noisy waveform in EEGLAB? [resolved]
jctaylor
Canada
Hi folks,
I'm hoping someone can help me with this. I'm using a cyton+daisy to try to record through LSL and Lab Recorder. I'll describe my typical procedure, but I've monkeyed around with variations without success so far.
Normally, I start the OpenBCI GUI and then plug in the dongle and turn on the headset to PC. I then start the GUI and start streaming in order to insure that no channels are railed. I have tried using the LSL stream from the GUI, but EEGLAB threw a bit of a fit when I tried to open that file. So I run openbci_lsl from an Anaconda (Python) window. I run LabRecorder.exe from another Anaconda window and save an .xdf file.
The way things look when I open the .xdf file using either the xdfimport or the Mobilab matlab extensions is as shown below:
The way the channels look in the GUI is more normal (to my untrained eye)
Does anyone have any clue what I'm doing wrong or how I might go about problem-solving this? At first I thought it was related to my marker stream from a psychopy program, but it happens even without that running (which is what I described above). I've messed around without success. Any help would be greatly appreciated.
Comments
JC, hi.
Your EEGLAB graph shows huge amounts of mains noise (60 or 50 Hz). This is normally filtered out by the OpenBCI_GUI, but you need to tell EEGLAB to do similar. The GUI uses a 'notch' filter at (60 Hz in your screenshot). The data recordings or stream from the GUI is NOT filtered, it is raw.
William
William,
Thank you very much for your comment! I wondered about wall current, but the blobs only appeared about 5-6 per second so thought it must be something else. I also thought that with the headset working off battery, wall current wouldn't be an issue. But I'm not surprised I am wrong about that. I will be back in the lab later today and will try out your suggestion. Thanks very much for taking your time on this!
Jamie
If you turn off the filtering in the GUI, you'll see the same large amounts of mains noise. Check the FFT window, it will show a large spike at the mains freq.
Hi William,
You are right - I see a prominent spike around 60 Hz in the FFT of the data. Filtering to the 0.1Hz - 30 Hz range yielded a much more reasonable set of channel plots. Now I can start sampling a longer series and try creating averaged ERPs. Many thanks for your help. I knew about mains noise, but I didn't realize it would be such a prominent issue in a battery-powered EEG headset, so I didn't imagine that could be the problem.
best
Jamie
Mains noise is ubiquitous in the environment. There are inexpensive EMF meters that can demonstrate this. Even some phone apps if your phone has a magnetometer.
https://www.google.com/search?q=emf+meter+app