Cognitive Radio Access to TV White Spaces:Spectrum Opportunities, Commercial Applications

and Remaining Technology ChallengesMaziar Nekovee

BT Innovate and Design, Polaris 134, Adastral Park,Martlesham, Suffolk, IP5 3RE, UK

andCentre for Computational Science, University College London

20 Gordon Street, London WC1H 0AJ, UKmaziar.nekovee@bt.com

Abstract— Cognitive radio is being intensively researched asthe enabling technology for secondary access to the so-calledTV White Spaces (TVWS), large portions of spectrum in theUHF/VHF bands which become available on a geographical basisafter digital switchover. Both in the US, and more recently, inthe UK the regulators have given conditional endorsement to thisnew mode of access. This paper reviews the state-of-the-art intechnology, regulation and standardisation of cognitive access toTVWS. It examines the spectrum opportunity and commercialuse cases associated with this form of secondary access.

I. INTRODUCTION

A cognitive radio [1] consists of a cognitive engine (CE),which contains algorithms and toolboxes for radio environ-ment sensing, machine-learning, and reasoning and decisionmaking, and a configurable radio platform, which could bea Software Defined Radio (SDR), that basically does what itis told by the CE. The concept of CR was first describedby Mitola and Maguire [2] as “transforming radio nodesfrom blind executors of pre-defined protocols to radio-domain-aware intelligent agents that search out ways to deliver theservices that the user wants even if that user does not knowhow to obtain them”. The ideal CR knows everything aboutthe user requirements, the capability of the radio device, thenetwork requirements and the external environment (includingthe radio environment). It will plan ahead and negotiate for thebest part of the spectrum to operate in and at the best power,modulation scheme etc, and manage these resources in realtime to satisfy the service and user demands. The ideal CRis currently at the early proof-of-concept stage research, withmost of the work taking place in universities.

A much more developed form of the CR technology iscognitive radio for dynamic spectrum access (DSA) [3]. Theaim here is to achieve device-centric interference control anddynamic re-use of radio spectrum based on the frequencyagility and intelligence offered by cognitive radio technology.This form of CR technology is currently being intenselyresearched. However, there is also already significant industryeffort towards prototyping, standardisation and commercialisa-

tion of the technology. Important industry players with activeR&D efforts in cognitive radio technology include Alcatel-Lucent, Ericsson and Motorola from the mobile equipmentindustry, BT and Orange from network operators, Philips andSamsung from the consumer electronics industry, HP and Dellfrom the computer industry, and Microsoft and Google fromthe Internet/software industry. Dynamic spectrum access maytake place in several ways: between a licensed primary systemand a licensed-exempt secondary system, e.g. secondary spec-trum access to digital TV or military spectrum, within the sameprimary system, e.g. micro-macro sharing of 3G spectrum infemtocells, and finally among two primary systems, e.g. real-time leasing and trading of spectrum between two cellularoperators.

The first form of dynamic spectrum access is arguablythe most disruptive application of the CR technology, asit enables licensed-exempt users (end-user devices and basestations) to act as spectrum scavengers. They can identifyunused portions of licensed spectrum (also called spectrumholes or White Spaces) and make use of this spectrum fortheir connectivity at times and/or locations where they are notused. Allowing the operation of such scavengers promises togreatly increase the efficiency of spectrum usage by preventingexclusively licensed spectrum from being wasted due to lowspatial or temporal usage. Mainly for this reason licensed-exempt cognitive access to certain licensed bands is beingkeenly promoted by the US regulator, the FCC (FederalCommunication Commission) [4], [5], and more recently alsoby Ofcom [6], [7], [8]. The rationale is to maximise the usageof licensed spectrum through secondary access by cognitiveradios and, at the same time, promote rapid introduction ofnew wireless technologies and services without the need forsetting aside any new spectrum for this purpose. Most mobileoperators see this from of cognitive access as highly disruptiveto their current business model.

To date both in the UK [8] and US [5] regulators havecommitted to licence-exempt cognitive access to the so-calledTV White Spaces (TVWS). The TVWS spectrum comprises

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

978-1-4244-5188-3/10/$26.00 ©2010 IEEE

large portions of the UHF/VHF spectrum that become avail-able on a geographical basis for cognitive access as a resultof the switchover from analogue to digital TV. The totalcapacity associated with TVWS is significant. According tomodelling studies commissioned by Ofcom over 50% oflocations in the UK are likely to have more than 150 MHZof interleaved spectrum and that even at 90% of locationsaround 100 MHz of interleaved spectrum might be availablefor cognitive access [8]. In addition to TVWS, the defencespectrum may provide another significant capacity opportunityfor license-exempt cognitive access. For example, around 30%of spectrum below 15 GHz is allocated to Defence in theUK. The UK MoD (Ministry of Defence) had until the late1990’s access to spectrum at no or a low cost. However,following the Cave Audit, the Government has committed toreleasing a ”significant proportion” of the MOD’s spectrumbetween 2008 and 2010. Results form a 2008 study by PAconsulting (commissioned jointly by MoD and Ofcom) suggestthat [9] there is significant scope for license-exempt use of thereleased spectrum using cognitive radio technology, both on aspatial and a temporal basis. For example, low power cognitivedevices could potentially share with radar if the radar sweepcan be detected and the transmission of the cognitive devicecan be timed to avoid interference.

This paper aims to review the state-of-the-art in technology,regulation and standardisation of cognitive radio access toTVWS. It also examines the spectrum opportunity, potentialbusiness applications and some of the open research challengesassociated with this new form of access, drawing lessons andconclusions from recent recent findings in the UK [26], [28],US [14], [15] and elsewhere. The rest of this paper is organisedas follows. Section II provides a brief overview of cognitiveradio access to TV White Spaces. In Section III the regulatorystatus and standardisation efforts are reviewed and some of theoutstanding research and technology challenges are discussed.In Section IV we discuss our results on quantifying theavailability, spatial variation and frequency decomposition ofTVWS spectrum for cognitive access in the UK. In section Vcommercially prominent candidate use cases of this spectrumare outlined and some prominent cases are examined. Weconclude this paper in Section VI.

II. COGNITIVE ACCESS TO TV WHITE SPACES

A. What are TV White Spaces?

Broadcast television services operate in licensed channelsin the VHF and UHF portions of the radio spectrum. The reg-ulatory rules in most countries prohibit the use of unlicenseddevices in TV bands, with the exception of remote control,medical telemetry devices and wireless microphones. In mostdeveloped countries regulators are currently in the processof requiring TV stations to convert from analogue to digitaltransmission. This Digital Switchover (DSO) is expected tobe completed in the US in 2009 and in the UK in 2012. Asimilar switchover process is also underway or being planned(or is already completed) in the rest of the EU and manyother countries around the world. After Digital Switchover a

portion of TV analogue channels become entirely vacant dueto the higher spectrum efficiency of digital TV (DTV). Thesecleared channels will then be reallocated by regulators to otherservices through auctions.

In addition to cleared spectrum, after the DTV transitionthere will be typically a number of TV channels in a givengeographic area that are not being used by DTV stations,because such stations would not be able to operate withoutcausing interference to co-channel or adjacent channel stations.However, a transmitter operating on such a locally vacant TVchannel at a much lower power level would not need a great(physical) separation from co-channel and adjacent channelTV stations to avoid causing interference. Low power devicescan therefore operate on vacant channels in locations thatcould not be used by TV stations due to interference planning.These vacant TV channels are known as TV White Spaces orInterleaved Spectrum in the language of the UK regulator.

B. Detection and incumbent protection

Secondary operation of cognitive radios in TV bands relieson the ability of cognitive devices to successfully detectTVWS, and is conditioned by regulators on the ability ofthese devices to avoid harmful interference to licensed users ofthese bands, which in addition to DTV include also wirelessmicrophones. Both the FCC and Ofcom have considered threemethods for ensuring that cognitive devices do not causeharmful interference to incumbent: beacons, geo-location com-bined with access to a database, and sensing. Currently, thedatabase approach seems to offers the best short-term solutionfor incumbent detection and interference avoidance. Both inthe US and UK regulatory and industry efforts is, therefore,underway to further develop the concepts, algorithms andregulatory framework necessary for this approach.

Beacons

With the beacon method, unlicensed devices only transmitif they receive a control signal (beacon) identifying vacantchannels within their service areas. The signal can be receivedfrom a TV station, FM broadcast station, or TV band fixedunlicensed transmitter. Without reception of this control signal,no transmissions are permitted. One issue with the controlsignal method is that it requires a beacon infrastructure to bein place, which needs to be maintained and operated, either bythe incumbent or a third party. Furthermore, beacon signals canbe lost due to mechanisms similar to the hidden node problemdescribed below.

Geo-location combined with database

In this method, an unlicensed device incorporates a GPSreceiver to determine its location and accesses a database todetermine the TV channels that are vacant at that location.There are at least three issues associated with this method.There is a need for a new (commercial) entity to build andmaintain the database. Devices need to know their locationwith a prescribed accuracy. For outdoor applications GPS canbe used to support these requirements, but in the case of indoor

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

Fig. 1. The hidden node problem of a sensing-based cognitive radio.

application there are issues with the penetration of GPS deepinside buildings. Finally, devices need additional connectivityin a different band in order to be able to access the databaseprior to any transmission in DTV bands.

Sensing

Finally, in the sensing method, unlicensed devices au-tonomously detect the presence of TV signals and only usethe channels that are not used by TV broadcaster. Detectionof the TV signal can be subject to the hidden node problem,which is depicted in Fig. 1. This problem can arise when thereis blockage between the unlicensed device and a TV station,but no blockage between the TV station and a TV receiverantenna and no blockage between the unlicensed device andthe same TV receiver antenna. In such a case, a cognitive radiomay not detect the presence of the TV signal and could startusing an occupied channel, causing harmful interference to theTV receiver.

C. Regulatory developments

US

In the US the FCC proposed to allow opportunistic accessto TV bands already in 2004 [4]. Prototype cognitive radiosoperating in this mode were put forward to FCC by Adaptrum,I2R, Microsoft, Motorola and Philips in 2008. After extensivetests the FCC adopted in November 2008 a Second Report andOrder that establishes rules to allow the operation of cognitivedevices in TVWS on a license-exempt basis [5]. In summarythese rules require cognitive devices to use both spectrumsensing and geolocation. In order to minimise the chance ofharmful interference due to the hidden node problem FCChas required that cognitive devices should be able to senseboth television signals and wireless microphones down to -114 dBm. They must also locate their position to within 50metres and then consult a database that will inform them aboutavailable spectrum in that location [5].

Mobile devices may transmit in a locally vacant TV channelat up to 100 mW unless they are using a channel adjacent toterrestrial television, in which case their transmission powercan only be 40 mW. Fixed devices (base stations or customer

premises) may transmit at a locally vacant channel at up to4W (EIRP). Devices without geolocation capabilities are alsoallowed if they are transmitting to a device that has determinedits location. In this case, one device would be acting as amaster for a network and the other slave devices would operatebroadly under its control in terms of the spectrum they woulduse. Devices that use sensing alone are allowed in principle;however, they must be submitted in advance to the FCC forlaboratory and field testing so the FCC can determine whetherthey are likely to cause harmful interference. The exact processthat the FCC will use to determine this has not been specified.

Importantly, the FCC report includes a detailed discussionabout whether cognitive access should be licensed, licence-exempt or subject to light licensing. It concludes that the bestway to facilitate innovative new applications is via licence-exemption and that licensing would not be practicable formany of the new applications envisaged. It also notes that anylicenses would be difficult to define and subject to change (e.g.if television coverage was re-planned), so the rights awardedwould be rather tenuous.

UK

In its Digital Dividend Review Statement released in De-cember 2007 the UK regulator, Ofcom, proposed to “allowlicence exempt use of interleaved spectrum for cognitive de-vices” [6]. Furthermore Ofcom stated that “We see significantscope for cognitive equipments using interleaved spectrumto emerge and to benefit from international economics ofscale [6]”. In a consultation published on 16 February 2009 [7]Ofcom proposed a number of technical parameters for licence-exempt cognitive use of interleaved spectrum which closelyfollow those suggested by FCC.

Subsequently, in a statement published on July 1 2009Ofcom proposed to allow sensing alone as well as geolocationfor incumbent detection [8]. However, it concludes that inthe short term the most important mechanism for spectrumdetection will be geolocation. Ofcom is suggesting that fur-ther work, possibly leading to a consultation specifically ongeolocation, is appropriate. Finally Ofcom states that it “willwork with stakeholders to further develop the concepts andalgorithms necessary for geolocation and expect to consultfurther on Geo-location later in 2009” [8].

Worldwide

Work on a pan-European specification for cognitive devicesis currently taking place within the SE34 working group ofCEPT (the European Conference of Postal and Telecommu-nications Administrations). An important aim of this groupis to define technical and operational requirements for theoperation of cognitive radio systems in TV White Spaces inorder to ensure the protection of incumbent services/systemsand to investigate the amount of spectrum across Europe thatis potentially available as White Spaces. Furthermore on aworldwide scale, agenda item 1.19 of the WRC-11 (WorldRadiocommunications Conference, 2011) will be consideringregulatory measures and their relevance, in order to enable

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

the introduction of software defined radio and cognitive radiosystems, based on the results of ongoing ITU-R studies.

D. Standardisation and industry effort

Industry-led research and development on cognitive radiotechnology has been so far mainly focused in the USA, and islargely driven by the desire of important new players, includ-ing Google and Microsoft, to get access to the TVWS whichwill be released this year. However, a number of EU-backedand industry-led collaborative projects are currently underwaythat aim at bringing cognitive radio technology in Europecloser to commercial exploitation. Two major standardisationefforts, which are currently at an advanced stage, are discussedbelow.

The Cognitive Networking Alliance (CogNeA) Standard

The Cognitive Networking Alliance (CogNea) [10] is anopen industry association. The Alliance intends to commer-cialise low power personal/portable Cognitive Radio platformsby enabling and promoting the rapid adoption, regulation, stan-dardisation and multi-vendor compliance and interoperabilityof CRs world wide. Alliance board members include ETRI,HP, Philips, Samsung Electro-Mechanics, Texas Instruments,and more recently BT. The initial geographical focus area isNorth America. The initial focus radio spectrum is TV WhiteSpaces.

The Alliance intends to promote TVWS spectrum regu-lations worldwide, and to establish a recognisable CogNeAbrand that indicates a device is CogNeA-compliant and cantherefore interoperate with other CogNeA-certified devicesfrom various manufacturers. The Alliance also develops spec-ifications for the Common Cognitive Radio Platform (CCRP)which supports multiple applications [11].

The Alliance intends to bring the standard to an internationalstatus, in collaboration with an existing Standards DefinitionOrganisation (SDO), to make it globally accepted. The primarytarget applications for the CogNeA standard are:

• in-home high definition multimedia networking and dis-tribution solutions that overcome the whole home cover-age problems inherent to solutions using ISM bands

• Unlicensed broadband wireless access for communi-ties/neighbourhoods/campuses

The standard is developing a Common Radio Platformconsisting of the Physical Layer (PHY) and the Media AccessLayer (MAC). The PHY consists of the Radio Front End, theBaseband and the Cognitive Entity, which contains a geo-location block, a sensing block and an Internet access andinterference map resources. The MAC carries the Communi-cation/Networking protocol, Air access rules, and interface forthe higher layers, such as network and application layers.

ECMA International is currently developing a high-speedwireless networking standard for use in the Television WhiteSpaces, based on the contribution from CogNeA. The standardwill employ cognitive radio sensing and database technolo-gies to avoid interference with licensed services and otherincumbent users in compliance with the FCC regulatory rules.

The first edition of the standard was is expected finalised andpublished in December 2009 1

The IEEE 802.22 Standards

The IEEE 802.22 Working Group [12] has defined anair interface (PHY and MAC) standard based on cognitiveradio techniques. The 802.22 standard is being developed forWireless Regional Area Networks (WRANs). The primarytarget application of the standard is licensed-exempt broad-band wireless access to rural areas in TVWS. The initialgeographical focus area is North America. The 802.22 systemspecifies a fixed point-to-multipoint wireless air interfacewhereby a base station (BS) manages its own cell and allassociated Consumer Premise Equipments (CPEs). The net-work architecture including MAC and PHY are derived fromIEEE 802.16 WiMAX. The 802.22 PHY layer is designedto support a system which uses vacant TVWS channels toprovide wireless communication access over distances of upto 100 Km. The PHY specification is based on OrthogonalFrequency Division Multiple Access (OFDMA) for both theupstream and downstream access.

The IEEE 802.22 standard supports incumbent detectionthrough spectrum sensing (the database approach is optional).The standard specifies inputs and outputs for the sensingfunction, as well as the performance requirements for thesensing algorithms implemented (e.g. probability of detection,incumbent detection threshold and probability of false alarms).These include energy detector, and cyclostationary and pilotsensing detectors for ATSC DTV signals, and an FFT-basedalgorithm for detection of wireless microphone signals [13].

The IEEE 802.22 defines a connection oriented and cen-tralised MAC layer. Two important capabilities are intro-duced in the 802.22 MAC layer to support reliable incum-bent detection based on sensing: network-wide quite periodsscheduled by each BS during which all transmissions aresuspended in order to allow reliable sensing, and channelmeasurement management to coordinate distributed channelmeasurement/incumbent detection by CPEs and their reportingto BS.

After an initial accelerated phase, the development of thestandard seems to have slowed down during the last year. Ac-cording to the IEEE 802.22 sources, the standard is currentlyat the Ballot stage. However, the final completion date forstandard is not known yet, and there have been no vendorcompanies so far to build equipment based on the IEEE802.22.

III. RESEARCH CHALLENGES

A. High-precision spectrum sensing

In order to minimise the chance of harmful interference dueto the hidden node problem both the FCC and Ofcom requirethat cognitive devices should be able to sense TV signalsat detection margins much lower than that of TV receivers

1. see www.ecma-international.org/publications/files/ECMA-ST/ECMA-329.pdf

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

(114 dBm for 6 MHz US channels and −120 dBm for 8MHz UK channels) [5], [8]. Such weak signals are wellbelow thermal noise, and cannot be detected using the energydetection algorithms that are implemented in the currentgeneration of wireless devices [16]. Recent research, however,shows that such sensing levels can be achieved by moresophisticated sensing algorithms that rely on certain features ofincumbent signals, such as cyclostationary properties and auto-correlation [17]. However, such advanced sensing algorithmsrequire considerable processing power which may be eithernot available or not desirable (due to power consumption) inhandheld devices.

There has been considerable recent research in cooperativedetection algorithms, where sensing measurements performedby multiple devices are combined (either on the physicallayer or application layer) in order to achieve higher sensingthresholds than is possible by single devices or to deal withthe hidden node problem [19], [18]. Interestingly, cooperativesensing was also considered in Ofcom’s consultation on cogni-tive access as a possible approach to the detection problem [7].

There are, however, two main issues with this approach. Oneissue is that the achievable detection level depends on severalfactors, including the number of cooperating cognitive devicesand their spatial arrangement [16,17]. Therefore, in general itwould be difficult to test and certify the detection capability ofsuch cooperating cognitive devices on an individual basis tocheck device compliance with regulatory requirements [18].Furthermore this method requires additional communicationoverhead since local measurements will be collected at systemlevel in order to make a decision, which is then broadcast toall cognitive radios involved.

A second problem with the sensing approach is that theability of cognitive devices to sense extremely weak TV sig-nals may eliminate the hidden node problem (false negatives)but at the same time it can lead to a situation where acognitive radio detects TV signals from transmitters that areperhaps hundreds of kilometres away (false positives), therebyremoving a considerable portion of usable White Spaces. Veryrecent studies in the US, for example, indicate that a thresholdof -114 dBm reduces the recoverable White Spaces by a factorof 3 [14]. Even worse, initial modelling studies performed atBT [20] show that in some UK locations, a cognitive devicewith a -114dBm sensitivity level will identify all DTT channelsas occupied, and therefore will have no White Space availablefor its operation if it relies on naive sensing only!

B. Agile transmission and spectrum pooling techniques

Physical layer transmission technique that are able to effec-tively deal with the fragmented nature of TVWS spectrumare a very important component of future cognitive radiosIn particular, these techniques must be sufficiently agile toenable unlicensed users to transmit in (locally) availableTVWS bands while not interfering with the incumbent usersoperating at adjacent bands. Moreover, to support throughput-intensive applications, these techniques should be able toachieve high data rates by pooling several (not necessarily

Fig. 2. In future commercial applications, such as BT FONs communityWiFi networks, the power levels of cognitive devices need to be controlledon a system-wide level in order to avoid aggregated interference to primaryreceivers.

contiguous TVWS channels). One technique that seems tomeet both these requirements is a variant of orthogonal fre-quency division multiplexing (OFDM) called non-contiguousOFDM (NC-OFDM) [21]. NC-OFDM is capable of deac-tivating subcarriers across its transmission bandwidth thatcould potentially interfere with the transmission of other users.Moreover, NC-OFDM can support a high aggregate data ratewith the remaining subcarriers, and simultaneously maintainan acceptable level of error robustness. In addition to NC-OFDM several other techniques have been proposed to enableagile waveforming over fragmented spectrum. One prominentexample is the use of filterbank multicarrier techniques forsuch cognitive radio applications [22].

C. Multiple antenna technologies for cognitive radio

The use of antenna diversity or MIMO antenna architecturecan provide a significant increase in the spectral efficiencyof wireless systems [24], [25]. However, the use of multipleantennas in cognitive radio networks is underdeveloped. Oneof the major objectives for cognitive radio is to improve thespectrum utilisation. With the advantages offered by MIMOsystems, it is therefore logical to exploit potentials in applyingthe MIMO antenna architecture to cognitive radio networks.Introducing multiple antenna technologies for cognitive radiowill extend the dimension of CR from the current frequencyband and time slot regime even further into spatial domain.Furthermore, the use of multiple antennas can significantly im-prove sensing capabilities of cognitive devices. One issue withthe use of multiple antennas in the context of cognitive accessto TVWS is that the typical wavelengths in the UHF bandsvary between 0.3-0.6 m. Optimal use of multiple antennas on asingle cognitive device, therefore may not be feasible in mostapplications due to the small footprints involved. However,fixed BWA applications similar to that considered in the IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

Fig. 3. The UK UHF bands after completion of Digital Switchover [7].

802.22 that involve large base stations and customer premisesmay greatly benefit from multiple antenna technologies.

D. System-level issues

Most of the research on cognitive radio applications so farhas focused on a single cognitive device accessing TVWSspectrum. However, the provision of commercial servicesbased on cognitive radio technologies, e.g. mobile broadbandor wireless home networks, will inevitably involve situationsinvolving multiple cognitive equipments that may belong toeither the same or different service providers. Some openresearch challenges associated with such service scenariosinclude the following:

• how to control aggregated interference impacts of cogni-tive devices [23] towards primary users

• politeness (etiquette) rules that achieve fair and efficientsharing of secondary spectrum among competing cogni-tive radios [7]

• spectrum sensing under aggregate interference conditions• quantitative understanding of secondary user performance

degradation as a function of the number of secondaryusers and the achievable service levels using secondaryaccess to TVWS.

IV. QUANTIFYING THE WHITE SPACE AVAILABILITY

Fig. 3 shows allocation of the UHF spectrum in the UK afterthe completion of DSO [8]. The 128 MHz of spectrum markedin green (16 bands) is the cleared spectrum which Ofcomplans to license through auctions. The 256 MHZ (32 channels)marked in purple is the interleaved spectrum which can beused on a geographical basis for license-exempt access byusing cognitive radio technology. Finally the channel markedin pink is licensed by Ofcom for exclusive access for wirelessmicrophones etc (PMSE).

From the above chart it appears that there is significantcapacity available for cognitive access in the UHF bands.However, due to its secondary nature the availability andfrequency decomposition of the UHF spectrum for cognitiveaccess is not the same at all locations and depends also on

the power levels used by cognitive devices [26], [14]. This isan important feature of license-exempt cognitive access to TVbands which distinguish it from, e.g. WiFi access to the ISMbands.

Potential commercial applications of TVWS devices willstrongly depend on how the availability of this spectrumvaries; both from location to location and as a function oftransmit power of cognitive devices. A number of recentstudies have investigated various aspect of TVWS spectrumin the US [15], [14]. Very recently we developed a modellingtool for quantifying the availability of TVWS spectrum in theUK and its variation with location and transmit power. Ourmodelling tool [26] makes use of the publicly available mapsof DTV coverage in the UK [27] which were generated viacomputer simulations from the Ofcoms database of location,transmit power, antenna height and transmit frequency ofUKs DTV transmitters. It combines these coverage maps withsimplified propagation modelling calculations to obtain upperbounds for the vacant TWVS frequencies at any given locationas well as a lower-bound estimate for the variations of TVWSspectrum with the transmit power of a cognitive devices.

The computer model for obtaining the upper bounds worksas follows [26]. We use the UK National Grid (NG) coordinatesystem in order to specify the geographical position of anylocation on the UK map. Given the NG coordinates of a UKlocation the computer code then maps this location onto theclosest grid point on the DTV coverage maps. For a givenDTV transmitter this grid point is then evaluated to determineif it falls within the coverage area of that transmitter. If this isthe case, then the frequencies associated with the transmitterare tagged as occupied at those locations, otherwise they aretagged as vacant. Repeating this procedure for coverage mapsof all DTV transmitters, we then obtain a list of vacant TVfrequencies at a given location that can be used by a low-powercognitive devices which is positioned in that location.

In the case of high power cognitive equipments, e.g. thoseconsidered within the 802.22 standard, the required computa-tions are very intensive. In order to reduce this computationaleffort, we approximate the actual DTV coverage areas bycircular disks which were constructed such that each of thementirely encompassed the coverage area of the associated trans-mitter while also having the minimum possible surface area.With this simplification, it is then computationally straightfor-ward to calculate from the vacant TV frequencies as a functionof both position and transmit power of cognitive devices.

We have used our modelling tools to investigate the vari-ations in TVWS as a function of the location and transmitpower of cognitive radios, and to examine how constraints onadjacent channel emissions of cognitive radios may affects theresults. This analysis provides a realistic view on the potentialspectrum opportunity associated with cognitive radio accessto TWVS in the UK, and also presents the first quantitativestudy of the availability and frequency composition of TWVSoutside the United States. Fig. 4 summarises in a bar-chart theavailability of TVWS channels for 18 major population centresin England, Wales and Scotland. The total number of channels

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

available at each location is shown as a blue bar. These resultsshow that there are considerable variations in the number ofTVWS channels as we move from one UK location to another.For any given location, however, a minimum of 12 channels(96 MHz) are accessible to low-power cognitive devices, whilethe averaged per location capacity is just over 150 MHz.

When a high power cognitive device operates in a vacantTV channel, energy leakage to adjacent channels may causeinterference to adjacent frequencies, which may be occupied.Ofcom had raised concerns that operation of low-power cog-nitive devices on a given channel may also cause adjacent-channel interference for mobile TV receivers that are in closevicinity. Consequently, even in some future use cases, cogni-tive devices may be constrained not to use vacant channelswhose immediate adjacent frequencies are used for mobileTV. The total number of available TVWS after imposing theabove adjacent channel constraint are shown as red bars inFig. 4. It can be seen that imposing the constraint greatlyreduces the amount of accessible spectrum in most locationsconsidered (on average the available capacity drops to justbelow 40 MHz/location).

Recent studies on quantifying the availability of TVWS inthe United States were reported in [15], [14], and the resultare in line with our findings for the UK. In particular adetailed study performed in [14] shows that in the US themain channels of relevance are the lower UHF channels where∼ 15 (90 MHz) channels per location/per person are availablefor low power cognitive access. However this number dropssignificantly (to ∼ 5) when adjacent channels also have to beprotected.

In addition to estimating total available TVWS, it is of im-portance to investigate channel composition of this spectrum.In Fig. 5 we show, as an example, channel composition ofTVWS in Central London. In this Figure vacant channels areshown as blue bars while occupied channels are left black.As can be seen from the figure, the available TVWS channelscan be highly non-contiguous. This feature may greatly restrictaccess to TVWS by most current wireless technologies, asmodulation schemes implemented in these technologies oftenrequire a contiguous portion of the spectrum. In the case ofLondon although a total of 96 MHz spectrum is in princi-ple available, only 16 MHz can be utilised for contiguousfrequency access. Figure 6 shows the maximum number ofcontiguous TVWS channels that are available at the above 18locations, and further highlights the non-contiguous nature ofTVWS spectrum. For example, in Edinburgh and Manchesteralthough the total available TVWS bandwidth is 104 MHz,only 24 MHz and 16 MHz, respectively, is available asone contiguous block. Consequently, at such locations non-contiguous channel bonding need to be performed by thecognitive radios in applications such as wireless distributionof multiple HDTV streams around home.

V. USE CASES

In addition to considerable capacity it offers, which isevident from the discussion in the previous section, an impor-

Fig. 4. Available TVWS capacity for low-power cognitive access in18 UK locations as obtained from coverage modelling. Results are shownboth without (green bars) and with (red bars) considering adjacent channelinterference constraint.

Fig. 5. TVWS channels available for cognitive access in Central London.

Fig. 6. Maximum contiguous TVWS bandwith available in 18 locations inthe UK.

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

tant reason why TVWS spectrum has attracted much interestis an exceptionally attractive combination of bandwidth andcoverage. Signals in the VHF/UHF TV bands travel muchfurther than both the WiFi and 3G signals and penetratebuildings more readily. This in turn means that these bandscan be used for a very wide range of potential new services.In addition to broadband wireless access to underserved areas,which is specifically targeted by the IEEE 802.22 standard,other technologically important applications of TVWS spec-trum which are also commercially of great interest include thefollowing:

• Wireless home networks• Smart metering• Femtocells• Mobile broadband• Vehicular communications for intelligent transport

In the following paragraphs we shall focus our attention onindoor use cases. These use case are of significant interest toboth fixed and mobile operators and because of the relativelylow transmit power levels involved can utilise a maximumnumber of TVWS channels [26].

A. Next generation home networks

Fuelled by the quick progress of wireless technologies,broadband adoption, and without the burden of spectrumlicences, home wireless networking has become in the lastfew years a pervasive technology. Between 2004 and 2006,home network adoption boomed across Europe, with growthrates surpassing Asia and North America. France and the U.K.both trebled the number of households with a home network,putting them slightly ahead of the U.S. Italy and Germany stilllagged behind but posted notable growth nonetheless. Morethan 54% of European households have a computer and a totalof 34% are using WiFi routers. The future wireless home willconsist not only of PC, laptops and PDAS wirelessly connectedto the Internet but also media servers (High Definition TV,video and audio), access points, computer electronics likewireless cameras and game consoles, as well as domesticapplications, and gas, electricity and water meters will in thefuture come equipped with radio receivers allowing control,monitoring and easy configuration.

Most of these devices and services support wireless con-nectivity using one or a number of short-range wirelesstechnologies, such as WiFi (IEEE 802.11) , Zigbee (IEEE802.15.4) etc all operating without the need for a licencein the already congested ISM bands. Home networks of thefuture operating exclusively in these bands are expected tosuffer severe capacity limitations resulting from interferencecaused by the high device density and limited spectrumavailability in the ISM bands. Furthermore, the aggregatedinterference resulting from these devices is bound to createa high interference burden on the WiFi-based provision ofbroadband wireless access in homes.

Additional capacity offered by secondary access to TVWShas the potential to solve this capacity limitation problemthereby contributing to increasing takeup of wireless home

networking and services, and spurring future technologicalinnovation and revenue generation. In particular, some of themost bandwidth-intensive home networks applications (such asmultimedia streaming) can be offloaded to TVWS bands hencefreeing up the ISM bands for other consumer applications.

Our recent system-wide simulation studies of such scenariosshow that [28] due to lower operation frequencies home accesspoints operating in the TWVS UHF frequencies can achievedata rates that are either higher or comparable to WiFi@5GHZ(802.11n) while using significantly lower transmit power levels(two order of magnitudes in mW) [28]. An additional benefithere is that a significant saving in energy consumption can beachieved in home networking scenarios by switching from theISM bands to TVWS bands [28].

Protection of DTV in such home networking scenarios canbe achieved using a master-slave architecture where function-alities of spectrum detection and/or geolocation and databaseaccess and spectrum assignment are all integrated into thehome access point. The access point monitors the availabilityof spectrum in the ISM and TV bands and instruct customerdevices which spectrum to use based on their bandwidth andQoS requirements 2. The proposed architecture is shown inFig.7.

B. Smart metering

Smart meters are the next generation of utility meters thatprovide accurate real time information on consumption to theuser as well as the utility company. The consumers can usethe information available through the new system to help themsave money and reduce the overall carbon emission. They willalso be given more choice to switch between energy suppliersmore easily. For utility companies, this will save costs in meterreading and allow them to offer grater range of tariff packages.They could also use smart meters to promote behaviour thatwould save energy. For example, use of renewable energysources, home and micro-generation and electric vehicles byutilising time of use tariffs. It is expected that smart meteringwill play an important role in transforming most developedcountries to a low-carbon economies. In the UK, for example,it is estimated that rolling out smart meters to all domestichouseholds by end 2020 will deliver net benefits between2.5bn and 3.6bn over the next 20 years and reduce carbonemissions of about 2.6 million tonnes per year by 2020 [30].Although there have been small scale attempts at deployingsmart meters in other parts of the world, such as in Italy,USA (California), Canada, Australia, New Zealand and severalother countries, nowhere is it scaled to the extent undertakenby UK. Although the detail design is not completed, basedon the requirements for smart metering in UK, it is apparentthere could be two levels of communication, as seen in Figure8:

• Communication between smart meters and devices in thehome via a Home Area Network (HAN)

2. Protection of wireless microphones is more problematic and may requiresthat the homehub perform periodic sensing to detect the activity of suchdevices [29].

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

• Communication delivered to the home via a Wide AreaNetwork (WAN).

Zigbee is currently being seriously considered as the de-facto standard for communication between meters and HANwhile the communication to and from the home has severalcontenders. Whatever the access technology, the WAN is alsolikely to have a Network Management System (NMS) in theback end co-ordinate the messages with the utility providers.

The use of Zigbee operating in the ISM bands for com-munication to/from smart meters faces at least two importantissues

• Coverage: This probably is one of the biggest challengesdue to the sheer variability of meter locations. Somemeters may be in outside garages, or in metal cagesand under stairs that makes guarantee of wireless prop-agation difficult. One possibility is to use Zigbee meshnetworking to allow the distant meters to connect withthe Neighbourhood Area Networks (NANs) to mitigatethe effect. Even this is unlikely to cover all cases.

• Interference issues A study carried out by the ZigbeeAlliance indicated that Zigbee when operated in residen-tial environments with typical WiFi usage patterns didnot affect performance for Zigbee. However when WiFiis used with higher power and higher duty cycle thanthose applications available today it can cause severedegradation on the performance of Zigbee. Althoughtheoretical limit for Zigbee and WiFi coexistence, areunlikely to be reached with the applications availabletoday. The interference issue may become more pressingwhen smart metering becomes a main stream applicationin homes.

Using TVWS frequencies for smart metering, possibly imple-mented in the Zigbee standard, can provided a feasible solutionto the above issues. In particular, given the relatively low datarates involved in smart metering communications within eachhome, a single (or even a small sub-channel) TVWS channelcould be assigned for such applications in order to providewhole-home coverage. One potential issue, however, is therather large resonance antenna sizes (∼ 30 − 50 cm) that arerequired to achieve maximal data rate in the UHF TV bands.Other potential use of TVWS for smart metering applicationsis for communication to/from home to the operator’s streetcabinet.

C. Cognitive femtocells

A femtocell is a small base station of 2G, 3G, or LTEtechnology, controlled by the mobile operator and placedinside the home /small office of the customer. Femtocells areuseful when a user experiences bad indoor coverage or itsapplication is too capacity-demanding for indoor conditions.The user may be already inside or going inside a building.Femtocells help maintaining a mobile broadband session orto allow it where it previously was not possible. Currentgeneration femtocells use the same frequencies as mobilenetworks, hence creating a potential source of interference

Fig. 7. Proposed architecture for HDTV distribution from a home hub usingTV White Space spectrum.

Fig. 8. Communication architecture proposed for smart metering.

that can be difficult to control since the user femtocell isnot controlled by the operator. Femtocells operating in TVWSwould be an alternative to femtocells proprietary technolo-gies that are appearing on the market for dedicated 2G and3G networks. The main advantage of CR based femto-cellscompared to traditional femtocells will be reduced or bettercontrolled interference into the operators’ network. Anothercase of great interest to operators is to use CR to backhaulingof femtocells (either traditional or CR-based themselves). Thisallows a mobile operator gain control into the home of theuser, should it be outside the DSL coverage or when the userhas another operator delivering broadband to his/her homenetwork.

VI. CONCLUSIONS

In this paper we surveyed the state-of-the-art in cognitiveradio access to TV White Spaces. We showed that a regulatory

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings

framework for secondary utilisation of TVWS spectrum is wellunderway both in the US and UK, and important steps in thisdirection are also being taken within the EU and worldwide.Using result from recent quantitative studies of the TVWSavailability in the UK and US we illustrated that cognitiveaccess to these bands provides a very significant spectrumopportunity for a range of indoor and outdoor applicationsand services. In addition to rural broadband, which is themain focus of the IEEE 802.22 standard, these include wirelesshome networks, smart metering, femtocells, mobile broadbandand wireless vehicular communications.

However, effective exploitation of this spectrum for suchcommercial services requires addressing an array of importanttechnology challenges. One of these, high-precision spectrumsensing, has been the subject of numerous research paperswhile others, including TVWS spectrum database, multiplesecondary access, aggregate interference control, and agiletransmission techniques, have not yet received the attentionwe believe they deserve. Furthermore, quantitative techno-economical studies of the commercial feasibility and costversus benefit associated with use cases of cognitive radio arecrucial in influencing the takeup of the technology by wirelessnetwork and service providers but are currently very limited.Our own research is currently focusing on some of the above-mentioned technology and business challenges of cognitiveaccess to TVWS.

ACKNOWLEDGEMENTS

The author wishes to acknowledge his colleagues at BT, M.Fitch, S. Kawade, K. Briggs, X. Gu, for their contributionsand stimulating discussions.

REFERENCES

[1] A. Wyglynski, M. Nekovee, T. Hou (Eds.), Cognitive Radio Communi-cations and Networks: Principle and Practice, Elsevier, January 2010.

[2] Mitola III, J., Maguire, G.Q,, Cognitive Radio: Making SoftwareRadios More Personal, IEEE Personal Communications, Vol 6, No 4,August 1999.

[3] M. Nekovee, Dynamic spectrum access concepts and future architec-tures, BT Technology Journal, Vol 24, No 2, pp 111-116, 2006.

[4] Federal Communications Commission, Spectrum Policy TaskForce, Report of the Spectrum Efficiency Working Group(http://www.fcc.gov/sptf/reports.html), November 15, 2004.

[5] Federal Communications Commission (FCC) 08-260: Second Reportand Order and Memorandum Opinion and Order in ET Docket Nos.02-380 (Additional Spectrum for Unlicensed Devices Below 900 MHzand in the 3 GHz Band) and 04-186 (Unlicensed Operation in the TVBroadcast Bands), Nov. 14, 2008.

[6] Ofcom, Digital Dividend Review, A statement on our approachtoawarding the digital dividend, 13 December 2007

[7] Ofcom Consultation: Digital Dividend: Cognitive Access, Feb 16,2009. http://www.ofcom.org.uk/consult/condocs/cognitive/summary/

[8] Ofcom, Statement on Cognitive Access to Interleaved Spectrum, 1 July2009.

[9] J. Bradford, T. Cook, D. Ramsbottom, and S. Jones, Optimising usageof spectrum below 15GHz used for defence in the UK, IET Seminaron Software Defined Radio and Cognitive Radio, London,Sept. 2008.

[10] CogNeA website at www.cognea.org/v01/

[11] K. Challapali, Philips Research North America (private communica-tion); FAQ on CogNeA Alliance, K. Kimyacioglu, Philips.

[12] IEEE 802.22 website at http://www.ieee802.org/22/

[13] C. Cordeiro, D. Cavalcanti and S. Shankar, Cognitive Radio forBroadband Wireless Access in TV Bands: The IEEE 802.22 Standards,in Cognitive Radio Communication and Networking: Principle andPractice, A. Wyglynski, M. Nekovee, T. Hou (Eds), Academic Press,2009.

[14] M Mishra, A Sahai, How much white space is there?, Technical report,Electrical Engineering and Computer Sciences, University of Californiaat Berkeley, January 2009.

[15] T. X. Brown and D. C. Sicker, Can cognitive radio support broadbandwireless access, In Proc. IEEE DySPAN07, pp 123-132, April 2007.

[16] A. Sahai and D. Cabric, Spectrum sensing: fundamental limits andpractical challenges, IEEE DySPAN 2005, Baltimore, MD, USA,November 2005. see also a recent unpublished research report fromA. Sahai on the same topics (2009).

[17] T. Yucek and H. Arslan, A survey of spectrum sensing algorithmsfor cognitive radio applications IEEE Communications Surveys &Tutorials, vol. 11, Issue: 1, pp. 116 130, First quarter 2009, andreferences therein.

[18] G. Ganesan, Y. Li, Cooperative spectrum sensing in cognitive radio,IEEE transactions on Wireless communications, June 2007.

[19] A Ghasemi, E.S. Sousa Collaborative spectrum sensing for opportunis-tic access in fading environments, IEEE DySPAN 2005, Baltimore,USA, November 2005.

[20] BT Response to Ofcom Consultation on Digital Dividend: CognitiveAccess.

[21] R. Rajbanshi, A. M. Wyglynski and G. Minden, An efficient imple-mentation of the NC-OFDM transceivers for cognitive radios, Proc.CrownCom 2006, Mykonos, Greece, June 2006; R. Rajbanshi, OFDM-based cognitive radio for DSA networks, PhD Thesis, University ofKansas, Lawrence, KS, USA, 2007.

[22] W. Rhee, C. Chuang, and L. J. Cimini, Performance comparison ofOFDM and multitone with polyphase filterbank for wireless commu-nications, IEEE VTC Conference, May 1998, pp 768-772.

[23] S. Shankar and C. Cordeiro, Analysis of Aggregated Interferenceat DTV receivers in TV bands, Proc. CrownCom 2008, May 2008,Singapore.

[24] M. Jankiraman, Space-Time Codes and MIMO Systems Artech House,2004.

[25] N. Devroye et al., Limits on communications in a cognitive radiochannel, IEEE Comms. Mgn., June 2006.

[26] M. Nekovee, Quantifying the availability of TV White Spaces forcognitive access in the UK, IEEE ICC 2009, June 2009, Dresden,Germany.

[27] UK FREE.TV website at www.ukfree.tv

[28] S. Kawade and M. Nekovee, Can cognitive radio access to TV WhiteSpaces support future home networks?, Proc. IEEE DySPAN, April2010, Singapore.

[29] M. Gosh, V. Gaddam, G. Turkenich, K. Challapali, Spectrum sensingprototype for sensing ATSC and wireless microphone signals, Crown-Com 2008, May 2008, Singapore.

[30] Energy metering: A consultation on smart metering for electricity andgas, UK Department of Energy and Climate Change, May 2009.

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE DySPAN 2010 proceedings