Connectivity Options for Wireless Sensor Networks
“What should I use — BLE? WiFi? NB-IoT? Or perhaps I should consider LoRA/Sigfox? How can I gather sensor data in a reliable, timely and efficient manner?”
OK — so you have zeroed in to your favourite cloud provider — which might be AWS, Azure or Bluemix, etc. And you have the required expertise in Machine Learning to solve a key problem faced by a section of industrial customers. While these major cloud providers fully support & document software needed to run on field gateways, they are for the large part, conspicuously silent about the actual wireless sensor network that you might end up needing to deploy in the factory. In a previous article, I shared my take on selecting a Gateway for an IoT PoC. In this article, I want to kickstart a discussion around selecting the right low power technology for wireless sensing networks in Smart Factory and Smart Agriculture applications.
So what are your options when you want to collect sensor data in a reliable, timely and efficient manner? And what factors should you consider when evaluating different options?
Perhaps one of the most important factors is range. How far apart are your sensor nodes likely to be in the factory? 5m? 10m? 50m? or 100m? What sort of environment are these likely to be in? Metal cabinets, concrete walls, and heavy machinery — these are all enemies of wireless range.
Are your sensors likely to run off batteries? Or will they be running on mains supply? If they run off batteries, what would be the desired battery life?
What would be the maximum data throughput for the wireless sensor node? 10Kbps? 50Kbps? Or perhaps 250Kbps? If you don’t know this yet, it might be important to take some time and work out the use-cases to arrive at a specification for maximum throughput.
How frequently be the sensor nodes transmitting data over the wireless link? Once a day? Once an hour? Or perhaps once every 30s?
What is the volume of deployment? Is there a cost budget for upfront payment? Is there one for data transmission?
Depending on the location of deployment, the ISM band if operating in SubGHz (<1GHz frequency) might change. The regulatory requirements will also vary across the globe. Read my other introductory article on the topic.
How much time and development resources do you have? (hopefully, your answer is ‘lots’ but I wouldn’t be too surprised if it isn’t).
The table below lists out a comparison I have between different wireless technologies. It is important to note that figures mentioned are more to get a sense for the order of magnitude and might not be accurate.
What would I use?
For industrial applications where I would expect to be transmitting a reasonable amount of data frequently, as of today — I would probably go with 6LowPAN or WiSUN option for my wireless sensor network operating in the subGHz ISM band. The reasons are:
- IP to the end-node: 6LowPAN provides an industry-standard adaptation layer between IPv4 and the LPWAN (Low Power Wireless Area Network) formed by the sensor nodes. This means, theoretically, every sensor node becomes accessible from the cloud.
- Scalable throughput and range— I have greater flexibility to trade-off throughput (bandwidth) vs range. The ADF7030–1 for example is a narrowband RF transceiver that can support throughputs from 0.1Kbps to 300Kbps. Using this transceiver in a wireless sensor node running the 6LowPAN stack would be a powerful option for a wide variety of use-cases.
- Ultra Low Power — for battery powered applications, I can design my wireless sensor network in a manner which lets the sensor node use a high throughput to transmit its data quickly and go to sleep. Thereby reducing the amount of time the sensor node consumes the higher level of Tx current. ADF7030–1 has an ultra low sleep current of 10nA. This coupled with a host MCU such as the ADuCM3029 — which consumes just 830nA in Hibernate mode, would result in an overall energy-efficient wireless sensor node.
For other applications where I wouldn’t expect as much data to be transmitted regularly, for example — smart agriculture, there might be a case for Sigfox. I would still prefer to re-use the same 6LowPAN wireless sensor node (perhaps until NB-IoT becomes more widely available). The required range/distance might be larger in case of smart agriculture, but then I can trade-off throughput vs range with a common wireless sensor node solution involving the ADF7030–1 and ADuCM3029 for instance.
Long Term Trend
Longer term (5 years or so), my view is that Bluetooth will continue to evolve and provide greater flexibility in supporting a wider range of throughput. Once it does that — it should be able to dominate the short range (<100m) wireless market. At the other end of the spectrum, NB-IOT once available globally, will probably dominate the long range (>100m, ~KMs) wireless market. Therefore, we will probably observe a convergence to Bluetooth and NB-IoT as the two main wireless options for battery powered applications. That is my personal take! For more info on Analog Devices’ high performance solutions in the wireless space, feel free to leave me a message. #AheadOfWhatsPossible!
Until then, adios and good luck with your projects!