Well, I'm finally subscribed to MQTT-SN topics over a radio distance of 20 cm...

I'm losing about 40% of messages output by my beacon publisher, mainly due to an occasional dropped message about only 10% of the time (due to the Bus Pirate end not being as reliable as the embedded Linux spidev) seeming to interrupt the 'flow' of MQTT. My bridge is currently dumb, doing nothing other than a checksum. I thought that MQTT-SN would be more robust than that, being designed for such links, so I probably should read the spec now! In the medium term that wont be a problem anymore once I can replace the bus Pirate completely.

We had a planning meeting over the long weekend and refined the architecture a bit, so the main diagrams are going to need to be updated.

We'll probably end up with three node types.

First, the farmstead, with Ethernet and no requirement for low power - we'll probably use a Raspberry Pi because these can be had at short lead time here.

Second, a more powerful fencepost with store and forward capability, as this topology maps best onto typical farms in this area. The Carambola 2 will be put to use in this capacity as the first deployed node, Linux powering store-and-forward with backup logging to USB, and a suite of additional sensors including USB goodies that wont necessarily be deployed to the leaves, and if required more than one radio. This will need a largish solar panel, but will be better for low power than a Raspberry Pi.

Currently the comms is all effectively dumb half duplex with no media access control, and initially for multiple spokes we'll drive multiple radios on the store and forward node. Longer term of course we need to upgrade to a proper link protocol, maybe LoRa or Dash7, but that can be done in software in the longer term future.

The leaf fencepost nodes we are now looking at coupling an ESP8266 (201) with the SX1276, they have enough grunt to drive it but are much cheaper than the Linux modules, and still provide wifi for a farmer to drive up and communicate with using a phone. At this stage the fencepost nodes still going to provide wifi to the ground sensors yet have enough GPIO be able to pivot to 433MHz.

Finally we are going to use Arduino Nanos or maybe the new uHex Atmel powered models for low power sensor applications where a MCU cant sleep (like rain gauges) slaved from the fencepost nodes.

All this is of course a work in progress and subject to further change!

Soon hopefully James will be providing an update one of the whether sensors we are designing.