US boffins propose yet another low-low power Wi-Fi for Things
A sober reflection on wireless comms
University of Washington boffins are touting extremely low-power Wi-Fi transmissions – if your application can put up with a maximum 11 Mbps capacity.
The students and boffins, Bryce Kellogg, Vamsi Talla, Shyamnath Gollakota and Joshua Smith of the University of Washington, have in mind the kind of low-power devices that will proliferate as household sensors in the Internet of Things era.
Gollakota has a long record in fooling around with backscatter. In 2013 with Sidhant Gupta, he demonstrated a scheme to use Wi-Fi for gesture recognition. He improved on that in 2014 (with Kellog and Rajalakshmi Nandakumar), by putting gesture recognition into unmodified laptops.
Last year, Gollakota and Smith – the academics in the author list – set up a company called Jeeva Wireles to pursue their interest in low power communications.
The scheme their students Kellog and Talla worked on has two key power-saving techniques. First, they offload all of Wi-Fi's carrier sensing activities to devices like routers, since these are always-on anyhow. Second, they've created a way to modulate 802.11 signals on the back-scatter from a device.
These approaches mean a device can be entirely passive and still communicate, albeit at a limited data rate – the group demonstrated communications at 1 Mbps (consuming 14.5 microwatts) and 11 Mbps (consuming 59.2 microwatts), which aren't going to excite people signed on for Netflix 4K services.
Their starting point is simple: analogue circuits like ADC/DACs and amplifiers are power hogs, and unlike digital circuits (for example CMOS) they don't benefit from digital scaling. The aim of “passive Wi-Fi” is to get rid of those analogue components.
Offloading carrier sensing to the Wi-Fi hotspot is straightforward: whether that hotspot is a smartphone or a broadband router, it will be sensing the radio channel whether or not the sensor is transmitting.
There's a lot more smarts in how the technology uses “backscatter” – the router's signal reflected back towards it – to create an 802.11 channel from the sensor.
The mathematics is in the linked paper, but the key point of manipulating backscatter is this: characteristics like frequencies in the reflected signal can be manipulated by changing the antenna that's doing the reflecting.
That can be done with a single digital switch “that toggles between two impedance states, resulting in a binary signal”, they write. So instead of the power budget of a radio transmitter, passive Wi-Fi-equipped switches only need to power that binary switch.
Everything in the group's demonstrator was made with off-the-shelf components, they note, but a custom comms stack is required for the plugged-in device to receive the sensor's communications.
The Register would note that there's at least one good reason why more work is needed for “passive Wi-Fi” to be a practical proposition: the University of Washington paper doesn't mention either security or encryption. ®