Previous discussion on HN (with participation from the inventor):
pCell, A Bold Scheme For Super-Fast Wireless Data
11–20 of 28 posts
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#12This 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.
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#13This 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.
Perlman demonstrates [0] some sort of 'adaptive algorithm' that pCell uses to track real-time a UE's position.
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#14For 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) :)
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#15For 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 Art…
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#16This 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.
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
I'm surprised at how rough the audience was for the presentation. I wonder where those emotions came from.
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#17I 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.
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#18secong 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...
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#19I 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?
Re: pCell, A Bold Scheme For Super-Fast Wireless Data
#20secong 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...
It is not the same thing. 11ac does normal beam forming not network/distributed MIMO.