pulsarbat.RadioSignal#
- class pulsarbat.RadioSignal(z, /, *, sample_rate, start_time=None, center_freq, chan_bw, freq_align='center', meta=None)[source]#
Class for heterodyned radio signals.
A radio signal is often heterodyned, so a center frequency (
center_freq
) and a channel bandwidth (chan_bw
) must be provided to determine the frequencies of each channel. Optionally,freq_align
is provided to indicate how the channel frequencies are positioned. Usually, the signal data here would be the result of some sort of filterbank.The signal data (
z
) must have at least 2 dimensions where the first two dimensions refer to time (axis=0
) and frequency (axis=1
), respectively.- Parameters:
- zarray-like
The signal data. Must be at least 2-dimensional with shape
(nsample, nchan, ...)
.- sample_rate
Quantity
The number of samples per second. Must be in units of frequency.
- start_time
Time
, optional The start time of the signal (that is, the time at the first sample of the signal). Default is None.
- center_freq
Quantity
The frequency at the center of the signal’s band. Must be in units of frequency.
- chan_bw
Quantity
The bandwidth of a channel. The total bandwidth is
chan_bw * nchan
. Must be in units of frequency.- freq_align{‘bottom’, ‘center’, ‘top’}, default: ‘center’
The alignment of channels relative to the
center_freq
. Default is ‘center’ (as with odd-length complex DFTs). ‘bottom’ and ‘top’ only have an effect whennchan
is even.- meta
dict
, optional Any metadata that the user might want to attach to the signal.
See also
Notes
The input signal array must have shape
(nsample, nchan, ...)
, wherensample
is the number of time samples, andnchan
is the number of frequency channels.The channels must be adjacent in frequency and of equal channel bandwidth, such that the frequency of a channel
i
is given by,:freq_i = center_freq + chan_bw * (i + a - nchan/2)
where
i
is in[0, ..., nchan - 1]
,nchan
is the number of channels (z.shape[1]
),chan_bw
is the bandwidth of a single channel, anda
is in{0, 0.5, 1}
depending on the value offreq_align
.An even-length complex-valued DFT (as implemented in
numpy.fft.fft()
) would havefreq_align = 'bottom'
with a center frequency of 0, whereas an odd-length DFT would havefreq_align = 'center'
due to symmetry.freq_align = 'top'
is used in cases where the DFT was conducted on the lower sideband of the signal instead.When
nchan
is odd, thefreq_align
attribute will internally be fixed to ‘center’ since the channels must be necessarily centered on thecenter_freq
.Attributes
axes_labels
Dictionary of axes labels.
bandwidth
Total bandwidth.
center_freq
Center frequency.
chan_bw
Channel bandwidth.
channel_freqs
Center frequencies of all channels.
data
The signal data.
dt
Sample spacing of the signal (
1 / sample_rate
).dtype
Data type of the signal.
freq_align
Alignment of channel frequencies.
max_freq
Frequency at top of the band.
meta
Signal metadata.
min_freq
Frequency at bottom of the band.
nchan
Number of frequency channels.
ndim
Number of dimensions in data.
sample_rate
Sample rate of the signal.
sample_shape
Shape of a sample.
shape
Shape of the signal.
start_time
Start time of the signal (Time at first sample).
stop_time
Stop time of the signal (Time at sample after the last sample).
time_length
Length of signal in time units.
Methods
compute
(**kwargs)Returns signal with computed data.
contains
(t, /)Whether time(s) are within the bounds of the signal.
get_axis
(axis)Axis number from an integer or axis label.
like
(obj[, z])Returns a signal object "like" given signal object.
persist
(**kwargs)Returns signal with data persisted in memory.
rechunk
([chunks])Rechunks signal data if stored as a Dask array.
Returns signal with data as a Dask array.
- compute(**kwargs)#
Returns signal with computed data.
Has no effect unless data is stored as a Dask Array.
kwargs
are passed on to thedask.compute()
function.
- contains(t, /)#
Whether time(s) are within the bounds of the signal.
- get_axis(axis)#
Axis number from an integer or axis label.
- classmethod like(obj, z=None, /, **kwargs)#
Returns a signal object “like” given signal object.
This classmethod inspects the class signature and creates a signal object like the given reference object. The signal data (as an array) must always be provided as well as any other arguments that are intended to be different than the reference object. All arguments required by the class signature that are not provided are acquired from attributes of the same name in the reference object.
- Parameters:
- obj
Signal
The reference signal object.
- zarray-like
The signal data.
- **kwargs
Additional keyword arguments to pass on to the target class.
- obj
- persist(**kwargs)#
Returns signal with data persisted in memory.
Has no effect unless data is stored as a Dask Array.
kwargs
are passed on to thedask.persist()
function.
- rechunk(chunks=None, **kwargs)#
Rechunks signal data if stored as a Dask array.
If the underlying data is not a Dask array, it is converted to one first. By default, there is no chunking along the first dimension and other dimensions are automatically chunked based on the default chunk size set in Dask’s configuration.
kwargs
are passed along todask.array.rechunk()
.
- to_dask_array()#
Returns signal with data as a Dask array.
Uses
dask.array.asanyarray()
to convert signal data to a Dask array.