Method apply_gradient_compensation

mne/io/base.py:310–357

Apply CTF gradient compensation. .. warning:: The compensation matrices are stored with single precision, so repeatedly switching between different of compensation (e.g., 0->1->3->2) can increase numerical noise, especially if d

(self, grade, verbose=None)

    @verbose
    def apply_gradient_compensation(self, grade, verbose=None):
        """Apply CTF gradient compensation.

        .. warning:: The compensation matrices are stored with single
                     precision, so repeatedly switching between different
                     of compensation (e.g., 0->1->3->2) can increase
                     numerical noise, especially if data are saved to
                     disk in between changing grades. It is thus best to
                     only use a single gradient compensation level in
                     final analyses.

        Parameters
        ----------
        grade : int
            CTF gradient compensation level.
        %(verbose)s

        Returns
        -------
        raw : instance of Raw
            The modified Raw instance. Works in-place.
        """
        grade = int(grade)
        current_comp = self.compensation_grade
        if current_comp != grade:
            if self.proj:
                raise RuntimeError(
                    "Cannot change compensation on data where projectors have been "
                    "applied."
                )
            # Figure out what operator to use (varies depending on preload)
            from_comp = current_comp if self.preload else self._read_comp_grade
            comp = make_compensator(self.info, from_comp, grade)
            logger.info(
                "Compensator constructed to change %d -> %d", current_comp, grade
            )
            set_current_comp(self.info, grade)
            # We might need to apply it to our data now
            if self.preload:
                logger.info("Applying compensator to loaded data")
                lims = np.concatenate(
                    [np.arange(0, len(self.times), 10000), [len(self.times)]]
                )
                for start, stop in zip(lims[:-1], lims[1:]):
                    self._data[:, start:stop] = np.dot(comp, self._data[:, start:stop])
            else:
                self._comp = comp  # store it for later use
        return self

    @property
    def _dtype(self):