pCell, A Bold Scheme For Super-Fast Wireless Data(readwrite.com)
readwrite.com
pCell, A Bold Scheme For Super-Fast Wireless Data
http://readwrite.com/2014/03/19/pcell-artemis-wireless-dido-steve-perlman
10 comments
This looks like it'd work for fixed broadband installations (say, for people in rural Montana or Ontario), but I'm less convinced it'd work well for cell phones because the tower would need to predict your movements in advance of you actually moving there in order for it to know where to build the constructive interference.
The receiver can determine the "Direction Of Arrival" of a received signal which it uses for beam forming (this is how it typically done, don't know anything about pCell).
Perlman demonstrates [0] some sort of 'adaptive algorithm' that pCell uses to track real-time a UE's position.
[0] http://youtu.be/wGAnDQEQJ_s
[0] http://youtu.be/wGAnDQEQJ_s
I'm surprised at how rough the audience was for the presentation. I wonder where those emotions came from.
Sigh
The biggest limiting factor of wireless is the lack of frequency in which to put you're data. Not interference.
pCell only works for small areas, what it does is essentially beamform a signal directly to the consumer (basically cupping your hands around your mouth when you shout at someone) I assume the "magic" comes from some sort of noise cancelling (yes just like the ear phones)
However in practice I assume they just use more than one beam former to create more than one beam.
It cannot be used for wide area systems, as the only reason this works is because its local. it provides bandwidth for one small area using the same frequency that the wider area is using.
The biggest limiting factor of wireless is the lack of frequency in which to put you're data. Not interference.
pCell only works for small areas, what it does is essentially beamform a signal directly to the consumer (basically cupping your hands around your mouth when you shout at someone) I assume the "magic" comes from some sort of noise cancelling (yes just like the ear phones)
However in practice I assume they just use more than one beam former to create more than one beam.
It cannot be used for wide area systems, as the only reason this works is because its local. it provides bandwidth for one small area using the same frequency that the wider area is using.
Interference is pretty much the only the limiting factor locally for information capacity for any given frequency band. What Artemis is doing is "spacetime" coding signal in such a way constructive interference forms at desired locations.
Part of the "magic" apparently comes from return channel analysis, nothing at all like noise cancelling earphones. This doesn't try to cancel existing interference in any way.
This is a superset of beamforming, a more generic solution with arbitrary positioning. Nothing is ever completely new.
I'm not sure what you mean by wide area systems, but I don't see any reason why the range can't be even hundreds of kilometers, limited just by line of sight and signal attenuation. Roundtrip delay might affect doppler compensation at greater distances.
Part of the "magic" apparently comes from return channel analysis, nothing at all like noise cancelling earphones. This doesn't try to cancel existing interference in any way.
This is a superset of beamforming, a more generic solution with arbitrary positioning. Nothing is ever completely new.
I'm not sure what you mean by wide area systems, but I don't see any reason why the range can't be even hundreds of kilometers, limited just by line of sight and signal attenuation. Roundtrip delay might affect doppler compensation at greater distances.
> The biggest limiting factor of wireless is the lack of frequency in which to put you're data. Not interference.
Isn’t the idea to send data to multiple devices at the same frequency simultaneously? Signal interferences are prevented by clever timing of the three senders, namely by creating interferences that add up to noise everywhere except for a small volume that hopefully contains the receiving device.
Isn’t the idea to send data to multiple devices at the same frequency simultaneously? Signal interferences are prevented by clever timing of the three senders, namely by creating interferences that add up to noise everywhere except for a small volume that hopefully contains the receiving device.
Thats spread spectrum. Which is what wifi uses. each device raises the noise floor, until you can't add more devices.
Can you elaborate why it wouldn't work in a wide area setting? In a demonstration, they had a lot of cellphones on a desk, but not a matching number of pCell antennas, so it seems they can support many devices with a single (or at least much less) antenna.
Or do you mean to say that the beam forming is not going to work over large distances, so phones must be in close proximity to the antennas?
Or do you mean to say that the beam forming is not going to work over large distances, so phones must be in close proximity to the antennas?
the system will work over long distances, indeed he claims in the paper (with nothing to back it up, that paper is extra-ordinarily vague.)
However it wont increase bandwidth. (I should have been more precise) The only reason why people are interested in this is because it offers the idea of greater bandwidth.
The way they increase bandwidth is effectively partitioning the local airspace off from the wider world. Each cell (be it TV, phone or $other) can deliver x bandwith.
The bigger the area covered by the cell, the less bandwidth per unit of volume. (conversely the smaller the more bandwidth per volume)
This is before we start doing clever things like spread spectrum or account for signal loss/noise/shannon's law
This is pretty much snake oil. Its beam forming with a marketing budget. There is nothing in that paper that suggests otherwise
However it wont increase bandwidth. (I should have been more precise) The only reason why people are interested in this is because it offers the idea of greater bandwidth.
The way they increase bandwidth is effectively partitioning the local airspace off from the wider world. Each cell (be it TV, phone or $other) can deliver x bandwith.
The bigger the area covered by the cell, the less bandwidth per unit of volume. (conversely the smaller the more bandwidth per volume)
This is before we start doing clever things like spread spectrum or account for signal loss/noise/shannon's law
This is pretty much snake oil. Its beam forming with a marketing budget. There is nothing in that paper that suggests otherwise
Just as a sidenote, LTE can also use beamforming, over rather large distances. How well implemented this is in current gear , I don't know.
I have seen some statements of Perlman and persons related to him or his companies, that this technology is not only applicable to communication networks. That, and the recent patent applications, make me think that he is talking about wireless power.
I really wonder if, and I have some doubts that, the underlying technology is at all feasible for such an application?
I really wonder if, and I have some doubts that, the underlying technology is at all feasible for such an application?
Here's the actual white paper, which is fairly readable:
http://www.rearden.com/DIDO/DIDO_White_Paper_110727.pdf
I wish he would have taken Apple's lead on having a big launch with a product that ships in a couple weeks rather than a dribble of presentations with a shipping: ??? slide.
If pCell works, getting press won't matter. If you can destroy the markets for mobile data + home internet you won't have to fight Stanford undergrads in some anonymous lecture hall. So, I am sorta skeptical.
If pCell works, getting press won't matter. If you can destroy the markets for mobile data + home internet you won't have to fight Stanford undergrads in some anonymous lecture hall. So, I am sorta skeptical.
Question: The article mentions that pCells use interference to their advantage. Would a network of these be able to operate in a public spectrum?
With the licensing environment that exists today the answer is a definite "uhhhh maybe?!"
If you wanted to do it with LTE gear no since it's fixed to non-public spectrum.
If you did some serious wifi hacking to use the same concepts you might be able to. But with the area that wifi covers and the generally terrible power efficiency of the amplifiers I'm not entirely sure you'd want to. I don't think you'd be able to call it pCell but you could probably use the DIDO ideas and make it happen if sufficiently motivated.
If you wanted to do it with LTE gear no since it's fixed to non-public spectrum.
If you did some serious wifi hacking to use the same concepts you might be able to. But with the area that wifi covers and the generally terrible power efficiency of the amplifiers I'm not entirely sure you'd want to. I don't think you'd be able to call it pCell but you could probably use the DIDO ideas and make it happen if sufficiently motivated.
I don't see any legal reason why you couldn't do something like this today. There are already WiFI routers that do beamforming and this just seems like an application of beam forming where your antennas are much further apart.
secong gen 802.11ac will do this too, MU-MIMO will use beam forming to send up to 4 streams to up to four clients simultanously.
http://chimera.labs.oreilly.com/books/1234000001739/ch04.htm...
http://chimera.labs.oreilly.com/books/1234000001739/ch04.htm...
It is not the same thing. 11ac does normal beam forming not network/distributed MIMO.
secong gen will be distributed
For all interested, here is a link to a recent talk Perlman gave at Stanford: https://www.youtube.com/watch?v=wGAnDQEQJ_s
Since many will not watch through the whole thing here is a basic synopsis.
If you are looking for pCell's strength to be in any sort of physical layer innovation, stop. It's not the actual antenna 'box' that you see in photos that is doing the heavy work. That is just a basic RF frontend that Artemis could care less if you broke it open and reverse engineered every component inside.
Perlman's talk confirms my suspicions that this is not an RF advancement at all. It's a software advancement. The real work is being done by the "datacenter". The overall RAN architecture is a CRAN (Cloud RAN) architecture, which means that it fails miserably to be innovative without it's special software.
We are not talking about protocol stack software (as found in LTE eNodeB's and LTE capable UEs). We are talking about Artemis's proprietary software (which must be some killer highly optimized, low-level, kick-ass mathematics and networking algorithms).
Artemis's datacenter seems to create a virtual enodeB for every device accessing the CRAN (the access points being the 'Cloud' of those antenna boxes you see in the photos/videos which work together).
Each device then communicates physically through the boxes but virtually with an enodeB "server" in the datacenter. Thus, it reads the environment AS IF it has an ENTIRE basestation (tower) all to itself. :P
Now if you know every device's information via uplink information provided by the protocol stack and every access points information provided by the already in-house data then you could theoretically play around heavily with wave-front mathematics.
But I think pCell's potential is being a bit overestimated. I remember Perlman saying something like pCell offering a 10x advantage or something. (Sorry don't remember every detail I watched the video late last night).
That advantage level is cool, but it seems Artemis is pushing (and for good reason) not some sort of instant HUGE advantage with pCell, but a long-term scalable advantage.
According to their claims, it seems that you should just be able to add more access points (as well as extend the processing capability of the datacenter) to increase the overall capability of the CRAN.
It would be neat to know what algorithms they are using, but they really seem to not want to say anything that might even come close to giving it away.
I am not sure what aspects of the channel quality and UE feedback they use from the protocol stacks (LTE, wi-fi, etc.) but they seem to have something working.
We shall see what happens.
EDIT: It's a funny way to think about it, but it seems that the 'innovation' exists outside of the protocol stack. It's like a "sub-physical layer" that "transports" the actual link between the device and the virtual enodeB by placing the wave-front 1cm around the antenna(s) of the device? (Open to critiques on anything I said) :)
Since many will not watch through the whole thing here is a basic synopsis.
If you are looking for pCell's strength to be in any sort of physical layer innovation, stop. It's not the actual antenna 'box' that you see in photos that is doing the heavy work. That is just a basic RF frontend that Artemis could care less if you broke it open and reverse engineered every component inside.
Perlman's talk confirms my suspicions that this is not an RF advancement at all. It's a software advancement. The real work is being done by the "datacenter". The overall RAN architecture is a CRAN (Cloud RAN) architecture, which means that it fails miserably to be innovative without it's special software.
We are not talking about protocol stack software (as found in LTE eNodeB's and LTE capable UEs). We are talking about Artemis's proprietary software (which must be some killer highly optimized, low-level, kick-ass mathematics and networking algorithms).
Artemis's datacenter seems to create a virtual enodeB for every device accessing the CRAN (the access points being the 'Cloud' of those antenna boxes you see in the photos/videos which work together).
Each device then communicates physically through the boxes but virtually with an enodeB "server" in the datacenter. Thus, it reads the environment AS IF it has an ENTIRE basestation (tower) all to itself. :P
Now if you know every device's information via uplink information provided by the protocol stack and every access points information provided by the already in-house data then you could theoretically play around heavily with wave-front mathematics.
But I think pCell's potential is being a bit overestimated. I remember Perlman saying something like pCell offering a 10x advantage or something. (Sorry don't remember every detail I watched the video late last night).
That advantage level is cool, but it seems Artemis is pushing (and for good reason) not some sort of instant HUGE advantage with pCell, but a long-term scalable advantage.
According to their claims, it seems that you should just be able to add more access points (as well as extend the processing capability of the datacenter) to increase the overall capability of the CRAN.
It would be neat to know what algorithms they are using, but they really seem to not want to say anything that might even come close to giving it away.
I am not sure what aspects of the channel quality and UE feedback they use from the protocol stacks (LTE, wi-fi, etc.) but they seem to have something working.
We shall see what happens.
EDIT: It's a funny way to think about it, but it seems that the 'innovation' exists outside of the protocol stack. It's like a "sub-physical layer" that "transports" the actual link between the device and the virtual enodeB by placing the wave-front 1cm around the antenna(s) of the device? (Open to critiques on anything I said) :)
Whenever I see something that claims to overcome a long tested theory like Shannon's law, it always makes me very suspicious. The explanations I have seen so far are long on claims and short on proof and remind me very much of "perpetual motion machines". So far nothing I have seen shows how they are "side stepping" Shannon. I smell a rat.
This doesn't violate the Shannon capacity theorem: Different channels, like MIMO channels, have different capacities.
pCell is basically a proprietary version of network MIMO. (Though maybe still valuable if it can get to market faster).
pCell is basically a proprietary version of network MIMO. (Though maybe still valuable if it can get to market faster).
They are sidestepping Shannon by taking advantage of spacial multiplexing.
If I'm user fiber-optic communications with 1064nm laser, there's a maximum bandwidth I can get out of a single fiber. But I can always just add more fibers.
You can do this with radio waves too. Just broadcast in such a way that a signal has maximum strength at the desired user's location, and less strength elsewhere.
You need multiple antennas, of course (like multiple fiber heads). I think if you wanted 100% bandwidth for each user, you'd need at least one antenna per user.
If I'm user fiber-optic communications with 1064nm laser, there's a maximum bandwidth I can get out of a single fiber. But I can always just add more fibers.
You can do this with radio waves too. Just broadcast in such a way that a signal has maximum strength at the desired user's location, and less strength elsewhere.
You need multiple antennas, of course (like multiple fiber heads). I think if you wanted 100% bandwidth for each user, you'd need at least one antenna per user.
https://news.ycombinator.com/item?id=7316606