Skip to content

Farm sensor architecture — Demory farm site

Purpose: Canonical reference for long-range field sensing on SITE_FARM (Demory) — approximately 120 acres, off-grid-first / resilience-first, with local survivability for basic telemetry and Starlink only as optional WAN for egress and Claxton home-site visibility. Does not specify parcel coordinates, tower sites, or SKUs—those require field survey and pilot evidence.

Assumptions (disconfirm with mapping + RF survey): ~120 ac production surface; terrain may break LOS for single-gateway LoRaWAN; Starlink present at a farm edge for WAN but must not be the only path for local sensor truth or safety interlocks.

Doctrine: Off-grid systems doctrine package — Demory · Mesh and field networking strategy — off-grid Demory.

Sources (official / operator cluster): Demory farm sensor layer — official and operator-grade source cluster.

Related pages: Gateway and controller patterns — Demory, ~120 acres · Sensor degraded modes and failure rules — Demory · Sensor power and duty-cycle assumptions — Demory · Sensor checklist matrix — Demory farm · LoRaWAN vs Wi‑Fi HaLow vs Meshtastic — Demory farm sensor layer.

Layering (field vs WAN)

Layer Role Starlink required?
S — Sensor leaf Soil, tank, weather, gate state, power telemetry (sparse uplinks) No
R — Mesh repeater (Meshtastic-class) Multi-hop where no single gateway sees all edges No
G — Field gateway LoRaWAN NS / ChirpStack, MQTT to local broker, buffer No for local delivery
H — HaLow / 802.11ah IP bridge IP-native sensors / bridges where LoRa throughput is insufficient No for LAN segment
Controller / edge Broker, SoR edge, optional k3s workloads (Pcrit tier) No
W — WAN edge Starlink CPE → internet Yes only for cloud / remote Claxton visibility

Rule: Basic sensing and welfare-critical local logic must not depend on Starlink—see Connectivity dependency map — Demory.

Starlink terminates WAN at the farm controller site—not in the field RF layer. Field devices speak LoRaWAN / mesh / HaLow to G/H; G connects east to MQTT/DB and north to WAN for remote operators at Claxton.

When WAN is down, local broker state and queues define what Claxton cannot see until reconnect—design SoR and alerting accordingly (Telemetry system of record — boundaries and authority).

Reference architecture (steady-state)

Legend: Solid = local-first path; Dashed = optional WAN.

```mermaid flowchart TB subgraph field["Field — Demory ~120 ac"] S1["S — LoRaWAN / Meshtastic leaf"] S2["S — sensor leaf"] R["R — mesh repeater optional"] H["H — HaLow IP bridge optional"] end

subgraph edge["Farm edge — Pcrit-backed controller"] G["G — LoRaWAN gateway + ChirpStack / NS"] BR["MQTT broker / SoR edge"] CTRL["Controller — automation + store"] end

subgraph wan["WAN — sheddable"] W["W — Starlink CPE"] end

subgraph home["Claxton SITE_HOME — operations"] OPS["Operator UI / VPN / dashboards"] end

S1 --> G S2 --> R --> G H --> CTRL G --> BR --> CTRL CTRL -.->|"optional egress"| W W -.->|"remote visibility"| OPS CTRL -.->|"site-to-site / VPN"| OPS ```

Local-only survivable (no Starlink): S → G → BR → CTRL and on-CTRL dashboards at farm.

Pilot architecture (Phase 0–1)

Reduce RF families and gateways until DR-5-class evidence exists (Off-grid infrastructure stop rules).

```mermaid flowchart LR subgraph pilot["Pilot scope"] S3["1–2 use cases e.g. tank + soil"] G1["One G — one gateway family"] end

subgraph local["Must work without WAN"] BR2["Local broker"] UIf["Farm-local UI or logs"] end

S3 --> G1 --> BR2 --> UIf ```

Degraded-mode architecture (WAN loss / gateway loss / power reduction)

```mermaid flowchart TB subgraph degrade["Failure classes"] F1["WAN / Starlink down"] F2["Field gateway G impaired"] F3["Power reduction — shed non-Pcrit"] end

subgraph local2["Still required"] L1["Local welfare / interlocks — wired or sight"] L2["Last-known broker state — not life-safety alone"] end

subgraph manual["Escalation"] M1["Manual rounds / physical verification"] end

F1 -.->|"queue / stale OK"| BR3["Broker buffers"] F2 -->|"partition"| M1 F3 -->|"drop RF tiers"| M1 L1 --> M1 L2 --> M1 ```

Runbooks: Runbook — broker or backhaul down · Runbook — sensor false-positive alert triage · Off-grid degraded modes — power and connectivity (Demory).

Cross-site visibility (Claxton)

Goal: Telemetry and operational visibility at Claxton without making Starlink a hidden dependency for field truth.

  • Preferred: SoR or broker replication / sync with explicit staleness labels when WAN flaps.
  • Avoid: Only cloud dashboard as authority for field state.

Two-site ops hub: Two-site smart farm operations.