def three_in_one(wanted='X', save='yes'):
path1 = r"C:\Users\Lenovo\Desktop\Internship\NoisyData" + "\\" + str(
wanted) + str(205) + ".npy"
path2 = r"C:\Users\Lenovo\Desktop\Internship\NoisyData" + "\\" + str(
wanted) + str(305) + ".npy"
path3 = r"C:\Users\Lenovo\Desktop\Internship\NoisyData" + "\\" + str(
wanted) + str(405) + ".npy"
data1 = np.load(path1)
data2 = np.load(path2)
data3 = np.load(path3)
tau1 = np.logspace(0, 4, 1000)
rate = [0.01, 1, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100]
fig = plt.figure(figsize=(12, 7)) #width , height
fig.suptitle(
"Allan Deviation for " + wanted +
" for different rates \n BLUE - 400 \n GREEN - 650 \n RED - 900")
fig.subplots_adjust(hspace=0.9, wspace=0.8) #height , width
for i, r in enumerate(rate):
(tau21, ad1, ade, adn) = allantools.oadev(data1,
rate=r,
data_type="freq",
taus=tau1)
(tau22, ad2, ade, adn) = allantools.oadev(data2,
rate=r,
data_type="freq",
taus=tau1)
(tau23, ad3, ade, adn) = allantools.oadev(data3,
rate=r,
data_type="freq",
taus=tau1)
plt.subplot(3, 4, i + 1)
plt.loglog(tau21, ad1, 'b')
plt.loglog(tau22, ad2, 'g')
plt.loglog(tau23, ad3, 'r')
plt.title("Rate " + str(r))
plt.xlabel('tau')
plt.ylabel('ADEV [V]')
plt.grid()
plt.show()
save_path = r"C:\Users\Lenovo\Desktop\Internship\NoisyData\\AllanAll" + wanted
if save == 'yes':
plot_path = save_path + ".png"
fig.savefig(plot_path)
def test_phasedat_oadev(self):
s32_rows = testutils.read_stable32( 'phase_dat_oadev_octave.txt' , 1.0 )
phase = testutils.read_datafile('PHASE.DAT')
(taus,devs,errs,ns) = allan.oadev(phase, taus=[s32['tau'] for s32 in s32_rows])
# CI computation
# alhpa +2,...,-4 noise power
# d 1 first-difference variance, 2 allan variance, 3 hadamard variance
# alpha+2*d >1
# m tau/tau0 averaging factor
# F filter factor, 1 modified variance, m unmodified variance
# S stride factor, 1 nonoverlapped estimator, m overlapped estimator (estimator stride = tau/S )
# N number of phase observations
los=[]
his=[]
for (d,t, n) in zip(devs, taus, ns):
edf2 = allan.edf_greenhall( alpha=0, d=2, m=int(t), N=len(phase), overlapping = True, modified=False )
(lo,hi) = allan.confidence_interval( dev=d, edf=edf2 )
los.append(lo)
his.append(hi)
# compare to Stable32
print("oadev()")
for (s32, t2, d2, lo2, hi2, n2) in zip(s32_rows, taus, devs, los, his, ns):
print("s32 %03d %03f %1.6f %1.6f %1.6f" % (s32['n'], s32['tau'], s32['dev_min'], s32['dev'], s32['dev_max']))
print("at %03d %03f %1.6f %1.6f %1.6f" % (n2, t2, round(lo2,5), round(d2,5), round(hi2,5) ))
testutils.check_approx_equal(s32['dev_min'], lo2, tolerance=1e-3)
testutils.check_approx_equal(s32['dev'], d2, tolerance=1e-4)
testutils.check_approx_equal(s32['dev_max'], hi2, tolerance=1e-3)
print("----")
def test_oadev_ci_and_noiseID(self):
""" ADEV with confidence intervals, including noise-ID """
change_to_test_dir()
s32rows = testutils.read_stable32(
resultfile='stable32_OADEV_octave.txt', datarate=1.0)
for row in s32rows:
phase = testutils.read_datafile('gps_1pps_phase_data.txt.gz')
(taus, devs, errs, ns) = allan.oadev(phase,
rate=rate,
data_type="phase",
taus=[row['tau']])
dev = devs[0]
#print(dev/row['dev'])
assert np.isclose(dev, row['dev'], rtol=1e-2, atol=0)
try:
# CI including noise-ID
(lo2,
hi2) = allan.confidence_interval_noiseID(phase,
dev,
af=int(row['m']),
dev_type="oadev",
data_type="phase")
print(" tau= %f lo/s32_lo = %f hi/s32_hi = %f " %
(row['tau'], lo2 / row['dev_min'], hi2 / row['dev_max']))
assert np.isclose(lo2, row['dev_min'], rtol=1e-2, atol=0)
assert np.isclose(hi2, row['dev_max'], rtol=1e-2, atol=0)
except NotImplementedError:
print("can't do CI for tau= %f" % row['tau'])
pass
def allan_deviation(data, rate):
"""Returns the Allan deviation. Makes use of `oadev` from the `allantools`
repository
Parameters
----------
data: array_like(float)
The data to be processed
rate: float
The sampling rate in Hz
Returns
-------
tau: list of float
The taus used in the Allan deviation
dev: list of float
The Allan deviations.
Examples
--------
>>>foo()
"""
tau, dev, dev_error, N = oadev(data, rate=rate, data_type='freq')
return tau, dev
def test_oadev_ci(self):
s32rows = testutils.read_stable32(resultfile='oadev_decade.txt',
datarate=1.0)
for row in s32rows:
data = testutils.read_datafile(data_file)
(taus, devs, errs, ns) = allan.oadev(data,
rate=rate,
taus=[row['tau']])
edf = allan.edf_greenhall(alpha=row['alpha'],
d=2,
m=row['m'],
N=len(data),
overlapping=True,
modified=False,
verbose=True)
(lo, hi) = allan.confidence_intervals(devs[0],
ci=0.68268949213708585,
edf=edf)
print("n check: ", testutils.check_equal(ns[0], row['n']))
print("dev check: ",
testutils.check_approx_equal(devs[0], row['dev']))
print(
"min dev check: ", lo, row['dev_min'],
testutils.check_approx_equal(lo,
row['dev_min'],
tolerance=1e-3))
print(
"max dev check: ", hi, row['dev_max'],
testutils.check_approx_equal(hi,
row['dev_max'],
tolerance=1e-3))
def allan_deviation(data, rate):
tau1 = np.logspace(0, 4, 1000) # tau values from 1 to 10000
(tau2, ad, ade, adn) = allantools.oadev(data,
rate=rate,
data_type="freq",
taus=tau1)
return tau2, ad
def test_oadev_ci(self):
s32rows = testutils.read_stable32(resultfile='oadev_octave.txt',
datarate=1.0)
for row in s32rows:
data = testutils.read_datafile(data_file)
data = allan.frequency2fractional(data, mean_frequency=1.0e7)
(taus, devs, errs, ns) = allan.oadev(data,
rate=rate,
data_type="freq",
taus=[row['tau']])
# NOTE! Here we use alhpa from Stable32-results for the allantools edf computation!
edf = allan.edf_greenhall(alpha=row['alpha'],
d=2,
m=row['m'],
N=len(data),
overlapping=True,
modified=False,
verbose=True)
(lo, hi) = allan.confidence_interval(devs[0], edf=edf)
print("n check: ", testutils.check_equal(ns[0], row['n']))
print(
"dev check: ", devs[0], row['dev'],
testutils.check_approx_equal(devs[0],
row['dev'],
tolerance=2e-3))
print(
"min dev check: ", lo, row['dev_min'],
testutils.check_approx_equal(lo,
row['dev_min'],
tolerance=2e-3))
print(
"max dev check: ", hi, row['dev_max'],
testutils.check_approx_equal(hi,
row['dev_max'],
tolerance=2e-3))
def AllanDeviation(self,id=29850,spacing=1.0):
self.curve = cs.CurveDB.objects.get(id=id)
dat = self.curve.data
l=len(dat)
rate = 1/float(spacing) # data rate in Hz
taus = np.logspace(np.log10(spacing),np.log10(spacing*l),5000)
# fractional frequency data
#(taus_used, adev, adeverror, adev_n) = allantools.adev(dat.values, rate, taus)
(taus_used, adev, adeverror, adev_n) = allantools.oadev(dat.values, rate, taus)
self.adev = cs.CurveDB.create(taus_used,adev,name="Allan deviation")
self.curve.add_child(self.adev)
def main_use_allantools() :
f_0 = 3e8 / 730e-9 # the frequency of 730nm light
timet, fract_data = read_data(r'./Ca_clock_transition_data.xls')
timet.pop()
fract_data.pop()
for i, elem in enumerate(fract_data) :
fract_data[i] = elem / f_0
(t2, ad, ade, adn) = allantools.oadev(fract_data, rate=11.2,
data_type="freq", taus = [i for i in range(1, len(timet) + 1)])
plt.loglog(t2, ad)
plt.savefig('ad.png', dpi=400)
def plotallan_phase(plt, y, rate, taus, style, label):
(t2, ad, ade, adn) = allantools.oadev(y,
data_type='phase',
rate=rate,
taus=taus)
plt.loglog(
t2,
ad,
style,
fillstyle='none',
label=label,
)
def test_oadev_ci(self):
s32rows = testutils.read_stable32(resultfile='oadev_decade.txt', datarate=1.0)
for row in s32rows:
data = testutils.read_datafile(data_file)
(taus, devs, errs, ns) = allan.oadev(data, rate=rate,
taus=[ row['tau'] ])
edf = allan.edf_greenhall(alpha=row['alpha'],d=2,m=row['m'],N=len(data),overlapping=True, modified = False, verbose=True)
(lo,hi) =allan.confidence_intervals(devs[0],ci=0.68268949213708585, edf=edf)
print("n check: ", testutils.check_equal( ns[0], row['n'] ) )
print("dev check: ", testutils.check_approx_equal( devs[0], row['dev'] ) )
print("min dev check: ", lo, row['dev_min'], testutils.check_approx_equal( lo, row['dev_min'], tolerance=1e-3 ) )
print("max dev check: ", hi, row['dev_max'], testutils.check_approx_equal( hi, row['dev_max'], tolerance=1e-3 ) )
def oadev(data, taus, scale=1 / 429228004229873, alpha=0):
"""Calculate overlapping Allan deviation with non-naive errors, from frequency data file.
Args:
data (array): array of fractional frequency data file whose first column is a timestamp in (s)
taus (list of float): list of tau-values for OADEV computation
scale (float, optional): scaling factor for fractional frequency. Defaults to 1/Sr87_BIPM_frequency
alpha (int, optional): +2,...,-4 noise type, either estimated or known. Defaults to 0 to match Stable 32
Returns:
t2 (list of floats): list of tau_values for which deviations were computed
ad_ff (list of floats): list of oadev deviations in fractional frequency units
err_lo_ff (list of floats): list of non-naive lower 1-sigma errors for each point over which
deviations were computed
err_hi_ff (list of floats): list of non-naive higher 1-sigma errors for each point over
which deviations were computed
adn (list): list of number of pairs in overlapping allan computation
"""
x = data[:, 0] # Get timestamps
y = data[:, 1] # Get fractional frequency data
avg_interval = (x[-1] - x[0]) / len(
x) # average interval between measurements
r = 1 / avg_interval # average sample rate in Hz of the input data
t = np.array(taus) * avg_interval # tau values on which to evaluate metric
(t2, ad, ade, adn) = allantools.oadev(
y, rate=r, data_type="freq",
taus=t) # normal ODEV computation, giving naive 1/sqrt(N) errors
# correct for deadtime ad/np.sqrt(B2*B3)
# https://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication1065.pdf | 5.15 Dead Time
# TODO
# Get correct (Stable32) errors
err_lo, err_hi = get_better_ade(t2,
ad,
avg_interval,
len(x),
alpha=alpha,
d=2,
overlapping=True,
modified=False)
ad_ff = ad * scale
err_lo_ff = err_lo * scale
err_hi_ff = err_hi * scale
return t2, ad_ff, err_lo_ff, err_hi_ff, adn
def allandeviation(wanted="X", number=0):
if wanted == "X":
path_allan = SavingAndReadingData.path + "\X" + str(number) + ".npy"
if wanted == "Y":
path_allan = SavingAndReadingData.path + "\Y" + str(number) + ".npy"
values = np.load(path_allan)
tau = []
for i in range(1, (len(values) + 1)):
tau.append(i)
tau1 = np.logspace(0, 4, 1000) # tau values from 1 to 1000
# values = np.random.randn(10000)
r = 20 # sampling rate
(tau2, ad, ade, adn) = allantools.oadev(values,
rate=20,
data_type="freq",
taus=tau1)
(tau22, ad2, ade2, adn2) = allantools.oadev(values,
rate=1,
data_type="freq",
taus=tau1)
fig = plt.figure(figsize=(8, 6))
# fig.subplots_adjust(hspace=0.5, wspace=0.3)
plt.subplot(1, 2, 1)
plt.loglog(tau2, ad, 'b')
plt.plot([10000], [10000])
plt.xlabel('tau')
plt.ylabel('ADEV [V]')
plt.grid()
plt.subplot(1, 2, 2)
plt.plot([10000], [10000])
plt.loglog(tau22, ad2, 'b')
plt.xlabel('tau')
plt.ylabel('ADEV [V]')
plt.grid()
plt.show()
def test_oadev_ci(self):
""" Overlapping ADEV with confidence intervals """
s32rows = testutils.read_stable32(resultfile='oadev_octave.txt', datarate=1.0)
for row in s32rows:
data = testutils.read_datafile(data_file)
data = allan.frequency2fractional(data, mean_frequency=1.0e7)
(taus, devs, errs, ns) = allan.oadev(data, rate=rate, data_type="freq",
taus=[ row['tau'] ])
# NOTE! Here we use alhpa from Stable32-results for the allantools edf computation!
edf = allan.edf_greenhall(alpha=row['alpha'],d=2,m=row['m'],N=len(data),overlapping=True, modified = False, verbose=True)
(lo, hi) =allan.confidence_interval(devs[0], edf=edf)
print("n check: ", testutils.check_equal( ns[0], row['n'] ) )
print("dev check: ", devs[0], row['dev'], testutils.check_approx_equal( devs[0], row['dev'], tolerance=2e-3 ) )
print("min dev check: ", lo, row['dev_min'], testutils.check_approx_equal( lo, row['dev_min'], tolerance=2e-3 ) )
print("max dev check: ", hi, row['dev_max'], testutils.check_approx_equal( hi, row['dev_max'], tolerance=2e-3 ) )
def test_phasedat_oadev(self):
(s32_taus, s32_devs, s32_devs_lo, s32_devs_hi,
s32_ns) = read_stable32('phase_dat_oadev_octave.txt', 1.0)
phase = read_datafile('PHASE.DAT')
(taus, devs, errs, ns) = allan.oadev(phase, taus=s32_taus)
# CI computation
# alhpa= +2,...,-4 noise power
# d= 1 first-difference variance, 2 allan variance, 3 hadamard variance
# alpha+2*d >1
# m = tau/tau0 averaging factor
# F filter factor, 1 modified variance, m unmodified variance
# S stride factor, 1 nonoverlapped estimator, m overlapped estimator (estimator stride = tau/S )
# N number of phase obs
los = []
his = []
for (d, t, n) in zip(devs, taus, ns):
#edf = greenhall_simple_edf( alpha=0, d=2, m=t, S=1, F=t, N=len(phase) )
edf2 = allan.edf_greenhall(alpha=0,
d=2,
m=int(t),
N=len(phase),
overlapping=True,
modified=False)
#print(edf,edf2,edf2/edf)
(lo,
hi) = allan.confidence_intervals(dev=d,
ci=0.68268949213708585,
edf=edf2) # 0.68268949213708585
#allan.uncertainty_estimate(len(phase), t, d,ci=0.683,noisetype='wf')
los.append(lo)
his.append(hi)
#print greenhall_simple_edf( alpha=0, d=2, m=1, S=1, F=1, N=len(phase) )
#print confidence_intervals( dev
# tau N edf chi2 chi2 dev_lo dev dev_hi
# 1 999 782.030 821.358 742.689 2.8515e-01 2.9223e-01 2.9987e-01
# 2 997 540.681 573.374 507.975 1.9520e-01 2.0102e-01 2.0738e-01
for (t1, d1, lo1, hi1, n1, t2, d2, lo2, hi2,
n2) in zip(s32_taus, s32_devs, s32_devs_lo, s32_devs_hi, s32_ns,
taus, devs, los, his, ns):
print("s32 %03d %03f %1.6f %1.6f %1.6f" % (n1, t1, lo1, d1, hi1))
print("at %03d %03f %1.6f %1.6f %1.6f" %
(n2, t2, round(lo2, 5), round(d2, 5), round(hi2, 5)))
approx_equal(lo1, lo2, tolerance=1e-3)
approx_equal(hi1, hi2, tolerance=1e-3)
print("----")
def oallan_plot(y,
plotaxis,
rate,
taus,
style,
marker,
x_lab,
yl_lab,
label,
title,
legends=True):
y, label = ratio_builder_allan(y, plotaxis, label)
colors = colormap_generator(plotaxis)
if np.size(y) != 0:
y_size = np.shape(y)[0]
else:
y_size = 0
plt.figure()
ax = plt.subplot(111,
xscale="log",
yscale="log",
xlabel=x_lab,
ylabel=yl_lab,
title=title)
for i in range(y_size):
yy = check_mask(y[i])
col = next(colors)
(t2, ad, ade, adn) = at.oadev(yy,
rate=rate,
data_type="freq",
taus='all')
mini = np.argmin(ad[:int(np.size(ad) / 2)])
allan_minimum = '{:1.1e}'.format(ad[mini])
time_minimum = '{:1.1e}'.format(t2[mini])
label[i] = label[i].replace(
"Amplitude", "") + " [" + allan_minimum + "; " + time_minimum + "]"
ax.loglog(t2,
ad,
"g.",
c=col,
ls=style[i % np.size(style)],
marker=marker[i % np.size(marker)],
label=label[i]) #label[i])
if legends == True:
ax.legend(loc='best')
plt.show()
def plotallan_oadev(plt, y, rate, taus, style):
(t2, ad, ade, adn) = allantools.oadev(y,
rate=rate,
data_type="freq",
taus=taus)
plt.loglog(t2, ad, style)
ad_min_y = round(ad.min(), 4) #save Allan minimum
ad_min_x = round(t2[np.argmin(ad)]) #save corresponding averaging time/tau
string = "Allan Minimum " + str(ad_min_y) + "\nIntegration time " + str(
ad_min_x)
plt.text(ad_min_x, ad_min_y + (ad.max() - ad_min_y) / 2, string)
plt.ylabel('Allan Deviation [per mil]')
plt.xlabel('Time [sec]')
ttext = plt.title('Allan Deviation')
plt.setp(ttext, size='large', color='r', weight='bold')
return t2, ad, ade, adn
def test_oadev_ci(self):
s32rows = testutils.read_stable32(resultfile="oadev_octave.txt", datarate=1.0)
for row in s32rows:
data = testutils.read_datafile(data_file)
data = allan.frequency2fractional(data, mean_frequency=1.0e7)
(taus, devs, errs, ns) = allan.oadev(data, rate=rate, data_type="freq", taus=[row["tau"]])
edf = allan.edf_greenhall(
alpha=row["alpha"], d=2, m=row["m"], N=len(data), overlapping=True, modified=False, verbose=True
)
(lo, hi) = allan.confidence_intervals(devs[0], ci=0.68268949213708585, edf=edf)
print("n check: ", testutils.check_equal(ns[0], row["n"]))
print("dev check: ", devs[0], row["dev"], testutils.check_approx_equal(devs[0], row["dev"], tolerance=2e-3))
print(
"min dev check: ", lo, row["dev_min"], testutils.check_approx_equal(lo, row["dev_min"], tolerance=2e-3)
)
print(
"max dev check: ", hi, row["dev_max"], testutils.check_approx_equal(hi, row["dev_max"], tolerance=2e-3)
)
def test_phasedat_oadev(self):
s32_rows = testutils.read_stable32('phase_dat_oadev_octave.txt', 1.0)
phase = testutils.read_datafile('PHASE.DAT')
(taus, devs, errs,
ns) = allan.oadev(phase, taus=[s32['tau'] for s32 in s32_rows])
# CI computation
# alhpa +2,...,-4 noise power
# d 1 first-difference variance, 2 allan variance, 3 hadamard variance
# alpha+2*d >1
# m tau/tau0 averaging factor
# F filter factor, 1 modified variance, m unmodified variance
# S stride factor, 1 nonoverlapped estimator, m overlapped estimator (estimator stride = tau/S )
# N number of phase observations
los = []
his = []
for (d, t, n) in zip(devs, taus, ns):
#edf = greenhall_simple_edf( alpha=0, d=2, m=t, S=1, F=t, N=len(phase) )
edf2 = allan.edf_greenhall(alpha=0,
d=2,
m=int(t),
N=len(phase),
overlapping=True,
modified=False)
#print(edf,edf2,edf2/edf)
(lo,
hi) = allan.confidence_intervals(dev=d,
ci=0.68268949213708585,
edf=edf2) # 0.68268949213708585
#allan.uncertainty_estimate(len(phase), t, d,ci=0.683,noisetype='wf')
los.append(lo)
his.append(hi)
for (s32, t2, d2, lo2, hi2, n2) in zip(s32_rows, taus, devs, los, his,
ns):
print("s32 %03d %03f %1.6f %1.6f %1.6f" %
(s32['n'], s32['tau'], s32['dev_min'], s32['dev'],
s32['dev_max']))
print("at %03d %03f %1.6f %1.6f %1.6f" %
(n2, t2, round(lo2, 5), round(d2, 5), round(hi2, 5)))
testutils.check_approx_equal(s32['dev_min'], lo2, tolerance=1e-3)
testutils.check_approx_equal(s32['dev_max'], hi2, tolerance=1e-3)
print("----")
def test_oadev_ci_and_noiseID(self):
""" ADEV with confidence intervals, including noise-ID """
change_to_test_dir()
s32rows = testutils.read_stable32(resultfile='stable32_OADEV_octave.txt', datarate=1.0)
for row in s32rows:
phase = testutils.read_datafile('gps_1pps_phase_data.txt.gz')
(taus, devs, errs, ns) = allan.oadev(phase, rate=rate, data_type="phase",
taus=[row['tau']])
dev = devs[0]
#print(dev/row['dev'])
assert np.isclose(dev, row['dev'], rtol=1e-2, atol=0)
try:
# CI including noise-ID
(lo2, hi2) = allan.confidence_interval_noiseID(phase, dev, af=int(row['m']), dev_type="oadev", data_type="phase")
print(" tau= %f lo/s32_lo = %f hi/s32_hi = %f "% (row['tau'], lo2/row['dev_min'], hi2/row['dev_max']))
assert np.isclose(lo2, row['dev_min'], rtol=1e-2, atol=0)
assert np.isclose(hi2, row['dev_max'], rtol=1e-2, atol=0)
except NotImplementedError:
print("can't do CI for tau= %f"%row['tau'])
pass
def plot_allan(samples):
taus = np.logspace(-3, 3, 100)
(t2, ad, ade, adn) = allantools.oadev(samples,
rate=100,
data_type='freq',
taus=taus)
(tau_n, n, line_n) = estimate_random_walk(t2, ad)
(tau_k, k, line_k) = estimate_rate_random_walk(t2, ad)
(tau_b, b, line_b) = estimate_bias_instability(t2, ad)
plt.subplot(111, xscale='log', yscale='log')
plt.loglog(t2, ad)
plt.grid(True)
plt.text(tau_n, n, 'N: ' + str(n))
plt.plot(t2, line_n)
plt.text(tau_k, k, 'K: ' + str(k))
plt.plot(t2, line_k)
plt.text(tau_b, b, 'B: ' + str(b))
plt.plot(t2, line_b)
plt.legend(['sigma', 'sigma_N', 'sigma_K', 'sigma_B'])
def get_arw(self,
seconds=60,
autophase=False,
autohome=True,
scale_factor=None,
rate=None):
"""
Performs a tombstone test of the gyro, but rather than returning a
:class:`pyfog.tombstone.Tombstone` object, it returns the ARW in units
of deg/√h. If this is not set, the gyro will run the
:func:`hardware.gyros.get_scale_factor` routine.
Args:
seconds (float): The number of seconds
minutes (float): The number of minutes
hours (float): The number of hours
rate (float): The sample rate
autophse (bool): If true, the gyro will return to the home position
before running the tombstone test.
scale_factor (float): The scale factor to use in order to convert
between lock-in amplifier voltage and rotation rate in units of
°/h/Volt. If this is not set, the gyro will run the
Returns:
float: The angular random walk in units of °/√h
"""
if not scale_factor:
scale_factor = self.get_scale_factor()
data = daq.read(seconds, rate=rate)
if not rate:
# duration isn't defined in this method, even in the original code
rate = len(data / duration)
_, dev, _, _ = oadev(data * scale_factor,
rate=rate,
data_type='freq',
taus=[1])
return dev[0] / 60
""" ########################## # test stable32plot.py # ########################## """ import allantools from pylab import figure, show, plot from stable32plot import sigmaplot, dataplot #import 2 functions: sigmaplot,dataplot """#------------generate random data and cal adev-----------------""" x1 = allantools.noise.white(1000) (taus, adevs, errors, ns) = allantools.adev(x1) (taust, adevst, errorst, nst) = allantools.tdev(x1) (tauso, adevso, errorso, nso) = allantools.oadev(x1) x2 = allantools.noise.white(1000, 0.6) (taus2, adevs2, errors2, ns2) = allantools.oadev(x2) x3 = allantools.noise.white(1000, 0.5) (taus3, adevs3, errors3, ns3) = allantools.oadev(x3) x4 = allantools.noise.white(1000, 0.4) (taus4, adevs4, errors4, ns4) = allantools.oadev(x4) x5 = allantools.noise.white(1000, 0.3) (taus5, adevs5, errors5, ns5) = allantools.oadev(x5) x6 = allantools.noise.white(1000, 0.2) (taus6, adevs6, errors6, ns6) = allantools.oadev(x6)
plt.loglog(f_fi[1:],[h0*v0*v0/(ff*ff) for ff in f_fi[1:]],label='h_0 * v0^2 * f^-2' )
plt.legend(framealpha=0.5)
plt.title('PSD of phase (radians)')
plt.xlabel('Frequeny / Hz')
plt.ylabel('one-sided PSD / S_fi(f)')
plt.figure()
plt.loglog(f_x,psd_x,label='numpy.fft()')
plt.loglog(f_x2,psd_x2,label='scipy.signal.welch()')
plt.loglog(f_x[1:],[h0/(2*math.pi*ff)**2 for ff in f_x[1:]],label='h0/(2*pi*f)**2' )
plt.legend(framealpha=0.5)
plt.title('PSD of phase (time)')
plt.xlabel('Frequeny / Hz')
plt.ylabel('one-sided PSD / S_x(f)')
plt.figure()
taus=np.logspace(-2.2,4,100)
(taus_y, devs_y, errs_y, ns_y) = allantools.oadev(frequency=y, rate=fs, taus=taus)
(taus_x, devs_x, errs_x, ns_x) = allantools.oadev(phase=x, rate=fs, taus=taus)
plt.loglog(taus_y,devs_y,'o',label='ADEV from y')
plt.loglog(taus_x,devs_x,'*',label='ADEV from x')
adev_y = [math.sqrt( 0.5*h0*(1.0/tt) ) for tt in taus]
plt.loglog(taus,adev_y,label='sqrt( 0.5*h0*tau^-1 )')
plt.xlim((8e-3,1e3))
plt.legend(framealpha=0.6)
plt.title('Allan deviation')
plt.xlabel('Tau / s')
plt.ylabel('Allan deviation')
plt.show()
def moving_allan(y, x, length, size, rate, plotaxis, yl_lab, title, label):
y_new, label = ratio_builder_allan(y, plotaxis, label)
if np.shape(y_new)[1] != 0:
y_size = np.shape(y_new)[0]
else:
y_size = 0
k = int((np.shape(y_new)[1] - length) / size) + 1
x_allan = (length / 2.0 + np.arange(0, k) * size) / rate + x[0]
minimas = np.zeros((2, y_size, k))
for i in range(y_size):
yy = check_mask(y_new[i])
for j in range(k):
(t2, ad, ade, adn) = at.oadev(yy[j * size:length + j * size],
rate=rate,
data_type="freq",
taus='all')
minimas[0][i][j] = (t2[np.argmin(ad[:int(length / 2)])])
minimas[1][i][j] = (np.min(ad[:int(length / 2)]))
H, xedges, yedges = (np.histogram2d(minimas[1][i],
minimas[0][i],
bins=10))
H /= k
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
X, Y = np.meshgrid(xedges[:-1] + 0.1 * xedges[0],
yedges[:-1] + 0.1 * yedges[0])
X = X.flatten('F')
Y = Y.flatten('F')
H = H.flatten()
Z = np.zeros_like(X)
dx = 0.8 * abs(xedges[1] - xedges[0])
dy = 0.8 * abs(yedges[1] - yedges[0])
ax.bar3d(X, Y, Z, dx, dy, H, zsort='average')
ax.set_xlabel('Allan minimum')
ax.set_ylabel('Integration time (sec)')
ax.set_zlabel('normalized frequency')
fig.show()
plt.figure()
ax1 = plt.subplot(211)
ax2 = plt.subplot(212, sharex=ax1)
ax3 = ax2.twinx()
x = np.linspace(0, np.size(yy) / rate, num=np.size(yy))
ax1.plot(x, yy)
# ax1.set_xlim
ax2.set_yscale('log')
ax2.errorbar(x_allan,
minimas[1][0],
xerr=length / (2 * rate),
c='r',
ls='',
label='Allan minima')
# ax2.set_xlim([min(x),max(x)])
ax3.plot(x_allan,
minimas[0][0],
c='b',
ls='',
marker='o',
label="Integration time")
ax1.set_ylabel(yl_lab)
ax3.set_ylabel("Time (sec)")
ax2.set_ylabel("Allan minimum")
lines2, labels2 = ax2.get_legend_handles_labels()
lines3, labels3 = ax3.get_legend_handles_labels()
ax3.legend(lines2 + lines3, labels2 + labels3, loc=4)
plt.title(title)
plt.show()
def main():
nr = 2**14 # number of datapoints in time-series
adev0 = 1.0e-11
#
# for each noise type generated with cn.noiseGen()
# compute
# - phase PSD and coefficient g_b
# - frequency PSD and coefficient h_a
# - ADEV
# - MDEV
#
tau0 = 0.1 # sample interval for simulated time-series
sample_rate = 1.0 / tau0
qd0 = qd1 = qd2 = qd3 = qd4 = pow(adev0,
2) # discrete variance for noiseGen()
# This scaling makes all g_i coefficients equal
# -> PSDs cross at 1 Hz.
qd1 = qd0 * 2 * np.pi / sample_rate
qd2 = qd1 * 2 * np.pi / sample_rate
qd3 = qd2 * 2 * np.pi / sample_rate
qd4 = qd3 * 2 * np.pi / sample_rate
print("generating timeseries..."),
x0 = cn.noiseGen(nr, qd0, 0) # white phase noise (WPM)
x1 = cn.noiseGen(nr, qd1, -1) # flicker phase noise (FPM)
x2 = cn.noiseGen(nr, qd2, -2) # white frequency noise (WFM)
x3 = cn.noiseGen(nr, qd3, -3) # flicker frequency noise (FFM)
x4 = cn.noiseGen(nr, qd4, -4) # random walk frequency noise (RWFM)
# compute frequency time-series
y0 = at.phase2frequency(x0, sample_rate)
y1 = at.phase2frequency(x1, sample_rate)
y2 = at.phase2frequency(x2, sample_rate)
y3 = at.phase2frequency(x3, sample_rate)
y4 = at.phase2frequency(x4, sample_rate)
print("done.")
print("computing PSDs..."),
# compute phase PSD
(f0, psd0) = at.noise.scipy_psd(x0, f_sample=sample_rate, nr_segments=4)
(f1, psd1) = at.noise.scipy_psd(x1, f_sample=sample_rate, nr_segments=4)
(f2, psd2) = at.noise.scipy_psd(x2, f_sample=sample_rate, nr_segments=4)
(f3, psd3) = at.noise.scipy_psd(x3, f_sample=sample_rate, nr_segments=4)
(f4, psd4) = at.noise.scipy_psd(x4, f_sample=sample_rate, nr_segments=4)
# compute phase PSD prefactor g_b
g0 = cn.phase_psd_from_qd(qd0, 0, tau0)
g1 = cn.phase_psd_from_qd(qd1, -1, tau0)
g2 = cn.phase_psd_from_qd(qd2, -2, tau0)
g3 = cn.phase_psd_from_qd(qd3, -3, tau0)
g4 = cn.phase_psd_from_qd(qd4, -4, tau0)
print g0
print g1
print g2
print g3
print g4
# compute frequency PSD
(ff0, fpsd0) = at.noise.scipy_psd(y0, f_sample=sample_rate, nr_segments=4)
(ff1, fpsd1) = at.noise.scipy_psd(y1, f_sample=sample_rate, nr_segments=4)
(ff2, fpsd2) = at.noise.scipy_psd(y2, f_sample=sample_rate, nr_segments=4)
(ff3, fpsd3) = at.noise.scipy_psd(y3, f_sample=sample_rate, nr_segments=4)
(ff4, fpsd4) = at.noise.scipy_psd(y4, f_sample=sample_rate, nr_segments=4)
# compute frequency PSD prefactor h_a
a0 = cn.frequency_psd_from_qd(qd0, 0, tau0)
a1 = cn.frequency_psd_from_qd(qd1, -1, tau0)
a2 = cn.frequency_psd_from_qd(qd2, -2, tau0)
a3 = cn.frequency_psd_from_qd(qd3, -3, tau0)
a4 = cn.frequency_psd_from_qd(qd4, -4, tau0)
print "a_i == g_i*(4pi^2)"
print a0, g0 * pow(2 * np.pi, 2)
print a1, g1 * pow(2 * np.pi, 2)
print a2, g2 * pow(2 * np.pi, 2)
print a3, g3 * pow(2 * np.pi, 2)
print a4, g4 * pow(2 * np.pi, 2)
print("done.")
print("computing ADEV/MDEV..."),
# compute ADEV
(t0, d0, e, n) = at.oadev(x0, rate=sample_rate)
(t1, d1, e, n) = at.oadev(x1, rate=sample_rate)
(t2, d2, e, n) = at.oadev(x2, rate=sample_rate)
(t3, d3, e, n) = at.oadev(x3, rate=sample_rate)
(t4, d4, e, n) = at.oadev(x4, rate=sample_rate)
# compute MDEV
(mt0, md0, e, n) = at.mdev(x0, rate=sample_rate)
(mt1, md1, e, n) = at.mdev(x1, rate=sample_rate)
(mt2, md2, e, n) = at.mdev(x2, rate=sample_rate)
(mt3, md3, e, n) = at.mdev(x3, rate=sample_rate)
(mt4, md4, e, n) = at.mdev(x4, rate=sample_rate)
print("done.")
plt.figure()
# Phase PSD figure
plt.subplot(2, 2, 1)
plt.loglog(f0, [g0 * pow(xx, 0.0) for xx in f0],
'--',
label=r'$g_0f^0 = {h_2\over4\pi^2}f^0$',
color='black')
plt.loglog(f0, psd0, '.', color='black')
plt.loglog(f1[1:], [g1 * pow(xx, -1.0) for xx in f1[1:]],
'--',
label=r'$g_{-1}f^{-1} = {h_1\over4\pi^2}f^{-1} $',
color='red')
plt.loglog(f1, psd1, '.', color='red')
plt.loglog(f2[1:], [g2 * pow(xx, -2.0) for xx in f2[1:]],
'--',
label=r'$g_{-2}f^{-2} = {h_0\over4\pi^2}f^{-2} $',
color='green')
plt.loglog(f2, psd2, '.', color='green')
plt.loglog(f3[1:], [g3 * pow(xx, -3.0) for xx in f3[1:]],
'--',
label=r'$g_{-3}f^{-3} = {h_{-1}\over4\pi^2}f^{-3} $',
color='pink')
plt.loglog(f3, psd3, '.', color='pink')
plt.loglog(f4[1:], [g4 * pow(xx, -4.0) for xx in f4[1:]],
'--',
label=r'$g_{-4}f^{-4} = {h_{-2}\over4\pi^2}f^{-4}$',
color='blue')
plt.loglog(f4, psd4, '.', color='blue')
plt.grid()
plt.legend(framealpha=0.9, fontsize=22)
plt.title(r'Phase Power Spectral Density')
plt.xlabel(r'Frequency (Hz)')
plt.ylabel(r' $S_x(f)$ $(s^2/Hz)$')
# frequency PSD figure
plt.subplot(2, 2, 2)
plt.loglog(ff0[1:], [a0 * pow(xx, 2) for xx in ff0[1:]],
'--',
label=r'$h_{2}f^{2}$',
color='black')
plt.loglog(ff0, fpsd0, '.', color='black')
plt.loglog(ff1[1:], [a1 * pow(xx, 1) for xx in ff1[1:]],
'--',
label=r'$h_{1}f^{1}$',
color='red')
plt.loglog(ff1, fpsd1, '.', color='red')
plt.loglog(ff2[1:], [a2 * pow(xx, 0) for xx in ff2[1:]],
'--',
label=r'$h_{0}f^{0}$',
color='green')
plt.loglog(ff2, fpsd2, '.', color='green')
plt.loglog(ff3[1:], [a3 * pow(xx, -1) for xx in ff3[1:]],
'--',
label=r'$h_{-1}f^{-1}$',
color='pink')
plt.loglog(ff3, fpsd3, '.', color='pink')
plt.loglog(ff4[1:], [a4 * pow(xx, -2) for xx in ff4[1:]],
'--',
label=r'$h_{-2}f^{-2}$',
color='blue')
plt.loglog(ff4, fpsd4, '.', color='blue')
plt.grid()
plt.legend(framealpha=0.9, fontsize=22)
plt.title(r'Frequency Power Spectral Density')
plt.xlabel(r'Frequency (Hz)')
plt.ylabel(r' $S_y(f)$ $(1/Hz)$')
# ADEV figure
plt.subplot(2, 2, 3)
f_h = 0.5 / tau0
K0 = 3.0 * f_h / (4.0 * pow(np.pi, 2))
plt.loglog(t0, [np.sqrt(K0 * a0) / xx for xx in t0],
'--',
label=r'$\sqrt{3f_H\over4\pi^2} \sqrt{h_{2}}\tau^{-3/2}$',
color='black')
plt.loglog(t0, d0, 'o', color='black')
# IEEE 1139-2008, table B.2
def K1(f_H, tau):
gamma = 1.038 / 2
return ((3.0 / 2.0) * np.log(2.0 * np.pi * f_h * tau) +
gamma) / (2.0 * pow(np.pi, 2))
plt.loglog(
t1, [np.sqrt(K1(f_h, xx) * a1) / xx for xx in t1],
'--',
label=
r'$\sqrt{ 3\ln{(2 \pi f_H \tau)} + \gamma \over 4\pi^2 }\sqrt{h_{1}}\tau^{-1}$',
color='red')
plt.loglog(t1, d1, 'o', color='red')
K2 = 0.5
plt.loglog(t2, [np.sqrt(K2 * a2) / math.sqrt(xx) for xx in t2],
'--',
label=r'$\sqrt{1\over2} \sqrt{h_{0}}\tau^{-1/2}$',
color='green')
plt.loglog(t2, d2, 'o', color='green')
K3 = 2 * np.log(2)
plt.loglog(t3, [np.sqrt(K3 * a3) for xx in t3],
'--',
label=r'$ \sqrt{2\ln{2}}\sqrt{h_{-1}}\tau^0$',
color='pink')
plt.loglog(t3, d3, 'o', color='pink')
K4 = (2 * np.pi**2.0) / 3.0
plt.loglog(t4, [np.sqrt(K4 * a4) * math.sqrt(xx) for xx in t4],
'--',
label=r'$\sqrt{2\pi^2\over3}\sqrt{h_{-2}}\tau^{+1/2}$',
color='blue')
plt.loglog(t4, d4, 'o', color='blue')
plt.legend(framealpha=0.9, loc='lower left', fontsize=22)
plt.grid()
plt.title(r'Allan Deviation')
plt.xlabel(r'$\tau$ (s)')
plt.ylabel(r'ADEV')
# MDEV
plt.subplot(2, 2, 4)
M0 = 3.0 / (8.0 * pow(np.pi, 2))
plt.loglog(t0, [np.sqrt(M0 * a0) / pow(xx, 3.0 / 2.0) for xx in t0],
'--',
label=r'$\sqrt{ 3\over 8\pi^2 } \sqrt{h_{2}}\tau^{-3/2}$',
color='black')
plt.loglog(mt0, md0, 'o', color='black')
M1 = (24.0 * np.log(2) - 9.0 * np.log(3)) / (8.0 * pow(np.pi, 2))
plt.loglog(
t1, [np.sqrt(M1 * a1) / xx for xx in t1],
'--',
label=r'$\sqrt{ 24\ln(2)-9\ln(3) \over 8\pi^2 } \sqrt{h_{1}}\tau^{-1}$',
color='red')
plt.loglog(mt1, md1, 'o', color='red')
M2 = 0.25
plt.loglog(t2, [np.sqrt(M2 * a2) / math.sqrt(xx) for xx in t2],
'--',
label=r'$ \sqrt{1\over4}\sqrt{h_{0}}\tau^{-1/2}$',
color='green')
plt.loglog(mt2, md2, 'o', color='green')
M3 = 27.0 / 20.0 * np.log(2)
plt.loglog(t3, [np.sqrt(M3 * a3) for xx in t3],
'--',
label=r'$\sqrt{ {27\over20}\ln(2) } \sqrt{h_{-1}}\tau^0$',
color='pink')
plt.loglog(mt3, md3, 'o', color='pink')
M4 = 11.0 / 20.0 * pow(np.pi, 2)
plt.loglog(t4, [np.sqrt(M4 * a4) * math.sqrt(xx) for xx in t4],
'--',
label=r'$ \sqrt{ {11\over20}\pi^2 } \sqrt{h_{-2}}\tau^{+1/2}$',
color='blue')
plt.loglog(mt4, md4, 'o', color='blue')
plt.legend(framealpha=0.9, loc='lower left', fontsize=22)
plt.grid()
plt.title(r'Modified Allan Deviation')
plt.xlabel(r'$\tau$ (s)')
plt.ylabel(r'MDEV')
plt.show()
########################## # test stable32plot.py # ########################## """ import allantools from pylab import figure,show,plot from stable32plot import sigmaplot,dataplot#import 2 functions: sigmaplot,dataplot """#------------generate random data and cal adev-----------------""" x1 = allantools.noise.white(1000) (taus, adevs, errors, ns) = allantools.adev(x1) (taust, adevst, errorst, nst) = allantools.tdev(x1) (tauso, adevso, errorso, nso) = allantools.oadev(x1) x2=allantools.noise.white(1000,0.6) (taus2,adevs2,errors2,ns2)=allantools.oadev(x2) x3=allantools.noise.white(1000,0.5) (taus3,adevs3,errors3,ns3)=allantools.oadev(x3) x4=allantools.noise.white(1000,0.4) (taus4,adevs4,errors4,ns4)=allantools.oadev(x4) x5=allantools.noise.white(1000,0.3) (taus5,adevs5,errors5,ns5)=allantools.oadev(x5) x6=allantools.noise.white(1000,0.2) (taus6,adevs6,errors6,ns6)=allantools.oadev(x6)
# S_x
plt.figure()
plt.loglog(f_x, psd_x, label='numpy.fft()')
plt.loglog(f_x2, psd_x2, label='scipy.signal.welch()')
plt.loglog(f_x, [h2 / (2 * math.pi)**2] * len(f_x), label='h2/(4*pi^2)')
plt.legend(framealpha=0.5)
plt.title('PSD of phase (time)')
plt.xlabel('Frequeny / Hz')
plt.ylabel('one-sided PSD / S_x(f)')
plt.grid()
# ADEV figure
plt.figure()
taus = [tt for tt in np.logspace(-7, 4, 100)]
(taus_y, devs_y, errs_y, ns_y) = allantools.oadev(y,
data_type='freq',
rate=fs,
taus=taus)
(taus_x, devs_x, errs_x, ns_x) = allantools.oadev(x,
data_type='phase',
rate=fs,
taus=taus)
plt.loglog(taus_y, devs_y, 'o', label='ADEV from y')
plt.loglog(taus_x, devs_x, '*', label='ADEV from x')
print(devs_y)
print(devs_x)
fh = 0.5 * fs # this restricts the noise power to below fh
adev_y = [
math.sqrt(3 * fh / (4 * math.pi**2) * h2 * (1.0 / tt**2)) for tt in taus
]
plt.loglog(taus, adev_y, label='sqrt( 3*fh/(4*pi^2) * h2*tau^-2 )')
mpu_sample_rate = len(data_mpu) / last_mpu_timestamp
adis_sample_rate = len(data_adis) / last_adis_timestamp
print("Last MPU Timestamp: " + str(last_mpu_timestamp))
print("Last ADIS Timestamp: " + str(last_adis_timestamp))
print("MPU Sample Rate: " + str(mpu_sample_rate))
print("ADIS Sample Rate: " + str(adis_sample_rate))
taus = [tt for tt in np.logspace(-1, 4, 100)]
if (allan):
if sensor == 'gyro':
(taus, devs_x, _, _) = allantools.oadev(data_adis[:, 0],
rate=adis_sample_rate,
data_type='freq',
taus=taus)
(taus, devs_y, _, _) = allantools.oadev(data_adis[:, 1],
rate=adis_sample_rate,
data_type='freq',
taus=taus)
(taus, devs_z, _, _) = allantools.oadev(data_adis[:, 2],
rate=adis_sample_rate,
data_type='freq',
taus=taus)
plt.loglog(taus, devs_x, label='ADIS Gyro X')
plt.loglog(taus, devs_y, label='ADIS Gyro Y')
plt.loglog(taus, devs_z, label='ADIS Gyro Z')
for (t, x, y, z) in zip(taus, devs_x, devs_y, devs_z):
def allandeviation(density='low', wanted="X", number=0, save='yes'):
if density == 'low':
if wanted == "X":
path_allan = r"C:\Users\Lenovo\Desktop\Internship\Data\LessDense" + "\X" + str(
number) + ".npy"
save_path = r"C:\Users\Lenovo\Desktop\Internship\Data\LessDense" + "\Xallan" + str(
number)
if wanted == "Y":
path_allan = r"C:\Users\Lenovo\Desktop\Internship\Data\LessDense" + "\Y" + str(
number) + ".npy"
save_path = r"C:\Users\Lenovo\Desktop\Internship\Data\LessDense" + "\Yallan" + str(
number)
if density == 'high':
if wanted == "X":
path_allan = r"C:\Users\Lenovo\Desktop\Internship\Data\MoreDense" + "\X" + str(
number) + ".npy"
save_path = r"C:\Users\Lenovo\Desktop\Internship\Data\MoreDense" + "\Xallan" + str(
number)
if wanted == "Y":
path_allan = r"C:\Users\Lenovo\Desktop\Internship\Data\MoreDense" + "\Y" + str(
number) + ".npy"
save_path = r"C:\Users\Lenovo\Desktop\Internship\Data\MoreDense" + "\Yallan" + str(
number)
if density == 'random':
wanted = "Random"
save_path = r"C:\Users\Lenovo\Desktop\Internship\Data\allanRandom" + str(
number)
values = np.load(path_allan)
# tau = []
# for i in range(1,(len(values)+1)):
# tau.append(i)
tau1 = np.logspace(0, 4, 1000) # tau values from 1 to 10000
# values = np.random.randn(10000)
rate = [0.01, 1, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100]
fig = plt.figure(figsize=(12, 6)) #width , height
# Global title
fig.suptitle("Allan Deviation for " + wanted + " for different rates")
for i, r in enumerate(rate):
(tau2, ad, ade, adn) = allantools.oadev(values,
rate=r,
data_type="freq",
taus=tau1)
fig.subplots_adjust(hspace=0.8, wspace=0.8) #height , width
plt.subplot(3, 4, i + 1)
plt.loglog(tau2, ad, 'b')
plt.title("Rate " + str(r))
plt.xlabel('tau')
plt.ylabel('ADEV [V]')
plt.grid()
# Calculating the characteristics
characteristics_calculation(tau2, ad, ade, adn, r)
plt.show()
if save == 'yes':
plot_path = save_path + ".png"
fig.savefig(plot_path)
plt.loglog(f_fi[1:], [h0 * v0 * v0 / (ff * ff) for ff in f_fi[1:]], label="h_0 * v0^2 * f^-2")
plt.legend(framealpha=0.5)
plt.title("PSD of phase (radians)")
plt.xlabel("Frequeny / Hz")
plt.ylabel("one-sided PSD / S_fi(f)")
plt.figure()
plt.loglog(f_x, psd_x, label="numpy.fft()")
plt.loglog(f_x2, psd_x2, label="scipy.signal.welch()")
plt.loglog(f_x[1:], [h0 / (2 * math.pi * ff) ** 2 for ff in f_x[1:]], label="h0/(2*pi*f)**2")
plt.legend(framealpha=0.5)
plt.title("PSD of phase (time)")
plt.xlabel("Frequeny / Hz")
plt.ylabel("one-sided PSD / S_x(f)")
plt.figure()
taus = [tt for tt in np.logspace(-2.2, 4, 100)]
(taus_y, devs_y, errs_y, ns_y) = allantools.oadev(y, data_type="freq", rate=fs, taus=taus)
(taus_x, devs_x, errs_x, ns_x) = allantools.oadev(x, data_type="phase", rate=fs, taus=taus)
plt.loglog(taus_y, devs_y, "o", label="ADEV from y")
plt.loglog(taus_x, devs_x, "*", label="ADEV from x")
adev_y = [math.sqrt(0.5 * h0 * (1.0 / tt)) for tt in taus]
plt.loglog(taus, adev_y, label="sqrt( 0.5*h0*tau^-1 )")
plt.xlim((8e-3, 1e3))
plt.legend(framealpha=0.6)
plt.title("Allan deviation")
plt.xlabel("Tau / s")
plt.ylabel("Allan deviation")
plt.show()
out=[]
for f in flist:
out.append( f/float(f0) - 1.0 )
return out
fname = "gps_1pps_phase_data.txt.gz"
phase = read_datafile(fname,column=0)
print len(phase), " values read: ", len(phase)/3600.0 , " hours"
#f = to_fractional(f10MHz, 10e6 ) # convert to fractional frequency
my_taus = numpy.logspace(1,6,60) # log-spaced tau values from 10s and upwards
rate = 1/float(1.0)
(adev_taus,adev_devs,adev_errs,ns) = allan.adev(phase=phase, rate=rate, taus=my_taus)
(oadev_taus,oadev_devs,oadev_errs,ns) = allan.oadev(phase=phase, rate=rate, taus=my_taus)
(hdev_taus,hdev_devs,hdev_errs,ns) = allan.hdev(phase=phase, rate=rate, taus=my_taus)
(ohdev_taus,ohdev_devs,ohdev_errs,ns) = allan.ohdev(phase=phase, rate=rate, taus=my_taus)
(mdev_taus,mdev_devs,mdev_errs,ns) = allan.mdev(phase=phase, rate=rate, taus=my_taus)
(totdev_taus,totdev_devs,totdev_errs,ns) = allan.totdev(phase=phase, rate=rate, taus=my_taus)
(tie_taus,tie_devs,tie_errs,ns) = allan.tierms(phase=phase, rate=rate, taus=my_taus)
#(mtie_taus,mtie_devs,mtie_errs,ns) = allan.mtie(phase=phase, rate=rate, taus=my_taus)
(tdev_taus,tdev_devs,tdev_errs,ns) = allan.tdev(phase=phase, rate=rate, taus=my_taus)
(tdev2_taus,tdev2_devs,tdev2_errs,ns2) = allan.tdev(frequency=allan.phase2frequency(phase,1.0), rate=rate, taus=my_taus)
plt.subplot(111, xscale="log", yscale="log")
def nbs14_test():
taus = [1, 2]
devs = []
tol = 1e-4
# first tests that call the _phase functions
print "nbs14 tests for phase data:"
(taus2,adevs2,aerrs2,ns2) = allan.adev_phase( nbs14_phase, 1.0, taus)
adevs = nbs14_devs[0]
assert( check_devs( adevs2[0], adevs[0] ) )
assert( check_devs( adevs2[1], adevs[1] ) )
print "nbs14 adev OK"
(taus2,adevs2,aerrs2,ns2) = allan.oadev_phase( nbs14_phase, 1.0, taus)
oadevs = nbs14_devs[1]
assert( check_devs( adevs2[0], oadevs[0] ) )
assert( check_devs( adevs2[1], oadevs[1] ) )
print "nbs14 oadev OK"
(taus2,adevs2,aerrs2,ns2) = allan.mdev_phase( nbs14_phase, 1.0, taus)
mdevs = nbs14_devs[2]
assert( check_devs( adevs2[0], mdevs[0] ) )
assert( check_devs( adevs2[1], mdevs[1] ) )
print "nbs14 mdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.totdev_phase( nbs14_phase, 1.0, taus)
totdevs = nbs14_devs[3]
assert( check_devs( adevs2[0], totdevs[0] ) )
assert( check_devs( adevs2[1], totdevs[1] ) )
print "nbs14 totdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.hdev_phase( nbs14_phase, 1.0, taus)
hdevs = nbs14_devs[4]
assert( check_devs( adevs2[0], hdevs[0] ) )
assert( check_devs( adevs2[1], hdevs[1] ) )
print "nbs14 hdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.tdev_phase( nbs14_phase, 1.0, taus)
tdevs = nbs14_devs[5]
assert( check_devs( adevs2[0], tdevs[0] ) )
assert( check_devs( adevs2[1], tdevs[1] ) )
print "nbs14 tdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.ohdev_phase( nbs14_phase, 1.0, taus)
ohdevs = nbs14_devs[6]
assert( check_devs( adevs2[0], ohdevs[0] ) )
assert( check_devs( adevs2[1], ohdevs[1] ) )
print "nbs14 ohdev OK"
# then the same tests for frequency data
print "nbs14 tests for frequency data:"
f_fract = [ float(f) for f in nbs14_f]
(taus2,adevs2,aerrs2,ns2) = allan.adev( f_fract, 1.0, taus)
adevs = nbs14_devs[0]
assert( check_devs( adevs2[0], adevs[0] ) )
assert( check_devs( adevs2[1], adevs[1] ) )
print "nbs14 freqdata adev OK"
(taus2,adevs2,aerrs2,ns2) = allan.oadev( f_fract, 1.0, taus)
oadevs = nbs14_devs[1]
assert( check_devs( adevs2[0], oadevs[0] ) )
assert( check_devs( adevs2[1], oadevs[1] ) )
print "nbs14 freqdata oadev OK"
(taus2,adevs2,aerrs2,ns2) = allan.mdev( f_fract, 1.0, taus)
mdevs = nbs14_devs[2]
assert( check_devs( adevs2[0], mdevs[0] ) )
assert( check_devs( adevs2[1], mdevs[1] ) )
print "nbs14 freqdata mdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.totdev( f_fract, 1.0, taus)
totdevs = nbs14_devs[3]
assert( check_devs( adevs2[0], totdevs[0] ) )
assert( check_devs( adevs2[1], totdevs[1] ) )
print "nbs14 freqdata totdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.hdev( f_fract, 1.0, taus)
hdevs = nbs14_devs[4]
assert( check_devs( adevs2[0], hdevs[0] ) )
assert( check_devs( adevs2[1], hdevs[1] ) )
print "nbs14 freqdata hdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.tdev( f_fract, 1.0, taus)
tdevs = nbs14_devs[5]
assert( check_devs( adevs2[0], tdevs[0] ) )
assert( check_devs( adevs2[1], tdevs[1] ) )
print "nbs14 freqdata tdev OK"
(taus2,adevs2,aerrs2,ns2) = allan.ohdev( f_fract, 1.0, taus)
ohdevs = nbs14_devs[6]
assert( check_devs( adevs2[0], ohdevs[0] ) )
assert( check_devs( adevs2[1], ohdevs[1] ) )
print "nbs14 freqdata ohdev OK"
print "nbs14 all test OK"
def to_fractional(flist, f0):
out = []
for f in flist:
out.append(f/float(f0) - 1.0)
return out
fname = "ocxo_frequency.txt"
f10MHz = read_datafile(fname, column=0)
f = to_fractional(f10MHz, 10e6) # convert to fractional frequency
# log-spaced tau values from 10s and upwards
my_taus = numpy.logspace(1, 5, 40)
rate = 1/float(10) # data collected with 10s gate time
(oadev_taus, oadev_devs, oadev_errs, ns) = allan.oadev(
f, data_type='freq', rate=rate, taus=my_taus)
(mdev_taus, mdev_devs, mdev_errs, ns) = allan.mdev(
f, data_type='freq', rate=rate, taus=my_taus)
(hdev_taus, hdev_devs, hdev_errs, ns) = allan.hdev(
f, data_type='freq', rate=rate, taus=my_taus)
(ohdev_taus, ohdev_devs, ohdev_errs, ns) = allan.ohdev(
f, data_type='freq', rate=rate, taus=my_taus)
plt.subplot(111, xscale="log", yscale="log")
plt.errorbar(oadev_taus, oadev_devs, yerr=oadev_errs, label='OADEV')
plt.errorbar(mdev_taus, mdev_devs, yerr=mdev_errs, label='MDEV')
plt.errorbar(hdev_taus, hdev_devs, yerr=hdev_errs, label='HDEV')
plt.errorbar(ohdev_taus, ohdev_devs, yerr=ohdev_errs, label='OHDEV')
plt.xlabel('Taus (s)')
def adev(self):
tau, dev, _, _ = oadev(self.rotation, rate=self.rate, data_type='freq')
return tau, dev
def main(argv):
vartypes = ['avar', 'oavar']
tau_axis = None
for arg in argv:
key = arg.split('=')[0]
value = arg.split('=')[1]
if key == 'taus':
tau_axis = value
elif key == 'var':
toKeep = value
toRemove = []
for i in range(0, len(vartypes)):
if vartypes[i] != toKeep:
toRemove.append(i)
for idx in toRemove[::-1]:
del vartypes[idx]
# allantools options
if not (tau_axis in [None, 'all', 'octave', 'decade']):
print('./plot.py taus=[all,octave,decade]')
return 0
# read input
x = readcsv("input.csv")
#x = allantools.noise.brown(16384, b2=1.0)
#x = allantools.noise.white(16384*100) # b2=1.0)
# gui
fig = plt.figure()
ax1 = plt.subplot(111)
plt.grid(True)
for vartype in vartypes:
print("VAR: {:s}".format(vartype))
for dtype in ['phase', 'freq']:
# generate model
if vartype == 'avar':
if tau_axis == 'all':
(taus, ym, errors, ns) = allantools.adev(x,
data_type=dtype,
taus='all')
elif tau_axis == 'decade':
(taus, ym, errors, ns) = allantools.adev(x,
data_type=dtype,
taus='decade')
elif tau_axis == 'octave':
(taus, ym, errors, ns) = allantools.adev(x,
data_type=dtype,
taus='octave')
else:
(taus, ym, errors, ns) = allantools.adev(x,
data_type=dtype)
elif vartype == 'oavar':
if tau_axis == 'all':
(taus, ym, errors, ns) = allantools.oadev(x,
data_type=dtype,
taus='all')
elif tau_axis == 'decade':
(taus, ym, errors, ns) = allantools.oadev(x,
data_type=dtype,
taus='decade')
elif tau_axis == 'octave':
(taus, ym, errors, ns) = allantools.oadev(x,
data_type=dtype,
taus='octave')
else:
(taus, ym, errors, ns) = allantools.oadev(x,
data_type=dtype)
ym = np.power(ym, 2) # adev compared to avar
ym = 20 * np.log10(ym)
ax1.semilogx(taus,
ym,
'+-',
label="{:s} '{:s}' model".format(vartype, dtype))
# output
y = readcsv("{:s}-{:s}.csv".format(vartype, dtype))
y = 20 * np.log10(y)
if tau_axis == None:
taus = powers_of_two_axis(len(y))
elif tau_axis == 'decade':
taus = powers_of_ten_axis(len(y))
else:
taus = np.linspace(0, len(y), len(y), dtype='int')
ax1.semilogx(taus,
y,
'+',
label="{:s} '{:s}'".format(vartype, dtype))
ax1.legend(loc='best')
# tb
error = 0
max_tol = 0.01
print("-------- {:s} test -------".format(dtype))
for i in range(0, min(len(ym), len(y))):
e = abs(y[i] - ym[i])
if e > max_tol:
error += 1
print("tau: {:d} error: {:.3e} dB".format(taus[i], e))
if error > 0:
print("failed")
else:
print("passed")
print("---------------------")
plt.show()
def drift(self):
tau, dev, _, _ = oadev(self.rotation, rate=self.rate, data_type='freq')
return min(dev)
def vank(objname, weightthresh=10.0,chithresh=0.0, sigmaithresh=0.0):
c = 299792458.0
if socket.gethostname() == 'Main':
filenames = glob.glob('/Data/kiwispec-proc/n20160[5,6]*/*' + objname + '*.chrec' + exten + '.npy')
else:
filenames = glob.glob('/n/home12/jeastman/minerva/data/n2016051[4-9]/*' + objname + '*.chrec' + exten + '.npy') +\
glob.glob('/n/home12/jeastman/minerva/data/n201605[2-3]?/*' + objname + '*.chrec' + exten + '.npy') +\
glob.glob('/n/home12/jeastman/minerva/data/n2016060?/*' + objname + '*.chrec' + exten + '.npy') +\
glob.glob('/n/home12/jeastman/minerva/data/n2016061[0-2]/*' + objname + '*.chrec' + exten + '.npy')
ntel = 4
nobs = len(filenames)*ntel
if nobs <= 3: return
vji = []
wji = []
chiji = []
ctsji = []
alphaji = []
ipsigmaji = []
slopeji = []
jdutcs = np.array([])
telescope = np.array([])
for filename in filenames:
st = os.stat(filename)
# if (st.st_size != 121120): continue
if (st.st_size == 0.0):
nobs -= 4
continue
chrec = np.load(filename)
fitsname = os.path.splitext(os.path.splitext(filename)[0])[0] + '.fits'
h = pyfits.open(fitsname,mode='update')
# # reject chunks with bad DSST (Step 1)
# bad = np.where(chrec['wt' + str(i)] == 0)
# chrec['z' + str(i)][bad] = np.nan
# reject chunks where fitter failed to converge (rchi2 == 0 or 100)
# (Step 2)
for i in range(1,ntel+1):
if h[0].header['FLUXMID' + str(i)] == 'UNKNOWN':
t = Time(h[0].header['DATE-OBS'], format='isot',scale='utc')
midflux = t.jd+h[0].header['EXPTIME']/2.0/86400.0
h[0].header['FLUXMID' + str(i)] = midflux
else:
midflux = h[0].header['FLUXMID' + str(i)]
# very noisy data can find great fits
bad = np.where(chrec['chi' + str(i)] <= 0.3)
chrec['z' + str(i)][bad] = np.nan
# bad fits will have bad chi^2
bad = np.where(chrec['chi' + str(i)] >= 5)
chrec['z' + str(i)][bad] = np.nan
# very noisy data can find great fits
bad = np.where((chrec['alpha' + str(i)] <= -0.95) | (chrec['alpha' + str(i)] >= 0.95))
chrec['z' + str(i)][bad] = np.nan
# very noisy data can find great fits
bad = np.where((chrec['sigma' + str(i)] <= 0.25) | (chrec['sigma' + str(i)] >= 1.25))
chrec['z' + str(i)][bad] = np.nan
# reject chunks with the lowest DSST weight (Step 3) or bad weights (Step 1)
lowweight = np.nanpercentile(chrec['wt' + str(i)],weightthresh)
bad = np.where((chrec['wt' + str(i)] <= lowweight) | (chrec['wt' + str(i)] == 0))
chrec['z' + str(i)][bad] = np.nan
if 'BARYCOR' + str(i) not in h[0].header.keys():
# do the barycentric correction
ra = h[0].header['TARGRA' + str(i)]
dec = h[0].header['TARGDEC' + str(i)]
try: pmra = h[0].header['PMRA' + str(i)]
except: pmra = 0.0
try: pmdec = h[0].header['PMDEC' + str(i)]
except: pmdec = 0.0
try: parallax = h[0].header['PARLAX' + str(i)]
except: parallax = 0.0
try: rv = h[0].header['RV' + str(i)]
except: rv = 0.0
result = utils.barycorr(midflux, ra, dec, pmra=pmra, pmdec=pmdec, parallax=parallax, rv=rv, zmeas=0.0)
zb = result/c
h[0].header['BARYCOR' + str(i)] = zb
elif h[0].header['BARYCOR' + str(i)] == 'UNKNOWN':
# do the barycentric correction
ra = h[0].header['TARGRA' + str(i)]
dec = h[0].header['TARGDEC' + str(i)]
try: pmra = h[0].header['PMRA' + str(i)]
except: pmra = 0.0
try: pmdec = h[0].header['PMDEC' + str(i)]
except: pmdec = 0.0
try: parallax = h[0].header['PARLAX' + str(i)]
except: parallax = 0.0
try: rv = h[0].header['RV' + str(i)]
except: rv = 0.0
result = utils.barycorr(midflux, ra, dec, pmra=pmra, pmdec=pmdec, parallax=parallax, rv=rv, zmeas=0.0)
zb = result/c
h[0].header['BARYCOR' + str(i)] = zb
else:
zb = h[0].header['BARYCOR' + str(i)]
# active = np.append(active ,h[0].header['FAUSTAT' + str(i)] == 'GUIDING')
if h[0].header['FAUSTAT' + str(i)] == 'GUIDING' or 'daytimeSky' in h[0].header['OBJECT' + str(i)]:
rvs = c*((1.0+chrec['z' + str(i)])*(1.0+zb)-1.0)
# if all chunks are bad, skip it
if len(np.where(np.isnan(rvs))[0]) == len(rvs):
nobs-=1
continue
jdutcs = np.append(jdutcs,midflux)
telescope = np.append(telescope,i)
if len(vji) == 0: vji = rvs
else: vji = np.vstack((vji,rvs))
if len(wji) == 0: wji = chrec['wt' + str(i)]
else: wji = np.vstack((wji,chrec['wt' + str(i)]))
if len(chiji) == 0: chiji = chrec['chi' + str(i)]
else: chiji = np.vstack((chiji,chrec['chi' + str(i)]))
if len(ctsji) == 0: ctsji = chrec['cts' + str(i)]
else: ctsji = np.vstack((ctsji,chrec['cts' + str(i)]))
if len(alphaji) == 0: alphaji = chrec['alpha' + str(i)]
else: alphaji = np.vstack((alphaji,chrec['alpha' + str(i)]))
if len(ipsigmaji) == 0: ipsigmaji = chrec['sigma' + str(i)]
else: ipsigmaji = np.vstack((ipsigmaji,chrec['sigma' + str(i)]))
if len(slopeji) == 0: slopeji = chrec['slope' + str(i)]
else: slopeji = np.vstack((slopeji,chrec['slope' + str(i)]))
else: nobs-=1
h.flush()
h.close()
nchunks = len(chrec)
vij = np.transpose(vji)
wij = np.transpose(wji)
chiij = np.transpose(chiji)
ctsij = np.transpose(ctsji)
alphaij = np.transpose(alphaji)
ipsigmaij = np.transpose(ipsigmaji)
slopeij = np.transpose(slopeji)
snr = np.sqrt(np.nansum(ctsij,axis=0))
# reject chunks with the worst fits (step 4)
chimed = np.nanmedian(chiij,axis=1)
hichi = np.nanpercentile(chimed,100.0-chithresh)
bad = np.where(chimed >= hichi)
vij[bad,:] = np.nan
# adjust all chunks to have the same RV zero points (step 5)
# subtract the mean velocity of all observations from each chunk
vij -= np.transpose(np.tile(np.nanmean(vij,axis=1),(nobs,1)))
# compute chunk weights (step 6):
# median velocities for each observation
vjmed = np.nanmedian(vij,axis=0) # eq 2.9
# compute the matrix of velocity differences
Deltaij = vij - np.tile(vjmed,(nchunks,1)) # eq 2.8
sigmai = np.nanstd(Deltaij,axis=1) # eq 2.6
# reject the highest sigmai (step 7)
bad = np.where(sigmai == 0.0)
sigmai[bad] = np.inf
hisigma = np.nanpercentile(sigmai,100.0-sigmaithresh)
bad = np.where(sigmai >= hisigma)
sigmai[bad] = np.inf
# compute rj (eq 2.7)
rj = np.nanmedian(np.abs(Deltaij)*np.transpose(np.tile(sigmai,(nobs,1))),axis=0) # eq 2.7
sigmaij = np.transpose(np.tile(sigmai,(nobs,1)))*np.tile(rj,(nchunks,1)) # eq 2.5
# prevent nans
bad = np.where(sigmaij == 0.0)
sigmaij[bad] = np.inf
wij = 1.0/sigmaij**2
vj = np.nansum(vij*wij,axis=0)/np.nansum(wij,axis=0) # eq 2.10
sigmavj = 1.0/np.sqrt(np.nansum(wij,axis=0)) # eq 2.12
print objname + " RMS: " + str(np.nanstd(vj))
# plot scatter vs time
colors = ['','r','g','b','orange']
for i in range(1,ntel+1):
match = np.where(telescope == i)
plt.plot(jdutcs[match]-2457389,vj[match],'o',color=colors[i])
plt.title(objname)
plt.xlabel('Days since UT 2016-01-01')
plt.ylabel('RV (m/s)')
plt.savefig(objname + '.' + exten + '.png')
plt.close()
nightlyrvs = {'all':np.array([]),
'T1':np.array([]),
'T2':np.array([]),
'T3':np.array([]),
'T4':np.array([])}
# bin per telescope per night
for jd in range(2457480,2457571):
for i in range(1,ntel+1):
match = np.where((jdutcs >= jd) & (jdutcs < (jd+1)) & np.isfinite(vj) & (telescope==i))
night = np.mean(jdutcs[match])-2457389
nightlyrv = np.sum(vj[match]/sigmavj[match]**2)/np.sum(1.0/sigmavj[match]**2)
plt.plot([night],[nightlyrv],'o',color=colors[i])
nightlyrvs['T' + str(i)] = np.append(nightlyrvs['T' + str(i)],nightlyrv)
match = np.where((jdutcs >= jd) & (jdutcs < (jd+1)) & np.isfinite(vj))
night = np.mean(jdutcs[match])-2457389
nightlyrv = np.sum(vj[match]/sigmavj[match]**2)/np.sum(1.0/sigmavj[match]**2)
plt.plot([night],[nightlyrv],'ko')
nightlyrvs['all'] = np.append(nightlyrvs['all'],nightlyrv)
plt.title(objname)
plt.xlabel('Days since UT 2016-01-01')
plt.ylabel('Nightly binned RV (m/s)')
plt.savefig(objname + '.' + exten + '.binned.png')
plt.close()
print objname + ' nightly binned RMS:'
print "All: " + str(np.nanstd(nightlyrvs['all']))
print "T1: " + str(np.nanstd(nightlyrvs['T1']))
print "T2: " + str(np.nanstd(nightlyrvs['T2']))
print "T3: " + str(np.nanstd(nightlyrvs['T3']))
print "T4: " + str(np.nanstd(nightlyrvs['T4']))
# create a histogram of scatter within a night
mindate = int(math.floor(min(jdutcs)))
maxdate = int(math.ceil(max(jdutcs)))
jdbin = []
rvbin = []
errbin = []
sigmabin = []
for i in range(mindate,maxdate):
match = np.where((jdutcs >= i) & (jdutcs < (i+1)) & (np.isfinite(vj)))
if len(match[0]) > 1:
jdbin.append(mindate)
rvbin.append(np.mean(vj[match]))
errbin.append(np.std(vj[match]))
sigmabin.append(np.mean(sigmavj[match]))
# print jdbin, rvbin, errbin, sigmabin
hist, bins = np.histogram(errbin, bins=20)
width = 0.7 * (bins[1] - bins[0])
center = (bins[:-1] + bins[1:]) / 2
plt.bar(center, hist, align='center', width=width)
plt.xlabel('Intranight RMS (m/s)')
plt.ylabel('# of Nights')
plt.savefig(objname + '.' + exten + '.hist.png')
plt.close()
# plot median sigma as a function of chunk number
for tel in range(1,ntel+1):
match = np.where(telescope == tel)
for chunk in range(np.shape(ipsigmaij)[0]):
plt.plot(chunk,np.nanmedian(ipsigmaij[chunk,match]),'o',color=colors[tel])
sigmawav.append()
sigmaall.append(np.nanmedian(ipsigmaij[chunk,match]))
sigmatel.append(tel)
plt.title(objname)
plt.xlabel('Chunk number')
plt.ylabel('sigma')
plt.savefig(objname + '.' + exten + '.sigmavchunk.png')
plt.close()
# plot median alpha as a function of chunk number
for tel in range(1,ntel+1):
match = np.where(telescope == tel)
for chunk in range(np.shape(alphaij)[0]):
plt.plot(chunk,np.nanmedian(alphaij[chunk,match]),'o',color=colors[tel])
plt.title(objname)
plt.xlabel('Chunk number')
plt.ylabel('alpha')
plt.savefig(objname + '.' + exten + '.alphavchunk.png')
plt.close()
# plot median slope as a function of chunk number
for tel in range(1,ntel+1):
match = np.where(telescope == tel)
for chunk in range(np.shape(slopeij)[0]):
plt.plot(chunk,np.nanmedian(slopeij[chunk,match]),'o',color=colors[tel])
plt.title(objname)
plt.xlabel('Chunk number')
plt.ylabel('slope')
plt.savefig(objname + '.' + exten + '.slopevchunk.png')
plt.close()
# plot sigma for chunk 150 over time
chunk = 150
for t in jdutcs:
for tel in range(1,ntel+1):
match = np.where((telescope == tel) & (jdutcs == t))
plt.plot(t-2457389,np.nanmedian(ipsigmaij[chunk,match]),'o',color=colors[tel])
plt.title(objname)
plt.xlabel('Days since UT 2016-01-01')
plt.ylabel('sigma')
plt.savefig(objname + '.' + exten + '.sigma150vtime.png')
plt.close()
# plot scatter within a night vs SNR
plt.plot(sigmabin, errbin,'bo')
plt.title(objname)
plt.xlabel('Sigma_v_j')
plt.ylabel('Intranight RMS (m/s)')
plt.savefig(objname + '.' + exten + '.SNR.png')
plt.close()
# subtract a linear trend from vj
good = np.where(np.isfinite(vj))
t0 = np.nanmedian(jdutcs[good])
coeffs = np.polyfit(jdutcs[good]-t0,vj[good],1)
vjtrend = coeffs[1] + (jdutcs[good]-t0)*coeffs[0]
# output text files of JD, rv, rverr
f = open(objname + '.dat','w')
for i in range(len(vj)):
if np.isfinite(vj[i]):
f.write('{0:f} {1:f} {2:f} T{3:d} {4:s}'.format(jdutcs[i],vj[i],sigmavj[i],int(telescope[i]),'\n'))
f.close()
# plot scatter minus trend vs time
plt.plot(jdutcs[good]-2457389,vj[good]-vjtrend,'bo')
plt.title(objname)
plt.xlabel('Days since UT 2016-01-01')
plt.ylabel('RV (m/s)')
plt.savefig(objname + '.' + exten + '.detrended.png')
plt.close()
print objname + " Detrended RMS: " + str(np.nanstd(vj[good]-vjtrend))
for i in range(1,ntel+1):
match = np.where((telescope == i) & np.isfinite(vj))
# rate = 1.0/(jdutcs[match]-np.roll(jdutcs[match],1))
# taus = np.logspace(0, 3, 50)
taus = np.arange(1,len(vj[match])/3)
(t2, ad, ade, adn) = allantools.oadev(vj[match], rate=1, data_type="freq", taus=taus)
plt.loglog(t2,ad,'o',color=colors[i])
# overplot the line
y = np.asarray([pow(tt,-0.5) for tt in taus])*ad[0]
plt.loglog(taus,y,'-',color=colors[i])
match = np.where(np.isfinite(vj))
taus = np.arange(1,len(vj[match])/3)
(t2, ad, ade, adn) = allantools.oadev(vj[match], rate=1, data_type="freq", taus=taus)
plt.loglog(t2,ad,'ko')
# overplot the line
y = np.asarray([pow(tt,-0.5) for tt in taus])*ad[0]
plt.loglog(taus,y,'k-')
plt.xlabel('N bin')
plt.ylabel('Precision (m/s)')
plt.savefig(objname + '.' + exten + '.allan.png')
plt.close()
def main():
nr = 2**14 # number of datapoints in time-series
tau0=1.0 # sampling interval, sets nyquist frequency to 0.5/tau0
adev0 = 1.0e-11
#
# for each noise type generated with cn.noiseGen()
# compute
# - phase PSD and coefficient g_b
# - frequency PSD and coefficient h_a
# - ADEV
# - MDEV
#
qd0 = qd1 = qd2 = qd3 = qd4 = pow(adev0, 2) # discrete variance for noiseGen()
tau0 = 1.0 # sample interval
sample_rate = 1/tau0
x0 = cn.noiseGen(nr, qd0, 0) # white phase noise (WPM)
x1 = cn.noiseGen(nr, qd1, -1) # flicker phase noise (FPM)
x2 = cn.noiseGen(nr, qd2, -2) # white frequency noise (WFM)
x3 = cn.noiseGen(nr, qd3 , -3) # flicker frequency noise (FFM)
x4 = cn.noiseGen(nr, qd4 , -4) # random walk frequency noise (RWFM)
# compute frequency time-series
y0 = at.phase2frequency(x0, sample_rate)
y1 = at.phase2frequency(x1, sample_rate)
y2 = at.phase2frequency(x2, sample_rate)
y3 = at.phase2frequency(x3, sample_rate)
y4 = at.phase2frequency(x4, sample_rate)
# compute phase PSD
(f0, psd0) = at.noise.scipy_psd(x0, fs=sample_rate, nr_segments=4)
(f1, psd1) = at.noise.scipy_psd(x1, fs=sample_rate, nr_segments=4)
(f2, psd2) = at.noise.scipy_psd(x2, fs=sample_rate, nr_segments=4)
(f3, psd3) = at.noise.scipy_psd(x3, fs=sample_rate, nr_segments=4)
(f4, psd4) = at.noise.scipy_psd(x4, fs=sample_rate, nr_segments=4)
# compute phase PSD prefactor g_b
g0 = cn.phase_psd_from_qd( qd0, 0, tau0 )
g1 = cn.phase_psd_from_qd( qd0, -1, tau0 )
g2 = cn.phase_psd_from_qd( qd0, -2, tau0 )
g3 = cn.phase_psd_from_qd( qd0, -3, tau0 )
g4 = cn.phase_psd_from_qd( qd0, -4, tau0 )
# compute frequency PSD
(ff0, fpsd0) = at.noise.scipy_psd(y0, fs=sample_rate, nr_segments=4)
(ff1, fpsd1) = at.noise.scipy_psd(y1, fs=sample_rate, nr_segments=4)
(ff2, fpsd2) = at.noise.scipy_psd(y2, fs=sample_rate, nr_segments=4)
(ff3, fpsd3) = at.noise.scipy_psd(y3, fs=sample_rate, nr_segments=4)
(ff4, fpsd4) = at.noise.scipy_psd(y4, fs=sample_rate, nr_segments=4)
# compute frequency PSD prefactor h_a
a0 = cn.frequency_psd_from_qd( qd0, 0, tau0 )
a1 = cn.frequency_psd_from_qd( qd1, -1, tau0 )
a2 = cn.frequency_psd_from_qd( qd2, -2, tau0 )
a3 = cn.frequency_psd_from_qd( qd3, -3, tau0 )
a4 = cn.frequency_psd_from_qd( qd4, -4, tau0 )
# compute ADEV
(t0,d0,e,n) = at.oadev(x0, rate=sample_rate)
(t1,d1,e,n) = at.oadev(x1, rate=sample_rate)
(t2,d2,e,n) = at.oadev(x2, rate=sample_rate)
(t3,d3,e,n) = at.oadev(x3, rate=sample_rate)
(t4,d4,e,n) = at.oadev(x4, rate=sample_rate)
# compute MDEV
(mt0,md0,e,n) = at.mdev(x0, rate=sample_rate)
(mt1,md1,e,n) = at.mdev(x1, rate=sample_rate)
(mt2,md2,e,n) = at.mdev(x2, rate=sample_rate)
(mt3,md3,e,n) = at.mdev(x3, rate=sample_rate)
(mt4,md4,e,n) = at.mdev(x4, rate=sample_rate)
plt.figure()
# Phase PSD figure
plt.subplot(2,2,1)
plt.loglog(f0,[ g0*pow(xx, 0.0) for xx in f0],'--',label=r'$g_0f^0$', color='black')
plt.loglog(f0,psd0,'.', color='black')
plt.loglog(f1[1:],[ g1*pow(xx, -1.0) for xx in f1[1:]],'--',label=r'$g_{-1}f^{-1}$', color='red')
plt.loglog(f1,psd1,'.' ,color='red')
plt.loglog(f2[1:],[ g2*pow(xx,-2.0) for xx in f2[1:]],'--',label=r'$g_{-2}f^{-2}$', color='green')
plt.loglog(f2,psd2,'.', color='green')
plt.loglog(f3[1:], [ g3*pow(xx,-3.0) for xx in f3[1:]],'--',label=r'$g_{-3}f^{-3}$', color='pink')
plt.loglog(f3, psd3, '.', color='pink')
plt.loglog(f4[1:], [ g4*pow(xx,-4.0) for xx in f4[1:]],'--',label=r'$g_{-4}f^{-4}$', color='blue')
plt.loglog(f4, psd4, '.', color='blue')
plt.grid()
plt.legend(framealpha=0.9)
plt.title(r'Phase Power Spectral Density')
plt.xlabel(r'Frequency (Hz)')
plt.ylabel(r' $S_x(f)$ $(s^2/Hz)$')
# frequency PSD figure
plt.subplot(2,2,2)
plt.loglog(ff0[1:], [ a0*pow(xx,2) for xx in ff0[1:]],'--',label=r'$h_{2}f^{2}$', color='black')
plt.loglog(ff0,fpsd0,'.', color='black')
plt.loglog(ff1[1:], [ a1*pow(xx,1) for xx in ff1[1:]],'--',label=r'$h_{1}f^{1}$', color='red')
plt.loglog(ff1,fpsd1,'.', color='red')
plt.loglog(ff2[1:], [ a2*pow(xx,0) for xx in ff2[1:]],'--',label=r'$h_{0}f^{0}$', color='green')
plt.loglog(ff2,fpsd2,'.', color='green')
plt.loglog(ff3[1:], [ a3*pow(xx,-1) for xx in ff3[1:]],'--',label=r'$h_{-1}f^{-1}$', color='pink')
plt.loglog(ff3,fpsd3,'.', color='pink')
plt.loglog(ff4[1:], [ a4*pow(xx,-2) for xx in ff4[1:]],'--',label=r'$h_{-2}f^{-2}$', color='blue')
plt.loglog(ff4,fpsd4,'.', color='blue')
plt.grid()
plt.legend(framealpha=0.9)
plt.title(r'Frequency Power Spectral Density')
plt.xlabel(r'Frequency (Hz)')
plt.ylabel(r' $S_y(f)$ $(1/Hz)$')
# ADEV figure
plt.subplot(2,2,3)
plt.loglog(t0, [ cn.adev_from_qd(qd0, 0, tau0)/xx for xx in t0],'--', label=r'$\sqrt{Eh_{2}}\tau^{-1}$', color='black')
plt.loglog(t0,d0,'o', color='black')
plt.loglog(t1, [ cn.adev_from_qd(qd1, -1, tau0)/xx for xx in t1],'--', label=r'$\sqrt{Dh_{1}}\tau^{-1}$', color='red')
plt.loglog(t1,d1,'o', color='red')
plt.loglog(t2, [ cn.adev_from_qd(qd2, -2, tau0)/math.sqrt(xx) for xx in t2],'--', label=r'$\sqrt{Ch_{0}}\tau^{-1/2}$', color='green')
plt.loglog(t2,d2,'o', color='green')
plt.loglog(t3, [ cn.adev_from_qd(qd3, -3, tau0)*1 for xx in t3],'--', label=r'$\sqrt{Bh_{-1}}\tau^0$', color='pink')
plt.loglog(t3,d3,'o', color='pink')
plt.loglog(t4, [ cn.adev_from_qd(qd4, -4, tau0)*math.sqrt(xx) for xx in t4],'--', label=r'$\sqrt{Ah_{-2}}\tau^{+1/2}$', color='blue')
plt.loglog(t4,d4,'o', color='blue')
plt.legend(framealpha=0.9, loc='lower left')
plt.grid()
plt.title(r'Allan Deviation')
plt.xlabel(r'$\tau$ (s)')
plt.ylabel(r'ADEV')
# MDEV
plt.subplot(2, 2, 4)
plt.loglog(t0, [ cn.adev_from_qd(qd0, 0, tau0)/pow(xx,3.0/2.0) for xx in t0],'--', label=r'$\sqrt{Eh_{2}}\tau^{-3/2}$', color='black')
plt.loglog(mt0,md0,'o', color='black')
plt.loglog(t1, [ cn.adev_from_qd(qd1, -1, tau0)/xx for xx in t1],'--', label=r'$\sqrt{Dh_{1}}\tau^{-1}$', color='red')
plt.loglog(mt1,md1,'o', color='red')
plt.loglog(t2, [ cn.adev_from_qd(qd2, -2, tau0)/math.sqrt(xx) for xx in t2],'--', label=r'$\sqrt{Ch_{0}}\tau^{-1/2}$', color='green')
plt.loglog(mt2,md2,'o', color='green')
plt.loglog(t3, [ cn.adev_from_qd(qd3, -3, tau0)**1 for xx in t3],'--', label=r'$\sqrt{Bh_{-1}}\tau^0$', color='pink')
plt.loglog(mt3,md3,'o', color='pink')
plt.loglog(t4, [ cn.adev_from_qd(qd4, -4, tau0)*math.sqrt(xx) for xx in t4],'--', label=r'$\sqrt{Ah_{-2}}\tau^{+1/2}$', color='blue')
plt.loglog(mt4,md4,'o', color='blue')
plt.legend(framealpha=0.9, loc='lower left')
plt.grid()
plt.title(r'Modified Allan Deviation')
plt.xlabel(r'$\tau$ (s)')
plt.ylabel(r'MDEV')
plt.show()
def noise(self):
_, dev, _, _ = oadev(self.rotation, rate=self.rate, data_type='freq')
return dev[0] / 60
def main():
nr = 2**14 # number of datapoints in time-series
adev0 = 1.0e-11
#
# for each noise type generated Noise class
# compute
# - phase PSD and coefficient g_b
# - frequency PSD and coefficient h_a
# - ADEV
# - MDEV
#
# discrete variance for noiseGen()
qd0 = qd1 = qd2 = qd3 = qd4 = pow(adev0, 2)
tau0 = 1.0 # sample interval
sample_rate = 1.0 / tau0
x0 = at.Noise(nr, qd0, 0) # white phase noise (WPM)
x0.generateNoise()
x1 = at.Noise(nr, qd1, -1) # flicker phase noise (FPM)
x1.generateNoise()
x2 = at.Noise(nr, qd2, -2) # white frequency noise (WFM)
x2.generateNoise()
x3 = at.Noise(nr, qd3, -3) # flicker frequency noise (FFM)
x3.generateNoise()
x4 = at.Noise(nr, qd4, -4) # random walk frequency noise (RWFM)
x4.generateNoise()
# compute frequency time-series
y0 = at.phase2frequency(x0.time_series, sample_rate)
y1 = at.phase2frequency(x1.time_series, sample_rate)
y2 = at.phase2frequency(x2.time_series, sample_rate)
y3 = at.phase2frequency(x3.time_series, sample_rate)
y4 = at.phase2frequency(x4.time_series, sample_rate)
# compute phase PSD
(f0, psd0) = at.noise.scipy_psd(x0.time_series,
f_sample=sample_rate,
nr_segments=4)
(f1, psd1) = at.noise.scipy_psd(x1.time_series,
f_sample=sample_rate,
nr_segments=4)
(f2, psd2) = at.noise.scipy_psd(x2.time_series,
f_sample=sample_rate,
nr_segments=4)
(f3, psd3) = at.noise.scipy_psd(x3.time_series,
f_sample=sample_rate,
nr_segments=4)
(f4, psd4) = at.noise.scipy_psd(x4.time_series,
f_sample=sample_rate,
nr_segments=4)
# compute phase PSD prefactor g_b
g0 = x0.phase_psd_from_qd(tau0)
g1 = x1.phase_psd_from_qd(tau0)
g2 = x2.phase_psd_from_qd(tau0)
g3 = x3.phase_psd_from_qd(tau0)
g4 = x4.phase_psd_from_qd(tau0)
# compute frequency PSD
(ff0, fpsd0) = at.noise.scipy_psd(y0, f_sample=sample_rate, nr_segments=4)
(ff1, fpsd1) = at.noise.scipy_psd(y1, f_sample=sample_rate, nr_segments=4)
(ff2, fpsd2) = at.noise.scipy_psd(y2, f_sample=sample_rate, nr_segments=4)
(ff3, fpsd3) = at.noise.scipy_psd(y3, f_sample=sample_rate, nr_segments=4)
(ff4, fpsd4) = at.noise.scipy_psd(y4, f_sample=sample_rate, nr_segments=4)
# compute frequency PSD prefactor h_a
a0 = x0.frequency_psd_from_qd(tau0)
a1 = x1.frequency_psd_from_qd(tau0)
a2 = x2.frequency_psd_from_qd(tau0)
a3 = x3.frequency_psd_from_qd(tau0)
a4 = x4.frequency_psd_from_qd(tau0)
# compute ADEV
(t0, d0, e, n) = at.oadev(x0.time_series, rate=sample_rate)
(t1, d1, e, n) = at.oadev(x1.time_series, rate=sample_rate)
(t2, d2, e, n) = at.oadev(x2.time_series, rate=sample_rate)
(t3, d3, e, n) = at.oadev(x3.time_series, rate=sample_rate)
(t4, d4, e, n) = at.oadev(x4.time_series, rate=sample_rate)
# compute MDEV
(mt0, md0, e, n) = at.mdev(x0.time_series, rate=sample_rate)
(mt1, md1, e, n) = at.mdev(x1.time_series, rate=sample_rate)
(mt2, md2, e, n) = at.mdev(x2.time_series, rate=sample_rate)
(mt3, md3, e, n) = at.mdev(x3.time_series, rate=sample_rate)
(mt4, md4, e, n) = at.mdev(x4.time_series, rate=sample_rate)
plt.figure()
# Phase PSD figure
plt.subplot(2, 2, 1)
plt.loglog(f0, [g0 * pow(xx, 0.0) for xx in f0],
'--',
label=r'$g_0f^0$',
color='black')
plt.loglog(f0, psd0, '.', color='black')
plt.loglog(f1[1:], [g1 * pow(xx, -1.0) for xx in f1[1:]],
'--',
label=r'$g_{-1}f^{-1}$',
color='red')
plt.loglog(f1, psd1, '.', color='red')
plt.loglog(f2[1:], [g2 * pow(xx, -2.0) for xx in f2[1:]],
'--',
label=r'$g_{-2}f^{-2}$',
color='green')
plt.loglog(f2, psd2, '.', color='green')
plt.loglog(f3[1:], [g3 * pow(xx, -3.0) for xx in f3[1:]],
'--',
label=r'$g_{-3}f^{-3}$',
color='pink')
plt.loglog(f3, psd3, '.', color='pink')
plt.loglog(f4[1:], [g4 * pow(xx, -4.0) for xx in f4[1:]],
'--',
label=r'$g_{-4}f^{-4}$',
color='blue')
plt.loglog(f4, psd4, '.', color='blue')
plt.grid()
plt.legend(framealpha=0.9)
plt.title(r'Phase Power Spectral Density')
plt.xlabel(r'Frequency (Hz)')
plt.ylabel(r' $S_x(f)$ $(s^2/Hz)$')
# frequency PSD figure
plt.subplot(2, 2, 2)
plt.loglog(ff0[1:], [a0 * pow(xx, 2) for xx in ff0[1:]],
'--',
label=r'$h_{2}f^{2}$',
color='black')
plt.loglog(ff0, fpsd0, '.', color='black')
plt.loglog(ff1[1:], [a1 * pow(xx, 1) for xx in ff1[1:]],
'--',
label=r'$h_{1}f^{1}$',
color='red')
plt.loglog(ff1, fpsd1, '.', color='red')
plt.loglog(ff2[1:], [a2 * pow(xx, 0) for xx in ff2[1:]],
'--',
label=r'$h_{0}f^{0}$',
color='green')
plt.loglog(ff2, fpsd2, '.', color='green')
plt.loglog(ff3[1:], [a3 * pow(xx, -1) for xx in ff3[1:]],
'--',
label=r'$h_{-1}f^{-1}$',
color='pink')
plt.loglog(ff3, fpsd3, '.', color='pink')
plt.loglog(ff4[1:], [a4 * pow(xx, -2) for xx in ff4[1:]],
'--',
label=r'$h_{-2}f^{-2}$',
color='blue')
plt.loglog(ff4, fpsd4, '.', color='blue')
plt.grid()
plt.legend(framealpha=0.9)
plt.title(r'Frequency Power Spectral Density')
plt.xlabel(r'Frequency (Hz)')
plt.ylabel(r' $S_y(f)$ $(1/Hz)$')
# ADEV figure
plt.subplot(2, 2, 3)
plt.loglog(t0, [x0.adev_from_qd(tau0, xx) / xx for xx in t0],
'--',
label=r'$\propto\sqrt{h_{2}}\tau^{-1}$',
color='black')
plt.loglog(t0, d0, 'o', color='black')
plt.loglog(t1, [x1.adev_from_qd(tau0, xx) / xx for xx in t1],
'--',
label=r'$\propto\sqrt{h_{1}}\tau^{-1}$',
color='red')
plt.loglog(t1, d1, 'o', color='red')
plt.loglog(t2, [x2.adev_from_qd(tau0, xx) / math.sqrt(xx) for xx in t2],
'--',
label=r'$\propto\sqrt{h_{0}}\tau^{-1/2}$',
color='green')
plt.loglog(t2, d2, 'o', color='green')
plt.loglog(t3, [x3.adev_from_qd(tau0, xx) * 1 for xx in t3],
'--',
label=r'$\propto\sqrt{h_{-1}}\tau^0$',
color='pink')
plt.loglog(t3, d3, 'o', color='pink')
plt.loglog(t4, [x4.adev_from_qd(tau0, xx) * math.sqrt(xx) for xx in t4],
'--',
label=r'$\propto\sqrt{h_{-2}}\tau^{+1/2}$',
color='blue')
plt.loglog(t4, d4, 'o', color='blue')
plt.legend(framealpha=0.9, loc='lower left')
plt.grid()
plt.title(r'Allan Deviation')
plt.xlabel(r'$\tau$ (s)')
plt.ylabel(r'ADEV')
# MDEV
plt.subplot(2, 2, 4)
plt.loglog(t0,
[x0.mdev_from_qd(tau0, xx) / pow(xx, 3.0 / 2.0) for xx in t0],
'--',
label=r'$\propto\sqrt{h_{2}}\tau^{-3/2}$',
color='black')
plt.loglog(mt0, md0, 'o', color='black')
plt.loglog(t1, [x1.mdev_from_qd(tau0, xx) / xx for xx in t1],
'--',
label=r'$\propto\sqrt{h_{1}}\tau^{-1}$',
color='red')
plt.loglog(mt1, md1, 'o', color='red')
plt.loglog(t2, [x2.mdev_from_qd(tau0, xx) / math.sqrt(xx) for xx in t2],
'--',
label=r'$\propto\sqrt{h_{0}}\tau^{-1/2}$',
color='green')
plt.loglog(mt2, md2, 'o', color='green')
plt.loglog(t3, [x3.mdev_from_qd(tau0, xx)**1 for xx in t3],
'--',
label=r'$\propto\sqrt{h_{-1}}\tau^0$',
color='pink')
plt.loglog(mt3, md3, 'o', color='pink')
plt.loglog(t4, [x4.mdev_from_qd(tau0, xx) * math.sqrt(xx) for xx in t4],
'--',
label=r'$\propto\sqrt{h_{-2}}\tau^{+1/2}$',
color='blue')
plt.loglog(mt4, md4, 'o', color='blue')
plt.legend(framealpha=0.9, loc='lower left')
plt.grid()
plt.title(r'Modified Allan Deviation')
plt.xlabel(r'$\tau$ (s)')
plt.ylabel(r'MDEV')
plt.show()
return p
def to_fractional(flist,f0):
out=[]
for f in flist:
out.append( f/float(f0) - 1.0 )
return out
fname = "ocxo_frequency.txt"
f10MHz = read_datafile(fname,column=0)
f = to_fractional(f10MHz, 10e6 ) # convert to fractional frequency
my_taus = numpy.logspace(1,5,40) # log-spaced tau values from 10s and upwards
rate = 1/float(10) # data collected with 10s gate time
(oadev_taus,oadev_devs,oadev_errs,ns) = allan.oadev(frequency=f, rate=rate, taus=my_taus)
(mdev_taus,mdev_devs,mdev_errs,ns) = allan.mdev(frequency=f, rate=rate, taus=my_taus)
(hdev_taus,hdev_devs,hdev_errs,ns) = allan.hdev(frequency=f, rate=rate, taus=my_taus)
(ohdev_taus,ohdev_devs,ohdev_errs,ns) = allan.ohdev(frequency=f, rate=rate, taus=my_taus)
plt.subplot(111, xscale="log", yscale="log")
plt.errorbar(oadev_taus, oadev_devs, yerr=oadev_errs, label='OADEV')
plt.errorbar(mdev_taus, mdev_devs, yerr=mdev_errs, label='MDEV')
plt.errorbar(hdev_taus, hdev_devs, yerr=hdev_errs, label='HDEV')
plt.errorbar(ohdev_taus, ohdev_devs, yerr=ohdev_errs, label='OHDEV')
plt.xlabel('Taus (s)')
plt.ylabel('ADEV')
fname = "gps_1pps_phase_data.txt.gz"
phase = read_datafile(fname, column=0)
print len(phase), " values read: ", len(phase) / 3600.0, " hours"
#f = to_fractional(f10MHz, 10e6 ) # convert to fractional frequency
my_taus = numpy.logspace(1, 6,
60) # log-spaced tau values from 10s and upwards
rate = 1 / float(1.0)
(adev_taus, adev_devs, adev_errs, ns) = allan.adev(phase=phase,
rate=rate,
taus=my_taus)
(oadev_taus, oadev_devs, oadev_errs, ns) = allan.oadev(phase=phase,
rate=rate,
taus=my_taus)
(hdev_taus, hdev_devs, hdev_errs, ns) = allan.hdev(phase=phase,
rate=rate,
taus=my_taus)
(ohdev_taus, ohdev_devs, ohdev_errs, ns) = allan.ohdev(phase=phase,
rate=rate,
taus=my_taus)
(mdev_taus, mdev_devs, mdev_errs, ns) = allan.mdev(phase=phase,
rate=rate,
taus=my_taus)
(totdev_taus, totdev_devs, totdev_errs, ns) = allan.totdev(phase=phase,
rate=rate,
def main(args):
rospy.init_node('allan_variance_node')
t0 = rospy.get_time()
"""""" """""" ""
" Parameters "
"""""" """""" ""
bagfile = rospy.get_param('~bagfile_path', '~/data/static.bag')
topic = rospy.get_param('~imu_topic_name', '/imu')
axis = rospy.get_param('~axis', 0)
sampleRate = rospy.get_param('~sample_rate', 100)
isDeltaType = rospy.get_param('~delta_measurement', False)
numTau = rospy.get_param('~number_of_lags', 1000)
resultsPath = rospy.get_param('~results_directory_path', None)
"""""" """""" """""" """""" ""
" Results Directory Path "
"""""" """""" """""" """""" ""
if resultsPath is None:
paths = rospkg.get_ros_paths()
path = paths[1] # path to workspace's devel
idx = path.find("ws/")
workspacePath = path[0:(idx + 3)]
resultsPath = workspacePath + 'av_results/'
if not os.path.isdir(resultsPath):
os.mkdir(resultsPath)
print "\nResults will be save in the following directory: \n\n\t %s\n" % resultsPath
"""""" """""" """"""
" Form Tau Array "
"""""" """""" """"""
taus = [None] * numTau
cnt = 0
for i in np.linspace(
-2.0, 5.0,
num=numTau): # lags will span from 10^-2 to 10^5, log spaced
taus[cnt] = pow(10, i)
cnt = cnt + 1
"""""" """""" """""
" Parse Bagfile "
""" """""" """""" ""
bag = rosbag.Bag(bagfile)
N = bag.get_message_count(topic) # number of measurement samples
data = np.zeros((6, N)) # preallocate vector of measurements
if isDeltaType:
scale = sampleRate
else:
scale = 1.0
cnt = 0
for topic, msg, t in bag.read_messages(topics=[topic]):
data[0, cnt] = msg.linear_acceleration.x * scale
data[1, cnt] = msg.linear_acceleration.y * scale
data[2, cnt] = msg.linear_acceleration.z * scale
data[3, cnt] = msg.angular_velocity.x * scale
data[4, cnt] = msg.angular_velocity.y * scale
data[5, cnt] = msg.angular_velocity.z * scale
cnt = cnt + 1
bag.close()
print "[%0.2f seconds] Bagfile parsed\n" % (rospy.get_time() - t0)
"""""" """""" """"""
" Allan Variance "
"""""" """""" """"""
if axis is 0:
currentAxis = 1 # loop through all axes 1-6
else:
currentAxis = axis # just loop one time and break
while (currentAxis <= 6):
(taus_used, adev, adev_err,
adev_n) = allantools.oadev(data[currentAxis - 1],
data_type='freq',
rate=float(sampleRate),
taus=np.array(taus))
randomWalkSegment = getRandomWalkSegment(taus_used, adev)
biasInstabilityPoint = getBiasInstabilityPoint(taus_used, adev)
randomWalk = randomWalkSegment[3]
biasInstability = biasInstabilityPoint[1]
"""""" """""" """
" Save as CSV "
""" """""" """"""
if (currentAxis == 1):
fname = 'allan_accel_x'
title = 'Allan Deviation: Accelerometer X'
elif (currentAxis == 2):
fname = 'allan_accel_y'
title = 'Allan Deviation: Accelerometer Y'
elif (currentAxis == 3):
fname = 'allan_accel_z'
title = 'Allan Deviation: Accelerometer Z'
elif (currentAxis == 4):
fname = 'allan_gyro_x'
title = 'Allan Deviation: Gyroscope X'
elif (currentAxis == 5):
fname = 'allan_gyro_y'
title = 'Allan Deviation: Gyroscope Y'
elif (currentAxis == 6):
fname = 'allan_gyro_z'
title = 'Allan Deviation: Gyroscope Z'
print "[%0.2f seconds] Finished calculating allan variance - writing results to %s" % (
rospy.get_time() - t0, fname)
f = open(resultsPath + fname + '.csv', 'wt')
try:
writer = csv.writer(f)
writer.writerow(('Random Walk', 'Bias Instability'))
writer.writerow((randomWalk, biasInstability))
writer.writerow(('Tau', 'AllanDev', 'AllanDevError', 'AllanDevN'))
for i in range(taus_used.size):
writer.writerow(
(taus_used[i], adev[i], adev_err[i], adev_n[i]))
finally:
f.close()
"""""" """""" """
" Plot Result "
""" """""" """"""
plt.figure(figsize=(12, 8))
ax = plt.gca()
ax.set_yscale('log')
ax.set_xscale('log')
plt.plot(taus_used, adev)
plt.plot([randomWalkSegment[0], randomWalkSegment[2]],
[randomWalkSegment[1], randomWalkSegment[3]], 'k--')
plt.plot(1, randomWalk, 'rx', markeredgewidth=2.5, markersize=14.0)
plt.plot(biasInstabilityPoint[0], biasInstabilityPoint[1], 'ro')
plt.grid(True, which="both")
plt.title(title)
plt.xlabel('Tau (s)')
plt.ylabel('ADEV')
for item in ([ax.title, ax.xaxis.label, ax.yaxis.label] +
ax.get_xticklabels() + ax.get_yticklabels()):
item.set_fontsize(20)
plt.show(block=False)
plt.savefig(resultsPath + fname)
currentAxis = currentAxis + 1 + axis * 6 # increment currentAxis also break if axis is not =0
inp = raw_input("Press Enter key to close figures and end program\n")
# S_x
plt.figure()
plt.loglog(f_x,psd_x,label='numpy.fft()')
plt.loglog(f_x2,psd_x2,label='scipy.signal.welch()')
plt.loglog(f_x,[h2/(2*math.pi)**2 ]*len(f_x),label='h2/(4*pi^2)' )
plt.legend(framealpha=0.5)
plt.title('PSD of phase (time)')
plt.xlabel('Frequeny / Hz')
plt.ylabel('one-sided PSD / S_x(f)')
plt.grid()
# ADEV figure
plt.figure()
taus=[tt for tt in np.logspace(-7,4,100)]
(taus_y, devs_y, errs_y, ns_y) = allantools.oadev(y,data_type='freq', rate=fs, taus=taus)
(taus_x, devs_x, errs_x, ns_x) = allantools.oadev(x,data_type='phase', rate=fs, taus=taus)
plt.loglog(taus_y,devs_y,'o',label='ADEV from y')
plt.loglog(taus_x,devs_x,'*',label='ADEV from x')
print devs_y
print devs_x
fh = 0.5*fs # this restricts the noise power to below fh
adev_y = [math.sqrt( 3*fh/(4*math.pi**2) * h2*(1.0/tt**2) ) for tt in taus]
plt.loglog(taus,adev_y,label='sqrt( 3*fh/(4*pi^2) * h2*tau^-2 )')
plt.xlim((1e-7,1e3))
plt.legend(framealpha=0.6)
plt.title('Allan deviation')
plt.xlabel('Tau / s')
plt.ylabel('Allan deviation')
plt.grid()