Folks, Forgive me if I have not gone through the video. This sounds like distributed digital beam forming.
Beam forming creates a steerable beam. This is essentially using two transmitters offset from each other to create interference patterns that positively coincide at a desired location. There is a lot of processing involved in this which makes the hardware expensive and complicated, pretty much implemented in bespoke silicon. In software you have less worries, processing is cheap, so why do it in silicon. This is called Software Defined Radio, although SDR which was originally conceived for handset and not base station optimisations. We have already seen cloud RAN technology where the radio node is stripped down and the RF processing is aggregated at a metro site. Trade off here is that you need dedicated fibre between the RF head and the RF processing, but in Metro areas this is possible. Mix beam forming, SDR and cloud RAN together and you can have large arrays of digital beam forming nodes, with central processing. The processing must be central to allow for coordination of multiple beams. As ever in RF there is no free lunch, you pay in back haul and processing. But you are up against the laws of physics, and something has to give in order for you to scale. Technology is already giving way to cheap fibre and cloud computing so at least this type of RF design is riding the right trends. The cheap fibre is in urban areas and we are trying to solve an urban customer density problem. If you now scale this up. You only need as much processing as you have active customers, these customers move, and in crowds, so why build the processing at the edge of the network in the base station where some locations will be massively overloaded and some empty. Better deploy cheap simple base stations and process near the edge. To quote Douglas Adams, this technology is fiendishly difficult, but there is a real problem to solve so the money is sure to follow. John Bourke on Mobile On 10 Dec 2014, at 02:53, Gord Slater <[email protected]<mailto:[email protected]>> wrote: On 9 December 2014 at 23:41, Ian Tomkins <[email protected]<mailto:[email protected]>> wrote: Whilst I only know what I have learned from watching a couple of the vids on the Artemis web site, I think we can say for sure that they aren’t proposing lots of smaller and smaller cells oriented around more and more transmitters, there also isn’t any suggestion of P2P. They seem to me to be requiring many base stations, they are attempting to virtualise the concept of cells, but their technique *requires* more than one base unit per handset, as a minimum, to work. 3 in the case of 8 handsets, 8 bases in the demio of 16 handsets (I found another vid on youtube https://www.youtube.com/watch?v=MZnKYHC6rn8 ) I'm not sure how many handsets any given number of bases can support, but I took it that they are looking to saturate an area with their base units to provide geographical spectrum re-use as well as active techniques to help. Yeah I was wrong about the P2P, sorry. I was assuming that he was trying to demonstrate near-field coupling of the device antennae "within millimetres of each other" (he laid them up against each other in a toppled stack-like arrangement - just on from here in the vid I watched first. https://www.youtube.com/watch?v=5bO0tjAdOIw#t=1699 Since P2P via nearfield was the only way to allow device-to-device comms to be useful at anything much from a delivery standpoint, it was all I could think of :) What this really needed was a simple statement or paper that the minimum rcvr separation is Xmm apart, because the sum of waves can only be calculated and synthesized to a resolution accuracy of Xmm for X devices that are stationary for a given period of X(milli?)seconds. I'd like to see how that scales in a street full of people moving at various velocities and vectors with other reflective objects like buses, trucks, cars, moving at typical velocities, in addition to the normal intense multipath environment of a city street (which naturally and in a very complex way creates the intense additive and subtractive phase difference environment that they are utilising, plaguing urban communications, acoustic as well as RF) It seems to me to take a lot of back-end grunt to work currently: 2 servers running to provide 8 devices, so the claim about "1mW of power" can only relate to the TX output power, which is far less impressive from an RF perspective. It was this specific statement I happened upon that led me to question the RF engineering experience of the presenter. I've also watched a second video where he seems to be confusing spectral bandwidth with throughput - https://www.youtube.com/watch?v=MZnKYHC6rn8 ; the claim also means nothing without accurately defining the throughput requirement of the video streams first. It could be extremely technically impressive, or maybe mostly hype. I do hope it isn't hype. If you're happy quantifying throughput and congestion performance by olympic-size swimming pools or London buses, feel free to celebrate. He may be just be dumbing down to a target audience. I'd rather he just got on with it so I can tell if it's bull or not. Until I see real figures and real tests, I'm not wasting my time on it. I'm not his target customer anyways and I doubt I'll be benefiting from it even if it is truly revolutionary for a while yet. My APs in this room tonight are around -10dB into a loss of about a further 6dB against an isotropic radiator, a tenth of a watt RF each at the final amplifier stage, attenuated by a further 6dB of intentional antenna system loss. My EIRP about 0.025Watts of RF for each AP. (This is for a very good reason: to enable channel re-use within the property to maximise throughput and therefore maximising spectrum utilisation all at the same time for different equipment in 3 dimensional space. I do not compute the service areas dynamically, I simply have different equipment running in the next room on the same channel at the same time. Where possible, I modify the mobile end of the link to reduce EIRP of the mobile transmitter commensurately, but this is not always needed in practice. I could also use directional antennae (by altering the phase difference of signals between elements of the antennae) to maximise SNRs and further reduce co-channel interference. This could also be done dynamically in real-time, using conventional methods, in practice, I simply rely on slight directionality and attenuation through the (thick) walls between the rooms. "Beamforming" with one AP is an example of a extant phasing method to enhance signal:noise ratios to/from for mobile devices. With two or more APs beamforming in a controlled manner, then zones of good SNR (due to additive In essence their technology seems to revolve around using software defined radios to dynamically create radio interference patterns by sending low power signals from multiple aerials. The interference pattern can be targeted so that there is a small area of positive interference that exactly coincides with where the receiving radio is. Even a single radio transmitter already creates natural interference patterns when the signal is reflected of objects. The larger the object compared to the wavelength of the signal and the more conductuve the object is, the more the reflections for interference patterns. Normally they are randomised and a problem ("multipath" is one of the phrases used to describe the problem that randomised interference patterns create) In addition, two or more radio transmitters, or a single transmitter feeding several antennae, either by the most direct route or by adding delays or extra distance in the transmission line or signal path between the transmitter and receiver also cause interference patterns. They are utilised where needed and we try to avoid them where they would be a problem. We can even dynamically generate and manipulate these patterns in the time domain to provide performance gains or specific effects. That is far from a new concept. Look at a PAVE PAWS radar site for an easily-verifiable 1980's example. Earlier techniques for other purposes go back decades before that. too, not just for radar. The human head processes acoustic signal in stereo to provide direction finding using the same techniques in receive mode. The self same dynamic method using interference patterns. A domestic hifi system uses the same interference patterns ot provide a stereo image, the perceived positioning of the sound depending on the SNR at the receiver, which can become very complex and degraded with unwanted interactions and reflections off nearby objects, especially ones that resonate at particular frequencies, though the medium is somewhat different, the same interference patterns occur for better or for worse, intentional or unintentional. SDR is not a prerequisite for this end result at radio frequencies, active phased array antennae have been used for a long time to achieve similar results for radar heads using phase differences. this follows on from initial research on interferometry and direction finding using phases sets of antennae in the first decade of the last century, building on Youngs work a hundred years earlier. indeed, I have used an entirely passive phase system to improve the SNR of a radio signal myself, by delaying the signal to produce the wanted responce without moving the antennae or the receiver position. I could easily dynamically change the phase shift to adjust the direction of best reception, and with a second receiver and phase adjustment network placed some distance away from mine (on a long baseline for maximum positional resolution) I could adjust the spatial position, now in distance terms as well as direction, of best reception of a particular transmitter site, to the detriment of other unwanted transmitters, without moving my antennae or my receiver too. The improvement in SNR could be as high as 60dB by nulling the unwanted signals. by peaking the wanted signal, the results are less spectacular, but approached 20dB improvement SNR at least. This was using 1940s technology and methods. The terminated beverage antenna, very popular in the early 1900s, could be null-steared or peaked by several tens of degrees too, using the same principle, without moving the antenna orientation or position itself. SDR at the base radio end simply makes some of this easier, (as well as introducing a few problems of it's own) but SDR has been revolutionising radiocomms since the early 90's and continues to do so for a variety of reasons (initial time delay subtraction and addition mathematically as a proof of concept was done in the early 80s). £6 to £8 on Ebay or Amazon can buy a very usable SDR receiver that can do amazing things when used correctly, intended to be used as a USB TV receiver dongle. SDR is not a panacea nor is it a requirement for interference phasing techniques by any means. The same technique can be used with direct conversion receivers, superhets, phasing-method demodulators (hint: related topic, phasing) as well as more esoteric receivers like superegens. For the transmitter side, the world is your oyster as long as it oscillates and produces the right power. It can also be used, of course, with phase modulation equipment (again, related) because it matters not what the signal is nor what it is used for in the slightest. It is simply a natural phenomena that can be exploited, having been exploited for almost 100 years, often on a automatic "dynamic" basis in the last 2 or 3 decades for a number of applications, mainly radio location (in many different forms) such as navigation, radar and Radio Direction Finding. This targeting is automatically tuned based on picking up the transmitted signal from the remote device – iPhones in their example videos. This whole system requires a bunch of real time processing on the back end using stacks of Linux servers, apparently this scales linearly so they can increase capacity by simply adding more tin. By extrapolation, they assume it scales linearly, as far as they have tested. 2000 people in one area are not beyond the bounds of imagination, see the original post in this thread for an example. Under current iterations of backend processing, that's 125 servers per thousand users. There will be eventual economies of scale and probably FPGA or similar optimisation I'm sure, but for now, these can only be extrapolations until real-world large-scale non-open-area-test-site trials are complete, especially in the urban or city centre environment, where the problem are likely to be the greatest. What seems particularly interesting is that they say they can independently transmit to multiple devices at full bandwidth over the same piece of spectrum simultaneously using this effect. In addition because individual devices don't share spectrum with each other they can all talk different protocols without causing each other issues. I'd say they absolutely have to share the same spectrum, they are claimed to be unmodified handsets, there are merely not in the same spatial positions. They may not share common service areas at any one given moment in time, but then again neither do current cell-sites. The same spectrum is reused over and over whenever co-channel interference constraints permit, due to path loss and taking advantage of terrain screening and antenna directionality. The devices, unless they plan on using other frequency allocations (not too difficult with SDR on the handsets but optimal antenna engineering for wideband) simply have to work on the same available spectrum. The system seems to rely on synthesis of wide aperture effects (not new by a long way) and adjusting them dynamically. Potential problems are that while positive effects occur at at least one position (and possibly one position only, if the system is highly optimised to get the phasing right) that inherently comes at the expense of subtractive effects at another location or locations, for any given spatial position of any one given handset at any one given moment in the time domain. A problem is that there may be, by quirk of mathematics or physical reflections, more than one spatial position at any given moment in time where a additive node is located in 3 dimensional space. That means that there will definitely be a practical limit, based in part on some reflections form objects within the coverage area as well as mathematical collisions, on how many devices can be located at nodes at any given time, while the rest of the handsets are also dynamically placed at the subtractive antinodes, to prevent a reduced SNR. There isn’t much talk about how much bandwidth the devices get for talking back and how that works, so that may be a little more of a traditional problem but there is a comment that the individual devices need less power to send and consequently get better battery life, which suggests that they are also able to dynamically focus the receiving antennas as part of the process. if the phasing technique used is consistent and applied to the antenna system itself, reciprocity will be met (meaning, yes, it works for transmit and receive in that case). If the timing of the difference of arrival is done too slowly ("TDOA" DF might be worth a google to see why this is relevant, and the methods used to achieve the additive and subtractive effect) and is done at the signal generation stage itself, whether done by SDR, pre-SDR or hybrid methods, then a separate system must be used for the handset>cellsite direction. The same effect can be achieved my passive, non-electronic methods, sucjh as exploiting cardioid or figure-of-8 responses of DF antennae such as the Bellini-Tosi loop systems before WW1. They work by phase in the same way, phase cancellation and addition, though the computation is done by human operators and is somewhat slower. Computational electronics (even using thermionic devices) have speeded up the computational methods dramatically in the last 70 years, with automatic systems developed in the 1940s using analog thermionic computers. Prior to that, mechanical computers and pre-computed lookup tables were used to solve the geometry problem in the 1930s for the same purpose. It's stated in the video I watched that no processing is done on the mobile devices themselves, and that they are standard phones, not modified or experimental hardware, so the handsets do not appear to be part of the "intelligence" of the system. Basically, this means that "yes, they do the base radios receive processing in the base units or backend servers", either by antenna phasing or dynamic mathematical methods in an SDR RX stage. Existing mobile handsets already use dynamic transmit power adjustment both to improve battery life and also, as a side effect, promote spectrum re-use in the spatial domain. So not breaking the laws of physics but certainly working around the limitations we are used to dealing with in innovative ways, and if all of the above is true then I think they are right when they say it is a revolutionary use of radio. Ignoring Young's 1800's work, phased antenna systems dating back 100 years, many TDOA techniques including WW1 stereo sound detection of aircraft, SAR, interferometry (both optical and radio frequency) and similar non-DSP/non-SDR systems that use the same techniques, yes. That's *if* scales linearly as they claim, in real-world use. In the same way that Marconi's radio equipment is seen as revolutionary by many people today, but the story is a lot deeper than that and has some hidden surprises when you look deeper. It's similar to the way that people are taught "Edison invented the Lightbulb". I live not too far from where Swan lived and worked and even nearer to where one of the most important milestones of domestic electricity use occurred. Revolutions are never simple - the victor rewrites the history books, not history itself. In the "lab demo 4"video I linked to above, there's more cell bases for more phones, but not a linear scaling. 8 base radios for 16 handsets, instead of 3 base radios for 8 handsets. The comparison is meaningless though without the test conditions and data to compare the results. Or maybe my linear scaling ruler is just bent, iDunno. I admit: may be a world first using iPhones though. That's if you ignore the DSP phasing enhancement of the microphones used to optimise the SNR of the user's voice when used as a phone or to record a memo or message; and the DSP that is used for the loudspeakers to optimise SNR at the listener's ears for receiving the call or playing music. The radio chipsets also dynamically adjust their output powers based on RSSI and continuously monitor it and are prepared to change it dynamically. GSM bases have done this since inception of the standard simply by ramping the AGC line and observing the results. # So yeah. ~Maybe~ revolutionary, time will tell. This revolution will not be televised due to white space being sold off to use for LTE :) Ultimately how receive the news all depends on where you stand and your viewpoint. I must get off to bed, my Aunty Node is visiting tomorrow. My Uncle Node may coincide at the same position too, but at a different time. I won't allow myself be phased by their arrival times, because life takes us all in different directions. He's a doppleganger for Maxwell and has been equated with him. When he moves you can hear him whistle - listening to his pitch - it's very moving. We're all just ripples in a pond you see, as if someone has dropped two Pebbles in a pond, close to the lillypads of connectivity. I merely have the front to question the tech. I'm just trying to give a summary, I'm not trying to be deliberately negative about it. There are up and downs to all dynamic arguments. QED QDM. Anyways, I must sine off for the night o/ waves -- sent via Gmail web interface, so please excuse my gross neglect of Netiquette This e-mail and any attachments are confidential and intended solely for the use of the recipient(s) to whom they are addressed. If you have received it in error, please destroy all copies and inform the sender. This email and any attachments are believed to be free from viruses but the Satellite Applications Catapult accepts no liability in connection therewith. Any views or opinions presented in this email are solely those of the author and do not necessarily represent those of Satellite Applications Catapult.
