The question usually arrives framed as a choice: should the sensors use LoRaWAN, 4G or Starlink? It is the wrong framing, because the three technologies solve different problems. A soil probe in a far paddock needs to send a few bytes every quarter of an hour, for years, on a small battery. A site office, a camera network or an edge compute node needs a steady stream of data to somewhere off the property. Most sites need both, and the designs that work treat them as two layers rather than one decision.
Two different jobs
The first job is getting readings in from the paddocks. The devices are small, numerous and spread out. They usually run on batteries, they sit wherever the thing being measured is, which is often low in the landscape at a trough, a creek crossing or a soil pit, and they send very little data. What matters here is range, battery life, cost per device and tolerance of awkward terrain.
The second job is backhaul: connecting the site as a whole to the outside world. There are only a few endpoints, the data volumes are much larger, power is available or at least budgeted, and you choose where the equipment goes. What matters here is throughput, latency, reliability and what happens when the link fails.
Most of the trouble comes from pushing one layer’s technology into the other’s job. A 4G modem on every trough sensor multiplies SIM plans and flattens batteries. Camera uploads over a sensor radio simply do not fit.
LoRaWAN: built for paddocks
LoRaWAN is a long-range, low-power radio standard designed for exactly the first job. It runs in shared spectrum rather than on a carrier’s network, so there is no mobile plan per device; use equipment that is approved and configured for Australia, and check the details with your supplier.
Range depends almost entirely on where the gateway sits. From a well-sited gateway, several kilometres is realistic in open country, more with a clear line of sight, and much less behind a ridge or in timbered gullies. Height beats transmit power: a gateway on a hilltop or a tall mast at the shed will outperform the same gateway on a verandah post by a wide margin.
The trade-off is data. Payloads are tiny, typically tens of bytes, which is plenty for a tank level, a soil moisture reading, a temperature or a gate switch, and nowhere near enough for images. In return, simple sensors can run for years on a small battery.
A few details catch people out:
- Most battery sensors are Class A devices, which can only receive a message in a short window just after they transmit. That is fine for reporting, and awkward for control. If you need to open a valve on demand, check the device class and expect delays.
- Devices and gateways must agree on the Australian channel plan. A sensor bought from an overseas supplier may be configured for a different region.
- The gateway does not get data off the property by itself. It forwards packets to a network server, and that server has to live somewhere. You can run your own gateway and server, or join a community or commercial network where one covers your area.
4G: convenient where it reaches
Where coverage is solid, 4G is the quickest way to connect a single device: a modem, a SIM and a plan. It suits a lone weather station at an out-station, a pump controller, or the second path for site backhaul.
Coverage maps are modelled, and they are more optimistic than a paddock usually is. Hills and timber create dead spots, and the house having good reception tells you little about the dam. Test on site with the actual modem and antenna you intend to use. An external directional antenna on a mast will often turn a marginal signal into a usable one.
LTE-M and NB-IoT are low-power cellular variants built for sensors, and they are available on parts of the Australian mobile networks. They can run on batteries, and in some places they reach further than standard 4G, but coverage differs by carrier and by technology, so confirm it for the specific location rather than the district.
Two practical notes. Running costs scale with device count, because each device carries its own plan. And Australia’s 3G networks have been switched off, so any older telemetry that relied on them, such as tank alarms, pump controllers or early weather stations, is either dead already or worth checking.
Starlink: backhaul, not a sensor radio
Starlink brings real bandwidth to anywhere with a clear view of the sky, and it is the natural primary backhaul for a remote site. It is not a sensor network. Each terminal draws tens of watts continuously, carries a monthly plan and needs a clear sky view. One dish serves the site; it does not serve forty sensors scattered across a property. Its job is to carry everything the site produces, from sensor readings to camera events and remote access, off the property.
How the layers fit together
On most properties the pieces end up arranged the same way:
- Sensors in the paddocks report over LoRaWAN to one or two gateways mounted as high as practical.
- The gateways connect to an edge node at the shed or house, by cable where distance allows or by a point-to-point wireless bridge where it does not. A pair of directional radios with clear line of sight will carry a gateway or a camera on a ridge a couple of kilometres away without difficulty.
- The edge node runs the network server and stores readings locally.
- The site router sends data off the property over Starlink as the primary path, with 4G behind it as a fallback.
Step 3 is the detail that matters most. If the network server lives in the cloud, a backhaul outage leaves the gateways with nowhere to deliver packets, and many gateways can hold only a little data locally. Readings from the paddocks are simply lost for the duration. With the network server on site, the sensors keep reporting to local storage and the data goes up when the link returns. Open-source network servers such as ChirpStack run comfortably on small ARM hardware.
Questions to ask before choosing
- How much data does each device send, and how often?
- Where will the devices be, and what terrain lies between them and the obvious gateway site? Walk it or run a terrain profile before buying anything.
- What power is available at each point?
- What does it cost you if a reading arrives an hour late? A day late?
- Does anyone need to send commands to the device, or only read from it?
- How many devices will there be in five years, not just this season?
As a rough guide: small, slow and many points to LoRaWAN; a few devices with moderate data where mobile coverage is proven points to 4G; the site’s link to the world points to Starlink, with 4G behind it.
Where Bizix Agritech fits
We design the backhaul and edge layer that paddock sensor networks depend on: SD-WAN built on VyOS over Starlink with a fallback path, and low-power ARM edge compute on site that keeps collecting when the link is down. The sensor radio is chosen to suit the paddocks, and the network is built so the data has somewhere to go. Designed and supported in Australia.