Still seems hard to believe you can scale up the math and deal with intermittent changes in the RF (e.g. cars passing by) but I guess now we'll find out if it works.
The point of the data center is to compute, on the fly, unique phase solutions for moving devices being served by multiple antennas, and to compute new solutions as the devices move. It's a huge workload, perfect for a modern, high-speed multiprocessor system to tell the cell-tower antennas which phase delay to apply to (a) each antenna, (b) each data packet, and (c) each device in a separate, changing location.
You don't need a "datacenter" per se, but you need a private control plane shared by the devices so that they can pass relevant information to each other. As long as the sites are interconnected and have some sort of addressing that is tied to their geographic location (i.e. devices know very quickly who their neighbor devices are by address), a central datacenter is not necessary.
TL;DR You need a shared bus for passing geospatial data between geographically local devices.
Perhaps, I'm just using the language from the linked article:
"To work properly, a company backing the pCell technology would need to build out a large data center in addition to deploying the transmitters. It’s in the data center where servers constantly crunch away on the algorithms that form the unique wireless stream aimed at each device."
> It's not terribly difficult provided that you've got enough directional antennas in your base station.
This scheme doesn't rely on directional antennas, it relies on computing an optimal set of phase solutions for multiple antennas to maximize reception quality. Each antenna is an ordinary omnidirectional dipole, but several of them working together, with calculated phase delays, provide a kind of directionality if you want to think about it that way.
How do you know it doesn't rely on directional antennas in addition to the multi-phase solution? If I were trying to do something like this I would ABSOLUTELY use multiple directional antennas in each base station AND phase things appropriately.
I don't mean trying to make a ball of yagis or other narrow beam antennas. But at the very least try and take advantage of splitting a base station into multiple (2, 4, 8, etc) slices. I can't see how that would hurt performance (though it would probably make things more expensive) and it would certainly make it easier to get more bandwidth.
Here's an excerpt from the patent:
[0120]FIG. 3 provides additional detail of one embodiment of the Base Station 200 and Client Devices 203-207 shown in FIG. 2. For the purposes of simplicity, the Base Station 300 is shown with only three antennas 305 and only three Client Devices 306-308. It will be noted, however, that the embodiments of the invention described herein may be implemented with a virtually unlimited number of antennas 305 (i.e., limited only by available space and noise) and Client Devices 306-308.
Obviously I haven't studied the patent in detail but that does sound to me like there are multiple antennas per base station.
> How do you know it doesn't rely on directional antennas in addition to the multi-phase solution?
Because:
1. The article claims it can use the existing cell tower system.
2. Cell tower antennas are simple dipoles, not directional antennas.
> If I were trying to do something like this I would ABSOLUTELY use multiple directional antennas in each base station AND phase things appropriately.
Yes, that would improve the performance, but this would make it a hard sell to cell companies who are trying to reduce the cost of their installed equipment. Also, to use directional antennas in a high-speed dynamic network, for a given beam width N, you would need 360/N directional antennas. For a beam width of 30 degrees, you would need 12 antennas where one exists now, and you would need a way to switch between antennas on each transmitted packet to multiple served devices. That would be a nightmare.
> Obviously I haven't studied the patent in detail but that does sound to me like there are multiple antennas per base station.
Yes, the basic idea requires multiple antennas whose relative phase can be adjusted. But not directional antennas.
> 2. Cell tower antennas are simple dipoles, not directional antennas.
This is where you went wrong. They are in fact directional antennas. From a wikipedia article:
"Due to the sectorized arrangement of antennas on a tower, it is possible to vary the strength and angle of each sector depending on the coverage of other towers in view of the sector."
That certainly sounds to me like a cell base station not only already has multiple antennas (which you can see with your own eyes) but that it makes use of said multiple antennas in some kind of semi-intelligent way. Obviously not as smart as the pCell, but sorta.
> For a beam width of 30 degrees, you would need 12 antennas where one exists now, and you would need a way to switch between antennas on each transmitted packet to multiple served devices. That would be a nightmare.
As we've established there are already multiple antennas. Maybe not 12, but several. And this is a trivial problem to solve actually. The same technology that tracks you from mast to mast can be used to track you from antenna to antenna. This is the job of the base station controller. Here's a clip:
"By using directional antennas on a base station, each pointing in different directions, it is possible to sectorise the base station so that several different cells are served from the same location. Typically these directional antennas have a beamwidth of 65 to 85 degrees. This increases the traffic capacity of the base station (each frequency can carry eight voice channels) whilst not greatly increasing the interference caused to neighboring cells (in any given direction, only a small number of frequencies are being broadcast). Typically two antennas are used per sector, at spacing of ten or more wavelengths apart. This allows the operator to overcome the effects of fading due to physical phenomena such as multipath reception. Some amplification of the received signal as it leaves the antenna is often used to preserve the balance between uplink and downlink signal."
I appreciate your enthusiasm for debate but please do some more research into cellular radio before you weigh in further. You've got a lot of the basic concepts right but not the particulars of the implementation. If that comes across harsh; I'm sorry. I don't know how to communicate this more gently.
>> 2. Cell tower antennas are simple dipoles, not directional antennas.
> This is where you went wrong. They are in fact directional antennas.
No, they (simple dipoles) are not. Read my quote above -- a simple dipole is not directional. Your reply says that an array of such dipoles can be made directional, which is true and a point I made as well, but it's a different topic.
> I appreciate your enthusiasm for debate but please do some more research into cellular radio before you weigh in further.
That's my advice to you -- before you change the subject, learn enough to realize that you're changing the subject.
> If that comes across harsh
Harsh? How about wrong? Dipoles are not directional. Cell tower antennas are vertically polarized dipoles, therefore they aren't directional. An array of dipoles is directional in a crude sense and as expressed in your linked article, but not remotely comparable to a phased array, the topic of the present discussion.
> I don't know how to communicate this more gently.
Let's see if you gently recognize your error. If you move the goal posts, obviously the game changes.
Okay I can't tell if you're trolling me or what. Let's start with one really simple thing and from that we can learn a lot.
You have claimed repeatedly that cell tower antennas are simple dipoles. What do you base this claim upon?
And when I say "base" I'd really like some links to in-depth explanations. I don't want to see a simple hexagonal cell system like this with A-G labeled cells and a repeating pattern (http://www.ofcom.org.uk/static/archive/ra/topics/mpsafety/sc...). That's how cell phones got started back when there weren't many users and there was a LOT of ground to cover. These days there are many, many more users and as such things have gotten more sophisticated. I included links to wikipedia articles that contain more current information.
If you can prove that they are in fact simple dipoles I have no problems accepting that they are not directional; a simple dipole is an omni. You learn that as a 3rd year EE.
> You have claimed repeatedly that cell tower antennas are simple dipoles. What do you base this claim upon?
The fact that the vast majority of cell tower antennas are simple dipoles -- it's the basis of the system. You want to argue -- go somewhere else. I was doing this before you were born.
> I don't want to see a simple hexagonal cell system ...
No, of course not -- that would imply that you were wrong. This discussion revolves around a scheme to phase ordinary cellular antennas to improve the performance of a classic cellular system, the kind of system you don't want to talk about.
The big advantage of the scheme under discussion is that it works with a classic cellular antenna system (plus some additional electronics and mathematical processing). That makes a classic cellular system the topic.
Typical collection of vertically polarized cellular dipole transmitting and receiving antennas:
In the first document you linked here's an excerpt:
"Wireless carriers have taken the reduce and reuse approach a step further with the use of directional antennas, illustrated in Figure 5. Rather than using a single omni-directional antenna that covers a circular radius around a tower, carriers introduced directional antennas, to further segment cell sizes and enable the reuse of additional frequencies. For example, placing three antennas operating in separate frequencies on a tower allows sectors to be created within a cell, essentially tripling capacity per cell."
"Cell towers frequently use a directional signal to improve reception in higher traffic areas. In the United States, the FCC limits omni-directional cell tower signals to 100 watts of power. If the tower has directional antennas, the FCC allows the cell operator to broadcast up to 500 watts of effective radiated power (ERP).[7]"
The picture that you linked to on the wikimedia site shows what are clearly referred to by another wikipedia page as "sector antennas". See for yourself. http://en.wikipedia.org/wiki/Sector_antenna
Compare those with the picture that you linked which you claim is of simple dipoles. I see very little difference between the two. Which is more likely? That they're making dipoles which just happen to look exactly like sector antennas AND that they're putting up well more than they need to (since dipoles are omni they'd only need one) or they're actually sector antennas?
Finally if you're going to invoke the "I was doing this before you were born" I'm more than happy to quit now. I had this foolish idea that you (as most people on HN do) wanted to understand the world as it really is. I see now that I was mistaken, you simply want to be right. That's fine, I can't fault you for it. It's human nature. But please take that kind of corrosive attitude elsewhere. The reason this community is so great is that people here listen to data or evidence. Please don't dilute those values for the sake of "winning" a stupid internet argument.
> Finally if you're going to invoke the "I was doing this before you were born" I'm more than happy to quit now. I had this foolish idea that you (as most people on HN do) wanted to understand the world as it really is.
I understand these issues perfectly well, and you are trolling. I proved my case, or weren't you paying attention? The conversation is about an unconventional scheme that relies on a conventional cellular system.
As I said before, go argue with someone else until your earlobes dry.
> 2. Cell tower antennas are simple dipoles, not directional antennas.
> Cell tower antennas are vertically polarized dipoles
> The fact that the vast majority of cell tower antennas are simple dipoles -- it's the basis of the system
You stated three times that cell tower antennas are simple dipoles. That's the problem. They're not simple dipoles. 20 or 30 years ago they absolutely were. Then a lot of users joined the system and the carriers all rolled out directional antennas in all but the most rural of places.
I'm trying to make sure that you understand that this is the point I'm arguing.
I fully agree that a 1/4 wave dipole is an omni. I fully agree that pCell probably uses omnis (at least initially). I fully agree that many years ago the cellular system used omnis.
Where I don't agree is the idea that the current cellular system uses omnis. By and large, anywhere suburban or urban, and even many rural base stations all use directional antennas. I've provided tons of links showing that the carriers use directional antennas to further break up the large cells that existed when they put all the towers up (and used omnis) into smaller cells today which can support more users.
> I proved my case, or weren't you paying attention?
Just asserting that some statement is true doesn't make it so. I've provided plenty of evidence that carriers today use multiple directional antennas to service extra customers from their existing towers. "Directional antenna" doesn't necessarily mean a 5 degree beam width, an antenna can have a 30 or 60 or 90 or 180 degree beam width and still be directional. The towers used to be in the middle of the hexagonal cells, now they're at the place where three hexagons meet.
Here is a patent by Nortel for a six-sector system in 2004.
A dipole is not directional? The radiation pattern of a resonant dipole consists of toroidal lobes normal the to axis of the dipole with nulls at the axis of the dipole.
Source: I do NEC simulations of my antennas for amateur radio.
A vertical dipole, something I mentioned a few posts ago, before msandford decided to turn this into a pissing contest. As a radio amateur, you will be familiar with the radiation pattern of a 1/4 wave vertical coupled to a ground plane.
Right, I couldn't agree more. It's a very clever delay system that ensures that the EM waves all arrive at a single point such that they constructively add together.
That means that when you're in the pCell pocket you get excellent reception because of everything adding together, and when you move away it drops off VERY quickly.
It's smart stuff but not impossible to comprehend. I'm surprised that none of the VCs "experts" could figure it out.
This is great, but I'm doubting the FCC will approve this unless he has found a way to not cause interference on the other frequencies his system uses.
> Despite such demonstrations, Perlman has been unable to tempt venture capitalists with the technology. “They invariably bring in experts who say it doesn’t really work,” he says. “I am showing them a demo, but they remain convinced that it’s something else.”
Are you actually claiming that subject-matter experts argued the technology in Apple's original iPhone was impossible and that the demos at its unveiling were just tricks? If so, your claim is pretty hard for me to believe, as I don't remember reading articles arguing anything like this at the time.
To be fair, the first iPhone demo didn't "really work". The order in which things were presented was determined by what was least likely to crash on stage. Longish article, but it's an interesting read:
you would think the military would love something like this, they could just blanket an area with easily disposable transmitters to keep their troops connected. With the processing power held behind safe lines the could even possibly determine who should have what information
I think that's the issue. The technology is cool and his demo is probably real but rolling this out in an urban environment seems like it would require massive computing power in addition to lots of new antennas.
sure makes it look like he is broadcasting 6 QPSK modulated video streams on the same frequency.
1\ it requires ridiculously precise location feedback from the receiving device
2\ it probably requires ridiculous amount of synchronized transmitters
3\ + lots of computation
Experts claiming it works, but not the way he sells it makes me think there is something fishy with the demo, like maybe he isnt really broadcasting on same freq or something.
No. The transmitters calculate signals such that constructive/destructive interference will combine to form a strong, clean signal in the precise location where the device is.
No, it's not the number of antennas (although there is always more than one), the method relies on the phase (or "time delay") of the signal at each antenna.
I say no because the system relies on multiple antennas, but beyond a few, the advantage evaporates while the workload increases. I wouldn't be surprised if the system limits itself to three well-placed antennas surrounding the target device, just to simplify the calculations.
> The more antennas you have, the cleaner you can keep the beam and minimize constructive interference happening in the wrong places.
Beyond three antennas surrounding a given device, this just isn't true -- more antennas don't produce a proportional increase in performance to compensate for the increased computation workload.
Not for one client device, but it definitely helps when handling lots of clients. The more antennas the fewer spots
with unwanted constructive interference.
> The more antennas the fewer spots with unwanted constructive interference.
Yes, true, but the entire system becomes more complex for a small improvement in performance. Also, most of the theoretical work in a scheme like this would focus on the least common denominator, which is three antennas.
Interestingly, for a roughly circular array of antennas, you end up with concentric rings of high signal strength as you approach the midpoint of the array. So even the thesis that you've eliminated unused hot spots breaks down in some cases.
Here's my diagram for three phase-controlled antennas (the blue dots):
The wavelengths in this diagram are longer than for a cell system, it's just to show how this idea works. Buy adjusting the phase at each antenna, you can move the high-strength lobes around to match up with the location of a given device.
The trick is getting that many antennas deployed in an urban area. They also have to be wired, and have reliable
low latency connections to the datacenter.
The problem is not the idea, but the word. Most people don't think of constructive phase adjustments as beamforming. It's not wrong, it's just an unfamiliar term.
Basically the base arial seems to focus a 1cm wide beam at each device. Works nicely in his lab but dunno how well that'll work using your phone in the car.
Without knowing the details (which are still secret for good reason), before transmitting a packet to a specific device, the system computes an ideal transmission wavefront from multiple antennas to maximize the signal quality at that specific device.
This is not a new idea -- phased array radars do essentially the same thing -- but it's a new application of the mathematical methods, and the complexity is much higher than for a modern radar because the system has to deal with many different devices, freely moving around, and calculate the optimal combination of radio signals to produce a high signal level at that devices's current location.
It is the high computation workload that makes Perlman emphasize the role of a data center, which would need to be added to an existing cell network, to make this method possible.
Cell systems already know how to determine the location of a given device using signal arrival times at multiple antennas. This method builds on that knowledge base to improve the outgoing signals, and possibly also to phase-adjust the reception antennas, to maximize signal quality in both directions.
In essence, the system computes a protocol for each outgoing packet, beyond the packet's data content. Each packet contains additional information for each antenna that will be used to transmit it, information about phase delay. Before the packet is transmitted, the required phase delay for each of the transmitting antennas is adjusted to correspond to the calculated best solution for that device at its present location. This process is performed for each packet, for each device, which explains the high computation workload.
If physically separate antennas are used for transmitting and receiving (usually true in cell service), a separate computation would be required to most efficiently receive packets from each device. I don't know if the system has this added level of complexity, but it's possible in principle.
> This is not a new idea -- phased array radars do essentially the same thing -- but it's a new application of the mathematical methods[...]
Yes, it's called pre-coding, every modern wireless system does it. Doing it across wireless stations is distributed beamforming which Perleman did not invent.
> Yes, it's called pre-coding, every modern wireless system does it.
Present cell systems don't adjust phase to maximize reception quality in a coordinated way for each of several transmitting antennas, for each packet, for each participating device. That's new.
> Doing it across wireless stations is distributed beamforming which Perleman did not invent.
That's certainly true -- phased array radars have been doing this for decades. But Perlman should be able to get a patent anyway, based on the combination of ideas this scheme represents. Multiple, physically separate antennas, the use of a data center to compute solutions on the fly to accommodate multiple antennas and different devices with different locations, that's new.
> Present cell systems don't adjust phase to maximize reception quality in a coordinated way for each of several transmitting antennas, for each packet, for each participating device. That's new.
You are just describing MIMO pre-coding which has existed for more than 20 years.
I don't see what is patentable: Distributed beamforming necessarily requires physically dispersed antennas. Adding a data center doesn't seem patentable. But I don't know anything about patents nor do I care that much.
Being a critic of software patents myself, and given how much this idea relies on software and mathematics (mathematics is definitely unpatentable), I agree that it's an issue.
I suspect that the entire method is patentable even though each element taken separately isn't.
Yea the novelty is essentially that he's (allegedly) managed to make MIMO work with "non-local" antennas, which is really tricky in practice due to the sub-nanosecond phase synchronization required.
Yes. The only reason I can say that is because current military phased-array radars do it. I would quibble with "at once", but they can certainly accommodate multiple targets, in series, in a short time span.
It's called multiple access and it doesn't really have to do with beamforming: There are many multiple access systems that do not rely on time division, CDMA being a prominent example.
> This process is performed for each packet, for each device, which explains the high computation workload.
Could this not be sped up in ASIC hardware? I'm not a hardware guy, but I would think a GPU could fit the bill for these sorts of geospatial on-the-fly calculations.
EDIT:
"Accelerating geospatial analysis on GPUs using CUDA"
If the wavelength at 700Mhz is 43cm, wouldn't this have to track the position of the phone to within ~5-10cm to get a benefit? Or at least relative distance between antennas.
The present cell system can locate a cell device within the same limits using time delay and phase detection of the received signal (this requires three antennas for a 2D fix). Many cell phones that claim to have GPS actually rely on the cell network to locate them, and the location method is reasonably accurate. That means the system already knows where each device is located using the same wavelength constraints required to make this idea work.
Quote: "Mobile phone tracking refers to the attaining of the current position of a mobile phone, stationary or moving. Localization may occur either via multilateration of radio signals between (several) radio towers of the network and the phone, or simply via GPS."
Reading Perlman's descriptions it does sound like distributed beamforming. Perlman is a smart guy, although I expect he would be more successful selling this technology initially to crowded venues (think large stadiums, or conference centers) which would allow those venues to both offer consistent high performance access to fans as well as high performance wireless channels for things like flying cameras (which is currently quite the mismash of technologies).
That said, I continue to remain dubious because his claims are fairly out there in terms of bandwidth. I don't know if Claude Shannon were alive today if he could figure out if there was an upper limit of directed traffic in a channel but that is the principle that feels like its being violated here. All of my training has taught me to think of a bit of spectrum like a single wire, and sure you can attach a bunch of things to that wire, but there isn't a lot of theory around how that becomes a bunch of separate but equivalent channels.
> That said, I continue to remain dubious because his claims are fairly out there in terms of bandwidth.
No additional bandwidth is required beyond what cell systems already provide. The idea behind this method is that each outgoing packet has some extra information about phase for each antenna that will be used to transmit it.
The extra information is calculated using Perlman's data center and added to the data structure of a specific packet meant for a specific device. The extra information tells the transmitting antennas which phase delay to apply to maximize the signal at the device's current location.
My point is that this method doesn't change the bandwidth required for the cell system -- in many ways everything is as it was before, except that each participating device experiences better reception.
This system relies on a relatively simple phase adjustment at each transmitting antenna, but because of the number of devices, the number of antennas and the math involved, a dedicated data center would have to be added to an existing cell system.
But there's no change in terms of bandwidth -- that remains the same.
"The technology would do away with wireless network congestion by giving each smartphone and tablet its own super-fast connection instead of asking these devices to share bandwidth pumped out by a cell tower."
"In demonstrations at his laboratory, Perlman showed off iPhones, Surface tablets, and TVs streaming massive files—the 4K UltraHD version of House of Cards from Netflix, for example—via his own wireless networking equipment"
These are both bandwidth narratives, not signal quality narratives. (granted they are the naive reporter's narratives but still). His breakthrough is that "all these devices can stream 4K ultra HD" that says that somehow he getting more bits to more machines, and that is a function of bandwidth not signal strength.
Lets assume for the moment that the basestation is connected to some multi-lambda super fiber with a 40G connection back to the Internet. He's going to give every LTE phone a 100Mbps to 1Gbps connection to that? That works for the first 400 or 40 people but then what? 5 bars and stuttering video?
This story, and others like it, have made the argument that the problem is congestion not signal clarity. Congestion is a function of channel bandwidth and operation rate. That is what leaves me dubious, I haven't seen how he can increase the bandwidth of the channel with his gizmo.
I think the journalist didn't understand what he was seeing. Clearly if signal strength is improved and phase distortion is removed, the error rate will go down and this will improve apparent bandwidth, but without requiring more bandwidth than the system already can provide in principle.
To me, the first sentence in your quote from the article is simply hyperbolic but has no connection with reality. The method doesn't do away with network congestion, it simply improves the signal at each participating device.
> Congestion is a function of channel bandwidth and operation rate.
I think the congestion this method addresses is that caused by unsophisticated signal treatment methods and an overall decline in performance. Obviously if the system can optimize the phase at each transmitting antenna and improve the received signal for each device, then more devices can use the system simultaneously. That's not how I would define "do away with network congestion", but again, I think the journalist just had no idea what he was describing.
I'm not sure if I'm reading you right (not my area of expertise), but my understanding of the system is that this does increase total bandwidth. Each cell phone gains access to the entire spectrum of the transmitting source, rather than all phones splitting/sharing the available spectrum. It's as if only a single cell phone was in a given tower cell, and the tower was dedicating all its bandwidth to that single phone.
No, I don't think so. If that were true, the cell system would have to upgrade its throughput. That would make this scheme unattractive to cell companies.
I think the reason for all the attention given to this method is because it greatly improves the performance of a cell system without requiring any bandwidth increase. Remember that dropped packets are at times a big limitation on cell system performance, and much existing bandwidth is often wasted on packets that don't get decoded properly because of weak signals.
> It's as if only a single cell phone was in a given tower cell, and the tower was dedicating all its bandwidth to that single phone.
If the system can adjust the phases of the transmitting antennas to maximize the signal strength to a particular device, for each transmitted packet, then for all practical purposes that's true -- the entire bandwidth of the system really does become available to that particular device, for the duration of each packet meant for it.
I don't know the statistics on dropped packets, but I think it's substantial in modern times for high-speed networks. The ratio of dropped packets essentially represents a bandwidth decrease. If that could be eliminated, it would improve throughput, and someone will be tempted to claim that bandwidth has increased. Only sort of.
AFAIK the Shannon limit applies to a channel with a single transmit antenna and a single receive antenna. MIMO can get a factor of N more throughput without breaking Shannon by using N antennas.
Shannon's channel capacity theorem applies to channels generally, including MIMO channels. It is a theoretical maximum which cannot be exceeded for a given channel. But a MIMO channel is a different channel than SISO and has a different capacity.
Not if the antennas are all carrying the same signal. In the described system, the antennas get phase adjustments, but they do not get different signals, and this idea doesn't rely on multiple, independent data channels. So the Nyquist-Shannon sampling theorem still applies.
I imagine two phones X and Y and one antenna sends X XOR Y while the other sends X XOR Y XOR Y. A phone getting both signals could determine X and Y, but it would take twice as much bandwidth. If you can XOR the signals in the air through superposition somehow, X might be received in one location, and Y in another.
So 2 antennas cant service more than 2 phones at full speed, but they can service those two phones at full bandwidth. This might not seem like an advantage over directional antennas, but it relaxes the physical constraint of making sure the phones are in two separate enough places to be serviceable by the two antennas. Like maybe your phone would have two antennas and get twice the data--not very feasible if your phone were somehow serviced by two directional antennas.
I haven't worked out how superposition could do something exactly like XOR, or if it maybe has a minimum of 3 antennas to start being possible, etc., but it seems plausible.
Anyway, the idea would be sending signal X and Y. You want X interfering with Y at some time offset to yield desired phone signal A, and X interfering with Y at some other time offset to yield phone signal B. The phones never receive X or Y, but some superposition of the two, The phones couldn't cheat and just record the two waves and timeshift internally to yield twice the bandwidth, because they don't receive the two waves separately, they only get the superposition of the two.
The beamforming means that each channel can use less power and get a higher SNR. Thinking of aimed signals as a single wire is wrong. The hardware analogy is a single wire in a highly-directional antenna aimed at each client. There's no real sharing of spectrum. The only difference is that you can use software plus 100 undirected antennas to simulate having directed antennas.
> although I expect he would be more successful selling this technology initially to crowded venues
I would expect he would be more successful patenting the hell out of it and licensing it to Ericsson, NSN, Qualcomm, Cisco, etc than trying to sell it to individual venues and customers.
It reads like a type of cooperative base station, which means geographically separate antennas are used to transmit signals that add up to a "clean" signal at the receiving antenna, in which interference is cancelled. That's equivalent to a form of MIMO, in which the antennas are widely distributed in space. Consequently, I'd expect it to be subject to the same limitations as MIMO.
I'm predicting that the breathless claims of "no limit" are based on being the only show in town. As long as the system is controlling every transmission in a volume of space it will do well, and you will be able to add users until you reach the same limits that a MIMO system would hit if it was the only transmitter. Start adding uncontrolled interference, and things will start to degrade.
Maybe the next step, would be for the distributed base stations to use receivers to sample the spectrum. This data could be fed back into the cloud and an attempt be made to predict and cancel the uncontrolled interferers. The FCC (or local equivalent) might take a dim view of this though, if it turned out that the cancellation was effectively jamming other users.
--- Edit.
Just adding, in my experience, if you want to understand the latest radio communications technology, speak to a radio astronomer. When MIMO first came out and people were trying to figure out its limits, it turned out that the radio astronomers already knew the limits, based on their knowledge of large antenna arrays. I'd guess the astronomers would look at this current technology and think "very long baseline interferometry".
You bet this is just like interferometry -- astronomers those days can actually perform it with thousands of miles of separation. Although they're not really forced into the economies required here.
Typical beamforming just boosts the signal quality, on a slice of the available spectrum. But it would be faster if you used the whole bandwidth. But then no one else would be able to use that bandwidth. But if you used a parabolic antenna pointed at each receiver, then each phone would get the full bandwidth. But if the phone is moving, you can use (tracking antennas or...) deliberate interference within a "mesh" of antennas.
An unfortunate title -- the method doesn't rely on hardware, on "machines", it uses hardware already in place but calculates, then generates, a unique signal for each device. The method depends on the calculation, the mathematical theory, not machines.
Calling this a machine is like calling the first car a horseless carriage.
I don't get it. Wouldn't it just be easier to install more base stations with a lower transmission strength (essentially leveraging the same transmission frequency more times)?
I can think of at least two reasons why this is superior:
1) You need to have multiple base stations installed roughly equally spaced. This seems to support heterogenous spacing of the stations
2) With lots of low-powered base-stations interference and obstructions start to become bigger issues. With many reachable antennas, a small fraction that have interference or obstructions should merely decrease SNR a bit rather than drop completely.
Of course advantage #2 seems (to me) to be theoretical; I'd like to see how this performs in the real-world.
The other issue I can see is moving receivers and/or moving reflectors. Imagine having this work, for example, when you are near a wind-farm, with many moving reflectors changing the apparent signal strength continuously.
I know someone who worked on a similar problem in the past, and temporal coherence in real-world situations was close enough to the feasible round-trip delay to the data-center. (e.g. if you need to recalculate every 150ms and it takes 125ms round-trip to the data center, you only get 1/6 the theoretical benefit).
An antenna not integrated into the system would see noise. The designated phone would see the exact interference pattern it needs to get the "picture."
Nice example with visual crypto, but does this scale? Are there algorithms able to generate arbitrary number of patterns where groups of two or more of those patterns form desired groups of signals?
Basically generate 10 visual crypto pictures and be able to get different messages by combining random pairs of pictures.
In this diagram the size of the high-signal-strength lobes is much larger than would be true at cellular frequencies, but it gives a sense of how this idea works. By adjusting the phase at each antenna, the lobes can be moved around to accommodate different devices within their range.
As I expected, increasing the number of antennas past three doesn't really improve the outcome enough to justify the increased complexity.
Yea, and MIMO is should be not only beamforming (which, best case scenario, gives you one strong channel); but it uses multipath to create essentially two channels at the same frequency.
My sole reason for posting the example was to show how a small number of individual antennas could produce focusing lobes within a 2D space. Phased array radars use many more elements, much higher control over the wavefront, and for a different purpose.
In a cellular antenna scenario, unlike phased array radar, you cannot avoid multiple lobes, but the advantage of phasing is still present. Adding antennas to the system (beyond three or four) becomes a case of diminishing returns.
> More antennas can enable you to narrow lobes more - you may have not been phasing them properly to get that effect.
There are paradoxical effects with more than a few antennas -- apart from the complexity issue. In some cases you will see lobes with strange shapes in locations having no benefit. With a circular array of antennas, you sometimes get concentric circles of high signal strength, not what you're after in the described system. Example:
I'd recommend you make your system scale appropriate. Considering that cellular transmissions are on the order of GHz, you're effectively simulating a circle of antennas around 1 m wide looking at what happens in the middle. That doesn't seem particularly relevant to the case of antennas being very far away (far-field vs near-field etc).
Would you mind going into a little more detail about what each color represents.
Intuitively, the diagram seems to say that behind (in front of?) each blue dot (antenna) the signal is really strong while on the other side it kinda drops down, only to strengthen again when we hit the middle of the triangle. But I really have no idea what's going on...
> Would you mind going into a little more detail about what each color represents.
It's a simple density plot of an electromagnetic field in which the brighter colors represent constructive interference and dark areas represent destructive interference. The three antennas constructively or destructively interfere, with varying effects at different locations within the field.
I want to say again that the diagram is only an example and uses a frequency much lower than that used by cellular systems, just so the lobes are a reasonable size, and only to give some idea of the principles at work.
> Intuitively, the diagram seems to say that behind (in front of?) each blue dot (antenna) the signal is really strong while on the other side it kinda drops down, only to strengthen again when we hit the middle of the triangle.
Yes, that's true -- the three external lobes happen to be places where the three antennas constructively interfere. One can do this using pencil and paper (as I did when I was a NASA Space Shuttle engineer many years ago, and before there were personal computers), but computers make it much easier to experiment with different configurations.
Here's a different antenna configuration, one that I doubt anyone would want to build -- a circle of antennas, giving circular intensity lobes:
105 comments
[ 5.2 ms ] story [ 211 ms ] threadTL;DR You need a shared bus for passing geospatial data between geographically local devices.
Perhaps, I'm just using the language from the linked article:
"To work properly, a company backing the pCell technology would need to build out a large data center in addition to deploying the transmitters. It’s in the data center where servers constantly crunch away on the algorithms that form the unique wireless stream aimed at each device."
In other words it could very well scale as O(n), n = the number of devices.
This scheme doesn't rely on directional antennas, it relies on computing an optimal set of phase solutions for multiple antennas to maximize reception quality. Each antenna is an ordinary omnidirectional dipole, but several of them working together, with calculated phase delays, provide a kind of directionality if you want to think about it that way.
I don't mean trying to make a ball of yagis or other narrow beam antennas. But at the very least try and take advantage of splitting a base station into multiple (2, 4, 8, etc) slices. I can't see how that would hurt performance (though it would probably make things more expensive) and it would certainly make it easier to get more bandwidth.
Here's an excerpt from the patent: [0120]FIG. 3 provides additional detail of one embodiment of the Base Station 200 and Client Devices 203-207 shown in FIG. 2. For the purposes of simplicity, the Base Station 300 is shown with only three antennas 305 and only three Client Devices 306-308. It will be noted, however, that the embodiments of the invention described herein may be implemented with a virtually unlimited number of antennas 305 (i.e., limited only by available space and noise) and Client Devices 306-308.
Obviously I haven't studied the patent in detail but that does sound to me like there are multiple antennas per base station.
Because:
1. The article claims it can use the existing cell tower system.
2. Cell tower antennas are simple dipoles, not directional antennas.
> If I were trying to do something like this I would ABSOLUTELY use multiple directional antennas in each base station AND phase things appropriately.
Yes, that would improve the performance, but this would make it a hard sell to cell companies who are trying to reduce the cost of their installed equipment. Also, to use directional antennas in a high-speed dynamic network, for a given beam width N, you would need 360/N directional antennas. For a beam width of 30 degrees, you would need 12 antennas where one exists now, and you would need a way to switch between antennas on each transmitted packet to multiple served devices. That would be a nightmare.
> Obviously I haven't studied the patent in detail but that does sound to me like there are multiple antennas per base station.
Yes, the basic idea requires multiple antennas whose relative phase can be adjusted. But not directional antennas.
This is where you went wrong. They are in fact directional antennas. From a wikipedia article:
"Due to the sectorized arrangement of antennas on a tower, it is possible to vary the strength and angle of each sector depending on the coverage of other towers in view of the sector."
http://en.wikipedia.org/wiki/Cell_site#Channel_reuse
That certainly sounds to me like a cell base station not only already has multiple antennas (which you can see with your own eyes) but that it makes use of said multiple antennas in some kind of semi-intelligent way. Obviously not as smart as the pCell, but sorta.
> For a beam width of 30 degrees, you would need 12 antennas where one exists now, and you would need a way to switch between antennas on each transmitted packet to multiple served devices. That would be a nightmare.
As we've established there are already multiple antennas. Maybe not 12, but several. And this is a trivial problem to solve actually. The same technology that tracks you from mast to mast can be used to track you from antenna to antenna. This is the job of the base station controller. Here's a clip:
"By using directional antennas on a base station, each pointing in different directions, it is possible to sectorise the base station so that several different cells are served from the same location. Typically these directional antennas have a beamwidth of 65 to 85 degrees. This increases the traffic capacity of the base station (each frequency can carry eight voice channels) whilst not greatly increasing the interference caused to neighboring cells (in any given direction, only a small number of frequencies are being broadcast). Typically two antennas are used per sector, at spacing of ten or more wavelengths apart. This allows the operator to overcome the effects of fading due to physical phenomena such as multipath reception. Some amplification of the received signal as it leaves the antenna is often used to preserve the balance between uplink and downlink signal."
http://en.wikipedia.org/wiki/Base_station_controller#Sectori...
I appreciate your enthusiasm for debate but please do some more research into cellular radio before you weigh in further. You've got a lot of the basic concepts right but not the particulars of the implementation. If that comes across harsh; I'm sorry. I don't know how to communicate this more gently.
> This is where you went wrong. They are in fact directional antennas.
No, they (simple dipoles) are not. Read my quote above -- a simple dipole is not directional. Your reply says that an array of such dipoles can be made directional, which is true and a point I made as well, but it's a different topic.
> I appreciate your enthusiasm for debate but please do some more research into cellular radio before you weigh in further.
That's my advice to you -- before you change the subject, learn enough to realize that you're changing the subject.
> If that comes across harsh
Harsh? How about wrong? Dipoles are not directional. Cell tower antennas are vertically polarized dipoles, therefore they aren't directional. An array of dipoles is directional in a crude sense and as expressed in your linked article, but not remotely comparable to a phased array, the topic of the present discussion.
> I don't know how to communicate this more gently.
Let's see if you gently recognize your error. If you move the goal posts, obviously the game changes.
You have claimed repeatedly that cell tower antennas are simple dipoles. What do you base this claim upon?
And when I say "base" I'd really like some links to in-depth explanations. I don't want to see a simple hexagonal cell system like this with A-G labeled cells and a repeating pattern (http://www.ofcom.org.uk/static/archive/ra/topics/mpsafety/sc...). That's how cell phones got started back when there weren't many users and there was a LOT of ground to cover. These days there are many, many more users and as such things have gotten more sophisticated. I included links to wikipedia articles that contain more current information.
If you can prove that they are in fact simple dipoles I have no problems accepting that they are not directional; a simple dipole is an omni. You learn that as a 3rd year EE.
The fact that the vast majority of cell tower antennas are simple dipoles -- it's the basis of the system. You want to argue -- go somewhere else. I was doing this before you were born.
http://www.unisonsite.com/pdf/resource-center/How%20Towers%2...
> I don't want to see a simple hexagonal cell system ...
No, of course not -- that would imply that you were wrong. This discussion revolves around a scheme to phase ordinary cellular antennas to improve the performance of a classic cellular system, the kind of system you don't want to talk about.
The big advantage of the scheme under discussion is that it works with a classic cellular antenna system (plus some additional electronics and mathematical processing). That makes a classic cellular system the topic.
Typical collection of vertically polarized cellular dipole transmitting and receiving antennas:
http://upload.wikimedia.org/wikipedia/commons/2/2d/Cell_Phon...
"Wireless carriers have taken the reduce and reuse approach a step further with the use of directional antennas, illustrated in Figure 5. Rather than using a single omni-directional antenna that covers a circular radius around a tower, carriers introduced directional antennas, to further segment cell sizes and enable the reuse of additional frequencies. For example, placing three antennas operating in separate frequencies on a tower allows sectors to be created within a cell, essentially tripling capacity per cell."
And a clip from a wikipedia page (http://en.wikipedia.org/wiki/Cellular_network#Directional_an...):
"Cell towers frequently use a directional signal to improve reception in higher traffic areas. In the United States, the FCC limits omni-directional cell tower signals to 100 watts of power. If the tower has directional antennas, the FCC allows the cell operator to broadcast up to 500 watts of effective radiated power (ERP).[7]"
Furthermore here is a diagram of how this is done: http://en.wikipedia.org/wiki/File:CellTowersAtCorners.gif
The picture that you linked to on the wikimedia site shows what are clearly referred to by another wikipedia page as "sector antennas". See for yourself. http://en.wikipedia.org/wiki/Sector_antenna
Here's what a single one looks like: http://en.wikipedia.org/wiki/File:Sector_antenna2.png
And here's what many of them at the top of a tower look like: http://en.wikipedia.org/wiki/File:Transmitting_tower_top_us....
Compare those with the picture that you linked which you claim is of simple dipoles. I see very little difference between the two. Which is more likely? That they're making dipoles which just happen to look exactly like sector antennas AND that they're putting up well more than they need to (since dipoles are omni they'd only need one) or they're actually sector antennas?
Finally if you're going to invoke the "I was doing this before you were born" I'm more than happy to quit now. I had this foolish idea that you (as most people on HN do) wanted to understand the world as it really is. I see now that I was mistaken, you simply want to be right. That's fine, I can't fault you for it. It's human nature. But please take that kind of corrosive attitude elsewhere. The reason this community is so great is that people here listen to data or evidence. Please don't dilute those values for the sake of "winning" a stupid internet argument.
I understand these issues perfectly well, and you are trolling. I proved my case, or weren't you paying attention? The conversation is about an unconventional scheme that relies on a conventional cellular system.
As I said before, go argue with someone else until your earlobes dry.
> Cell tower antennas are vertically polarized dipoles
> The fact that the vast majority of cell tower antennas are simple dipoles -- it's the basis of the system
You stated three times that cell tower antennas are simple dipoles. That's the problem. They're not simple dipoles. 20 or 30 years ago they absolutely were. Then a lot of users joined the system and the carriers all rolled out directional antennas in all but the most rural of places.
I'm trying to make sure that you understand that this is the point I'm arguing.
I fully agree that a 1/4 wave dipole is an omni. I fully agree that pCell probably uses omnis (at least initially). I fully agree that many years ago the cellular system used omnis.
Where I don't agree is the idea that the current cellular system uses omnis. By and large, anywhere suburban or urban, and even many rural base stations all use directional antennas. I've provided tons of links showing that the carriers use directional antennas to further break up the large cells that existed when they put all the towers up (and used omnis) into smaller cells today which can support more users.
> I proved my case, or weren't you paying attention?
Just asserting that some statement is true doesn't make it so. I've provided plenty of evidence that carriers today use multiple directional antennas to service extra customers from their existing towers. "Directional antenna" doesn't necessarily mean a 5 degree beam width, an antenna can have a 30 or 60 or 90 or 180 degree beam width and still be directional. The towers used to be in the middle of the hexagonal cells, now they're at the place where three hexagons meet.
Here is a patent by Nortel for a six-sector system in 2004.
http://www.freepatentsonline.com/6745051.html
You'll notice the familiar triangle shape holding the antennas and that there are a total of 12 antennas.
http://www.freepatentsonline.com/6745051-0-large.jpg
What does that look an awful lot like? To me it looks exactly like the middle antenna array in this picture:
http://upload.wikimedia.org/wikipedia/commons/2/2d/Cell_Phon...
I don't know how to make it any clearer than that.
Source: I do NEC simulations of my antennas for amateur radio.
A vertical dipole, something I mentioned a few posts ago, before msandford decided to turn this into a pissing contest. As a radio amateur, you will be familiar with the radiation pattern of a 1/4 wave vertical coupled to a ground plane.
That means that when you're in the pCell pocket you get excellent reception because of everything adding together, and when you move away it drops off VERY quickly.
It's smart stuff but not impossible to comprehend. I'm surprised that none of the VCs "experts" could figure it out.
Then it's not an adequate demo?
But they said that of the iPhone as well.
There's always going to be skeptical people, even if your demo is perfect and your product is a machine that turns "crap into gold"
The ratio of rejections to acceptances unfortunately is very high even for a great product (and it's possibly very low for bad products as well)
http://www.nytimes.com/2013/10/06/magazine/and-then-steve-sa...
It was demoable in a lab setting, but you certainly couldn't have given one to a consumer and expected it to go well.
sure makes it look like he is broadcasting 6 QPSK modulated video streams on the same frequency.
1\ it requires ridiculously precise location feedback from the receiving device
2\ it probably requires ridiculous amount of synchronized transmitters
3\ + lots of computation
Experts claiming it works, but not the way he sells it makes me think there is something fishy with the demo, like maybe he isnt really broadcasting on same freq or something.
Is this distributed beamforming?
> The more antennas you have, the cleaner you can keep the beam and minimize constructive interference happening in the wrong places.
Beyond three antennas surrounding a given device, this just isn't true -- more antennas don't produce a proportional increase in performance to compensate for the increased computation workload.
Yes, true, but the entire system becomes more complex for a small improvement in performance. Also, most of the theoretical work in a scheme like this would focus on the least common denominator, which is three antennas.
Interestingly, for a roughly circular array of antennas, you end up with concentric rings of high signal strength as you approach the midpoint of the array. So even the thesis that you've eliminated unused hot spots breaks down in some cases.
Here's my diagram for three phase-controlled antennas (the blue dots):
http://i.imgur.com/tyyVh0j.png
The wavelengths in this diagram are longer than for a cell system, it's just to show how this idea works. Buy adjusting the phase at each antenna, you can move the high-strength lobes around to match up with the location of a given device.
Basically the base arial seems to focus a 1cm wide beam at each device. Works nicely in his lab but dunno how well that'll work using your phone in the car.
This is not a new idea -- phased array radars do essentially the same thing -- but it's a new application of the mathematical methods, and the complexity is much higher than for a modern radar because the system has to deal with many different devices, freely moving around, and calculate the optimal combination of radio signals to produce a high signal level at that devices's current location.
It is the high computation workload that makes Perlman emphasize the role of a data center, which would need to be added to an existing cell network, to make this method possible.
Cell systems already know how to determine the location of a given device using signal arrival times at multiple antennas. This method builds on that knowledge base to improve the outgoing signals, and possibly also to phase-adjust the reception antennas, to maximize signal quality in both directions.
In essence, the system computes a protocol for each outgoing packet, beyond the packet's data content. Each packet contains additional information for each antenna that will be used to transmit it, information about phase delay. Before the packet is transmitted, the required phase delay for each of the transmitting antennas is adjusted to correspond to the calculated best solution for that device at its present location. This process is performed for each packet, for each device, which explains the high computation workload.
If physically separate antennas are used for transmitting and receiving (usually true in cell service), a separate computation would be required to most efficiently receive packets from each device. I don't know if the system has this added level of complexity, but it's possible in principle.
Yes, it's called pre-coding, every modern wireless system does it. Doing it across wireless stations is distributed beamforming which Perleman did not invent.
Present cell systems don't adjust phase to maximize reception quality in a coordinated way for each of several transmitting antennas, for each packet, for each participating device. That's new.
> Doing it across wireless stations is distributed beamforming which Perleman did not invent.
That's certainly true -- phased array radars have been doing this for decades. But Perlman should be able to get a patent anyway, based on the combination of ideas this scheme represents. Multiple, physically separate antennas, the use of a data center to compute solutions on the fly to accommodate multiple antennas and different devices with different locations, that's new.
You are just describing MIMO pre-coding which has existed for more than 20 years.
I don't see what is patentable: Distributed beamforming necessarily requires physically dispersed antennas. Adding a data center doesn't seem patentable. But I don't know anything about patents nor do I care that much.
Being a critic of software patents myself, and given how much this idea relies on software and mathematics (mathematics is definitely unpatentable), I agree that it's an issue.
I suspect that the entire method is patentable even though each element taken separately isn't.
Could this not be sped up in ASIC hardware? I'm not a hardware guy, but I would think a GPU could fit the bill for these sorts of geospatial on-the-fly calculations.
EDIT:
"Accelerating geospatial analysis on GPUs using CUDA"
http://www.zju.edu.cn/jzus/opentxt.php?doi=10.1631/jzus.C110...
Link: http://en.wikipedia.org/wiki/Mobile_phone_tracking
Quote: "Mobile phone tracking refers to the attaining of the current position of a mobile phone, stationary or moving. Localization may occur either via multilateration of radio signals between (several) radio towers of the network and the phone, or simply via GPS."
That said, I continue to remain dubious because his claims are fairly out there in terms of bandwidth. I don't know if Claude Shannon were alive today if he could figure out if there was an upper limit of directed traffic in a channel but that is the principle that feels like its being violated here. All of my training has taught me to think of a bit of spectrum like a single wire, and sure you can attach a bunch of things to that wire, but there isn't a lot of theory around how that becomes a bunch of separate but equivalent channels.
No additional bandwidth is required beyond what cell systems already provide. The idea behind this method is that each outgoing packet has some extra information about phase for each antenna that will be used to transmit it.
The extra information is calculated using Perlman's data center and added to the data structure of a specific packet meant for a specific device. The extra information tells the transmitting antennas which phase delay to apply to maximize the signal at the device's current location.
My point is that this method doesn't change the bandwidth required for the cell system -- in many ways everything is as it was before, except that each participating device experiences better reception.
This system relies on a relatively simple phase adjustment at each transmitting antenna, but because of the number of devices, the number of antennas and the math involved, a dedicated data center would have to be added to an existing cell system.
But there's no change in terms of bandwidth -- that remains the same.
"In demonstrations at his laboratory, Perlman showed off iPhones, Surface tablets, and TVs streaming massive files—the 4K UltraHD version of House of Cards from Netflix, for example—via his own wireless networking equipment"
These are both bandwidth narratives, not signal quality narratives. (granted they are the naive reporter's narratives but still). His breakthrough is that "all these devices can stream 4K ultra HD" that says that somehow he getting more bits to more machines, and that is a function of bandwidth not signal strength.
Lets assume for the moment that the basestation is connected to some multi-lambda super fiber with a 40G connection back to the Internet. He's going to give every LTE phone a 100Mbps to 1Gbps connection to that? That works for the first 400 or 40 people but then what? 5 bars and stuttering video?
This story, and others like it, have made the argument that the problem is congestion not signal clarity. Congestion is a function of channel bandwidth and operation rate. That is what leaves me dubious, I haven't seen how he can increase the bandwidth of the channel with his gizmo.
To me, the first sentence in your quote from the article is simply hyperbolic but has no connection with reality. The method doesn't do away with network congestion, it simply improves the signal at each participating device.
> Congestion is a function of channel bandwidth and operation rate.
I think the congestion this method addresses is that caused by unsophisticated signal treatment methods and an overall decline in performance. Obviously if the system can optimize the phase at each transmitting antenna and improve the received signal for each device, then more devices can use the system simultaneously. That's not how I would define "do away with network congestion", but again, I think the journalist just had no idea what he was describing.
I think the reason for all the attention given to this method is because it greatly improves the performance of a cell system without requiring any bandwidth increase. Remember that dropped packets are at times a big limitation on cell system performance, and much existing bandwidth is often wasted on packets that don't get decoded properly because of weak signals.
> It's as if only a single cell phone was in a given tower cell, and the tower was dedicating all its bandwidth to that single phone.
If the system can adjust the phases of the transmitting antennas to maximize the signal strength to a particular device, for each transmitted packet, then for all practical purposes that's true -- the entire bandwidth of the system really does become available to that particular device, for the duration of each packet meant for it.
I don't know the statistics on dropped packets, but I think it's substantial in modern times for high-speed networks. The ratio of dropped packets essentially represents a bandwidth decrease. If that could be eliminated, it would improve throughput, and someone will be tempted to claim that bandwidth has increased. Only sort of.
That could certainly accommodate a situation of everybody in a room streaming the same 4k content.
I imagine two phones X and Y and one antenna sends X XOR Y while the other sends X XOR Y XOR Y. A phone getting both signals could determine X and Y, but it would take twice as much bandwidth. If you can XOR the signals in the air through superposition somehow, X might be received in one location, and Y in another.
So 2 antennas cant service more than 2 phones at full speed, but they can service those two phones at full bandwidth. This might not seem like an advantage over directional antennas, but it relaxes the physical constraint of making sure the phones are in two separate enough places to be serviceable by the two antennas. Like maybe your phone would have two antennas and get twice the data--not very feasible if your phone were somehow serviced by two directional antennas.
I haven't worked out how superposition could do something exactly like XOR, or if it maybe has a minimum of 3 antennas to start being possible, etc., but it seems plausible.
Anyway, the idea would be sending signal X and Y. You want X interfering with Y at some time offset to yield desired phone signal A, and X interfering with Y at some other time offset to yield phone signal B. The phones never receive X or Y, but some superposition of the two, The phones couldn't cheat and just record the two waves and timeshift internally to yield twice the bandwidth, because they don't receive the two waves separately, they only get the superposition of the two.
XOR is just a primitive way of making an orthogonal code.
I would expect he would be more successful patenting the hell out of it and licensing it to Ericsson, NSN, Qualcomm, Cisco, etc than trying to sell it to individual venues and customers.
I'm predicting that the breathless claims of "no limit" are based on being the only show in town. As long as the system is controlling every transmission in a volume of space it will do well, and you will be able to add users until you reach the same limits that a MIMO system would hit if it was the only transmitter. Start adding uncontrolled interference, and things will start to degrade.
Maybe the next step, would be for the distributed base stations to use receivers to sample the spectrum. This data could be fed back into the cloud and an attempt be made to predict and cancel the uncontrolled interferers. The FCC (or local equivalent) might take a dim view of this though, if it turned out that the cancellation was effectively jamming other users.
--- Edit.
Just adding, in my experience, if you want to understand the latest radio communications technology, speak to a radio astronomer. When MIMO first came out and people were trying to figure out its limits, it turned out that the radio astronomers already knew the limits, based on their knowledge of large antenna arrays. I'd guess the astronomers would look at this current technology and think "very long baseline interferometry".
See: http://blog.airtightnetworks.com/bang-for-the-buck-with-expl...
Typical beamforming just boosts the signal quality, on a slice of the available spectrum. But it would be faster if you used the whole bandwidth. But then no one else would be able to use that bandwidth. But if you used a parabolic antenna pointed at each receiver, then each phone would get the full bandwidth. But if the phone is moving, you can use (tracking antennas or...) deliberate interference within a "mesh" of antennas.
http://groups.csail.mit.edu/netmit/wordpress/projects/spectr...
Calling this a machine is like calling the first car a horseless carriage.
Um, they did do that, you know. I agree with your point, but the analogy is broken.
So it's an excellent description which leverages an analogy everyone is currently familiar with? :P
1) You need to have multiple base stations installed roughly equally spaced. This seems to support heterogenous spacing of the stations
2) With lots of low-powered base-stations interference and obstructions start to become bigger issues. With many reachable antennas, a small fraction that have interference or obstructions should merely decrease SNR a bit rather than drop completely.
Of course advantage #2 seems (to me) to be theoretical; I'd like to see how this performs in the real-world.
The other issue I can see is moving receivers and/or moving reflectors. Imagine having this work, for example, when you are near a wind-farm, with many moving reflectors changing the apparent signal strength continuously.
I know someone who worked on a similar problem in the past, and temporal coherence in real-world situations was close enough to the feasible round-trip delay to the data-center. (e.g. if you need to recalculate every 150ms and it takes 125ms round-trip to the data center, you only get 1/6 the theoretical benefit).
His white paper probably addresses the more inquisitive and technical readers' questions: http://www.rearden.com/DIDO/DIDO_White_Paper_110727.pdf
802.11ac beamforming justs boosts signal quality. It does not take advantage of more bandwidth.
I don't think the math is actually that complicated to explain. I think of it as the electromagnetic equivalent of visual cryptography: http://datagenetics.com/blog/november32013/index.html
An antenna not integrated into the system would see noise. The designated phone would see the exact interference pattern it needs to get the "picture."
Because this is whats needed for this to work.
Mind. Blown.
http://i.imgur.com/tyyVh0j.png
The antenna locations are marked with blue dots.
In this diagram the size of the high-signal-strength lobes is much larger than would be true at cellular frequencies, but it gives a sense of how this idea works. By adjusting the phase at each antenna, the lobes can be moved around to accommodate different devices within their range.
As I expected, increasing the number of antennas past three doesn't really improve the outcome enough to justify the increased complexity.
https://en.wikipedia.org/wiki/Phased_array
More antennas can enable you to narrow lobes more - you may have not been phasing them properly to get that effect.
In a cellular antenna scenario, unlike phased array radar, you cannot avoid multiple lobes, but the advantage of phasing is still present. Adding antennas to the system (beyond three or four) becomes a case of diminishing returns.
> More antennas can enable you to narrow lobes more - you may have not been phasing them properly to get that effect.
There are paradoxical effects with more than a few antennas -- apart from the complexity issue. In some cases you will see lobes with strange shapes in locations having no benefit. With a circular array of antennas, you sometimes get concentric circles of high signal strength, not what you're after in the described system. Example:
http://i.imgur.com/hK9BZnF.png
Intuitively, the diagram seems to say that behind (in front of?) each blue dot (antenna) the signal is really strong while on the other side it kinda drops down, only to strengthen again when we hit the middle of the triangle. But I really have no idea what's going on...
Thanks.
It's a simple density plot of an electromagnetic field in which the brighter colors represent constructive interference and dark areas represent destructive interference. The three antennas constructively or destructively interfere, with varying effects at different locations within the field.
I want to say again that the diagram is only an example and uses a frequency much lower than that used by cellular systems, just so the lobes are a reasonable size, and only to give some idea of the principles at work.
> Intuitively, the diagram seems to say that behind (in front of?) each blue dot (antenna) the signal is really strong while on the other side it kinda drops down, only to strengthen again when we hit the middle of the triangle.
Yes, that's true -- the three external lobes happen to be places where the three antennas constructively interfere. One can do this using pencil and paper (as I did when I was a NASA Space Shuttle engineer many years ago, and before there were personal computers), but computers make it much easier to experiment with different configurations.
Here's a different antenna configuration, one that I doubt anyone would want to build -- a circle of antennas, giving circular intensity lobes:
http://i.imgur.com/hK9BZnF.png
Not very useful, but interesting.
мʏ вυɖɖʏ'ѕ ℮х-աιғ℮ мαĸ℮ѕ $66 нօυʀʟʏ օɴ тн℮ ʟαքтօք. ѕн℮ нαѕ в℮℮ɴ ғιʀ℮ɖ ғʀօм աօʀĸ ғօʀ 9 мօɴтнѕ вυт ʟαѕт мօɴтн н℮ʀ ƈн℮ƈĸ աαѕ $15718 ʝυѕт աօʀĸιɴɢ օɴ тн℮ ʟαքтօք ғօʀ α ғ℮ա нօυʀѕ. ɢ℮т тн℮ ғαƈтѕ SaveJury.com
I wonder what spammers will come up with next.