Battery is the business end of IoT
Battery is very important in wireless sensor networks. The quality of an internet of things project has a big impact on battery life. Hardware is cheap. Getting the software right is vital.
Everything is built for low power
Low power means a slow network. 802.15.4, which is the physical network, has a low bitrate and runs between 20 Kbps and 400 Kbps. Devices can come and go from the network. Low power consumption is at the heart of everything. This research paper from Korea puts battery life of a 6LoWPAN on a TI CC2530 board at 53 days for two small AA Duracell batteries. In the video in my first post, Vint Cerf reports his cisco equipment has battery life over a year on AAs. The Waspmote battery life is said to be up to five years depending on configuration. The quality of the IoT project determines battery life; various configurations and software choices impact power consumption. The common button cell CR2032, 3cm2 batteries, may only provide about a months worth of power to a Zigbee microcontroller unit with 128 KB of flash memory with up to 8 KB of RAM. A sensor hardware bill of materials is only a couple of dollars. Hardware is only part of the story about battery life. Software is the other part.
Building systems for low power
Low power wireless sensors have low power microcontrollers, without virtual memory. Android Linux and Embedded Linux have no memory paging. Linux has the developer support – protocol libraries, tooling, builds, etc. Because of this, the Linux kernel is the economical choice. That said, the more tasks running, the more battery consumed so you need only add necessary libraries to an embedded application running on a mote. There are ways to combine Linux with the very efficient TinyOS. You get the best of both worlds: the TinyOS scheduler and the Linux support. TinyOS puts the microcontroller of a mote to sleep when no tasks are scheduled.
Other options are available, these simple microcontrollers usually don’t need to run much more than basic process abstractions.
Developments in ambient power
Since the 1970s backscatter has been proposed as the solution to power chips. This year saw the first working backscatter system, which is exciting. The communication channel offered by ambient backscatter is tiny. If such energy harvesting improved, a whole stack of protocols would need to be published to get something like this into the mainstream.