Apologies out of the gate here as I'm a CS person with not a huge amount of experience in electronics, so this might be a pretty basic question. I've built basic EEGs before, basically electrodes together with a series of active filters.

My approach in the past has always been to play with my filters until I don't see a peak at 60Hz, and I don't see huge swings when someone moves their eyes. Once that's done there's usually some kind of wave that gets spit out, but I've never really been convinced that I was seeing a real signal, and not just picking up on some other random noise, which is extremely likely.

I'm now a bit older and more foolish and I wanted to take a more intentional stab at this, but I was wondering if there's any reasonable way of testing this. My naive approach was going to be attempting to generate a signal in the nV realm as input, then trying to re-create it after processing. There are two main reasons why I think this doesn't make sense:

  1. Electrodes/Skin conduct introduce noise that I can't model this way
  2. I'm very unclear on exactly why an EEG works. It feels very strange to me that you'd get actual signal by strapping electrodes to your head

So how exactly do I test an EEG, and have confidence that an EEG I create is working?


Aside from generating a signal of a few 10's of microvolts to run through your amps to make sure they're working and verify your frequency cutoffs, you should be able to see clear differences between EEG with eyes open and eyes closed. Surface EEG should be open from about 0.1 Hz to 100 Hz, so if you absolutely must have a 60Hz filter, make sure the notch is tight. You also expect to see blink artifacts.

See the relevant figure in http://iopscience.iop.org/article/10.1088/1741-2560/10/3/036006, where the author is validating an electrode. The eyes are closed at 20 seconds.

enter image description here

Lastly, when you build an amplifier for EEG, you will certainly need something with a very high common mode rejection ratio, which is very difficult if you're using only op amps.

For MUCH of the homemade Brain-Machine-Interface stuff, I'm almost convinced that much of what one sees is actually due to surface EMG of the scalp and underlying muscles, or some sort of modulated motion artifact from the electrodes. (In fact, once, many moons ago, the Society for Neuroscience sent out an advertising postcard with a photo of me arguing just this point in front of an author's poster on it. The community, of course, has come a long way in making sure these artifacts aren't poisoning the data -- but that's professional systems, and not homemade!).

In short, you're right to approach this with skepticism.

  • \$\begingroup\$ This is perfect! Exactly what I was looking for. The auditory and visual signaling they use in the paper can help me get into the right ballpark and then I can shift to more practical metrics. I realize that op-amps aren't really going to work here, but other than custom ICs I'm not really sure what else I should be looking at. Should probably be a different question, but if you've got a recommendation it would be much appreciated. \$\endgroup\$ – Slater Victoroff Mar 1 '17 at 18:07
  • 1
    \$\begingroup\$ @SlaterTyranus You should start with a first stage of a modest gain instrumentation amplifier able to handle about 200mV of offset from the electrodes without saturating, filter out any DC offsets, then continue to amplify while removing bias and noise in stages \$\endgroup\$ – Scott Seidman Mar 1 '17 at 18:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.