def process_vis(self, vis_range=(390, 720), min_scan=None,
max_scan=None, sigma=2.3, para_angle=45):
data_file = self.file
wls = self.vis_wls()
t = data_file['t']
d = -data_file['data_Stresing CCD'][min_scan:max_scan, ...]
if 'rot' in data_file:
rot = data_file['rot'][min_scan:max_scan]
para_idx = (abs(np.round(rot) - para_angle) < 3)
else:
n_ir_cwl = data_file['wl_Remote IR 32x2'].shape[0]
para_idx = np.repeat(np.array([False, True], dtype='bool'), n_ir_cwl)
dpm = sigma_clip(d[para_idx, ...], axis=0, sigma=sigma, max_iter=10)
dsm = sigma_clip(d[~para_idx, ...], axis=0, sigma=sigma, max_iter=10)
dp = dpm.mean(0)
dps = dpm.std(0)
ds = dsm.mean(0)
dss = dsm.std(0)
para = TimeResSpec(wls, t, dp[0, :, 0, ...], freq_unit='nm',
disp_freq_unit='nm')
perp = TimeResSpec(wls, t, ds[0, :, 0, ...], freq_unit='nm',
disp_freq_unit='nm')
pol = PolTRSpec(para, perp)
pol = pol.cut_freq(*vis_range, invert_sel=True)
return pol.para, pol.perp, pol
def test_pol_plot():
ds = TimeResSpec(wl, t, data)
ds = ds.bin_freqs(50)
ds = PolTRSpec(para=ds, perp=ds)
ds = ds.bin_freqs(50)
ds.plot.trans([550])
ds.plot.spec([2, 10])
ds.plot.trans([550], norm=1)
ds.plot.trans([550], norm=1, marker='o')
def test_pol_plot():
ds = TimeResSpec(wl, t, data)
ds = ds.bin_freqs(50)
ds = PolTRSpec(para=ds, perp=ds)
ds = ds.bin_freqs(50)
ds.plot.trans([550])
ds.plot.spec([2, 10])
ds.plot.trans([550], norm=1)
ds.plot.trans([550], norm=1, marker='o')
def test_plot():
ds = TimeResSpec(wl, t, data)
ds = ds.bin_freqs(50)
ds.plot.trans([550])
ds.plot.spec([2, 10])
ds.plot.trans([550], norm=1)
ds.plot.trans([550], norm=1, marker='o')
ds.plot.map(plot_con=0)
ds.plot.svd()
def test_plot():
ds = TimeResSpec(wl, t, data)
ds = ds.bin_freqs(50)
ds.plot.trans([550])
ds.plot.spec([2, 10])
ds.plot.trans_integrals((1e7 / 550, 1e7 / 600))
ds.plot.trans_integrals((1e7 / 600, 1e7 / 500))
ds.plot.trans([550], norm=1)
ds.plot.trans([550], norm=1, marker='o')
ds.plot.map(plot_con=0)
ds.plot.svd()
def make_pol_ds(self, sigma=None) -> PolTRSpec:
para = np.nanmean(self.par_data, axis=0)
ds_para = TimeResSpec(self.wavelength,
self.t,
1000 * para,
disp_freq_unit='cm')
perp = np.nanmean(self.per_data, axis=0)
ds_perp = TimeResSpec(self.wavelength,
self.t,
1000 * perp,
disp_freq_unit='cm')
return PolTRSpec(ds_para, ds_perp)
def test_sas():
from skultrafast.kinetic_model import Model
ds = TimeResSpec(wl, t, data)
x0 = [0.1, 0.1, 1, 1000]
out = ds.fit_exp(x0)
m = Model()
m.add_transition('S1', 'S1*', 'k1')
m.add_transition('S1', 'zero', 'k2')
out.make_sas(m, {})
m2 = Model()
m2.add_transition('S1', 'S1*', 'k1', 'qy1')
m2.add_transition('S1', 'zero', 'k2')
out.make_sas(m2, {'qy1': 0.5})
def test_concat():
ds = TimeResSpec(wl, t, data)
ds = ds.bin_freqs(50)
n = ds.wavelengths.size // 2
ds1 = ds.cut_freq(ds.wavelengths[n], np.inf)
ds2 = ds.cut_freq(-np.inf, ds.wavelengths[n])
dsc = ds1.concat_datasets(ds2)
assert (np.allclose(dsc.data, ds.data))
pol_ds = PolTRSpec(ds, ds)
a = PolTRSpec(ds1, ds1)
b = PolTRSpec(ds2, ds2)
pol_dsc = a.concat_datasets(b)
for p in 'para', 'perp', 'iso':
assert (np.allclose(getattr(pol_dsc, p).data, getattr(pol_ds, p).data))
def test_methods():
ds = TimeResSpec(wl, t, data)
bds = ds.bin_freqs(300)
assert (len(bds.wavelengths) == 300)
nds = ds.cut_freqs([(400, 600)])
assert (np.all(nds.wavelengths > 600))
nds = ds.cut_times([(-100, 1)])
assert (np.all(nds.t > .99))
nds = ds.bin_times(5)
assert (nds.t.size == np.ceil(ds.t.size / 5))
ds.mask_freqs([(400, 600)])
assert (np.all(ds.data.mask[:, ds.wl_idx(550)]))
def process_ir(self,
t0=0,
min_scans=0,
max_scans=None,
subtract_background=True,
center_ch=16,
disp=14,
sigma=3) -> PolTRSpec:
data_file = self.file
t = data_file['t'] - t0
wli = data_file['wl_Remote IR 32x2']
print(wli[:, 16, None])
wli = -(disp * (np.arange(32) - center_ch)) + wli[:, 16, None]
wli = 1e7 / wli
d = data_file['data_Remote IR 32x2'][min_scans:max_scans]
print(d.shape)
dp = sigma_clip(d[1::2, ...], axis=0, sigma=sigma)
dpm = dp.mean(0)
ds = sigma_clip(d[0::2, ...], axis=0, sigma=sigma)
dsm = ds.mean(0)
if subtract_background:
dsm -= dsm[:, :10, ...].mean(1, keepdims=True)
dpm -= dpm[:, :10, ...].mean(1, keepdims=True)
para = TimeResSpec(wli[0],
t,
dpm[0, :, 0, ...],
freq_unit='cm',
disp_freq_unit='cm')
perp = TimeResSpec(wli[0],
t,
dsm[0, :, 0, ...],
freq_unit='cm',
disp_freq_unit='cm')
para.plot.spec(1, n_average=20)
for i in range(1, wli.shape[0]):
para_t = TimeResSpec(wli[i],
t,
dpm[i, :, 0, ...],
freq_unit='cm',
disp_freq_unit='cm')
para_t.plot.spec(1, n_average=20)
para = para.concat_datasets(para_t)
perp = perp.concat_datasets(
TimeResSpec(wli[i],
t,
dsm[i, :, 0, ...],
freq_unit='cm',
disp_freq_unit='cm'))
both = PolTRSpec(para, perp)
return both
def test_das_pol_plots():
ds = TimeResSpec(wl, t, data)
pds = PolTRSpec(ds, ds) # fake pol
x0 = [0.1, 0.1, 1, 1000]
out = pds.fit_exp(x0)
pds.plot.das()
pds.plot.edas()
def test_pol_tr():
ds = TimeResSpec(wl, t, data)
ds2 = TimeResSpec(wl, t, data)
ps = PolTRSpec(para=ds, perp=ds2)
out = ps.bin_freqs(10)
assert (out.para.wavenumbers.size == 10)
assert (out.perp.wavenumbers.size == 10)
assert_almost_equal(out.perp.data, out.para.data)
ps.subtract_background()
ps.mask_freqs([(400, 550)])
print(ps.para.data.mask, ps.para.data.mask[1, ps.para.wl_idx(520)])
assert (ps.para.data.mask[1, ps.para.wl_idx(520)])
out = ps.cut_freqs([(400, 550)])
assert (np.all(out.para.wavelengths >= 550))
assert (np.all(out.perp.wavelengths >= 550))
ps.bin_times(6)
def test_methods():
ds = TimeResSpec(wl, t, data)
bds = ds.bin_freqs(300)
assert(len(bds.wavelengths) == 300)
nds = ds.cut_freqs([(400, 600)])
assert(np.all(nds.wavelengths > 600))
nds = ds.cut_times([(-100, 1)])
assert(np.all(nds.t > .99))
nds = ds.bin_times(5)
assert(nds.t.size == np.ceil(ds.t.size/5))
ds.mask_freqs([(400, 600)])
assert(np.all(ds.data.mask[:, ds.wl_idx(550)]))
def test_sas_pol_plots():
from skultrafast.kinetic_model import Model
ds = TimeResSpec(wl, t, data)
pds = PolTRSpec(ds, ds) # fake pol
x0 = [0.1, 0.1, 1, 1000]
out = pds.fit_exp(x0)
m = Model()
m.add_transition('S1', 'S1*', 'k1')
m.add_transition('S1', 'zero', 'k2')
pds.plot.sas(m)
def integrate_pump(self,
lower: float = -np.inf,
upper: float = np.inf) -> TimeResSpec:
"""
Calculate and return 1D Time-resolved spectra for given range.
Parameters
----------
lower : float
Lower pump wl
upper : float
upper pump wl
Returns
-------
TimeResSpec
The corresponding 1D Dataset
"""
pu_idx = inbetween(self.pump_wn, lower, upper)
data = np.trapz(self.spec2d[:, :, pu_idx],
self.pump_wn[pu_idx],
axis=-1)
return TimeResSpec(self.probe_wn, self.t, data, freq_unit='cm')
def test_integrate():
ds = TimeResSpec(wl, t, data)
ds.wn_i(15000, 20000)
def test_das_plots():
ds = TimeResSpec(wl, t, data)
x0 = [0.1, 0.1, 1, 1000]
out = ds.fit_exp(x0)
ds.plot.das()
ds.plot.edas()
def test_error_calc():
ds = TimeResSpec(wl, t, data)
x0 = [0.1, 0.1, 1, 1000]
out = ds.fit_exp(x0)
out.calculate_stats()
def test_fitter():
ds = TimeResSpec(wl, t, data)
x0 = [0.1, 0.1, 1, 1000]
out = ds.fit_exp(x0)
def test_est_disp():
ds = TimeResSpec(wl, t, data)
ds.auto_plot = False
for s in ['abs', 'diff', 'gauss_diff', 'max']:
ds.estimate_dispersion(heuristic=s)
def average_scans(self,
sigma=3,
max_iter=3,
min_scan=0,
max_scan=None,
disp_freq_unit=None):
"""
Calculate the average of the scans. Uses sigma clipping, which
also filters nans. For polarization resolved measurements, the
function assumes that the polarisation switches every scan.
Parameters
----------
sigma : float
sigma used for sigma clipping.
max_iter: int
Maximum iterations in sigma clipping.
min_scan : int or None
All scans before min_scan are ignored.
max_scan : int or None
If `None`, use all scan, else just use the scans up to max_scan.
disp_freq_unit : 'nm', 'cm' or None
Sets `disp_freq_unit` of the created datasets.
Returns
-------
dict or TimeResSpec
TimeResSpec or Dict of DataSets containing the averaged datasets. If
the first delay-time are identical, they are interpreted as
background and their mean is subtracted.
"""
if max_scan is None:
sub_data = self.data
else:
sub_data = self.data[..., min_scan:max_scan]
num_wls = self.data.shape[0]
t = self.t
if disp_freq_unit is None:
disp_freq_unit = "nm" if self.wl.shape[0] > 32 else "cm"
kwargs = dict(disp_freq_unit=disp_freq_unit)
if not self.is_pol_resolved:
data = sigma_clip(sub_data, sigma=sigma, max_iter=max_iter, axis=-1)
mean = data.mean(-1)
std = data.std(-1)
err = std / np.sqrt((~data.mask).sum(-1))
if self.valid_channel in [0, 1]:
mean = mean[..., self.valid_channel]
std = std[..., self.valid_channel]
err = err[..., self.valid_channel]
out = {}
if num_wls > 1:
for i in range(num_wls):
ds = TimeResSpec(self.wl[:, i], t, mean[i, ..., :], err[i, ...],
**kwargs)
out[self.pol_first_scan + str(i)] = ds
else:
out = TimeResSpec(self.wl[:, 0], t, mean[0, ...], err[0, ...],
**kwargs)
return out
else:
raise NotImplementedError("TODO")
elif self.is_pol_resolved and self.valid_channel in [0, 1]:
assert self.pol_first_scan in ["para", "perp"]
data1 = sigma_clip(sub_data[..., self.valid_channel, ::2],
sigma=sigma,
max_iter=max_iter,
axis=-1)
mean1 = data1.mean(-1)
std1 = data1.std(-1, ddof=1)
err1 = std1 / np.sqrt(np.ma.count(data1, -1))
data2 = sigma_clip(sub_data[..., self.valid_channel, 1::2],
sigma=sigma,
max_iter=max_iter,
axis=-1)
mean2 = data2.mean(-1)
std2 = data2.std(-1, ddof=1)
err2 = std2 / np.sqrt(np.ma.count(data2, -1))
out = {}
for i in range(num_wls):
wl, t = self.wl[:, i], self.t
if self.pol_first_scan == "para":
para = mean1[i, ...]
para_err = err1[i, ...]
perp = mean2[i, ...]
perp_err = err2[i, ...]
elif self.pol_first_scan == "perp":
para = mean2[i, ...]
para_err = err2[i, ...]
perp = mean1[i, ...]
perp_err = err1[i, ...]
para_ds = TimeResSpec(wl, t, para, para_err, **kwargs)
perp_ds = TimeResSpec(wl, t, perp, perp_err, **kwargs)
out["para" + str(i)] = para_ds
out["perp" + str(i)] = perp_ds
iso = 1/3*para + 2/3*perp
out["iso" + str(i)] = TimeResSpec(wl, t, iso, **kwargs)
self.av_scans_ = out
return out
else:
raise NotImplementedError("Iso correction not supported yet.")
def test_methods():
ds = TimeResSpec(wl, t, data)
bds = ds.bin_freqs(300)
ds2 = TimeResSpec(1e7 / wl, t, data, freq_unit='cm', disp_freq_unit='cm')
bds2 = ds2.bin_freqs(50)
assert (np.all(np.isfinite(bds2.data)))
assert (len(bds.wavelengths) == 300)
nds = ds.cut_freq(400, 600)
assert (np.all(nds.wavelengths > 600))
nds = ds.cut_time(-100, 1)
assert (np.all(nds.t > .99))
nds = ds.bin_times(5)
assert (nds.t.size == np.ceil(ds.t.size / 5))
ds.mask_freqs([(400, 600)])
assert (np.all(ds.data.mask[:, ds.wl_idx(550)]))
ds2 = ds.scale_and_shift(2, t_shift=1, wl_shift=10)
assert_almost_equal(2 * ds.data, ds2.data)
assert_almost_equal(ds.t + 1, ds2.t)
assert_almost_equal(ds.wavelengths + 10, ds2.wavelengths)
assert_almost_equal(1e7 / ds2.wavelengths, ds2.wavenumbers)
# another one containing the delay times and one two dimensional array containing # the data in mOD. wavelengths, t_ps, data_mOD = data_io.load_example() # %% # Lets look at the constructor of the `TimeResSpec` (Time resolved Spectra) class, # which is the main object when working with single data: print(TimeResSpec.__init__.__doc__) # %% # As we see, we can supply all the required parameters. # Since the `freq_unit` defaults to 'nm' we don't need to supply this argument. ds = TimeResSpec(wavelengths, t_ps, data_mOD) # %% # The TimeResSpec object simply consists of data itself and methods using that # data. The attributes containing the data can be accessed under `ds.data`, # `ds.wavenumbers`, `ds.wavelengths` and `ds.t`. print(ds.data.shape, ds.t.shape, ds.wavelengths.shape) # %% # The TimeResSpec object also has some helper methods to work with the data. # These functions find the index of the nearest value for a given number, e.g. to find # the position in the time array where the time is zero we can call the `t_idx` # method
def test_merge():
ds = TimeResSpec(wl, t, data)
nds = ds.merge_nearby_channels(10)
assert (nds.wavelengths.size < ds.wavelengths.size)
def test_plot():
ds = TimeResSpec(wl, t, data)
ds = ds.bin_freqs(50)
ds.trans([550])
ds.spec([2, 10])
ds.trans([550], norm=1)
ds.trans([550], norm=1, marker='o')
ds.plot.map(plot_con=0)
ds.plot.svd()
def test_est_disp():
ds = TimeResSpec(wl, t, data)
ds.auto_plot = False
for s in ['abs', 'diff', 'gauss_diff', 'max']:
ds.estimate_dispersion(heuristic=s)
def test_fitter():
ds = TimeResSpec(wl, t, data)
x0 = [0.1, 0.1, 1, 1000]
out = ds.fit_exp(x0)