pub const FIT_PPM_COMPLEX_SPACING_TOLERANCE_S: f64 = 2e-5;Expand description
How far, in seconds, each echo spacing may be from the first for fit_ppm_complex and
MEDI’s noise map (fit_ppm_complex_noise) to count the echoes as evenly spaced: 0.02 ms.
Fit_ppm_complex.m models the phase in echo index and takes no echo times, so it has no
tolerance of its own; the check is QSM.rs’s, to refuse trains the model does not describe.
The tolerance is absolute and set by how echo times are stored, not by the spacing: BIDS
sidecars and DICOM headers round them, typically to 0.01 ms, so each spacing can be off by
0.01 ms and two spacings of an evenly spaced train can differ by 0.02 ms (for example
[4.92, 9.84, 14.77, 19.68] ms, spacings [4.92, 4.93, 4.91] ms). A relative 1e-3 of the
spacing, used before, is 0.005 ms on a 5 ms train and refused such sidecars. A train that
really is uneven (a different first interval, a missing echo) is off by far more.