Choosing a Board for the Field Network

There is no single best board for this project. The Pico, Raspberry Pi 5, and Khadas VIM3 solve different parts of the problem. A reliable design gives each board the job it is best suited to do.

Match the Board to the Job

NeedBest starting pointReason
Small battery node with sensorsPico or Pico 2 + LoRa HATLow power, simple hardware, and direct control of GPIO and radio timing
Wireless field node with local WiFi optionPico W or Pico 2 W + LoRa HATBuilt-in WiFi/Bluetooth when supported by the firmware; still verify the power budget
Always-on Meshtastic gatewayPi 5 + USB Pico radioDebian, Python tools, storage, community support, and easy remote administration
Local analysis or specialized processingKhadas VIM3eMMC, networking, and an NPU for supported edge-inference workloads
Large multi-service base stationPi 5 16 GB or VIM3 plus storageExtra memory and storage headroom for dashboards, databases, and processing

Separate the Radio from the Heavy Work

A useful architecture keeps the battery node simple and moves expensive work to a powered machine:

Field Network Board Responsibilities Comparison diagram showing the responsibilities of a battery-powered Pico node versus a mains-powered Pi 5 gateway. The Pico handles sensors and LoRa transmission, while the Pi handles databases, MQTT, and dashboards. Field Node (Pico) ⚡ Battery / Solar Sensor Readings Sleep Management LoRa Mesh Gateway (Pi 5) 🔌 Mains Power SQLite MQTT Dashboards & Logs
Figure 1: By separating responsibilities, the fragile battery nodes can remain simple and reliable, while the gateway handles the complex, power-hungry database and networking tasks.

This also makes failures easier to understand. If the gateway is offline, field nodes can still communicate with one another. If a field node battery dies, the gateway and the rest of the mesh remain available.

Questions to Ask Before Choosing

  1. Does the board need to run Linux, or is one embedded program enough?
  2. Will it sleep between radio transmissions, or stay awake as a relay?
  3. Does it need eMMC, NVMe, or only a few configuration files?
  4. Will it be inside a protected building or exposed to cold, moisture, and solar heat?
  5. Does it need an NPU or other accelerator, or would that add unnecessary complexity?
  6. Can the exact LoRa radio, antenna, firmware target, and regional settings be supported together?

Use the 4 GB and 16 GB Pi 5 Boards Differently

The 4 GB boards are a good fit for dedicated gateways, serial radio bridges, SQLite logging, and lightweight services. The 16 GB board gives more room for development tools, multiple dashboards, larger databases, containers, and experiments—but memory alone does not improve a weak antenna, poor storage, or unstable power supply.

Build in Stages

Stage 1: one Pico and one radio, direct bench test
Stage 2: two nodes, private channel, outdoor range test
Stage 3: Pi 5 or VIM3 gateway with local logging
Stage 4: elevated relay node and scheduled telemetry
Stage 5: sensors, dashboards, remote maintenance, and power monitoring

Each stage should have a known-good configuration before the next one is added. This is slower than assembling everything at once, but much faster than debugging a complete field system with several unknowns.

References