A light switch with no battery, no wires, and no conventional electronics inside — just plastic, a gear, and Wi-Fi. It sounds like a pitch deck slide. It isn't.
The technology is called backscatter, it's twelve years old, it has measured numbers and shipping silicon, and this month the IEEE working group standardizing it puts its first draft to a vote. Here's how a device transmits data without having a transmitter — and the one detail that almost every popular explanation of this gets wrong.
Why ordinary Wi-Fi needs a battery
Start with a question that sounds trivial: why is Wi-Fi so power-hungry in the first place?
Moore's law has squeezed digital logic to absurdity. A microcontroller that counts button presses draws nanoamps and costs pennies. But a radio is not just digital. To get on the air, a device needs an analog chain: a frequency synthesizer to generate the 2.4 GHz carrier, and a power amplifier to push the signal into the antenna.
Those two blocks do not scale with Moore's law. Analog RF has barely gotten cheaper in energy terms over the same decades that digital got millions of times cheaper. That's why transmitting over Wi-Fi from a sensor still costs hundreds of milliwatts — and that, not some inefficiency in the protocol, is what keeps Wi-Fi out of the world of tiny sensors. BLE and ZigBee are thriftier, but they hit the same wall: they need their own oscillator too.
Which leads to the obvious question: what if you didn't generate a carrier at all?
A mirror, not a flashlight
The analogy that explains backscatter best is a signal mirror.
To send a light signal at night you need a flashlight: power source, lamp, the whole apparatus. In daylight, a small mirror is enough. It produces zero photons of its own; it only redirects someone else's light. A heliograph works exactly this way — the sun does the shining, the operator only changes the angle. The energy cost of "transmitting" is whatever it takes to move your wrist.
Radio works the same way. The air around us is full of other people's waves. Every object reflects them. The only question is whether you can control how it reflects — and turn that into data.
You can. An antenna reflects an incident wave differently depending on what's connected to it: shorted or open, impedance-matched or not. Put a single transistor switch between the antenna and ground, toggle it, and the reflected wave becomes modulated. One means reflect, zero means absorb. This is how passive RFID tags in badges and anti-theft labels work; backscatter is their direct descendant that learned to speak Wi-Fi.
The key point: what toggles is a digital switch at baseband, not an analog oscillator at 2.4 GHz. Somebody else makes the carrier. You only pay for the switching — and that's microwatts instead of milliwatts.
Three generations in three years
The technology came out of Shyam Gollakota's and Joshua Smith's lab at the University of Washington, and its history is a fight for range and throughput.
2013 — Ambient Backscatter. First demonstration of the principle: two devices talk to each other by parasitizing TV tower signals. No power source of any kind. It worked, slowly and at arm's length.
2014 — Wi-Fi Backscatter. First time a battery-free device connected to ordinary Wi-Fi infrastructure. Conceptually a breakthrough; practically modest: hundreds of bits per second at 2–4 feet.
2016 — Passive Wi-Fi. This is where it broke open. Instead of riding a narrowband side channel on top of someone else's Wi-Fi signal, Bryce Kellogg, Vamsi Talla, Gollakota and Smith figured out how to generate full 802.11b packets by reflection — packets that any existing receiver decodes. Any router, laptop, phone. No modifications on the receiving side.
Measured on their fabricated IC:
| Metric | Value |
|---|---|
| Power at 1 Mbps | 14.5 µW |
| Power at 11 Mbps | 59.2 µW |
| Range | 30–100 feet, including through walls |
| vs. existing Wi-Fi chipsets | 10,000× lower |
| vs. Bluetooth LE and ZigBee | 1,000× lower |
To feel the scale: 14.5 microwatts is a level you can harvest out of thin air. Not figuratively — literally: that's roughly what you can pull off an antenna collecting ambient RF, or off a tiny photovoltaic cell under indoor lighting. The battery becomes unnecessary not because it was stretched thin, but because the energy required is now less than what's lying around for free.
The work was named one of MIT Technology Review's 10 Breakthrough Technologies of 2016 and took best paper awards at NSDI and SIGCOMM.
The part popular explanations get wrong
Here's the detail worth going to the primary source for.
Nearly every popular write-up describes the setup like this: "the router constantly emits radio waves, and the object reflects them differently." Intuitive, tidy, and wrong for Passive Wi-Fi.
The real architecture splits in two. The passive device handles only the digital baseband: coding, modulation, toggling the switch. Everything expensive in energy terms — the frequency synthesizer and the amplifier — is moved off the device onto a separate, plugged-in unit. That unit doesn't transmit data. Its only job is to radiate a single-frequency tone, a flat unmodulated signal that nearby passive tags reflect.
So there are three participants, not two: a plugged-in tone source, the passive tag, and an ordinary receiver (router or phone) that ends up reading the packet. One source serves every tag in range, so the economics work — but nothing works without it. The router sitting on your shelf today does not emit that tone.
This isn't pedantry about wording. That requirement for a dedicated tone source is exactly what kept the technology in labs for twelve years: it demands special equipment pre-installed in the room. As long as that's true, the story is about warehouses and industrial sites, not a button in your bedroom.
And that is precisely what's changing now.
Where things stand in 2026
Three separate threads converged over the past year.
Silicon. Canada's HaiLa shipped the BSC2000, a Wi-Fi-compatible connectivity RFIC operating in the ~40 microwatt range. The company's engineers are targeting single-digit microwatts for the third generation. That's a part you can design into a product, not an FPGA on a lab bench.
Research. In August 2025, Privitera, Ballo, Ahmed, Grasso and Alioto published a transmitter with neither a battery nor its own oscillator: it extracts the frequency directly from the incident wave through second-order intermodulation of a two-tone signal. The same incoming wave does three jobs at once — energy harvesting, data transmission, and motion detection. Reported sensitivity down to −19 dBm, consumption below one microwatt.
Standardization. The important one. Since March 2024, IEEE 802.11 has had Task Group bp writing an amendment called Ambient Power (AMP) — communication for devices running on harvested energy. This is not a hobbyist effort: chair from Sanechips, vice chairs from Qualcomm and Infineon, secretary from Ericsson, technical editor from OPPO.
The group's official milestones:
| Milestone | Date |
|---|---|
| D1.0 ready for WG letter ballot | Sep 2026 |
| D2.0 WG LB recirculation | Mar 2027 |
| Initial SA ballot (D4.0) | Jul 2027 |
| Final 802.11 WG approval | Jan 2028 |
| RevCom and SASB approval | May 2028 |
That first row is this month. And at the July plenary the group demonstrated a Wi-Fi device communicating with an ambient power tag powered solely by that device's own RF emissions — the missing piece, with the tone source built into the access point itself.
In parallel, the cellular world is doing the same thing: 3GPP Release 19 standardizes Ambient IoT, where a "device 1" is a tag with roughly 1 µW peak consumption, backscattering a carrier wave supplied externally. OOK and BPSK modulation, indoor inventory and command use cases.
When both major wireless ecosystems independently write backscatter into their standards, it stops being a lab curiosity.
About that gear
One detail from the original demo still looks strange against all of the above: the mechanics. A tooth touches a contact, the signal distorts; no contact, no change.
It's honest, and rather elegant. Since transmission reduces to toggling a switch, the source of those toggles can be anything that opens and closes a circuit. A chip is optional. A gear turned by a finger closes a contact in a specific rhythm, and an encoded sequence goes out over the air. Not "Morse code for Wi-Fi" as a metaphor — a literal mechanical encoder.
Which generalizes nicely: the mechanism becomes the sensor. Water flow spins an impeller, so water is flowing. Wind turns a vane, so there's wind. A lid rotates, so the jar was opened. No electronics in the measurement path at all, energy from motion, communication from reflection.
So why isn't it in your house yet
The honest answer: the problem isn't physics, it's infrastructure and economics.
The physics was solved a decade ago and has only improved since. But for a battery-free button to work in an ordinary apartment, the access point itself has to supply the tone — which means it has to be in the standard and in the firmware of routers people actually buy. That's what 802.11bp is for, and that's why the date in its plan is May 2028, not "already".
Plus the unglamorous open questions: collision handling when there are hundreds of tags in a room; interference in an already-crowded 2.4 GHz band; how much range degrades in a real apartment with concrete and furniture rather than a lab hallway.
But the trajectory is clear. In twelve years this went from hundreds of bits per second at arm's length to shipping silicon and a draft standard. If IEEE holds its schedule, the first devices carrying a Wi-Fi logo and no battery compartment show up around 2028–2029 — not a revolution in a day, but a very quiet replacement of billions of batteries with nothing.
Originally published at klukyanov.ru.
Top comments (0)