Mind the gaps

Making better use of gaps in the radio spectrum could enable a spate of new applications. But it's not going to be as easy as some people think, according to E&T.

The radio spectrum is a valuable resource. Even if we're unlikely to see another government benefit to the tune of £22bn, as happened in the UK's 3G auction of 2000, access to key spectrum bands around the world routinely sells for billions of pounds, dollars or euros.

Yet tests suggest that only about 20 per cent of a key band of spectrum is ever in use at any given location. The rest may be unused for various reasons, such as the lack of an immediate need (the Ministry of Defence keeps some spectrum for use during military training exercises, for example), or because it provides a guard zone between two uses of the same frequency. The value of the spectrum and the fact that so much seems unused has led some to conclude that finding a way to access the remaining 80 per cent would generate enormous value.

The technologists' solution is to propose an intelligent or 'cognitive' radio system that can work out which parts of the spectrum around it are unused and adapt its transmissions to fit. The radio then makes short transmissions, pausing regularly to confirm that the spectrum is still unused.

The idea has gained some influential supporters, including Microsoft, Google, Intel, Motorola, Philips and others. They say that cognitive radio access could enable networks of connected devices, allowing all sorts of electronic equipment to wirelessly exchange messages and, they suggest, leading to a range of applications that we cannot yet imagine. Others perceive that cognitive devices could be used to provide broadband services to rural communities, enabling universal access.

Cognitive radio and the hidden node

The first problem that the backers of cognitive radio systems faced was that their devices were illegal. To transmit in radio spectrum you need a licence, or your device has to be exempted from licensing by a regulator such as Ofcom or America's Federal Communi-cations Commission (FCC).

Regulators are cautious about allowing cognitive radio. Despite recognising the potential benefits, they also realise that if a cognitive device does not work as planned it could interfere with transmissions from other licence holders. In particular, if a cognitive device deduces that a piece of spectrum is free when it is actually in use it could cause potentially serious interference. In the US, broadcasters are worried that cognitive devices might interfere strongly with their TV signals.

The problem of determining whether a channel is really unused turns out to be extremely difficult. Consider the diagram on page 69. A house receives a TV signal using a rooftop directional aerial mounted clear of surrounding buildings (path 1). A nearby mobile cognitive device is attempting to detect the same signal at street level, but it is blocked by surrounding buildings (path 2) and therefore much reduced in strength. The cognitive device might conclude that there are no transmissions and hence no active nearby receivers, transmit on the same band and cause harmful interference to the rooftop aerial (path 3). There are many similar situations for what has become known as the 'hidden node' issue.

Because of the risk of interference, regulators have not yet allowed cognitive radio access to spectrum that is wholly owned by a licence holder. However, the TV bands are a special case. TV transmitters in many countries are licensed on a location-by-location basis. Gaps are left between two transmitters using the same frequency to avoid interference and these gaps (or 'white spaces' on the coverage maps that are coloured to show frequency allocations) are not licensed to anyone. Hence, the regulator has a freer hand to allow cognitive radio. TV bands are also in spectrum with excellent propagation characteristics and are something like 300MHz wide - much wider than, for example, nearby cellular bands.

The FCC and Ofcom have both been looking at how a cognitive device might work in these TV bands, helped by substantial input and research from proponents of cognitive devices and from licence holders such as broadcasters and the users of radio microphones. A number of problems have emerged.

The first is that to reliably solve the hidden node problem by listening alone would require devices that can detect extra-ordinarily weak signals. This appears virtually impossible, especially in real-world situations where there are other signals in nearby bands and fluctuations in all signal levels. It would be like trying to see a very faint star at night when car headlights are flashing all around and clouds are obscuring the sky.

The second problem is that cognitive devices could prevent a licence holder from changing how they use a band, for example from second to third-generation cellular. This is because the cognitive device needs to be 'tuned' to the characteristics of the signal it is trying to detect in order to stand any chance of detecting that the band is in use. If a new technology with different characteristics is then introduced the cognitive device would usually be worse at detecting it, causing more interference and possibly making it too risky to deploy a new service. In just the same way that it would be virtually impossible to recall all Wi-Fi devices, recalling or modifying cognitive devices would likely not be possible. Some even argue that cognitive access could prevent any further improvement in wireless technology.

There may be a solution to these problems. Instead of the cognitive device trying to work out spectrum availability for itself, it could consult a central database. If the device could work out where it was (or 'geolocate'), it could then ask the database which spectrum was available locally. As long as the database is accurate this solves the hidden node problem as well as enabling new technologies to use existing spectrum: the database could be updated to accommodate these technologies, or even to exclude all cognitive devices from a band by showing that it was in use everywhere.

In November 2008 the US's FCC put forward legislation to allow cognitive access to TV bands for devices that could geolocate and consult a database. It also forms a key component of a consultation on cognitive access issued by Ofcom in early 2009.

Determining spectrum usage

Work underway on understanding cognitive access suggests that the idea that only 20 per cent of the spectrum is in use at once is probably misleading. Analysis shows that any device that is monitoring a band of spectrum is unlikely to be able to detect many of the ways it is being used, resulting in underestimates of its usage. Even if the spectrum really were unused, it looks likely that much of it will have to remain so in order to avoid interference. Imagine a snapshot of a road that showed cars only occupying 20 per cent of the tarmac - there are gaps, but they're necessary to avoid accidents, so a 20 per cent 'occupancy' may be the maximum safe usage of the road.

The work also tells us that it will be very difficult, perhaps impractical, for a cognitive device to reliably determine spectrum usage by sensing alone: it would need such a sensitive receiver that its cost would probably exceed what the market would pay.

The alternative of using a database looks more viable. Such databases can reliably report whether spectrum is unused. However, the scheme only works well when the licensed usage of spectrum changes relatively slowly. It would be less useful, for example, in sharing emergency service frequencies where an emergency could suddenly require full access to all channels. With a database approach there would be a time lag associated with loading the database with the updated licensed usage and then disseminating it to the cognitive devices, which could take hours. This is a problem even in the TV bands. Although TV transmissions rarely change, the band is also used by wireless microphones, which are often deployed at very short notice, particularly by news-gathering teams.

A database approach has other drawbacks. It requires devices to locate themselves and transmit that location through a conventional channel to the database, adding cost and complexity to the device. It also makes operations indoors difficult as most location systems do not work well inside buildings.

No revolution just yet

The initial promise of cognitive devices appears to have lost some of its shine. There is not as much free spectrum as was first thought, and accessing it will require expensive and complex devices. These devices will be competing with well-established technologies such as Wi-Fi and cellular networks, which are cheap and increasingly ubiquitous. Add in the increased device cost, the difficulty of indoor use, and the reduced battery consumption of cognitive devices and their shine becomes tarnished.

There are wider implications, too. It has often been suggested that managing the spectrum and issuing licences will become less relevant as devices become better at finding and exploiting gaps on a very short-term basis. But experience with cognitive devices so far suggests that they will find it very difficult to understand where those gaps are for the foreseeable future. The database solution shows more promise, and could be extended to larger areas of spectrum, enabling a more dynamic allocation and reservation of resources for a wide range of applications. It also offers potential for novel ways to manage spectrum.

Regulators are not concerned primarily with whether a new technology will succeed - although, of course, they hope it will. Instead, they see their role as enabling as wide a range of technologies as possible to flourish, leaving it to the market to determine which ones win out. So expect Ofcom and other regulators to work hard to set conditions that enable cognitive access. It's far less clear when we might see commercially successful devices in use, but it is just possible that some of the mechanisms developed to enable cognitive access will revolutionise spectrum use in the longer term.

Professor William Webb is head of R&D at Ofcom

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles