OTD ACADEMY
Sign up free
Library

LIBRARYHow to Build an EEG Brain-Computer Interface

A first-principles roadmap to a real DIY EEG brain-computer interface: from what the signal is, through the electronics that capture it, to turning brainwaves into a command.

A brain-computer interface (BCI) turns measured brain activity into a command, with no muscles in the loop (Wolpaw et al., 2002). The most accessible non-invasive way in is the EEG: electrodes on the scalp that read the brain's electrical activity. This guide is the map for building a real one from first principles, and it is honest about what that does and does not mean. An EEG BCI is not mind-reading. It is pattern detection on very small, very noisy signals: microvolts, buried in interference. Everything hard about building one follows from that single fact.

The five-stage EEG BCI pipeline

Every EEG BCI is the same five-stage pipeline. Electrodes pick up microvolt voltages on the scalp. An analog front-end amplifies them and rejects the noise they are drowning in. An analog-to-digital converter turns them into numbers. Signal processing filters the data and extracts features (which rhythms, which electrodes). A classifier maps those features to a small set of intents, which become commands. Master each stage and the whole thing works; skimp on one, usually noise, and nothing downstream can recover it.

Safety first: isolation is non-negotiable

NEVER connect a mains-powered rig (a USB device plugged into a wall charger, a bench supply) directly to electrodes on a person. A fault could drive mains current through the body. Power the front-end from a battery, or use a properly isolated supply. This is not optional caution. It is the one rule that keeps DIY EEG safe.

StageWhat it doesThe hard part
ElectrodesPick up µV-scale scalp voltagesContact impedance, placement (10-20 system)
Analog front-endAmplify µV signals, reject common-mode noiseNoise floor + 50/60 Hz mains rejection
ADCDigitize each channelResolution + simultaneous sampling
Signal processingFilter + extract features (bands, spatial patterns)Separating signal from artifact
ClassifierMap features to intent to commandLow, noisy bandwidth; per-session variability

How to build one, step by step

  1. Understand the signal: what EEG is, and what it isn't.
  2. Place electrodes: the 10-20 system, wet vs dry, the reference and bias electrodes.
  3. Build the analog front-end: amplify microvolts without amplifying noise (this is where the ADS1299 comes in).
  4. Kill the noise: 50/60 Hz mains, the right-leg-drive / bias loop, shielding.
  5. Read the rhythms: delta/theta/alpha/beta/gamma and what they mean.
  6. Pick a control paradigm: motor imagery, SSVEP, or attention.
  7. Turn signal into a command: features (CSP) and classifiers (EEGNet).
  8. Control something real: close the loop on a robot or a drone.
Deep dive· Go deeper: why scalp EEG is so hard

The brain's electrical activity has to cross the cortex, skull, and scalp before it reaches an electrode, a process called volume conduction that smears and attenuates it down to microvolts. Meanwhile the body acts as an antenna for 50/60 Hz mains hum that is orders of magnitude larger, and every muscle twitch, eye blink, and cable movement adds artifact. So the whole front-end is a fight for signal-to-noise: high common-mode rejection, a low noise floor, a clean reference, and an active bias/right-leg-drive loop that cancels the mains interference before it ever reaches the ADC.

Each stage explained

References

Ready to build the hardware instead of just reading about it? The OTD Academy EEG front-end project walks the full design, schematic to bring-up, of a real multi-channel front-end: the board at the center of this whole path.

One Thousand Drones Academy · reviewed June 2026

Coming soon

8-Channel EEG Front-End on ESP32

Design the analog board that reads real brainwaves: the BCI.