A microphone,
not a radar.

Deciding on the node

Hearing is the easy half. The work is in telling a drone apart from everything else outdoors — wind, rain, a chainsaw, a motorbike on the next road — well enough that an alert is worth acting on. An early-warning system that cries wolf gets switched off within a week.

That decision is made by a small neural network running continuously on a low-power processor inside the node. Audio never leaves the device. Only a short alert does, which keeps the cellular data requirement small enough for the cheapest plan available and keeps the system useful where bandwidth is poor.

Why sound, not radio

Radio detection only works while the drone is talking, and more and more of them are not. Pre-programmed routes, autonomous navigation and fibre-optic control all leave the radio spectrum silent. The propellers cannot be made silent without grounding the aircraft.

Why emitting is a liability

An active radar can be detected, located and targeted from well outside its own detection range. Its transmission is a beacon, and switching it on is a decision with consequences. A microphone has no such trade-off: there is nothing on the air for hostile radio reconnaissance to find while the node listens, so it can stay on permanently.

The same reasoning shapes the enclosure. A nest box on a tree is not worth a second look.

What acoustic does not solve

Being straight about the limits is part of the design. Range depends on the target. The engine of an attack drone carries for kilometres, while a small electric quadcopter is heard only at short range, and Nest is built for the former. Strong wind, heavy rain and loud places such as motorways shrink the range, and mounting high helps. A single node gives presence, not a track; direction and speed come from several nodes reporting together.

None of this is a reason to wait. A line of affordable nodes that hears what flies low is worth more than a perfect picture that arrives too late.