Artifact of unknown origin in MEG data, turns out to be the light camera

[schoffelen]

Describe the issue
MEG data seems to suffer from a 20 Hz artifact (+harmonics), already for a while now. This should be investigated and fixed.

Describe yourself

1. Jan-Mathijs Schoffelen
2. Staff scientist
3. MR physics

Audience
Everybody using MEG.

Test data
Any recently recorded dataset.

To Reproduce
Steps to reproduce the behavior:

1. Read in the data
2. Chunk it in segments, e.g. of 5 seconds
3. Compute power spectra
4. Visualize the spectra

Expected behavior
We would expect well-behaved spectra from emptyroom measurements, with only spectral lines at 50 + harmonics, not at 20 + harmonics as well

Screenshots

Environment and versions (please complete the following information):

• n.a.

Additional context
JM teams up with Uriel and Robert to address this. This includes identification of the cause, fixing the cause, and documenting what can (and needs to) be done to clean existing data.

[schoffelen]

Update: as per some investigations by Uriël and myself, we have identified the source of the artifact to come from the light camera (https://github.com/Donders-Institute/meg_quality/blob/master/meg_quality_artifact20Hz_20211021.m), which is a device that is operated by a raspberry pi, which is in the MSR. Looking back in time, it seems that the artifact has been introduced somewhere in july 2018 (https://github.com/Donders-Institute/meg_quality/blob/master/meg_quality_artifact20Hz_history.m).

[schoffelen]

I will write and document some example code that can be used to alleviate the artifact, demonstrating:
-how to do 3d order gradient balancing (->ft_denoise_synthetic)
-how to do adaptive balancing (->ft_denoise_pca)