Example: the first time alpha showed up
An example build log written by the site's author to show the shape. Not the learner's work. The failures in it are the typical ones.
What I set out to do
Project B in Phase 1: one channel of EEG with an INA128, a driven-right-leg circuit, and a 0.5 to 30 Hz band-pass, into an Arduino Nano and the Web Serial plotter. The goal was the eyes-closed test: a bump at 10 Hz that disappears when the eyes open.
What I actually did
Breadboarded the INA128 with a 500 Ω gain resistor for a gain of about 100, then a second stage with a TL072 for another 20×, total about 2000. Two 9 V batteries for a ±9 V supply, mid-point tied to the Arduino’s analog reference through a 10 kΩ / 10 kΩ divider so the output sat at 2.5 V. Gold cups with Ten20 paste at O1 (back left of the head) and the reference on the left earlobe. DRL electrode on the forehead.
First trace: a 60 Hz sine wave, several volts peak to peak at the output, everything else invisible.
What went wrong
- The laptop was plugged in. The Arduino’s ground was the laptop’s ground was the wall. Unplugging the charger cut the hum by about ten times. This is also the safety rule, which I had read and then ignored because the amplifier was “on batteries.” The Arduino was not.
- Gain resistor too small for the first stage. With 2000× and a 50 mV electrode offset, the second stage saturated. Adding a 0.5 Hz high-pass between stages (1 µF and 330 kΩ) fixed it.
- The reference cup dried out. After twenty minutes the trace got noisy and jumped when I moved. More paste, and pressing the cup down, brought it back.
- I looked for alpha at the front of the head first. Nothing. Moving the electrode to O1 made it obvious within a minute. The explainer says why; I had not read it yet, which was the point.
What I measured
With the input shorted through a 10 kΩ resistor, the output noise was about 30 mV RMS at a gain of 2000, so an input-referred noise of roughly 15 µV RMS over 0.5 to 30 Hz. That is high (a good system is under 1 µV) and most of it was the Arduino’s ADC and the breadboard. Alpha, eyes closed, was around 20 to 40 µV peak, so visible above the noise but not by a lot.
The spectrum showed a peak at 9.8 Hz that appeared within two seconds of closing my eyes and vanished within one second of opening them. I saved the plot; it is the artifact for this project.
What I would change
Solder it. The breadboard was the noise. Add a proper 60 Hz-free power path for the Arduino (a USB battery pack). Put the DRL electrode on the other earlobe instead of the forehead so blinks do not drive it.
Things that surprised me
- How much of the signal was not brain. Blinks were ten times the size of alpha.
- The DRL made a bigger difference than the gain did.
- Seeing my own alpha appear on command does not stop being strange.