By William Webb

brain-phone1

In the first instalment of the exclusive serialisation of his book Being Mobile on Telecoms.com, Ofcom’s William Webb looks at the potential of cognitive radio and white space devices.

The thinking behind cognitive devices has undergone much evolution over the last five years. The original idea of a cognitive device was one that, when turned on in a new environment, would scan the radio spectrum and make intelligent (or “cognitive”)  deductions about which spectrum was in use and which was free. It would then be able to decide which frequencies it might use to make transmissions and adapt its transmission parameters, such as bandwidth and modulation type, to best fit the available spectrum.

As the idea developed further, proponents focussed on UHF TV spectrum as the first band where the concept might be tried. Maps of TV coverage in the US tended to colour the country where there was TV coverage and leave other areas white. As a result, cognitive devices intended for commercial use in the TV bands came to be known as ‘white space devices’.

There is a snag with this approach, however, often referred to as the “hidden node problem”. Under certain circumstances, the cognitive device may not be able to detect a signal because of local topography and hence might transmit in error, causing interference. Resolving this problem directly either requires the primary user of the spectrum to accept occasional interference, or the cognitive device to be sufficiently sensitive that it can detect the signal even when the topography is problematic.

Primary users, unfortunately, are not inclined to accept interference. Television broadcasters argue that their viewers will not tolerate occasional loss of signal, for example, while radio microphone operators point out the highly problematic consequences of a radio microphone suffering interference during a major stage show.

The alternative is to make the cognitive device sufficiently sensitive that the hidden terminal problem occurs so rarely as to be considered resolved. In the USA and UK substantial work has been undertaken to determine what sensitivity might be needed. Definitively determining sensitivity is not possible because it depends on real-world geometries, obstacles and topographies, but substantial measurement and modelling have provided good guidance. The answer is that cognitive devices need to be extraordinarily sensitive, so much so that it is unclear whether such devices can actually be realised, let alone at a price that would appeal to consumers.

If the cognitive device cannot accurately determine the available spectrum, an alternative is for a central infrastructure to make this determination and transmit the information to the device. In such an approach a cognitive device determines its location, perhaps using GPS, and then reports this to a central database, perhaps via a cellular data channel. The database returns information on the frequencies available in the vicinity, possibly including additional information such as the transmit power that the device might employ or the likely duration of the channel’s availability.

For such an approach to work, the database must be supplied with detailed information about the licensed use including transmitter locations, transmitter power, antenna orientation and so on. It must also be aware of the operating parameters of the licensed systems such as the carrier-to-interference (C/I) level that the service needs to operate successfully. With this information it can derive a model of possible receiver locations and signal strengths and then, for every given point on a map, it can determine what signal strength a cognitive device could transmit with without causing interference to the licensed service.

While this resolves the problems with detection, it requires considerable organisation and agreement. One or more databases need to be created and maintained and protocols designed for communicating with them. The cognitive devices now need to have a means of determining their location and also of communicating to the database using non-cognitive wired or wireless systems. In addition, licensed users need to update the database whenever their use changes, which can be frequently in the case of uses such as wireless microphones and cameras.

Applications

The development of white space technologies could be considered as something of a “solution in search of a problem”. The solution was a way to access additional spectrum but the need for that access was not clear. So let’s have a look at some applications that might make use of white space.

Cognitive access is a somewhat uncertain process. In any given location there may be more or less spectrum available, conceivably even none. Available spectrum must be shared with other cognitive users with the result that congestion might occur. Such uncertain access often discourages network deployment because of the risk that the significant capital expenditure needed might not be recovered if spectrum availability declines. Furthermore, it is difficult to offer any kind of quality of service guarantees when the access to the spectrum itself is uncertain. This suggests that cognitive access will likely not be used for applications such as cellular or broadcast networks. Transmit powers also need to be relatively low to avoid interference to licensed users some distance away – in the UHF TV bands, transmit powers of less than around 4W have been suggested. These are well below those for base stations or broadcast systems.

Proponents of cognitive devices have taken two different approaches to discussing applications. The first is the “build it and they will come” approach. Here, proponents suggest that, while predicting applications is prone to failure, history suggests that if spectrum is made open and available then novel applications will emerge to use it. They cite WiFi and BlueTooth as examples. The second approach is to try to guess at a starting list of applications, noting that others may emerge. This list of applications includes:

A key question for cognitive access is whether or not the additional spectrum access enabled is sufficiently valuable to overcome the extra costs and complexities of a cognitive device compared to current unlicensed technologies. This is still very much an open question.

The current focus for database cognitive solutions is the UHF TV bands for some of the reasons set out above. However, there is no reason why the database should not be extended further. In principle, it could be made to cover all useful spectrum and become the default manner in which spectrum was managed.

This raises a philosophical issue. As mentioned above, many licence holders have nationwide licenses. It is arguable that any white spaces in their frequency bands belong to them, to dispose of or use as they see fit. A licence holder might argue that they should be able to profit from the use of their white spaces, in the same way that a home owner would expect to profit from the occupancy of their home while they were away for an extended period. Alternatively, the regulator (or others) might argue that if the spectrum is unused then there is no cost to the license holder in others accessing it and that this would provide benefit to the country. There is no right answer to this issue.

Database access could, in principle, be taken even further. Licensed use could be removed from multiple frequency bands and all access could be via a database. This is the “spectrum commons” utopia advocated by some.

However, many applications require certainty of access. Few operators will invest in costly widespread infrastructures if they are uncertain as to whether they will be able to access spectrum once deployed, or whether interference might reduce the capacity of their network. A system of cognitive access without any ability to reserve spectrum might discourage investment. Making all spectrum users adopt cognitive access on an unreserved basis does not appear appropriate or likely. A reservation capability could be built into any database, but if long-term reservations were made then this would then essentially be the same as a license for the spectrum.

Verdict

Cognitive access is less of a new technology than a new approach. Cognitive devices, likely using database access, need to implement a location sensor (eg GPS), a communications channel (eg GPRS) and a mechanism for tuning across a range of frequencies according to the instructions received – but none of these require new technology. Instead, a new approach to managing devices and regulating spectrum access must be adopted in order to implement and update the database. This is clearly possible technically and procedurally but might take some time and harmonisation before it can come into practice.

Cognitive applications are unclear but seem likely to be mostly restricted to unlicensed access where the market already has multiple successful technologies to select from. At best, cognitive might facilitate a range of new unforeseen applications that benefit from the greater range at lower frequencies or the access to more spectrum. At worst the added complexity of cognitive access might increase the cost of devices to a level where they cannot compete with existing unlicensed solutions. It is clearly worth facilitating cognitive access to see which of these outcomes will transpire.