Life Oracle

Installation protocol: soil probe

Where

  • Representative spot, not the wettest corner, not the field gate: away from ditches, drains, gateways and tramlines. With a node the probes sit 1.5 m south of the post on their own 2 m leads, so the post itself has to stand somewhere representative. South because the panel's shadow always falls north. The bias sources are all sub-metre: a drip line reaches 0.3 to 0.5 m, the disturbed ground round a driven post 0.2 to 0.3 m.
  • One probe per field at 10 cm. For deep-rooting crops (maize, miscanthus, trees) add a second at 30 cm. Register each as its own device with its depthCm.
  • Note the position to a metre (phone GPS) and mark it: a short stake 30 cm away, or a buried magnet plus a note in the field map, so machinery does not hit it and you can find it in five years.

How

  1. Dig a hole wider than 20 cm to just below the target depth. Keep the soil in order of layers.
  2. Insert the prongs horizontally into the undisturbed side wall at the target depth, fully, slowly, without rocking. Do not hammer. Stones break probes; if you hit one, move 20 cm.
  3. Backfill in the original order and firm to natural density. Air gaps give low readings for months.
  4. Route the 2 m lead back to the box through a downward gland, with the slack pinned down under a peg or a short length of conduit where it crosses ground that gets mown. A stand-alone LoRaWAN probe has its transmitter on a short stake instead, above mowing height, antenna vertical.
  5. Wait 24 to 48 hours before trusting the readings; the soil has to settle around the prongs.

Wiring into the node (standard)

Each probe keeps its own factory lead and goes straight through its own gland into the box: 5 V and ground from a USB breakout on the Pi, so the probes are powered exactly when the Pi is, and the RS485 pair to the adapter. No extension cable, no junction box. Set each probe to its own Modbus address (10 cm and 30 cm) before it goes in the ground, because reaching it afterwards means digging.

  1. Register both probes as their own devices on the place, each with its depthCm, and write the two device tokens into the node with node/provision.sh.
  2. Bench-test before you dig. life-soil-agent --scan walks the Modbus addresses and prints the raw registers next to the decoded values, which is how you catch a register map that does not match its datasheet, and life-soil-agent --set-address 2 --addr 1 writes the new address, with only that probe on the bus. Both probes leave the factory on address 1 and will collide. The order of work is in the build plan.
  3. In the field, connect, then watch one reading arrive. The node posts every 20 minutes and buffers to disk when the modem is down, so a gap in coverage is a delay and not a loss.
  4. Note both depths and the position on the place. "Last seen" moves within one interval.

LoRaWAN join (stand-alone probes)

  1. Before you go to the field, add the device in The Things Stack (Sandbox or your operator) from the device repository (Dragino, Seeed, Milesight, Decentlab are all there) with its DevEUI, AppEUI/JoinEUI and AppKey from the label.
  2. Set the uplink interval to 20 to 30 minutes. That is about 10 s of airtime a day at SF7 to SF9, well inside the fair-use policy, and gives battery life of five years and more.
  3. In the field, power on and watch the join in the console. No join: move the gateway, not the probe. A window-sill indoor gateway reaches 1 to 2 km; an outdoor one on a mast 3 to 8 km on flat land.
  4. Add the webhook (see data protocols). The oracle shows "last seen" on the device within one uplink interval.

Calibration

Factory calibration is for mineral soil. For peat or very sandy soil, note the soil type in the device notes; the oracle uses relative change over time (direction), so a constant offset does not change a reading. If you have a soil texture analysis, put it in the notes: it will be used when the texture-specific moisture curve lands.

Maintenance

None planned. Check "last seen" quarterly. A wired probe on a node has no battery; a LoRaWAN probe needs one every 5 to 10 years depending on the model. If "last seen" goes stale on a node, both streams go stale together, which points at the box or the modem rather than at the probe.