sigpyproc.timeseries

sigpyproc.timeseries#

Classes for handling 1D time series data.

This module contains the TimeSeries class for handling 1D time series data.

Classes

TimeSeries

Container for 1-D time series data.

class sigpyproc.timeseries.TimeSeries(data, header)[source]#

Bases: object

Container for 1-D time series data.

Parameters:

dataArrayLike: 1-D time series array.
headerHeader: Header object containing metadata.

Attributes:

data: Time series data array.
header: Metadata header object.
nsamples: Number of samples.

Methods

`apply_boxcar`(width)	Apply a square-normalized boxcar filter to the time series.
`correlate`(other)	Perform cross correlation with another time series using FFTs.
`deredden`([method, window, fast])	Remove low-frequency red noise using a moving filter.
`downsample`(factor[, filter_method])	Downsample the time series.
`fold`(period[, accel, nbins, nints])	Fold time series into discrete phase and sub-integration bins.
`from_dat`(datfile[, inffile])	Read a Presto format `.dat` file.
`from_tim`(timfile)	Read a sigproc format `.tim` file.
`normalise`([loc_method, scale_method])	Normalise/standardise the time series.
`pad`(npad[, mode])	Pad a time series with mean valued data.
`resample`(accel)	Perform time domain resampling to remove acceleration.
`rfft`([fftn])	Perform 1-D real to complex forward FFT.
`to_dat`([basename])	Write time series in presto `.dat` format.
`to_tim`([filename])	Write time series in sigproc format.

property data#

Time series data array.

Returns:

NDArray[float32]: 1-D time series array.

property header#

Metadata header object.

Returns:

Header: Header object containing metadata.

property nsamples#

Number of samples.

Returns:

int: Number of samples in the time series.

normalise(loc_method='mean', scale_method='std')[source]#

Normalise/standardise the time series.

Normalisation is performed by subtracting the loc estimate, and dividing by the scale estimate of the data.

Parameters:

loc_method{“mean”, “median”}, optional: Method to estimate location to subtract, by default “mean”.
scale_method{“std”, “iqr”, “mad”}, optional: Method to estimate scale to divide by, by default “std”.

Returns:

TimeSeries: Normalised time series.

downsample(factor, filter_method='mean')[source]#

Downsample the time series.

Returned time series is of size nsamples // factor.

Parameters:

factorint: Factor by which to downsample the time series.
filter_method{“mean”, “median”}, optional: Method to downsample, by default ‘mean’.

Returns:

TimeSeries: Downsampled time series.

pad(npad, mode='mean', **pad_kwargs)[source]#

Pad a time series with mean valued data.

Parameters:

npadint: Number of bins to add at the end of the time series.
modestr, optional: Padding mode (as used by numpy.pad), by default ‘mean’.
**pad_kwargsdict: Keyword arguments for numpy.pad.

Returns:

TimeSeries: Padded time series.

deredden(method='mean', window=0.5, *, fast=False)[source]#

Remove low-frequency red noise using a moving filter.

Parameters:

method{‘mean’, ‘median’}, optional: Moving filter function to use, by default ‘mean’.
windowint, optional: Width of moving filter window in seconds, by default 0.5 seconds.
fastbool, optional: Use a faster but less accurate method, by default False.

Returns:

TimeSeries: De-reddened time series.

Raises:

ValueError: If window < 0.

fold(period, accel=0, nbins=50, nints=32)[source]#

Fold time series into discrete phase and sub-integration bins.

Parameters:

periodfloat: Period to fold (in seconds).
accelfloat, optional: Acceleration to fold, by default 0.
nbinsint, optional: Number of phase bins in output, by default 50.
nintsint, optional: Number of sub-integrations in output, by default 32.

Returns:

FoldedData: Data cube containing the folded data.

Raises:

ValueError: If nbins * nints is too large for length of the data.

rfft(fftn=None)[source]#

Perform 1-D real to complex forward FFT.

Time series is zero-padded to the next good size for the FFT.

Parameters:

fftnCallable[[np.ndarray], np.ndarray], optional: The fft function to use. Own fft implementations can be used, e.g, pyfftw.interfaces.numpy_fft.rfft, or mkl_fft.interfaces.numpy_fft.rfft, by default None.

Returns:

FourierSeries: Fourier transform of the time series.

apply_boxcar(width)[source]#

Apply a square-normalized boxcar filter to the time series.

Parameters:

widthint: Width of boxcar to apply in bins.

Returns:

TimeSeries: Filtered time series.

Raises:

ValueError: If width < 1.

Notes

Time series returned is normalized in units of S/N.

resample(accel)[source]#

Perform time domain resampling to remove acceleration.

Parameters:

accelfloat: Acceleration to remove from the time series.

Returns:

TimeSeries: Resampled time series.

correlate(other)[source]#

Perform cross correlation with another time series using FFTs.

This method implements correlation equivalent to scipy.signal.correlate with mode='full' and method='fft'. Correlation lags will be np.arange(-len(other) + 1, nsamples).

Parameters:

otherTimeSeries | np.ndarray: Array to correlate with.

Returns:

TimeSeries: Time series containing the full correlation.

Raises:

OSError: If input array other is not TimeSeries or ndarray.

sigpyproc.timeseries

On this page

sigpyproc.timeseries#