Simulation-based evaluation of spectrum opportunities in UMTS cellular networks

Simulation-based

Evaluation

of

Spectrum

Opportunities

in

UMTS

Cellular

Networks

Artur Pereira*,Atilio Gameiro*t

*Instituto deTelecomunica,oes, Universidade de Aveiro, 3810 Aveiro, Portugal

*tDepartamento deElectronica&

Telecomunica,6es,

Universidade de Aveiro,3810Aveiro, Portugal

Abstract-A scenario based on an UMTS TDD opportunistic cellular system that operates over UMTS FDD licensed cellular networks is considered. Therefore we develop a simulation tool that addresses thegoalofanalysis and assessment of UMTS TDD opportunistic radio system in a coexistence environment with UMTS FDD primary cellular networks. The communication presents the scenario considered, the main features of the tool, discusses and proposes metrics to evaluate the communication opportunities in UMTS FDD primary cellular networks, and presents numerical simulation results. These show that a moderate number of UMTS TDD available frequencies and a reasonable UMTS TDD opportunistic radio transmission power will allow the deployment of UMTS TDD opportunistic radio networks in a coexistence environment with existing licensed systems.

I. INTRODUCTION

With the increasing interest of consumers in new wireless applications, demand for the radio spectrum is expected togrow dramaticallyinthecomingyears. Today,access toradiospectrum is restricted by an old radio regulatory regime of command and control thatemerged overthe last 100 years. Large parts ofour

radio spectrum areallocatedtolicensed radio services and only a small fraction of radio spectrum is open for free, unlicensed operation. Spectrumutilization studies have shown thatmostofthe assignedspectrumis under-utilized [1], that considerablespectrum is available when both the dimensions of space and time are considered. Hence the problem is in most cases a problem of inefficient spectrum management rather than spectrum shortage. The traditional methods of command and control in spectrum management are nolonger optimal, and thus, analternate remedy

tospectrumscarcity istoallow othersystemsto accesssuch under-utilized licensed bandsdynamically

[2].

Unlicensed devices designed for efficiently using shared spectrumwith no or minimal interference to legacy systems are

generally referredto as "cognitive radios" [3]. In abroadsense, a

cognitive radio isaradio thatcanchange its transmitterparameters basedoninteraction with the environmentinwhich itoperates. In

this communication we consider however a narrower definition where the environment awareness considered is restricted to spectrumawarenessandterm itasopportunistic radios (OR).

Inthis communicationweconsiderascenario, whereinaddition

to releasing some new spectrum bands, the regulator allows a

secondary market usage, that is secondary users can use the licensed spectrum provided that they do not cause harmful interference to the owners of the licensed band i.e. the primary

users [4]. This case study fits with the functional scenario "detection offading opportunities", that is, the opportunities are generated by the propagation losses conditions and shadowing between theFDDprimaryusers and theTDDopportunistic radio terminal. In such scenario the opportunistic radios autonomously coordinate theusageof shared spectrum. They actively search the radio spectrum for available frequency bands, and use them dynamically adjusting their transmissions to avoid harmful interference with licensed Primary Users (PU). The specific scenariowe consider here is thecasewhereanUMTS TDDuser canopportunisticallyuse(ifno resources areavailableintheTDD

band) the spectrumofFDDsystems. Forthis scenario evaluation and independently of the specific algorithms or protocols used, resorting to simulations is indispensable for the opportunities evaluation of theORnetwork and its impactonthe licensedone.

Thereforewedevelop a simulation tool that allows assessing the coexistence ofUMTS TDD OR as a secondary system together withprimaryUMTSFDDcellularsystems.

The simulation platform is basedon a system level simulator, designed for the simulation of packet based for UMTS FDD

mobile communication systems with extensions to support opportunistic radio systems. The systemlevel simulator considers that thereareNcellularsystemsinageographicalareaoperatingat

frequency bands

ft,n

fnN

, and a network ofORs which can

operate on any of these frequency bands, but whose communication isasecondpriorityone,i.e. theTDDORcanonly transmit in the band

f,

ifguaranteed that it does not degrade the performance oftheFDDcellular licensedusers.Foreachfrequency

f,

the systemlevel simulatorcomputes, during aduration call, the interference that would be causedbyaTDDOR oneachuserofthe cellular network. These values arethen usedto compute simple statistics for the OR communication opportunities evaluation, in function of the ORtransmission powerand number of available

frequencies.

The remainder ofthispaperisorganizedasfollows.Insection2, the scenario considered for thedevelopmentof the simulation tool is described. In section 3 we describe the main features of the simulation tool and the metrics for the spectrum opportunities evaluation and thenin section4we givesome numerical results, andfinally the main conclusions ofthispaperaregiveninsection5.

II. CELLULAR SCENARIO DESCRIPTION

We considera scenario where theprimary cellular network is an UMTS FDD system and the secondary network, a

centralized cellular network based on the UMTS TDD

standard with extra sensing features and able to switch its carrier frequency to UMTS FDD frequencies. This is motivated bythe fact that 3G radio frequency spectrumis an expensive commodity in many countries, and therefore the opportunistic use of these bands could be one way for the ownersof the licensestomakeextra revenues.

The UMTS FDD usepaired uplink and downlink bands and is intended for highmobility applicationsin macrocell environments with dataratesofup to384Kbit/s and because of thepaired bands is suited forsymmetric traffic. So, an OR systembasedon UMTS TDDcould takeadvantage ofeventual opportunitiesonthese bands and increase the overallUMTSsystemcapacity.

The figure 1 illustrates the concept proposed inthis work. Notice that there isanUMTSFDD DL channel with half load occupancy, theUMTS TDD ORdetects this opportunity and accommodatesan extracapacity.

PSD 5 MHz

5 MHz

1900 1920 1980 2010 2025 2110 2170 MHz [1 UMTSTDD(UL+DL) 0 OR: UMTSTDD(UL+DL)

R UMTSFDD(UL) El UMTSFDD(DL)

Fig. 1. UsinganUMTSFDDULfrequency band inanopportunisticway

WedefineaPUas asubscriber of thesystemthatlegallyowns

theUMTSFDD frequency band.FDDPUsare notORaware,i.e. thereare no meanstoexchange information betweenFDDprimary and TDD secondary users provided by a primary system. Specifically,FDDPUsdonotprovide special signallinginorderto accesstheirfrequency band. Onthe other hand theTDDORisan

entity that is ableto usethespectrumoflicensedusers,provided the measurements orinformation available about this spectrum usage guaranteesit willnotdegrade the quality ofPUscommunication.

TDDORshave constrainedaccess to aPU frequency band. A TDDORcanonly transmitinthefrequency

f

ifguaranteed that it does not degrade the performance of the UMTS FDD cellular licensed users, i.e. if for a reliable operation PU(i) requires SIR>y(i), then it must be guaranteed that the extra interference causedbya TDD useroperatingintheopportunistic mode doesnot

leadto aviolation ofthis condition.

Itis clear from the previous discussion thatifa TDDterminal needs to operate in the OR mode

(because

no resources are

availableinthe licensedTDDband) it needstoaccuratelycompute the interference itmaycause onanyof theUMTSFDDPUs. The acquisition of this information which can be done through

collaborative sensingoreventually throughathirdpartyisbeyond

thescopeofthis communication. Weconsider that it is availableto

theORandconcentrate onthe statisticsconcerning theopportunity

communication availableto an OR assuming such a information available.

The scenario considered inthis work is illustrated in figure

2, where there are N cellular systems overlapped in a

geographical area and operating at frequency bands

f',...,

fN.

The operation of the TDD OR terminals is explained as

follows. Ifthe TDDORterminal has datato transmit/receive,

the base station check ifradio resources are available within the TDD licensed band.

(y?)

Fig.2. Multisystem cellularnetworks

If not, it searches the radio spectrum for available FDD frequency bands, and use them dynamically without decreasing theSIR atFDDPUs locations belowatargetvalue

(y).

Let us consider PU unit i, whose target is a SIR > y(i). We consider that the attenuation between theFDDBSandFDDPUiis L(i), theninthe absence of TDDORstheFDDPUSIRisgiven by,

SIR

P(i)xL()

)

(1)

IiQ)+

N0

whereNois the noisepowerandI(i) the cellular interference (intra-cell andinter-cell).

A TDD OR(k) locatedatpositionsothat there isanattenuation

L,(k,i)

between itself and theFDDPU(i).P0(i)will contributeto a noise level of

P,(k).L,(ki)

atFDDPUi. Then the FDD PUSIR conditiongets,

P(i)xL(i)

>

JO

l(i)+N-A+

P

(k)x

Lojk,

i)

(2)

whichmeanthatTDDOR(k) isonly allowedtotransmitif, P(i)x L(i)>

I(i)

+ N + Pr (k)x L,(k,i)

y(i)

(3)

for all mobilesiactiveat agiven time.

As the sharedspectrumis accessedina nonexclusiveway, the maximum allowableTDDORtransmissionpowerdepends

onpropagation losses conditions and shadowing between the

TDD OR and the FDD PU terminal. However, inthis work,

we assume that the power transmitted by the TDD OR is constant, simplifying, inthat way, the implementation, butin the other hand increasing the number of transitions (hops) between available frequencies.

III. SIMULATION TOOL DESCRIPTION

In this section we give an overview of the system level simulation structure, channel models and metrics for the

assessmentofspectrumopportunities. A. Structure

Thestructureof thesystemlevel simulator is showninfigure3. Thesystemlevel simulator interfaces with thelinklevel simulator through look uptables (LUTs)and therefore canbe used with a

large number of different PHY layers, provided the appropriate LUTs are present asinput to the simulator. Several propagation and traffic modelsare available, and the simulatorcomputesthe entire channel losses(fast and slow fading), thereby ensuringaccuracyin the system level parameters computed. The outputs are the parameters that usually characterize packet transmissions: BER, FER,packet loss, packet delay,ORmetrics,etc,and globalsystem and individual statistics can be calculated using these output parameters.

The current simulationplatform packet system is based on the High-Speed Downlink Packet Access (HSDPA) MAC layer mechanisms.

BER FERPacket delayOR metrics

Fig. 3. Structure of system level simulator

At the beginning of each run,mobile stations arecreated and remain active for the all run duration. Theseusers are randomly uniformly distributedover ahexagonal network ofomni directional cells. Foreachcouple ofFDD PUand BS, arandom shadowing value is drawn, the fast fading is generated and the path loss calculated. Positions, shadowing, path loss and fast fading values

arecalculated for each time transmissioninterval, i.e. theFDDPUs aremoving. Notices thatifanUMTS FDDPUleaves thesystem, i.e. exceed the celllimit, a newFDDPUis inserted in arandom position. Therefore, the number ofPUs present inthe system is always K. Inthat case, even ifa TDD ORis transmitting inthe frequencyf, thenewFDDPUinsertedinthe cellassumesthat the frequency is available andstartsits transmission withoutinforming the TDD OR (theFDD PUs and the TDD OR do not exchange information).

B. Channel Models

The fast fading is generated using Jakes model for fast generation of independent Rayleigh faders [5]. Bi-dimensional shadowing is generated at the beginning of each run. The shadowing is log-normal. The shadowingSH(x,yj) indB between

oneMS(x,y)andoneBSjis thesumoftwovariables,

SH(x,

y,)=

X5

X(F

(x,

y)

+F

(x,y))

(4)

whereFO, andF, are spatial functions generated using the method describedin [6],fulfilling the requirements listed hereafter.FO, and

F, have a Gaussian distribution, with mean equal to ca (the

shadowing standard deviation in dB), and a spatial correlation given by,

-In(2)xd

R(d)=e

D

(5)

where d is the distance betweentwopointsonthenetwork, andD is the shadowing de-correlation length.

The path loss is expressed with the COST231-Walfish-Ikegami model [7] arguments taking into account base station and receiver heights. The expression ofthe path loss Lb in dB is given by,

Lb =73.24+r24.5-

1.5fc

I~

logl0 fc

+

381ogl0

d

(6)

925)

wheref,

denote the carrierfrequencyinMHz, and d is the distance betweentwopointsonthe networkgiveninkm.

Notice that the platform selects the right speed, carrier frequency, path loss, shadowing and fast fading channelparameters accordingtothe environment chosen for the simulation.

C. Metrics

Inorderto share thespectrumbetween primary and secondary users, the simulator distributes K FDD PUs in each of the N availablefrequencies and introduces anORinthatgivencoverage area. At the beginning, the OR tries to start communicating on

frequencyfl.

Ifthe conditionSIR>y(i) is satisfied for1i1... K,the OR transmits in that frequency. Otherwise, the OR tries to

communicateon

frequencyf2,fi,

andso on.Thisfrequency search stopswhen theORfindsacommunication/spectrum opportunityor

whenno morefrequenciesareavailable.

Inthis context, we have defined the probability of transition, which is the probability ofa TDD OR to change its operating frequency during the ORcall. If

P.

represents theprobability of transition, then

IJ N

ENt

j=O

(7)

where

Nt#

denote the number oftransitions betweenstate iandstate jduring theORcall.

This metric allow us to define transition diagrams, and then assuming that the system can be described as a Markov state

diagram, as shown in figure 4. We can completely describe the systembehaviourusing such diagram.

POC

PO1 P10

PNO

PON

Fig. 4. Markov diagram for theTDDORcarrierfrequency transition

ConsideringNfrequencies for the TDD ORcommunication and assumingnoradioresourcesavailableintheUMTSTDDnetwork

* S0: No opportunities are available in the FDD systems and theTDDORterminal isOFF.

* S1: TheTDDcommunicatesthroughanFDD

bandf.

*

P.:

Probability of transition from frequency

f,

to

frequencyf.

With sucha characterization the statistics of theopportunities can

be calculatedusing the transitiondiagram.Let r, = probability that

terminal isin state

S,

thenwehave,

T(SA)=

1

Tti

Decomposingstate SA as

SpuS2,

asshowninfigure6,

(1 -c) (1-1)

1;To

~~=LZJF

(8)

where P is thetransitionprobability,i.e. the elementi1jof P is

P.

[8].

Thus,we candefined the communication probability, obtainedas N

p =1-7ro =

EI7c1

i=l (9)

Other statistics of interestarethe time that the systemspendsin

So,becausenoopportunitiesareavailable, and theaverageduration

ofopportunities. However even when there are opportunitiesthe

systemmayhavetoswitch fromfrequenciesand another statistics of interest is the time between these frequency transitions. To derive such statistics weconsiderthe diagram showninfigure 5,

where D variable representsaniteration(i.e.:T1).

a)D OD

Fig.5. Equivalent Markovdiagram model for theTDDORcarrierfrequency transition

Weareinterested ineventsofthetype such that the systemstarts as So and then returns to So and compute the time spent in the

differentstates. For that andsimilarly totheproceduredone with convolutional codes to characterize the distance profile, we

consider the initial and final states to be So. In the diagram we

consider SA=S1US2 (only two available frequencies). The use of

convolutionaltheoryofMarkovdiagramsallowsustocompute the averagenumber of iterations until the systemreturnsbacktoSoas,

Noo

=1+

I

-¢x nb oftransitions spentinS0 + 1-I nb oftransitions spentinSA (10)

whereoc and representstotheprobabilityof transition fromSoto

So

andSAtoSA, respectively. Therefore theaveragetimespentinSo beforeentering SAis

a

T(So)

= Tt

1-

t-¢x

(1 1)

and the average time spent in SA (i.e. with opportunities to

communicate)is

Fig. 6. StateSAof the equivalent Markovdiagram model for theTDDOR carrierfrequency transition

it is possible to show that during the time T(SA) the number of transitionsbetween

frequenciesf,

andf2,isgiven by

q

int I_ A

Thus, theaveragetime betweenfrequencyswitchis is

At

_

T(SA)

UDTti

1-ID _

Tti

Ntint 1-,fl flq q

(13)

(14) The switching turn-on turn-off times are considered to be

instantaneous.

IV. NUMERICAL RESULTS

Table 1. Mainparametersused for the simulations

Parameter Name UMTSFDD system

Time transmission interval(Ta) Celltype

Cellradius Datarate(Rb) Eb/NOtarget

SIRtarget(y) Spreadingfactor Channel Model Carrierfrequency Channel model

Mobile terminalsvelocity FDDPrimary User(PU)

Number of primary user(s) per

cell/frequency (K) Sensibility/Powerreceived TDDOpportunistic Radio (OR) Number ofopportunistic radio(s) inthe cellcoverage area

(M)

Maximum/Minimum power

transmitted(Po [maxmin]) Durationcall(Ted)

Value 2ms Omni 577m 12.2kbps 9 dB -16 dB 16 Urban 2GHz ITU vehicularA 30 km/h 5 -117dBm[9] 1 10/-44dBm[10] 90s(45000

Tj)

Simulationswerecarriedout tocompute the statistics definedin

the previous section, related to the OR communication opportunities. We consider for the reported simulations that the

(12)

informationnecessary to SIRcalculation is perfect, andwedonot

deal directly with the algorithmstoacquire them, sinceouraim isto

analyse the existing opportunities. The mainparameters used for the simulations are summarized in table 1. We furthermore consider load characteristics identicalineveryUMTSFDDcellular

system, asreferred above, and thefrequenciesare closeenoughso

that thesamestatistical modelsapply.

In order to analyse the impact of the available licensed frequencies on an TDD OR duration call, simulations have been performed considering differentsPO, a TDDORduration callequal

to90seconds andaFDDandTDDterminalsvelocity equalto30 km/h.

The table 2 contains the results related to the probability of each carrierfrequency transition, consideringN=2 andP0=-30, -20, -10,0 and 10dBm.

Table 2. Probability for each possible carrier frequency transition,

consideringN=2and

P0=[-30;10]

dBm

0(dBm)

-30 -20 -10 0 10 Transition p0, 0 0.2864 0.5011 0.5285 0.8069 PoI 0 0.3166 0.2396 0.2287 0.1008

PO2

0 0.3970 0.2593 0.2428 0.0923 Pll 0.9998 0.9796 0.9348 0.6844 0.4454

0lo

0 0.0130 0.0242 0.2252 0.4460 P12 0.0002 0.0074 0.0410 0.0904 0.1087

P22

0.9998 0.9771 0.9402 0.7006 0.4191

^20

0 0.0157 0.0234 0.2173 0.4589 P21 0.0002 0.0072 0.0364 0.0821 0.1220 Table 2 shows that theprobability of continuing instate S1 or

S2 is low for high values of PO; since there are few

communication opportunities in the available frequencies, the TDDOR needto frequently change its carrier frequency.

Basedonthe transitionsprobabilities (table 2),we cansolve

expression (8) for N=2, P00

P1O

p20

P₀₁ P pi1 P21 02 0 10 12 ,1r = 71l '22_ 2__I 2_ (15)

andthrough (9), determine the probability of communication

P,.

Therefore, for P0=-10 and 10 dBm the probability of communication isequalto95.45 and29.93%,respectively;we

show that the lower thepowertransmittedby the TDDOR the

higher the probability of a TDD OR being able to

communicate.

The table 3 shows the results related to the average time

T(SA) and At 's2 .Accordingto Markov diagram Fc=P00, (1

-q)=P11=P22 and f=P1±+P12=P22+P21. Since inpractice,because

of finite simulationtimes, the measurement given F22, we

use P11=P22= (Pl +

P21

2 and P12

=P21=

(2

+121)/

2 Table 3 shows that the average communication time T(SA) decreases with the increase ofpowertransmittedby the OR. Toachieve

anaveragecommunication timeequalto the durationcall,the TDD OR must transmit at most -30 dBm. In this case, the transitions between available frequencies occur every 10 seconds.

Table 3. Average time in state SA and betweenSIand S2, considering

P0=[-30;10]dBm

Av.time

(is)

T(SA)

AtS17S2

PO (dBm) \ -30 90x103 IOx103 -20 137,3728 270,0411 -10 82,0336 50,4496 0 7,0395 18,0580 10 2,4204 9,4937 V. CONCLUSIONS

Inthis paper, we have considered the challenge of

spectrum-sharing between licensed cellular mobile communications system

andvirtually unlicensedusersthat share the licensedspectrumbut have a lower priority and are not allowed to degrade the performance ofthePUs. Arealistic evaluation ofthe coexistence of such systems can only be done resorting to simulations, and therefore a system level simulator, where N cellular systems

operatingatdifferentfrequencies allow secondaryUMTSTDDOR users to share their spectrum was developed. In a general way,

simulation results have shown, considering realistic propagation environments, thatamoderate number of availablefrequencies and

a reasonable TDD OR transmission power will allow the deployment of UMTS TDD OR networks in a coexistence environment withexisting licensedsystems.

REFERENCES

[1] Shared Spectrum Company, "New York City Spectrum Occupancy

MeasurementsSeptember 2004".

[2] A. Ghasemi and E. S. Sousa, "Collaborative Spectrum Sensing for Opportunistic Access in Fading Environments," New Frontiers in

Dynamic Spectrum Access Networks, First IEEE International Symposium,page131-136, November 2005.

[3] J. Mitola, III, "Cognitive Radio for Flexible Mobile Multimedia Communications," Mobile Multimedia Communications, IEEE

InternationalWorkshop,page3, 1999.

[4] R.Brodersenetal, "CORVUS:ACognitive Radio Approach forUseof Virtual UnlicensedSpectrum",UniversityofBerkeley.

[5] Y.LiandX. Huang,"The Generation ofIndependent Rayleigh Faders" Proceedings of IEEE, ICC'00, 2000.

[6] X. Cai, MemberIEEE, and G. B. Giannakis, FellowIEEE, "A

two-Dimensional Channel Simulation Model for Shadowing Processes".

IEEETransactionsOnVehicularTechnology, vol. 52, November 2003. [7] COST 23 IFinal Report. Available:

[8] Ross, Sheldon M. (1997), Introduction to Probability Models, 6th edition, Berkeley,CA:AcademicPress.

[9] 3GPP, "Technical Specification Group Radio Access Network; User

Equipment (UE) radio transmission andreception (FDD) (Release 6)," TS 25.101.v6.13.0, October 2006.

[10] 3GPP, "Technical Specification Group Radio Access Network; User

Equipment (UE) radio transmission andreception (TDD) (Release 6)," TS 25.102.v6.5.0, October 2006.