Build the instrument
Design, build, and characterize a research-grade EEG amplifier.
In Phase 1 you used an amplifier. Here you build the whole signal chain around the same chip that sits inside research systems, and then you measure what it actually does. The difference between "I made an EEG" and "I characterized an EEG" is the difference between a hobbyist and an engineer.
Phase 1 was about a memory. Phase 2 is about a number. By the end of it you will be able to say what your instrument’s noise floorNoise floor What your instrument records when the input is shorted or connected to a perfectly quiet source; everything smaller than this is invisible to you. Related: Thermal (Johnson) noiseThermal (Johnson) noise The random voltage every resistor produces just by being warm; about 13 nanovolts per root hertz for 10 kilohms, which matters at microvolt levels. , CharacterizationCharacterization Measuring what your instrument actually does, such as its noise floor, bandwidth, and rejection, instead of assuming the datasheet. . Glossary entry is, how much hum it rejects, and where its bandwidth ends, and you will have measured those things rather than read them off a datasheetDatasheet The manufacturer's document for a chip: pinout, limits, typical performance, and the reference circuit you should start from. Glossary entry.
There are two routes to the same place. Building the ADS1299ADS1299 Texas Instruments' 8-channel, 24-bit biopotential front-end chip used in OpenBCI and many research amplifiers. Related: Delta-sigma ADCDelta-sigma ADC A converter that samples very fast with a crude 1-bit decision and averages, achieving 24-bit resolution at modest rates; what the ADS1299 uses. , OpenBCIOpenBCI The open-source EEG hardware company whose Cyton (ADS1299) and Ganglion boards are the default hobby and research-lite amplifiers. . Glossary entry board from a KiCadKiCad Free, open-source software for drawing schematics and laying out printed circuit boards. Related: Printed circuit board (PCB)Printed circuit board (PCB) The board that holds and connects your components; designing one is how a breadboard prototype becomes an instrument. , JLCPCBJLCPCB A Chinese board house that fabricates and assembles small-run PCBs for a few dollars, the standard for student projects. . Glossary entry schematic through JLCPCBJLCPCB A Chinese board house that fabricates and assembles small-run PCBs for a few dollars, the standard for student projects. Related: Printed circuit board (PCB)Printed circuit board (PCB) The board that holds and connects your components; designing one is how a breadboard prototype becomes an instrument. . Glossary entry assembly to working firmwareFirmware The program running on the microcontroller, which configures the chip, reads samples on time, and ships them to the computer. Related: SPISPI A simple four-wire bus for talking to chips like the ADS1299 from a microcontroller. . Glossary entry teaches the most and takes a semester of evenings. Buying an OpenBCIOpenBCI The open-source EEG hardware company whose Cyton (ADS1299) and Ganglion boards are the default hobby and research-lite amplifiers. Related: ADS1299ADS1299 Texas Instruments' 8-channel, 24-bit biopotential front-end chip used in OpenBCI and many research amplifiers. , BrainFlowBrainFlow A library that talks to OpenBCI and most consumer EEG headsets with one API, in Python and other languages. . Glossary entry Ganglion and reverse-engineering it while you use it gets you to decoding a month sooner. Either way, the phantomPhantom A physical stand-in for tissue, such as saline gelatin with embedded wires, used to test an instrument against known signals. Related: CharacterizationCharacterization Measuring what your instrument actually does, such as its noise floor, bandwidth, and rejection, instead of assuming the datasheet. . Glossary entry head and the characterizationCharacterization Measuring what your instrument actually does, such as its noise floor, bandwidth, and rejection, instead of assuming the datasheet. Related: Noise floorNoise floor What your instrument records when the input is shorted or connected to a perfectly quiet source; everything smaller than this is invisible to you. , PhantomPhantom A physical stand-in for tissue, such as saline gelatin with embedded wires, used to test an instrument against known signals. , Common-mode rejection ratio (CMRR)Common-mode rejection ratio (CMRR) How much better an amplifier responds to a difference between its inputs than to a signal shared by both, in decibels; 100 dB means 100,000 times. . Glossary entry report are not optional. They are the phase.
What attaches here
The theory in this phase is the theory of instruments: instrumentation amplifiersInstrumentation amplifier A precision differential amplifier with very high input impedance and high CMRR, built from three op-amps and sold as a single chip such as the INA128 or AD8232. Related: Differential measurementDifferential measurement Measuring the difference between two electrodes rather than one electrode against ground, so anything common to both cancels. , Common-mode rejection ratio (CMRR)Common-mode rejection ratio (CMRR) How much better an amplifier responds to a difference between its inputs than to a signal shared by both, in decibels; 100 dB means 100,000 times. , Input impedanceInput impedance How hard the amplifier's inputs are to push current into; it must be far higher than the electrode impedance or the signal is lost in the divider. . Glossary entry, the electrodeElectrode The conductor that turns ion currents in tissue into electron currents in a wire; its chemistry and contact quality set the noise you live with. Related: Silver/silver-chloride (Ag/AgCl)Silver/silver-chloride (Ag/AgCl) The standard electrode material for biopotentials, because its chemistry produces a stable, low-noise, low-drift interface with salty gel. , Electrode impedanceElectrode impedance How hard it is to push current through the skin-electrode contact, typically 5 to 50 kilohms with gel and megohms dry, higher at low frequencies. , Half-cell potentialHalf-cell potential The DC voltage that appears wherever a metal meets an electrolyte; two electrodes with different half-cell potentials produce an offset that can swamp the signal. . Glossary entry-electrolyte interface, analog-to-digital conversion, noise, electrical safety, digital filtering, artifactArtifact Anything in the recording that did not come from the brain: blinks, muscle, heartbeat, movement, sweat, mains hum, electrode pops. Related: Blink artifactBlink artifact The large, slow deflection an eye blink produces in frontal channels, often ten times the size of the brain signal. , Independent component analysis (ICA)Independent component analysis (ICA) A method that unmixes multichannel data into statistically independent sources, so blinks and heartbeat can be identified and removed. , Artifact subspace reconstruction (ASR)Artifact subspace reconstruction (ASR) An automatic method that detects short high-variance bursts and reconstructs them from clean data; common in real-time pipelines. . Glossary entry removal, and spatial referencing. Every one of these explainers opens with something you will have just seen on your bench.
Projects
Choose one. Do a second if hooked.Project A: The ADS1299 board
Eight channels of 24-bit EEG around the chip inside most research and consumer systems. Schematic in KiCad, fabricated and assembled by JLCPCB, firmware on an ESP32, streaming into BrainFlow. A semester of evenings, and the whole signal chain becomes yours.
ProjectProject B: Reverse-engineer a bought board
If you buy an OpenBCI Ganglion or Cyton instead of building, read its schematic, trace the signal chain, read its firmware, and characterize it as if it were yours. Faster to decoding, nearly as educational, and no longer the cheaper path.
ProjectProject C: Make your own electrodes
Silver/silver-chloride by chloriding a silver wire, dry electrodes from conductive silicone or printed pins, and an active electrode with a buffer on it. Then measure the impedance of each versus frequency and find out why gel exists.
ProjectProject D: The phantom head
Saline gelatin with buried wire dipoles driven at known amplitudes. Now you can measure your amplifier's noise floor, crosstalk, and rejection instead of guessing. The difference between 'I made an EEG' and 'I characterized an EEG.'
ProjectProject E: The eight-channel EMG armband
Eight electrodes around the forearm, a classifier, and hand gestures recognized from muscle activity. The wristband idea, and the most dependable 'control something with your nervous system' demo there is.
ProjectProject F: The software pipeline
From a raw file to cleaned epochs to a topographic map, in MNE-Python, on public data. Build the pipeline before you have your own data so you can trust it when you do.
Explainers
Theory that attaches to what you just built. Read after doing.Instrumentation amplifiers, properly
What is inside the chip you have been trusting. Three op-amps, four matched resistors, and the numbers that decide whether the board records brain or room.
ExplainerThe electrode-electrolyte interface
What happens where metal meets salty skin. Half-cell potentials, why silver chloride is stable, why impedance depends on frequency, and why gel takes half an hour and dry electrodes are a compromise.
ExplainerAnalog to digital
Sampling, aliasing, quantization, and why the ADS1299 gets 24 bits out of a one-bit decision made very fast. The four ideas you need to read any converter's datasheet.
ExplainerNoise, and where the microvolt goes
Thermal, 1/f, amplifier, and electrode noise, added up for a real EEG channel. Why the electrode dominates, why the breadboard was so bad, and how to read a noise floor.
ExplainerElectrical safety, with the numbers
What IEC 60601 actually limits, why a battery satisfies it and a wall supply must be designed for it, how isolation works, and what the 390 kΩ resistor is doing.
ExplainerDigital filters, and what they do to time
FIR versus IIR, Butterworth and its cousins, phase distortion, zero-phase filtering, and the ordering mistakes that quietly wreck analyses.
ExplainerArtifact removal, three ways
Prevent, reject, or model. Regression against EOG, independent component analysis, and artifact subspace reconstruction, with the rule for when each is honest.
ExplainerSpatial filters and montages
There is no neutral spot on the head. Referential, bipolar, average reference, Laplacian, and why the same recording looks different under each.
Skills
Hand skills and software skills picked up along the way.KiCad to JLCPCB
Schematic, layout, and a fabricated, assembled board for about the price of a textbook. The workflow, the layout rules that matter for microvolt signals, and the mistakes that cost a reorder.
SkillFirmware and SPI for a biopotential chip
Reading a register, reading a frame on data-ready, unpacking 24-bit two's complement, and shipping samples to a computer without losing any. The four things every acquisition firmware does.
SkillLab Streaming Layer and PsychoPy
How every BCI lab synchronizes a recording with the stimuli that caused it. Streams, timestamps, markers, and a stimulus program with millisecond timing, in an afternoon.
SkillThe characterization report
One page that says what your instrument does, with numbers, methods, and a plot per number. The document that turns a project into an instrument and a student into an engineer.
Ready for Phase 3 when
Checks save on this deviceWhat an interviewer would ask
If you can answer these, you are done here- Why does the ADS1299 use a delta-sigma converter, and what does 24 bits buy you when the signal is 50 microvolts in a 4.5 volt range?
- Two electrodes have impedances of 5 and 50 kilohms. Your amplifier's CMRR is 110 dB. What limits your hum rejection now?
- What is the leakage current limit for a patient-connected device, roughly, and how does a battery satisfy it?
- You filtered your epochs after cutting them. What went wrong?
Say it out loud, then reveal what a good answer contains.