def mtspec(
self,
time_bandwidth=2,
sample_count: Optional[int] = None,
to_dft: bool = False,
**kwargs,
) -> "mopy.SpectrumGroup":
"""
Return a SpectrumGroup by calculating amplitude spectra via mtspec.
Parameters
----------
time_bandwidth
The time bandwidth used in mtspec, see its docstring for more
details.
sample_count
The number of samples in the time series to use in the
transformation. If greater than the length of the data columns
it will be zero padded, if less the data will be cropped.
Defaults to the next fast length.
to_dft
If True, convert the power spectral density calculated by mtspec to
the discrete fourier transform
Notes
-----
The parameter time_bandwidth and the kwargs are passed directly to
mtspec.mtspec, see its documentation for details:
https://krischer.github.io/mtspec/multitaper_mtspec.html
"""
mtspec = optional_import("mtspec")
# get prepared input array
if sample_count is None:
sample_count = next_fast_len(len(self.data.columns))
ar = pad_or_trim(self.data.values, sample_count=sample_count)
# collect kwargs for mtspec
delta = 1.0 / self.sampling_rate
kwargs = dict(kwargs)
kwargs.update({"delta": delta, "time_bandwidth": time_bandwidth})
# unfortunately we may need to use loops here
out = [mtspec.mtspec(x, **kwargs) for x in ar]
array = np.array([x[0] for x in out])
freqs = out[0][1]
df = pd.DataFrame(array, index=self.data.index, columns=freqs)
# normalize to number of non-zero samples
norm = len(self.data.columns) / self.stats["npts"]
df = df.multiply(norm, axis=0)
df.columns.name = "frequency"
# # collect and return
sg_kwargs = dict(data=df,
stats_group=self.stats_group,
spectra_type="psd")
sg = mopy.SpectrumGroup(**sg_kwargs)
if to_dft:
sg = sg.to_spectra_type("dft")
return sg
def dft(self, sample_count: Optional[int] = None) -> "mopy.SpectrumGroup":
"""
Apply a discrete Fourier transform to the data to create a SpectrumGroup
Parameters
----------
sample_count
The number of samples in the time series to use in the
transformation. If greater than the length of the data columns
it will be zero padded, if less the data will be cropped.
"""
data = self.data
if sample_count is None:
sample_count = next_fast_len(len(self.data.columns))
spec = np.fft.rfft(data.values, n=sample_count, axis=-1)
freqs = np.fft.rfftfreq(len(data.columns), 1.0 / self.sampling_rate)
df = pd.DataFrame(spec, index=data.index, columns=freqs)
df.columns.name = "frequency"
# create SpectrumGroup
kwargs = dict(data=df,
stats_group=self.stats_group,
spectra_type="dft")
return mopy.SpectrumGroup(**kwargs)
def source_group_node_session(node_trace_group):
""" Return a source group with node data. """
sg = mopy.SpectrumGroup(node_trace_group)
assert not hasattr(sg.stats, "process") or not sg.stats.processing
assert not (sg.data == 0).all().all()
return sg