Ex-Apple inventor Steve Perlman wants to persuade the world that his recently unveiled pCell technology really will revolutionize wireless communications and alleviate the impending data crunch for mobile operators. pCell, which stands for personal Cell, or according to some Perlman Cell, once had the working and more technically descriptive title of Distributed Input Output (DIDO) wireless. Perlman has been working on it for 10 years and the launch by his company called Artemis Networks has attracted a lot of positive reviews after a successful demonstration.
Yet closer inspection reveals that the demos have not themselves achieved anything that could not be done using existing LTE wireless or WiFi technology, while the extravagant claims have yet to be substantiated and beg a number of questions. The positive aspect is that the work does indicates that there is huge potential to boost capacity of wireless networks by perhaps several orders of magnitude within existing spectrum, but perhaps with different technology than Perlman is proposing.
pCell’s big claim is to enable massive scaling of wireless capacity through exploitation of constructive interference between radio signals from multiple antennas serving each user in a cell. Perlman admits that in practice scaling can never be infinite, but he claims that pCell will enable up to 1000 users to communicate simultaneously while enjoying the full radio spectrum of the cell without any contention between them. At present all forms of radio access communication, including WiFi and cellular 3G/4G/LTE, effectively carve the spectrum up into time slices to avoid interference between users. This means that if there are say two users transmitting data at the same time, each will have access to only 50% of the available bandwidth and hence be restricted to at most 50% of the maximum potential data rate. Similarly each of three users would get only 33% and so on.
Various techniques have been developed to extract more carrying capacity from a given amount of spectrum. Most notably beam forming has been deployed using MIMO (Multiple Input Multiple Output), which creates several streams, typically 4 or even 8, using various techniques such as line of sight propagation and spatial multiplexing to increase signal propagation for a given amount of spectrum. This can be regarded as creating multiple channels that do not interfere with each other. However, at least at the time Perlman started working on pCell, it looked like beam forming had only limited scope to increase capacity by a factor of just a few and this he quite rightly saw would be overwhelmed by the eventual increase in demand for mobile data capacity.
His big idea was to deploy totally coordinated constructive interference within radio cells by adding an additional component, a server that calculates how to create waveforms that will interfere constructively at each user’s device. It exploits the fact that if each user receives multiple waves that are exactly in phase, the effective signal power is increased. For this reason pCell does not contravene Shannon’s Law, which is a limit on the amount of information that can be transmitted over a channel with a given spectral capacity. For once it is worth looking at the formula, because this is critical for pCell. It is:
See pdf version for formula
Here C is the information carrying capacity, which pCell is designed to increase. B is the bandwidth measured in Herz, which is the immovable object set by the amount of spectrum available. S is the signal strength, while N is the noise caused in the case of radio by background interference and other ambient effects. The key point is that if you increase S, the signal strength, you can carry more information, which can also be done by decreasing N, the noise.
It is worth noting in passing that with vectoring, deployed on DSL lines to increase capacity by mitigating cross talk interference, N is reduced while leaving S the same. In the case of pCell, the effect of the constructive interference is to increase S, the signal strength, while leaving N alone.
pCell and vectoring share one important property. They both rely on intensive processing to coordinate signal output across all channels in a given vicinity, within a radio cell for the former and a bundle of twisted pair copper wires for the latter. The maths involved is totally different in detail but somewhat similar in broad concept. Both are computationally intensive, but this is all right because the required processing power is now affordable.
pCell does though have one additional complexity that may cause it some problems when scaling up to real commercial deployments. That is the fact client devices are mobile rather than still, which could cause loss of performance in the event users are moving at any speed, as in cars or trains. The system relies on calculating precise waveforms for each user by taking account of their location. So it would have to be constantly and rapidly re-calculated for moving phones.
pCell relies on use of multiple antennas to create signal diversity, with each antenna being independent of all the others, having a unique signal path. This uniqueness allows the pCell data center to distinguish between each device and then determine the precise waveform for each user. Each device then receives a unique waveform that is the sum total of multiple individual waveforms transmitted by the local antenna network. Because each user is at a different position relative to each antenna, the waveform is different. Each device is effectively surrounded by a sphere of coherence, within which it will be able to decode its own data. These spheres therefore must be sufficiently small that they do not overlap between neighboring devices.
The key task performed by the pCell data center is then to ensure that all of the waveforms interfere constructively at every client, which is mathematically possible, but computationally hard. The other critical mathematical component is an algorithm that allows each client to demodulate its own data from the composite waveform.
The technology is impressive, but we think it is disingenuous to suggest that the underlying mathematics is incredibly innovative. It is more of a signal processing and computational achievement than a mathematical one. But back to the practical elements.
Firstly the recent demo involving only a small number of devices did not break any radical new ground, merely showing that it was possible for a small number of clients to collectively enjoy close to the advertised peak throughput of a cellular system. This is a reasonable achievement but could be done with the help of existing MIMO technology.
This leaves the promise of much greater aggregate throughput to come, and that’s where the questions begin. Above all it remains to be seen whether pCell will scale and work in real commercial networks. Some suspicions must be raised by the fact Perlman still appears to be looking for backers, making us wonder why Qualcomm, Broadcom and others are not jumping at the technology if it is so revolutionary. Maybe they will, but they haven’t yet.
There are other promises, such as robustness and avoidance of no go signal areas, both of which look like they could stack up. Equally there seems no reason why the technology would not work with existing wireless standards, such as LTE and WiFi, again as Perlman has promised.
Less convincing though is the claim that the technology will be cheap to deploy, when it looks like the entire infrastructure, including data centers, base stations and wireless routers would have to be updated, even if existing devices would work with the system. This would be a major task, even assuming that the technology is adopted, and it would take a year or two at least for pCell to be incorporated into 802.11 or LTE, assuming that this is readily achievable.
The other major caveat is that pCell is far from the only game in town when it comes to potential massive improvements in wireless spectral efficiency. There has been a lot of progress with dense MIMO configurations, while techniques such as WiFi Direct promise to exploit peer to peer communications, in effect to create smaller dense cells of use, both for low bit rate communications among devices within IoT (Internet of Things) applications and also for transmitting multimedia content, such as video, locally without consuming spectrum across the whole cell.
More radically there is the promise of exploiting the laws of physics more effectively, through for example the orbital angular momentum (OAM) of electromagnetic waves, to boost capacity enormously. All electromagnetic waves, including light and radio waves, has momentum both in the direction of movement (linear) and also rotationally around its position. OAM is a component of this angular momentum that is determined by the spatial distribution of the waves and is therefore in a sense related to MIMO. OAM has been proposed as a complement or alternative to MIMO for effectively creating more channels within given spectrum and tests have shown it has the potential for a huge boost in capacity. OAM can be applied to both wired and wireless communications and again does not break Shannon’s law because it creates additional channels within a given portion of spectrum, rather than encoding more data in a single channel.
pCell then faces considerable competition, as well as a big task convincing the key industry players that it does represent the way forward for mobile data communications. But Perlman has made millions of dollars, if not billions, out of selling his ideas on, so perhaps this will have a taker, but it is hard to imagine what that would look like ‘ perhaps a large real world demonstration, in an entirely new network? We shall see.