When applying SSS and movement compensation, the MNE maxwell_filter implementation (v1.11) introduces broadband noise into the spectra which is not seen using (I think) the equivalent MaxFilter command. Is there a difference in the Maxwell-filter implementation that would account for this?
MNE code:
import mne
import matplotlib.pyplot as plt
raw = mne.io.read_raw('test_raw.fif')
fig = raw.compute_psd().plot()
fig.savefig('psd_raw.png')
plt.close()
amp = mne.chpi.compute_chpi_amplitudes(raw)
locs = mne.chpi.compute_chpi_locs(raw.info, amp)
pos = mne.chpi.compute_head_pos(raw.info, locs)
bads_noisy, bads_flat = mne.preprocessing.find_bad_channels_maxwell(raw)
raw.info['bads'] = bads_noisy + bads_flat
raw_sss = mne.preprocessing.maxwell_filter(raw, origin='auto', coord_frame='head', head_pos=pos)
raw_sss.save('mne_sss.fif', overwrite=True)
fig = raw_sss.compute_psd().plot()
fig.savefig('psd_maxwell_filter.png')
plt.close()
megin_sss = mne.io.read_raw('megin_sss.fif')
fig = megin_sss.compute_psd().plot()
fig.savefig('psd_megin_sss.png')
plt.close()
MaxFilter command:
/neuro/bin/util/maxfilter \
-f /neuro/data/sinuhe/test_raw.fif \
-o /neuro/data/sinuhe/megin_sss.fif \
-movecomp \
-autobad on \
-frame head \
-origin fit \
-hp /neuro/data/sinuhe/test_hp.pos -v > /neuro/data/sinuhe/test_sss.log 2>&1
Sytem information
Platform macOS-26.2-arm64-arm-64bit-Mach-O
Python 3.14.3 | packaged by conda-forge | (main, Feb 9 2026, 22:09:14) [Clang 20.1.8 ]
Executable /Users/rh687/miniconda3/envs/mne/bin/python
CPU Apple M2 (8 cores)
Memory 24.0 GiB
Core
├☑ mne 1.11.0 (latest release)
├☑ numpy 2.4.2 (unknown linalg bindings (threadpoolctl module not found: No module named 'threadpoolctl'))
├☑ scipy 1.17.0
└☑ matplotlib 3.10.8 (backend=macosx)