where the shutter is
shutter_open, shutter_closed, duty_cycle = nudie.determine_shutter_shots(
data)
# tag the phases
shutter_info = {
'last open idx': shutter_open,
'first closed idx': shutter_closed
}
tags = nudie.tag_phases(start_idxs,
period,
tags=phase_cycles,
nframes=data.shape[1],
shutter_info=shutter_info)
nudie.remove_incomplete_t1_waveforms(tags, phase_cycles)
data_t = np.zeros((data.shape[1], data.shape[0]), dtype=float)
# verify synchronization of applied field to Dazzler
for t1, k in it.product(range(waveforms_per_table), phase_cycles):
# check that every second frame in analog channel 2 is either high field, or low field, but not both
idx_open = tags[nrepeat - 1][t1][k]['shutter open']
idx_closed = tags[nrepeat - 1][t1][k]['shutter closed']
# Offset 0, every second frame has high field
ao = a2[idx_open][0::2] > min_field
# Offset 0, every second frame has low field
bo = a2[idx_open][0::2] <= min_field
# Offset 1, every second frame has high field
co = a2[idx_open][1::2] > min_field
# Offset 1, every second frame has low field
do = a2[idx_open][1::2] <= min_field
# Offset 0, every second frame has high field
ac = a2[idx_closed][0::2] > min_field
# Offset 0, every second frame has low field
bc = a2[idx_closed][0::2] <= min_field
# Offset 1, every second frame has high field
cc = a2[idx_closed][1::2] > min_field
# Offset 1, every second frame has low field
dc = a2[idx_closed][1::2] <= min_field
assert np.logical_xor(np.all(ao), np.all(bo)), \
'detected stark synchronization error, offset 0, phase %s' % str(k)
assert np.logical_xor(np.all(co), np.all(do)), \
'detected stark synchronization error, offset 1, phase %s' % str(k)
assert np.logical_xor(np.all(ac), np.all(bc)), \
'detected stark synchronization error, offset 0, phase %s' % str(k)
assert np.logical_xor(np.all(cc), np.all(dc)), \
'detected stark synchronization error, offset 1, phase %s' % str(k)
assert np.logical_xor(np.all(ao), np.all(co)), \
'detected stark synchronization error, offset 0, phase %s' % str(k)
assert np.logical_xor(np.all(bo), np.all(do)), \
'detected stark synchronization error, offset 1, phase %s' % str(k)
# determine stark-on indexes
stark_idx = np.where(a2 > min_field)[0]
nostark_idx = np.where(a2 <= min_field)[0]
# subtract the average of closed shutter shots from shutter open data
for t1, k in it.product(range(waveforms_per_table), phase_cycles):
idx_open = tags[nrepeat - 1][t1][k]['shutter open']
idx_closed = tags[nrepeat - 1][t1][k]['shutter closed']
so = np.intersect1d(idx_open, stark_idx, assume_unique=True)
sc = np.intersect1d(idx_closed, stark_idx, assume_unique=True)
no = np.intersect1d(idx_open, nostark_idx, assume_unique=True)
nc = np.intersect1d(idx_closed, nostark_idx, assume_unique=True)
data_t[so, :] = (data[:, so].T - data[:, sc].mean(axis=1))
data_t[no, :] = (data[:, no].T - data[:, nc].mean(axis=1))
t = np.fft.fftfreq(pad_to, df)
lo_window = gaussian2(lo_width, prd_est, t)
dc_window = gaussian2(dc_width, 0, t)
# fft everything
fdata = np.fft.fft(data_t, axis=1, n=pad_to)
if all([plot, table == 0, t2 == 0]):
tmp_fft = fdata[tags[0][0][phase_cycles[1]]['shutter open'][0], :]
plot_windows(t, lo_window, dc_window, tmp_fft, prd_est)
# spectral interferometry
rIprobe = np.fft.ifft(fdata * dc_window, axis=1)[:, :npixels]
rIlo = np.fft.ifft(fdata * lo_window, axis=1)[:, :npixels]
rEprobe = np.sqrt(np.abs(rIprobe))
'''
tag_none1 = tags[0][0]['none1']['shutter open']
nnone1 = rEprobe[tag_none1].mean(axis=0)
none1 = rIlo[tag_none1].mean(axis=0)
tag_zero = tags[0][0]['zero']['shutter open']
nzero = rEprobe[tag_zero].mean(axis=0)
zero = rIlo[tag_zero].mean(axis=0)
tag_none2 = tags[0][0]['none2']['shutter open']
nnone2 = rEprobe[tag_none2].mean(axis=0)
none2 = rIlo[tag_none2].mean(axis=0)
tag_pipi = tags[0][0]['pipi']['shutter open']
npipi = rEprobe[tag_pipi].mean(axis=0)
pipi = rIlo[tag_pipi].mean(axis=0)
'''
stark_tag = lambda x: np.intersect1d(x, stark_idx, assume_unique=True)
nostark_tag = lambda x: np.intersect1d(
x, nostark_idx, assume_unique=True)
tag_none1 = tags[0][0]['none1']['shutter open']
stag_none1 = stark_tag(tag_none1)
ntag_none1 = nostark_tag(tag_none1)
#snone1 = rEprobe[stag_none1]
#nnone1 = rEprobe[ntag_none1]
snone1 = rIlo[stag_none1].mean(axis=0)
nnone1 = rIlo[ntag_none1].mean(axis=0)
tag_zero = tags[0][0]['zero']['shutter open']
stag_zero = stark_tag(tag_zero)
ntag_zero = nostark_tag(tag_zero)
snzero = rEprobe[stag_zero].mean(axis=0)
nnzero = rEprobe[ntag_zero].mean(axis=0)
szero = rIlo[stag_zero].mean(axis=0)
nzero = rIlo[ntag_zero].mean(axis=0)
tag_none2 = tags[0][0]['none2']['shutter open']
stag_none2 = stark_tag(tag_none2)
ntag_none2 = nostark_tag(tag_none2)
#nnone2 = rEprobe[tag_none2]
snone2 = rIlo[stag_none2].mean(axis=0)
nnone2 = rIlo[ntag_none2].mean(axis=0)
tag_pipi = tags[0][0]['pipi']['shutter open']
stag_pipi = stark_tag(tag_pipi)
ntag_pipi = nostark_tag(tag_pipi)
snpipi = rEprobe[stag_pipi].mean(axis=0)
nnpipi = rEprobe[ntag_pipi].mean(axis=0)
spipi = rIlo[stag_pipi].mean(axis=0)
npipi = rIlo[ntag_pipi].mean(axis=0)
#TG = 1/(nzero + npipi)*(zero/nzero + pipi/npipi) \
# - (1/(nnone1 + nnone2)*(none1/nnone1 + none2/nnone2))
#TG = ((zero + pipi - none1 - none2)/(nzero + npipi)).mean(axis=0)
stark_axis = np.array([a2[stark_idx].mean(), a2[nostark_idx].mean()])
#TG = ((zero + pipi - none1 - none2)/(nzero + npipi))
sTG = ((szero + spipi - snone1 - snone2) / (snzero + snpipi))
nTG = ((nzero + npipi - nnone1 - nnone2) / (nnzero + nnpipi))
'''
if all([plot, table==0, t2==0]):
plot_phasing_tg(f, TG)
plot_phasing_tg(f, (nzero + npipi).mean(axis=0))
'''
with h5py.File(str(save_path), 'a') as sf:
# save data at current t2
# for some reason, I need -1j now -> 1j to make the real part the
# absorptive and -1 to do the 2D convention
#sf['raw transient-grating'][t2, 0] = -1j*sTG
#sf['raw transient-grating'][t2, 1] = -1j*nTG
sf['raw transient-grating'][t2, 0] = sTG
sf['raw transient-grating'][t2, 1] = nTG
with h5py.File(str(save_path), 'a') as sf:
# write out meta data
sf.attrs['batch_name'] = tg_info['batch_name']
sf.attrs['batch_no'] = tg_info['batch_no']
sf.attrs['batch_path'] = tg_info['batch_path']
sf.attrs['job_name'] = tg_info['job_name']
sf.attrs['nt2'] = tg_info['nt2']
sf.attrs['when'] = tg_info['when']
sf.attrs['detection axis pad to'] = pad_to
sf.attrs['probe lo delay estimate'] = prd_est
sf.attrs['analysis timestamp'] = arrow.now().format('DD-MM-YYYY HH:mm')
sf.attrs['nudie version'] = nudie.version
sf.attrs['experiment type'] = 'stark transient-grating'
# write out axes
gaxes = sf.require_group('axes')
gaxes.create_dataset('stark axis', data=stark_axis)
freq_dataset = gaxes.create_dataset('detection frequency', data=f)
freq_dataset.attrs['df'] = df
gaxes.create_dataset('detection wavelength',
data=nudie.spectrometer.speed_of_light / f)
gaxes.create_dataset('t2', data=tg_info['t2'][:, 1])
# add dimension scales and attach them
rdata = sf['raw transient-grating']
rdata.dims.create_scale(gaxes['t2'], 'population time / fs')
rdata.dims[0].label = 'population time'
rdata.dims[0].attach_scale(gaxes['t2'])
rdata.dims.create_scale(gaxes['stark axis'], 'average field / kV')
rdata.dims[1].label = 'average voltage'
rdata.dims[1].attach_scale(gaxes['stark axis'])
rdata.dims.create_scale(gaxes['detection frequency'],
'frequency / 1000 THz')
rdata.dims.create_scale(gaxes['detection wavelength'],
'wavelength / nm')
rdata.dims[2].label = 'detection axis'
rdata.dims[2].attach_scale(gaxes['detection frequency'])
rdata.dims[2].attach_scale(gaxes['detection wavelength'])
def run(pp_name,
pp_batch,
when='today',
wavelengths=None,
plot=False,
exclude=[],
analysis_path='analyzed',
datapath=None):
'''run the batch job to analyze pump-probe data
Performs a shot-to-shot pump-probe analysis.
```wavelengths``` is a path to a wavelengths file
```plot``` is a Boolean parameter controlling whether to show matplotlib
plots
'''
# tags for phase cycles
phase_cycles = ['none1', 'zero', 'none2', 'pipi']
nrepeat = 1 # how many times each waveform is repeated in the camera file. Assumed to be one
nwaveforms = 1 # number of dazzler waveforms in pump probe is 1
npixels = 1340 # length of detection axis
# change datapath
if datapath is None:
datapath = nudie.data_folder
# load up pp data to use
pp_info = next(
nudie.load_job(job_name=pp_name,
batch_set=[pp_batch],
when=when,
data_path=datapath))
# set current batch directory
current_path = Path(pp_info['batch_path'])
# generate hdf filename based on data date
analysis_folder = Path(analysis_path)
# create folder if it doesn't exist
if not analysis_folder.exists():
analysis_folder.mkdir()
save_path = analysis_folder / (pp_info['batch_name'] + '.h5')
# remove data file if it exists
if save_path.exists(): save_path.unlink()
with h5py.File(str(save_path), mode='w') as sf:
# initialize groups
sf.create_group('axes')
loops = np.array(list(
filter(lambda x: x not in exclude, pp_info['loop_range'])),
dtype=int)
shape = (pp_info['nt2'], len(loops), npixels)
raw_pp = sf.create_dataset('raw pump-probe', shape, dtype=float)
avg_pp = sf.create_dataset('averaged pump-probe', (shape[0], shape[2]),
dtype=float)
try:
wl = load_wavelengths(current_path.parent / wavelengths)
except FileNotFoundError as e:
nudie.log.error('Could not load wavelength calibration file!')
raise e
for t2, table, loop in it.product(pp_info['t2_range'],
pp_info['table_range'], loops):
# first file requires special synchronization
# this is the rule that determines that it is the first file
first = 'first' if all([t2 == 0, table == 0, loop == 0]) else False
# Load everything for the given t2, table, and loop value
analogs = nudie.load_analogtxt(pp_info['job_name'], current_path, t2,
table, loop)
cdata = nudie.load_camera_file(pp_info['job_name'],
current_path,
t2,
table,
loop,
force_uint16=True)
# Synchronize it, trimming some frames in the beginning and end
data, (a1, a2) = nudie.trim_all(*nudie.synchronize_daq_to_camera(\
cdata, analog_channels=analogs, which_file=first))
data = data.astype(float) # convert to float64 before manipulating it!
# interpolate to even frequency spacing
f, data, df = nudie.wavelen_to_freq(wl, data, ret_df=True, ax=0)
start_idxs, period = nudie.detect_table_start(a1)
# determine where the shutter is
shutter_open, shutter_closed, duty_cycle = nudie.determine_shutter_shots(
data)
# tag the phases
shutter_info = {
'last open idx': shutter_open,
'first closed idx': shutter_closed
}
tags = nudie.tag_phases(start_idxs,
period,
tags=phase_cycles,
nframes=data.shape[1],
shutter_info=shutter_info)
# the first two indexes should be zero, so ignore them
tags = tags[0][0]
# truncate everything to the same length
trunc_idx = slice(
min([len(tags[k]['shutter open']) for k in phase_cycles]))
# subtract scatter from shutter closed shots
zero = ((data[:, tags['zero']['shutter open']][:, trunc_idx]).T \
- data[:, tags['zero']['shutter closed']].mean(axis=-1)).T
pipi = ((data[:, tags['pipi']['shutter open']][:, trunc_idx]).T \
- data[:, tags['pipi']['shutter closed']].mean(axis=-1)).T
none1 = ((data[:, tags['none1']['shutter open']][:, trunc_idx]).T \
- data[:, tags['none1']['shutter closed']].mean(axis=-1)).T
none2 = ((data[:, tags['none2']['shutter open']][:, trunc_idx]).T \
- data[:, tags['none2']['shutter closed']].mean(axis=-1)).T
# to match frank's code, except with minus signs
A3 = (0.5 * zero - 0.5 * none1)
B3 = (0.5 * pipi - 0.5 * none2)
# FIXME: unsure about the proper signs for C3!
C3 = (0.25 * zero + 0.25 * none1 + 0.25 * pipi + 0.25 * none2)
C3 = ma.masked_less(C3, 1e-7 * C3.max())
S3 = np.mean((0.5 * A3 + 0.5 * B3) / ma.sqrt(C3), axis=1)
pp = ma.filled(S3 / np.max(S3), fill_value=0)
with h5py.File(str(save_path), mode='a') as sf:
loop_idx = np.argmin(np.abs(loops - loop))
sf['raw pump-probe'][t2, loop_idx] = pp
with h5py.File(str(save_path), mode='a') as sf:
if plot:
for t2, i in it.product(pp_info['t2_range'], range(len(loops))):
mpl.plot(f,
sf['raw pump-probe'][t2, i],
label='t2 {:.1f} loop {:d}'.format(
pp_info['t2'][t2][1], loops[i]))
mpl.legend()
mpl.show()
sf['averaged pump-probe'][:] = np.array(
sf['raw pump-probe']).mean(axis=1)
# write out meta data
gaxes = sf['axes']
gaxes['detection wavelength'] = nudie.spectrometer.speed_of_light / f
gaxes['t2'] = pp_info['t2'][:, 1] # take only the time position
gaxes['loop'] = loops
gaxes['detection frequency'] = f
gaxes['detection frequency'].attrs['df'] = df
sf.attrs['batch_name'] = pp_info['batch_name']
sf.attrs['batch_no'] = pp_info['batch_no']
sf.attrs['batch_path'] = pp_info['batch_path']
sf.attrs['job_name'] = pp_info['job_name']
sf.attrs['nt2'] = pp_info['nt2']
sf.attrs['when'] = pp_info['when']
sf.attrs['nloop'] = pp_info['loop_range'].stop
sf.attrs['analysis timestamp'] = arrow.now().format('DD-MM-YYYY HH:mm')
sf.attrs['nudie version'] = nudie.version
sf.attrs['experiment type'] = 'pump probe'
# add dimension scales
rpp = sf['raw pump-probe']
rpp.dims.create_scale(gaxes['loop'], 'loop number')
rpp.dims.create_scale(gaxes['t2'], 'population time / fs')
rpp.dims.create_scale(gaxes['detection frequency'],
'detection frequency / 1000 THz')
rpp.dims.create_scale(gaxes['detection wavelength'],
'detection wavelength / nm')
rpp.dims[0].label = 'population time'
rpp.dims[0].attach_scale(gaxes['t2'])
rpp.dims[1].label = 'loop number'
rpp.dims[1].attach_scale(gaxes['loop'])
rpp.dims[2].label = 'detection axis'
rpp.dims[2].attach_scale(gaxes['detection frequency'])
rpp.dims[2].attach_scale(gaxes['detection wavelength'])
app = sf['averaged pump-probe']
app.dims[0].label = 'population time'
app.dims[0].attach_scale(gaxes['t2'])
app.dims[1].label = 'detection axis'
app.dims[1].attach_scale(gaxes['detection frequency'])
app.dims[1].attach_scale(gaxes['detection wavelength'])
def run(dd_name,
dd_batch,
when='today',
wavelengths=None,
plot=False,
pump_chop=False,
central_wl=None,
phaselock_wl=None,
pad_to=2048,
waveforms_per_table=40,
prd_est=850.,
lo_width=200,
dc_width=200,
gaussian_power=2.,
analysis_path='./analyzed',
detrend_t1=False,
datapath=None):
if plot:
import matplotlib as mpl
import matplotlib.pyplot as plt
mpl.rcParams['figure.figsize'] = (16, 12)
mpl.use('Qt5Agg')
nrepeat = 1 # how many times each waveform is repeated in the camera file. Assumed to be one
npixels = 1340
trim_to = 3, -3
# change datapath
if datapath is None:
datapath = nudie.data_folder
# load up 2d data to use
dd_info = next(
nudie.load_job(job_name=dd_name,
batch_set=[dd_batch],
when=when,
data_path=datapath))
if pump_chop == True:
phase_cycles = [(0., 0.), (1., 1.), (0, 0.6), (1., 1.6), (0, 1.3),
(1., 2.3), 'chop']
else:
phase_cycles = [(0., 0.), (1., 1.), (0, 0.6), (1., 1.6), (0, 1.3),
(1., 2.3)]
# set current batch directory
current_path = Path(dd_info['batch_path'])
# generate hdf filename based on data date
analysis_folder = Path(analysis_path)
# create folder if it doesn't exist
if not analysis_folder.exists():
analysis_folder.mkdir(parents=True)
save_path = analysis_folder / (dd_info['batch_name'] + '.h5')
# remove data file if it exists
if save_path.exists(): save_path.unlink()
with h5py.File(str(save_path), 'w') as sf:
# initialize groups
sf.create_group('axes')
shape = (dd_info['nt2'], dd_info['nt1'], npixels)
sf.create_dataset('raw rephasing', shape, dtype=complex)
sf.create_dataset('raw non-rephasing', shape, dtype=complex)
sf.create_dataset('raw transient-grating', shape, dtype=complex)
try:
wl = load_wavelengths(current_path.parent / wavelengths)
except FileNotFoundError as e:
nudie.log.error('Could not load wavelength calibration file!')
raise e
# define gaussian
def gaussian2(w, x0, x):
c = 4 * np.log(2)
ξ = x - x0
return np.exp(-c * (ξ / w)**(2 * gaussian_power))
## Make phase-cycling coefficient matrix
# subspaces
if pump_chop == True:
sub1 = np.array(
[[np.exp(1j * np.pi * (x - y)),
np.exp(-1j * np.pi * (x - y)), 1]
for x, y in phase_cycles[0:-1:2]])
sub2 = np.array(
[[np.exp(1j * np.pi * (x - y)),
np.exp(-1j * np.pi * (x - y)), 1]
for x, y in phase_cycles[1:-1:2]])
else:
sub1 = np.array(
[[np.exp(1j * np.pi * (x - y)),
np.exp(-1j * np.pi * (x - y)), 1]
for x, y in phase_cycles[0::2]])
sub2 = np.array(
[[np.exp(1j * np.pi * (x - y)),
np.exp(-1j * np.pi * (x - y)), 1]
for x, y in phase_cycles[1::2]])
perm = np.array([[1, 0, 0, 0, 0, 0], [0, 0, 0, 1, 0,
0], [0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0], [0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 1]])
C = np.kron(np.array([[1, 0], [0, 0]]), sub1) + np.kron(
np.array([[0, 0], [0, 1]]), sub2)
Cinv = np.linalg.inv(C)
Cperm_inv = np.dot(Cinv, perm.T)
for loop, t2, table in it.product(dd_info['loop_range'],
dd_info['t2_range'],
dd_info['table_range']):
if loop != 0:
raise NotImplementedError(
'Code is not setup to handle multiple loops')
# first file requires special synchronization
# this is the rule that determines that it is the first file
first = 'first' if all([t2 == 0, table == 0, loop == 0]) else False
# Load everything for the given t2, table, and loop value
analogs = nudie.load_analogtxt(dd_info['job_name'], current_path, t2,
table, loop)
cdata = nudie.load_camera_file(dd_info['job_name'],
current_path,
t2,
table,
loop,
force_uint16=True)
# Synchronize it, trimming some frames in the beginning and end
data, (a1, a2) = nudie.trim_all(*nudie.synchronize_daq_to_camera(\
cdata, analog_channels=analogs, which_file=first),
trim_to=slice(*trim_to))
data = data.astype(float) # convert to float64 before manipulating it!
start_idxs, period = nudie.detect_table_start(a1)
f, data, df = nudie.wavelen_to_freq(wl, data, ret_df=True, ax=0)
# determine where the shutter is
shutter_open, shutter_closed, duty_cycle = nudie.determine_shutter_shots(
data)
# tag the phases
shutter_info = {
'last open idx': shutter_open,
'first closed idx': shutter_closed
}
tags = nudie.tag_phases(start_idxs,
period,
tags=phase_cycles,
nframes=data.shape[1],
shutter_info=shutter_info)
nudie.remove_incomplete_t1_waveforms(tags, phase_cycles)
data_t = np.zeros((data.shape[1], data.shape[0]), dtype=float)
# subtract the average of closed shutter shots from shutter open data
for t1, k in it.product(range(waveforms_per_table), phase_cycles):
idx_open = tags[nrepeat - 1][t1][k]['shutter open']
idx_closed = tags[nrepeat - 1][t1][k]['shutter closed']
data_t[idx_open, :] = data[:, idx_open].T \
- data[:, idx_closed].mean(axis=1)
t = np.fft.fftfreq(pad_to, df)
lo_window = gaussian2(lo_width, prd_est, t)
dc_window = gaussian2(dc_width, 0, t)
# fft everything
fdata = np.fft.fft(data_t, axis=1, n=pad_to)
if all([plot, table == 0, t2 == 0]):
tmp_fft = fdata[tags[0][0][phase_cycles[0]]['shutter open'][0], :]
plot_windows(t, lo_window, dc_window, tmp_fft, prd_est)
# spectral interferometry
rIprobe = np.fft.ifft(fdata * dc_window, axis=1)[:, :npixels]
rIlo = np.fft.ifft(fdata * lo_window, axis=1)[:, :npixels]
if pump_chop:
nudie.log.warning('pump chopping code is not well tested and will' +\
' probably result in wrong values!')
nudie.log.warning('It currently doesn\'t mask invalid reference ' +\
'values')
# Pump chop before division
rEsig = np.zeros_like(rIlo)
for t1, k in it.product(range(waveforms_per_table),
phase_cycles[:-1]):
phase_idx = tags[nrepeat - 1][t1][k]['shutter open']
chop_idx = tags[nrepeat - 1][t1]['chop']['shutter open']
# FIXME: not sure about sign. in rEsig, also whether to take
# the square of rEprobe
rEprobe = np.sqrt(
np.abs(0.5 * rIprobe[phase_idx, :] +
0.5 * rIprobe[chop_idx, :]))
rEsig[phase_idx, :] = -(rIlo[phase_idx, :] -
rIlo[chop_idx, :]) / rEprobe
else:
rEprobe = np.sqrt(np.abs(rIprobe))
masked_probe = ma.masked_less(rEprobe, 1e-7 * rEprobe.max())
rEsig = ma.filled(rIlo / masked_probe, fill_value=0)
# average each phase together
if pump_chop:
avg = np.zeros(
(len(phase_cycles[:-1]), waveforms_per_table, npixels),
dtype=complex)
for t1, (ik, k) in it.product(range(waveforms_per_table),
enumerate(phase_cycles[:-1])):
avg[ik,
t1, :] = rEsig[tags[0][t1][k]['shutter open']].mean(axis=0)
else:
avg = np.zeros((len(phase_cycles), waveforms_per_table, npixels),
dtype=complex)
for t1, (ik, k) in it.product(range(waveforms_per_table),
enumerate(phase_cycles)):
avg[ik,
t1, :] = rEsig[tags[0][t1][k]['shutter open']].mean(axis=0)
r1, nr1, tg1, r2, nr2, tg2 = np.tensordot(Cperm_inv,
avg,
axes=[[
1,
], [
0,
]])
R = 0.5 * (r1 + r2)
NR = 0.5 * (nr1 + nr2)
TG = 0.5 * (tg1 + tg2)
if all([plot, table == 0, t2 == 0]):
phasing_tg = 0.5 * (R[0] + NR[0])
plot_phasing_tg(f, phasing_tg)
#plot_phasing_tg(f, masked_probe)
with h5py.File(str(save_path), 'a') as sf:
# save data at current t2
t1_slice = slice(table * waveforms_per_table,
(table + 1) * waveforms_per_table)
sf['raw rephasing'][t2, t1_slice] = R
sf['raw non-rephasing'][t2, t1_slice] = NR
sf['raw transient-grating'][t2, t1_slice] = TG
del data, fdata, data_t, avg, rIprobe, rIlo, rEprobe, rEsig
with h5py.File(str(save_path), 'a') as sf:
if detrend_t1:
R = sf['raw rephasing'][:]
NR = sf['raw non-rephasing'][:]
TG = sf['raw transient-grating'][:]
R = detrend(R, axis=1, type='constant')
NR = detrend(NR, axis=1, type='constant')
TG = detrend(TG, axis=1, type='constant')
sf['raw rephasing'][:] = R
sf['raw non-rephasing'][:] = NR
sf['raw transient-grating'][:] = TG
# write out meta data
sf.attrs['batch_name'] = dd_info['batch_name']
sf.attrs['batch_no'] = dd_info['batch_no']
sf.attrs['batch_path'] = dd_info['batch_path']
sf.attrs['job_name'] = dd_info['job_name']
sf.attrs['nt1'] = dd_info['nt1']
sf.attrs['nt2'] = dd_info['nt2']
sf.attrs['when'] = dd_info['when']
sf.attrs['detection axis pad to'] = pad_to
sf.attrs['probe lo delay estimate'] = prd_est
sf.attrs['analysis timestamp'] = arrow.now().format('DD-MM-YYYY HH:mm')
sf.attrs['nudie version'] = nudie.version
sf.attrs['experiment type'] = '2d'
# write out axes
gaxes = sf.require_group('axes')
freq_dataset = gaxes.create_dataset('detection frequency', data=f)
freq_dataset.attrs['df'] = df
gaxes.create_dataset('detection wavelength',
data=nudie.spectrometer.speed_of_light / f)
gaxes.create_dataset('t1', data=dd_info['t1'])
gaxes.create_dataset('t2', data=dd_info['t2'][:, 1])
# add dimension scales
rdata = sf['raw rephasing']
rdata.dims.create_scale(gaxes['t2'], 'population time / fs')
rdata.dims.create_scale(gaxes['t1'], 'pump delay time / fs')
rdata.dims.create_scale(gaxes['detection frequency'],
'frequency / 1000 THz')
rdata.dims.create_scale(gaxes['detection wavelength'],
'wavelength / nm')
# attach them
for tmp in [
sf['raw rephasing'], sf['raw non-rephasing'],
sf['raw transient-grating']
]:
tmp.dims[0].label = 'population time'
tmp.dims[0].attach_scale(gaxes['t2'])
tmp.dims[1].label = 'pump delay time'
tmp.dims[1].attach_scale(gaxes['t1'])
tmp.dims[2].label = 'detection axis'
tmp.dims[2].attach_scale(gaxes['detection frequency'])
tmp.dims[2].attach_scale(gaxes['detection wavelength'])
def run(dd_name,
dd_batch,
when='today',
wavelengths=None,
plot=False,
central_wl=None,
phaselock_wl=None,
pad_to=2048,
waveforms_per_table=40,
prd_est=850.,
lo_width=200,
dc_width=200,
gaussian_power=2,
analysis_path='./analyzed',
min_field=0.2,
detrend_t1=False,
datapath=None):
nrepeat = 1 # how many times each waveform is repeated in the camera file. Assumed to be one
nstark = 2
npixels = 1340
trim_to = 3, -3
# change datapath
if datapath is None:
datapath = nudie.data_folder
# load up 2d data to use
dd_info = next(
nudie.load_job(job_name=dd_name,
batch_set=[dd_batch],
when=when,
data_path=datapath))
phase_cycles = [(0., 0.), (1., 1.), (0, 0.6), (1., 1.6), (0, 1.3),
(1., 2.3)]
# set current batch directory
current_path = Path(dd_info['batch_path'])
# generate hdf filename based on data date
analysis_folder = Path(analysis_path)
# create folder if it doesn't exist
if not analysis_folder.exists():
analysis_folder.mkdir()
save_path = analysis_folder / (dd_info['batch_name'] + '.h5')
# remove data file if it exists
if save_path.exists(): save_path.unlink()
with h5py.File(str(save_path), 'w') as sf:
# initialize groups
sf.create_group('axes')
sf.attrs['nstark'] = nstark
shape = (dd_info['nt2'], nstark, dd_info['nt1'], npixels)
sf.create_dataset('raw rephasing', shape, dtype=complex)
sf.create_dataset('raw non-rephasing', shape, dtype=complex)
sf.create_dataset('raw transient-grating', shape, dtype=complex)
try:
wl = load_wavelengths(current_path.parent / wavelengths)
except FileNotFoundError as e:
nudie.log.error('Could not load wavelength calibration file!')
raise e
def gaussian2(w, x0, x):
c = 4 * np.log(2)
ξ = x - x0
return np.exp(-c * (ξ / w)**(2 * gaussian_power))
## Make phase-cycling coefficient matrix
# subspaces
sub1 = np.array(
[[np.exp(1j * np.pi * (x - y)),
np.exp(-1j * np.pi * (x - y)), 1] for x, y in phase_cycles[0::2]])
sub2 = np.array(
[[np.exp(1j * np.pi * (x - y)),
np.exp(-1j * np.pi * (x - y)), 1] for x, y in phase_cycles[1::2]])
perm = np.array([[1, 0, 0, 0, 0, 0], [0, 0, 0, 1, 0,
0], [0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0], [0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 1]])
C = np.kron(np.array([[1, 0], [0, 0]]), sub1) + np.kron(
np.array([[0, 0], [0, 1]]), sub2)
Cinv = np.linalg.inv(C)
Cperm_inv = np.dot(Cinv, perm.T)
for loop, t2, table in it.product(dd_info['loop_range'],
dd_info['t2_range'],
dd_info['table_range']):
if loop != 0:
raise NotImplemented('code is not setup to handle multiple loops')
# first file requires special synchronization
# this is the rule that determines that it is the first file
first = 'first' if all([t2 == 0, table == 0, loop == 0]) else False
# Load everything for the given t2, table, and loop value
analogs = nudie.load_analogtxt(dd_info['job_name'], current_path, t2,
table, loop)
cdata = nudie.load_camera_file(dd_info['job_name'],
current_path,
t2,
table,
loop,
force_uint16=True)
# Synchronize it, trimming some frames in the beginning and end
data, (a1, a2) = nudie.trim_all(*nudie.synchronize_daq_to_camera(\
cdata, analog_channels=analogs, which_file=first),
trim_to=slice(*trim_to))
data = data.astype(float) # convert to float64 before manipulating it!
start_idxs, period = nudie.detect_table_start(a1)
# data comes out [npixels, nframes]
f, data, df = nudie.wavelen_to_freq(wl, data, ret_df=True, ax=0)
# determine where the shutter is
shutter_open, shutter_closed, duty_cycle = nudie.determine_shutter_shots(
data)
# tag the phases
shutter_info = {
'last open idx': shutter_open,
'first closed idx': shutter_closed
}
tags = nudie.tag_phases(start_idxs,
period,
tags=phase_cycles,
nframes=data.shape[1],
shutter_info=shutter_info)
# data_t is transposed data, [nframes, npixels]
data_t = np.zeros((data.shape[1], data.shape[0]), dtype=float)
# verify synchronization of applied field to Dazzler
for t1, k in it.product(range(waveforms_per_table), phase_cycles):
# check that every second frame in analog channel 2 is either high field, or low field, but not both
idx_open = tags[nrepeat - 1][t1][k]['shutter open']
idx_closed = tags[nrepeat - 1][t1][k]['shutter closed']
# Offset 0, every second frame has high field
ao = a2[idx_open][0::2] > min_field
# Offset 0, every second frame has low field
bo = a2[idx_open][0::2] <= min_field
# Offset 1, every second frame has high field
co = a2[idx_open][1::2] > min_field
# Offset 1, every second frame has low field
do = a2[idx_open][1::2] <= min_field
# Offset 0, every second frame has high field
ac = a2[idx_closed][0::2] > min_field
# Offset 0, every second frame has low field
bc = a2[idx_closed][0::2] <= min_field
# Offset 1, every second frame has high field
cc = a2[idx_closed][1::2] > min_field
# Offset 1, every second frame has low field
dc = a2[idx_closed][1::2] <= min_field
assert np.logical_xor(np.all(ao), np.all(bo)), \
'detected stark synchronization error, offset 0, phase %s' % str(k)
assert np.logical_xor(np.all(co), np.all(do)), \
'detected stark synchronization error, offset 1, phase %s' % str(k)
assert np.logical_xor(np.all(ac), np.all(bc)), \
'detected stark synchronization error, offset 0, phase %s' % str(k)
assert np.logical_xor(np.all(cc), np.all(dc)), \
'detected stark synchronization error, offset 1, phase %s' % str(k)
assert np.logical_xor(np.all(ao), np.all(co)), \
'detected stark synchronization error, offset 0, phase %s' % str(k)
assert np.logical_xor(np.all(bo), np.all(do)), \
'detected stark synchronization error, offset 1, phase %s' % str(k)
# determine stark-on indexes
stark_idx = np.where(a2 > min_field)[0]
nostark_idx = np.where(a2 <= min_field)[0]
for t1, k in it.product(range(waveforms_per_table), phase_cycles):
idx_open = tags[nrepeat - 1][t1][k]['shutter open']
idx_closed = tags[nrepeat - 1][t1][k]['shutter closed']
so = np.intersect1d(idx_open, stark_idx, assume_unique=True)
sc = np.intersect1d(idx_closed, stark_idx, assume_unique=True)
no = np.intersect1d(idx_open, nostark_idx, assume_unique=True)
nc = np.intersect1d(idx_closed, nostark_idx, assume_unique=True)
data_t[so, :] = (data[:, so].T - data[:, sc].mean(axis=1))
data_t[no, :] = (data[:, no].T - data[:, nc].mean(axis=1))
t = np.fft.fftfreq(pad_to, df)
lo_window = gaussian2(lo_width, prd_est, t)
dc_window = gaussian2(dc_width, 0, t)
# fft everything
fdata = np.fft.fft(data_t, axis=1, n=pad_to)
if all([plot, table == 0, t2 == 0]):
tmp_fft = fdata[tags[0][0][phase_cycles[0]]['shutter open'][0], :]
plot_windows(t, lo_window, dc_window, tmp_fft, prd_est)
# spectral interferometry
rIprobe = np.fft.ifft(fdata * dc_window)[:, :npixels]
rIlo = np.fft.ifft(fdata * lo_window)[:, :npixels]
rEprobe = np.sqrt(np.abs(rIprobe))
rEsig = rIlo / rEprobe
# average each phase together
stark_axis = np.array([a2[stark_idx].mean(), a2[nostark_idx].mean()])
avg = np.zeros(
(len(phase_cycles), nstark, waveforms_per_table, npixels),
dtype=complex)
for t1, (ik, k) in it.product(range(waveforms_per_table),
enumerate(phase_cycles)):
# create indexes for stark shots
idx_open = tags[nrepeat - 1][t1][k]['shutter open']
so = np.intersect1d(idx_open, stark_idx, assume_unique=True)
no = np.intersect1d(idx_open, nostark_idx, assume_unique=True)
avg[ik, 0, t1, :] = rEsig[so].mean(axis=0)
avg[ik, 1, t1, :] = rEsig[no].mean(axis=0)
r1, nr1, tg1, r2, nr2, tg2 = np.tensordot(Cperm_inv,
avg,
axes=[[
1,
], [
0,
]])
R = 0.5 * (r1 + r2)
NR = 0.5 * (nr1 + nr2)
TG = 0.5 * (tg1 + tg2)
if all([plot, table == 0, t2 == 0]):
# plot the no-stark TG
phasing_tg = 0.5 * (R + NR)
plot_phasing_tg(f, phasing_tg[1][0])
plot_phasing_tg(f, phasing_tg[0][0])
#mpl.contourf(np.abs(phasing_tg), 50)
#mpl.show()
with h5py.File(str(save_path), 'a') as sf:
# save data at current t2
t1_slice = slice(table * waveforms_per_table,
(table + 1) * waveforms_per_table)
sf['raw rephasing'][t2, :, t1_slice] = R
sf['raw non-rephasing'][t2, :, t1_slice] = NR
sf['raw transient-grating'][t2, :, t1_slice] = TG
nudie.log.debug('Shapes of data:')
nudie.log.debug('R: {!s}, NR: {!s}, TG: {!s}'.format(
R.shape, NR.shape, TG.shape))
del data, fdata, data_t, avg, rIprobe, rIlo, rEprobe, rEsig
with h5py.File(str(save_path), 'a') as sf:
if detrend_t1:
R = sf['raw rephasing'][:]
NR = sf['raw non-rephasing'][:]
TG = sf['raw transient-grating'][:]
R = detrend(R, axis=2, type='constant')
NR = detrend(NR, axis=2, type='constant')
TG = detrend(TG, axis=2, type='constant')
sf['raw rephasing'][:] = R
sf['raw non-rephasing'][:] = NR
sf['raw transient-grating'][:] = TG
# write out meta data
sf.attrs['batch_name'] = dd_info['batch_name']
sf.attrs['batch_no'] = dd_info['batch_no']
sf.attrs['batch_path'] = dd_info['batch_path']
sf.attrs['job_name'] = dd_info['job_name']
sf.attrs['nt1'] = dd_info['nt1']
sf.attrs['nt2'] = dd_info['nt2']
sf.attrs['when'] = dd_info['when']
sf.attrs['probe lo delay estimate'] = prd_est
sf.attrs['analysis timestamp'] = arrow.now().format('MM-DD-YYYY HH:mm')
sf.attrs['nudie version'] = nudie.version
sf.attrs['experiment type'] = 'stark 2d'
# write out axes
gaxes = sf.require_group('axes')
freq_dataset = gaxes.create_dataset('detection frequency', data=f)
freq_dataset.attrs['df'] = df
gaxes.create_dataset('detection wavelength',
data=nudie.spectrometer.speed_of_light / f)
gaxes.create_dataset('t1', data=dd_info['t1'])
gaxes.create_dataset('t2', data=dd_info['t2'][:, 1])
gaxes.create_dataset('stark axis', data=stark_axis)
# add dimension scales
rdata = sf['raw rephasing']
rdata.dims.create_scale(gaxes['t2'], 'population time / fs')
rdata.dims.create_scale(gaxes['t1'], 'pump delay time / fs')
rdata.dims.create_scale(gaxes['detection frequency'],
'frequency / 1000 THz')
rdata.dims.create_scale(gaxes['detection wavelength'],
'wavelength / nm')
rdata.dims.create_scale(gaxes['stark axis'], 'average field / kV')
# attach them
for tmp in [
sf['raw rephasing'], sf['raw non-rephasing'],
sf['raw transient-grating']
]:
tmp.dims[0].label = 'population time'
tmp.dims[0].attach_scale(gaxes['t2'])
tmp.dims[1].label = 'average voltage'
tmp.dims[1].attach_scale(gaxes['stark axis'])
tmp.dims[2].label = 'pump delay time'
tmp.dims[2].attach_scale(gaxes['t1'])
tmp.dims[3].label = 'detection axis'
tmp.dims[3].attach_scale(gaxes['detection frequency'])
tmp.dims[3].attach_scale(gaxes['detection wavelength'])