Utils

baseband_utils

get_file_from_timestamp

get_file_from_timestamp(ts, dir_parent, search_type, force_ts=False, acclen=393216)

Given a timestamp, return the file inside of which that timestamp lies. The function works with both baseband and direct spectra files.

Parameters:

Name	Type	Description	Default
ts	int or str	The timestamp you're interested in. (ctime)	required
dir_parent	str	The directory which contains 5 digit folders.	required
search_type	d or f	'd' (directory) if you are using this function for direct spectra, 'f' (file) for baseband.	required
force_ts	bool	Return the next available file if a file containing ts is not found.	False

Returns:

Type	Description
str	Absolute path to the file which contains your timestamp.

Raises:

Type	Description
NotFoundError	If there is no matching file, the user is required to start their integration from the next available timestamp that's helpfully suggested.

time2fnames

time2fnames(time_start, time_stop, dir_parent, search_type, fraglen=5, mind_gap=False)

Gets a list of filenames within specified time-rage.

Given a start and stop ctime, retrieve list of corresponding files. This function assumes that the parent directory has the directory structure /<5-digit coarse time fragment>/<10-digit fine time stamp>.

Parameters:

Name	Type	Description	Default
time_start	int	start time in ctime	required
time_stop	int	stop time in ctime	required
dir_parent	str	parent directory, e.g. /path/to/data_100MHz	required
fraglen	int	number of digits in coarse time fragments	5

Returns:

Type	Description
list of str	List of files in specified time range.

get_init_info

get_init_info(init_t, end_t, dir_parent, force_ts=False)

Get relevant indices from timestamps.

Returns the index of file in a folder and the index of the spectra in that file corresponding to init_timestamp

Parameters:

Name	Type	Description	Default
init_t		Start timestamp of time we are interested in. (ctime)	required
end_t		End of time window we are interested in. (ctime)	required
parent_dir		The directory we search in.	required

Returns:

Name	Type	Description
idxstart	int	The index of the starting row of interest to us within a baseband data file.
fileidx	int	The index of the file in the sorted directory.
files	list of str	Sorted list of path strings to all files in 'parent_dir'.

get_plot_lims

get_plot_lims(pol, acclen)

Get limits for display settings for pretty plots!

Parameters:

Name	Type	Description	Default
pol		Baseband from a baseline. (channelized data.) Is this a 2d array??	required
acclen		Depricated, doesn't get used.	required

Returns:

Name	Type	Description
med	float	Mean power in pol.
vmin	float	Two standard deviations below mean.
vmax	float	Two standard deviations above mean.

plot_4bit

plot_4bit(pol00, pol11, pol01, channels, acclen, time_start, vmin, vmax, opath, minutes=False, logplot=True)

Waterfall plotting routine for 4-bit spectrum integrated data.

Plots and saves figure.

Parameters:

Name	Type	Description	Default
pol00		Channelized autocorr data from pol0.	required
pol11		Channelized autocorr data from pol1.	required
pol01		Channelized x-corr data (pol0 x pol1).	required
channels		Indices of channels of interest.	required
acclen		??	required
time_start		??In what format??	required
vmin		Plotting parameter. Two standard deviations below mean. Used to set colorbar scale.	required
vmax		Plotting parameter. Two standard deviations above mean. Used to set colorbar scale.	required
opath		Ouput path. Path to image of figure to be saved.	required
minutes		Whether to display time in minutes. Defaults to False, displaying seconds.	False
logplot		Defaults to True. Logarithmic scale on y-axis.	True

plot_1bit

plot_1bit(pol01, channels, acclen, time_start, opath, minutes=False, logplot=False)

Waterfall plotting routine for 1-bit spectrum integrated data.

Plots and saves figure.

Parameters:

Name	Type	Description	Default
pol01		Channelized x-corr data (pol0 x pol1).	required
channels		Indices of channels of interest.	required
acclen		??	required
time_start		??In what format; is this relative time or absolute timestamp?	required
opath		Ouput path. Path to image of figure to be saved.	required
minutes		Whether to display time in minutes. Defaults to False, displaying seconds.	False
logplot		Depricated. Defaults to True. Logarithmic scale on y-axis.	False

load_antenna_power

load_antenna_power(init_t, end_t, dir_parent, chans=None, tags=['pol00', 'pol11'])

Load direct spectra for a given antenna. Internally, this function finds all files between two timestamps, loads specified pols (00,11, 01r, 01i) for each file and returns an array of dimensions (npol,nchan,ntime), where ntime ~ (end_t - init_t)/6.44.

Parameters:

Name	Type	Description	Default
init_t		Unix C-time for data start.	required
end_t		Indices of channels of interest.	required
dir_parent		Parent directory for direct data (should contain 5 digit dirs)	required
chans		List of frequency channels in range [0, 2048) that need to be loaded. Default None, (set to the full range)	None
tags		List of polarizations that need to be read. Default ['pol00', 'pol11']	['pol00', 'pol11']

Returns:

Name	Type	Description
large_arr	ndarray	(npol,nchan,ntime) array of all data found between two timestramps
ctime_arr	ndarray	(ntime,) array of C timestamps corresponding to each of the ntime spectra in the data.

get_present_files

get_present_files(t_start, t_end, ant_path_list, T_SCAN=10, tolerance=70)

Given a path to several antenna, will tell you when data is present for each antenna

Args: t_start (int): starting time of when we're scanning t_end (int): ending time of when we're scanning ant_path_list (list of strings): just the directory location for all antenna data T_SCAN (int): the time interval between scans, basically just the dt tolerance (int): maximum time that we tolerate between a point in time and the last present file

Returns: arr (array): array of binary entries, of shape (ntimes, nants). 1 means data is present, 0 means it is absent ntimes is in units of T_SCAN

fig (figure): just visualizes arr, with time on y axis, in human time

get_simul_files

get_simul_files(arr, time_start, dt, desired_ant_indices)

Returns a list of [t_start, t_end] for times during which data is present for all antenna in desired_ant_indices

Parameters:

Name	Type	Description	Default
arr	array	array shape (ntimes, nants) of binary entries, with 1 meaning data is present and 0 meaning data is not present	required
time_start		starting unix time of array, point at which we start counting	required
dt	int	difference in time (seconds) between entries in array	required
desired_ant_indices	list	the antenna (index on arr) for which we want data to be present. If it's -1, it considers ALL antenna in arr.	required

Returns:

Name	Type	Description
runs	list	list of two-element lists [t_start, t_end] which indicate the times of simultaneous continuous data in all desired antenna.

check_data_holes

check_data_holes(t_start, t_end, dir, min_filesize=500001224, tol=60, verbose=False, force_ts=False)

This is a measure for abundance of caution when opening up files. Returns True if there is data missing, or any holes between the data. Returns False if there are no problems (i.e. no data holes)

sat_utils

get_MAD

get_MAD(data, axis=None)

Find real median absolute deviation

get_haversine_dist

get_haversine_dist(p1, p2, radius=6371000)

Vectorized Haversine distance using NumPy.

Parameters:

Name	Type	Description	Default
p1	array-like of shape (..., 2)	lat, lon in degrees	required
p2	array-like of shape (..., 2)	lat, lon in degrees	required
radius	float	Earth radius (km by default)	6371000

Returns:

Name	Type	Description
distances	ndarray	Distance(s) in the same unit as radius.

get_bline_dist

get_bline_dist(coord1, coord2)

returns the physical distance between two coordinates in meters (i.e. magnitude of baseline vectors)

get_risen_sats2

get_risen_sats2(tle_file, coords, t_start, satlist, dt=5, niter=560, good=None, altitude_cutoff=1)

Get all satellites risen at a particular point on earth at a list of epochs. Epochs start at t_start and a list of risen satellites is returned for every t_start + i * dt epoch The satellites are read form a TLE file.

Parameters:

Name	Type	Description	Default
coords	tuple of floats	(latitude, longitude, elevation) of the position on Earth. Elevation is measured in meteres.	required
t_start	float	Start timestamp (ctime). Converted to JD internally.	required
dt	float	Delta between epochs, by default None which sets it to 6.44 seconds internally (accumulation time of direct spectra).	5
niter	int	Number of iterations, by default 560	560
elevation_cutoff	float	Altitude cutoff (in degrees) above which a satellite is considered risen, by default 1 degree.	required

Returns:

Name	Type	Description
risen_sats	list of lists	One list of risen satellites per epoch. Each epoch's list carries the name of the risen satellite at that epoch. E.g. [["FM118","NOAA15"], ["NOAA15"]]

get_detections

get_detections(cx, snr_array, temp_satmap)

determines detections given a coarse-cross correlation and SNR of a satellite pass

for each frequency channel, we sort the SNR across all correlations (uncorrected and per-sat beamform) we skip the channel if uncorrected (cx[0]) has highest SNR (i.e. if beamforming didn't improve SNR) if beamform SNR improves uncorr SNR by factor of (5 * √2) or more, count as detection get detections for each chan for each sat get reliability ratio from the cxcorr peak shape

note that we limit each channel to one detection (i.e. multiple sats cannot be detected in the same chan)

Parameters:

Name	Type	Description	Default
Returns			required
detected_sats			required
detected_peaks			required
detected_snrs			required
rel_ratios			required

get_vis

get_vis(pulse_start_t, pulse_end_t, paths, offsets, T_SPECTRA=4096 / 250000000.0, v_acclen=5000)

Computes visibilities for one baseline for a set interval, given a specnumoffset.

Different to version in helper_finetiming as this must load up the data as BFI object.

Parameters:

Name	Type	Description	Default
Returns			required

Notes

note that this is just a regular CPU visibility computation, mainly useful for sanity checks also note that this should be tested with two non-ref antenna (usually run with one ref one non ref)

unix_to_lst_with_fix

unix_to_lst_with_fix(start_unix, end_unix, coords)

Assume that interval can be no longer than 24 hrs, so that we can unwrap safely by 360

snr2db

snr2db(snrarr)

Sends snr as ratio to dB By convention of snr_times output, sends snr of zero to 0 dB.

sat_utils_gpu

get_cxcorr_many_sats

get_cxcorr_many_sats(p0_ref, p0_nref, tle_path, times, sats_present, satmap, coords, dN, T_SPECTRA=4096 / 250000000.0, c_acclen=10 ** 6)

get the coarse cross-correlation for a given chunk of data, for a given set of sats

Parameters:

Name	Type	Description	Default
p0_ref		Data array of reference/non-reference antenna, shape (c_acclen, nchans). 'Reference' means it is the second in spectrum number, and lower in index. By convention, we always write specnumoffset (and therefore delays) as (nref-ref).	required
tle_path		Directory path of TLE file containing satellite positions for pulse time we care about This allows us to beamform accurately.	required
times		List of [start_time, end_time] for the pulse, in unix time. Lets us know where to beamform	required
sats_present		List of length nsats which tells us which satellites to try and beamform on. Sats are in index form, of legacy hard-coded satlist [28654,25338,33591,57166,59051,44387].	required
satmap		Dictionary that maps each sat_idx to it satID, and vice versa. Lets us move between these forms for utility and ease, depending on which is needed. To perhaps phase out later down the line	required
coords		List of the two sets of coordinates [ref_coords, nref_coords]. Each coord is in form [lat, lon, alt]	required
dN		Gives how many spectra of shift (per direction) we actually include in our cxcorr data. So for example, if dN is 10e5, we will have 2*10e5 different attempted offsets	required
T_SPECTRA		period of spectra	4096 / 250000000.0
c_acclen		Number of spectra in each chunk. Integration time of the cxcorr.	10 ** 6

Returns:

Name	Type	Description
cx	list of arrays	A list of the coarse cross-correlations for each sat in sats_present, and also for no beamforming. Uncorrected always given first. ['Uncorrected', sat1, sat2] Each coarse xcorr has shape (nchans, 2*dN)

get_vis_gpu

get_vis_gpu(pulse_start_t, pulse_end_t, paths, offsets, T_SPECTRA=4096 / 250000000.0, v_acclen=30000)

computes visibilities for one baseline for a set period, given a specnumoffset

note that this is just a regular CPU visibility computation, mainly useful for sanity checks also note that this should be tested with two non-ref antenna (usually run with one ref one non ref)

get_chunk_data

get_chunk_data(files, idxs, chanstart, chanend, nchunks=None, c_acclen=10 ** 6)

Get chunk data for two antenna given times and paths, using bfi.

Purpose of wrapping in a function is to minimize cached memory usage (which is very heavy for bfi files) for modules using large loops.

Parameters:

Name	Type	Description	Default
files		Paths for the two antenna, in form [ref_path, nref_path]	required
idxs		List of starting indices within the first file for the timestream. In form [ref_idx, nref_idx]	required
chanstart		starting/ending channel index in baseband file index	required
c_acclen		length of chunk in spectra	10 ** 6
ref_bool		tells you if you have to switch the order of files and idx lists. used when fitting for coordinates. If true, ref ant is fit ant, so all good. If False, then must switch, since convention is that we write ref ant first, fitting ant first.	required

Returns:

Name	Type	Description
	p0_ref_copy/p0_nref_copy: cupy array	array shape (c_acclen, nchans), with one chunk of data for ref/nref antenna
specnum_offset	int	initial offset of absolute spectrum numbers between the data streams. i.e. for this set of data (this load), what the clock tells you the offset is. The calculated relative offset for THIS chunk load will then update this specnum offset, and give the actual absolute offset between timestreams.

get_snr_from_coords

get_snr_from_coords(fit_coords, nfit_coords, fit_p0, nfit_p0, times, satmap, satID, chan_s_idx, dN=100000, T_SPECTRA=4096 / 250000000.0, c_acclen=10 ** 6)

Get the SNR of a pulse for ONE pulse on ONE baseline using ONE chunk.

Parameters:

Name	Type	Description	Default
Returns			required

get_snr_from_coords_many

get_snr_from_coords_many(fit_coords, fit_path, other_coords, other_paths, pulse_times, pulse_chans, pulse_sats, satmap, dN=1000000.0, T_SPECTRA=4096 / 250000000.0, c_acclen=10000000.0)

Get the total summed SNR for all desired pulses for all desired blines, using a certain fixed fitting antenna.

Purpose is to fit for that desired fitting antenna's coordinates.

Parameters:

Name	Type	Description	Default
Returns			required

orbcomm_utils

ctime2mjd

ctime2mjd(tt=None, type='Dublin')

Return Various Julian Dates given ctime. Options include Dublin, MJD, JD

get_coarse_xcorr

get_coarse_xcorr(f1, f2, Npfb=4096)

Get coarse xcorr of each channel of two channelized timestreams. The xcorr is 0-padded, so length of output is twice the original length (shape[0]).

Parameters:

Name	Type	Description	Default
f1	ndarray of complex64	First and second timestreams. Both n_spectrum x n_channel complex array.	required
f2	ndarray of complex64	First and second timestreams. Both n_spectrum x n_channel complex array.	required
chans		Channels (columns) of f1 and f2 that should be correlated.	required

Returns:

Type	Description
ndarray of complex128	xcorr of each channel's timestream. 2*n_spectrum x n_channel complex array.

get_coarse_xcorr_fast

get_coarse_xcorr_fast(f1, f2, dN, chans=None, Npfb=4096)

Get coarse xcorr of each channel of two channelized timestreams. The xcorr is 0-padded, so length of output is twice the original length (shape[0]).

Parameters:

Name	Type	Description	Default
f1	ndarray of complex64	First and second timestreams. Both n_spectrum x n_channel complex array.	required
f2	ndarray of complex64	First and second timestreams. Both n_spectrum x n_channel complex array.	required
chans		Channels (columns) of f1 and f2 that should be correlated.	None

Returns:

Type	Description
ndarray of complex128	xcorr of each channel's timestream. 2*n_spectrum x n_channel complex array.

get_interp_xcorr

get_interp_xcorr(coarse_xcorr, chan, sample_no, coarse_sample_no)

Get a upsampled xcorr from coarse_xcorr by adding back the carrier frequency.

Parameters:

Name	Type	Description	Default
coarse_xcorr		1-D array of coarse xcorr of one channel.	required
chan	int	Channel for the passed coarse xcorr.	required
osamp		Number of times to over sample over the default 4 ns time-resolution. E.g. osamp=4 means 1 ns time-resolution.	required

Returns:

Name	Type	Description
final_xcorr_cwave	ndarray	Complex upsampled xcorr.

get_interp_xcorr_fast

get_interp_xcorr_fast(coarse_xcorr, chan, sample_no, coarse_sample_no, shift, out=None, shift_phase=False)

Get a upsampled xcorr from coarse_xcorr by adding back the carrier frequency.

Parameters:

Name	Type	Description	Default
coarse_xcorr		1-D array of coarse xcorr of one channel.	required
chan	int	Channel for the passed coarse xcorr.	required
osamp		Number of times to over sample over the default 4 ns time-resolution. E.g. osamp=4 means 1 ns time-resolution.	required

Returns:

Name	Type	Description
final_xcorr_cwave	ndarray	Complex upsampled xcorr.

gauss_smooth

gauss_smooth(data, sigma=5)

Gaussian smooth an N-dim signal. The user should take care about 0-padding. This function will simply smooth whatever is provided with a gaussian kernel of the same size/dimensions with an N-dim FFT.

Parameters:

Name	Type	Description	Default
data	ndarray	Data to smooth	required
sigma	int	width of gaussian in no. of samples, by default 5	5

Returns:

Type	Description
ndarray	Smoothened data

get_risen_sats

get_risen_sats(tle_file, coords, t_start, dt=None, niter=560, good=None, altitude_cutoff=1)

Get all satellites risen at a particular point on earth at a list of epochs. Epochs start at t_start and a list of risen satellites is returned for every t_start + i * dt epoch The satellites are read form a TLE file (currently hardcoded).

Parameters:

Name	Type	Description	Default
coords	tuple of floats	(latitude, longitude, elevation) of the position on Earth. Elevation is measured in meteres.	required
t_start	float	Start timestamp (ctime). Converted to JD internally.	required
dt	float	Delta between epochs, by default None which sets it to 6.44 seconds internally (accumulation time of direct spectra).	None
niter	int	Number of iterations, by default 560	560
elevation_cutoff	float	Altitude cutoff (in degrees) above which a satellite is considered risen, by default 1 degree.	required

Returns:

Name	Type	Description
risen_sats	list of lists	One list of risen satellites per epoch. Each epoch's list carries the name of the risen satellite at that epoch. E.g. [["FM118","NOAA15"], ["NOAA15"]]

find_pulses

find_pulses(x, cond='==', thresh=None, pulses=True)

Given a signal "x", find locations where the signal is ON. Whether or not the signal is ON is determined by the comparison condition passed. Comparison done is x cond 0.

Parameters:

Name	Type	Description	Default
x	ndarray(float or int)	Signal timestream	required
cond	str	One of "==", ">", "<", ">=", "<=", "!=" Consider the effect of round-off error if you are using "eq" with a float array.	'=='
thresh	float	If passed, (x - thresh) is compared against the condition, by default None	None

get_per_ant_sat_delay

get_per_ant_sat_delay(antpos, tle_path, time_start, niter, satnorad, dt=1)

Generate a delay timestream for each antenna for a given satellite.

Parameters:

Name	Type	Description	Default
antpos	list of array-like	List of (latitude, longitude, altitude) for each antenna in decimal degrees. For one antenna, must be a single-element list.	required
tle_path	str	Path to directory with TLE files.	required
time_start	float	Start unix-time of the pass.	required
niter	int	Number of iterations in the pass. Default 1 second per iteration.	required
satnorad	int	Satellite NORAD catalog number.	required
dt	float	delta-time (in seconds) per iteration, by default 1.0	1

Returns:

Type	Description
ndarray	(n_ant x n_time) array of delays

pred

pred(coord1, coord2, start_t, end_t, channel, satID, T_SPECTRA=4096 / 250000000.0, v_acclen=30000)

predicted phase given satellite

orbcomm_utils_gpu

apply_delay

apply_delay(arr, delay, freqs, out=None, copy=True)

Apply a time-dependent exponential phase to a timestream. Timestream can be E-field, or visibility, or anything else a user desires. Output = Input * exp(-j 2 pi freq tau)

Parameters:

Name	Type	Description	Default
arr	ndarray	Usually n_time x n_chan	required
delay	ndarray	Delay timestream of length n_time	required
freqs	ndarray	Frequency array of length n_chan	required
out	ndarray	Write to this array, should be of the same shape as input array, by default None. Overrides copy, if passed.	None
copy	bool	Make a copy of the input array for the output, by default True	True

Returns:

Type	Description
ndarray	out array

finetiming_utils

get_MAD

get_MAD(data, axis=None)

Find real median absolute deviation

haversine

haversine(p1, p2, radius=6371000)

Vectorized Haversine distance using NumPy.

Parameters:

Name	Type	Description	Default
p1	array-like of shape (..., 2)	lat, lon in degrees	required
p2	array-like of shape (..., 2)	lat, lon in degrees	required
radius	float	Earth radius (km by default)	6371000

Returns:

Name	Type	Description
distances	ndarray	Distance(s) in the same unit as radius.

apply_delay

apply_delay(arr, out, delay, freqs)

Apply phase delay to visibilities.

Parameters:

Name	Type	Description	Default
arr	np.ndarry shape (nspec, nchan)	Input array	required
out	np.ndarray shape (nspec, nchan)	Output array	required
delay	float	Time delay between the timestreams, in seconds.	required
freqs	np.ndarray shape (nchan)	Frequencies corresponding to each channel	required

Returns:

Name	Type	Description
out	np.ndarray shape ()	Same object but now with delayed data from 'arr' in it.

xcorr_avg

xcorr_avg(arr1, arr2, acclen)

Computes cross-correlation between two timestreams.

Takes chunks of size 'acclen' and computes averaged conjugate product for each chunk.

Parameters:

Name	Type	Description	Default
arr1	np.ndarray shape (nspec, nchan)	First array for correlation	required
arr2	np.ndarray shape (nspec, nchan)	Second array for correlation	required
acclen	int	number of spectra to include in single chunk of correlated data	required

Returns:

Type	Description
np.ndarray shape (nspec//acclen, nfreq)	Cross-correlated data.

xcorr_avg_1d

xcorr_avg_1d(arr1, arr2, acclen)

Computes cross-correlation between two timestreams, but for only one channel.

Takes chunks of size 'acclen' and computes averaged conjugate product for each chunk.

Parameters:

Name	Type	Description	Default
arr1	np.ndarray shape (nspec)	First array for correlation	required
arr2	np.ndarray shape (nspec)	Second array for correlation	required
acclen	int	number of spectra to include in single chunk of correlated data	required

Returns:

Name	Type	Description
out	np.ndarray shape (nspec//acclen)	Cross-correlated data.

beamformed_xcorr

beamformed_xcorr(data, a1, a2, p1, p2, delay, freqs, acclen)

Compute beamformed visibilities for single baseline for specific polarizations.

Parameters:

Name	Type	Description	Default
data	np.ndarray shape (nblines, npol, nspec, nchans)	Raw baseband data	required
a1	int	Antenna 1 index	required
a2	int	Antenna 2 index	required
p1	int	Polarization 1 index	required
p2	int	Polarization 2 index	required
delay	np.ndarray shape (nspec)	Time delays between antenna 1 and 2	required
freqs	np.ndarray shape (nchans)	Frequencies for each channel	required
acclen	int	Accumulation length for single visibility	required

Returns:

Name	Type	Description
V	np.npdarray shape (nspec//acclen, nchans)	Beamformed visibility data for single baseline and single polarization pair.

get_vis

get_vis(data, satID, freqs, pstart, pend, antpos, ant_idxs, tle_path, T_SPECTRA, acclen)

Computes beamformed visibilities for N antenna over time.

Different to version in sat_utils because this uses data already present on hard drive after upsampling.

Parameters:

Name	Type	Description	Default
data	np.ndarray shape (nblines, npols, nspectra, nchans)	Contains channelized data for each baseline, for the same corresponding spectra	required
satID	int	NORAD satellite ID of the satellite we want to beamform onto	required
freqs	np.ndarray shape (nchans)	Frequencies in Hz corresponding to each channel	required
pstart	int	Pulse start time in 10 digit UNIX timestamp	required
pend	int	Pulse end time in 10 digit UNIX timestamp	required
antpos	list	List of coordinates for each antenna, each entry in form [lat, lon, alt]	required
ant_idxs	list	List of indicies of antenna that we wish to use. Some antenna may have corrupted data.	required
tle_path	string	File path leading to TLE files, which contain satellite position information.	required
T_SPECTRA	float	Period of spectrum, e.g. 1.04 ms for x64 upsampled baseband data	required
acclen	int	Number of spectra to include in each visibility chunk	required

Returns:

Type	Description
np.ndarray shape (nblines, ntimes, nchans)	Visibility Data.

func

func(x, data, freq, weights, fit_phi=False)

Get residuals for given tau at single time sample, for all blines

Parameters:

Name	Type	Description	Default
x	np.ndarray shape (nant-1)	The time offsets for each non-reference antenna	required
data	np.ndarray shape (nblines, nchan)	Visibility data for single timestamp	required
freq	np.npdarray shape (nchan)	Frequencies corresponding to each channel	required
weights	np.npdarray shape (nblines)	Weights (from phase noise) for each baseline.	required
fit_phi	bool	If you want to have an phase offset as well	False

Returns:

Name	Type	Description
	np.ndarray shape (2*nblines*nchan)	Residuals with real and imaginary parts separated for fitting.
Notes
	Reference antenna has zero delay, all other ants are with respect to it.
	That's why tau_t has length nant-1.
	Convention is reference - nonreference.
	Weights can also be per-channel, so shape (nblines, nchan).

func2

func2(x, data, freq, weights, fit_phi=False)

Get residuals for given tau (and optionally phase offsets) at a single time sample.

Parameters:

Name	Type	Description	Default
x	ndarray	If fit_phi=False: x = tau, shape (nant-1,) If fit_phi=True: x = [tau, phi], shape (2*(nant-1),)	required
data	ndarray	Visibility phases for a single timestamp, flattened to shape (nblines * nchan,).	required
freq	ndarray	Frequencies corresponding to each channel.	required
weights	ndarray	Baseline or baseline/channel weights.	required
fit_phi	bool	Whether to fit a constant phase offset for each antenna.	False

Returns:

Type	Description
ndarray	Residual vector with real and imaginary parts concatenated.

jac

jac(tau_t, data, freq, weights)

Get the Jacobian for given tau at single time sample, for all blines

Parameters:

Name	Type	Description	Default
tau_t	np.ndarray shape (nant-1)	Delays for each non-reference antenna	required
data	np.npdarray shape (nblines, nchan)	Visibility data for single timestamp	required
freq	np.npdarray shape (nchans)	Frequency values for each channel	required
weights		Weights for each baseline determined by phase noise.	required

Returns:

Type	Description
np.npdarray shape (2*nblines*nchan, nant-1)	Jacobian of objective function at given tau values.

See also

helper_finetiming.func() The parent objective function.

jac2

jac2(x, data, freq, weights, fit_phi=False)

Jacobian for a single time sample.

Parameters:

Name	Type	Description	Default
x	ndarray	If fit_phi=False: x = tau, shape (nant-1,) If fit_phi=True: x = [tau, phi], shape (2*(nant-1),)	required

find_signal_channels

find_signal_channels(data, sat_type)

Cuts channels by taking median power in time. Isolates signal channels for later.

Parameters:

Name	Type	Description	Default
data	np.ndarray shape (nblines, ntimes, nchan)	Visibility data	required
sat_type	str	Type of satellite (determines bandwidth)	required

Returns:

Type	Description
slice	Slice of data array in channel axis, of corresponding bandwidth to sat_type

find_highest_amp

find_highest_amp(data, chan_slice, window_size=120)

Determine contiguous window of highest median SNR (in terms of amplitude).

Parameters:

Name	Type	Description	Default
data	np.ndarray shape (nblines, ntimes, nchans)	Visibility data	required
chan_slice	slice	Slice along frequency axis where signal is found. Determined by find_signal_channels()	required
window_size	int	Desired number of visibility times inside of which to maximize median SNR (amplitude)	120

Returns:

Type	Description
(start, end)	Start and end indices of visibility times corresponding to optimal window

find_lowest_noise

find_lowest_noise(noise, window_size=120)

Determine continuous window of lowest median phase noise for multi-baseline visibility data.

Parameters:

Name	Type	Description	Default
noise	np.npdarray shape (nblines, ntimes)	Array of phase noise for all baselines at all times.	required
window_size	int	Desired number of visibility times inside of which to minimize median phase noise.	120

Returns:

Type	Description
(int, int)	The best starting and ending time indices (from array noise) Give window of size specified in window_size

discrep_cutting

discrep_cutting(V, satID, acclen=1024)

Full function that goes from beamformed visibilities to final amplitude cut

Eventually phase out.

cost_surface

cost_surface(data, guesses, freqs_normalized, wts, antidx=0, timeidx=0, N1=10001, N2=40, err=None)

Compute and plot the cost surface for func() given initial offsets, and a single antenna direction to vary.

Also includes: weightings, vertical line showing guess value, shaded region showing error on guess fit. Will also determine the offset value of the nearest cost minimum (trough).

Parameters:

Name	Type	Description	Default
data	np.ndarray shape (nblines, ntimes, nchans)	Visibility data	required
guesses	np.ndarray shape (nant-1)	Offset values for each non-reference antenna (units of ns)	required
freqs_normalized	np.ndarray shape (nchans)	Frequencies for each channel (units of GHz)	required
wts	np.npdarray shape (nblines)	Weights corresponding to phase noise	required
antidx	int	Antenna index for which we want to vary tau (the x-axis variable)	0
timeidx	int	The timestamp for which we want to look at the data	0
N1	int	Number of data points to have in the figures	10001
N2	int	Units of tau (in ns) to zoom into, either side of the guess	40
err	float	Error in the guess tau we want to plot (units of ns)	None

Returns:

Name	Type	Description
cost1	np.ndarray shape (N1)	Zoomed out cost surface
cost2	np.ndarray shape (N1)	Zoomed in cost surface.
trough	float	Location of nearest cost minimum to guess value
fig	matplotlib figure	Figure with zoomed out and zoomed in cost surfaces.

get_mask

get_mask(vis, tol=1.5)

get_thermal_noise

get_thermal_noise(vis, ant_idxs, mask=None, T_SPECTRA=4096 / 250000000.0 * 64, acclen=1024)

Computes thermal noise for visibility data, and also plots all visibility phase angles.

Parameters:

Name	Type	Description	Default
vis	np.npdarray shape (nblines, ntimes, nchans)	Visibiilty Data	required
ant_idxs	list	Indices of antenna that we want to use	required
mask	np.npdarray shape (nblines, ntimes, nchans)	Optional way to mask high noise regions out of the main data (RFI masking)	None
T_SPECTRA	float	Single spectrum sampling period	4096 / 250000000.0 * 64
acclen	int	Number of spectra in single visibility time sample	1024

Returns:

Name	Type	Description
thermal_noise	np.npdarray shape (nblines, ntimes)	Phase noise for all blines and all times
fig	matplotlib figure	Figure of all baseline visibility plots

get_prediction_error

get_prediction_error(res, spectra)

Get the error on the predicted UTC discrepancy given residual matrix

symmetrize

symmetrize(A)

Symmetrize a matrix

get_grammian

get_grammian(nant)

Determine grammian for nant anntena.

get_AtA_Atd

get_AtA_Atd(data, Ag, noise_var, nu, nant, nfreq, ntime, fit_constant=True)

Determine both A^T A and A^T d for OLS fitting.

Parameters:

Name	Type	Description	Default
data	np.ndarray shape (ntimes, nchans, nblines)	Data input. Different form than usual, as in column-major ordering	required
Ag		XXXX	required
noise_var		XXXX	required
nu		XXXX	required
nant	int	Number of antenna in entire system (including reference antenna)	required
nfreqs	int	Number of channels	required
ntime	int	Number of time samples we are fitting over	required
fit_constant	bool	XXXX	True

pfb_utils

StreamingPFB

StreamingPFB(nant, npol, timestream_size=50000, lblock=4096, ntap=4, window='hamming', dtype='float')

Works for arbitrary timestream sizes. Can be less than lblock.

StreamingPFB_

StreamingPFB_(nant, npol, timestream_size=50000, lblock=4096, ntap=4, window='hamming', dtype='float')

Works for arbitrary timestream sizes. Can be less than lblock.

cupy_ipfb

cupy_ipfb(dat, matft, thresh=0.0)

On-device ipfb. Expects the data to be iPFB'd to live in GPU memory.

Parameters:

Name	Type	Description	Default
dat	ndarray	nspec x nchan array of complex64	required
matft	ndarray	lblock x nspec array of float32. Note that this is transpose of Jon's original convention for speed reasons. [w0 wN w2N w3N 0 0 . . . ][w1 . 0 0 ] [ . . . . ][wN-1 . . . w4N-1 . 0 ]	required
thresh	float	Wiener filter threshold, by default 0.0	0.0

Returns:

Type	Description
ndarray	nspec*nchan timestream values as a C-major matrix of shape (nspec x nchan)

math_utils