# Continuous EEG: Where does pronounced power peak at 38 Hz come from?

I'm pretty new to EEG analysis. I measured EEG on six scalp positions over night. When I plot the power of the frequencies, I get the following:

What I want to explain (and didn't manage to) is the origin of the 38 Hz peak in my case.

The most important steps to arrive at my graph were (using Python-MNE):

1. band-pass filter of data (1 to 45 Hz)
2. resampling from 512 Hz to 200 Hz

Then for every of the six channels:

1. cut recording into consecutive segments of 2 sec duration

exclude segments if:

1. maximum allowed voltage step of 50 µV is exceeded
2. activity is lower than 0.5 µV
3. maximum and minimum amplitude exceed +/- 200 µV
4. maximum absolute difference of values in the segment > 200 µV

further:

1. retain only segments that are ok for all electrodes
2. randomly choose 15 such good segments (all electrodes) from every participant
3. calculate FFT for every 2 sec time segment for every electrode and participant
4. calculate mean of all these FFTs
5. plot

My question is: is someone familiar with such an artifact at 38 Hz? Where could it come from? These are the things I already checked:

• it is pronounced in every participant
• it is pronounced in every electrode
• eye movements are unlikely since this peak is there after filtering out horizontal eye movements via ICA
• electrical devises are unlikely since they should peak at 50 Hz
• it is pronounced when segments are 6 sec instead of 2 sec
• I also used a notch filter at 50 Hz instead of the bandpass filter, with the following result (here the individual 15 segments are shown, but the 38 Hz peak is still there):

Does someone have an idea where this feature could come from?