Hi everyone,
I’m not sure if this is the right place to post, but I’d like to ask if anyone has experienced a similar issue.
Recently, I recorded EEG data using the BrainVision system with a 64-channel EEG cap.
The preparation included measuring head circumference, applying exfoliating gel and alcohol to clean the scalp, and using conductive gel. I ensured that all electrode impedances were below 5 kΩ (with ground and reference ≤ 1 kΩ). At the beginning of the recording, everything looked fine—the signals appeared normal without any obvious artifacts.
However, about halfway through the experiment, I began to notice something unusual. A strange electrical noise started appearing in the data, and it gradually became stronger over time. What’s even stranger is in each block I checked the impedances page and found out all the electrodes still below 5 kΩ (with ground and reference ≤ 1 kΩ)
This clearly shows that line noise at 60Hz and harmonics increased dramatically in the second half. Which frequency range are you interested in? If you focus on something like 1 to 35Hz, you could filter the signal and then compare the time courses as well as PSDs again.
It’s a bit weird that this would also show in bipolar montages (as your variable name suggests), but it happens. I would suggest to inspect the average-referenced data first before continuing with your analysis.
Ah sorry about the confusing variable name. raw_bipolar here simply refers to subtracting the EOG electrodes (e.g., LEOG – REOG, and so on).
As for the 60 Hz artifact, I suspect it’s related to DC offset. But that comes a question—why does it become so much stronger in the latter part of the recording? After all, the amplifier is connected to a power supply (battery). During experiments, we unplug the charge of it. But I think doing an A/B test with the power supply might be worth exploring (in the setup, two amplifiers share one power supply, so I’m not sure if that could have an effect).
On the other hand, you mentioned about the band pass filter and the PSD of it. I’m not entirely sure whether this type of noise/artifact can really be resolved with a band-pass filter……
Here is the part of the data in raw.plot() and its creepy…. The data point is ~2000.
Regarding the two PSDs, it would be helpful if you zoomed into the filtered range so that we can see the frequency range of interest more clearly.
The 60Hz artifact should not be related to the DC offset – can you explain your thinking? Also, I would always run the amplifier on battery without any connection to the grid, so yes, make sure to unplug the charger.
However, even when running on battery you will pick up line noise through your cables. It is also perfectly normal that the amount of interference changes over time, which would explain the different PSDs. You could also look for devices that might interfere, e.g., did you turn on/off the lights in the middle of the measurement? Or some other electronic device? The point is, line noise will almost always be present, but usually it does not really get in the way of your analysis (if for example you focus on 1–35Hz anyway).
Regarding the time course you are showing, maybe the scaling is just off? Can you try zooming (press the +/- keys multiple times) until you get nice signals?
Finally, the peaks you are showing are harmonics of the 60Hz power line signal. The smaller ones are aliased back into the 0–500Hz range because you sample at 1000Hz.
Hi hi~
Sorry for the late reply, things have been too busy lately🥲
Regarding the DC offset, you’re absolutely right. I searched it again, and DC offset can indeed be any constant voltage shift. Even if it changes over time, the variation should be quite slow. That means the 60 Hz component is most likely power-line noise.
Although you mentioned that a BPF could remove such noise, I’m concerned that if the data contain strong high-frequency oscillations that exceed the recording voltage range, it could cause clipping, making those epochs unusable.
A few days ago, I tried recording again using a 32-channel cap, with no extra electronic devices turned on, and the cable setup was exactly the same as before. Interestingly, I didn’t observe the same issue this time. For safety, I replaced both the battery and the amplifier, and after changing the hardware, the problem didn’t appear again. I’m guessing I’ll need to borrow EEG equipment from other labs and run a series of A/B tests to find out which component was causing the issue……
Thank you so much not only for your detailed reply but also for patiently teaching me skills I hadn’t learned before! XDD
