Join 3,556 readers in helping fund MetaFilter (Hide)


AM? FM? BM?
June 29, 2011 11:06 AM   Subscribe

In perhaps the lowest key announcement of a world-changing breakthrough ever, Steve Perlman, CEO of technology research incubator Rearden Labs (and the once controversial, now successful, OnLive gaming network), claimed during a presentation to Columbia Engineering students on June 6, 2011 that Shannon's Law ... need not apply.

Shannon's Law dictates how much information can be sent at a given time at a given transmission frequency. It's why voice calls sound tinny (to save bandwidth), why TV/radio towers needed to be spaced widely apart (to avoid interference), and why putting too many cell phones in the same room means no one can place a call (frequency congestion).

It also dictates the value -- in billions of dollars -- of one of the most precious resources of the Internet Era: our wireless spectrum.

The presentation only gives scattered hints as to how it is done, but if the claim is correct then Rearden's technology may uncap the world's wireless data.
posted by seanmpuckett (98 comments total) 27 users marked this as a favorite

 
(shoots random firearms into the air)
posted by Afroblanco at 11:08 AM on June 29, 2011 [79 favorites]


Looks like someone didn't disambiguate.
posted by tommasz at 11:10 AM on June 29, 2011 [4 favorites]


Meat starts at 55:13 into the video.
posted by seanmpuckett at 11:10 AM on June 29, 2011 [1 favorite]


BM?

Heh heh heh. I see what you did there.

Otherwise, if this is true it is freaking awesome.
posted by Mister Fabulous at 11:11 AM on June 29, 2011


Thanks seampuckett I was about to faint
posted by elpapacito at 11:11 AM on June 29, 2011 [1 favorite]


This will make their merger with Taggart Transcontinental all the more valuable!
posted by griphus at 11:15 AM on June 29, 2011 [25 favorites]


So, Reardon Labs has developed a radical, unproven new technology that could completely revolutionize our infrastructure. Are we 100% sure this isn't some sort of viral ad for the next Atlas Shrugged movie or should we wait to see if Perlman suddenly disappears into an Objectivist Wonderland first?
posted by Copronymus at 11:16 AM on June 29, 2011 [8 favorites]


Text messages will still be overpriced.
posted by furiousxgeorge at 11:17 AM on June 29, 2011 [15 favorites]


Gigapan pics or it didn't happen.
posted by fuq at 11:17 AM on June 29, 2011


"I know there's going to be an article published that..." (aggravated noise)
posted by boo_radley at 11:18 AM on June 29, 2011


You know how they do?

VOLUME.
posted by blue_beetle at 11:19 AM on June 29, 2011 [1 favorite]


Ok, so the guy *says* they have a radio system running at 10x the Shannon limit for the shared channel but doesn't provide any explanation of how it works or any demo of it working? Yeah...
posted by Juffo-Wup at 11:19 AM on June 29, 2011 [1 favorite]


I dont see the meat here. There's a couple slides announcing a radical new thing which promises to change everything and...thats it.
posted by vacapinta at 11:20 AM on June 29, 2011


I believe it. Verizon is no longer installing FIOS (fiber to the curb), the wireless technology is going to leapfrog it.
posted by stbalbach at 11:22 AM on June 29, 2011


A CEO is the best person to deliver news of world-changing and hugely lucrative technologies. They have no incentive to hype the shit out of it and leave out caveats.

(See Codename: Ginger)
posted by benzenedream at 11:24 AM on June 29, 2011 [6 favorites]


So, Reardon Labs has developed a radical, unproven new technology that could completely revolutionize our infrastructure. Are we 100% sure this isn't some sort of viral ad for the next Atlas Shrugged movie or should we wait to see if Perlman suddenly disappears into an Objectivist Wonderland first?

I was curious and Wikipedia says his company was originally named Rearden Steel. So, yeah, there's that
posted by crayz at 11:25 AM on June 29, 2011 [7 favorites]


Even if he's right, ATT and Verizon would never deploy the technology in the US, and will crush the life out of any other company that tried to end run around them. Rent seekers seek rent, and those rentiers are the biggest rent seekingest rent seekers that ever sought rent.
posted by T.D. Strange at 11:26 AM on June 29, 2011 [14 favorites]


The Congressional bill he's talking about at around 1:01:00 passed last week. Obama hasn't signed it into law yet, but it looks like a done deal, which is crazy considering it appears almost blatantly unconstitutional.
posted by nushustu at 11:26 AM on June 29, 2011


To be fair to Rearden Labs, when they announced OnLive there were people who insisted that it was technically impossible / highly implausible and that actually works pretty well.
Still, this claim is of another order altogether. Frankly the fact that this is an offhand mention in a lecture given to students makes it more likely to be an actual thing than if this was a glitzy product announcement.
posted by atrazine at 11:26 AM on June 29, 2011 [1 favorite]


So on one hand, we have a theory proposed by one of the greatest minds of the 20th century, widely verified by experiment and observation, and proven in multiple ways with regards to existing theory.

On the other hand, we have the unverified claim of an "entrepreneur and inventor devoted to pioneering Internet, entertainment, multimedia, consumer electronics and communications technologies and services," given during a presentation to a bunch of students, apropos of nothing.

Yeeeeah. This is my skeptical face.

Not that I wouldn't like it to be true, but I think it's more likely that Mr Perlman is either misunderstanding the Shannon-Hartley theorem (which is possible as it has some nuances; e.g. the most common form of it assumes Gaussian noise), or somewhere along the line somebody misread the data.
posted by Kadin2048 at 11:28 AM on June 29, 2011 [12 favorites]


Look. You can't beat chemistry. Because you can't beat physics. Any you know why you can't beat physics?

Because you can never. ever. beat math.
posted by GuyZero at 11:31 AM on June 29, 2011 [7 favorites]


Because you can never. ever. beat math.

Ahem.

GOD is applied POWER
which is applied GOVERNMENT
which is applied POLITICS
which is applied ADVERTISING
which is applied SOCIOLOGY
which is applied PSYCHOLOGY
which is applied BIOLOGY
which is applied CHEMISTRY
which is applied PHYSICS
which is applied MATH
which is applied PHILOSOPHY
which is applied BULLSHIT

Bullshit > Philosophy > Math.
posted by Mister Fabulous at 11:34 AM on June 29, 2011 [6 favorites]


We've seen this kind of claim before. A few years ago there was a startup who claimed they had come up with a new compression algorithm which could yield drastically better compression than Huffman Encoding.

Huffman Encoding can approach within a few percent of the Shannon limit, and these guys were claiming they could break past the Shannon limit. They didn't provide any details, or give any demonstrations, and we never heard anything about them again.
posted by Chocolate Pickle at 11:37 AM on June 29, 2011 [1 favorite]


To be fair to Rearden Labs, when they announced OnLive there were people who insisted that it was technically impossible / highly implausible and that actually works pretty well.

That is true only for extremely restricted values of "pretty well". It's fine for some things and it works much better than people thought it would (which was not at all) but in my limited experience it has many of the lag and latency issues that people said it would have. And I have no idea if it's "successful" as sated in the FPP or not -- at least economically. I think the jury is still out.
posted by The Bellman at 11:39 AM on June 29, 2011 [2 favorites]


The comments at YouTube are devotional. That's probably a bad sign.
posted by Lemurrhea at 11:39 AM on June 29, 2011 [4 favorites]


I don't really understand the technology involved, but keep in mind that he doesn't say that shannon's limit is false, only that rearden has managed to work around it somehow. also keep in mind the previous metafilter thread about onlive in which there were probably too many casually dismissive skeptics. :) I think perlman is a pretty smart guy. it is pretty creepy that the company is named after an atlas shrugged character though.
posted by Post-it Goat at 11:41 AM on June 29, 2011


Possible Apocryphal Story Time (wherein I may be misremembering things):

When I took my Communications I and II (EE, not Speech Comm) we went over in agonizing detail all the math underlying FM radio. My instructor said that at first engineers thought with FM you could squeeze ever more data into an ever vanishing amount of bandwidth. And indeed, if you look at the frequency plot it appears as if the only thing limiting the capabilities of FM is your imagination.

But it's a trick. If you analyze the math more carefully, you see that FM actually requires more bandwidth to transmit the same signal as AM. Actually, it requires infinite bandwidth but outside a certain frequency range the power is negligible.

I suspect something similar is happening here.
posted by sbutler at 11:43 AM on June 29, 2011 [2 favorites]


Given the connection between Shannon's Law and the laws of thermodynamics this is ... interesting
posted by fallingbadgers at 11:44 AM on June 29, 2011 [4 favorites]


There isn't any way to violate the Shannon–Hartley theorem per se, but one might always :

(1) Find more unimportant information to ignore, i.e. improved even-more-lossy compression. Or prioritize information so that recipients may ignore parts or enjoy it sooner, ala a jpg displaying a grainy image before all the details complete downloading, maybe one might call this "interactive lossy compression".

(2) find ways to squeeze more information from your physical channel, as opposed to the mathematical abstraction of a channel discussed by the Shannon–Hartley theorem, like say using unpopular parts of the spectrum.

If they've identified new tricks for now showing stuff your brain ignores anyways, well that's great, although the audiophiles won't like it. If they're using more spectrum, well maybe that's great, or maybe it's illegal.
posted by jeffburdges at 11:45 AM on June 29, 2011 [2 favorites]


Great. License it to Atheros and then it will end up in Ubiquiti, and then we'll actually be getting somewhere.

And rent seeking telcos won't stop it outright, just contract exclusively for the technology.

Although, my local broadcast station might be due for a decent bump in valuation if they can stuff more HD broadcasts into their frequency space. Over the air cable TV lineups anyone?
posted by dglynn at 11:45 AM on June 29, 2011 [1 favorite]


Ok, bugger all this speculation. I'm going to pull his patent applications. There's no way a serial inventor like him wouldn't have filed before talking about this. (Especially and ironically with the upcoming change to first-to-file that he mentioned in his talk)
posted by atrazine at 11:47 AM on June 29, 2011 [2 favorites]


Still, this claim is of another order altogether.

Yes, extraordinary claims demand extraordinary proof. Shannon-Hartley has a huge body of proof, both mathematical and practical, behind it. Most of the so-called disproofs involved schemes that were more error prone than claimed, or used more bandwidth than they realized.


So, yeah, I don't believe it, certainly not with only one video as the evidence.
posted by eriko at 11:48 AM on June 29, 2011 [2 favorites]


Look, he might be wrong about what he can do. At this point he almost certainly is wrong. But the video is part of a lecture to engineering students, and it currently has 750 views. He's not saying what he's saying to make a big PR impact. Or if that was his aim, he's terrible at it. He genuinely believes he's figured out how to do this crazy, almost definitely impossible thing.
posted by penduluum at 11:50 AM on June 29, 2011


I'm still puzzling it through. If it isn't AM and it isn't FM then it has to be some other modifiable attribute of a photonic signal. Polarization? Phase? Or even lower down... wavelets? That'd be some crazy shit right there. I'm no EE, though. He says the antennas and electronics are simple, which kind of argues in favour of wavelets or some other kind of packetizing. He also says the stuff travels around the earth's curvature, so whatever it is it is either boundary coupled or bounces off the ionosphere.

The guy's presentation was basically not about this announcement but, "how does one innovate?" So if they stopped asking the question "how do we squeeze more data into an FM band" and started asking the question "what is a new way we can modulate RF photons in order to carry data" then I am thinking he may be on to something.
posted by seanmpuckett at 11:50 AM on June 29, 2011


There's a recent series of patents to do with Distributed-Input, Distributed-Output communications, something to do with user clustering?
Maybe he's doing some RF equivalent of multicast for commonly demanded content?
posted by atrazine at 11:50 AM on June 29, 2011 [4 favorites]


Evidence or GTFO.
posted by rmd1023 at 11:52 AM on June 29, 2011


It could be a form of spatial multiplexing. That really does let you evade the Shannon limit, by effectively increasing the number of channels between two points.
posted by topynate at 11:52 AM on June 29, 2011 [1 favorite]


His recent patent applications look at the late 2010 and 2011 applications.

(God do I ever hate patent-ese)
posted by atrazine at 11:52 AM on June 29, 2011 [1 favorite]


I am putting out an Amber Alert for Shannon's Law.
posted by fourcheesemac at 11:53 AM on June 29, 2011 [2 favorites]


An interesting data point to consider is compressed sensing

Basically it states that you can reconstruct a signal fully from a set number of random points in the frequency domain if your signal has certain characteristics, mainly if it mostly sparse over the entire frequency domain.

Here's a list of good resources for the mathematical.
posted by The Power Nap at 11:54 AM on June 29, 2011 [2 favorites]


I dont see the meat here. There's a couple slides announcing a radical new thing which promises to change everything and...thats it.

It's a scooter that doesn't tip over.
posted by steambadger at 11:55 AM on June 29, 2011 [11 favorites]


whoops, forgot to add the single pixel camera
posted by The Power Nap at 11:55 AM on June 29, 2011 [2 favorites]


A CEO is the best person to deliver news of world-changing and hugely lucrative technologies

I worked with Steve for several years, and yes, he's a pitch man that Billy Mays would have been proud of, but at his core he is an engineer, and a pretty sharp one at that. You've heard of Quicktime? At WebTV he proved to be as shrewd at the business side of things as the tech -- he designed the Solo2 system-on-chip used in the diminutive set-top boxes which, in 1995, had an OS, full network stack, and HTML engine supporting js, audio and video all running in just 2MB of RAM -- and sold the company for the highest value tech acquisition at the time, to Microsoft.

When he left to start Rearden Steel, a lot of the great talent from WebTV went with him, while others started offshoot companies which you may be familiar with that he supported either directly or indirectly.

He's always starting or selling something under the Rearden umbrella, and I take everything he says with a grain of salt, but I wouldn't totally put these claims past him.
posted by ancillary at 11:57 AM on June 29, 2011 [4 favorites]


It could be a form of spatial multiplexing. That really does let you evade the Shannon limit, by effectively increasing the number of channels between two points.

Oho. I bet that you're right.

Here's a a patent granted in 2010:

System and method for spatial-multiplexed tropospheric scatter communications
A method is described comprising: transmitting a training signal from each antenna of a base station to each of a plurality of client devices utilizing tropospheric scatter, each of the client devices analyzing each training signal to generate channel characterization data, and transmitting the channel characterization data back to the base station utilizing tropospheric scatter; storing the channel characterization data for each of the plurality of client devices; receiving data to be transmitted to each of the client devices; and precoding the data using the channel characterization data associated with each respective client device to generate precoded data signals for each antenna of the base station; and transmitting the precoded data signals through each antenna of the base station to each respective client device.
posted by atrazine at 11:57 AM on June 29, 2011 [10 favorites]


Also, he helped develop Broadcom software modems? I hated this guy and didn't even know who he was.
posted by dglynn at 11:59 AM on June 29, 2011 [3 favorites]


maybe one might call this "interactive lossy compression".

That is brilliant.

I've always wondered why the default image-rendering behavior in a web browser had to be top down rather than from the spreading out from the center of the image. Think of how typically the top edge is clouds or sky or trees, for example.
posted by fourcheesemac at 11:59 AM on June 29, 2011 [1 favorite]


fourcheesemac,

There's actually even better ways of doing it. Some compression algorithms will render low resolutions first and the resolution gets better as the full data is downloaded. I've always wanted to see that in a video streaming algorithm, degrade the resolution and keep playing without an irritating break to buffer.
posted by atrazine at 12:05 PM on June 29, 2011 [3 favorites]


I've always wondered why the default image-rendering behavior in a web browser had to be top down

It's not the browser, but the image encoding. A camera will produce a sequential image because it's quicker to do and (probably) corresponds to the way it sampled the sensor. But standard jpeg libraries will let you encode it as progressive instead, and a great many image manipulation programs will let you select that for output. This was more popular in modem days, as it allowed users to see a low-fi version of the entire image early in the download process, which progressively got better. However, it's not popular now because for any given final quality the file size is actually larger.
posted by George_Spiggott at 12:09 PM on June 29, 2011


It sounds like it is some form of spatial multiplexing: current approaches to doing that require multiple aerials to get the multiple channels. If they've come up with a way to do it from a single aerial then that's genuinely new.
posted by pharm at 12:12 PM on June 29, 2011


If they've identified new tricks for now showing stuff your brain ignores anyways, well that's great, although the audiophiles won't like it. If they're using more spectrum, well maybe that's great, or maybe it's illegal.

He seems to make it pretty clear he's not talking about compression. He claims he's talking about a fundamentally different way to "use" spectrum -- could be a new multiplexing protocol like (like TDMA/CDMA) but I think from that patent application it's something else. He says it has range beyond the curvature of the Earth, which is consistent (from this total layman's reading of an article in Wikipedia) with the tropospheric scatter reference in the patent application. But if that Wiki article is at all accurate a system like this is going to have some trouble remaining "clean" as it scales. Note that, as far as I can tell, if that's what he's talking about, he's not "breaking" the Shannon limit, he's redefining what a "unit" of spectrum is.

Disclaimer: I have no idea what I'm talking about and I'm just looking shit up on the Internet. I'm the farthest thing from an engineer. Someone will come along and tell me this soon.
posted by The Bellman at 12:13 PM on June 29, 2011


So on one hand, we have a theory proposed by one of the greatest minds of the 20th century, widely verified by experiment and observation, and proven in multiple ways with regards to existing theory.

This is the same logic that was applied when Svante Arrhenius said salt dissociated completely in water, rather than simply disolving, a la sugar. (I mean look at these melting points!)

Of course, Arrhenius had a whole PhD disertation with data and stuff (which beats three powerpoint presentations and a pair of aces) so that's not to say this isn't pure fiction with a layer of wild fantasy thrown in.
posted by Kid Charlemagne at 12:13 PM on June 29, 2011


I will not believe Perlman's claims that Rearden Labs has created a cold-fusion device until they transparently demonstrate the science behind how they produce 10x more energy than the input!
posted by StrangerInAStrainedLand at 12:13 PM on June 29, 2011


There's actually even better ways of doing it. Some compression algorithms will render low resolutions first and the resolution gets better as the full data is downloaded. I've always wanted to see that in a video streaming algorithm, degrade the resolution and keep playing without an irritating break to buffer.

Netflix does that.
posted by kafziel at 12:15 PM on June 29, 2011 [1 favorite]


Please oh please let this be the death knell for regional monopolies like AT&T and Comcast.
posted by BrotherCaine at 12:15 PM on June 29, 2011 [1 favorite]


There may be an issue here with power requirements. However, Cold Fusion should solve that.
posted by seanyboy at 12:20 PM on June 29, 2011 [4 favorites]


There are several possible approaches to your desired video streaming algorithm, atrazine.

You might imaging designing the algorithm so that, no matter what UDP packet gets dropped, some reasonable lower quality image appears. I'd expect such algorithms would first fill-in the missing information from other recent frames. And I vaguely suspect it'd cost you additional bandwidth over the long haul if you went much beyond doing that, but maybe not, or not much.

Alternatively, you might design your algorithm to take advantage of the network's quality-of-service tools, i.e. tell the backbone routers which UDP packets are less important.

There is definitely room for better lossy compression technology using either approach.
posted by jeffburdges at 12:34 PM on June 29, 2011


I think we are overlooking something crucial here, and that is that this was discovered by a corporation named "Readen Steel" which means that if this is real it will be in the control of someone who named their business after something in "Atlas Shrugged", not to mention the same people responsible for the creepy bypassing-the-first-sale-doctrine OnLive business model. If I have erred in my facts, let me know.

I really hope this turns out to be bullshit, though.
posted by This, of course, alludes to you at 12:39 PM on June 29, 2011 [3 favorites]


Ok, I'm neither a mathematician nor a communications engineer, but I believe Shannon's Law is about as absolute as you get. For that law to be violated, something would have to be deeply wrong with our understanding of how photons work, or how math works.

Someone casually claiming that they've violated Shannon's Law should be looked at roughly with the same level of skepticism as someone claiming an antigravity device.

Actually, since we don't understand gravity, it would be more accurate to say that you should trust the idea of an antigravity device MORE than this idea.

In the overall rankings of impossible things, this is where this supposed invention would fit, from less likely to more likely:

1. A perpetual motion machine.
2. A violation of Shannon's Law.
3. An antigravity device.

If this guy was claiming to have a machine that could float via gravity repulsion to the Moon, you wouldn't believe him unless you saw him hop into said machine and start floating, and even then you wouldn't completely buy it until he was waving from the lunar surface.

This claim is even less likely than that.
posted by Malor at 12:40 PM on June 29, 2011 [4 favorites]


I can't contribute too much to the discussion, other than to say that OnLive is less of a failure than I'd expected it to be.

Tangentially to that: I'd like to see a website that dedicates itself to tracking extraordinary claims of science that will "change everything". So often stuff is posted here (potential fusion methods, fossil fuel generating microbes, new solar panel fabrication methods) that promises to revolutionize a feild in shoddy, breathless mass-market science writing, and then you never hear anything else about it. I'd like it if there was a blog that picked up on promising inventions, and used scientific journal subscriptions, news hounding, and investigative journalism to track these things from start to finish. That's a lot of work, though, for probably not much reward.
posted by codacorolla at 12:40 PM on June 29, 2011 [1 favorite]


Malor, your MIMO 802.11n wireless access point already exceeds the Shannon limit for communication over a single channel: It does so by using multiple aerials & multipath communication to create multiple channels between the sender & receiver, doing an effective end-run around the Shannon limit.
posted by pharm at 12:45 PM on June 29, 2011


Someone casually claiming that they've violated Shannon's Law should be looked at roughly with the same level of skepticism as someone claiming an antigravity device.

Again, he's not actually claiming to have broken Shannon's Law. He's claiming to have done an end-run around it. Like he said: "the bad news is that Shannon was right; the good news is that he asked the wrong question".
posted by vorfeed at 12:47 PM on June 29, 2011


not to mention the same people responsible for the creepy bypassing-the-first-sale-doctrine OnLive business model. If I have erred in my facts, let me know.

You've erred in your facts. OnLive is subscription based. World of Warcraft doesn't violate the first sale doctrine either. Paying a subscription to play a whole basket of games works well for the kind of person who only ever plays the newest games and doesn't replay them much.

I really hope this turns out to be bullshit, though.

God forbid someone comes up with something clever!
posted by atrazine at 12:47 PM on June 29, 2011 [1 favorite]


Multipath multiplexing is not breaking Shannon's law indeed. And while I'm sure there are some clever ways to do free-space multipath mutiplexing I remain certain that they're really fucking complex and that they will break down under lots of common conditions.

But if he has a reliable, novel multiplexing technique, more power to the guy. That's cool.
posted by GuyZero at 12:50 PM on June 29, 2011 [1 favorite]


When he left to start Rearden Steel

Look, that is the first problem right there. When an inventor names his company after an Ayn Rand character, it implies a kind of limited, simplistic view of the world.

Also, there is the thing where he claims to have overcome a physical law. I don't care if he helped implement Quicktime, that's claiming the power of the gods.

At WebTV he proved to be as shrewd at the business side of things as the tech

And we all know where WebTV is today. Well wait, we may not. It's now called MSN TV, and Microsoft doesn't actually sell the hardware anymore, although they do still make available the proprietary service.

Writing a web browser for a set-top box is not such a feat of engineering skill, although I suppose there is a kind of business acumen in that they actually sold some to befuddled elderly folk who were scared about this internet thing but didn't want to buy a PC and go "all out" for AOL. That's not what I could call a Randian creation of value as much as bilking seniors out of their pension money.

Malor, your MIMO 802.11n wireless access point already exceeds the Shannon limit for communication over a single channel:

It does?

It does so by using multiple aerials & multipath communication to create multiple channels between the sender & receiver, doing an effective end-run around the Shannon limit.

Oh, then I guess it doesn't! ARGH.
posted by JHarris at 12:51 PM on June 29, 2011 [2 favorites]


Also, Shannon didn't ask the wrong question. Cripes, that's like saying Plato asked the wrong question about the nature of sensory perception versus reality because it's not a donut recipe.
posted by GuyZero at 12:51 PM on June 29, 2011 [3 favorites]


There isn't any way to violate the Shannon–Hartley theorem per se, but one might always :

Misinterpret its application to the real world.

1962: "Shannon's limit conclusively proves that copper will never support greater than 300 baud."
1994: "Shannon's limit conclusively proves that copper will never support greater than 33.6k baud."
1998: "Shannon's limit conclusively proves that copper will never support greater than 56k baud."
2000: "Shannon's limit conclusively proves that ADSL on copper will never support greater than 8Mbit."
2010: "Shannon's limit conclusively proves that IDSL on copper will never go beyond 40Mbit."

1996-2011: "Shannon's limit conclusively proves that Xmhz wireless will never go beyond YMbit."

Honestly, Shannon's limit has been mis-applied time after time after time. It's translation to the real world appears to have confused three generations of physics students.

Check back in 2014, 2016, 2018, and 2020 for this same conversation.
posted by Tell Me No Lies at 12:58 PM on June 29, 2011 [15 favorites]


Ok, so in a standard MIMO system you have a receiver and a transmitter, both with multiple antennas. You can theoretically reach N * the Shannon limit where N is the # of antennas.

This system uses multiple antennas on the base station and one antenna per mobile station to achieve a similar effect. So a base station with a 10 antennas and 10 mobiles with one antenna each, the signal transmitted by the antenna array is computed using a set of constantly updated channel vectors, one per mobile. If there are more handsets than base station antennas (usually) then additional CDMA, TDMA, etc. multiplexing schemes are used on top of the spatial multiplexing.

Further patents add the idea of coördinating base stations with overlapping coverage areas to get even bigger arrays (and with better spatial separation) and add a whole bunch of implementation stuff that I'm not going to bother reading.

Potential implementation problems that I can see:
1) Adding additional antennas and computing power to base stations is not super expensive, but neither is it free.
2) Nearby phones will have very similar spatial characterisation (one of the patents says that one way to solve that is to assign nearby phones to different slots using TDMA etc.) Will that be an issue as easy to solve as they think?
3) One way of keeping the signals trained is to update them ever 100ms or so. That's fine if the only change is movement relative to the base station because the changes will be semi-robustly linear and small, but if other people moving around you with or without their own devices have an effect on those signal parameters then you might have a situation where the spatial channel parameters are jumping around very quickly. Can you still get the system to work under those circumstances?

Finally of course this only increases downstream bandwidth, but that is probably fine for most people.
posted by atrazine at 1:01 PM on June 29, 2011 [2 favorites]


Stephen was right, it is a matter of asking the right questions. Actively canceling out interference at the transmitter, instead of just maximizing received signal, is a wickedly cool new way to look at the problem of channel utilization.

I found what I think is the relevant patent application, and it is very cool stuff indeed. It's a form of spatial multiplexing, with some additional math on top of it. They derive a set of coefficients describing the channel from every transmitter to every receiver. Then they tweak the values of the channels in a matrix such that they aim the nulls in transmission for a given channel at receivers who don't want to receive it.

This allows you to send more signal to those who want it, while avoiding sending it to those who don't want it. This is done with multiple transmitters, to multiple receivers, all at the same time, and the coefficient matrix is kept updated along the way.

Channel capacity is proportional to S/N. If a single transmitted signal is 20 db over the noise floor, and you can reduce the level of interference below that value, you could get 100 times the original bandwidth for a given system... and I think this technology can do just that.

I can't wait to see this get applied in the real world, it should free up quite a bit of bandwidth for us all.
posted by MikeWarot at 1:06 PM on June 29, 2011 [9 favorites]


Also, there is the thing where he claims to have overcome a physical law.

Dude, it's a theory, not a law.
posted by Fidel Cashflow at 1:10 PM on June 29, 2011


There may be an issue here with power requirements. However, Cold Fusion should solve that.

Well, it runs Metafilter.
posted by phong3d at 1:11 PM on June 29, 2011 [6 favorites]


Huh. That is pretty cool, it makes some sense of what he said in the talk about new ways to think about RF energy transmission.
posted by atrazine at 1:12 PM on June 29, 2011


He explains it by analogy at 1:18 in the video.
posted by orthogonality at 1:15 PM on June 29, 2011 [1 favorite]


Netflix does that.

Yeah, that is just this: Silverlight Smooth Streaming demo
posted by smackfu at 1:33 PM on June 29, 2011


Fidel Cashflow: "Dude, it's a theory, not a law."

A theorem. Not a theory.
posted by BungaDunga at 1:39 PM on June 29, 2011 [3 favorites]


Malor, your MIMO 802.11n wireless access point already exceeds the Shannon limit for communication over a single channel: It does so by using multiple aerials & multipath communication to create multiple channels between the sender & receiver, doing an effective end-run around the Shannon limit.
That's like saying someone "violates the speed limit" when they take half as much time to reach their destination... because they only drove half as far.

However, Objectivists suck.
posted by delmoi at 1:45 PM on June 29, 2011 [1 favorite]


@atrazine I didn't say "violating"-- I said "bypassing". I think that your comparison to World of Warcraft* is pretty good, though!

All I'm saying is, while I don't begrudge what I'm sure is a nice, decent guy his achievements, I kind of wish that maybe somebody in Angola or something had been able to profit from it instead, you know?

*I don't really like WoW and think that games-as-subscription-products are exploitative and borderline rent-seeking, to appropriate more terminology.
posted by This, of course, alludes to you at 1:47 PM on June 29, 2011


Dude, it's a theory, not a law.
From the wikipedia article you linked too: In information theory, the Shannon–Hartley theorem (also known as Shannon's law)

The difference between a "theory" and a "law" is just something grade school science teachers tell their students. It's not a "real" distinction. And anyway, Newton's laws are still called laws even though we know they're wrong. And they were replaced by the Theory of Relativity. So in that case a "Theory" is more correct then a "Law". They're just random titles that get applied.

And then of course "theorem" usually applies to mathematics and refers to things that have been completely proven.
posted by delmoi at 1:50 PM on June 29, 2011 [1 favorite]


Thanks for the explanation. Duh, I've seen that.
posted by fourcheesemac at 2:24 PM on June 29, 2011


Okay, so some one of you nerds break this down for me. Is this or isn't this going to lead to my goddamned flying car?
posted by jph at 2:40 PM on June 29, 2011 [2 favorites]


No flying car, I'm afraid. It's more like when the newly-minted agricultural society hits a population threshold and has to develop new agricultural methods to support the larger population. Just as they still get to eat, and it's awesome, you'll still get to watch you-tube videos of cats on your Android smart-phone. Or something like that.

This sounds interesting, and it will be great to see how it holds up in practice... I read a bit of the DIDO wireless patent. Signals theory isn't my area (by a long shot), but it sounds like there's plenty of interesting work to be done there.
posted by kaibutsu at 2:48 PM on June 29, 2011


I wonder whether this is more of less efficient in terms of watts per bit?
posted by GuyZero at 3:02 PM on June 29, 2011


I guess this is a method that lets you receive more information at the end point, so, um, the efficiency is all at that end? I guess it would take more power to transmit a greater volume of data.
posted by Kevin Street at 3:08 PM on June 29, 2011


I was one of the people who thought that Onlive was impossible, primarily due to speed of light issues. I now play Onlive.

I didn't know it was Reardon, and well, color me fucking impressed. I am going to not only think twice before saying something he thinks is possible is impossible, but actually wait for the demo.
posted by Freen at 3:51 PM on June 29, 2011


That really does let you evade the Shannon limit, by effectively increasing the number of channels between two points.

Which means you are increasing the bandwidth, therefore, you are not reaching the Shannon limit. As mentioned above, doing this with one antenna would be a wifty trick, but doesn't violate physics. Even if the bandwidth gain is only at one end (three xmitters, one receiver), it's still three channels -- and using more than one channel to send data is old news -- ditto adaptively adding and subtracting channels on the fly.

Whenever you hear someone claim that they've beaten Shannon's Limit, you look for two things.

1) How have they expanded bandwidth?
2) Where are they losing bits.

The first means you can send more bits. The second means you are sending less bits. Either alone or in combination, it means that you're under the limit.

As to people saying 9600 baud modems were as fast as we could go? We were nowhere near Shannon's Limit with a 64KHz channel at 9600 baud, so of *course* we could go faster. We were near the practical limits of a non-error corrected channel, which V.32 signaling was, but once we implemented error correction, we could then go faster, and did.

But none of that ever exceeded Shannon's Limit -- it beat what we thought the wire could carry, but everyone knew that we could *theoretically* get 38KHz out of a analog signal on POTS, and 64K out of a digital signal, it just took us a long time to get close to it.

And then, of course, we just upped the hell out of the bandwidth with DSL and Cable.
posted by eriko at 5:24 PM on June 29, 2011 [1 favorite]


How does online actually work? The first, and most obvious method would be to simply have the servers in a geographical location really close to the user. Like hosted at their ISP's data centers per city. That could probably get the lag down to 10-20ms right there. Way more then you need.

The other, I guess, would be to encode an 'interactive optical flow' so instead of having one optical flow channel, you encode several based on potential inputs. Actually you'd need to send branching optical flow, but you could do it. That seems unlikely.

Also, how much bandwidth does onlive actually take? it seems like it might run into problems with bandwidth caps. And video games rendered locally look a *lot* better then then compressed video of video games. A game like starcraft doesn't compress that well. Then again it seems like tons of people just don't notice or care about video quality at all.
posted by delmoi at 5:39 PM on June 29, 2011


I get the sense that Rearden's new radios will be capable of treating other, nearby radios as sources not of noise but of information. The base station has access to all of the information being exchanged with all its client radios; if it can accurately model the effect of that aggregate signal as perceived from any given client, then it can work with difference signals between that and the information exchanged with the client. The noise that the Shannon limit applies to would then be limited to modelling errors plus whatever is coming from outside the network as a whole.
posted by flabdablet at 6:15 PM on June 29, 2011


As someone whose name is Shannon, I say this man must be stopped.
posted by brundlefly at 6:55 PM on June 29, 2011 [2 favorites]


Onlive 'works' because human beings can compensate for pretty huge latencies. Back in the old days of dialup a 100ms ping was phenomenal, and yet people managed to play first person shooters. There was never any question that onlive would work from a technical perspective (all it is is remote desktop with a modern video codec), what people were and are skeptical about is whether it is economical for either the producer or consumer. On one hand, time-shared machines in huge server farms can take advantage of some significant economies of scale, but on the other, bandwidth and especially video encoding add non-negligible costs, and you have to consider the investment in hardware which will be rendered obsolete within two years.
posted by Pyry at 7:20 PM on June 29, 2011 [1 favorite]


"I believe it. Verizon is no longer installing FIOS (fiber to the curb), the wireless technology is going to leapfrog it."

Holy shit, this might explain what the contractors were doing this spring for Verizon that they wouldn't talk about and left all these strange looking, kind of right angled flat antennae all over the top of the telephone/power poles.
posted by digitalprimate at 7:26 PM on June 29, 2011


I publicly predicted developments like this 15 years ago, and was resoundingly mocked by engineering "experts" in my local community and in the press for stating so. Here's to open minds, and innovation!
posted by Vibrissae at 8:02 PM on June 29, 2011


"I believe it. Verizon is no longer installing FIOS (fiber to the curb), the wireless technology is going to leapfrog it."

Wireless is going to leapfrog optical cable? More then 100 terabits per second per line and an unlimited number of lines? I find that kind of hard to believe.

FIOS is pretty slow compared to what you can get in other countries, which already have 1gbps on offer.
posted by delmoi at 10:11 PM on June 29, 2011


Delmoi, the problem is that a cheap "good enough" last mile wireless will probably crush the value of fiber to the home (where good enough =~30Mbps Internet + scads of tv channels). If Verizon spends a boatload of money on digging up the street for fiber and a wireless competitor comes along and beats them down on monthly price they'll never recover their investment.
posted by BrotherCaine at 11:24 PM on June 29, 2011


Those patent applications might run into Ruckus's patents.

I'm seeing that radios can spatial multiplex, time divide, request feedback from other radios as to their perception of signal quality, and then map cooperatively the potential capacity of throughput in an array, to add capacity beyond existing protocols.

Isn't this evolutionary and not a revocation of physics?

Again, license it to Atheros, Ubiquiti makes it a field product, and then I bitch about trying to optimize implementations.
posted by dglynn at 11:28 PM on June 29, 2011


And then of course "theorem" usually applies to mathematics and refers to things that have been completely proven

Well, yeah. Shannon's law is a mathematical result, and a brilliantly simple one. It's not descriptive (like relativity or evolution). To get around it you pretty much have to change your definition of information or your definition of a channel. Since people have been touting MIMO as OMG BYPASSES SHANNON'S LAW for years and years, I assume this guy is doing something along those same lines.

You presumably haven't read Shannon's papers. They're fairly short, very readable, available in an inexpensive paperback edition, and might give you a better chance at appearing to know what you're talking about.
posted by hattifattener at 12:42 AM on June 30, 2011 [1 favorite]


FIOS is pretty slow compared to what you can get in other countries, which already have 1gbps on offer.

You can actually get gigabit right now in the US. You just have to live in Chattanooga(fully deployed), or Kansas City, though I'm not sure how far along the Google fiber project is.

And Sonic.net is running a test fiber deployment to a few hundred homes in tiny little Sebastopol, California... they'll be charging $70/mo for gigabit. If it works out, they'll roll out the service over a much broader area.

So Chattanooga, Kansas City, or Sebastopol. Marvel at the breadth of choices on offer. :)
posted by Malor at 12:56 AM on June 30, 2011


Fuck Ayn Rand.
posted by stavrosthewonderchicken at 4:14 AM on June 30, 2011 [1 favorite]


You can actually get gigabit right now in the US. You just have to live in Chattanooga(fully deployed), or Kansas City, though I'm not sure how far along the Google fiber project is.

Yeah. It's not that expensive when you do it all at once. The reason we don't have gigabit connections is because the phone companies figure there's more money to be made in rent seeking, which they can do more easily with phone, then actually rolling out the internet. And of course the content providers don't want people to get faster internet anyway, why make it even easier for people to pirate content?
posted by delmoi at 5:01 AM on June 30, 2011


Fuck Ayn Rand.
posted by stavrosthewonderchicken


What, man, do I look like Nathaniel Branden?
posted by COBRA! at 5:15 AM on June 30, 2011 [1 favorite]


« Older [Bill] Kerlina said his two years [as principal of...  |  Serenity trailer... with ponie... Newer »


This thread has been archived and is closed to new comments