Mrs. Sable, my English teacher, had a top-floor door that worried her — nobody used it, but she couldn’t hear if someone came through it at night. She asked if I could build something. I said yes, because I’d do anything to get out of prep time.
The system
Multiple nodes scattered across the house — PIR motion sensors at doors and entry points, one master node near the bedroom that actually drove the alarm. The alarm itself was a motorbike horn — loud enough to wake the dead, powered by a 12V SLA battery. The master listened for trigger messages from the sensor nodes over 433 MHz.
The accidental mesh
I hit an immediate problem: some sensor nodes couldn’t reach the master directly because of wall thickness and distance. The obvious fix at the time was “move the master” — which didn’t work either — and I didn’t know there was a whole field of network engineering called “mesh networking” with established solutions to this. So I reinvented a version of it from first principles.
I programmed each node to listen for packets from other nodes as well as generate its own, and to rebroadcast any packet it saw if it hadn’t already. With a hop count to prevent infinite loops. That meant a sensor on the far side of the house could route a trigger through an intermediate node to the master, and the alarm would fire. Packet loss was ugly at first — I added a lightweight acknowledgement scheme with retries.
Where it led
This was where I learned that mesh networks exist, because a year later I read about them and went, “oh, I made one of those.” Mrs. Sable’s alarm ran for years as far as I know. The patterns I figured out here showed up later in Arzen’s master/slave cinema sync.