Wi-Fi: how CSIRO beat the echo

The full story

This is the episode's narration, word for word. Headings jump to that point in the video.

Told honestly 0:00

This season is about things Australians invented or discovered, told honestly. What each one solved, how it works, who shared the credit, and what came before. We'll start with the famous one. You've probably heard that Australia invented Wi-Fi. That's an overstatement. The true story is narrower, and better. It's about one stubborn problem: how do you send data fast through a room full of echoes?

Listening for an exploding black hole 0:30

It starts in radio astronomy. In 1974, Stephen Hawking predicted that black holes slowly leak energy, and that a small one could end in an explosion. Martin Rees suggested such an explosion might give off a brief flash of radio waves. In the Netherlands, a Sydney-trained engineer called John O'Sullivan, with Ron Ekers and Peter Shaver, went looking for flashes lasting millionths of a second. They found none, and in 1978 they published the limits instead.

What he brought home 1:05

What he did bring home to CSIRO, Australia's national science agency, in 1983, was a skill. Radio astronomers lean heavily on a piece of maths called the Fourier transform, and he had mastered it on paper and in hardware. With a company called Austek, his group built a chip that did Fourier transforms very fast. The official story, in O'Sullivan's prize citation, is that the black hole search led to Wi-Fi. The more accurate version is that the Fourier toolkit did.

What a Fourier transform does 1:43

So what does a Fourier transform do? Any signal, however messy, can be built by adding up pure, smooth waves of different frequencies. The Fourier transform runs that in reverse. Give it the messy signal, and it tells you which pure waves are in it, and how strong each one is. It's like hearing a chord and naming every note. A fast way to calculate it was published by Cooley and Tukey in 1965. Hold on to that idea, because it comes back in a big way.

Cut the cable 2:18

By the late 1980s, O'Sullivan was thinking about computer networks. Office computers were tied to their desks by cables. If you could just cut the wires, he said later, there would be huge potential. The CSIRO team set out to match the speed of the best wired networks. A wireless office network Motorola sold managed only about 3 to 6 megabits a second, according to CSIRO's patent. So what was holding everyone back?

A room full of echoes 2:51

Echoes. Radio waves bounce off walls, floors, ceilings and furniture, so a receiver gets the direct signal, then a string of weaker copies arriving a little later. How much later? Radio travels at the speed of light, about 30 centimetres in a billionth of a second. The CSIRO patent puts typical indoor echo delays at around 50 nanoseconds. That's a bounce that travels about 15 metres further than the direct path.

Why echoes cap the speed 3:25

Fifty billionths of a second sounds harmless. It isn't. Data is sent as a stream of symbols, one after another. If each symbol is shorter than the echo delay, the echoes of one land on top of the next. The team was aiming for about 100 megabits a second, where each bit lasts just 10 nanoseconds. So every bit would be smeared across the next four or five. It's like talking fast in a cathedral. Slowing down fixes it, but throws away the speed you wanted.

Many slow lanes 4:02

Others tried filters that subtract echoes, or switching between directional aerials, and the patent argues both were too bulky and costly. CSIRO's idea was to split the fast stream into many slow ones, and send them all at once, each on its own radio frequency. Each lane is slow enough that its symbols outlast the echoes, but together they carry the full speed. In the patent's example, twelve lanes stretch each symbol to about 120 nanoseconds, comfortably longer than the echoes.

Lanes that overlap 4:39

You'd think those lanes need empty space between them. They don't, and this part isn't Australian. In 1966, Robert Chang at Bell Labs described, and filed a patent on, lanes whose frequencies overlap without interfering. They're arranged so that when the receiver listens for one lane, all the others add up to exactly zero. It's now called orthogonal frequency division multiplexing, or OFDM. In 1971, Weinstein and Ebert showed a Fourier transform could make all the lanes at once.

Fourier does the heavy lifting 5:18

This is where the Fourier toolkit pays off. The transmitter runs an inverse Fourier transform, adding every lane into one combined wave in a single calculation. The receiver runs a Fourier transform and pulls the lanes apart again, like naming the notes in that chord. A team that had built fast Fourier hardware for telescopes was well placed to make that practical.

A buffer for stragglers 5:45

There's one more trick. Even slow symbols get their start smudged by the tail of the one before. So each symbol is padded with a short copy of part of itself. In Wi-Fi, that's a copy of its ending, stuck on the front. Late echoes fall into the padding, and the receiver ignores it. It's called a guard interval, and in the patent's design it was a quarter as long as the symbol. Using a copy, rather than a gap, keeps the overlapping lanes from bleeding into each other.

Holes in the spectrum 6:20

Echoes cause a second problem. At some frequencies, the echoes arrive out of step with the direct signal and cancel it. So the signal has deep dips across the band, and any lane in a dip is nearly wiped out. Worse, the dips move whenever someone walks across the room. Telephone lines have dips too, but theirs drift slowly enough to correct for.

Shuffle before you deal 6:46

So CSIRO added two well-known tricks. One is error correction: extra bits that let the receiver rebuild a few that go missing. The other is interleaving: shuffling the bits so that neighbours in a message travel on lanes far apart. When a dip wipes out a couple of lanes, each message loses only a scattered bit or two, and the code repairs them. It's like shuffling a deck before dealing, so losing one hand doesn't cost anyone a whole suit.

What the patent claims 7:20

CSIRO filed for an Australian patent in November 1992, and the US patent was granted in January 1996. The inventors were John O'Sullivan, Graham Daniels, Terry Percival, Diet Ostry and John Deane. The claims the courts relied on describe a combination: many slow lanes with symbols longer than the echoes, plus error correction, plus interleaving, in a confined space full of echoes. CSIRO says their system took indoor wireless from 10 megabits a second to more than 50.

Into the standard 8:02

Meanwhile, the IEEE, an engineering standards body, was writing the standard we now call Wi-Fi. In 1999, a version called 802.11a adopted OFDM, and in 2003, 802.11g brought it to the 2.4 gigahertz band. 802.11a sends 48 data lanes, each symbol four millionths of a second long including its guard, which works out to a top speed of 54 megabits a second. In 1998, CSIRO told the standards group its patent might be needed, and promised licences on reasonable terms.

Taking on the industry 8:49

In 2005, CSIRO sued the network maker Buffalo in Texas over its 802.11a and 802.11g products. Intel, Dell, Microsoft, Hewlett-Packard and Netgear went to court to have the patent struck down. In 2006 the judge ruled for CSIRO, and in 2007 he barred Buffalo from selling those products in the United States. In 2008, the appeals court agreed Buffalo infringed, but said whether the invention was obvious needed a trial.

The settlements 9:31

In April 2009 the fight finally reached a Texas jury, but it never got to a verdict. Hewlett-Packard settled first, then, part-way through the trial, more than a dozen others, including Microsoft, Intel and Dell. Reports put the total at over 200 million Australian dollars. In 2012, the big American phone carriers and several tech firms settled for more than 220 million. CSIRO puts its total income from its Wi-Fi licences at around 430 million. Critics called it Australia's biggest patent troll. Announcing the 2012 deal, the science minister said companies selling billions of devices should pay for the technology they used.

What came from elsewhere 10:22

So, did Australia invent Wi-Fi? No. OFDM came from Bell Labs. Lanes plus error correction plus interleaving were already in a 1989 paper on digital radio for cars, and CSIRO's own expert acknowledged it. That paper wasn't about indoors, and whether that step was obvious was never settled by a jury. The Wi-Fi standard came from the IEEE, with roots in WaveLAN, a project NCR and AT&T ran in the Netherlands. And the name Wi-Fi was invented by a branding firm, and stands for nothing.

What was Australian 11:05

What a CSIRO team did was find a way to beat indoor echoes at wired-network speeds, build it, and patent it. A spin-off company, Radiata, made what O'Sullivan's prize citation calls the world's first 802.11a chips. O'Sullivan won the Prime Minister's Prize for Science in 2009, and the team a European Inventor Award in 2012. So here's the honest sentence. Australia didn't invent Wi-Fi, but a CSIRO team found, and patented, a key way to make it fast indoors.

Next time 11:48

Next time, a different kind of listening. When a plane goes down and nobody survives, how do investigators find out what happened in the cockpit?

Sources

Every factual claim in the episode is tied to one of these. Spotted an error? Tell us.

1. US Court of Appeals for the Federal Circuit, CSIRO v. Buffalo Technology (USA),… — cafc.uscourts.gov
2. US Patent 3,488,445, R. W. Chang, "Orthogonal frequency multiplex data transmission… — patents.google.com
3. C. Hetting, "Vic Hayes & Bruce Tuch inducted into the Wi-Fi NOW Hall of Fame", Wi-Fi… — wifinowglobal.com
4. S. W. Hawking, "Black hole explosions?", Nature 248, 30–31 (1974) — doi.org
5. M. J. Rees, "A better way of searching for black-hole explosions?", Nature 266,… — doi.org
6. Prime Minister's Prizes for Science 2009: "John O'Sullivan: How astronomy freed the… — web.archive.org
7. J. D. O'Sullivan, R. D. Ekers, P. A. Shaver, "Limits on cosmic radio bursts with… — doi.org
8. J. P. Hamaker, J. D. O'Sullivan, J. E. Noordam, "Image sharpness, Fourier optics, and… — doi.org
9. A. Moses, "CSIRO to reap 'lazy billion' from world's biggest tech companies", The Age,… — theage.com.au
10. HyperPhysics (R. Nave, Georgia State Univ.): "Fourier Analysis and Synthesis" — hyperphysics.phy-astr.gsu.edu
11. J. W. Cooley, J. W. Tukey, "An algorithm for the machine calculation of complex… — doi.org
12. US Patent 5,487,069, "Wireless LAN", J. D. O'Sullivan, G. R. Daniels, T. M. P.… — patents.google.com
13. NIST CODATA: "speed of light in vacuum" (299 792 458 m/s) — physics.nist.gov
14. R. W. Chang, "Synthesis of band-limited orthogonal signals for multichannel data… — doi.org
15. Wikipedia: "IEEE 802.11a-1999" (52/48/4 sub-carriers, 312.5 kHz, 4 µs symbol incl. 0.8… — en.wikipedia.org
16. S. B. Weinstein, P. M. Ebert, "Data transmission by frequency-division multiplexing… — doi.org
17. H. Takanashi, M. Morikura (NTT), R. van Nee (Lucent), "OFDM Physical Layer… — ieee802.org
18. CSIRO: "Bringing fast wireless connectivity to the world" (WLAN history, team, A$430m,… — csiro.au
19. CSIRO news release, 3 Aug 2016: "Revolutionary Wifi invention selected for World in… — csiro.au
20. IEEE 802.11 Working Group: "Official IEEE 802.11 Working Group Project Timelines"… — ieee802.org
21. Wikipedia: "IEEE 802.11g-2003" ("operates in the 2.4 GHz microwave band"; "The… — en.wikipedia.org
22. D. N. Cooper (CSIRO), letter to V. Hayes, Chair IEEE P802.11, 4 Dec 1998 (IEEE patent… — standards.ieee.org
23. A. Moses, "CSIRO wins landmark legal battle", The Age, 16 Nov 2006 — theage.com.au
24. "Buffalo allowed to sell wireless products again (for now)", TechCrunch, 3 Dec 2008… — techcrunch.com
25. C. Duckett, "Microsoft, Fujitsu, Asus settle with CSIRO", ZDNet, 15 Apr 2009 — zdnet.com
26. CourtListener record: CSIRO v. Buffalo Technology Inc., 492 F. Supp. 2d 600 (E.D.… — courtlistener.com
27. R. Miller, "CSIRO's patent lawsuits conclude with the final 13 companies settling",… — engadget.com
28. J. Mullin, "How the Aussie government 'invented WiFi' and sued its way to $430… — arstechnica.com
29. M. Schubert, "Australian scientists cash in on Wi-Fi invention", Sydney Morning… — smh.com.au
30. R. Chirgwin, "Champagne at CSIRO after WiFi patent settlement", The Register, 1 Apr 2012 — theregister.com
31. "CSIRO wins legal battle over wi-fi patent", ABC News, 1 Apr 2012 — abc.net.au
32. B. Popper, "Australia's Biggest Patent Troll Goes After AT&T, Verizon and T-Mobile",… — web.archive.org
33. Wikipedia: "Wi-Fi", History section (WaveLAN 1991; 802.11 1997 at 2 Mbit/s; 802.11b… — en.wikipedia.org
34. C. Doctorow quoting Phil Belanger (founding member, Wi-Fi Alliance), "WiFi isn't short… — boingboing.net
35. European Patent Office, "EPO's European Inventor Award 2012 goes to outstanding… — web.archive.org

Researched and scripted with AI assistance, fact-checked claim by claim, with synthetic narration and diagrams drawn in code. How we make episodes.