LIBRARYThe ADS1299 Explained.

Why the Texas Instruments ADS1299 (an 8-channel, 24-bit, simultaneous-sampling biopotential front-end) became the de-facto chip for DIY and research EEG.

The Texas Instruments ADS1299 is the chip most DIY and research EEG systems are built around. It's an 8-channel, 24-bit, simultaneous-sampling delta-sigma analog front-end (AFE) for biopotential measurement: each channel pairs a low-noise programmable-gain amplifier with its own 24-bit ADC, and the chip throws in an internal reference, oscillator, bias-drive amplifier, and lead-off detection. In one SPI device it replaces a sprawling discrete instrumentation-amp-plus-ADC front-end you'd otherwise design by hand.

Why EEG needs a specialized chip

Why does EEG need something this specialized? Scalp EEG is a microvolt signal sitting under interference that can be thousands of times larger. To recover it you need an extremely low noise floor, very high common-mode rejection (to kill mains hum), and a converter resolution fine enough to see µV steps. The ADS1299 is the low-noise, EEG-targeted member of TI's ADS129x family. Its ADS1294/6/8 siblings are aimed at ECG, whose signals are far larger and more forgiving.

Fig 1

Spec	ADS1299
Channels	8 (also -6 and -4 variants)
Resolution	24-bit
ADC architecture	Delta-sigma, simultaneous-sampling (one ADC per channel)
PGA gain settings	1, 2, 4, 6, 8, 12, 24
Input-referred noise	~1 µVpp (70 Hz BW)
Sample rate	250 SPS – 16 kSPS
Interface	SPI
Reference	Internal 4.5 V
Supplies	AVDD ~5 V (or ±2.5 V) · DVDD 1.8–3.6 V

The bias / right-leg-drive loop

The ADS1299's most EEG-critical feature is its on-chip bias-drive amplifier, the same idea as ECG's (right-leg drive). It senses the body's common-mode voltage and drives a corrective signal back through a bias electrode, actively cancelling 50/60 Hz mains interference before it reaches the channels. In practice it isn't optional: the community's classic '60 Hz everywhere' bug almost always traces to a missing or poor bias connection.

The registers worth knowing

You don't program the ADS1299 pin by pin. You write its registers over SPI. The ones worth knowing by purpose: CONFIG1–3 set the global data rate, reference, and test-signal source and enable the bias amp; CH1SET–CH8SET set each channel's PGA gain and input source (normal, shorted, internal test signal, or bias measurement); BIAS_SENSP/N choose which channels feed the bias-drive loop; LOFF configures lead-off (disconnected-electrode) detection; and the SRB1 switch routes one shared reference electrode to every channel for a referential montage. Two pins run the timing: START begins conversions and DRDY (data-ready) tells the host a fresh sample is waiting on the SPI bus.

▸Deep dive· Go deeper: why not just roll your own front-end?

You can build an EEG front-end from discrete instrumentation amplifiers and a generic ADC, but then you personally own the noise floor, the common-mode rejection, channel-to-channel matching, and a lot of board complexity, and microvolt EEG punishes every shortcut. The ADS1299 hands you a validated low-noise multichannel chain for the price of learning its register map, which is why it anchors open hardware like OpenBCI's Cyton. A peer-reviewed evaluation of the ADS1299 even found no statistically significant difference from a gold-standard clinical system for low-frequency EEG (Sensors, 2018). The tradeoff is integration-and-ecosystem versus full control.

Gotchas the community hits

A microvolt front-end will faithfully digitize your power-supply noise, so it needs a clean, low-noise supply. Electrode-skin impedance, not the chip, usually dominates real-world quality. And the ADS1299 is an instrumentation part, NOT a medical device: a DIY EEG built on it is for learning and experimentation, not diagnosis, and must be isolated from mains when near AC-powered equipment.

Checkpoint

Quick check

What is the ADS1299?