def check_state_vec(state_channel, gpsb, gpse, threshold,
condition='greater_equal'):
if condition == 'greater_equal':
cond_function = np.greater_equal
else:
if condition == 'equal':
cond_function = np.equal
else:
if condition == 'greater':
cond_function = np.greater
else:
if condition == 'less_equal':
cond_function = np.less_equal
else:
if condition == 'less':
cond_function = np.less
else:
raise ValueError('Condition must be either '
'"greater_equal", "greater",'
'"equal", "less_equal" or "less"')
print "Acquiring state vector"
with vrg.getChannel(state_channel, gpsb, gpse - gpsb) as sttvct:
print "Evaluating state vector and generating state flag"
sttflag = (cond_function(sttvct.data, threshold)).astype('int')
# finding the points where the flag changes state
u = np.diff(np.concatenate(([0], sttflag, [0])))
ups = np.where(u == 1, u, 0)
# shifting the ones (raising state) by one step,
# removing length 1 segments
u = (u - ups + np.roll(ups, 1))[1:]
starts = np.argwhere(u == 1)
stops = np.argwhere(u == -1)
segments = np.hstack((starts, stops))/sttvct.fsample + gpsb
print len(segments)
return segments
def process_channel(ch_p, src, segments):
step_instances = []
for step in ch_p.steps.itervalues():
step_instances.append(step['class'](step))
brms_buffer = []
for (j, (gpsb, gpse)) in enumerate(segments):
with vrg.getChannel(src, ch_p.channel, gpsb, gpse - gpsb) as r_data:
# Decimate the channel data
ch_data = decimator_wrapper(r_data, ch_p.ds_freq)
fft_per_segment = ch_p.compute_n_fft(len(ch_data))
# Estimate the time vector of the brms samples for this seg
t_disc = ch_p.compute_time_vector(fft_per_segment, gpsb)
# Reset the buffers if there is a time discontinuity between
# this segment and the previous one.
if t_disc:
for step in step_instances:
step.reset()
# make a function with the spectra as output?
# il ciclo su fft_per segment lo terrei cmq qui. o no?
for i in xrange(fft_per_segment):
freqs, s = sig.welch(
ch_data[i * ch_p.n_over_points:i * ch_p.n_over_points +
ch_p.n_points],
fs=ch_p.ds_freq,
window='hanning',
nperseg=ch_p.n_points)
pipe = (freqs, s)
for step in step_instances:
print "Processing step {} for channel {}".format(
step, ch_p.channel)
pipe = step(pipe)
# todo:use an hdf5 buffer between segments?probably unnecessary
brms_buffer.append(pipe[1])
real_bands = pipe[0]
return brms_buffer, real_bands
def process_channel(ch_p, src, segments):
step_instances = []
for step in ch_p.steps.itervalues():
step_instances.append(step['class'](step))
brms_buffer = []
step_times = np.zeros(len(step_instances))
fft_time = 0
append_time = 0
for (j, (gpsb, gpse)) in enumerate(segments):
print 'segment {} of {}'.format(j, len(segments))
with vrg.getChannel(src, ch_p.channel, gpsb, gpse - gpsb) as r_data:
# Decimate the channel data
ch_data = decimator_wrapper(ch_p.ds_freq, r_data)
fft_per_segment = ch_p.compute_n_fft(len(ch_data))
print fft_per_segment
# Estimate the time vector of the brms samples for this seg
t_disc = ch_p.compute_time_vector(fft_per_segment, gpsb)
# Reset the buffers if there is a time discontinuity between
# this segment and the previous one.
if t_disc:
for step in step_instances:
step.reset()
# make a function with the spectra as output?
# il ciclo su fft_per segment lo terrei cmq qui. o no?
for i in xrange(fft_per_segment):
if i % 100 == 1:
print "FFt number {} of segment {} of {}".format(
i, j, len(segments))
tot_time = fft_time + append_time + np.sum(step_times)
perc_time = float(tot_time) / 100
print "FFT time: {:.2f}%, step time = {:.2f}%, append time" \
" = {:.2f}%".format(fft_time/perc_time,
np.sum(step_times)/perc_time,
append_time/perc_time)
t_0 = time()
freqs, s = sig.welch(
ch_data[i * ch_p.n_over_points:i * ch_p.n_over_points +
ch_p.n_points],
fs=ch_p.ds_freq,
window='hanning',
nperseg=ch_p.n_points)
fft_time += time() - t_0
pipe = (freqs, s)
for k, step in enumerate(step_instances):
if i % 100 == 1:
print step
print "step fraction time:{:.2f}".format(
float(step_times[k] * 100) / np.sum(step_times))
t_0 = time()
pipe = step(pipe)
step_times[k] += time() - t_0
# todo:use an hdf5 buffer between segments?probably unnecessary
t_0 = time()
brms_buffer.append(pipe[1])
real_bands = pipe[0]
append_time += time() - t_0
# transpose the buffer in order for it to be used
brms_buffer = np.array(brms_buffer).T.tolist()
return brms_buffer, real_bands
def getRawData(ch, gps, dt):
"""Read data from RAW file:
ch = channel name
gps = gps time
dt = duration
"""
with getChannel('raw', ch, gps, dt) as x:
return x.data, x.fsample
def read_virgo_timeseries(source, channel, gstart, gstop_or_dur):
"""quick and dirty function to read virgo data as timeseries"""
# TODO use gwpy
from virgotools import getChannel
with getChannel(source, channel, gstart, gstop_or_dur) as data:
return TimeSeries(data.data,
unit=data.unit,
t0=gstart,
dt=data.dt,
channel=channel)
def get_spectrum(gps_start, duration):
chan = 'V1:Hrec_hoft_16384Hz'
x = getChannel('raw', chan, gps_start, duration)
Fs = x.fsample
# Define parameters for FFT
stride = 5.0 # FFT stride in seconds
overlap = stride / 2 # overlap in seconds (50%)
Pxx, freq, t = specgram(x.data,
NFFT=int(stride * Fs),
Fs=int(Fs),
noverlap=int(overlap * Fs))
Axx = sqrt(Pxx)
ASD = median(Axx, 1) / sqrt(log(2))
return ASD, freq
'figure.figsize': fig_size,
'font.family': "serif",
'font.serif': ["Times"]
}) #,
# 'xtick.major.pad':'8'})
# Define parameters for FFT
stride = 6.0 # FFT stride in seconds
overlap = 3.0 # overlap in seconds (50%)
#### Grab Hrec data
gps_start = 1180111697
duration = 100
chan = 'V1:Hrec_hoft_16384Hz'
x = getChannel('raw', chan, gps_start, duration)
Fs = x.fsample
# for science segment, get PSD.
# 6-second FFTs, 50% overlap, hann window (default)
Pxx, V1freq, t = specgram(x.data,
NFFT=int(stride * Fs),
Fs=int(Fs),
noverlap=int(overlap * Fs))
Axx = sqrt(Pxx)
V1_ASD_NIWI = median(Axx, 1) / sqrt(log(2))
#### Grab Hrec data
gps_start = 1179865427
duration = 100
err_chan = 'V1:LSC_DARM_ERR'
# grab data in steps, 50% overlap
step_duration = 50
t_DARM = arange(0, duration, step_duration / 2)
datapoints = zeros(len(t_DARM))
demod_I = zeros(len(t_DARM))
demod_Q = zeros(len(t_DARM))
for i in range(len(t_DARM)):
gps_start = start_time + i * step_duration / 2
x = getChannel('raw', err_chan, gps_start, step_duration)
Fs = x.fsample
ts = arange(0, duration, 1. / Fs)
pre_line, t_dec = data_demod(ts,
x.data,
f0,
Fs,
step_duration,
lp=2500.0,
Fs0=10000.0)
datapoints[i] = median(pre_line)
print i, gps_start, datapoints[i]
fignum = 0
def read_virgo_timeseries(source,
channel,
t0,
gstop_or_dur,
mask=False,
fill_value=np.nan,
remote=False):
"""Function to read virgo data as timeseries.
This should one day be included in gwpy.
Parameters
----------
source : `str`
Frame file, either a full path to a ffl or gwf file, or an
abbreviation like 'raw', 'trend', which are looked up in
the stancard location. If omitted, defaults to 'raw',
but this default value is deprecated.
channel : `str`
Source datastream for these data.
If missing, a prefix 'V1:' is added.
t0 : `~gwpy.time.LIGOTimeGPS`, `float`, `str`
GPS epoch corresponding to starting time,
any input parsable by `~gwpy.time.to_gps` is fine
gstop_or_dur: `~gwpy.time.LIGOTimeGPS`, `float`, `str`
GPS epoch corresponding to end time,
any input parsable by `~gwpy.time.to_gps` is fine
If a `float` < 1e6 is provided, it corresponds to a duration
in seconds from `t0`.
mask : `bool`, optional
If mask is False, missing samples will be replaced
by fill_value. If it is True, the returned FrVect will
have an attribute missing, which is a mask vector that
is zero for missing samples.
Default : False
fill_value : `float`, optional
Value that is used for missing samples if mask is False.
Default: np.nan
remote: `bool` optional
If False, use PythonVirgoTools to parse raw data files.
If True, use gwpy.TimeSeries.get(), but takes longer.
Default : False, assuming the script is ran on Virgo server.
Examples
--------
>>> from virgotools import getChannel
Load the data from channel 'INJ_IMC_TRA_DC', from
Sep 14 2015 09:50:45.391, and a duration of 10s
>>> x = getChannel('raw', 'INJ_IMC_TRA_DC', 'Sep 14 2015 09:50:45.391', 10)
That can be simply visualise:
>>> import matplotlib.pyplot as plt
>>> plt.plot(x.time, x.data)
>>> plt.show()
Same, using 2 GPS times:
>>> x = getChannel('raw', 'INJ_IMC_TRA_DC', 1126259462.3910, 1126259472.3910)
"""
# Convert to gps times in seconds
# Use the Seconds and NanoSeconds instead of ns()
# Because otherwise one needs to multiply by 1e-9
# And this can cause rounding approximation
sec = to_gps(t0).gpsSeconds
nsec = to_gps(t0).gpsNanoSeconds
tstart = str(sec) + '.' + str(nsec)
gstart = float(tstart)
sec = to_gps(gstop_or_dur).gpsSeconds
nsec = to_gps(gstop_or_dur).gpsNanoSeconds
tend = str(sec) + '.' + str(nsec)
if float(tend) < 1e6:
gstop = gstart + float(tend)
else:
gstop = float(tend)
# If the script is running on Virgo's server.
if not remote:
from virgotools import getChannel
# Parse Virgo files
with getChannel(source,
channel,
gstart,
gstop,
mask=mask,
fill_value=fill_value) as data:
data = TimeSeries(data.data,
unit=data.unit,
t0=gstart,
dt=data.dt,
channel=channel)
else:
# If not running the script on Virgo's server. Takes longer
# Query is working, but crashes when computing q_transform.
# Data might not be the same format as with PythonVirgoTools.
# Further checks required.
if channel[:3] not in ['V1:', 'H1:', 'L1:']:
print(
'When accessing the data outside the virgo server, the channel must start with `V1:`, `H1:` or `L1:` '
)
data = TimeSeries.get(channel, gstart, gstop)
return data
PRCL_phi = array([])
SSFS_phi = array([])
PRCL_UGF = array([])
SSFS_UGF = array([])
DARM_UGF = array([])
MICH_UGF = array([])
B4_112mag = array([])
for i in range(len(seg_stop)):
if duration[i] < 100:
continue
print i, seg_start[i], seg_stop[i], duration[i]
chan = 'LSC_B2_8MHz_DPHI'
x = getChannel('rds',chan,seg_start[i],duration[i]-10)
PRCL_phi = append(PRCL_phi,x.data)
chan = 'LSC_B4_56MHz_DPHI'
x = getChannel('rds',chan,seg_start[i],duration[i]-10)
SSFS_phi = append(SSFS_phi,x.data)
chan = 'LSC_DARM_UGF'
x = getChannel('rds',chan,seg_start[i],duration[i]-10)
DARM_UGF = append(DARM_UGF,x.data)
chan = 'LSC_SSFS_UGF'
x = getChannel('rds',chan,seg_start[i],duration[i]-10)
SSFS_UGF = append(SSFS_UGF,x.data)
chan = 'LSC_MICH_UGF'
def get_channel_data(self, data_source, channel, gpsb, gpse):
# TODO: extract other stuff from ch i.e. the units?
with vrg.getChannel(data_source, channel, gpsb, gpse - gpsb) as ch:
data = decimator_wrapper(self.down_freq, ch)
return data
#
# Find segments and save seglist
#
################################################
# C8 start and end times
gps_start = 1178028018
# prior to 17:00 May 5 there were some times when DQ_META_ITF_Mode was not defined, start at 17:00
#gps_start = 1178038817
gps_stop = 1178262018
duration = gps_stop - gps_start
chan = 'DQ_META_ITF_Mode'
x = getChannel('rds', chan, gps_start, duration)
Fs = x.fsample
t = arange(0, duration, 1. / Fs) + gps_start
seg_start = []
seg_stop = []
seg_flag = False
for i in range(len(x.data)):
# If a segment has just started
if x.data[i] == 1 and seg_flag == False:
seg_start.append(t[i])
seg_flag = True