[Eeglablist] Persistent 20 Hz Harmonic Line Noise in EEG Data

[Scott Makeig]

Maeghal -

First, I would examine the raw data to see whether the 25-Hz signal appears
throughout or not.  If so, I would suggest that after removing any
truly 'bad' ('going their own way') channels and performing average
referencing, you perform an ICA decomposition using AMICA or infomax
(RUNICA).  Then examine the time courses of the resulting independent
components (ICs) (from the EEGLAB GUI) to see if you find the one or more
ICs with a 25-Hz train of sharp waves. This noise is  not line noise (at 50
or 60 Hz), but comes from some other nearby electronic source - possibly a
nearby computer screen(?). Unfortunately, participant head movements may
change the receptivity ('scalp map') of the induced 25-Hz signal in the EEG
channels, in which case ICA cannot find a single IC (with a static scalp
projection map) that accounts for all of the 25-Hz 'artifact' (or,
non-brain source activity) in the data. In best/easiest cases, one IC will
separate out all or most of this source data. Then, simply removing the
IC's scalp-projected activity from the scalp data will clean the data of
the unwanted signal. If this is not the case, then consider whether the
event-related or other measure(s) you intend to use to model the data will
be affected by the remaining 25-Hz signal - or not.

Scott Makeig

On Fri, May 8, 2026 at 1:58 PM Maeghal Jain via eeglablist <
eeglablist at sccn.ucsd.edu> wrote:

> Dear EEGLAB List,
>
> I am currently dealing with persistent line-noise contamination in my EEG
> recordings, specifically strong peaks around 20 Hz and its harmonics (40
> Hz, 60 Hz, etc.). The contamination remains substantial even after multiple
> preprocessing attempts.
>
> Raw Data PSD for reference:
> https://urldefense.com/v3/__https://ibb.co/5WbNqYCW__;!!Mih3wA!EfdhaFV_tpprT4Wc8_HcfK7KDVTXA0O0NPLW3Ws732nxP-TmPxHUXI_qXevcPIO8hfndJk8euI6BWbB1_dCuBxHG_A$
>
> I have already tried:
>
>    -
>
>    Average re-referencing before notch filtering
>    -
>
>    Average re-referencing after notch filtering
>    -
>
>    Multiple full-rank average reference implementations, including:
>    -
>
>       Makoto’s fullRankAveRef()
>       -
>
>       Manual full-rank average reference using (nchan + 1) denominator
>
> Despite this, a large amount of the harmonic noise is still present in the
> spectra.
>
> For reference, the issue looks somewhat similar to this discussion:
>
> https://urldefense.com/v3/__https://mne.discourse.group/t/line-noise-eeg/11817__;!!Mih3wA!EfdhaFV_tpprT4Wc8_HcfK7KDVTXA0O0NPLW3Ws732nxP-TmPxHUXI_qXevcPIO8hfndJk8euI6BWbB1_dB2ZPVDPg$
>
> I would greatly appreciate advice on how people in the EEGLAB community
> typically handle this kind of persistent harmonic contamination.
>
> Would you recommend CleanLine, Zapline-plus, spectrum interpolation, ASR,
> ICA-based approaches, or something else?
>
> Thank you very much for your help.
>
> Best regards,
> Maeghal Jain
> BISE Lab
> _______________________________________________
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu or visit
> https://sccn.ucsd.edu/mailman/listinfo/eeglablist .



-- 
Scott Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation, University of
California San Diego, La Jolla CA 92093-0559, http://sccn.ucsd.edu/~scott