# Aero Signal R01

Aero Signal is a non-invasive EEG research concept in an over-ear headphone form. The printed headset is an inert fit fixture with two removable cups, a headband gauge, and a place for sensor guides. The electronics prototype is for an artificial signal generator. It has no microphone, speaker amplifier, active electrode drive or intended contact with a person.

## Why around the ear

Reference studies detected resting EEG and auditory evoked responses with around-ear electrodes under controlled conditions. A direct comparison to scalp EEG showed that signal characteristics vary with placement. Speech, jaw movement, eye motion, hair, fit and contact quality are important confounders. Neither result establishes usable intent decoding on Aero's headset. [Debener et al., 2015](https://doi.org/10.1038/srep16743) · [Bleichner et al., 2016](https://doi.org/10.1088/1741-2560/13/6/066004)

## Architecture

- Fit prototype: cups with electrical separation between future acquisition and audio. The eight guide positions allow the team to compare candidate positions on an inert headform. A compact four-channel acquisition configuration is planned first. A reference and bias electrode geometry would require separate applied-part and scientific review.
- Desk electronics: the R01 PCB is a passive differential phantom adapter with two channels and configurable shielding. It is for the internal test sources and inputs of a supported bench AFE or an oscilloscope. The ADS1299-4 is a candidate 24-bit, four-channel acquisition device starting at 250 samples per second. Its published noise is device characterization under named conditions. Board noise, channel matching, contact impedance, common-mode rejection and safety are system measurements. [Texas Instruments ADS1299 data sheet](https://www.ti.com/lit/ds/symlink/ads1299.pdf)
- Compute: send bounded, sequenced raw windows to the paired phone or laptop. Four signed 24-bit channels at 250 samples per second need 3,000 payload bytes per second before framing; eight need 6,000. The prototype processes samples in the phone or laptop browser. It learns no biometric identity. A future BLE bridge needs sample counters, sender timestamps, gap flags, per-owner authentication and encrypted transport.
- Signal processing: mark missing or saturated samples before filtering. Estimate spectra from stationary windows, monitor 50/60 Hz interference, and reject suspect windows. Show the quality report before proposing an experimental control. Raw EEG stays local. No device action is taken on a sample.

## Bench and research gates

1. Prove correct amplitude scaling, channel order, rate and timestamp continuity using an artificial signal. Include blinded sine, noise, packet gap, drift, clipping and shorted-input cases.
2. With a fixture short and then known sources, record input-referred noise and 50/60 Hz common-mode rejection. Repeat with display, radio and audio on and off. The host demo is a software check and provides no hardware noise result.
3. Establish an isolation and risk plan with qualified biomedical engineering review before any body-connected electronics. Do not connect the R01 fixture to skin. It has no patient isolation or body-contact safety evidence. Keep USB, charging, programming and bench instrumentation away from any human-connected setup. A suitable validated research acquisition system and approved protocol are prerequisites for participant work.
4. If later approved, preregister a task, a baseline sensor, motion controls, false trigger rate, calibration duration and held-out sessions. Look for repeatable, task-specific input with a reliable abstain mode. Avoid interpreting mood, diagnosing health conditions, claiming unrestricted thought decoding, or controlling anything consequential from a single signal.

The fit design, PCB and software are separately testable. The launch stage is an inert enclosure and phantom experiment.
