The office is a separate building with a metal-foam sandwich roof and a steel-skinned door. The house has metal-framed, metal-barred windows. One Zigbee coordinator indoors cannot serve both sides of that masonry reliably, and the interesting question was not “what is the best link?” but “what is the worst link?”
The setup
Five devices: a coordinator, two routers, and two battery sensors. The coordinator is a network-attached SLZB-06M. I tested four positions and two antenna orientations, recording LQI and RSSI per device. The rule was to score each configuration by its lowest LQI, never by its best device.
| Config | Coordinator position | Antenna | Worst LQI | Worst RSSI |
|---|---|---|---|---|
| A | Bedroom | vertical | 68 | −94 dBm |
| B | Living room | vertical | 88 | −89 dBm |
| C | By the office door, outdoors | horizontal | 68 | −94 dBm |
| D | Bedroom | horizontal | 88 | −89 dBm |
The best individual link in config A was 204 LQI. The worst was 68. A configuration that looks excellent on one device can still be one hop away from dropping another.
What the numbers showed
- Antenna orientation was worth up to about 16 dB, and it was a trade, not a free gain. Rotating the coordinator improved the office plug from 112 to 176 LQI, while moving about 19 dB out of a living-room plug.
- The steel door cost about 6–7 dB. The office link was marginal when the door was closed and healthy when the house masonry was out of the path.
- Occupancy matters. A long instrumented window showed a 6–8 dB spread that was not orientation and not thermal drift; people and animals were moving through the 2.4 GHz paths.
- The battery sensors report roughly hourly. A stale-sensor alarm needs a 2–3 hour window, and nothing time-critical should depend on them.
The lesson
Design to the floor, not the mean. An empty house is the unrepresentative case; a normally used house is the one the system has to survive. The target design is two hops instead of one impossible hop: the coordinator indoors, a mains Zigbee router at the wall penetration, and the antenna aimed at the shed. Every radio goes above head height and out of the walkways.
The full measurement protocol is in the homelab notes so the next sweep is comparable with this one.