def add_Q_i_err(df):
df['Q_i_err'] = [
qi_error(row.Q, row.Q_err, row.Q_e_real, row.Q_e_real_err,
row.Q_e_imag, row.Q_e_imag_err)
for index, row in df.iterrows()
]
def __init__(self,sweep_filename,sweep_group_index=0,timestream_filename=None,timestream_group_index=0,
resonator_index=0,low_pass_cutoff_Hz=4.0,
dac_chain_gain = -49, delay_estimate=None,
deglitch_threshold=5, cryostat=None, mask_sweep_indicies=None, use_sweep_group_timestream=False,
conjugate_data=False):
"""
sweep_filename : str
NetCDF4 file with at least the sweep data. By default, this is also used for the timestream data.
sweep_group_index : int (default 0)
Index of the sweep group to process
timestream_filename : str or None (optional)
If None, use sweep_filename for the timestream data
otherwise, this is the NetCDF4 file with the timestream data
timestream_group_index : int (default 0)
Index of the timestream group to process
resonator_index : int (default 0)
index of the resonator to process data for
low_pass_cutoff_Hz : float
Cutoff frequency used for low pass filtering the timeseries for display
dac_chain_gain : float
Estimate of the gain from output of DAC to device.
Default value of -49 represents -2 dB loss intrinsic to analog signal conditioning
board, 7 dB misc cable loss and 40 dB total cold
attenuation.
delay_estimate : float
Estimate of basic cable delay to help fitting proceed more smoothly. Default value
is appropriate for the wideband noise firmware
deglitch_threshold : float
Threshold for glitch detection in units of median absolute deviation.
cryostat : str
(Optional) Override the cryostat used to take this data. By default, the cryostat
is guessed based on the machine you are processing data on. The guess is made by
the get_experiment_info_at function
"""
if type(sweep_filename) is str:
self.sweep_filename = sweep_filename
if timestream_filename:
self.timestream_filename = timestream_filename
else:
self.timestream_filename = self.sweep_filename
self._sweep_file = None
self._timestream_file = None
self._close_file = True
else:
# TODO: Fix this to allow different timestream file
self._sweep_file = sweep_filename
self._timestream_file = sweep_filename
self._close_file = False
self.sweep_filename = self._sweep_file.filename
self.timestream_filename = self._timestream_file.filename
self.sweep_group_index = sweep_group_index
self.timestream_group_index = timestream_group_index
self._use_sweep_group_timestream = use_sweep_group_timestream
self.sweep_epoch = self.sweep.start_epoch
pkg1,pkg2,load1,load2 = get_temperatures_at(self.sweep.start_epoch)
self.sweep_primary_package_temperature = pkg1
self.sweep_secondary_package_temperature = pkg2
self.sweep_primary_load_temperature = load1
self.sweep_secondary_load_temperature = load2
self.start_temp = self.sweep_primary_package_temperature
self.resonator_index = resonator_index
info = experiments.get_experiment_info_at(self.sweep_epoch, cryostat=cryostat)
self.experiment_description = info['description']
self.experiment_info = info
self.chip_name = info['chip_id']
self.is_dark = info['is_dark']
self.optical_state = info['optical_state']
self.dac_chain_gain = dac_chain_gain
self.mmw_atten_turns = self._sweep_file.mmw_atten_turns
try:
self.atten, self.total_dac_atten = self._sweep_file.get_effective_dac_atten_at(self.sweep_epoch)
self.power_dbm = dac_chain_gain - self.total_dac_atten
except:
print "failed to find attenuator settings"
self.atten = np.nan
self.total_dac_atten = np.nan
self.power_dbm = np.nan
# This uses the error calculation in readoutnc.SweepGroup
self.sweep_freqs_MHz, self.sweep_s21, self.sweep_errors = self.sweep.select_by_index(resonator_index)
self.conjugate_data = conjugate_data
if conjugate_data:
self.sweep_s21 = np.conj(self.sweep_s21)
# find the time series that was measured closest to the sweep frequencies
# this is a bit sloppy...
if self._use_sweep_group_timestream:
if delay_estimate is None:
self.delay_estimate_microseconds = self._sweep_file.get_delay_estimate()*1e6
else:
self.delay_estimate_microseconds = delay_estimate
if mask_sweep_indicies is None:
rr = fit_best_resonator(self.sweep_freqs_MHz,self.sweep_s21,errors=self.sweep_errors,
delay_estimate=self.delay_estimate_microseconds)
else:
mask = np.ones(self.sweep_s21.shape,dtype=np.bool)
mask[mask_sweep_indicies] = False
rr = fit_best_resonator(self.sweep_freqs_MHz[mask],self.sweep_s21[mask],errors=self.sweep_errors[mask],
delay_estimate=self.delay_estimate_microseconds)
timestream_index = np.argmin(abs(self.timestream.measurement_freq-rr.f_0))
else:
timestream_index = np.argmin(abs(self.timestream.measurement_freq-self.sweep_freqs_MHz.mean()))
self.timestream_index = timestream_index
original_timeseries = self.timestream.get_data_index(timestream_index)
if conjugate_data:
original_timeseries = np.conj(original_timeseries)
self.adc_sampling_freq_MHz = self.timestream.adc_sampling_freq[timestream_index]
self.noise_measurement_freq_MHz = self.timestream.measurement_freq[timestream_index]
self.nfft = self.timestream.nfft[timestream_index]
self.timeseries_sample_rate = self.timestream.sample_rate[timestream_index]
self.timestream_modulation_duty_cycle = self.timestream.modulation_duty_cycle[timestream_index]
self.timestream_modulation_freq = self.timestream.modulation_freq[timestream_index]
self.timestream_modulation_phase = self.timestream.modulation_phase[timestream_index]
self.timestream_modulation_period_samples = self.timestream.modulation_period_samples[timestream_index]
if not self._use_sweep_group_timestream:
self.timestream_mmw_source_freq = self.timestream.mmw_source_freq[timestream_index]
old_style_source_modulation_freq = self.timestream.mmw_source_modulation_freq[timestream_index]
else:
self.timestream_mmw_source_freq = np.nan
old_style_source_modulation_freq = np.nan
if (np.isfinite(old_style_source_modulation_freq) and
(old_style_source_modulation_freq != self.timestream_modulation_freq) and
(old_style_source_modulation_freq != 0)):
print ("found old style modulation frequency", old_style_source_modulation_freq,
"which doesn't match the new style",
self.timestream_modulation_freq,"using the old style value")
self.timestream_modulation_freq = old_style_source_modulation_freq
self.timestream_modulation_period_samples = int(self.timeseries_sample_rate/old_style_source_modulation_freq)
self.timestream_modulation_duty_cycle = 0.5
self.timestream_epoch = self.timestream.epoch[timestream_index]
self.timestream_duration = original_timeseries.shape[0]/self.timeseries_sample_rate
# The following hack helps fix a long standing timing bug which was recently fixed/improved
if self.timestream_epoch < 1399089567:
self.timestream_epoch -= self.timestream_duration
# end hack
self.timestream_temperatures_sample_times = np.arange(self.timestream_duration)
pkg1,pkg2,load1,load2 = get_temperatures_at(self.timestream_epoch + self.timestream_temperatures_sample_times)
self.timestream_primary_package_temperature = pkg1
self.timestream_secondary_package_temperature = pkg2
self.timestream_primary_load_temperature = load1
self.timestream_secondary_load_temperature = load2
self.end_temp = self.timestream_primary_package_temperature[-1]
# We can use the timestream measurement as an additional sweep point.
# We average only the first 2048 points of the timeseries to avoid any drift.
if False:
self.sweep_freqs_MHz = np.hstack((self.sweep_freqs_MHz,[self.noise_measurement_freq_MHz]))
self.sweep_s21 = np.hstack((self.sweep_s21,[original_timeseries[:2048].mean()]))
self.sweep_errors = np.hstack((self.sweep_errors,
[original_timeseries[:2048].real.std()/np.sqrt(2048)
+1j*original_timeseries[:2048].imag.std()/np.sqrt(2048)]))
# Now put all the sweep data in increasing frequency order so it plots nicely
order = self.sweep_freqs_MHz.argsort()
self.sweep_freqs_MHz = self.sweep_freqs_MHz[order]
self.sweep_s21 = self.sweep_s21[order]
self.sweep_errors = self.sweep_errors[order]
if delay_estimate is None:
self.delay_estimate_microseconds = self._sweep_file.get_delay_estimate()*1e6
else:
self.delay_estimate_microseconds = delay_estimate
if mask_sweep_indicies is None:
rr = fit_best_resonator(self.sweep_freqs_MHz,self.sweep_s21,errors=self.sweep_errors,
delay_estimate=self.delay_estimate_microseconds)
else:
mask = np.ones(self.sweep_s21.shape,dtype=np.bool)
mask[mask_sweep_indicies] = False
rr = fit_best_resonator(self.sweep_freqs_MHz[mask],self.sweep_s21[mask],errors=self.sweep_errors[mask],
delay_estimate=self.delay_estimate_microseconds)
self._resonator_model = rr
self.Q_i = rr.Q_i
self.Q_i_err = qi_error(rr.result.params['Q'].value, rr.result.params['Q'].stderr,
rr.result.params['Q_e_real'].value, rr.result.params['Q_e_real'].stderr,
rr.result.params['Q_e_imag'].value, rr.result.params['Q_e_imag'].stderr)
self.fit_params = rr.result.params
decimation_factor = self.timeseries_sample_rate/low_pass_cutoff_Hz
normalized_timeseries = rr.normalize(self.noise_measurement_freq_MHz,original_timeseries)
self.low_pass_normalized_timeseries = low_pass_fir(normalized_timeseries, num_taps=1024, cutoff=low_pass_cutoff_Hz,
nyquist_freq=self.timeseries_sample_rate, decimate_by=decimation_factor)
self.normalized_timeseries_mean = normalized_timeseries.mean()
projected_timeseries = rr.project_s21_to_delta_freq(self.noise_measurement_freq_MHz,normalized_timeseries,
s21_already_normalized=True)
# calculate the number of samples for the deglitching window.
# the following will be the next power of two above 1 second worth of samples
window = int(2**np.ceil(np.log2(self.timeseries_sample_rate)))
# reduce the deglitching window if we don't have enough samples
if window > projected_timeseries.shape[0]:
window = projected_timeseries.shape[0]//2
self.deglitch_window = window
self.deglitch_threshold = deglitch_threshold
if deglitch_threshold:
deglitched_timeseries = deglitch_window(projected_timeseries,window,thresh=deglitch_threshold)
else:
deglitched_timeseries = projected_timeseries
# TODO: should nyquist_freq be half the sample rate?
self.low_pass_projected_timeseries = low_pass_fir(deglitched_timeseries, num_taps=1024, cutoff=low_pass_cutoff_Hz,
nyquist_freq=self.timeseries_sample_rate, decimate_by=decimation_factor)
self.low_pass_timestep = decimation_factor/self.timeseries_sample_rate
self.normalized_model_s21_at_meas_freq = rr.normalized_model(self.noise_measurement_freq_MHz)
self.normalized_model_s21_at_resonance = rr.normalized_model(rr.f_0)
self.normalized_ds21_df_at_meas_freq = rr.approx_normalized_gradient(self.noise_measurement_freq_MHz)
self.sweep_normalized_s21 = rr.normalize(self.sweep_freqs_MHz,self.sweep_s21)
self.sweep_model_freqs_MHz = np.linspace(self.sweep_freqs_MHz.min(),self.sweep_freqs_MHz.max(),1000)
self.sweep_model_normalized_s21 = rr.normalized_model(self.sweep_model_freqs_MHz)
self.sweep_model_normalized_s21_centered = self.sweep_model_normalized_s21 - self.normalized_timeseries_mean
fractional_fluctuation_timeseries = deglitched_timeseries / (self.noise_measurement_freq_MHz*1e6)
self._fractional_fluctuation_timeseries = fractional_fluctuation_timeseries
fr,S,evals,evects,angles,piq = iqnoise.pca_noise(fractional_fluctuation_timeseries,
NFFT=None, Fs=self.timeseries_sample_rate)
self.pca_freq = fr
self.pca_S = S
self.pca_eigvals = evals
self.pca_eigvects = evects
self.pca_angles = angles
self.pca_piq = piq
self.freqs_coarse,self.prr_coarse,self.pii_coarse = self.get_projected_fractional_fluctuation_spectra(NFFT=2**12)
self._normalized_timeseries = normalized_timeseries[:2048].copy()
self._close_files()
def add_Q_i_err(df):
df['Q_i_err'] = [qi_error(row.Q, row.Q_err, row.Q_e_real, row.Q_e_real_err, row.Q_e_imag, row.Q_e_imag_err)
for index, row in df.iterrows()]
def __init__(self,
sweep_filename,
sweep_group_index=0,
timestream_filename=None,
timestream_group_index=0,
resonator_index=0,
low_pass_cutoff_Hz=4.0,
dac_chain_gain=-49,
delay_estimate=None,
deglitch_threshold=5,
cryostat=None,
mask_sweep_indicies=None,
use_sweep_group_timestream=False,
conjugate_data=False):
"""
sweep_filename : str
NetCDF4 file with at least the sweep data. By default, this is also used for the timestream data.
sweep_group_index : int (default 0)
Index of the sweep group to process
timestream_filename : str or None (optional)
If None, use sweep_filename for the timestream data
otherwise, this is the NetCDF4 file with the timestream data
timestream_group_index : int (default 0)
Index of the timestream group to process
resonator_index : int (default 0)
index of the resonator to process data for
low_pass_cutoff_Hz : float
Cutoff frequency used for low pass filtering the timeseries for display
dac_chain_gain : float
Estimate of the gain from output of DAC to device.
Default value of -49 represents -2 dB loss intrinsic to analog signal conditioning
board, 7 dB misc cable loss and 40 dB total cold
attenuation.
delay_estimate : float
Estimate of basic cable delay to help fitting proceed more smoothly. Default value
is appropriate for the wideband noise firmware
deglitch_threshold : float
Threshold for glitch detection in units of median absolute deviation.
cryostat : str
(Optional) Override the cryostat used to take this data. By default, the cryostat
is guessed based on the machine you are processing data on. The guess is made by
the get_experiment_info_at function
"""
if type(sweep_filename) is str:
self.sweep_filename = sweep_filename
if timestream_filename:
self.timestream_filename = timestream_filename
else:
self.timestream_filename = self.sweep_filename
self._sweep_file = None
self._timestream_file = None
self._close_file = True
else:
# TODO: Fix this to allow different timestream file
self._sweep_file = sweep_filename
self._timestream_file = sweep_filename
self._close_file = False
self.sweep_filename = self._sweep_file.filename
self.timestream_filename = self._timestream_file.filename
self.sweep_group_index = sweep_group_index
self.timestream_group_index = timestream_group_index
self._use_sweep_group_timestream = use_sweep_group_timestream
self.sweep_epoch = self.sweep.start_epoch
pkg1, pkg2, load1, load2 = get_temperatures_at(self.sweep.start_epoch)
self.sweep_primary_package_temperature = pkg1
self.sweep_secondary_package_temperature = pkg2
self.sweep_primary_load_temperature = load1
self.sweep_secondary_load_temperature = load2
self.start_temp = self.sweep_primary_package_temperature
self.resonator_index = resonator_index
info = experiments.get_experiment_info_at(self.sweep_epoch,
cryostat=cryostat)
self.experiment_description = info['description']
self.experiment_info = info
self.chip_name = info['chip_id']
self.is_dark = info['is_dark']
self.optical_state = info['optical_state']
self.dac_chain_gain = dac_chain_gain
self.mmw_atten_turns = self._sweep_file.mmw_atten_turns
try:
self.atten, self.total_dac_atten = self._sweep_file.get_effective_dac_atten_at(
self.sweep_epoch)
self.power_dbm = dac_chain_gain - self.total_dac_atten
except:
print "failed to find attenuator settings"
self.atten = np.nan
self.total_dac_atten = np.nan
self.power_dbm = np.nan
# This uses the error calculation in readoutnc.SweepGroup
self.sweep_freqs_MHz, self.sweep_s21, self.sweep_errors = self.sweep.select_by_index(
resonator_index)
self.conjugate_data = conjugate_data
if conjugate_data:
self.sweep_s21 = np.conj(self.sweep_s21)
# find the time series that was measured closest to the sweep frequencies
# this is a bit sloppy...
if self._use_sweep_group_timestream:
if delay_estimate is None:
self.delay_estimate_microseconds = self._sweep_file.get_delay_estimate(
) * 1e6
else:
self.delay_estimate_microseconds = delay_estimate
if mask_sweep_indicies is None:
rr = fit_best_resonator(
self.sweep_freqs_MHz,
self.sweep_s21,
errors=self.sweep_errors,
delay_estimate=self.delay_estimate_microseconds)
else:
mask = np.ones(self.sweep_s21.shape, dtype=np.bool)
mask[mask_sweep_indicies] = False
rr = fit_best_resonator(
self.sweep_freqs_MHz[mask],
self.sweep_s21[mask],
errors=self.sweep_errors[mask],
delay_estimate=self.delay_estimate_microseconds)
timestream_index = np.argmin(
abs(self.timestream.measurement_freq - rr.f_0))
else:
timestream_index = np.argmin(
abs(self.timestream.measurement_freq -
self.sweep_freqs_MHz.mean()))
self.timestream_index = timestream_index
original_timeseries = self.timestream.get_data_index(timestream_index)
if conjugate_data:
original_timeseries = np.conj(original_timeseries)
self.adc_sampling_freq_MHz = self.timestream.adc_sampling_freq[
timestream_index]
self.noise_measurement_freq_MHz = self.timestream.measurement_freq[
timestream_index]
self.nfft = self.timestream.nfft[timestream_index]
self.timeseries_sample_rate = self.timestream.sample_rate[
timestream_index]
self.timestream_modulation_duty_cycle = self.timestream.modulation_duty_cycle[
timestream_index]
self.timestream_modulation_freq = self.timestream.modulation_freq[
timestream_index]
self.timestream_modulation_phase = self.timestream.modulation_phase[
timestream_index]
self.timestream_modulation_period_samples = self.timestream.modulation_period_samples[
timestream_index]
if not self._use_sweep_group_timestream:
self.timestream_mmw_source_freq = self.timestream.mmw_source_freq[
timestream_index]
old_style_source_modulation_freq = self.timestream.mmw_source_modulation_freq[
timestream_index]
else:
self.timestream_mmw_source_freq = np.nan
old_style_source_modulation_freq = np.nan
if (np.isfinite(old_style_source_modulation_freq)
and (old_style_source_modulation_freq !=
self.timestream_modulation_freq)
and (old_style_source_modulation_freq != 0)):
print("found old style modulation frequency",
old_style_source_modulation_freq,
"which doesn't match the new style",
self.timestream_modulation_freq, "using the old style value")
self.timestream_modulation_freq = old_style_source_modulation_freq
self.timestream_modulation_period_samples = int(
self.timeseries_sample_rate / old_style_source_modulation_freq)
self.timestream_modulation_duty_cycle = 0.5
self.timestream_epoch = self.timestream.epoch[timestream_index]
self.timestream_duration = original_timeseries.shape[
0] / self.timeseries_sample_rate
# The following hack helps fix a long standing timing bug which was recently fixed/improved
if self.timestream_epoch < 1399089567:
self.timestream_epoch -= self.timestream_duration
# end hack
self.timestream_temperatures_sample_times = np.arange(
self.timestream_duration)
pkg1, pkg2, load1, load2 = get_temperatures_at(
self.timestream_epoch + self.timestream_temperatures_sample_times)
self.timestream_primary_package_temperature = pkg1
self.timestream_secondary_package_temperature = pkg2
self.timestream_primary_load_temperature = load1
self.timestream_secondary_load_temperature = load2
self.end_temp = self.timestream_primary_package_temperature[-1]
# We can use the timestream measurement as an additional sweep point.
# We average only the first 2048 points of the timeseries to avoid any drift.
if False:
self.sweep_freqs_MHz = np.hstack(
(self.sweep_freqs_MHz, [self.noise_measurement_freq_MHz]))
self.sweep_s21 = np.hstack(
(self.sweep_s21, [original_timeseries[:2048].mean()]))
self.sweep_errors = np.hstack((self.sweep_errors, [
original_timeseries[:2048].real.std() / np.sqrt(2048) +
1j * original_timeseries[:2048].imag.std() / np.sqrt(2048)
]))
# Now put all the sweep data in increasing frequency order so it plots nicely
order = self.sweep_freqs_MHz.argsort()
self.sweep_freqs_MHz = self.sweep_freqs_MHz[order]
self.sweep_s21 = self.sweep_s21[order]
self.sweep_errors = self.sweep_errors[order]
if delay_estimate is None:
self.delay_estimate_microseconds = self._sweep_file.get_delay_estimate(
) * 1e6
else:
self.delay_estimate_microseconds = delay_estimate
if mask_sweep_indicies is None:
rr = fit_best_resonator(
self.sweep_freqs_MHz,
self.sweep_s21,
errors=self.sweep_errors,
delay_estimate=self.delay_estimate_microseconds)
else:
mask = np.ones(self.sweep_s21.shape, dtype=np.bool)
mask[mask_sweep_indicies] = False
rr = fit_best_resonator(
self.sweep_freqs_MHz[mask],
self.sweep_s21[mask],
errors=self.sweep_errors[mask],
delay_estimate=self.delay_estimate_microseconds)
self._resonator_model = rr
self.Q_i = rr.Q_i
self.Q_i_err = qi_error(rr.result.params['Q'].value,
rr.result.params['Q'].stderr,
rr.result.params['Q_e_real'].value,
rr.result.params['Q_e_real'].stderr,
rr.result.params['Q_e_imag'].value,
rr.result.params['Q_e_imag'].stderr)
self.fit_params = rr.result.params
decimation_factor = self.timeseries_sample_rate / low_pass_cutoff_Hz
normalized_timeseries = rr.normalize(self.noise_measurement_freq_MHz,
original_timeseries)
self.low_pass_normalized_timeseries = low_pass_fir(
normalized_timeseries,
num_taps=1024,
cutoff=low_pass_cutoff_Hz,
nyquist_freq=self.timeseries_sample_rate,
decimate_by=decimation_factor)
self.normalized_timeseries_mean = normalized_timeseries.mean()
projected_timeseries = rr.project_s21_to_delta_freq(
self.noise_measurement_freq_MHz,
normalized_timeseries,
s21_already_normalized=True)
# calculate the number of samples for the deglitching window.
# the following will be the next power of two above 1 second worth of samples
window = int(2**np.ceil(np.log2(self.timeseries_sample_rate)))
# reduce the deglitching window if we don't have enough samples
if window > projected_timeseries.shape[0]:
window = projected_timeseries.shape[0] // 2
self.deglitch_window = window
self.deglitch_threshold = deglitch_threshold
if deglitch_threshold:
deglitched_timeseries = deglitch_window(projected_timeseries,
window,
thresh=deglitch_threshold)
else:
deglitched_timeseries = projected_timeseries
# TODO: should nyquist_freq be half the sample rate?
self.low_pass_projected_timeseries = low_pass_fir(
deglitched_timeseries,
num_taps=1024,
cutoff=low_pass_cutoff_Hz,
nyquist_freq=self.timeseries_sample_rate,
decimate_by=decimation_factor)
self.low_pass_timestep = decimation_factor / self.timeseries_sample_rate
self.normalized_model_s21_at_meas_freq = rr.normalized_model(
self.noise_measurement_freq_MHz)
self.normalized_model_s21_at_resonance = rr.normalized_model(rr.f_0)
self.normalized_ds21_df_at_meas_freq = rr.approx_normalized_gradient(
self.noise_measurement_freq_MHz)
self.sweep_normalized_s21 = rr.normalize(self.sweep_freqs_MHz,
self.sweep_s21)
self.sweep_model_freqs_MHz = np.linspace(self.sweep_freqs_MHz.min(),
self.sweep_freqs_MHz.max(),
1000)
self.sweep_model_normalized_s21 = rr.normalized_model(
self.sweep_model_freqs_MHz)
self.sweep_model_normalized_s21_centered = self.sweep_model_normalized_s21 - self.normalized_timeseries_mean
fractional_fluctuation_timeseries = deglitched_timeseries / (
self.noise_measurement_freq_MHz * 1e6)
self._fractional_fluctuation_timeseries = fractional_fluctuation_timeseries
fr, S, evals, evects, angles, piq = iqnoise.pca_noise(
fractional_fluctuation_timeseries,
NFFT=None,
Fs=self.timeseries_sample_rate)
self.pca_freq = fr
self.pca_S = S
self.pca_eigvals = evals
self.pca_eigvects = evects
self.pca_angles = angles
self.pca_piq = piq
self.freqs_coarse, self.prr_coarse, self.pii_coarse = self.get_projected_fractional_fluctuation_spectra(
NFFT=2**12)
self._normalized_timeseries = normalized_timeseries[:2048].copy()
self._close_files()