New Wireless Technologies: LIFI and DIDO

by on August 4, 2011 at 7:34 am in Science, Web/Tech | Permalink

Two new wireless technologies were unveiled to the public in the last few weeks. Harald Haas, speaking at TED Global, demonstrated wireless through LED lighting–basically fiber optics without the fiber. The Haas demonstration is impressive and the technology appears to be close to market for local operations, such as a building, although it is unclear to me how the signal is intended to reach the socket over distance  (broadband over power lines has proven difficult to implement). Here is Haas at TED:

Meanwhile Steve Perlman of Rearden Labs (yes, that Rearden) says his new DIDO technology breaks Shannon’s Law by at least 100 times–do note that these are fighting words, it isn’t called a law for nothing. What appears to be most impressive about the Perlman technology is that it is capable of being used over significant distances. We have yet to see a demonstration at large scale, however.

Business Week has a puffy but still fascinating profile of Perlman which includes this bit:

…his parents were both physicians, and they denied him an Apple II computer. They feared he would spend all day playing video games on the blasted machine—and they were right. “I was forced to build my own computer and create a graphics display for it and then write video games that I could play,” Perlman says….

Perlman would use this self-taught ability to understand electronics and computers as a way of getting out of jams. During his senior year in high school, he skipped so many classes that he was in danger of not graduating. So he built an illuminated marquee for the drama department to secure an English credit. Then he designed a computer simulation of the forces behind swings in the U.S. economy during the 19th century for a history credit. Later, while attending Columbia University, Perlman says he took a computer-programming class and taught himself Pascal during the open-book, midterm exam.

TallDave August 4, 2011 at 8:08 am

Fascinating talk, thanks for sharing. Here’s a link that has another video and a link to the whitepaper. I haven’t been able to read it yet.

http://www.tomsguide.com/us/DIDO-Steve-Perlman-OnLive-distributed-input-distributed-output,news-12014.html

Perlman says he took a computer-programming class and taught himself Pascal during the open-book, midterm exam.

Brings back memories, when I was interviewing after getting my M.S. (this was back at the peak of the Internet bubble) this was actually one of the tests a prospective employer gave us — they asked us to write a several programs in a language they knew we’d never seen before. Those of us who could figure it out were given offers. Their offers were better than those from any of the other companies I’d interviewed with.

TallDave August 4, 2011 at 10:53 am

Ahhh, they are programmatically adjusting for the interference at every location; the magic is in the algorithm.

IIRC his is something that people have been speculating about doing for some time, but it was considered to be very very difficult. Looks like maybe they cracked it. Very elegant tech, if this is accurate. This could be the most profitable algorithm since Brin’s.

Gunnar Tveiten August 4, 2011 at 8:33 am

Shannons law is math. Breaking it, is about as likely as making 2+2 equal to 5. That is to say, not very. Slightly more likely is, however, finding ways of exploiting the fact that commonly experienced noise is not, infact, perfectly gaussian. (which Shannons Law assumes that it is)

There’s been a long history of people who *believe* they’ve broken Shannons law, similar to the history of people who believe, or atleast claim, to have invented a perpetum mobile.

Mike Huben August 4, 2011 at 8:47 am

Reading the DIDO white paper, it looks as if the system doesn’t break Shannon’s Law, but instead produces a holographic image around the receivers with the appropriate signal at each receiver. It can do this because there is a separate broadcast antenna at a different location for each receiver at a different location.

As the article explains, Shannon’s law applies to single channels within some bandwidth. Creating holographic images within some bandwidth is quite a different matter.

We might expect some problems as antennae and receivers get really close together.

The article makes no mention of the term “holographic”. That is my own inference.

Bob Knaus August 4, 2011 at 9:29 am

Holograph is a good analogy. The method is very compute-intensive, and must respond to dynamic changes in the noise enviroment.

It certainly doesn’t break Shannon’s Law. There is likely a mathematical limit to the bandwidth that can be extracted given a particular noise profile, just as there is a limit to the resolution of a hologram given a particular loss of image data.

Sean August 4, 2011 at 10:45 am

I had to Google ‘Shannon’s Law’. This version is more entertaining:
http://en.wikipedia.org/wiki/Shannon%27s_law_(Arizona)

Silas Barta August 4, 2011 at 11:03 am

Interestingly, the Arizona Shannon’s Law does _not_ depend on whether your random discharging of a firearm into the air follows a Gaussian distribution.

Tim August 4, 2011 at 1:29 pm

It sounds like they are doing a MIMO system, but instead of doing the matrix inversion at each receiver, the transmitter is precoding the signal so that it works itself out when it gets there.

There must be substantial overhead transferring the channel information up to the transmitter all the time though, especially if the receiver is in a dynamic environment (either the receiver is moving or things around it are moving)

Someone August 4, 2011 at 10:05 pm

Wouldn’t surprise me if there wasn’t overhead from maintaining client ‘locations’ (wouldn’t surprise me if the algorithm never actually calculates position and instead just reverses distortion of a test signal to start). As a client or its environment moves the system will have a strong change in in the signals the APs receive. I hesitate to call this noise because its source would be the client signal distorted by the movement from which they can compensate for the change. The whitepaper’s too handwavy for me to be sure this is how they do it but if I’m right about this then the overhead of updating for a dynamic environment would scale based on the amount of movement.

Sam August 5, 2011 at 10:52 pm

Inverse transfer functions, people… but much more complicated than others I’ve heard of since the APs aren’t in regular arrays. (e.g. using stereo speakers to produce all-around sound, active electrically scanned array radars which can form multiple beams to paint multiple targets simultaneously),

I suspect they have to have very good time synchronization so that the various APs transmit with proper interference.

Andrew' August 4, 2011 at 9:52 am

I wonder if people should co-locate based on content preferences.

babar August 4, 2011 at 10:01 am

that’s the real reason people who like the same media are friends with each other.

Alex R August 4, 2011 at 10:59 am

You know, if you say your invention violates well-understood principles of mathematics or engineering, you’re guaranteed to bring out the cranks…

DIDO, as described by the white paper, violates no such principles, though how practical it is I don’t know. Despite his efforts to differentiate his invention from MIMO, it is a classic MIMO (multiple input multiple output) technique, using the spatial separation between transmitters and between receivers to create multiple channels. The limitations will depend on the processing power available, the accuracy of time synchronization, the distance between the receivers, and that sort of thing.

What really matters is not the theory, which is straightforward and unexceptional, but the quality of the implementation. I wish them luck.

TallDave August 4, 2011 at 2:58 pm

Looks like the main difference is the centralization of signal processing.

http://en.wikipedia.org/wiki/DIDO_%28network%29#Multi-user_to_multi-user

I’ll be interested to see how it scales.

FYI August 4, 2011 at 11:13 am

On the video he mentiones a new patent law and how that will impact our start-ups. Does anyone here know exactly what the core problem with the new law is?

Cliff August 4, 2011 at 12:40 pm

Well the law has not gone through reconciliation, so we don’t know exactly what it will be yet. However some of the changes include a move to granting patent rights to the first to file a patent application rather than first to invent. There is also an additional procedure for challenging patents. These are generally things that hurt small entities and start-ups, and patents are often the biggest assets start-ups have and a major magnet for financing. .

Thomas Bayes August 4, 2011 at 11:34 am

These are not fighting words, but, instead, they are words intended to attract attention to Perlman’s company.

Let me give an analogy: when applied to a standard phone line, Shannon’s Law says that the fastest speed at which a computer can communicate is about 56 thousand bits per second. With DSL you can communicate much faster over your phone line. Does DSL break Shannon’s Law? Absolutely not! It is a different technology, so the application of ‘the Law’ gives a different limit.

The mathematical fact that Shannon discovered doesn’t have anything to do with a particular technology. We can — and do — apply the fact to particular technologies, but changing technologies is not the same as ‘breaking Shannon’s Law’.

Perlman’s technology may be revolutionary, but I wish he wouldn’t resort to ridiculous hyperbole to promote it. (I am assuming Perlman actually understands that he hasn’t found a way to violate The Mathematical Theory of Communication.)

Professor, if you haven’t done so already, please read this sometime . . .

http://www.amazon.com/Mathematical-Theory-Communication-Claude-Shannon/dp/0252725484

Thomas Bayes August 4, 2011 at 11:37 am

As an interesting side note, Shannon titled his original paper “A Mathematical Theory of Communication.” For this book, the word ‘A’ was changed to ‘The’. I wonder if Perlman understands why?

Scoop August 4, 2011 at 1:58 pm

Don’t you read Tyler? Internet communications are irrelevant to human welfare. This means nothing.

Rahul August 4, 2011 at 3:43 pm

Interesting that so many comments about DIDO but none about LIFI. Does the reflect lesser utility? I sure thought LIFI had a “cool” factor but can’t think of much utility.

Rahul August 4, 2011 at 4:29 pm

Philosophically, the DIDO stuff resembles centralized planning efficiencies in some sense. If me and my neighbor both run our own private, totally independent, Access Points we reduce our usable bandwidth to half. OTOH DIDO requires us to cooperate and combine our signals in a centralized calculation unit and then use both our access points to get double the bandwidth of the original independent transmissions. Collaboration increases efficiency. It does raise the issue of who would I trust to run the DIDO processor; there are obvious security and data confidentiality issues.

Steve Perlman’s whitepaper made interesting reading and diagrams about this.

http://dl.dropbox.com/u/28994529/110727-DIDO-A%20New%20Approach%20to%20Multiuser%20Wireless-FINAL.pdf

Thank goodness that it was a whitepaper and not an esoteric journal article.

Colin August 5, 2011 at 1:50 am

LIFI: Sounds niche, at best, because of a complete change in infrastructure (and not just adding infrastructure to existing lighting, but a swapping out of existing lighting for new).

DIDO: As already pointed out, no law is broken. Perlman. Is. WRONG.

What I haven’t seen pointed out yet, is that to properly form the channels you need as many AP’s as users (read the white paper). Did that sink in to anyone else? For a stadium with 100,000 people you need…100,000 AP’s. If an AP is $100 then that’s $10M just for the stadium and they must be just as geographically distributed in the stadium as the users (think one under every seat). And for Manhattan? Yikes. If you keep it fixed at one AP then you are absolutely bound to Shannon’s law and every user added decreases bandwidth.

Both wireless solutions are doing the same thing: more infrastructure for more capacity. Very, very old trick with novel solutions (kinda). Nothing to really see here, move along.

Comments on this entry are closed.

Previous post:

Next post: