Artemis is building a 4G network in SF to prove its pCell tech works(gigaom.com)
gigaom.com
Artemis is building a 4G network in SF to prove its pCell tech works
https://gigaom.com/2015/02/23/artemis-is-building-a-4g-network-in-sf-to-prove-its-pcell-tech-works/
8 comments
I totally agree - they even mention Webpass in the linked article, though only in passing. Webpass is an established niche ISP in San Francisco (they sell 100Mbit symmetric home internet service for $55/mo, but serve a very limited area).
A cell-phone service that only covered San Francisco isn't worth very much to me, but I'm sure many residents would jump at the chance to ditch Comcast for their home internet service, and Webpass is probably just the partnership Artemis needs - assuming the technology pans out.
A cell-phone service that only covered San Francisco isn't worth very much to me, but I'm sure many residents would jump at the chance to ditch Comcast for their home internet service, and Webpass is probably just the partnership Artemis needs - assuming the technology pans out.
Just started reading the whitepaper, and something does not make sense. Their experiments are done in a controlled room with multiple antennas that are being directed towards one pCell. That makes sense. But when this technology goes outdoors, where the RF propagation is extremely hard to accurately predict (especially because of the 'changing environment'), how is the pCell created (assuming you have to have accurate location and propagation information to create the correct interference patterns from multiple antennas)? Something smells here.
I am not expert in this area, but I watched their presentation in NYC (or was it Columbia U) about a year ago. The way he avoided answering any technical questions made me think these guys are either full of shit, or they have such a great technology that don't need to convince anybody with their world, people will see it in action and believe them. I am now inclined to believe it is not magic. But time will tell.
The 5G capabilities of the technology look amazing on paper. Hope it works out in the real world.
Those interested in more technical details should check out the whitepaper: http://www.rearden.com/artemis/An-Introduction-to-pCell-Whit...
Page 81 talks about pCell VR, their ideas about cloud based wireless VR tech.
Those interested in more technical details should check out the whitepaper: http://www.rearden.com/artemis/An-Introduction-to-pCell-Whit...
Page 81 talks about pCell VR, their ideas about cloud based wireless VR tech.
Last time I asked there were some doubts regarding this technology (see https://news.ycombinator.com/item?id=7732821). Is the general consensus still that this is snake oil?
I went to school for EE/CE (double major) and I ended up getting a masters as well. I don't believe that there's anything that prevents this technology from working.
The main problem is that most of the "rules of thumb" or whatever you learn in school are based on a few assumptions that have served us well over the years. Antennas aren't directional, or are only marginally so. My signal is your noise and vise versa. Things like that.
What this technology does (along with several others I've read about) is challenge those underlying assumptions and by doing so, gets performance in excess of "what is possible" only so long as those underlying assumptions hold. But because a lot of people weren't taught WHY those assumptions were made, they believe the conclusions that result are fundamental laws of the universe rather than a good model for understanding.
For example, the general noise floor in the 1-2GHz range is about 1000 times higher than GPS signals and according to classical models, you can't recover it. http://www.gpssource.com/faqs/15 But as it turns out there's a LOT of redundancy and such built into the signal and through a method called "process gain" you can in fact recover it, even though it's way below what you'd originally think is possible.
Similarly radio controlled cars/planes/helicopters/boats/etc have limited frequency bands in which they can operate. This poses a problem because you can't generally fly the airplane around your house, but instead you go to a R/C airfield, where unfortunately there are a bunch of other enthusiasts there too, all vying for the same spectrum. If my radio interferes with yours, there goes both of our hard work crashing to the ground. This problem has caused HUGE uptake of DSSS technology whereby we can both transmit in the same band, but neither of us interferes with one another. This is done by coding both signals with a pseudorandom sequence such that my signal looks like noise to you, and your signal looks like noise to me, and our systems are capable of working even though there's some noise.
http://en.wikipedia.org/wiki/Direct-sequence_spread_spectrum
What the pCell is doing is basically like DSSS. But where DSSS is done strictly through time, the pCell is done through both TIME and SPACE. My signal for me is coherent only where I am, and your signal for you is only coherent where you are, and anywhere other than those places, it's just random noise.
The main problem is that most of the "rules of thumb" or whatever you learn in school are based on a few assumptions that have served us well over the years. Antennas aren't directional, or are only marginally so. My signal is your noise and vise versa. Things like that.
What this technology does (along with several others I've read about) is challenge those underlying assumptions and by doing so, gets performance in excess of "what is possible" only so long as those underlying assumptions hold. But because a lot of people weren't taught WHY those assumptions were made, they believe the conclusions that result are fundamental laws of the universe rather than a good model for understanding.
For example, the general noise floor in the 1-2GHz range is about 1000 times higher than GPS signals and according to classical models, you can't recover it. http://www.gpssource.com/faqs/15 But as it turns out there's a LOT of redundancy and such built into the signal and through a method called "process gain" you can in fact recover it, even though it's way below what you'd originally think is possible.
Similarly radio controlled cars/planes/helicopters/boats/etc have limited frequency bands in which they can operate. This poses a problem because you can't generally fly the airplane around your house, but instead you go to a R/C airfield, where unfortunately there are a bunch of other enthusiasts there too, all vying for the same spectrum. If my radio interferes with yours, there goes both of our hard work crashing to the ground. This problem has caused HUGE uptake of DSSS technology whereby we can both transmit in the same band, but neither of us interferes with one another. This is done by coding both signals with a pseudorandom sequence such that my signal looks like noise to you, and your signal looks like noise to me, and our systems are capable of working even though there's some noise.
http://en.wikipedia.org/wiki/Direct-sequence_spread_spectrum
What the pCell is doing is basically like DSSS. But where DSSS is done strictly through time, the pCell is done through both TIME and SPACE. My signal for me is coherent only where I am, and your signal for you is only coherent where you are, and anywhere other than those places, it's just random noise.
The theory isn't snake oil, but it isn't novel (it's called network MIMO) and I'm not sure if performance has been established in practice.
What they have demoed and are building out now is a pretty incredible advance in modern radio systems. So there is healthy room for skepticism.
It seems like there would be some major engineering challenges ahead, with predicting localization and movement of each receiver to adjust the location of the 'cell', and dealing with multipath. So this might explain why they are going slowly with building it out.
But I think that it is not snake oil, just a disruptive innovation that is slowly trying to get a foothold after coming out of relative secrecy in development for years. If they get it right, though, then it will absolutely reinvent the cellular industry.
It seems like there would be some major engineering challenges ahead, with predicting localization and movement of each receiver to adjust the location of the 'cell', and dealing with multipath. So this might explain why they are going slowly with building it out.
But I think that it is not snake oil, just a disruptive innovation that is slowly trying to get a foothold after coming out of relative secrecy in development for years. If they get it right, though, then it will absolutely reinvent the cellular industry.
It's still snake oil. What they're designing is called a thinned phased array, or sparse phased array, see figure 19 in the white paper. Such design suffers from a major flaw known to any radar engineer: The Thinned Array curse. It even has a wikipedia article:
https://en.wikipedia.org/wiki/Thinned-array_curse
In layman terms, >99% of the power transmitted is lost to sidelobes and doesn't reach the intended users (1-a/A power is lost, if you want to get technical). Such design can carry a small number of users, operating very close to the environment noise floor. The maximum allowed number of users is approximately proportional to the number of transmitters. Each user added above this limit decrease the SNR for all users in the system, killing the communication for everyone. Not a sound design for a cellular system.
Phased-array design is complex and not always intuitive. Whoever invested in this company hasn't done proper due diligence.
https://en.wikipedia.org/wiki/Thinned-array_curse
In layman terms, >99% of the power transmitted is lost to sidelobes and doesn't reach the intended users (1-a/A power is lost, if you want to get technical). Such design can carry a small number of users, operating very close to the environment noise floor. The maximum allowed number of users is approximately proportional to the number of transmitters. Each user added above this limit decrease the SNR for all users in the system, killing the communication for everyone. Not a sound design for a cellular system.
Phased-array design is complex and not always intuitive. Whoever invested in this company hasn't done proper due diligence.
Sorry, I don't think you are correct. What they are building is not just another phased-array system, and I don't think they are even doing any beamforming at each transmitter at all.
Instead, they are looking at a much more difficult task, of using constructive and destructive interference from distributed transmitters to only cohere the signal at a single point around the receiving antenna. Think CDMA, but spatially (like in 3d space).
> Radio frequency design and phased array design is complex and not always intuitive. Whoever invested in this company hasn't done proper due diligence.
One of the problems with what they are doing is that it is so new and flies in the face of decades of radio theory, which as you state is already incredibly difficult. Please give it another look, this video as well: https://www.youtube.com/watch?v=5bO0tjAdOIw
Instead, they are looking at a much more difficult task, of using constructive and destructive interference from distributed transmitters to only cohere the signal at a single point around the receiving antenna. Think CDMA, but spatially (like in 3d space).
> Radio frequency design and phased array design is complex and not always intuitive. Whoever invested in this company hasn't done proper due diligence.
One of the problems with what they are doing is that it is so new and flies in the face of decades of radio theory, which as you state is already incredibly difficult. Please give it another look, this video as well: https://www.youtube.com/watch?v=5bO0tjAdOIw
I'm not agreeing or disagreeing about this being snake oil... but how do you think phased-array and beam forming work if not by controlling phase delays between antennas so as to create constructive and destructive interference at a desired point, or points, in space?
Even if the air interface works the main reason I can see for carriers not deploying this is backhaul. It would be exponentially more expensive (35x if the air interface is indeed 35x faster) to provide bandwidth to all devices. Right now carriers actually rely on the LTE total cap limitation to save money on backhaul costs.
Even if the air interface works the main reason I can see for carriers not deploying this is backhaul. It would be exponentially more expensive (35x if the air interface is indeed 35x faster) to provide bandwidth to all devices. Right now carriers actually rely on the LTE total cap limitation to save money on backhaul costs.
> how do you think phased-array and beam forming work if not by controlling phase delays between antennas so as to create constructive and destructive interference at a desired point, or points, in space?
I consider DIDO separate from a phased array in that each transmitter has its own separately transmitted signal to do the interfering, instead of one phase-shifted copy of the signal.
Artemis/pCell is definitely not a phased array in that sense.
I consider DIDO separate from a phased array in that each transmitter has its own separately transmitted signal to do the interfering, instead of one phase-shifted copy of the signal.
Artemis/pCell is definitely not a phased array in that sense.
This could make the backhaul problem even worse- because in order to get that 35x bandwidth , you have to transport ALL that data to a large group of base-stations,unlike today where you transport data only to the relevant base-station.
On the other hand, if a company invented a new wireless technology, it's probably smart enough to be aware of the back haul problem.
On the other hand, if a company invented a new wireless technology, it's probably smart enough to be aware of the back haul problem.
I don't know how loosely you want to apply the term "beamforming" but there are ways to send data over multiple antennas for multiple users that allows the signal to be recovered for each user independently even if the signals nominally appear to be interfering.
"Constructive and destructive interference from distributed transmitters" is basically the definition of a sparse phased array... and it suffers from the Thinned Array Curse. Every radio engineer sooner or later re-discovers this classic "curse" :)
These aren't naive phased arrays systems and if the performance really were worse as you imply no one would use MIMO.
There are no results in the white paper showing how this technology works with moving terminals. It seems to me that if this were to be true, the requirements on real-time processing would be incredibly humongous.
The real problem with this type of technology is that you'll see great performance with a small number of subscribers.
So unless they will be able to get enough subscribers the results will still be questionable.
So unless they will be able to get enough subscribers the results will still be questionable.
But the goal of this experiment isn't to be profitable, just prove the technology works so they can sell it. I'm sure they'll be able to entice enough people in tech-savvy SF to buy and use one of their SIM cards for the sake of making their case.
Screw that. Build one in my town to prove 4G exists.
I think this is possible.
Also, some small cellular networks don't offer roaming and only work in very limited geographic areas. Many have successfully found niches among people who don't travel much or have alternatives when they are traveling.