Noise at Nyquist Frequency (125Hz) [resolved]
Hi! I've been experimenting with independent referencing with tests using 8 channels at fs of 250Hz. The first 6 channels are EEG, 7 is empty, and 8 is EMG. When I plotted the raw data in Matlab, I wasn't able to see any different frequencies because of large amplitude 125Hz noise on all the channels. The amplitude does vary between channels (although all still abnormally large) with the empty channel 7 having the largest amp of 89200 microvolts and the emg the smallest of about 500 microvolts. I checked back on the GUI to see the same thing when I turned off smoothing and filters. I did not have this problem in previous tests. What could be causing this/is there a way to fix this?
Comments
JTBME, hi.
Can you clarify? You are only seeing this issue in Matlab, not the GUI?
Unused channels should be switched off, as the values are noise and unpredictable.
EEG is normally filtered first, for example with a bandpass from say .5 Hz to 40 Hz. GUI lets you do this. EEG above 40Hz is very low amplitude and frequently contaminated by scalp muscle EMG noise. You cannot record 125 Hz EEG from the scalp, and samples at the Nyquist frequency (125 Hz) do not make any sense because of aliasing effects. You generally want 4 to 5 samples per waveform: 250/5 = 50 Hz max. If only 2 samples are made per waveform (250/2 = 125Hz), huge aliasing errors are induced when the sample times shift from peaks to null points in the signal. Does that make sense?
Your EMG channel should be disconnected from the SRB2 bus (using the Settings panel), and connected differentially with both plus and minus pins.
Regards, William
I am seeing the same issue in the GUI. Channel 7 was turned off, but the raw data txt file still includes it and its high amplitude and similar noise to the rest of the channels made me wonder if it was a part of the problem. Could be something in the environment? When I filter the data, the noise does "go away", but I am worried about the aliasing effects as you've said, and that I am just creating a signal that really isn't there. The filtering also causes a large jump in the beginning and end of the feature that makes me have to zoom in excessively to see it. In past tests, there's been some typical 60 Hz noise, but I was still able to see other frequencies behind it before filtering. The EMG was connected as you've said.
Thank you,
JTBME
Yes, it's possible that EMF environmental noise could be contributing to the 125 Hz artifacts you are seeing. Try moving your setup other rooms. Stay away from other electrical equipment, power supplies, power cords, conduits (wall, ceiling, floor), etc.
No, my reference to 'aliasing' was only to explain that sampling waveforms with only two samples per cycle, is inherently full of alias effects. You are not 'creating' bad signals. Filtering removes all this spurious junk.
William
Hi, just wanted to let you know that the "noise" had to do with the # of channels and whether or not the daisy board was connected. Whenever I used 8 channels on the gui but had the daisy connected, the raw data had that 125 hz spike on the fft plot. When I disconnected the daisy, this went away. I further checked with reconnecting the daisy and specifying 16 channels on the gui and the spike also disappeared. Why does this happen? Thank you so much for your help and patience!
JT, hi. Thanks for your detective work on this.
It's possible that some of the data collection / signal processing done by Brainflow or the GUI, misbehaves if Daisy is connected but only 8 channels are being displayed. I did not realize the GUI could be configured like this. Obviously the short term solution is to just not do this; and to configure the GUI for the proper number of channels.
Mentioning Richard @retiutut.
William