The key to understanding the possible limitation of pcell which is most likely some form of CoMP, massive MIMO or CloudRAN is the need for very low latency backhaul and scaling.
Yes, he states scaling is linear which is good up to a point. From the patent filings it appears that for N users they need M antennas. As N increases M increases linearly. As long as the ratio is decent this scales economically.
The other potential bottleneck is low latency backhaul requirement. It may be that they need direct line of sight microwave backhaul with few hops not because of economy or speed but because the fiber backhauls to towers introduce too much latency for them to do the calculations required. This is the traditional Achilles heel of network mimo.
Probably they have solved 90% of the problems AND gotten it to work on traditional LTE phones and that's killer. But if the need for low latency backhaul and a large number of antennas it's not going to be as simple a deployment as putting one of their antennas on existing towers...even those upgraded recently to a fiber connection from older T1 copper.
The key to understanding the possible limitation of pcell which is most likely some form of CoMP, massive MIMO or CloudRAN is the need for very low latency backhaul and scaling.
Yes, he states scaling is linear which is good up to a point. From the patent filings it appears that for N users they need M antennas. As N increases M increases linearly. As long as the ratio is decent this scales economically.
The other potential bottleneck is low latency backhaul requirement. It may be that they need direct line of sight microwave backhaul with few hops not because of economy or speed but because the fiber backhauls to towers introduce too much latency for them to do the calculations required. This is the traditional Achilles heel of network mimo.
Probably they have solved 90% of the problems AND gotten it to work on traditional LTE phones and that's killer. But if the need for low latency backhaul and a large number of antennas it's not going to be as simple a deployment as putting one of their antennas on existing towers...even those upgraded recently to a fiber connection from older T1 copper.
It appears that low-latency backhaul is addressed by Artemis' "Serendipitous Deployment" where the deployment of their Transceivers is largely determined by access to backhaul (line of sight Microwave or fiber cable) and cost of deployment.
Because Artemis takes advantage of interference it appears that target devices can be served by placing their transceivers anywhere -- not limited to existing towers, And these transceivers can broadcast at more power and for longer distances than current cell placements. Also, I suspect that a future, more robust Artemis transceiver could perform partial [distance/location/tracking ?] calculations on, say, the 50-100 nearest pCell devices; then transmit that to the servers for simpler [faster] trilateration and pCell intersection wave calculation -- reducing the latency and scaling requirements of the servers.
If the pCell can also be used to charge/maintain the battery power in the target device -- then the target device, itself, could take the place of the "more robust transceiver" discussed above. I am experimenting with trilateration calculations on iDevices with iBeacons -- and the calculations are fast. An iPhone with an A7 and an M7 could efficient perform these calculations, as needed -- as indicated by the M7 when movement or direction changes.
Artemis claims that a native pCell radio in the device (or the LTE radio supporting pCell) uses a fraction of the power required by a WiFi -- and, it appears, that it provides faster and more precise location positioning.
I am trying to look at this pCell system from a skeptically optimistic perspective -- If doable, it has some amazing potential.
It appears that low-latency backhaul is addressed by Artemis' "Serendipitous Deployment" where the deployment of their Transceivers is largely determined by access to backhaul (line of sight Microwave or fiber cable) and cost of deployment.
It's only serendipitous if you can find it. Regions with lots of tall buildings with good LOS to each other is serendipitous. Regions where LOS is limited is not so serendipitous.
This is called handwaving. The issue here is dependent on the existing backhaul topology. Daisy chain is probably worst because of the number of hops increasing your latency but that's what many 2G and 3G backhaul networks were doing. Hub and spoke would be best (and probably closest tot their microwave topology) but few deployments are this way. Mesh would be next best or perhaps Tiered.
If you have a lot of tall buildings like SF or NY great. If you are in suburbia probably not so great. It depends on the latency requirements for p-cell to work and the average and worst case latency for the backhaul topology you have to work with.
Quote:
Because Artemis takes advantage of interference it appears that target devices can be served by placing their transceivers anywhere -- not limited to existing towers,
If you are deploying in suburbia then you can't place your transcievers anywhere because there are few sites with good LOS with each other or the central hub. You ARE limited to the existing towers because that's where the backhaul is. If you are deploying to a region where they ripped out the older daisy chain topology and replaced it with a topology with sufficiently low latency to let pcell work then that's fine.
Some markets will be in a more favorable state for massive MIMO deployment. Others won't.
Quote:
And these transceivers can broadcast at more power and for longer distances than current cell placements. Also, I suspect that a future, more robust Artemis transceiver could perform partial [distance/location/tracking ?] calculations on, say, the 50-100 nearest pCell devices; then transmit that to the servers for simpler [faster] trilateration and pCell intersection wave calculation -- reducing the latency and scaling requirements of the servers.
Transmitting a bearing and range estimate vs just signal and bearing isn't going to greatly reduce your latency requirements to the central servers doing the calculations for beam forming and traffic encoding. As only one tower you can't get more than bearing and range estimate.
The scaling requirements are not so important for the servers. The scaling requirements are important with relation to the number of antennas to service users.
Quote:
If the pCell can also be used to charge/maintain the battery power in the target device -- then the target device, itself, could take the place of the "more robust transceiver" discussed above. I am experimenting with trilateration calculations on iDevices with iBeacons -- and the calculations are fast. An iPhone with an A7 and an M7 could efficient perform these calculations, as needed -- as indicated by the M7 when movement or direction changes.
Artemis claims that a native pCell radio in the device (or the LTE radio supporting pCell) uses a fraction of the power required by a WiFi -- and, it appears, that it provides faster and more precise location positioning.
I am trying to look at this pCell system from a skeptically optimistic perspective -- If doable, it has some amazing potential.
Yes, the geolocation feature is very useful and could provide more position data to the server. As is it the iPhones are most likely using existing LTE features to return signal strength data. As far as doing more processing on the device it still depends on how fast that data gets back to the central server. Again, you're limited by backhaul and its latency unless you're not moving that much.
/shrug
Everybody hopes stuff like this works. The theory behind it seems reasonably well understood, it's just the engineering challenges have been significant for deployment. It looks like they may be years ahead of other folks working in this domain in terms of deployable technology/software. If so great. If not then we'll get pretty much the same thing when 5G technology deploys.
Here's the deal though...they're dealing with incumbents without a huge interest in disruption. Something he points out himself. It requires someone like Apple or Google with billions to fund such disruption if the incumbents balk because telcom CAPEX would run even Apple's warchest dry in short order. TMobile or (brrrr...Sprint...yuck) might make a good partner and they're most likely to jump on board.
It appears that low-latency backhaul is addressed by Artemis' "Serendipitous Deployment" where the deployment of their Transceivers is largely determined by access to backhaul (line of sight Microwave or fiber cable) and cost of deployment.
It's only serendipitous if you can find it. Regions with lots of tall buildings with good LOS to each other is serendipitous. Regions where LOS is limited is not so serendipitous.
How are these areas [limited LOS] being served currently? Cell towers... Whatever? If this is the case, couldn't the pCell transceiver be mounted on the same "whatever" and broadcast stronger [partial] signals for a longer range and share the backhaul available? Wouldn't that be at least as good or better than the present service? In some rural areas Artemis is using SkyWave to bounce signals off the ionosphere to mitigate LOS problems. I read about a remote village that has a cell tower in the highest tree in the area.
This is called handwaving. The issue here is dependent on the existing backhaul topology. Daisy chain is probably worst because of the number of hops increasing your latency but that's what many 2G and 3G backhaul networks were doing. Hub and spoke would be best (and probably closest tot their microwave topology) but few deployments are this way. Mesh would be next best or perhaps Tiered.
I am familiar with most network topologies (sold/installed my first star topology LAN in 1980) or I can surf to find them, but I can't find a good explanation of "Tiered" -- got a link?
If you have a lot of tall buildings like SF or NY great. If you are in suburbia probably not so great. It depends on the latency requirements for p-cell to work and the average and worst case latency for the backhaul topology you have to work with.
Because Artemis takes advantage of interference it appears that target devices can be served by placing their transceivers anywhere -- not limited to existing towers,
If you are deploying in suburbia then you can't place your transcievers anywhere because there are few sites with good LOS with each other or the central hub. You ARE limited to the existing towers because that's where the backhaul is. If you are deploying to a region where they ripped out the older daisy chain topology and replaced it with a topology with sufficiently low latency to let pcell work then that's fine.
Some markets will be in a more favorable state for massive MIMO deployment. Others won't.
And these transceivers can broadcast at more power and for longer distances than current cell placements. Also, I suspect that a future, more robust Artemis transceiver could perform partial [distance/location/tracking ?] calculations on, say, the 50-100 nearest pCell devices; then transmit that to the servers for simpler [faster] trilateration and pCell intersection wave calculation -- reducing the latency and scaling requirements of the servers.
Transmitting a bearing and range estimate vs just signal and bearing isn't going to greatly reduce your latency requirements to the central servers doing the calculations for beam forming and traffic encoding. As only one tower you can't get more than bearing and range estimate.
The scaling requirements are not so important for the servers. The scaling requirements are important with relation to the number of antennas to service users.
I assumed that each pCell transceiver would provide the bearing and range for the n nearest target pCell devices and send these to the servers. The servers would select the subset that are nearest to the target pCell (likely 3-5), do the trilateration and wave intersect calculations and send those out through the selected subset transceivers.
If the pCell can also be used to charge/maintain the battery power in the target device -- then the target device, itself, could take the place of the "more robust transceiver" discussed above. I am experimenting with trilateration calculations on iDevices with iBeacons -- and the calculations are fast. An iPhone with an A7 and an M7 could efficient perform these calculations, as needed -- as indicated by the M7 when movement or direction changes. Artemis claims that a native pCell radio in the device (or the LTE radio supporting pCell) uses a fraction of the power required by a WiFi -- and, it appears, that it provides faster and more precise location positioning. I am trying to look at this pCell system from a skeptically optimistic perspective -- If doable, it has some amazing potential.
Yes, the geolocation feature is very useful and could provide more position data to the server. As is it the iPhones are most likely using existing LTE features to return signal strength data. As far as doing more processing on the device it still depends on how fast that data gets back to the central server. Again, you're limited by backhaul and its latency unless you're not moving that much.
/shrug
Well, now, it seems that, at least in the US, the cell carriers/providers are currently (or soon to be) required to identify (track?) your location to satisfy mobile 911 calls. I suspect that it will be easier to continuously monitor (track?) your location -- than to check for a 911 call and then , attempt to initiate detection of your location.
Everybody hopes stuff like this works. The theory behind it seems reasonably well understood, it's just the engineering challenges have been significant for deployment. It looks like they may be years ahead of other folks working in this domain in terms of deployable technology/software. If so great. If not then we'll get pretty much the same thing when 5G technology deploys.
Here's the deal though...they're dealing with incumbents without a huge interest in disruption. Something he points out himself. It requires someone like Apple or Google with billions to fund such disruption if the incumbents balk because telcom CAPEX would run even Apple's warchest dry in short order. TMobile or (brrrr...Sprint...yuck) might make a good partner and they're most likely to jump on board.
Didn't Edison require/promote direct current for his light bulbs and motors?
Edison's company had invested heavily in DC technology and was vigorously defending its DC based patents. George Westinghouse saw AC as a way to get into the business with his own patented competing system and set up the Westinghouse Electric Company to design and build it. The Westinghouse company also purchased the patents for alternating current devices from inventors in Europe and licensed patents from Nikola Tesla. In spite of a protracted anti-AC campaign waged by the Edison company, the economics of the alternating current system prevailed. Alternating current was selected in 1893 for transmission of power from Niagara Falls to Buffalo, New York - the technical and economic success of this project lead the way for the adoption of alternating current as the preferred electrical system.
Well, if this is real, it is a technology whose time has come (overdue, actually) -- and I don't think anyone will be able to stop it ... As long as governments can find a way to tax and regulate it.
Edit: This is exactly the type of discussion that can benefit from the broad range of talent and experience on AI forums. I, personally, am learning more than I ever cared to know about EMR, cell radios, network topology, backhauls, politics involved and practicality of improvement in wireless/wired data transmission.
The key to understanding the possible limitation of pcell which is most likely some form of CoMP, massive MIMO or CloudRAN is the need for very low latency backhaul and scaling.
This was the main issue with their cloud streaming tech. They needed low enough latency to allow real-time interaction with games. To get 30FPS, the latency for the entire round trip should be below 1 frame or 1/30 second = 33 millisecond latency - same reason computer displays have to be low latency too, they are normally below 10ms now (older IPS display were 20ms or higher so not suitable for 60FPS gaming that needs <16ms). Some home internet connections had enough bandwidth but if they had 50-100ms latency, the game lags behind the player controlling it.
In the case of a mobile phone pcell connection, the latency would mostly affect movement while using the device. If you walk at 4kph (400,000cm/h = 111cm/s), they need to have latency of under 9 milliseconds for the phone not to move more than 1cm before the calculation round trip finished although the 1cm size could just be the interference bubble formed and not the range of movement possible. The claim is that they already have sub-millisecond latency, so say they have 0.99ms (0.00099s), that means travelling 4kph moves 0.1cm before the calculation is done. Moving at 40kph is 1cm. This might make it less ideal for use in fast transport but they might be able to compensate for it because these devices have accelerometers and gyros so it'll know how fast it was moving in any direction.
As for the latency of the backhaul, the signals pass through buildings and they said the transmitters can work 30 miles away (possible up to 250 miles) so I don't think they need to be as densely located because of the backhaul but rather the number of users and potentially, the pcells can bounce signals to nearby pcells at the speed of light. What would be ideal is if they could put the algorithms they use on the computers into silicon and do the calculations in hardware rather than software. That way a single pcell can calculate what's needed using data from other pcells. The downside is that this makes it hard to improve the network later but if it's able to manage the performance needed, that shouldn't be much of a requirement.
The biggest use for this would be in the home. You can have a pcell located somewhere at the end of the street or on the side of your house. This would give you 100% signal everywhere in your house and garden with full bandwidth.
We'll see what issues there are when they start to deploy it (possibly Q4 2014). It'll probably have a limited roll out and it doesn't need to instantly replace existing networks - they were talking about being able to jump from pcell networks to legacy ones seamlessly, I can't recall if they said it was possible but if it has a 30 mile range, they'd just have to make sure the software didn't connect when users were near the edge.
The key to understanding the possible limitation of pcell which is most likely some form of CoMP, massive MIMO or CloudRAN is the need for very low latency backhaul and scaling.
This was the main issue with their cloud streaming tech. They needed low enough latency to allow real-time interaction with games. To get 30FPS, the latency for the entire round trip should be below 1 frame or 1/30 second = 33 millisecond latency - same reason computer displays have to be low latency too, they are normally below 10ms now (older IPS display were 20ms or higher so not suitable for 60FPS gaming that needs <16ms). Some home internet connections had enough bandwidth but if they had 50-100ms latency, the game lags behind the player controlling it.
In the case of a mobile phone pcell connection, the latency would mostly affect movement while using the device. If you walk at 4kph (400,000cm/h = 111cm/s), they need to have latency of under 9 milliseconds for the phone not to move more than 1cm before the calculation round trip finished although the 1cm size could just be the interference bubble formed and not the range of movement possible. The claim is that they already have sub-millisecond latency, so say they have 0.99ms (0.00099s), that means travelling 4kph moves 0.1cm before the calculation is done. Moving at 40kph is 1cm. This might make it less ideal for use in fast transport but they might be able to compensate for it because these devices have accelerometers and gyros so it'll know how fast it was moving in any direction.
Perlman may have addressed fast movement when he discussed the Doppler Effect. He said that pCell (or any ?) transceivers along a highway would have trouble with a car driving at highway speeds. To resolve this he said Artemis would use transceivers further down the highway ["into town"]. The effect would be less drastic the more distant the transceiver -- remembering that pCell transceivers can work at greater distance and more power because with pCell interference is expected rather than avoided.
I couldn't, quickly, locate a link -- but will post one later if I can find it ... too many damn videos :???:
As for the latency of the backhaul, the signals pass through buildings and they said the transmitters can work 30 miles away (possible up to 250 miles) so I don't think they need to be as densely located because of the backhaul but rather the number of users and potentially, the pcells can bounce signals to nearby pcells at the speed of light. What would be ideal is if they could put the algorithms they use on the computers into silicon and do the calculations in hardware rather than software. That way a single pcell can calculate what's needed using data from other pcells. The downside is that this makes it hard to improve the network later but if it's able to manage the performance needed, that shouldn't be much of a requirement.
Yes! And, eventually into the target pCell device itself. Again, the matrix math to triangulate an iPhone from 3 signals can be accomplished in a fraction of a second. Here's a test I just ran on the iPhone 5S:
====================================
Sphere 1: 0 0 0, radius 78.1
Sphere 2: 80 0 0, radius 67.09
Sphere 3: 0 100 0, radius 64.04 2014-03-03 11:01:23.243 iOS_Trilaterate_1[974:60b] ViewController.m:241 doit Before Trilateration
====================================
t1 = p3 - p1, t2 = ex (ex . (p3 - p1))........t1: 0 100 0
t1 = p3 - p1, t2 = ex (ex . (p3 - p1))........t2: 0 0 0 0
***************ey = (t1 - t2), t = |t1 - t2| ey: 0 100 0
***************ey = (t1 - t2), t = |t1 - t2| t: 100
***************ey = (t1 - t2), t = |t1 - t2| t > maxzero ey: 0 1 0
***************ey = (t1 - t2), t = |t1 - t2| t > maxzero j: 100
cross product 0 0 1
*************** z: 1.19504
*************** t2: 49.9909 59.9924 0 2014-03-03 11:01:23.244 iOS_Trilaterate_1[974:60b] ViewController.m:243 doit After Trilateration
====================================
Solution 1: 49.9909 59.9924 1.19504
Distance to sphere 1 is 78.1 (radius 78.1)
Distance to sphere 2 is 67.09 (radius 67.09)
Distance to sphere 3 is 64.04 (radius 64.04)
Solution 2: 49.9909 59.9924 -1.19504
Distance to sphere 1 is 78.1 (radius 78.1)
Distance to sphere 2 is 67.09 (radius 67.09)
Distance to sphere 3 is 64.04 (radius 64.04)
Note: this is doing 3D trilateration and could be refined and optimized significantly.
So, an iPhone 5S can trilaterate 3 signals in ~ .001 second!
IDK what is involved in calculating the signal intersects and generating the signals.
The biggest use for this would be in the home. You can have a pcell located somewhere at the end of the street or on the side of your house. This would give you 100% signal everywhere in your house and garden with full bandwidth.
We'll see what issues there are when they start to deploy it (possibly Q4 2014). It'll probably have a limited roll out and it doesn't need to instantly replace existing networks - they were talking about being able to jump from pcell networks to legacy ones seamlessly, I can't recall if they said it was possible but if it has a 30 mile range, they'd just have to make sure the software didn't connect when users were near the edge.
I agree that it will be fantastic in the home.
But, if it can maintain/charge the mobile's battery -- that alone would be a winner for most users -- all that and bandwidth too!
How are these areas [limited LOS] being served currently? Cell towers... Whatever? If this is the case, couldn't the pCell transceiver be mounted on the same "whatever" and broadcast stronger [partial] signals for a longer range and share the backhaul available? Wouldn't that be at least as good or better than the present service?
Maybe. Lets first be clear on three points:
I'm not a subject matter expert (SME) in this domain
I think there is fundamental research that shows promise so their claims (Ignoring some misdirection/marketspeak regarding shannon) are possible
I'm not poo-pooing their accomplishments just because some of these techniques may have papers or lab trials elsewhere. The neckbeard in the audience trying to poo-poo the presentation is an ass and the response entirely appropriate: if you think it's trivial show me a working example.
Okay so where I come from is a background in writing software for network gear in a past life and knowing some folks that do SDR. I haven't talked to any of them so my opinion is that of a layperson.
I looked at the literature surrounding coordinated multipoint/cooperative MIMO. My guess is their technique is along these lines given their patent filings and the musings of folks on forums more knowledgable than I.
IF so then backhaul latency is potentially critical to be able to beamform sufficiently quickly to adapt to both movement and changing environmental factors. You need updated channel state information (CSI) very quickly to coordinate multiple antennas sending signal to converge on the device and create a local maxima (aka virtual cell). Do it wrong and that 1 cm virtual cell is in the wrong place. They obviously do it very well...at least in the confines of an auditorium.
Read these and see if you come to the same conclusions I came to regarding backhaul requirements:
"In the case of deployment of remote radio units connected to a centralized baseband processing unit via Ethernet or fiber links, COMP backhaul requirements should also be no obstacle."
That's the caveat. They assume 0.1-20 ?s delay because of direct connection via fiber or 150 ?s/hop from microwave. Backhaul latency in real towers to that central server is often much higher than this. 10ms round trip for LTE but much higher for 2G and 3G towers.
Something DOCOMO attempts to mitigate using clustering:
"We investigate coordinated multipoint (CoMP) multiuser multiple-input-and-multiple-output (multiuser-MIMO) downlink transmissions over mobile access networks yielding different backhaul constraints. The larger number of base stations (BSs) participating in CoMP, the higher user throughput can be expected if there is no constraint in mobile backhaul networks. Limited capacity and latency in mobile backhaul networks impose constraints on the number of BSs that can actually participate in CoMP. We propose a CoMP system architecture with multiple clustering steps that that takes into account these backhaul constraints and enables to avoid selecting BSs which do not have enough backhaul network capability."
What the actual requirements of p-cell is unknown. Maybe this is a non-issue.
Quote:
I am familiar with most network topologies (sold/installed my first star topology LAN in 1980) or I can surf to find them, but I can't find a good explanation of "Tiered" -- got a link?
AKA tree. Hub and Spoke with branches with other hub and spokes.
if it can maintain/charge the mobile's battery -- that alone would be a winner for most users -- all that and bandwidth too!
I was also forgetting that it means the end to cellular calls because it would just use VOIP, which is a big improvement. This means they can put an end to phone numbers. You can use text strings for people instead. Skype and Facetime could be used more too, although WhatsApp probably less as there's no extra charges. Of course you're going to get people live streaming their life to some online services, that'll be the next celebrity-maker. Then people will walk up to these 'celebrities' as they are live streaming. Sooner or later it'll get to a point where they realise that going out and meeting normal people in the street is what people used to do in the first place all those years ago before technology took over.
if it can maintain/charge the mobile's battery -- that alone would be a winner for most users -- all that and bandwidth too!
I was also forgetting that it means the end to cellular calls because it would just use VOIP, which is a big improvement. This means they can put an end to phone numbers. You can use text strings for people instead. Skype and Facetime could be used more too, although WhatsApp probably less as there's no extra charges. Of course you're going to get people live streaming their life to some online services, that'll be the next celebrity-maker. Then people will walk up to these 'celebrities' as they are live streaming. Sooner or later it'll get to a point where they realise that going out and meeting normal people in the street is what people used to do in the first place all those years ago before technology took over.
Oooohhhh ... I hadn't thought of eliminating cel calls -- tho Perlman alludes to it, saying that something like an iPod (thinner than an iPhone) could have a native pCell radio and operate on the same network as LTE pCel devices.
This also would mean that mobile devices would no longer need to include/license cell radios -- so device cost could go down as well as access/usage costs!
I was born in 1939. My dad was very much into radio tech. For my 10th BDay (supposedly) I/we got a 7" Hallicrafters TV. We were the first household for blocks around that had a TV. We had a console Philco radio that was 8 x larger than the TV.
It had buttons for channels 1-13 (Channel 1?), We lived in a suburb of Minneapolis and only 2 channels (4 and 5) broadcast and then only for a few hours in the evenings...
Yeah, conversation and meeting people is a forgotten art.
How are these areas [limited LOS] being served currently? Cell towers... Whatever? If this is the case, couldn't the pCell transceiver be mounted on the same "whatever" and broadcast stronger [partial] signals for a longer range and share the backhaul available? Wouldn't that be at least as good or better than the present service?
Maybe. Lets first be clear on three points:
I'm not a subject matter expert (SME) in this domain
I think there is fundamental research that shows promise so their claims (Ignoring some misdirection/marketspeak regarding shannon) are possible
I'm not poo-pooing their accomplishments just because some of these techniques may have papers or lab trials elsewhere. The neckbeard in the audience trying to poo-poo the presentation is an ass and the response entirely appropriate: if you think it's trivial show me a working example.
I agree about the asshole -- he walked in the middle of the preso -- then immediately began haranguing the presenter (he had a compatriot, too).
Okay so where I come from is a background in writing software for network gear in a past life and knowing some folks that do SDR. I haven't talked to any of them so my opinion is that of a layperson.
I looked at the literature surrounding coordinated multipoint/cooperative MIMO. My guess is their technique is along these lines given their patent filings and the musings of folks on forums more knowledgable than I.
IF so then backhaul latency is potentially critical to be able to beamform sufficiently quickly to adapt to both movement and changing environmental factors. You need updated channel state information (CSI) very quickly to coordinate multiple antennas sending signal to converge on the device and create a local maxima (aka virtual cell). Do it wrong and that 1 cm virtual cell is in the wrong place. They obviously do it very well...at least in the confines of an auditorium.
Read these and see if you come to the same conclusions I came to regarding backhaul requirements:
Ha! The discussions about "interference" and what to do about it remind me of the CSMA/CD vs CSMA/CA (collision detection vs collision avoidance) wired LAN arguments of the 1980s. That was a tough read, but I found it was not necessarily in conflict with what we know about pCell. Perlman appears to understand the problems -- whether he has a practical solution is another matter.
"<span style="font-size:16px;line-height:1.4em;">In the case of </span>
<span style="font-size:16px;line-height:1.4em;">deployment of remote radio units connected to a </span>
<span style="font-size:16px;line-height:1.4em;">centralized baseband processing unit via Ethernet </span>
<span style="font-size:16px;line-height:1.4em;">or fiber links, COMP backhaul requirements </span>
<span style="font-size:16px;line-height:1.4em;">should also be no obstacle."</span>
That's the caveat. They assume 0.1-20 ?s <span style="font-size:16px;line-height:1.4em;">delay because of direct connection via fiber or 150</span> ?s<span style="font-size:16px;line-height:1.4em;">/hop from microwave. Backhaul latency in real towers to that central server is often much higher than this. 10ms round trip for LTE but much higher for 2G and 3G towers.</span>
Something DOCOMO attempts to mitigate using clustering:
<p style="margin-left:40px;">"<span style="font-size:16px;line-height:1.4em;">We investigate coordinated multipoint (CoMP) </span>
<span style="font-size:16px;line-height:1.4em;">multiuser multiple-input-and-multiple-output (multiuser-MIMO) </span>
<span style="font-size:16px;line-height:1.4em;">downlink transmissions over mobile access networks yielding </span>
<span style="font-size:16px;line-height:1.4em;">different backhaul constraints. The larger number of base </span>
<span style="font-size:16px;line-height:1.4em;">stations (BSs) participating in CoMP, the higher user throughput </span>
<span style="font-size:16px;line-height:1.4em;">can be expected if there is no constraint in mobile backhaul </span>
<span style="font-size:16px;line-height:1.4em;">networks. Limited capacity and latency in mobile backhaul </span>
<span style="font-size:16px;line-height:1.4em;">networks impose constraints on the number of BSs that can </span>
<span style="font-size:16px;line-height:1.4em;">actually participate in CoMP. We propose a CoMP system </span>
<span style="font-size:16px;line-height:1.4em;">architecture with multiple clustering steps that that takes into </span>
<span style="font-size:16px;line-height:1.4em;">account these backhaul constraints and enables to avoid selecting </span> <span style="font-size:16px;line-height:1.4em;">BSs which do not have enough backhaul network capability."</span>
</p>
What the actual requirements of p-cell is unknown. Maybe this is a non-issue.
I am familiar with most network topologies (sold/installed my first star topology LAN in 1980) or I can surf to find them, but I can't find a good explanation of "Tiered" -- got a link?
AKA tree. Hub and Spoke with branches with other hub and spokes.
Oh ... A tree I understand from my old IMS/DLI hierarchical database experience. This is not a very efficient way to traverse a path.
Oh, sorry. I have a IEEE account. Its sitting behind the IEEE paywall and I can't find an open copy.
No problem!
Hmm ... Your definition of a tier topology being similar to a tree topology made me regress to the old hierarchical db topology ... There was this thing called "intersection data" that was unique and only existed at the logical intersection of 2 db segments -- at the intersection of an order line item segment and a part number segment there existed an unique order qty which belonged in neither, but only existed because if the intersection.
In modern relational, tabular databases, the same thing exists in a many-to-many relationship.
I bring this up, because it is very fast and efficient to locate this data with a simple SQL query.
And, it occurs to me that this is very similar to what the pCell servers must do to locate the target pCell and generate the intersecting signals.
And, both OSX and iOS support SQLite databases ...
Oooohhhh ... I hadn't thought of eliminating cel calls -- tho Perlman alludes to it, saying that something like an iPod (thinner than an iPhone) could have a native pCell radio and operate on the same network as LTE pCel devices.
This also would mean that mobile devices would no longer need to include/license cell radios -- so device cost could go down as well as access/usage costs!
I was born in 1939. My dad was very much into radio tech. For my 10th BDay (supposedly) I/we got a 7" Hallicrafters TV. We were the first household for blocks around that had a TV. We had a console Philco radio that was 8 x larger than the TV.
It had buttons for channels 1-13 (Channel 1?), We lived in a suburb of Minneapolis and only 2 channels (4 and 5) broadcast and then only for a few hours in the evenings...
Yeah, conversation and meeting people is a forgotten art.
Technology changes so fast it's easy to forget the effect it has on kids. I saw this earlier on today:
[VIDEO]
Pretty funny to see them trying to figure out how they'd text on a rotary phone. Some of them weren't sure what payphones are, you don't think about things like payphones going away but with so many people using mobiles, they might one day disappear:
Comments
[@]Corrections[/@] maybe you can tell a friend about this topic? Should allow for a couple of "this all happened before"'s 8-)
Yes, he states scaling is linear which is good up to a point. From the patent filings it appears that for N users they need M antennas. As N increases M increases linearly. As long as the ratio is decent this scales economically.
The other potential bottleneck is low latency backhaul requirement. It may be that they need direct line of sight microwave backhaul with few hops not because of economy or speed but because the fiber backhauls to towers introduce too much latency for them to do the calculations required. This is the traditional Achilles heel of network mimo.
Probably they have solved 90% of the problems AND gotten it to work on traditional LTE phones and that's killer. But if the need for low latency backhaul and a large number of antennas it's not going to be as simple a deployment as putting one of their antennas on existing towers...even those upgraded recently to a fiber connection from older T1 copper.
It appears that low-latency backhaul is addressed by Artemis' "Serendipitous Deployment" where the deployment of their Transceivers is largely determined by access to backhaul (line of sight Microwave or fiber cable) and cost of deployment.
Because Artemis takes advantage of interference it appears that target devices can be served by placing their transceivers anywhere -- not limited to existing towers, And these transceivers can broadcast at more power and for longer distances than current cell placements. Also, I suspect that a future, more robust Artemis transceiver could perform partial [distance/location/tracking ?] calculations on, say, the 50-100 nearest pCell devices; then transmit that to the servers for simpler [faster] trilateration and pCell intersection wave calculation -- reducing the latency and scaling requirements of the servers.
If the pCell can also be used to charge/maintain the battery power in the target device -- then the target device, itself, could take the place of the "more robust transceiver" discussed above. I am experimenting with trilateration calculations on iDevices with iBeacons -- and the calculations are fast. An iPhone with an A7 and an M7 could efficient perform these calculations, as needed -- as indicated by the M7 when movement or direction changes.
Artemis claims that a native pCell radio in the device (or the LTE radio supporting pCell) uses a fraction of the power required by a WiFi -- and, it appears, that it provides faster and more precise location positioning.
I am trying to look at this pCell system from a skeptically optimistic perspective -- If doable, it has some amazing potential.
It appears that low-latency backhaul is addressed by Artemis' "Serendipitous Deployment" where the deployment of their Transceivers is largely determined by access to backhaul (line of sight Microwave or fiber cable) and cost of deployment.
It's only serendipitous if you can find it. Regions with lots of tall buildings with good LOS to each other is serendipitous. Regions where LOS is limited is not so serendipitous.
This is called handwaving. The issue here is dependent on the existing backhaul topology. Daisy chain is probably worst because of the number of hops increasing your latency but that's what many 2G and 3G backhaul networks were doing. Hub and spoke would be best (and probably closest tot their microwave topology) but few deployments are this way. Mesh would be next best or perhaps Tiered.
If you have a lot of tall buildings like SF or NY great. If you are in suburbia probably not so great. It depends on the latency requirements for p-cell to work and the average and worst case latency for the backhaul topology you have to work with.
If you are deploying in suburbia then you can't place your transcievers anywhere because there are few sites with good LOS with each other or the central hub. You ARE limited to the existing towers because that's where the backhaul is. If you are deploying to a region where they ripped out the older daisy chain topology and replaced it with a topology with sufficiently low latency to let pcell work then that's fine.
Some markets will be in a more favorable state for massive MIMO deployment. Others won't.
Transmitting a bearing and range estimate vs just signal and bearing isn't going to greatly reduce your latency requirements to the central servers doing the calculations for beam forming and traffic encoding. As only one tower you can't get more than bearing and range estimate.
The scaling requirements are not so important for the servers. The scaling requirements are important with relation to the number of antennas to service users.
Artemis claims that a native pCell radio in the device (or the LTE radio supporting pCell) uses a fraction of the power required by a WiFi -- and, it appears, that it provides faster and more precise location positioning.
I am trying to look at this pCell system from a skeptically optimistic perspective -- If doable, it has some amazing potential.
Yes, the geolocation feature is very useful and could provide more position data to the server. As is it the iPhones are most likely using existing LTE features to return signal strength data. As far as doing more processing on the device it still depends on how fast that data gets back to the central server. Again, you're limited by backhaul and its latency unless you're not moving that much.
/shrug
Everybody hopes stuff like this works. The theory behind it seems reasonably well understood, it's just the engineering challenges have been significant for deployment. It looks like they may be years ahead of other folks working in this domain in terms of deployable technology/software. If so great. If not then we'll get pretty much the same thing when 5G technology deploys.
Here's the deal though...they're dealing with incumbents without a huge interest in disruption. Something he points out himself. It requires someone like Apple or Google with billions to fund such disruption if the incumbents balk because telcom CAPEX would run even Apple's warchest dry in short order. TMobile or (brrrr...Sprint...yuck) might make a good partner and they're most likely to jump on board.
How are these areas [limited LOS] being served currently? Cell towers... Whatever? If this is the case, couldn't the pCell transceiver be mounted on the same "whatever" and broadcast stronger [partial] signals for a longer range and share the backhaul available? Wouldn't that be at least as good or better than the present service? In some rural areas Artemis is using SkyWave to bounce signals off the ionosphere to mitigate LOS problems. I read about a remote village that has a cell tower in the highest tree in the area.
http://arstechnica.com/information-technology/2014/02/cellulars-open-source-future-is-latched-to-tallest-tree-in-the-village/
I am familiar with most network topologies (sold/installed my first star topology LAN in 1980) or I can surf to find them, but I can't find a good explanation of "Tiered" -- got a link?
I assumed that each pCell transceiver would provide the bearing and range for the n nearest target pCell devices and send these to the servers. The servers would select the subset that are nearest to the target pCell (likely 3-5), do the trilateration and wave intersect calculations and send those out through the selected subset transceivers.
Well, now, it seems that, at least in the US, the cell carriers/providers are currently (or soon to be) required to identify (track?) your location to satisfy mobile 911 calls. I suspect that it will be easier to continuously monitor (track?) your location -- than to check for a 911 call and then , attempt to initiate detection of your location.
http://www.fcc.gov/guides/wireless-911-services
http://firstaid.about.com/od/callingforhelp/bb/cell911.htm
Didn't Edison require/promote direct current for his light bulbs and motors?
http://en.wikipedia.org/wiki/War_of_Currents
Well, if this is real, it is a technology whose time has come (overdue, actually) -- and I don't think anyone will be able to stop it ... As long as governments can find a way to tax and regulate it.
Edit: This is exactly the type of discussion that can benefit from the broad range of talent and experience on AI forums. I, personally, am learning more than I ever cared to know about EMR, cell radios, network topology, backhauls, politics involved and practicality of improvement in wireless/wired data transmission.
Thank you!
This was the main issue with their cloud streaming tech. They needed low enough latency to allow real-time interaction with games. To get 30FPS, the latency for the entire round trip should be below 1 frame or 1/30 second = 33 millisecond latency - same reason computer displays have to be low latency too, they are normally below 10ms now (older IPS display were 20ms or higher so not suitable for 60FPS gaming that needs <16ms). Some home internet connections had enough bandwidth but if they had 50-100ms latency, the game lags behind the player controlling it.
In the case of a mobile phone pcell connection, the latency would mostly affect movement while using the device. If you walk at 4kph (400,000cm/h = 111cm/s), they need to have latency of under 9 milliseconds for the phone not to move more than 1cm before the calculation round trip finished although the 1cm size could just be the interference bubble formed and not the range of movement possible. The claim is that they already have sub-millisecond latency, so say they have 0.99ms (0.00099s), that means travelling 4kph moves 0.1cm before the calculation is done. Moving at 40kph is 1cm. This might make it less ideal for use in fast transport but they might be able to compensate for it because these devices have accelerometers and gyros so it'll know how fast it was moving in any direction.
As for the latency of the backhaul, the signals pass through buildings and they said the transmitters can work 30 miles away (possible up to 250 miles) so I don't think they need to be as densely located because of the backhaul but rather the number of users and potentially, the pcells can bounce signals to nearby pcells at the speed of light. What would be ideal is if they could put the algorithms they use on the computers into silicon and do the calculations in hardware rather than software. That way a single pcell can calculate what's needed using data from other pcells. The downside is that this makes it hard to improve the network later but if it's able to manage the performance needed, that shouldn't be much of a requirement.
The biggest use for this would be in the home. You can have a pcell located somewhere at the end of the street or on the side of your house. This would give you 100% signal everywhere in your house and garden with full bandwidth.
We'll see what issues there are when they start to deploy it (possibly Q4 2014). It'll probably have a limited roll out and it doesn't need to instantly replace existing networks - they were talking about being able to jump from pcell networks to legacy ones seamlessly, I can't recall if they said it was possible but if it has a 30 mile range, they'd just have to make sure the software didn't connect when users were near the edge.
Perlman may have addressed fast movement when he discussed the Doppler Effect. He said that pCell (or any ?) transceivers along a highway would have trouble with a car driving at highway speeds. To resolve this he said Artemis would use transceivers further down the highway ["into town"]. The effect would be less drastic the more distant the transceiver -- remembering that pCell transceivers can work at greater distance and more power because with pCell interference is expected rather than avoided.
I couldn't, quickly, locate a link -- but will post one later if I can find it ... too many damn videos :???:
Yes! And, eventually into the target pCell device itself. Again, the matrix math to triangulate an iPhone from 3 signals can be accomplished in a fraction of a second. Here's a test I just ran on the iPhone 5S:
Note: this is doing 3D trilateration and could be refined and optimized significantly.
So, an iPhone 5S can trilaterate 3 signals in ~ .001 second!
IDK what is involved in calculating the signal intersects and generating the signals.
I agree that it will be fantastic in the home.
But, if it can maintain/charge the mobile's battery -- that alone would be a winner for most users -- all that and bandwidth too!
How are these areas [limited LOS] being served currently? Cell towers... Whatever? If this is the case, couldn't the pCell transceiver be mounted on the same "whatever" and broadcast stronger [partial] signals for a longer range and share the backhaul available? Wouldn't that be at least as good or better than the present service?
Maybe. Lets first be clear on three points:
Okay so where I come from is a background in writing software for network gear in a past life and knowing some folks that do SDR. I haven't talked to any of them so my opinion is that of a layperson.
I looked at the literature surrounding coordinated multipoint/cooperative MIMO. My guess is their technique is along these lines given their patent filings and the musings of folks on forums more knowledgable than I.
IF so then backhaul latency is potentially critical to be able to beamform sufficiently quickly to adapt to both movement and changing environmental factors. You need updated channel state information (CSI) very quickly to coordinate multiple antennas sending signal to converge on the device and create a local maxima (aka virtual cell). Do it wrong and that 1 cm virtual cell is in the wrong place. They obviously do it very well...at least in the confines of an auditorium.
Read these and see if you come to the same conclusions I came to regarding backhaul requirements:
http://www.easy-c.com/publications/Irmer_ComMag_2011.pdf
"In the case of deployment of remote radio units connected to a centralized baseband processing unit via Ethernet or fiber links, COMP backhaul requirements should also be no obstacle."
That's the caveat. They assume 0.1-20 ?s delay because of direct connection via fiber or 150 ?s/hop from microwave. Backhaul latency in real towers to that central server is often much higher than this. 10ms round trip for LTE but much higher for 2G and 3G towers.
Something DOCOMO attempts to mitigate using clustering:
"We investigate coordinated multipoint (CoMP) multiuser multiple-input-and-multiple-output (multiuser-MIMO) downlink transmissions over mobile access networks yielding different backhaul constraints. The larger number of base stations (BSs) participating in CoMP, the higher user throughput can be expected if there is no constraint in mobile backhaul networks. Limited capacity and latency in mobile backhaul networks impose constraints on the number of BSs that can actually participate in CoMP. We propose a CoMP system architecture with multiple clustering steps that that takes into account these backhaul constraints and enables to avoid selecting BSs which do not have enough backhaul network capability."
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=06139719
What the actual requirements of p-cell is unknown. Maybe this is a non-issue.
AKA tree. Hub and Spoke with branches with other hub and spokes.
I was also forgetting that it means the end to cellular calls because it would just use VOIP, which is a big improvement. This means they can put an end to phone numbers. You can use text strings for people instead. Skype and Facetime could be used more too, although WhatsApp probably less as there's no extra charges. Of course you're going to get people live streaming their life to some online services, that'll be the next celebrity-maker. Then people will walk up to these 'celebrities' as they are live streaming. Sooner or later it'll get to a point where they realise that going out and meeting normal people in the street is what people used to do in the first place all those years ago before technology took over.
Oooohhhh ... I hadn't thought of eliminating cel calls -- tho Perlman alludes to it, saying that something like an iPod (thinner than an iPhone) could have a native pCell radio and operate on the same network as LTE pCel devices.
This also would mean that mobile devices would no longer need to include/license cell radios -- so device cost could go down as well as access/usage costs!
I was born in 1939. My dad was very much into radio tech. For my 10th BDay (supposedly) I/we got a 7" Hallicrafters TV. We were the first household for blocks around that had a TV. We had a console Philco radio that was 8 x larger than the TV.
It had buttons for channels 1-13 (Channel 1?), We lived in a suburb of Minneapolis and only 2 channels (4 and 5) broadcast and then only for a few hours in the evenings...
Yeah, conversation and meeting people is a forgotten art.
I agree about the asshole -- he walked in the middle of the preso -- then immediately began haranguing the presenter (he had a compatriot, too).
Ha! The discussions about "interference" and what to do about it remind me of the CSMA/CD vs CSMA/CA (collision detection vs collision avoidance) wired LAN arguments of the 1980s. That was a tough read, but I found it was not necessarily in conflict with what we know about pCell. Perlman appears to understand the problems -- whether he has a practical solution is another matter.
This link is bad.
Oh ... A tree I understand from my old IMS/DLI hierarchical database experience. This is not a very efficient way to traverse a path.
This link is bad.
Oh, sorry. I have a IEEE account. Its sitting behind the IEEE paywall and I can't find an open copy.
No problem!
Hmm ... Your definition of a tier topology being similar to a tree topology made me regress to the old hierarchical db topology ... There was this thing called "intersection data" that was unique and only existed at the logical intersection of 2 db segments -- at the intersection of an order line item segment and a part number segment there existed an unique order qty which belonged in neither, but only existed because if the intersection.
In modern relational, tabular databases, the same thing exists in a many-to-many relationship.
I bring this up, because it is very fast and efficient to locate this data with a simple SQL query.
And, it occurs to me that this is very similar to what the pCell servers must do to locate the target pCell and generate the intersecting signals.
And, both OSX and iOS support SQLite databases ...
I need to ruminate on this for a bit!
Technology changes so fast it's easy to forget the effect it has on kids. I saw this earlier on today:
[VIDEO]
Pretty funny to see them trying to figure out how they'd text on a rotary phone. Some of them weren't sure what payphones are, you don't think about things like payphones going away but with so many people using mobiles, they might one day disappear:
http://www.usatoday.com/story/news/nation/2013/12/17/pay-phone-decline/4049599/
Telephone wires and corresponding poles can be removed eventually too.