breaking change: functions operating on arrays that also operate on NEURO objects are now private (_*) and available via NeuroAnalyzer._*
breaking change: load and save were renamed to load_neuro and save_neuro
breaking change: t2s now returns round(t·fs) + 1, so time 0 is sample 1 and t2s/s2t are exact inverses; results are one sample higher than before (e.g. t2s(1.0, 256) is now 257, was 256). s2t no longer rounds to 4 decimals; for durations use the new t2n/n2t: a 10 s window is t2s(t0, fs):(t2s(t0, fs) + t2n(10, fs) - 1)
breaking change: internally channel positions and time points are stored as exact values; rounding to 2 decimal places is done for presentation only
analyze: fooof: parameterize neural power spectra into aperiodic and periodic components
plot: plot: epochs can be manually selected and deleted
plot: plot: improved segment deletion for continuous signal
plot: ch_locs arguments for plot_psd, plot_phs, plot_erp, plot_gfp: if true, draws a head outline with the locations of the plotted channels (position given in ch_locs_pos argument)
plot: legend in multichannel plots is placed outside the plot
plot: axis is now properly log10-scaled
plot: axes are unified across similar plot types
plot: improved head outlines
plot: gain can be changed while vieving the signal plot
plot: plot_env, plot_ispc, plot_pli, plot_itpc, plot_itpc_spec, plot_itpc_frq moved from plugins
process: sphinterpolate_channel: interpolate EEG channel(s) using spherical splines
process: remove_powerline is now >50× faster
process: apply_ssp_projectors, generate_ssp_projectors: projector built in SI units and converted to data units
process: maxwell_filter: Maxwell filtering (SSS), temporal SSS (tSSS) and head movement compensation for Neuromag/MEGIN Vectorview/TRIUX data; information-based regularization, fine calibration and cross-talk correction
process: chpi_head_positions: continuous head position estimation from cHPI signals
process: read_fine_calibration, read_cross_talk: read sss_cal.dat and ct_sparse.fif
process: reference_custom modifies location of bipolar montage channels
io: better handling of non-integer sampling rates
io: improved labels cleaning
io: import_recording: improved automatic file format detection
io: import_fiff: updated FIFF importer:
allow_maxshield (IAS/MaxShield raw data refused by default) and channel_locs (:fiff, :layout, :none) arguments
channel locations taken from the FIFF file by default (fallback: bundled 306-channel layout)
channel labels are generated from FIFF channel names (including MNE long names)
EEG-only FIFF files are imported as EEG recordings
calibration raw × range × cal (unit multiplier no longer applied, as in MNE-Python)
import_locs_fiff: channel locations are imported from FIFF files
FFTW now uses single thread, which is vastly faster when processing channels in separate
You mention that the Raspberry Pi is supported and that a requirement is OpenGL 3.3 or later, which the PI does not have. How do you work around these conflicting requirements?
I haven’t touched an RPi for long time and many things were changed in NA meanwhile (switch from Plots.jl to GLMakie.jl among others). OpenGL is required for GLMakie visualizations, forcing software rendering should help: LIBGL_ALWAYS_SOFTWARE=1 MESA_GL_VERSION_OVERRIDE=3.3 julia