mmWave presence detection for building automation
Radar instead of PIR, because PIR cannot see a person sitting still.
Outcome Reliable occupancy detection for stationary people — the case PIR fails outright.
Hero: the sensor node installed in a ceiling or wall position, in a real room with real light. Context sells this one — a bare PCB says nothing about presence detection.
PIR sensors detect change in infrared, which means they detect movement. A person sitting still reading is, to a PIR, indistinguishable from an empty room.
Everybody has experienced the consequence: lights going out in an occupied meeting room, heating dropping back while someone is at a desk. The usual fix is a longer timeout, which trades energy waste for irritation and solves nothing.
Why radar
mmWave radar measures reflections, and a stationary human body still reflects — including the micro-movement of breathing. That makes it possible to distinguish an occupied still room from an empty one, which is the case PIR simply cannot address.
| PIR | mmWave radar | |
|---|---|---|
| Moving occupant | Reliable | Reliable |
| Stationary occupant | Fails | Detects, including respiration |
| Through thin materials | No | Possible, depending on material |
| Zone definition | Fixed lens cone | Configurable in range and angle |
| Cost | Very low | Higher, and falling |
| False triggers | HVAC airflow, sunlight, thermal drift | Fans, moving curtains, adjacent-room bleed |
Radar is not strictly better. It trades one set of false triggers for another, and the new set has to be engineered out — zone limiting, sensitivity tuning and sensible hysteresis. The advantage is that radar’s failure modes are addressable in design, where PIR’s central limitation is physical.
System integration
The presence sensing fed a wider building automation system: lighting, climate control, cameras and door locks, so occupancy could drive several subsystems from one reliable signal rather than each guessing separately.
Antenna keep-out is the whole board
A radar front-end dictates the layout. The antenna region needs its keep-out respected on every layer, a clean ground reference, and no metalwork in the field of view — including the enclosure and the mounting hardware. Get that wrong and the sensor underperforms in a way that looks like a firmware problem and is not.
Further images to produce
These slots are laid out and waiting for assets. Each carries its own
brief — see shotList in this project’s source file.
Coverage diagram: vector drawing showing detection zone versus a PIR cone, ideally with the stationary-occupant case marked. This is the whole argument in one image.
The board: top-down render or photo with the radar front-end and antenna region visible. Annotate the keep-out area.
Bench data: a screenshot or plot from your own measurements comparing radar and PIR detection of a stationary person. If you have this data, it is the most valuable asset on the entire site — it becomes an article as well.