BaroAware

Intrusion detection from air pressure. One module per floor. No line of sight, no microphone, no camera, no cloud.


A building is a leaky box. Open a hole in it, and the pressure inside changes — everywhere at once.

That last part is the whole product. At the frequencies involved the wavelength is many times the size of the building, so there is no wave travelling and nothing to localise: the enclosed air behaves as one volume. Open the envelope anywhere and the entire volume responds together — through walls, around corners, into the room where the module sits.

This is not acoustic detection. The module is not listening for a sound arriving from somewhere; it watches the pressure of an enclosed volume change when that volume is breached. It sees a door unlocked quietly with a key just as well as one forced.


Why this is not already everywhere

Measuring the pressure is the easy half. The other half is that wind against a façade produces swings larger than the event you are looking for. So does a ventilation system starting up. In amplitude, the signal-to-noise ratio is below one.

That kills the obvious approach outright. There is no threshold that catches a window and ignores the weather — not a cleverer threshold, not a sharper filter, not a different band. The two are separated by the shape of the disturbance in time and frequency, not by its size.

So it is learned, not written

BaroAware runs a convolutional neural network trained on real recordings from an occupied building — days and nights, in wind and rain, with people living normally inside. It runs entirely on the microcontroller. No cloud, no gateway, no subscription.

What the module sees

Spectral record from a BaroAware module: two bright horizontal bands marking two door openings, against a dark quiet background
An unedited record from a BaroAware module in an occupied building. Time runs upward. Two doors were opened — each one writes a bright band across the record. Everything between them is the building breathing on its own.

Nobody had to tell the module those two moments were different from the rest. That is the part that took the field data.

Block diagram: barometric sensor to microcontroller running a neural network, out to UART, ALARM and TROUBLE lines
Sensor → neural network → UART and two hardware lines. The module decides on board.

What that buys the installer


BA-CORE-01 — the module

BaroAware Core module, component side BaroAware Core module, solder side
BaroAware Core — component side and solder side. 24 × 24 mm.
Size24 × 24 mm
Supply3.3 V only — all 5 V circuitry stays on the carrier board
InterfaceUART, plus two logic outputs: ALARM and TROUBLE
Supply currentapproximately 11 mA
Product codeBA-CORE-01

It connects to whatever the integrator already has: an existing alarm panel, an ESP32 or Raspberry Pi, a PLC, a fieldbus node. The module decides on board; the host only has to listen.


What it does not do

One limitation, stated plainly: the module tells you that something was opened — not which. If you need to know which window, you still need a contact on that window. BaroAware tells you to go and look, and it does that from one point for the whole volume.

It complements an alarm system. It does not replace door contacts or motion detectors; it covers a class of entry those miss.

Status

Engineering samples. The detection technology is validated in the field — it has been running for months on our own hardware platform, on several units in a real, occupied building, continuously. BaroAware Core is that technology in its final, compact form.

Evaluation units are available to integrators and distributors. The serial interface specification is available on request.