In a previous PolicyTracker blog post (DIDO, the Shannon Law, and an antenna for every citizen), I discussed the idea that using a vast number of distributed antennas (eg one for every citizen) could enable the spectrum to be used many times over, effectively permitting each user access to all the spectrum, all the time.

Whilst the cost and implementation of such a scheme would be as vast as the number of antennas involved, a more practical approach to re-using the same spectrum for multiple users has recently been proposed by the Alcatel-Lucent chair on Flexible radio, at Supelec, France (Vandermonde frequency division multiplexing for cognitive radio).  The idea is named after French mathematician Alexandre-Théophile Vandermonde, not because of a re-discovery of some long lost mathematical process, but for his work on linear algebra, and in particular the Vandermonde matrix.  More of this later.

The concept is, that for any given user, there is a defined set of frequency-domain paths from every transmitter that is close enough to provide a signal.  Those paths can be modelled as a set of frequency dependent path profile co-efficients (lost you yet?)  If you know the path profile from a wanted transmitter to the user it is serving, and you know the path profile from any unwanted transmitter to that same user, it is possible to calculate a set of frequencies that can be transmitted from the unwanted transmitter that will not cause interference to that from the wanted transmitter.

Let's take an example, say the path from a wanted transmitter to a user had a frequency profile that looked like the blue line in the graph below, and the path from  an interfering transmitter to that same user had a profile that looked like the orange line. For a received signal to be strong enough to be used by the user, it must be of a strength over the dotted green line.

There are two areas of spectrum (those highlighted in orange) where the wanted signal is too weak (ie below the green line) to be able to provide the user with a service, and hence this spectrum could be used by the 'orange' transmitter with impunity, knowing that it would not cause additional interference to the user's service, regardless of how strong its signal was.

There is a further area, shaded blue, in which the signals from the orange transmitter are sufficiently below the wanted signal from the 'blue' transmitter that they would also not cause interference.

Thus, some of the spectrum could be re-used to provide a service from the orange transmitter to a different user, without causing interference to the service from the blue transmitter. The path profiles would be different for each user, and would have to be calculated frequently, especially if users are highly mobile.  If the users were fixed (such as is the case for television receivers for example), the situation becomes much more straightforward.

The part that Vandermonde's matrices play, is providing a mathematical mechanism for taking the different path profiles and working out which frequencies can be re-used by a potentially interfering transmitter.  The resulting concept has been termed 'Vandermonde Frequency Division Multiplexing' (VFDM) and is being proposed as a means of re-using spectrum on a cognitive basis.  As with the distributed antenna concept, it requires a lot of maths in both receivers and transmitters, but maths is, after all, what microprocessors are best at.

Whether VFDM, or for that matter distributed antennas, become reality is, perhaps, not the real story.  The fact that people are looking at extreme methods of frequency re-use in order to overcome the squeeze on spectrum availability perhaps reflects the headaches caused by their experiences of regulatory lethargy.  The cost of implementing such complex techniques clearly has to be counterbalanced with the stiction of some spectrum users to find means to become more spectrally efficient themselves.  Surely it is more cost effective to find a means to reduce the spectrum used for, for example, radars, than to introduce VFDM and the like into wireless networks.  Or maybe Vandermonde provides a means to use the spectrum allocated to more unwieldy users in a mathematically safe way, providing aspirin both for wireless networks clamoring to provide more capacity, and for those users sitting on spectrum which they are unwilling or unable to share without being guaranteed protection from interference.

P.S. Perhaps the time is right for an English mathematician to have their theories transposed into spectrum engineering for a change. I propose more research into the use of the Womersley number, named after my namesake John Womersley, to solve some currently undiscovered wireless viscosity problem!

The European Commission has recently agreed he Radio Spectrum Policy Programme (RSPP https://www.policytracker.com/headlines/eu-telecoms-ministers-approve-rspp) which will run until 2015, though in principle it will continue well beyond this. One of the key objectives of the RSPP is to: "make at least 1200 MHz of spectrum available for wireless broadband services in the Union by 2015, following an assessment based on the new spectrum inventory".

The 1200 MHz of spectrum is not 1200 MHz of new spectrum, but includes that which is already available for mobile services.  This includes:

This little lot already comprises 625 MHz or over half of that being sought.  If we add in bands which are already being touted for future mobile connectivity (eg by the ITU), or which are already available in some EU Member States but not necessarily all, we can add the following to our inventory:

This adds another 350 MHz making our total so far 975 MHz. In order to achieve the RSPP objectives, another 225 MHz of spectrum is therefore needed. Simply  harmonising those bands listed in the table above across Europe would bring the RSPG pretty close to their 1200 MHz target - not that agreeing such harmonisation would be simple, but there are precedents in some EU Member States which could be expanded more widely.

If we broaden slightly our definition of 'wireless broadband services' to include WiFi, where public, commercial networks already operate albeit not with the same degree of protection as those in fully licensed spectrum, we should add the following to our list:

Achieving wider use of just one of the 5 GHz bands alone would get things looking up. Whilst the target of finding 1200 MHz of spectrum may seem ambitious, there are plenty of bands which, with some effort (and the amount of effort should not be underestimated) could meet the target. 

The spectrum inventory currently being undertaken by the RSPG will certainly assist in understanding the extent to which these bands might be able to be used more widely across Member States, and may even throw up other opportunities not listed above.  If more than 1200 MHz could be found, it might provide flexibility for those countries where some of the bands cannot be made available because of, for example, significant use by spectrum incumbents such as public sector users.

The bigger question is possibly, "Is 1200 MHz enough?"  Certainly in a 2015 timescale, it would seem to be, especially given the time-to-market between any spectrum being released and networks being rolled-out.  Perhaps one of the goals of the RSPG should be to determine exactly how much spectrum is likely to be required for 'wireless broadband services' into the future, which, tied with the results of the inventory might serve to give an indication of how much, and from where, any future spectrum requirements above the initial 1200 MHz target might come.

A recent white paper by Steve Perlman and Antonio Forenza of Rearden (http://www.rearden.com/DIDO/DIDO_White_Paper_110727.pdf) discusses a technique they call 'Distributed Input, Distributed Output' or DIDO. In essence, DIDO is a form of MIMO, but which differs from MIMO in two specific ways:

• Firstly, the 'antennas' on the network side are intended to be placed anywhere (and everywhere), so could, for example, be replacements for your home WiFi hubs as easily as they could be positioned on towers.
• Secondly, the signal processing necessary to determine the waveform transmitted from each antenna is done centrally, rather than on a base-station by base-station basis.

The logic seems to be that, by using a central processing unit to do the complex MIMO calculations, the cost of the base stations can be reduced (because they become 'dumb transceivers') and further that the user equipment does not need any inherent signal processing at all, as it is all done in the network.  Though the white paper is a little nebulous in some areas (there is no talk of how the connection from users to the network is managed, nor of how mobility is dealt with), the principles seem reasonable.

It is claimed that DIDO provides connectivity that exceeds the limits set by the Shannon Law (the law which determines the theoretical maximum amount of data that can be transmitted over any piece of radio spectrum), because all the spectrum can be re-used, all the time, for every user.  But measured from the perspective of any user in the network, this is not the case.  Each user's connection is still bound by the restrictions enumerated in the law. 

Overlapping Fixed Links

As an analogy, the same could be said for two fixed links whose paths cross.  Each may achieve a service which approaches the theoretical maximum connections speed achievable, but at the point the paths overlap, the same spectrum is being used twice.  So at that point, surely the spectrum is carrying twice as much data as the theory states.  This is not the case, as it is the channel capacity which is bound by the law, not the capacity of the spectrum itself.

So does DIDO really do anything new?  Well the idea of distributed antennas for a MIMO network is interesting, however the paper alludes to the fact that for each new user on the network, a new antenna is required.  Whether the computing power required to calculate the waveforms for the sheer number of sites that would be required is feasible is a bit of an unknown.  But the more interesting point hidden in the text is the need for an antenna for every user.  It 'feels' reasonable that if every user on a network had an antenna dedicated to them, then the service that they would receive would be excellent. 

But what of the cost?  Is it at all economically feasible to have, say, 60 million 'antennas' in the UK, noting that each one is also a transceiver and needs a broadband internet connection.  On the face of it, if this were a regular terrestrial mobile network, then no.  But if every cordless phone, WiFi hub and other connected device were to form a DIDO network, then possibly.  The point here though is not that DIDO is the solution but that having so many points of connection could offer a means of providing ubiquitous, high quality, wireless broadband.  It's already clear that to satiate the demand for data, the only realistic solution available to mobile operators is to increase the number of sites they have.  Having so many would clearly solve that particular problem.  If you count WiFi hubs and hotspots as cell sites, one wonders just how many points of connection there already are.

What the white paper raises, is the spectre that, if demand for data grows in the way that many are predicting, the number of wireless points of connection that will be needed may be way beyond the most frightening nightmares of current network planners!

There are currently a number of ongoing national consultations in Europe over the future use of the 3.4 to 3.8 GHz band (as the PolicyTracker web-site will attest).  Amongst the various noises being made by interested parties, the UMTS Forum are making a strong play for the allocation of spectrum from 3.4 to 3.8 GHz to wireless broadband services.  They claim that there is 'no significant use of the band by satellite'.  This is a rather misleading statement. 

The 3.4 to 3.8 GHz band is known as the extended C-band by satellite operators and there are many satellites who use these frequencies to deliver broadcast television and other services.  It is true that there are no satellite services in the band whose primary target area is Europe (with the exception of Greenland), however that does not mean that the band is not used.  There are a wide range of services which are American, Arabic or African in origin for which use in Europe will fall into two main categories:

(1) Reception by professional users, either for re-distribution on cable networks, or re-uplink onto other (Ku-band) satellites for direct to home reception as part of various mulit-channel television packages.
(2) Direct to home/office reception by diaspora of the countries concerned, keen to keep up with news from home.

Of course there will be enthusiasts and others who make a hobby out of receiving obscure satellite services, but these numbers will be very small and (whilst recognising the can of worms that this opens up) arguably of low economic value.  In addition there are two-way links between Europe and these other regions for business, telecommunications and other purposes, whose economic value may be more significant.

Whilst it is truer to say that the sub-band 3.4 to 3.6 GHz is very lightly used (and in many countries the decision to hand this over to terrestrial wireless networks has already been taken), it isn't strictly true to blithely state that there is 'no significant use' unless the amount of use has been properly quantified.  For professional use, this ought to be relatively straightforward.  Cable networks and other multi-channel television head-ends are usually either registered or are well known to authorities.  Establishing direct to home use may be more complex.  It is highly likely, for example, that the embassy or high commission of a country will have a dish to receive their domestic television programming.

C-band dishes are relatively large and in many countries planning permission may need to be granted for their installation offering a means of identifying how many there are (though I note that one of my neighbours has a 1.8 metre dish installed on the side of their house for reception of South American programmes and has never received a knock from the authorities, despite it having no permission).  It therefore ought to be possible to get a feel for how many large dishes are installed and by applying a few simple metrics, be used as a proxy for the number of C-band receivers in use.  Unless figures for such use can be reliably established, filling the band with terrestrial wireless broadband services and consequentially eliminating the possibility of receiving C-band satellite services could not possibly be sanctioned by any sane regulatory authority.

But all is not necessarily lost for the terrestrial wireless broadband community.  There are solutions to both professional and domestic use.  In the case of professional use, the most straightforward means of addressing the problem (and one even suggested by the UMTS forum) is to put an exclusion zone around satellite earth stations.  In the UK, for example, there are probably a dozen or so large earth stations (such as the ones at Goonhilly, London Docklands, Crawley Court and Chilworth).  With the exception of the one in Docklands, most of these are in relatively sparsely populated areas and the loss of ability to offer services in these areas may not be particularly significant.  Much depends upon the size of the exclusion zone that would have to be established around the apposite dishes.

For domestic or semi-professional users, the situation is not as simple to address.  Putting exclusion zones around smaller dishes is doubly difficult.  Firstly in establishing where they are, and secondly because the dishes themselves being smaller will have a wider field of vision and will receive a weaker signal and thus will be more susceptible to interference.  But if the UMTS forum is willing to part with cash to solve the problem, there is a means at hand.  Simply downlink all the C-band services visible over Europe (at one of the aforementioned earth stations for example) and re-uplink these onto a Ku-band service with the necessary pan-European coverage.  Thus domestic users could install a Ku-band dish instead, leaving the C-band frequencies free to be used for wireless broadband.  Managing the rights and access to programmes might be complicated but not beyond the wit of man.

C-Band coverage map of satellite Yamal 202

So with a few concessions and a bit of expenditure, the extended C-band might be opened up for wireless broadband services across Europe.  But herein lies a small irony.  Looking through the list of services available over Europe on such satellites shows that there are a number of transponders dedicated to the provision of satellite Internet access (eg on Yamal 202).  C-band is well suited to use for V-SAT terminals in remote locations (and those subject to heavy rainfall) and though the swathe of new superfast Ka-band satellites being launched could provide alternative means of connection, the wider footprint and unique characteristics of C-band satellites might mean they are the most economically efficient means of providing services in the small pockets of Europe that other satellites don't reach (Greenland for example!)  Moreover, satellite operators have not yet begun to think about how else the C-band might be used.  It is feasible that it might provide the spectrum needed to introduce more mobile satellite services (with the necessary tweaks to allocation policy at the ITU), or that it could offer feasible mobile broadcasting to reasonable sized handsets to complement terrestrial broadband coverage.

What's left of Coldingham Abbey after the Vikings visited it

As the nuns at the monastery in Coldingham learnt when the marauding Vikings decided to rampage through their convent (and habits), despite them having cut off their noses to make themselves less attractive, 'cutting of your nose to spite your face' is a simple, cautionary tale whose moral may be applicable here too.  Handing C-band over to terrestrial wireless broadband services at the expense of what could prove to be innovative and socially beneficial satellite services may prove to be an equally bad judgement call.  A sensible, level-headed discussion between terrestrial and satellite operators needs to take place and who knows, maybe they might even find some common ground?

At the 6th European Spectrum Management conference in Brussels this week, a number of speakers focussed on the difficulties which are caused by the asymmetry of mobile data, as well as on possible solutions for it.  The problem is this: the vast majority of use of the Internet is asymmetric, and in particular the amount of data which users download is much greater than the amount of data which users upload.  This has been recognised at a European level. Early drafts of Europe's Digital Agenda cited the need for 30 Mbps, symmetric connections. But the published version now only talks about 30 Mbps download speeds and reference to symmetry is gone.  The asymmetry of mobile data is acknowledged within the majority of broadband radio standards which almost universally offer faster downlinks to uplinks (LTE for example, has a maximum downlink connection speed of around 100 Mbps but a maximum uplink of speed of only 50 Mbps).  Of course, part of the reason for this is the difference in quality of connection supported by the weaker signals transmitted from handsets compared to base stations, but you can be sure that those who develop the standards would have tried harder to make uplink and downlink speeds more evenly balanced had they believed there would be a need to do so.

But the extent of the asymmetry is being shown to be much greater than the 2:1 ratio inherent to the standards.  Plum (in some work done for Qualcomm http://www.plumconsulting.co.uk/pdfs/Plum_June2011_Benefits_of_1.4GHz_spectrum_for_multimedia_services.pdf) indicated that the downlink to uplink asymmetry is in the ratio of around 8:1 and other studies put the figure as high as 10:1.  Streaming video (seen by many as one of the major consumers of Internet bandwidth in the future) is almost unidirectional.

With the exception of spectrum being used for TDD purposes, all currently licensed mobile FDD spectrum is split 50:50 between uplink and downlink.  Qualcomm were touting their proposed supplementary downlink to add capacity for downloads using the 1.4 GHz spectrum they won at auction in the UK (and which is still available in most other countries in the world due to it being officially set-aside for L-Band DAB services which have not taken off).  Even the European Broadcasting Union (EBU) were keen to highlight the potential additional downlink capacity that broadcast networks could offer if they worked more closely with mobile operators (instead of being at loggerheads with them over the valuable UHF spectrum they inhabit).

But this smacks of a case of 'closing the gate after the horse has bolted'.  If the asymmetry of data is really 8:1, and technology has an inherent built in disparity of 2:1, it seems logical that there should be around 4 times as much spectrum given over to downlink compared with uplink.  Within existing mobile bands (eg 900, 1800 and 2100 MHz) there is little to no scope to do things differently, but have regulators made a boob in thinking about how to split up new bands?  The Digital Dividend (790 to 862 MHz) is currently split into 2 symmetric blocks of 30 MHz plus the necessary guard bands. A 4 to 1 split would instead have suggested that 12 MHz should have been given to uplink with 48 MHz used for downlink. 

The situation at 2.6 GHz is even more acute.  As it stands, there are 2 equal blocks of 70 MHz (plus 50 MHz of TDD spectrum).  Surely this should have been more like 30 MHz uplink and 120 MHz downlink with TDD squished in the left over bits (which could also be used as downlink only).

Different ways in which the 2.6 GHz band could be split between FDD and TDD

 Of course, proponents of TDD technologies would argue that their services have the flexibility to re-assign up and down link capacity on an 'as needed' basis and that the correct solution would be to assign all new spectrum for TDD use. But this may not be efficient, as the need for additional 5 MHz guard bands between operators would reduce the effectiveness of the arrangement.  One of the most beneficial uses for TDD spectrum (and one which gets around the need for 5 MHz guard bands) would be to use it only for downlink.  If all base stations spent 100% of their time transmitting and 0% receiving (a perfectly valid combination for a TDD network), guard bands become unecessary and the TDD system becomes downlink only, which is kind of win-win for both FDD and TDD operators.

Ironically, in the UK, Ofcom's initial proposals for the auction of the 2.6 GHz band could have yielded results in which the amount of FDD spectrum was not 2 x 70 MHz as per international norms, but that this could be reduced in favour of assigning more spectrum to TDD.  One of the flaws in this design was the need for inefficient guard bands between FDD and TDD networks (and between TDD and other TDD networks).  But if these TDD blocks were downlink only the guard band problem goes away and the auction could have yielded the kind of 4:1 split in downlink and uplink spectrum that may be necessary - it would have at least opened up the opportunity for asymmetric up and downlink assignments.

So what can be done?  As things stand very little.  Current arrangements will inevitably lead to inefficiencies where uplink spectrum is underused compared to downlink spectrum.  Perhaps using the 1.4 GHz band or using broadcast networks to provide additional data capacity might help a bit, but unless the realisation that asymmetric allocations are necessary takes hold quickly amongst policymakers, we will be left with suboptimal allocations that will inevitably lead to poorer or more expensive services as operators use other methods to solve the capacity crunch.  Perhaps any 'Digital Dividend 2' should be downlink only (a.k.a. broadcast) to redress the balance a little - now there's a thought...

I count myself privileged to have been speaking at the Every European Digital conference in Brussels this week (http://www.eu-ems.com/summary.asp?event_id=94) alongside such industry denizens as Commissioner Neelie Kroes and the US Ambassador to the EU William Kennard (formerly chairman of the FCC).

The topic I was asked to address was 'Can spectrum provide the answer?' in this case meaning the answer to Europe's stated objectives to bring high speed broadband to every citizen.  My simple answer was 'no', but it might help a little.

A recent study conducted for Ofcom which showed LTE being just over 3 times more spectrally efficient than UMTS means that technology will, if we assume mobile data doubling each year, provide room for about 18 months growth. 

If regulators manage to find an additional 500 MHz of spectrum to add to the 500 MHz or so already available in most EU countries (such as is being proposed in Sweden, the UK and by several other regulators), this will deal with another 12 months of  data growth. 

So where is the rest of the growth going to go?  The answer is most likely in infrastructure. New cell sites are needed, but this will be difficult against a backdrop where operators are trying to consolidate their site portfolio to have less sites, but of higher quality.

The same is probably true of spectrum.  Operators would rather have a consolidated set of quality spectrum assignments rather than bit of spectrum real estate scattered hither and thither. And it's equally important to remember that all wireless networks, whether terrestrial or satellite, need secure access to spectrum.

So, 'can spectrum provide the answer'?  Not really, but it is definitely an essential part of the solution.

The full text of my conference speech is available on the Helios web-site: http://www.askhelios.com/rvw-speech-may11.html