Added the aps calculator where blocks are neglectd below a certain threshold. Warning, this is sweep specific and is currently the method implemented in lasp_measurement. Fine for muZ, not necessarily for other stuff
This commit is contained in:
parent
9caf5fe387
commit
a986a6b9cd
@ -10,7 +10,8 @@ PowerSpectra::PowerSpectra(const us nfft, const Window::WindowType w)
|
||||
: PowerSpectra(Window::create(w, nfft)) {}
|
||||
|
||||
PowerSpectra::PowerSpectra(const vd &window)
|
||||
: nfft(window.size()), _fft(nfft), _window(window) {
|
||||
: nfft(window.size()), _fft(nfft), _window(window)
|
||||
{
|
||||
|
||||
d win_pow = arma::sum(window % window) / window.size();
|
||||
|
||||
@ -21,7 +22,8 @@ PowerSpectra::PowerSpectra(const vd &window)
|
||||
DEBUGTRACE_PRINT(win_pow);
|
||||
}
|
||||
|
||||
ccube PowerSpectra::compute(const dmat &input) {
|
||||
ccube PowerSpectra::compute(const dmat &input)
|
||||
{
|
||||
|
||||
/// Run very often. Silence this one.
|
||||
/* DEBUGTRACE_ENTER; */
|
||||
@ -34,11 +36,13 @@ ccube PowerSpectra::compute(const dmat &input) {
|
||||
cmat rfft = _fft.fft(input_tmp);
|
||||
|
||||
ccube output(rfft.n_rows, input.n_cols, input.n_cols);
|
||||
for (us i = 0; i < input.n_cols; i++) {
|
||||
for (us i = 0; i < input.n_cols; i++)
|
||||
{
|
||||
/// This one can be run in parallel without any problem. Note that it is
|
||||
/// the inner loop that is run in parallel.
|
||||
#pragma omp parallel for
|
||||
for (us j = 0; j < input.n_cols; j++) {
|
||||
for (us j = 0; j < input.n_cols; j++)
|
||||
{
|
||||
output.slice(j).col(i) =
|
||||
_scale_fac * (rfft.col(i) % arma::conj(rfft.col(j)));
|
||||
|
||||
@ -52,37 +56,46 @@ ccube PowerSpectra::compute(const dmat &input) {
|
||||
AvPowerSpectra::AvPowerSpectra(const us nfft, const Window::WindowType w,
|
||||
const d overlap_percentage,
|
||||
const d time_constant_times_fs)
|
||||
: _ps(nfft, w) {
|
||||
: _ps(nfft, w)
|
||||
{
|
||||
|
||||
DEBUGTRACE_ENTER;
|
||||
if (overlap_percentage >= 100 || overlap_percentage < 0) {
|
||||
if (overlap_percentage >= 100 || overlap_percentage < 0)
|
||||
{
|
||||
throw rte("Overlap percentage should be >= 0 and < 100");
|
||||
}
|
||||
|
||||
_overlap_keep = (nfft * overlap_percentage) / 100;
|
||||
DEBUGTRACE_PRINT(_overlap_keep);
|
||||
if (_overlap_keep >= nfft) {
|
||||
if (_overlap_keep >= nfft)
|
||||
{
|
||||
throw rte("Overlap is too high. Results in no jump. Please "
|
||||
"choose a smaller overlap percentage or a higher nfft");
|
||||
}
|
||||
if (time_constant_times_fs < 0) {
|
||||
if (time_constant_times_fs < 0)
|
||||
{
|
||||
_mode = Mode::Averaging;
|
||||
} else if (time_constant_times_fs == 0) {
|
||||
}
|
||||
else if (time_constant_times_fs == 0)
|
||||
{
|
||||
_mode = Mode::Spectrogram;
|
||||
} else {
|
||||
}
|
||||
else
|
||||
{
|
||||
_mode = Mode::Leaking;
|
||||
_alpha = d_exp(-static_cast<d>((nfft - _overlap_keep)/time_constant_times_fs));
|
||||
_alpha = d_exp(-static_cast<d>((nfft - _overlap_keep) / time_constant_times_fs));
|
||||
DEBUGTRACE_PRINT(_alpha);
|
||||
}
|
||||
}
|
||||
void AvPowerSpectra::reset() {
|
||||
void AvPowerSpectra::reset()
|
||||
{
|
||||
_timeBuf.reset();
|
||||
_est.reset();
|
||||
n_averages=0;
|
||||
|
||||
_n_averages = 0;
|
||||
}
|
||||
|
||||
ccube AvPowerSpectra::compute(const dmat &timedata) {
|
||||
ccube AvPowerSpectra::compute(const dmat &timedata)
|
||||
{
|
||||
|
||||
DEBUGTRACE_ENTER;
|
||||
DEBUGTRACE_PRINT(timedata.n_rows);
|
||||
@ -91,32 +104,42 @@ ccube AvPowerSpectra::compute(const dmat &timedata) {
|
||||
_timeBuf.push(timedata);
|
||||
|
||||
bool run_once = false;
|
||||
while (_timeBuf.size() >= _ps.nfft) {
|
||||
while (_timeBuf.size() >= _ps.nfft)
|
||||
{
|
||||
|
||||
DEBUGTRACE_PRINT((int)_mode);
|
||||
|
||||
dmat samples = _timeBuf.pop(_ps.nfft, _overlap_keep);
|
||||
|
||||
switch (_mode) {
|
||||
switch (_mode)
|
||||
{
|
||||
case (Mode::Spectrogram):
|
||||
DEBUGTRACE_PRINT("Spectrogram mode");
|
||||
_est = _ps.compute(samples);
|
||||
|
||||
break;
|
||||
case (Mode::Averaging):
|
||||
DEBUGTRACE_PRINT("Averaging mode");
|
||||
n_averages++;
|
||||
if (n_averages == 1) {
|
||||
_n_averages++;
|
||||
if (_n_averages == 1)
|
||||
{
|
||||
_est = _ps.compute(samples);
|
||||
} else {
|
||||
_est = (static_cast<d>(n_averages - 1) / n_averages) * _est +
|
||||
_ps.compute(samples) / n_averages;
|
||||
|
||||
}
|
||||
else
|
||||
{
|
||||
_est = (static_cast<d>(_n_averages - 1) / _n_averages) * _est +
|
||||
_ps.compute(samples) / _n_averages;
|
||||
}
|
||||
break;
|
||||
case (Mode::Leaking):
|
||||
DEBUGTRACE_PRINT("Leaking mode");
|
||||
if (arma::size(_est) == arma::size(0, 0, 0)) {
|
||||
if (arma::size(_est) == arma::size(0, 0, 0))
|
||||
{
|
||||
_est = _ps.compute(samples);
|
||||
} else {
|
||||
}
|
||||
else
|
||||
{
|
||||
_est = _alpha * _est + (1 - _alpha) * _ps.compute(samples);
|
||||
}
|
||||
break;
|
||||
@ -128,6 +151,135 @@ ccube AvPowerSpectra::compute(const dmat &timedata) {
|
||||
/// Othewise, we return an empty ccube.
|
||||
return run_once ? _est : ccube();
|
||||
}
|
||||
ccube AvPowerSpectra::get_est() const {
|
||||
ccube AvPowerSpectra::get_est() const
|
||||
{
|
||||
return _est;
|
||||
}
|
||||
|
||||
AvSweepPowerSpectra::AvSweepPowerSpectra(const us nfft, const Window::WindowType w,
|
||||
const d overlap_percentage,
|
||||
const d time_constant_times_fs)
|
||||
: _nfft(nfft), _ps(nfft, w)
|
||||
{
|
||||
|
||||
DEBUGTRACE_ENTER;
|
||||
if (overlap_percentage >= 100 || overlap_percentage < 0)
|
||||
{
|
||||
throw rte("Overlap percentage should be >= 0 and < 100");
|
||||
}
|
||||
|
||||
_overlap_keep = (nfft * overlap_percentage) / 100;
|
||||
DEBUGTRACE_PRINT(_overlap_keep);
|
||||
if (_overlap_keep >= nfft)
|
||||
{
|
||||
throw rte("Overlap is too high. Results in no jump. Please "
|
||||
"choose a smaller overlap percentage or a higher nfft");
|
||||
}
|
||||
if (time_constant_times_fs < 0)
|
||||
{
|
||||
_mode = Mode::Averaging;
|
||||
}
|
||||
else if (time_constant_times_fs == 0)
|
||||
{
|
||||
_mode = Mode::Spectrogram;
|
||||
}
|
||||
else
|
||||
{
|
||||
_mode = Mode::Leaking;
|
||||
_alpha = d_exp(-static_cast<d>((nfft - _overlap_keep) / time_constant_times_fs));
|
||||
DEBUGTRACE_PRINT(_alpha);
|
||||
}
|
||||
}
|
||||
void AvSweepPowerSpectra::reset()
|
||||
{
|
||||
_timeBuf.reset();
|
||||
_est.reset();
|
||||
_n_averages.reset();
|
||||
}
|
||||
|
||||
ccube AvSweepPowerSpectra::compute(const dmat &timedata)
|
||||
{
|
||||
|
||||
DEBUGTRACE_ENTER;
|
||||
DEBUGTRACE_PRINT(timedata.n_rows);
|
||||
DEBUGTRACE_PRINT(_ps.nfft);
|
||||
|
||||
_timeBuf.push(timedata);
|
||||
|
||||
bool run_once = false;
|
||||
while (_timeBuf.size() >= _ps.nfft)
|
||||
{
|
||||
|
||||
DEBUGTRACE_PRINT((int)_mode);
|
||||
|
||||
dmat samples = _timeBuf.pop(_ps.nfft, _overlap_keep);
|
||||
|
||||
us ncols = timedata.n_cols;
|
||||
|
||||
switch (_mode)
|
||||
{
|
||||
case (Mode::Spectrogram):
|
||||
DEBUGTRACE_PRINT("Spectrogram mode");
|
||||
_est = _ps.compute(samples);
|
||||
break;
|
||||
case (Mode::Averaging):
|
||||
{
|
||||
DEBUGTRACE_PRINT("Averaging mode");
|
||||
|
||||
ccube temp = _ps.compute(samples);
|
||||
us nfreq = arma::size(temp)[0];
|
||||
|
||||
if (_est.size() == 0) {
|
||||
// Initialize empty
|
||||
_est = ccube(nfreq, ncols, ncols, arma::fill::zeros);
|
||||
_n_averages = vd(nfreq, arma::fill::zeros);
|
||||
}
|
||||
|
||||
d threshold = arma::mean(arma::real(temp.slice(0).col(0)));
|
||||
|
||||
for (us f=0; f < nfreq; f++)
|
||||
{
|
||||
if (real(temp(f, 0, 0)) > threshold)
|
||||
{
|
||||
_n_averages[f]++;
|
||||
if (_n_averages[f] == 1)
|
||||
{
|
||||
_est.row(f) = temp.row(f);
|
||||
// _est.subcube(f, 0, 0, f, ncols - 1, ncols - 1) = temp.subcube(f, 0, 0, f, ncols - 1, ncols - 1);
|
||||
}
|
||||
else
|
||||
{
|
||||
_est.row(f) = (static_cast<d>(_n_averages[f] - 1) / _n_averages[f]) * _est.row(f) +
|
||||
temp.row(f) / _n_averages[f];
|
||||
// _est.subcube(f, 0, 0, f, ncols - 1, ncols - 1) = (static_cast<d>(_n_averages[f] - 1) / _n_averages[f]) * _est.subcube(f, 0, 0, f, ncols - 1, ncols - 1) +
|
||||
// temp.subcube(f, 0, 0, f, ncols - 1, ncols - 1) / _n_averages[f];
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
break;
|
||||
|
||||
case (Mode::Leaking):
|
||||
DEBUGTRACE_PRINT("Leaking mode");
|
||||
if (arma::size(_est) == arma::size(0, 0, 0))
|
||||
{
|
||||
_est = _ps.compute(samples);
|
||||
}
|
||||
else
|
||||
{
|
||||
_est = _alpha * _est + (1 - _alpha) * _ps.compute(samples);
|
||||
}
|
||||
break;
|
||||
} // end switch mode
|
||||
run_once = true;
|
||||
}
|
||||
|
||||
/// Return a copy of current estimator in case we have done one update.
|
||||
/// Othewise, we return an empty ccube.
|
||||
return run_once ? _est : ccube();
|
||||
}
|
||||
|
||||
ccube AvSweepPowerSpectra::get_est() const
|
||||
{
|
||||
return _est;
|
||||
}
|
||||
|
@ -81,7 +81,7 @@ class AvPowerSpectra {
|
||||
|
||||
Mode _mode;
|
||||
d _alpha; // Only valid in case of 'Leaking'
|
||||
us n_averages = 0; // Only valid in case of 'Averaging'
|
||||
us _n_averages = 0; // Only valid in case of 'Averaging'
|
||||
|
||||
PowerSpectra _ps;
|
||||
/**
|
||||
@ -165,3 +165,107 @@ public:
|
||||
us exactOverlapSamples() const { return _ps.nfft - _overlap_keep; }
|
||||
};
|
||||
/** @} */
|
||||
|
||||
|
||||
/**
|
||||
* @brief Estimate cross-power spectra using Welch' method of spectral
|
||||
* estimation. The exact amount of overlap in Welch' method is rounded up to a
|
||||
* certain amount of samples. This class is specifically for sweeps and will ignore
|
||||
* fft blocks where the auto-power of the "reference" signal is below the average
|
||||
* when in averaging mode.
|
||||
*/
|
||||
class AvSweepPowerSpectra {
|
||||
|
||||
enum class Mode {
|
||||
Averaging = 0, // Averaging all time date
|
||||
Leaking = 1, // Exponential weighting of an "instantaneous cps"
|
||||
Spectrogram = 2 // Instantenous spectrum, no averaging
|
||||
};
|
||||
|
||||
Mode _mode;
|
||||
d _alpha; // Only valid in case of 'Leaking'
|
||||
us _nfft;
|
||||
vd _n_averages { 1, arma::fill::zeros}; // Only valid in case of 'Averaging'
|
||||
|
||||
PowerSpectra _ps;
|
||||
/**
|
||||
* @brief Current estimate of cross-spectral density
|
||||
*/
|
||||
ccube _est;
|
||||
|
||||
/**
|
||||
* @brief Buffer of storage of time data.
|
||||
*/
|
||||
TimeBuffer _timeBuf;
|
||||
/**
|
||||
* @brief The amount of samples to keep in the overlap
|
||||
*/
|
||||
us _overlap_keep;
|
||||
|
||||
public:
|
||||
/**
|
||||
* @brief Initalize averaged power spectra computer. If a time constant is
|
||||
* given > 0, it is used in a kind of exponential weighting.
|
||||
*
|
||||
* @param nfft The fft length
|
||||
* @param w The window type.
|
||||
* @param overlap_percentage A number 0 < overlap_percentage <= 100. It
|
||||
* determines the amount of overlap used in Welch' method. A typical value is
|
||||
* 50 %, i.e. 50.
|
||||
* @param fs_tau Value should either be < 0, indicating that the
|
||||
* estimate is averages over all time data.
|
||||
* For a value = 0 the instantaneous power spectrum is returned, which can be
|
||||
* interpreted as the spectrogram. For a value > 0 a exponential forgetting is
|
||||
* used, where the value is used as the time constant such that the decay
|
||||
* follows approximately the trend exp(-n/fs_tau), where n is the
|
||||
* sample number in the power spectra. To choose 'fast' time weighting, set
|
||||
* time_constant to the value of fs*0.125, where fs denotes the sampling
|
||||
* frequency.
|
||||
**/
|
||||
AvSweepPowerSpectra(const us nfft = 2048,
|
||||
const Window::WindowType w = Window::WindowType::Hann,
|
||||
const d overlap_percentage = 50.,
|
||||
const d fs_tau = -1);
|
||||
|
||||
AvSweepPowerSpectra(const AvSweepPowerSpectra &) = delete;
|
||||
AvSweepPowerSpectra &operator=(const AvSweepPowerSpectra &) = delete;
|
||||
|
||||
/**
|
||||
* @brief Reset to empty state. Clears the time buffer and sets estimator to
|
||||
* empty.
|
||||
*/
|
||||
void reset();
|
||||
|
||||
/**
|
||||
* @brief Compute an update of the power spectra based on given time data.
|
||||
* Note that the number of channels is determined from the first time this
|
||||
* function is called. If a later call has an incompatible number of
|
||||
* channels, a runtime error is thrown.
|
||||
*
|
||||
* @param timedata
|
||||
*
|
||||
* @return a copy of the latest estimate of the power spectra. an
|
||||
* update is only given if the amount of new time data is enough to compute a
|
||||
* new estimate. if no new estimate is available, it returns an empty ccube.
|
||||
* Note that the latest available estimate can be obtained using get_est().
|
||||
* */
|
||||
ccube compute(const dmat &timedata);
|
||||
|
||||
/**
|
||||
* @brief Returns the latest estimate of cps (cross-power spectra.
|
||||
*
|
||||
* @return a copy of the latest estimate of the power spectra. an
|
||||
* update is only given if the amount of new time data is enough to compute a
|
||||
* new estimate. If no estimate is available, it returns an empty ccube.
|
||||
* */
|
||||
ccube get_est() const;
|
||||
|
||||
/**
|
||||
* @brief The overlap is rounded to a certain amount of time samples. This
|
||||
* function returns that value.
|
||||
*
|
||||
* @return The amount of samples in overlapping.
|
||||
*/
|
||||
us exactOverlapSamples() const { return _ps.nfft - _overlap_keep; }
|
||||
};
|
||||
/** @} */
|
||||
|
@ -129,6 +129,29 @@ void init_dsp(py::module &m) {
|
||||
return CubeToNpy<c>(est);
|
||||
});
|
||||
|
||||
/// AvSweepPowerSpectra
|
||||
py::class_<AvSweepPowerSpectra> asps(m, "AvSweepPowerSpectra");
|
||||
asps.def(py::init<const us, const Window::WindowType, const d, const d>(),
|
||||
py::arg("nfft") = 2048,
|
||||
py::arg("windowType") = Window::WindowType::Hann,
|
||||
py::arg("overlap_percentage") = 50.0,
|
||||
py::arg("time_constant") = -1);
|
||||
|
||||
asps.def("compute", [](AvSweepPowerSpectra &asps, dpyarray timedata) {
|
||||
std::optional<ccube> res;
|
||||
dmat timedata_mat = NpyToMat<d, false>(timedata);
|
||||
{
|
||||
py::gil_scoped_release release;
|
||||
res = asps.compute(timedata_mat);
|
||||
}
|
||||
|
||||
return CubeToNpy<c>(res.value_or(ccube(0, 0, 0)));
|
||||
});
|
||||
asps.def("get_est", [](const AvSweepPowerSpectra &sps) {
|
||||
ccube est = sps.get_est();
|
||||
return CubeToNpy<c>(est);
|
||||
});
|
||||
|
||||
py::class_<SLM> slm(m, "cppSLM");
|
||||
|
||||
slm.def_static("fromBiquads", [](const d fs, const d Lref, const us ds,
|
||||
|
@ -15,7 +15,7 @@ from scipy.io import wavfile
|
||||
import os, time, wave, logging
|
||||
from .lasp_common import SIQtys, Qty, getFreq
|
||||
from .lasp_version import LASP_VERSION_MAJOR, LASP_VERSION_MINOR
|
||||
from .lasp_cpp import Window, DaqChannel, AvPowerSpectra
|
||||
from .lasp_cpp import Window, DaqChannel, AvPowerSpectra, AvSweepPowerSpectra
|
||||
from typing import List
|
||||
from functools import lru_cache
|
||||
from .lasp_config import ones
|
||||
@ -814,9 +814,11 @@ class Measurement:
|
||||
channels = list(range(self.nchannels))
|
||||
|
||||
nchannels = len(channels)
|
||||
aps = AvPowerSpectra(nfft, window, overlap)
|
||||
freq = getFreq(self.samplerate, nfft)
|
||||
|
||||
# aps = AvPowerSpectra(nfft, window, overlap)
|
||||
aps = AvSweepPowerSpectra(nfft, window, overlap)
|
||||
freq = getFreq(self.samplerate, nfft)
|
||||
|
||||
for data in self.iterData(channels, **kwargs):
|
||||
CS = aps.compute(data)
|
||||
|
||||
@ -1005,7 +1007,8 @@ class Measurement:
|
||||
|
||||
# Convert range to [-1, 1]
|
||||
# TODO: this is wrong for float data where full scale > 1
|
||||
sensone = np.ones_like(self.sensitivity)
|
||||
sensone = np.ones(data.shape[1])
|
||||
|
||||
data = scaleBlockSens(data, sensone)
|
||||
|
||||
if dtype == "int16" or dtype == "int32":
|
||||
|
Loading…
Reference in New Issue
Block a user